diff --git a/AGENTS.md b/AGENTS.md index 33e987b93c6..204811c4cd8 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -29,7 +29,7 @@ Running tests: see [Testing](#testing) below (run the most focused test for chan ## Project at a Glance -**OmniRoute** — unified AI proxy/router. One endpoint, 383 LLM providers, auto-fallback. +**OmniRoute** — unified AI proxy/router. One endpoint, 384 LLM providers, auto-fallback. - API Routes — `src/app/api/v1/` (Next.js App Router entry points) - Handlers — `open-sse/handlers/` (request processing: chat, embeddings, etc) @@ -37,7 +37,7 @@ Running tests: see [Testing](#testing) below (run the most focused test for chan - Translators — `open-sse/translator/` (format conversion OpenAI↔Claude↔Gemini) - Transformer — `open-sse/transformer/` (Responses API ↔ Chat Completions) - Services — `open-sse/services/` (combo routing, rate limits, caching, etc) -- Database — `src/lib/db/` (SQLite domain modules, 230 migrations) +- Database — `src/lib/db/` (SQLite domain modules, 232 migrations) - Domain/Policy — `src/domain/` (policy engine, cost rules, fallback logic) - MCP Server — `open-sse/mcp-server/` (110 tools: 45 canonical + memory/skill/GitHub/pool/gamification/plugin/Notion/Obsidian/local-corpus/RTK modules; 3 transports: stdio / SSE / Streamable HTTP; 33 scopes) - A2A Server — `src/lib/a2a/` (JSON-RPC 2.0 agent protocol) diff --git a/README.md b/README.md index 7a540c87c17..b6798961cfc 100644 --- a/README.md +++ b/README.md @@ -9,7 +9,7 @@ **Created by [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Never stop coding. Every AI tool → 383 providers — 150+ free — through one endpoint. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity into FREE Claude / GPT / Gemini with auto-fallback. RTK + Caveman stacked compression saves 15–95% tokens (~89% avg) — never hit limits. 383 AI providers · 150+ free tiers · ~1.62B free tokens/mo · 20 routing strategies · $0 to start. +OmniRoute — Never stop coding. Every AI tool → 384 providers — 150+ free — through one endpoint. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity into FREE Claude / GPT / Gemini with auto-fallback. RTK + Caveman stacked compression saves 15–95% tokens (~89% avg) — never hit limits. 384 AI providers · 150+ free tiers · ~1.62B free tokens/mo · 20 routing strategies · $0 to start. @@ -67,7 +67,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------- | :-----------------------: | :---------: | :-------------------: | -| 🌐 Providers | 352 | **358** | 383 | +| 🌐 Providers | 352 | **358** | 384 | | 🧠 Unique chat model IDs | 1320 | **1374** | 1491 | | 🖼️ Modality Bridge | 🆕 vision + audio + video | ✓ | ✓ | | 📡 Radar free catalog | 🆕 opt-in | ✓ | ✓ | @@ -108,7 +108,7 @@ ⚙️ Features 🎯 Combos - 🌐 Providers + 🌐 Providers 🔌 CLI & MCP @@ -244,7 +244,7 @@ curl http://localhost:20128/v1/chat/completions \ -The Promise — One endpoint and 383 providers. Automatic fallback keeps routing while another healthy target is available. Six pillars: resilient fallback across 383 providers · up to 95% token savings on eligible workloads · $0 to start with 150+ free tiers and 52 recurring/keyless free-forever providers · 39 CLI/agent integrations through one config · OpenAI, Claude, Gemini and Responses API compatibility at /v1 · production controls including circuit breakers, TLS stealth, MCP 110 tools, A2A, memory, guardrails, evals and 56,000+ static test declarations across 7,800+ tracked test files. +The Promise — One endpoint and 384 providers. Automatic fallback keeps routing while another healthy target is available. Six pillars: resilient fallback across 384 providers · up to 95% token savings on eligible workloads · $0 to start with 150+ free tiers and 52 recurring/keyless free-forever providers · 39 CLI/agent integrations through one config · OpenAI, Claude, Gemini and Responses API compatibility at /v1 · production controls including circuit breakers, TLS stealth, MCP 110 tools, A2A, memory, guardrails, evals and 56,000+ static test declarations across 7,800+ tracked test files.

@@ -504,7 +504,7 @@ All **20** strategies — mix & match per combo step: -What sets OmniRoute apart — a dated feature snapshot vs 9router, OpenRouter, CLIProxyAPI and LiteLLM across 13 capabilities. OmniRoute: 383 providers, 150+ free tiers built in, 20 routing strategies, 12-engine token compression, built-in MCP server with 110 tools, A2A agent protocol, persistent memory, guardrails, cloud agents, TLS fingerprint stealth, Desktop/Termux/PWA and 67 i18n UI locales. OmniRoute is MIT-licensed and self-hostable. Competitor capabilities and counts may change; see the linked methodology. +What sets OmniRoute apart — a dated feature snapshot vs 9router, OpenRouter, CLIProxyAPI and LiteLLM across 13 capabilities. OmniRoute: 384 providers, 150+ free tiers built in, 20 routing strategies, 12-engine token compression, built-in MCP server with 110 tools, A2A agent protocol, persistent memory, guardrails, cloud agents, TLS fingerprint stealth, Desktop/Termux/PWA and 67 i18n UI locales. OmniRoute is MIT-licensed and self-hostable. Competitor capabilities and counts may change; see the linked methodology. 📊 Full methodology & per-feature detail vs 9router, OpenRouter, CLIProxyAPI & LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -608,7 +608,7 @@ the current catalog at **[radar.omniroute.online/planos](https://radar.omniroute - **🖼️ New endpoints** — `/v1/ocr` (Mistral OCR) and `/v1/audio/translations` (Whisper-style) round out the media surface. → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 Image / video / audio generation** — one API for media: xAI Grok Imagine & Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind, and speech providers such as ElevenLabs. → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 Deployment & ops** — reverse-proxy `basePath`, browser-language auto-detect, per-key device tracking, root-less MITM trust, zh-TW localization. → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 More providers & agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) with browser + OAuth login, Ollama first-class card, Claude Opus 5 & Sonnet 5, Kimi official partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… and a refreshed **383-provider catalog**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 More providers & agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) with browser + OAuth login, Ollama first-class card, Claude Opus 5 & Sonnet 5, Kimi official partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… and a refreshed **384-provider catalog**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 Routing transparency** — every response carries an `X-OmniRoute-Decision` header naming the strategy/provider/latency that served it, a new `cache-optimized` combo strategy + Auto-Combo `cacheAffinity` factor route repeat requests back to the connection holding the cached prefix, and a read-only `/v1/auto-combo/{channel}/candidates` endpoint exposes an `auto/*` channel's live candidate pool. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Local performance & infra** — one-click local Redis, Cloudflare Workers / Deno Deploy relay deployers, Bifrost & Mux as supervised embedded services. → [Embedded Services](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Also in the box** — plugin framework + marketplace, Omni/Agent/GitHub skills frameworks, Obsidian vault integration (22 MCP tools), OpenAI-compatible Batch & Files APIs, semantic response cache, gamification with leaderboards, ACP agent discovery (15 built-in agents), scheduled log export to BigQuery, `auto/chaos` parallel multi-model fan-out, a Telegram bot bridge, an in-app version manager and LMArena-ELO free-provider rankings. → [Docs](docs/README.md) @@ -692,11 +692,11 @@ of your shell history. → [CLI Integrations](docs/guides/CLI-INTEGRATIONS.md)
-## 🌐 383 AI Providers — 153 Catalog-Marked Free +## 🌐 384 AI Providers — 153 Catalog-Marked Free
-> **383 registered providers** across the canonical chat, media, search, local, cloud-agent and system collections, including **153 carrying `hasFree: true` discovery metadata**. The chat model registry covers **239 providers / 3,059 distinct provider-model pairs / 1,475 raw model IDs**; the separate free-budget catalog has **471 per-model rows**, **34 recurring pools** and **52 recurring/keyless free-forever providers**. These are different denominators by design; definitions and pool-deduped calculations live in the [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) and [Free Tiers](docs/reference/FREE_TIERS.md). +> **384 registered providers** across the canonical chat, media, search, local, cloud-agent and system collections, including **153 carrying `hasFree: true` discovery metadata**. The chat model registry covers **239 providers / 3,059 distinct provider-model pairs / 1,475 raw model IDs**; the separate free-budget catalog has **471 per-model rows**, **34 recurring pools** and **52 recurring/keyless free-forever providers**. These are different denominators by design; definitions and pool-deduped calculations live in the [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) and [Free Tiers](docs/reference/FREE_TIERS.md).
@@ -1313,7 +1313,7 @@ Métricas canônicas em 2026-10-18: **1.976 vídeos únicos** · **20.220.289 vi RuntimeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 LanguageTypeScript 6.0 — 100% TypeScript across src/ and open-sse/ (zero any in core since v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Databasebetter-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — 147 domain modules, 230 migrations + Databasebetter-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — 147 domain modules, 232 migrations MemorySQLite FTS5 full-text + int8-quantized vector embeddings, typed decay SchemasZod 4 — MCP tool I/O validation + API contracts ProtocolsMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/changelog.d/features/14833-agent-sessions-project-attribution.md b/changelog.d/features/14833-agent-sessions-project-attribution.md new file mode 100644 index 00000000000..07693cfabf4 --- /dev/null +++ b/changelog.d/features/14833-agent-sessions-project-attribution.md @@ -0,0 +1 @@ +- **feat(usage):** attribute proxied requests to coding-agent sessions and projects for team usage reporting, recording session token counters, priced cost, working directory, and project repo ([#14833](https://github.com/diegosouzapw/OmniRoute/pull/14833)) — thanks @fouadSalkini diff --git a/changelog.d/features/18356-paraloncloud-provider.md b/changelog.d/features/18356-paraloncloud-provider.md new file mode 100644 index 00000000000..8f9b00ce2e7 --- /dev/null +++ b/changelog.d/features/18356-paraloncloud-provider.md @@ -0,0 +1 @@ +- **feat(providers):** add ParalonCloud API-key provider (chat + embeddings) ([#18356](https://github.com/cheaper-inference/OmniRoute/pull/18356)) — thanks @robertjitaru diff --git a/docs/architecture/ARCHITECTURE.md b/docs/architecture/ARCHITECTURE.md index 63b066d39cb..da00e36e225 100644 --- a/docs/architecture/ARCHITECTURE.md +++ b/docs/architecture/ARCHITECTURE.md @@ -17,7 +17,7 @@ It provides a single OpenAI-compatible endpoint (`/v1/*`) and routes traffic acr Core capabilities: -- OpenAI-compatible API surface for CLI/tools (383 providers, 159 executor modules) +- OpenAI-compatible API surface for CLI/tools (384 providers, 159 executor modules) - Request/response translation across provider formats - Model combo fallback (multi-model sequence) - Structured combo steps (`provider + model + connection`) with runtime ordering by `compositeTiers` diff --git a/docs/architecture/CODEBASE_DOCUMENTATION.md b/docs/architecture/CODEBASE_DOCUMENTATION.md index 1c16ac947c2..f25a6e86998 100644 --- a/docs/architecture/CODEBASE_DOCUMENTATION.md +++ b/docs/architecture/CODEBASE_DOCUMENTATION.md @@ -501,7 +501,7 @@ helpers (tool schemas, fetch/TTS adapters, quirk handlers). Executors include: (shared identity helper) and `index.ts` (registry). > Note: providers not listed here are served by `default.ts` using the generic -> OpenAI-compatible executor. The full provider catalog (383 providers) lives in +> OpenAI-compatible executor. The full provider catalog (384 providers) lives in > `src/shared/constants/providers.ts`. ### 4.3 `open-sse/translator/` diff --git a/docs/architecture/REPOSITORY_MAP.md b/docs/architecture/REPOSITORY_MAP.md index baa0bc4d16d..0e8d235dfec 100644 --- a/docs/architecture/REPOSITORY_MAP.md +++ b/docs/architecture/REPOSITORY_MAP.md @@ -244,7 +244,7 @@ src/ | -------------------------------- | --------------------------------------------------------------------------------------------------- | | `components/` | Dashboard component API: adapters over `packages/ui-kit` — see `docs/architecture/DESIGN_SYSTEM.md` | | `constants/iconMap.ts` | `ICON_MAP`: icon names used by the screens → Lucide components (``) | -| `constants/providers.ts` | **383 providers** with Zod validation (source of truth) | +| `constants/providers.ts` | **384 providers** with Zod validation (source of truth) | | `constants/cliTools.ts` | External CLI tool registry | | `constants/routingStrategies.ts` | **20 routing strategies** with priorities | | `constants/publicApiRoutes.ts` | Routes that require Bearer (vs management) auth | @@ -414,7 +414,7 @@ open-sse/ | `CLI-TOOLS.md` | External CLI integrations + Internal OmniRoute CLI | | `I18N.md` | i18n architecture, adding a language, 67 locales | | `UNINSTALL.md` | Clean uninstall steps | -| `PROVIDER_REFERENCE.md` | **Auto-generated** catalog of 383 providers (regen: `npm run gen:provider-reference`) | +| `PROVIDER_REFERENCE.md` | **Auto-generated** catalog of 384 providers (regen: `npm run gen:provider-reference`) | ### Subsystem deep-dives diff --git a/docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md b/docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md index aade780eb61..499381a65f9 100644 --- a/docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md +++ b/docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md @@ -41,7 +41,7 @@ Objective feature comparison vs popular open-source AI routers. ## When to choose OmniRoute -- You self-host and want **maximum provider coverage** (383 providers, 155 free/no-auth catalog entries) +- You self-host and want **maximum provider coverage** (384 providers, 155 free/no-auth catalog entries) - You need a **built-in MCP server** (LLM tools, memory, skills exposed as tools) - You need **A2A protocol** for agent-to-agent workflows - You want **fingerprint stealth** (JA3/JA4) to avoid detection by upstream CAPTCHAs diff --git a/docs/diagrams/cli-terminal.svg b/docs/diagrams/cli-terminal.svg index 65c6a50b883..99eb8185148 100644 --- a/docs/diagrams/cli-terminal.svg +++ b/docs/diagrams/cli-terminal.svg @@ -1,5 +1,5 @@ - Animated terminal demoing the OmniRoute CLI: omniroute providers list (383 providers registered, anthropic, codex, glm, kimi shown active), omniroute combo list (always-on priority, cost-saver, fusion-panel, context-relay) and omniroute health (healthy, 18412 requests in 24h, p95 412ms, circuit breakers 24 closed, 1 half-open, 0 open), cycling over 89 top-level commands: providers, oauth, keys, combo, nodes, models, cache, compression, cost, usage, quota, health, resilience, telemetry, logs, audit, mcp, a2a, cloud, memory, skills, eval, doctor, repl, tunnel, backup, sync, webhooks, policy, pricing, translator, simulate and more. + Animated terminal demoing the OmniRoute CLI: omniroute providers list (384 providers registered, anthropic, codex, glm, kimi shown active), omniroute combo list (always-on priority, cost-saver, fusion-panel, context-relay) and omniroute health (healthy, 18412 requests in 24h, p95 412ms, circuit breakers 24 closed, 1 half-open, 0 open), cycling over 89 top-level commands: providers, oauth, keys, combo, nodes, models, cache, compression, cost, usage, quota, health, resilience, telemetry, logs, audit, mcp, a2a, cloud, memory, skills, eval, doctor, repl, tunnel, backup, sync, webhooks, policy, pricing, translator, simulate and more. Compact animated terminal cycling three real OmniRoute CLI commands with a typewriter effect and a scrolling subcommand ticker; the first frame shows the completed providers-list screen. OmniRoute — ኮድ መጻፍን በፍጹም አታቁም። እያንዳንዱ AI መሣሪያ → 383 አቅራቢዎች — ከ150+ ነጻ — በአንድ ኤንድፖይንት በኩል። Claude Code፣ Codex፣ Cursor፣ Cline፣ Copilot እና Antigravity ወደ ነጻ Claude / GPT / Gemini በራስ-ሰር fallback። RTK + Caveman የተደራረበ መጭመቂያ ከ15–95% ቶከኖችን (~89% በአማካይ) ይቆጥባል — ገደቦችን በፍጹም አትመታ። 383 AI አቅራቢዎች · 150+ ነጻ ደረጃዎች · ~1.62B ነጻ ቶከኖች/ወር · 20 የማዞሪያ ስልቶች · ለመጀመር $0። +OmniRoute — ኮድ መጻፍን በፍጹም አታቁም። እያንዳንዱ AI መሣሪያ → 384 አቅራቢዎች — ከ150+ ነጻ — በአንድ ኤንድፖይንት በኩል። Claude Code፣ Codex፣ Cursor፣ Cline፣ Copilot እና Antigravity ወደ ነጻ Claude / GPT / Gemini በራስ-ሰር fallback። RTK + Caveman የተደራረበ መጭመቂያ ከ15–95% ቶከኖችን (~89% በአማካይ) ይቆጥባል — ገደቦችን በፍጹም አትመታ። 384 AI አቅራቢዎች · 150+ ነጻ ደረጃዎች · ~1.62B ነጻ ቶከኖች/ወር · 20 የማዞሪያ ስልቶች · ለመጀመር $0።
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------- | :----------------: | :---------: | :-------------------: | -| 🌐 አቅራቢዎች | 352 | **358** | 383 | +| 🌐 አቅራቢዎች | 352 | **358** | 384 | | 🧠 ልዩ የውይይት ሞዴል መለያዎች | 1320 | **1374** | 1491 | | 🖼️ የሞዳሊቲ ድልድይ | 🆕 ራዕይ + ድምጽ + ቪዲዮ | ✓ | ✓ | | 📡 የRadar ነፃ ካታሎግ | 🆕 በፈቃደኝነት ምርጫ | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ ባህሪያት 🎯 ጥምረቶች - 🌐 አቅራቢዎች + 🌐 አቅራቢዎች 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -ተስፋው — አንድ ኤንድፖይንት እና 383 አቅራቢዎች። ሌላ ጤናማ መዳረሻ እስካለ ድረስ ራስ-ሰር መተካት ማዘዋወሩን ይቀጥላል። ስድስት መሠረታዊ አምዶች፦ በ383 አቅራቢዎች መካከል ጽኑ የመተካት አቅም · ብቁ በሆኑ የሥራ ጫናዎች ላይ እስከ 95% የቶከን ቁጠባ · ከ150+ ነፃ ደረጃዎች እና 52 በየጊዜው የሚሰጡ/ቁልፍ-አልባ ለዘላለም ነፃ አቅራቢዎች ጋር በ$0 መጀመር · በአንድ ቅንብር በኩል 39 CLI/agent ውህደቶች · OpenAI፣ Claude፣ Gemini እና Responses API ተኳሃኝነት በ/v1 · የወረዳ ቆራጮች፣ TLS stealth፣ MCP 110 መሣሪያዎች፣ A2A፣ ማስታወሻ፣ guardrails፣ evals እና በ7,800+ የተከታተሉ የሙከራ ፋይሎች ውስጥ ከ56,000+ የማይንቀሳቀሱ የሙከራ መግለጫዎችን ጨምሮ የምርት አካባቢ ቁጥጥሮች። +ተስፋው — አንድ ኤንድፖይንት እና 384 አቅራቢዎች። ሌላ ጤናማ መዳረሻ እስካለ ድረስ ራስ-ሰር መተካት ማዘዋወሩን ይቀጥላል። ስድስት መሠረታዊ አምዶች፦ በ384 አቅራቢዎች መካከል ጽኑ የመተካት አቅም · ብቁ በሆኑ የሥራ ጫናዎች ላይ እስከ 95% የቶከን ቁጠባ · ከ150+ ነፃ ደረጃዎች እና 52 በየጊዜው የሚሰጡ/ቁልፍ-አልባ ለዘላለም ነፃ አቅራቢዎች ጋር በ$0 መጀመር · በአንድ ቅንብር በኩል 39 CLI/agent ውህደቶች · OpenAI፣ Claude፣ Gemini እና Responses API ተኳሃኝነት በ/v1 · የወረዳ ቆራጮች፣ TLS stealth፣ MCP 110 መሣሪያዎች፣ A2A፣ ማስታወሻ፣ guardrails፣ evals እና በ7,800+ የተከታተሉ የሙከራ ፋይሎች ውስጥ ከ56,000+ የማይንቀሳቀሱ የሙከራ መግለጫዎችን ጨምሮ የምርት አካባቢ ቁጥጥሮች።

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRouteን የሚለየው — በ13 ችሎታዎች ላይ ከ9router፣ OpenRouter፣ CLIProxyAPI እና LiteLLM ጋር የተደረገ የቀን ማህተም ያለው የባህሪያት ንጽጽር። OmniRoute፦ 383 አቅራቢዎች፣ ከ150 በላይ ነፃ የአጠቃቀም ደረጃዎች አብረው የተካተቱ፣ 20 የማዞሪያ ስልቶች፣ 12-ሞተር የቶከን መጨመቂያ፣ 110 መሣሪያዎች ያሉት አብሮገነብ MCP አገልጋይ፣ A2A የኤጀንት ፕሮቶኮል፣ ቋሚ ማህደረ ትውስታ፣ የመከላከያ መመሪያዎች፣ የደመና ኤጀንቶች፣ የTLS አሻራ ምስጢራዊነት፣ Desktop/Termux/PWA እና 67 የi18n የተጠቃሚ በይነገጽ አካባቢያዊ ትርጉሞች። OmniRoute በMIT ፈቃድ የተሰጠው እና በራስዎ ማስተናገድ የሚችሉት ነው። የተወዳዳሪዎቹ ችሎታዎችና ቁጥሮች ሊቀየሩ ይችላሉ፤ የተገናኘውን ዘዴያዊ ሰነድ ይመልከቱ። +OmniRouteን የሚለየው — በ13 ችሎታዎች ላይ ከ9router፣ OpenRouter፣ CLIProxyAPI እና LiteLLM ጋር የተደረገ የቀን ማህተም ያለው የባህሪያት ንጽጽር። OmniRoute፦ 384 አቅራቢዎች፣ ከ150 በላይ ነፃ የአጠቃቀም ደረጃዎች አብረው የተካተቱ፣ 20 የማዞሪያ ስልቶች፣ 12-ሞተር የቶከን መጨመቂያ፣ 110 መሣሪያዎች ያሉት አብሮገነብ MCP አገልጋይ፣ A2A የኤጀንት ፕሮቶኮል፣ ቋሚ ማህደረ ትውስታ፣ የመከላከያ መመሪያዎች፣ የደመና ኤጀንቶች፣ የTLS አሻራ ምስጢራዊነት፣ Desktop/Termux/PWA እና 67 የi18n የተጠቃሚ በይነገጽ አካባቢያዊ ትርጉሞች። OmniRoute በMIT ፈቃድ የተሰጠው እና በራስዎ ማስተናገድ የሚችሉት ነው። የተወዳዳሪዎቹ ችሎታዎችና ቁጥሮች ሊቀየሩ ይችላሉ፤ የተገናኘውን ዘዴያዊ ሰነድ ይመልከቱ። 📊 ሙሉ ዘዴያዊ ሰነድ እና በእያንዳንዱ ባህሪ ላይ ዝርዝር ንጽጽር ከ9router፣ OpenRouter፣ CLIProxyAPI እና LiteLLM ጋር → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar በምርጫ የሚነቃ ሲሆን GET ጥያቄዎችን ብቻ ይጠ - **🖼️ አዲስ ኤንድፖይንቶች** — `/v1/ocr` (Mistral OCR) እና `/v1/audio/translations` (እንደ Whisper ያለ) የሚዲያ ተግባራትን ያሟላሉ። → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 የምስል / ቪዲዮ / ድምፅ ማመንጨት** — ለሚዲያ አንድ API፦ xAI Grok Imagine እና Novita AI ቪዲዮ፣ ComfyUI፣ Magnific፣ Adobe Firefly፣ Segmind እና እንደ ElevenLabs ያሉ የንግግር ፕሮቫይደሮች። → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 ማሰማራት እና ኦፕሬሽኖች** — የሪቨርስ-ፕሮክሲ `basePath`፣ በአሳሽ ቋንቋ ራስ-ሰር ማወቅ፣ በእያንዳንዱ ቁልፍ የመሣሪያ ክትትል፣ ሩት-አልባ MITM መተማመን፣ zh-TW አካባቢያዊ ትርጉም። → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 ተጨማሪ ፕሮቫይደሮች እና ወኪሎች** — የደመና ወኪሎች (Codex Cloud, Cursor, Devin, Jules)፣ Grok Build (xAI) ከአሳሽ + OAuth መግቢያ ጋር፣ የመጀመሪያ-ደረጃ Ollama ካርድ፣ Claude Opus 5 እና Sonnet 5፣ የKimi ይፋዊ አጋርነት (Code/Web/Moonshot)፣ Zed፣ Requesty፣ SenseNova፣ Yuanbao፣ Agnes AI… እና እንደገና የተዘጋጀ **383-ፕሮቫይደር ካታሎግ**። → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 ተጨማሪ ፕሮቫይደሮች እና ወኪሎች** — የደመና ወኪሎች (Codex Cloud, Cursor, Devin, Jules)፣ Grok Build (xAI) ከአሳሽ + OAuth መግቢያ ጋር፣ የመጀመሪያ-ደረጃ Ollama ካርድ፣ Claude Opus 5 እና Sonnet 5፣ የKimi ይፋዊ አጋርነት (Code/Web/Moonshot)፣ Zed፣ Requesty፣ SenseNova፣ Yuanbao፣ Agnes AI… እና እንደገና የተዘጋጀ **384-ፕሮቫይደር ካታሎግ**። → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 ግልጽ የራውቲንግ ሂደት** — እያንዳንዱ ምላሽ ያገለገለውን ስትራቴጂ/ፕሮቫይደር/መዘግየት የሚጠቅስ `X-OmniRoute-Decision` ራስጌ ይይዛል፤ አዲስ `cache-optimized` የኮምቦ ስትራቴጂ + Auto-Combo `cacheAffinity` ምክንያት ተደጋጋሚ ጥያቄዎችን የተሸጎጠውን ቅድመ-ቅጥያ ወደያዘው ግንኙነት ይመልሳል፤ እንዲሁም ተነባቢ-ብቻ `/v1/auto-combo/{channel}/candidates` ኤንድፖይንት የአንድ `auto/*` ቻናል ቀጥታ የእጩዎች ስብስብን ያሳያል። → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ የአካባቢ አፈጻጸም እና መሠረተ ልማት** — በአንድ ጠቅታ የሚሠራ የአካባቢ Redis፣ Cloudflare Workers / Deno Deploy የማስተላለፊያ ማሰማሪያዎች፣ Bifrost እና Mux እንደሚቆጣጠሩ የተካተቱ አገልግሎቶች። → [የተካተቱ አገልግሎቶች](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 እንዲሁም በጥቅሉ ውስጥ** — የፕለጊን ማዕቀፍ እና የገበያ መድረክ፣ Omni/Agent/GitHub የክህሎት ማዕቀፎች፣ የObsidian vault ውህደት (22 MCP መሣሪያዎች)፣ ከOpenAI ጋር ተኳሃኝ የBatch እና Files APIs፣ የትርጉም ምላሽ መሸጎጫ፣ የደረጃ ሰሌዳዎች ያሉት የጨዋታ አቀራረብ፣ ACP የወኪል ግኝት (15 አብሮ የተሰሩ ወኪሎች)፣ ወደ BigQuery በመርሐ-ግብር የሚደረግ የሎግ ማስወጣት፣ `auto/chaos` በትይዩ የሚሠራ የብዝሃ-ሞዴል fan-out፣ የTelegram bot ድልድይ፣ በመተግበሪያው ውስጥ ያለ የስሪት አስተዳዳሪ እና የLMArena-ELO ነፃ አቅራቢዎች ደረጃ አሰጣጦች። → [ሰነዶች](docs/README.md) @@ -688,11 +688,11 @@ omniroute configure codex # እንዲሁም: claude opencode qwen aider
-## 🌐 383 የAI አቅራቢዎች — 153 በካታሎግ ነፃ ተመዝግበዋል +## 🌐 384 የAI አቅራቢዎች — 153 በካታሎግ ነፃ ተመዝግበዋል
-> በመደበኛ የውይይት፣ የሚዲያ፣ የፍለጋ፣ የአካባቢ፣ የክላውድ-ኤጀንት እና የሲስተም ስብስቦች ውስጥ **383 የተመዘገቡ አቅራቢዎች** አሉ፤ ከእነዚህም **153 `hasFree: true` የግኝት ሜታዳታ የያዙ** ናቸው። የውይይት ሞዴል መዝገቡ **239 አቅራቢዎች / 3,059 የተለያዩ አቅራቢ-ሞዴል ጥንዶች / 1,475 ጥሬ የሞዴል መለያዎች** ይሸፍናል፤ የተለየው የነፃ በጀት ካታሎግ ደግሞ **471 በሞዴል የተደራጁ ረድፎች**፣ **34 ተደጋጋሚ ገንዳዎች** እና **52 ተደጋጋሚ/ቁልፍ-አልባ ለዘላለም ነፃ አቅራቢዎች** አሉት። እነዚህ በንድፍ የተለያዩ መለኪያዎች ናቸው፤ ትርጓሜዎቹ እና በገንዳ የተባዙ ስሌቶቹ በ [የአቅራቢ ማጣቀሻ](docs/reference/PROVIDER_REFERENCE.md) እና [ነፃ ደረጃዎች](docs/reference/FREE_TIERS.md) ውስጥ ይገኛሉ። +> በመደበኛ የውይይት፣ የሚዲያ፣ የፍለጋ፣ የአካባቢ፣ የክላውድ-ኤጀንት እና የሲስተም ስብስቦች ውስጥ **384 የተመዘገቡ አቅራቢዎች** አሉ፤ ከእነዚህም **153 `hasFree: true` የግኝት ሜታዳታ የያዙ** ናቸው። የውይይት ሞዴል መዝገቡ **239 አቅራቢዎች / 3,059 የተለያዩ አቅራቢ-ሞዴል ጥንዶች / 1,475 ጥሬ የሞዴል መለያዎች** ይሸፍናል፤ የተለየው የነፃ በጀት ካታሎግ ደግሞ **471 በሞዴል የተደራጁ ረድፎች**፣ **34 ተደጋጋሚ ገንዳዎች** እና **52 ተደጋጋሚ/ቁልፍ-አልባ ለዘላለም ነፃ አቅራቢዎች** አሉት። እነዚህ በንድፍ የተለያዩ መለኪያዎች ናቸው፤ ትርጓሜዎቹ እና በገንዳ የተባዙ ስሌቶቹ በ [የአቅራቢ ማጣቀሻ](docs/reference/PROVIDER_REFERENCE.md) እና [ነፃ ደረጃዎች](docs/reference/FREE_TIERS.md) ውስጥ ይገኛሉ።
@@ -1295,7 +1295,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1ም ይዘለዋ የማስኬጃ አካባቢNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ቋንቋTypeScript 6.0 — በsrc/ እና open-sse/ ውስጥ 100% TypeScript (ከv2.0 ጀምሮ በዋናው ኮድ ውስጥ ዜሮ any) ፍሬምወርክNext.js 16 + React 19 + Tailwind CSS 4 - የውሂብ ጎታbetter-sqlite3 (SQLite፣ WAL ጆርናሊንግ) + LowDB (የJSON ውርስ) — 147 የዶሜይን ሞጁሎች፣ 230 ማይግሬሽኖች + የውሂብ ጎታbetter-sqlite3 (SQLite፣ WAL ጆርናሊንግ) + LowDB (የJSON ውርስ) — 147 የዶሜይን ሞጁሎች፣ 232 ማይግሬሽኖች ማህደረ ትውስታየSQLite FTS5 ሙሉ-ጽሑፍ + int8-quantized የቬክተር ኢምቤዲንጎች፣ የተየበ መቀነስ ስኪማዎችZod 4 — የMCP መሣሪያ I/O ማረጋገጫ + የAPI ውሎች ፕሮቶኮሎችMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/am/llm.txt b/docs/i18n/am/llm.txt index c4f6c0f568a..dad5072a8b8 100644 --- a/docs/i18n/am/llm.txt +++ b/docs/i18n/am/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ar/README.md b/docs/i18n/ar/README.md index 817f625c7f0..e532cf23b57 100644 --- a/docs/i18n/ar/README.md +++ b/docs/i18n/ar/README.md @@ -15,7 +15,7 @@ **من إنشاء [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — لا تتوقف عن البرمجة. كل أدوات الذكاء الاصطناعي → 383 موفرًا — أكثر من 150 موفرًا مجانيًا — عبر نقطة نهاية واحدة. Claude Code وCodex وCursor وCline وCopilot وAntigravity إلى Claude / GPT / Gemini مجانًا مع التحويل التلقائي عند التعذر. يوفر الضغط المتسلسل RTK + Caveman نسبة 15–95% من الرموز (بمتوسط ~89%) — لا تتجاوز الحدود أبدًا. 383 موفرًا للذكاء الاصطناعي · أكثر من 150 مستوى مجانيًا · ~1.62B رمز مجاني شهريًا · 20 استراتيجية توجيه · ابدأ بتكلفة $0. +OmniRoute — لا تتوقف عن البرمجة. كل أدوات الذكاء الاصطناعي → 384 موفرًا — أكثر من 150 موفرًا مجانيًا — عبر نقطة نهاية واحدة. Claude Code وCodex وCursor وCline وCopilot وAntigravity إلى Claude / GPT / Gemini مجانًا مع التحويل التلقائي عند التعذر. يوفر الضغط المتسلسل RTK + Caveman نسبة 15–95% من الرموز (بمتوسط ~89%) — لا تتجاوز الحدود أبدًا. 384 موفرًا للذكاء الاصطناعي · أكثر من 150 مستوى مجانيًا · ~1.62B رمز مجاني شهريًا · 20 استراتيجية توجيه · ابدأ بتكلفة $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | -------------------------------- | :-------------------------: | :---------: | :-------------------: | -| 🌐 المزوّدون | 352 | **358** | 383 | +| 🌐 المزوّدون | 352 | **358** | 384 | | 🧠 معرّفات نماذج الدردشة الفريدة | 1320 | **1374** | 1491 | | 🖼️ جسر الوسائط | 🆕 الرؤية + الصوت + الفيديو | ✓ | ✓ | | 📡 كتالوج Radar المجاني | 🆕 الاشتراك الاختياري | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ الميزات 🎯 التركيبات - 🌐 المزوّدون + 🌐 المزوّدون 🔌 CLI وMCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -الوعد — نقطة نهاية واحدة و383 موفّرًا. يواصل الرجوع التلقائي التوجيه ما دام هناك هدف آخر سليم متاح. ست ركائز: رجوع مرن عبر 383 موفّرًا · توفير يصل إلى 95% من الرموز في أعباء العمل المؤهلة · البدء بتكلفة 0 دولار مع أكثر من 150 مستوى مجانيًا و52 موفّرًا مجانيًا دائمًا متكررًا/بلا مفاتيح · 39 تكاملًا مع CLI/الوكلاء من خلال إعداد واحد · التوافق مع OpenAI وClaude وGemini وResponses API عبر /v1 · عناصر تحكم للإنتاج تشمل قواطع الدارات، وتخفّي TLS، و110 أدوات MCP، وA2A، والذاكرة، والضوابط الوقائية، والتقييمات، وأكثر من 56,000 تصريح اختبار ثابت عبر أكثر من 7,800 ملف اختبار متعقَّب. +الوعد — نقطة نهاية واحدة و384 موفّرًا. يواصل الرجوع التلقائي التوجيه ما دام هناك هدف آخر سليم متاح. ست ركائز: رجوع مرن عبر 384 موفّرًا · توفير يصل إلى 95% من الرموز في أعباء العمل المؤهلة · البدء بتكلفة 0 دولار مع أكثر من 150 مستوى مجانيًا و52 موفّرًا مجانيًا دائمًا متكررًا/بلا مفاتيح · 39 تكاملًا مع CLI/الوكلاء من خلال إعداد واحد · التوافق مع OpenAI وClaude وGemini وResponses API عبر /v1 · عناصر تحكم للإنتاج تشمل قواطع الدارات، وتخفّي TLS، و110 أدوات MCP، وA2A، والذاكرة، والضوابط الوقائية، والتقييمات، وأكثر من 56,000 تصريح اختبار ثابت عبر أكثر من 7,800 ملف اختبار متعقَّب.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -ما الذي يميز OmniRoute — لقطة مؤرخة للميزات مقارنةً بـ 9router وOpenRouter وCLIProxyAPI وLiteLLM عبر 13 قدرة. يضم OmniRoute: ‏383 موفرًا، وأكثر من 150 مستوى مجانيًا مدمجًا، و20 استراتيجية توجيه، وضغطًا للرموز عبر 12 محركًا، وخادم MCP مدمجًا مع 110 أدوات، وبروتوكول وكلاء A2A، وذاكرة دائمة، وضوابط حماية، ووكلاء سحابيين، وتخفّيًا لبصمة TLS، وDesktop/Termux/PWA، و67 لغة محلية لواجهة المستخدم. OmniRoute مرخّص بموجب MIT وقابل للاستضافة الذاتية. قد تتغير إمكانات المنافسين وأعدادهم؛ راجع المنهجية المرتبطة. +ما الذي يميز OmniRoute — لقطة مؤرخة للميزات مقارنةً بـ 9router وOpenRouter وCLIProxyAPI وLiteLLM عبر 13 قدرة. يضم OmniRoute: ‏384 موفرًا، وأكثر من 150 مستوى مجانيًا مدمجًا، و20 استراتيجية توجيه، وضغطًا للرموز عبر 12 محركًا، وخادم MCP مدمجًا مع 110 أدوات، وبروتوكول وكلاء A2A، وذاكرة دائمة، وضوابط حماية، ووكلاء سحابيين، وتخفّيًا لبصمة TLS، وDesktop/Termux/PWA، و67 لغة محلية لواجهة المستخدم. OmniRoute مرخّص بموجب MIT وقابل للاستضافة الذاتية. قد تتغير إمكانات المنافسين وأعدادهم؛ راجع المنهجية المرتبطة. 📊 المنهجية الكاملة والتفاصيل حسب كل ميزة مقارنةً بـ 9router وOpenRouter وCLIProxyAPI وLiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar اختياري ويستخدم طلبات GET فقط. لا يرفع عمي - **🖼️ نقاط نهاية جديدة** — يكمل `/v1/ocr` (Mistral OCR) و`/v1/audio/translations` (بنمط Whisper) نطاق الوسائط. → [مرجع API](docs/reference/API_REFERENCE.md) - **🎨 توليد الصور / الفيديو / الصوت** — واجهة API واحدة للوسائط: xAI Grok Imagine وNovita AI للفيديو، وComfyUI، وMagnific، وAdobe Firefly، وSegmind، ومزوّدو الكلام مثل ElevenLabs. → [مرجع API](docs/reference/API_REFERENCE.md) - **🌍 النشر والعمليات** — `basePath` للوكيل العكسي، والكشف التلقائي عن لغة المتصفح، وتتبع الأجهزة لكل مفتاح، وثقة MITM دون صلاحيات الجذر، وتوطين zh-TW. → [البيئة](docs/reference/ENVIRONMENT.md) -- **🤝 مزوّدون ووكلاء إضافيون** — وكلاء سحابيون (Codex Cloud وCursor وDevin وJules)، وGrok Build (xAI) مع تسجيل الدخول عبر المتصفح وOAuth، وبطاقة Ollama من الدرجة الأولى، وClaude Opus 5 وSonnet 5، وشراكة Kimi الرسمية (Code/Web/Moonshot)، وZed وRequesty وSenseNova وYuanbao وAgnes AI… و**فهرس محدّث يضم 383 مزوّدًا**. → [المزوّدون](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 مزوّدون ووكلاء إضافيون** — وكلاء سحابيون (Codex Cloud وCursor وDevin وJules)، وGrok Build (xAI) مع تسجيل الدخول عبر المتصفح وOAuth، وبطاقة Ollama من الدرجة الأولى، وClaude Opus 5 وSonnet 5، وشراكة Kimi الرسمية (Code/Web/Moonshot)، وZed وRequesty وSenseNova وYuanbao وAgnes AI… و**فهرس محدّث يضم 384 مزوّدًا**. → [المزوّدون](docs/reference/PROVIDER_REFERENCE.md) - **📡 شفافية التوجيه** — تحمل كل استجابة ترويسة `X-OmniRoute-Decision` التي تسمّي الاستراتيجية/المزوّد/زمن الاستجابة الذي خدمها، وتعيد استراتيجية الدمج الجديدة `cache-optimized` + عامل `cacheAffinity` في Auto-Combo توجيه الطلبات المتكررة إلى الاتصال الذي يحتفظ بالبادئة المخزنة مؤقتًا، كما تكشف نقطة النهاية للقراءة فقط `/v1/auto-combo/{channel}/candidates` عن مجموعة المرشحين الحية لقناة `auto/*`. → [الدمج التلقائي](docs/routing/AUTO-COMBO.md) - **⚡ الأداء المحلي والبنية التحتية** — خدمات Redis محلية بنقرة واحدة، وأدوات نشر مرحّل Cloudflare Workers / Deno Deploy، وBifrost وMux كخدمات مضمّنة خاضعة للإشراف. → [الخدمات المضمّنة](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 متضمّن أيضًا** — إطار عمل الإضافات + سوقها، وإطارا مهارات Omni/Agent/GitHub، وتكامل خزنة Obsidian (22 أداة MCP)، وواجهتا Batch وFiles المتوافقتان مع OpenAI، وذاكرة تخزين مؤقت دلالية للاستجابات، وعناصر تحفيزية مع لوحات المتصدرين، واكتشاف وكلاء ACP (15 وكيلًا مضمّنًا)، وتصدير مجدول للسجلات إلى BigQuery، وتوزيع متوازٍ متعدد النماذج `auto/chaos`، وجسر روبوت Telegram، ومدير إصدارات داخل التطبيق، وتصنيفات مزوّدي LMArena-ELO المجانيين. → [الوثائق](docs/README.md) @@ -689,11 +689,11 @@ omniroute configure codex # أيضًا: claude opencode qwen aider goos
-## 🌐 383 مزوّدًا للذكاء الاصطناعي — 153 منها معلّمة في الفهرس كخدمات مجانية +## 🌐 384 مزوّدًا للذكاء الاصطناعي — 153 منها معلّمة في الفهرس كخدمات مجانية
-> **383 مزوّدًا مسجّلًا** ضمن مجموعات الدردشة والوسائط والبحث والمحلية ووكلاء السحابة والنظام الأساسية، بما في ذلك **153 مزوّدًا يحمل بيانات وصفية للاكتشاف `hasFree: true`**. يغطي سجل نماذج الدردشة **239 مزوّدًا / 3,059 زوجًا مميزًا من المزوّد-النموذج / 1,475 معرّف نموذج خام**؛ بينما يحتوي فهرس الميزانية المجانية المنفصل على **471 صفًا لكل نموذج**، و**34 مجموعة متكررة**، و**52 مزوّدًا مجانيًا إلى الأبد ومتكررًا/دون مفتاح**. هذه مقامات مختلفة عمدًا؛ وتوجد التعريفات والحسابات بعد إزالة تكرار المجموعات في [مرجع المزوّدين](docs/reference/PROVIDER_REFERENCE.md) و[الخطط المجانية](docs/reference/FREE_TIERS.md). +> **384 مزوّدًا مسجّلًا** ضمن مجموعات الدردشة والوسائط والبحث والمحلية ووكلاء السحابة والنظام الأساسية، بما في ذلك **153 مزوّدًا يحمل بيانات وصفية للاكتشاف `hasFree: true`**. يغطي سجل نماذج الدردشة **239 مزوّدًا / 3,059 زوجًا مميزًا من المزوّد-النموذج / 1,475 معرّف نموذج خام**؛ بينما يحتوي فهرس الميزانية المجانية المنفصل على **471 صفًا لكل نموذج**، و**34 مجموعة متكررة**، و**52 مزوّدًا مجانيًا إلى الأبد ومتكررًا/دون مفتاح**. هذه مقامات مختلفة عمدًا؛ وتوجد التعريفات والحسابات بعد إزالة تكرار المجموعات في [مرجع المزوّدين](docs/reference/PROVIDER_REFERENCE.md) و[الخطط المجانية](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # يؤدي CI=1 أيضً بيئة التشغيلNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 اللغةTypeScript 6.0 — TypeScript بنسبة 100% ضمن src/ وopen-sse/ (لا يوجد any في النواة منذ v2.0) إطار العملNext.js 16 + React 19 + Tailwind CSS 4 - قاعدة البياناتbetter-sqlite3 (SQLite، تسجيل WAL) + LowDB (JSON قديم) — 147 وحدة نطاق، و230 عملية ترحيل + قاعدة البياناتbetter-sqlite3 (SQLite، تسجيل WAL) + LowDB (JSON قديم) — 147 وحدة نطاق، و232 عملية ترحيل الذاكرةبحث SQLite FTS5 في النص الكامل + تضمينات متجهية مكمّمة إلى int8، مع اضمحلال مُنَمَّط المخططاتZod 4 — التحقق من إدخال/إخراج أدوات MCP + عقود API البروتوكولاتMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ar/llm.txt b/docs/i18n/ar/llm.txt index 59fe274b4d5..46ba72f67dc 100644 --- a/docs/i18n/ar/llm.txt +++ b/docs/i18n/ar/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/az/README.md b/docs/i18n/az/README.md index 410cee49e90..0fa7dcd1332 100644 --- a/docs/i18n/az/README.md +++ b/docs/i18n/az/README.md @@ -15,7 +15,7 @@ **[Diego Souza](https://github.com/diegosouzapw) tərəfindən yaradılıb** -OmniRoute — Kod yazmağı heç vaxt dayandırmayın. Hər AI aləti → 383 provayder — 150+ pulsuz — bir endpoint vasitəsilə. Claude Code, Codex, Cursor, Cline, Copilot və Antigravity → PULSUZ Claude / GPT / Gemini ilə avtomatik ehtiyat keçidi. RTK + Caveman yığılmış sıxışdırması 15–95% tokenə (~89% orta hesabla) qənaət edir — limitlərə heç vaxt çatmayın. 383 AI provayderi · 150+ pulsuz tarif · ayda ~1.62B pulsuz token · 20 marşrutlaşdırma strategiyası · Başlamaq üçün $0. +OmniRoute — Kod yazmağı heç vaxt dayandırmayın. Hər AI aləti → 384 provayder — 150+ pulsuz — bir endpoint vasitəsilə. Claude Code, Codex, Cursor, Cline, Copilot və Antigravity → PULSUZ Claude / GPT / Gemini ilə avtomatik ehtiyat keçidi. RTK + Caveman yığılmış sıxışdırması 15–95% tokenə (~89% orta hesabla) qənaət edir — limitlərə heç vaxt çatmayın. 384 AI provayderi · 150+ pulsuz tarif · ayda ~1.62B pulsuz token · 20 marşrutlaşdırma strategiyası · Başlamaq üçün $0.
@@ -73,7 +73,7 @@ OMNIROUTE sizə pula qənaət etməyə və işinizi asanlaşdırmağa kömək ed | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------------- | :------------------------: | :---------: | :--------------------: | -| 🌐 Təchizatçılar | 352 | **358** | 383 | +| 🌐 Təchizatçılar | 352 | **358** | 384 | | 🧠 Unikal çat model ID-ləri | 1320 | **1374** | 1491 | | 🖼️ Modallıq körpüsü | 🆕 görüntü + audio + video | ✓ | ✓ | | 📡 Radar pulsuz kataloqu | 🆕 qoşulma seçimi | ✓ | ✓ | @@ -114,7 +114,7 @@ OMNIROUTE sizə pula qənaət etməyə və işinizi asanlaşdırmağa kömək ed ⚙️ Xüsusiyyətlər 🎯 Kombinasiyalar - 🌐 Provayderlər + 🌐 Provayderlər 🔌 CLI və MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Vəd — Bir endpoint və 383 provayder. Avtomatik ehtiyat keçidi başqa bir sağlam hədəf mövcud olduğu müddətdə yönləndirməni davam etdirir. Altı sütun: 383 provayder üzrə etibarlı ehtiyat keçidi · uyğun iş yüklərində tokenlərə 95%-ə qədər qənaət · 150-dən çox pulsuz tarif və 52 təkrarlanan/açarsız, həmişəlik pulsuz provayderlə başlamaq üçün $0 · bir konfiqurasiya vasitəsilə 39 CLI/agent inteqrasiyası · /v1 ünvanında OpenAI, Claude, Gemini və Responses API uyğunluğu · circuit breaker-lər, TLS stealth, 110 MCP aləti, A2A, yaddaş, guardrails, evals və 7.800-dən çox izlənən test faylında 56.000-dən çox statik test deklarasiyası daxil olmaqla istehsal idarəetmə imkanları. +Vəd — Bir endpoint və 384 provayder. Avtomatik ehtiyat keçidi başqa bir sağlam hədəf mövcud olduğu müddətdə yönləndirməni davam etdirir. Altı sütun: 384 provayder üzrə etibarlı ehtiyat keçidi · uyğun iş yüklərində tokenlərə 95%-ə qədər qənaət · 150-dən çox pulsuz tarif və 52 təkrarlanan/açarsız, həmişəlik pulsuz provayderlə başlamaq üçün $0 · bir konfiqurasiya vasitəsilə 39 CLI/agent inteqrasiyası · /v1 ünvanında OpenAI, Claude, Gemini və Responses API uyğunluğu · circuit breaker-lər, TLS stealth, 110 MCP aləti, A2A, yaddaş, guardrails, evals və 7.800-dən çox izlənən test faylında 56.000-dən çox statik test deklarasiyası daxil olmaqla istehsal idarəetmə imkanları.

@@ -510,7 +510,7 @@ Bütün **20** strategiya — hər kombo addımında qarışdırıb uyğunlaşd -OmniRoute-u fərqləndirən cəhətlər — 13 imkan üzrə 9router, OpenRouter, CLIProxyAPI və LiteLLM ilə tarix qeydli xüsusiyyət müqayisəsi. OmniRoute: 383 provayder, daxili 150-dən çox pulsuz tarif, 20 marşrutlaşdırma strategiyası, 12 mühərrikli token sıxışdırması, 110 alətə malik daxili MCP serveri, A2A agent protokolu, daimi yaddaş, qoruyucu məhdudiyyətlər, bulud agentləri, TLS fingerprint gizliliyi, Desktop/Termux/PWA və 67 i18n UI lokalizasiyası. OmniRoute MIT lisenziyalıdır və öz serverinizdə yerləşdirilə bilər. Rəqib məhsulların imkanları və sayları dəyişə bilər; keçid verilmiş metodologiyaya baxın. +OmniRoute-u fərqləndirən cəhətlər — 13 imkan üzrə 9router, OpenRouter, CLIProxyAPI və LiteLLM ilə tarix qeydli xüsusiyyət müqayisəsi. OmniRoute: 384 provayder, daxili 150-dən çox pulsuz tarif, 20 marşrutlaşdırma strategiyası, 12 mühərrikli token sıxışdırması, 110 alətə malik daxili MCP serveri, A2A agent protokolu, daimi yaddaş, qoruyucu məhdudiyyətlər, bulud agentləri, TLS fingerprint gizliliyi, Desktop/Termux/PWA və 67 i18n UI lokalizasiyası. OmniRoute MIT lisenziyalıdır və öz serverinizdə yerləşdirilə bilər. Rəqib məhsulların imkanları və sayları dəyişə bilər; keçid verilmiş metodologiyaya baxın. 📊 9router, OpenRouter, CLIProxyAPI və LiteLLM ilə tam metodologiya və hər xüsusiyyət üzrə təfərrüatlar → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar-a qoşulmaq ixtiyaridir və o, yalnız GET sorğularından istifadə edir. - **🖼️ Yeni endpoint-lər** — `/v1/ocr` (Mistral OCR) və `/v1/audio/translations` (Whisper üslubunda) media imkanlarını tamamlayır. → [API arayışı](docs/reference/API_REFERENCE.md) - **🎨 Şəkil / video / audio yaradılması** — media üçün vahid API: xAI Grok Imagine və Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind və ElevenLabs kimi nitq provayderləri. → [API arayışı](docs/reference/API_REFERENCE.md) - **🌍 Yerləşdirmə və əməliyyatlar** — əks proksi `basePath`, brauzer dilinin avtomatik aşkarlanması, açar üzrə cihaz izlənməsi, root-suz MITM etibarı, zh-TW lokallaşdırması. → [Mühit](docs/reference/ENVIRONMENT.md) -- **🤝 Daha çox provayder və agent** — bulud agentləri (Codex Cloud, Cursor, Devin, Jules), brauzer + OAuth girişi olan Grok Build (xAI), birinci dərəcəli Ollama kartı, Claude Opus 5 və Sonnet 5, Kimi ilə rəsmi tərəfdaşlıq (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… və yenilənmiş **383 provayderli kataloq**. → [Provayderlər](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Daha çox provayder və agent** — bulud agentləri (Codex Cloud, Cursor, Devin, Jules), brauzer + OAuth girişi olan Grok Build (xAI), birinci dərəcəli Ollama kartı, Claude Opus 5 və Sonnet 5, Kimi ilə rəsmi tərəfdaşlıq (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… və yenilənmiş **384 provayderli kataloq**. → [Provayderlər](docs/reference/PROVIDER_REFERENCE.md) - **📡 Marşrutlaşdırma şəffaflığı** — hər cavab onu təqdim edən strategiyanın/provayderin/gecikmənin adını göstərən `X-OmniRoute-Decision` başlığını daşıyır; yeni `cache-optimized` kombinasiya strategiyası və Auto-Combo `cacheAffinity` faktoru təkrarlanan sorğuları keşlənmiş prefiksi saxlayan bağlantıya geri yönləndirir; yalnız oxunan `/v1/auto-combo/{channel}/candidates` endpoint-i isə `auto/*` kanalının canlı namizədlər hovuzunu təqdim edir. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Lokal performans və infrastruktur** — bir kliklə lokal Redis, Cloudflare Workers / Deno Deploy relay yerləşdiriciləri, nəzarət olunan daxili xidmətlər kimi Bifrost və Mux. → [Daxili xidmətlər](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Qutuda olanlar** — plagin çərçivəsi və marketplace, Omni/Agent/GitHub bacarıq çərçivələri, Obsidian vault inteqrasiyası (22 MCP aləti), OpenAI-uyğun Batch və Files API-ləri, semantik cavab keşləmə, lider cədvəlləri ilə qamifikasiya, ACP agent kəşfi (15 daxili agent), BigQuery-yə planlaşdırılmış jurnal ixracı, `auto/chaos` paralel çoxmodel­li fan-out, Telegram bot körpüsü, tətbiqdaxili versiya meneceri və LMArena-ELO pulsuz provayder reytinqləri. → [Sənədlər](docs/README.md) @@ -689,11 +689,11 @@ shell tarixçənizdən kənarda saxlayır. → [CLI İnteqrasiyaları](docs/guid
-## 🌐 383 AI provayderi — kataloqda qeyd olunan 153 pulsuz provayder +## 🌐 384 AI provayderi — kataloqda qeyd olunan 153 pulsuz provayder
-> **383 qeydiyyatdan keçmiş provayder** kanonik çat, media, axtarış, lokal, bulud-agent və sistem kolleksiyaları üzrə mövcuddur; bunlardan **153-ü `hasFree: true` kəşf metadatasını daşıyır**. Çat model reyestrində **239 provayder / 3,059 fərqli provayder-model cütü / 1,475 xam model ID-si** əhatə olunur; ayrıca pulsuz büdcə kataloqunda **471 model üzrə sətir**, **34 təkrarlanan hovuz** və **52 təkrarlanan/açar tələb etməyən həmişəlik pulsuz provayder** var. Bunlar məqsədli şəkildə fərqli məxrəclərdir; təriflər və hovuzların dublikatları çıxıldıqdan sonra aparılan hesablamalar [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) və [Free Tiers](docs/reference/FREE_TIERS.md) bölmələrində təqdim olunur. +> **384 qeydiyyatdan keçmiş provayder** kanonik çat, media, axtarış, lokal, bulud-agent və sistem kolleksiyaları üzrə mövcuddur; bunlardan **153-ü `hasFree: true` kəşf metadatasını daşıyır**. Çat model reyestrində **239 provayder / 3,059 fərqli provayder-model cütü / 1,475 xam model ID-si** əhatə olunur; ayrıca pulsuz büdcə kataloqunda **471 model üzrə sətir**, **34 təkrarlanan hovuz** və **52 təkrarlanan/açar tələb etməyən həmişəlik pulsuz provayder** var. Bunlar məqsədli şəkildə fərqli məxrəclərdir; təriflər və hovuzların dublikatları çıxıldıqdan sonra aparılan hesablamalar [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) və [Free Tiers](docs/reference/FREE_TIERS.md) bölmələrində təqdim olunur.
@@ -1295,7 +1295,7 @@ bir portda təqdim olunur, buna görə hazırda ayrıca yalnız CLI üçün pake İşləmə mühitiNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 DilTypeScript 6.0 — src/ və open-sse/ üzrə 100% TypeScript (v2.0-dan etibarən nüvədə sıfır any) FreymvorkNext.js 16 + React 19 + Tailwind CSS 4 - Verilənlər bazasıbetter-sqlite3 (SQLite, WAL jurnallaşdırması) + LowDB (JSON legacy) — 147 domen modulu, 230 miqrasiya + Verilənlər bazasıbetter-sqlite3 (SQLite, WAL jurnallaşdırması) + LowDB (JSON legacy) — 147 domen modulu, 232 miqrasiya YaddaşSQLite FTS5 tam mətn axtarışı + int8-kvantlaşdırılmış vektor daxiletmələri, tipləndirilmiş zəifləmə SxemlərZod 4 — MCP alətləri üçün I/O yoxlaması + API müqavilələri ProtokollarMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/az/llm.txt b/docs/i18n/az/llm.txt index f07c0fdd355..ac45dbb1c56 100644 --- a/docs/i18n/az/llm.txt +++ b/docs/i18n/az/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/bg/README.md b/docs/i18n/bg/README.md index 093e3ed5bc2..86ae0f7e924 100644 --- a/docs/i18n/bg/README.md +++ b/docs/i18n/bg/README.md @@ -15,7 +15,7 @@ **Създадено от [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Никога не спирайте да пишете код. Всеки AI инструмент → 383 доставчици — 150+ безплатни — през една крайна точка. Claude Code, Codex, Cursor, Cline, Copilot и Antigravity към БЕЗПЛАТНИ Claude / GPT / Gemini с автоматичен резервен вариант. Компресията RTK + Caveman на няколко нива спестява 15–95% от токените (~89% средно) — никога не достигайте лимитите. 383 AI доставчици · 150+ безплатни нива · ~1,62 млрд. безплатни токена/месец · 20 стратегии за маршрутизиране · $0 за начало. +OmniRoute — Никога не спирайте да пишете код. Всеки AI инструмент → 384 доставчици — 150+ безплатни — през една крайна точка. Claude Code, Codex, Cursor, Cline, Copilot и Antigravity към БЕЗПЛАТНИ Claude / GPT / Gemini с автоматичен резервен вариант. Компресията RTK + Caveman на няколко нива спестява 15–95% от токените (~89% средно) — никога не достигайте лимитите. 384 AI доставчици · 150+ безплатни нива · ~1,62 млрд. безплатни токена/месец · 20 стратегии за маршрутизиране · $0 за начало.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ---------------------------------------- | :-------------------------: | :---------: | :-------------------: | -| 🌐 Доставчици | 352 | **358** | 383 | +| 🌐 Доставчици | 352 | **358** | 384 | | 🧠 Уникални идентификатори на чат модели | 1320 | **1374** | 1491 | | 🖼️ Мост на модалностите | 🆕 визуално + аудио + видео | ✓ | ✓ | | 📡 Безплатен каталог на Radar | 🆕 с включване по избор | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Функции 🎯 Комбинации - 🌐 Доставчици + 🌐 Доставчици 🔌 CLI и MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Обещанието — една крайна точка и 383 доставчици. Автоматичното резервно превключване поддържа маршрутизирането, докато има друга работеща цел. Шест стълба: устойчиво резервно превключване между 383 доставчици · до 95% спестяване на токени при допустими работни натоварвания · $0 за начало със 150+ безплатни плана и 52 повтарящи се/безключови доставчици, безплатни завинаги · 39 CLI/агентски интеграции чрез една конфигурация · съвместимост с OpenAI, Claude, Gemini и Responses API на /v1 · средства за управление в продукционна среда, включително прекъсвачи на веригата, TLS stealth, MCP с 110 инструмента, A2A, памет, защитни ограничения, оценки и 56 000+ декларации за статични тестове в 7 800+ проследявани тестови файла. +Обещанието — една крайна точка и 384 доставчици. Автоматичното резервно превключване поддържа маршрутизирането, докато има друга работеща цел. Шест стълба: устойчиво резервно превключване между 384 доставчици · до 95% спестяване на токени при допустими работни натоварвания · $0 за начало със 150+ безплатни плана и 52 повтарящи се/безключови доставчици, безплатни завинаги · 39 CLI/агентски интеграции чрез една конфигурация · съвместимост с OpenAI, Claude, Gemini и Responses API на /v1 · средства за управление в продукционна среда, включително прекъсвачи на веригата, TLS stealth, MCP с 110 инструмента, A2A, памет, защитни ограничения, оценки и 56 000+ декларации за статични тестове в 7 800+ проследявани тестови файла.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -Какво отличава OmniRoute — актуална към определена дата снимка на функциите в сравнение с 9router, OpenRouter, CLIProxyAPI и LiteLLM по 13 възможности. OmniRoute: 383 доставчици, над 150 вградени безплатни нива, 20 стратегии за маршрутизиране, компресиране на токени с 12 машини, вграден MCP сървър със 110 инструмента, A2A протокол за агенти, постоянно запаметяване, защитни механизми, облачни агенти, скрит режим с TLS fingerprint, Desktop/Termux/PWA и 67 локализации на интерфейса с i18n. OmniRoute е лицензиран по MIT и може да се хоства самостоятелно. Възможностите и броят функции на конкурентите могат да се променят; вижте методологията на посочената връзка. +Какво отличава OmniRoute — актуална към определена дата снимка на функциите в сравнение с 9router, OpenRouter, CLIProxyAPI и LiteLLM по 13 възможности. OmniRoute: 384 доставчици, над 150 вградени безплатни нива, 20 стратегии за маршрутизиране, компресиране на токени с 12 машини, вграден MCP сървър със 110 инструмента, A2A протокол за агенти, постоянно запаметяване, защитни механизми, облачни агенти, скрит режим с TLS fingerprint, Desktop/Termux/PWA и 67 локализации на интерфейса с i18n. OmniRoute е лицензиран по MIT и може да се хоства самостоятелно. Възможностите и броят функции на конкурентите могат да се променят; вижте методологията на посочената връзка. 📊 Пълна методология и подробности за всяка функция в сравнение с 9router, OpenRouter, CLIProxyAPI и LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -613,7 +613,7 @@ Radar се използва само по желание и поддържа е - **🖼️ Нови крайни точки** — `/v1/ocr` (Mistral OCR) и `/v1/audio/translations` (в стил Whisper) допълват медийната функционалност. → [Справочник на API](docs/reference/API_REFERENCE.md) - **🎨 Генериране на изображения / видео / аудио** — един API за медии: xAI Grok Imagine и видео от Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind и доставчици на гласови услуги като ElevenLabs. → [Справочник на API](docs/reference/API_REFERENCE.md) - **🌍 Разгръщане и операции** — `basePath` за reverse proxy, автоматично разпознаване на езика на браузъра, проследяване на устройствата за всеки ключ, доверие в MITM без root права и локализация на zh-TW. → [Среда](docs/reference/ENVIRONMENT.md) -- **🤝 Повече доставчици и агенти** — облачни агенти (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) с вход чрез браузър + OAuth, пълноценна карта за Ollama, Claude Opus 5 и Sonnet 5, официално партньорство с Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… и обновен **каталог с 383 доставчици**. → [Доставчици](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Повече доставчици и агенти** — облачни агенти (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) с вход чрез браузър + OAuth, пълноценна карта за Ollama, Claude Opus 5 и Sonnet 5, официално партньорство с Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… и обновен **каталог с 384 доставчици**. → [Доставчици](docs/reference/PROVIDER_REFERENCE.md) - **📡 Прозрачност на маршрутизирането** — всеки отговор съдържа заглавка `X-OmniRoute-Decision`, която посочва стратегията/доставчика/латентността, обслужили заявката; новата стратегия за комбинации `cache-optimized` + факторът `cacheAffinity` на Auto-Combo насочва повтарящите се заявки обратно към връзката, която притежава кеширания префикс; а достъпната само за четене крайна точка `/v1/auto-combo/{channel}/candidates` показва актуалния пул от кандидати на канал `auto/*`. → [Автоматични комбинации](docs/routing/AUTO-COMBO.md). - **⚡ Локална производителност и инфраструктура** — локален Redis с едно щракване, инструменти за внедряване на relay към Cloudflare Workers / Deno Deploy, Bifrost и Mux като управлявани вградени услуги. → [Вградени услуги](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Също включено** — framework за плъгини и marketplace, framework-и за умения Omni/Agent/GitHub, интеграция с хранилища на Obsidian (22 MCP инструмента), съвместими с OpenAI Batch и Files API, семантичен кеш за отговори, геймификация с класации, откриване на ACP агенти (15 вградени агента), планиран износ на логове към BigQuery, паралелно разпределяне към множество модели чрез `auto/chaos`, мост към Telegram бот, вграден мениджър на версии и класации на безплатни доставчици по LMArena-ELO. → [Документация](docs/README.md) @@ -697,11 +697,11 @@ omniroute configure codex # също: claude opencode qwen aider goose
-## 🌐 383 доставчици на AI — 153 маркирани в каталога като безплатни +## 🌐 384 доставчици на AI — 153 маркирани в каталога като безплатни
-> **383 регистрирани доставчици** в каноничните колекции за чат, медии, търсене, локални услуги, облачни агенти и системи, включително **153 с метаданни за откриване `hasFree: true`**. Регистърът на чат моделите обхваща **239 доставчици / 3 059 различни двойки доставчик-модел / 1 475 необработени идентификатора на модели**; отделният каталог за безплатен бюджет съдържа **471 реда за отделни модели**, **34 повтарящи се пула** и **52 повтарящи се/без ключ безплатни доставчици завинаги**. Това са различни знаменатели по замисъл; дефинициите и изчисленията с дедупликация на пуловете са в [Справочник за доставчиците](docs/reference/PROVIDER_REFERENCE.md) и [Безплатни нива](docs/reference/FREE_TIERS.md). +> **384 регистрирани доставчици** в каноничните колекции за чат, медии, търсене, локални услуги, облачни агенти и системи, включително **153 с метаданни за откриване `hasFree: true`**. Регистърът на чат моделите обхваща **239 доставчици / 3 059 различни двойки доставчик-модел / 1 475 необработени идентификатора на модели**; отделният каталог за безплатен бюджет съдържа **471 реда за отделни модели**, **34 повтарящи се пула** и **52 повтарящи се/без ключ безплатни доставчици завинаги**. Това са различни знаменатели по замисъл; дефинициите и изчисленията с дедупликация на пуловете са в [Справочник за доставчиците](docs/reference/PROVIDER_REFERENCE.md) и [Безплатни нива](docs/reference/FREE_TIERS.md).
@@ -1303,7 +1303,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 също го п Среда за изпълнениеNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ЕзикTypeScript 6.0 — 100% TypeScript в src/ и open-sse/ (нула any в ядрото от v2.0 насам) ФреймуъркNext.js 16 + React 19 + Tailwind CSS 4 - База данниbetter-sqlite3 (SQLite, WAL журналиране) + LowDB (JSON наследена система) — 147 домейн модула, 230 миграции + База данниbetter-sqlite3 (SQLite, WAL журналиране) + LowDB (JSON наследена система) — 147 домейн модула, 232 миграции ПаметПълнотекстово търсене с SQLite FTS5 + векторни embeddings с int8 квантизация, типизирано затихване СхемиZod 4 — валидиране на I/O на MCP инструментите + API договори ПротоколиMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/bg/llm.txt b/docs/i18n/bg/llm.txt index 63b03c21fc1..c3f77c110c9 100644 --- a/docs/i18n/bg/llm.txt +++ b/docs/i18n/bg/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/bn/README.md b/docs/i18n/bn/README.md index 0037966b9ae..97bb9c9f2b5 100644 --- a/docs/i18n/bn/README.md +++ b/docs/i18n/bn/README.md @@ -15,7 +15,7 @@ **তৈরি করেছেন [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — কখনও কোড লেখা থামাবেন না। প্রতিটি AI টুল → 383টি প্রোভাইডার — 150টিরও বেশি বিনামূল্যে — একটি এন্ডপয়েন্টের মাধ্যমে। Claude Code, Codex, Cursor, Cline, Copilot ও Antigravity এখন স্বয়ংক্রিয় ফ্যালব্যাকসহ বিনামূল্যের Claude / GPT / Gemini-তে। RTK + Caveman স্তরভিত্তিক কম্প্রেশন 15–95% টোকেন সাশ্রয় করে (~89% গড়) — কখনও সীমায় পৌঁছাবেন না। 383টি AI প্রোভাইডার · 150টিরও বেশি বিনামূল্যের টিয়ার · মাসে প্রায় 1.62B বিনামূল্যের টোকেন · 20টি রাউটিং কৌশল · শুরুতে $0। +OmniRoute — কখনও কোড লেখা থামাবেন না। প্রতিটি AI টুল → 384টি প্রোভাইডার — 150টিরও বেশি বিনামূল্যে — একটি এন্ডপয়েন্টের মাধ্যমে। Claude Code, Codex, Cursor, Cline, Copilot ও Antigravity এখন স্বয়ংক্রিয় ফ্যালব্যাকসহ বিনামূল্যের Claude / GPT / Gemini-তে। RTK + Caveman স্তরভিত্তিক কম্প্রেশন 15–95% টোকেন সাশ্রয় করে (~89% গড়) — কখনও সীমায় পৌঁছাবেন না। 384টি AI প্রোভাইডার · 150টিরও বেশি বিনামূল্যের টিয়ার · মাসে প্রায় 1.62B বিনামূল্যের টোকেন · 20টি রাউটিং কৌশল · শুরুতে $0।
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------ | :--------------------: | :---------: | :-------------------: | -| 🌐 প্রোভাইডার | 352 | **358** | 383 | +| 🌐 প্রোভাইডার | 352 | **358** | 384 | | 🧠 অনন্য চ্যাট মডেল আইডি | 1320 | **1374** | 1491 | | 🖼️ মোডালিটি ব্রিজ | 🆕 ভিশন + অডিও + ভিডিও | ✓ | ✓ | | 📡 Radar ফ্রি ক্যাটালগ | 🆕 অপ্ট-ইন | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ বৈশিষ্ট্য 🎯 কম্বো - 🌐 প্রদানকারী + 🌐 প্রদানকারী 🔌 CLI ওamp; MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -প্রতিশ্রুতি — একটি এন্ডপয়েন্ট এবং 383টি প্রদানকারী। আরেকটি সুস্থ লক্ষ্য উপলভ্য থাকা পর্যন্ত স্বয়ংক্রিয় ফলব্যাক রাউটিং সচল রাখে। ছয়টি স্তম্ভ: 383টি প্রদানকারী জুড়ে স্থিতিস্থাপক ফলব্যাক · যোগ্য ওয়ার্কলোডে সর্বোচ্চ 95% টোকেন সাশ্রয় · 150+টি ফ্রি টিয়ার এবং 52টি পুনরাবৃত্ত/কী-বিহীন চিরস্থায়ী বিনামূল্যের প্রদানকারীসহ $0 দিয়ে শুরু · একটি কনফিগারেশনের মাধ্যমে 39টি CLI/এজেন্ট ইন্টিগ্রেশন · /v1-এ OpenAI, Claude, Gemini এবং Responses API সামঞ্জস্য · সার্কিট ব্রেকার, TLS স্টেলথ, MCP-এর 110টি টুল, A2A, মেমরি, গার্ডরেল, ইভ্যাল, এবং 7,800+ ট্র্যাক করা টেস্ট ফাইলজুড়ে 56,000+ স্ট্যাটিক টেস্ট ডিক্লারেশনসহ প্রোডাকশন নিয়ন্ত্রণ। +প্রতিশ্রুতি — একটি এন্ডপয়েন্ট এবং 384টি প্রদানকারী। আরেকটি সুস্থ লক্ষ্য উপলভ্য থাকা পর্যন্ত স্বয়ংক্রিয় ফলব্যাক রাউটিং সচল রাখে। ছয়টি স্তম্ভ: 384টি প্রদানকারী জুড়ে স্থিতিস্থাপক ফলব্যাক · যোগ্য ওয়ার্কলোডে সর্বোচ্চ 95% টোকেন সাশ্রয় · 150+টি ফ্রি টিয়ার এবং 52টি পুনরাবৃত্ত/কী-বিহীন চিরস্থায়ী বিনামূল্যের প্রদানকারীসহ $0 দিয়ে শুরু · একটি কনফিগারেশনের মাধ্যমে 39টি CLI/এজেন্ট ইন্টিগ্রেশন · /v1-এ OpenAI, Claude, Gemini এবং Responses API সামঞ্জস্য · সার্কিট ব্রেকার, TLS স্টেলথ, MCP-এর 110টি টুল, A2A, মেমরি, গার্ডরেল, ইভ্যাল, এবং 7,800+ ট্র্যাক করা টেস্ট ফাইলজুড়ে 56,000+ স্ট্যাটিক টেস্ট ডিক্লারেশনসহ প্রোডাকশন নিয়ন্ত্রণ।

@@ -605,7 +605,7 @@ Radar ব্যবহার ঐচ্ছিক এবং এটি কেবল - **🖼️ নতুন endpoint** — `/v1/ocr` (Mistral OCR) এবং `/v1/audio/translations` (Whisper-style) media surface সম্পূর্ণ করে। → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 Image / video / audio generation** — media-এর জন্য একটি API: xAI Grok Imagine ও Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind এবং ElevenLabs-এর মতো speech provider। → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 Deployment ও ops** — reverse-proxy `basePath`, browser-language auto-detect, প্রতি-কী device tracking, root-less MITM trust, zh-TW localization। → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 আরও প্রোভাইডার ও এজেন্ট** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login-সহ Grok Build (xAI), Ollama first-class card, Claude Opus 5 ও Sonnet 5, Kimi official partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… এবং নতুন করে সাজানো **383-provider catalog**। → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 আরও প্রোভাইডার ও এজেন্ট** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login-সহ Grok Build (xAI), Ollama first-class card, Claude Opus 5 ও Sonnet 5, Kimi official partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… এবং নতুন করে সাজানো **384-provider catalog**। → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 রাউটিং স্বচ্ছতা** — প্রতিটি response-এ একটি `X-OmniRoute-Decision` header থাকে, যা এটি পরিবেশন করা strategy/provider/latency-এর নাম জানায়; নতুন `cache-optimized` combo strategy এবং Auto-Combo `cacheAffinity` factor পুনরাবৃত্তি হওয়া অনুরোধগুলোকে cached prefix ধারণকারী connection-এ ফেরত পাঠায়; আর একটি read-only `/v1/auto-combo/{channel}/candidates` endpoint কোনো `auto/*` channel-এর live candidate pool প্রকাশ করে। → [Auto-Combo](docs/routing/AUTO-COMBO.md)। - **⚡ স্থানীয় পারফরম্যান্স ও অবকাঠামো** — এক-ক্লিকে স্থানীয় Redis, Cloudflare Workers / Deno Deploy রিলে ডিপ্লয়মেন্টকারী, এবং তত্ত্বাবধানে চলা এমবেডেড পরিষেবা হিসেবে Bifrost ও Mux। → [এমবেডেড পরিষেবা](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 আরও যা অন্তর্ভুক্ত রয়েছে** — প্লাগইন ফ্রেমওয়ার্ক ও মার্কেটপ্লেস, Omni/Agent/GitHub স্কিলস ফ্রেমওয়ার্ক, Obsidian ভল্ট ইন্টিগ্রেশন (22টি MCP টুল), OpenAI-সামঞ্জস্যপূর্ণ Batch ও Files API, সেমান্টিক রেসপন্স ক্যাশ, লিডারবোর্ডসহ গ্যামিফিকেশন, ACP এজেন্ট আবিষ্কার (15টি বিল্ট-ইন এজেন্ট), BigQuery-তে নির্ধারিত সময়ে লগ এক্সপোর্ট, `auto/chaos` সমান্তরাল মাল্টি-মডেল ফ্যান-আউট, একটি Telegram বট ব্রিজ, ইন-অ্যাপ ভার্সন ম্যানেজার এবং LMArena-ELO বিনামূল্যের প্রোভাইডার র্যাঙ্কিং। → [ডকুমেন্টেশন](docs/README.md) @@ -688,11 +688,11 @@ omniroute configure codex # এছাড়াও: claude opencode qwen
-## 🌐 383টি AI প্রদানকারী — ক্যাটালগে চিহ্নিত 153টি বিনামূল্যে +## 🌐 384টি AI প্রদানকারী — ক্যাটালগে চিহ্নিত 153টি বিনামূল্যে
-> **383টি নিবন্ধিত প্রদানকারী** canonical chat, media, search, local, cloud-agent এবং system collection জুড়ে রয়েছে, যার মধ্যে **153টি `hasFree: true` discovery metadata বহন করে**। Chat model registry-তে রয়েছে **239টি প্রদানকারী / 3,059টি স্বতন্ত্র provider-model pair / 1,475টি raw model ID**; পৃথক free-budget catalog-এ রয়েছে **471টি per-model row**, **34টি recurring pool** এবং **52টি recurring/keyless free-forever provider**। নকশা অনুযায়ী এগুলো ভিন্ন denominator; সংজ্ঞা এবং pool-deduped হিসাব [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) ও [Free Tiers](docs/reference/FREE_TIERS.md)-এ রয়েছে। +> **384টি নিবন্ধিত প্রদানকারী** canonical chat, media, search, local, cloud-agent এবং system collection জুড়ে রয়েছে, যার মধ্যে **153টি `hasFree: true` discovery metadata বহন করে**। Chat model registry-তে রয়েছে **239টি প্রদানকারী / 3,059টি স্বতন্ত্র provider-model pair / 1,475টি raw model ID**; পৃথক free-budget catalog-এ রয়েছে **471টি per-model row**, **34টি recurring pool** এবং **52টি recurring/keyless free-forever provider**। নকশা অনুযায়ী এগুলো ভিন্ন denominator; সংজ্ঞা এবং pool-deduped হিসাব [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) ও [Free Tiers](docs/reference/FREE_TIERS.md)-এ রয়েছে।
@@ -1294,7 +1294,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1-ও এটি এ রানটাইমNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ভাষাTypeScript 6.0 — src/ এবং open-sse/ জুড়ে 100% TypeScript (v2.0 থেকে কোরে শূন্য any) ফ্রেমওয়ার্কNext.js 16 + React 19 + Tailwind CSS 4 - ডেটাবেসbetter-sqlite3 (SQLite, WAL জার্নালিং) + LowDB (JSON লিগ্যাসি) — 147টি ডোমেইন মডিউল, 230টি মাইগ্রেশন + ডেটাবেসbetter-sqlite3 (SQLite, WAL জার্নালিং) + LowDB (JSON লিগ্যাসি) — 147টি ডোমেইন মডিউল, 232টি মাইগ্রেশন মেমরিSQLite FTS5 পূর্ণ-পাঠ্য অনুসন্ধান + int8-কোয়ান্টাইজড ভেক্টর এমবেডিং, টাইপযুক্ত ক্ষয় স্কিমাZod 4 — MCP টুল I/O ভ্যালিডেশন + API কনট্র্যাক্ট প্রোটোকলMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/bn/llm.txt b/docs/i18n/bn/llm.txt index c9ea4ddd1a0..c11f4740821 100644 --- a/docs/i18n/bn/llm.txt +++ b/docs/i18n/bn/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/bs/README.md b/docs/i18n/bs/README.md index 7df36dae652..0ff5ce47d69 100644 --- a/docs/i18n/bs/README.md +++ b/docs/i18n/bs/README.md @@ -15,7 +15,7 @@ **Kreirao [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Nikada ne prestajte programirati. Svaki AI alat → 383 pružalac — više od 150 besplatnih — putem jedne krajnje tačke. Claude Code, Codex, Cursor, Cline, Copilot i Antigravity prema BESPLATNIM modelima Claude / GPT / Gemini uz automatski povratak na alternativu. RTK + Caveman složena kompresija štedi 15–95% tokena (~89% u prosjeku) — nikada nećete dosegnuti ograničenja. 383 AI pružalac · više od 150 besplatnih nivoa · ~1,62 milijarde besplatnih tokena mjesečno · 20 strategija rutiranja · $0 za početak. +OmniRoute — Nikada ne prestajte programirati. Svaki AI alat → 384 pružalac — više od 150 besplatnih — putem jedne krajnje tačke. Claude Code, Codex, Cursor, Cline, Copilot i Antigravity prema BESPLATNIM modelima Claude / GPT / Gemini uz automatski povratak na alternativu. RTK + Caveman složena kompresija štedi 15–95% tokena (~89% u prosjeku) — nikada nećete dosegnuti ograničenja. 384 AI pružalac · više od 150 besplatnih nivoa · ~1,62 milijarde besplatnih tokena mjesečno · 20 strategija rutiranja · $0 za početak.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------------- | :-------------------------: | :---------: | :-------------------: | -| 🌐 Pružatelji | 352 | **358** | 383 | +| 🌐 Pružatelji | 352 | **358** | 384 | | 🧠 Jedinstveni ID-ovi chat modela | 1320 | **1374** | 1491 | | 🖼️ Most modaliteta | 🆕 vizija + zvuk + video | ✓ | ✓ | | 📡 Besplatni katalog radara | 🆕 dobrovoljno uključivanje | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funkcije 🎯 Kombinacije - 🌐 Dobavljači + 🌐 Dobavljači 🔌 CLI i MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Obećanje — Jedna krajnja tačka i 383 dobavljač. Automatski rezervni odabir održava usmjeravanje dok je dostupno drugo zdravo odredište. Šest stubova: otporno rezervno usmjeravanje kroz 383 dobavljača · do 95% uštede tokena na podržanim radnim opterećenjima · $0 za početak uz više od 150 besplatnih nivoa i 52 dobavljača koji su besplatni zauvijek uz periodičnu obnovu ili bez ključa · 39 CLI/agentskih integracija kroz jednu konfiguraciju · kompatibilnost s OpenAI, Claude, Gemini i Responses API na /v1 · produkcijske kontrole koje uključuju prekidače strujnog kola, TLS prikrivanje, MCP sa 110 alata, A2A, memoriju, zaštitne mehanizme, evaluacije i više od 56.000 statičkih deklaracija testova kroz više od 7.800 praćenih datoteka testova. +Obećanje — Jedna krajnja tačka i 384 dobavljač. Automatski rezervni odabir održava usmjeravanje dok je dostupno drugo zdravo odredište. Šest stubova: otporno rezervno usmjeravanje kroz 384 dobavljača · do 95% uštede tokena na podržanim radnim opterećenjima · $0 za početak uz više od 150 besplatnih nivoa i 52 dobavljača koji su besplatni zauvijek uz periodičnu obnovu ili bez ključa · 39 CLI/agentskih integracija kroz jednu konfiguraciju · kompatibilnost s OpenAI, Claude, Gemini i Responses API na /v1 · produkcijske kontrole koje uključuju prekidače strujnog kola, TLS prikrivanje, MCP sa 110 alata, A2A, memoriju, zaštitne mehanizme, evaluacije i više od 56.000 statičkih deklaracija testova kroz više od 7.800 praćenih datoteka testova.

@@ -510,7 +510,7 @@ Svih **20** strategija — kombinirajte ih po želji za svaki korak kombinacije: -Po čemu se OmniRoute izdvaja — datirani pregled funkcionalnosti u odnosu na 9router, OpenRouter, CLIProxyAPI i LiteLLM kroz 13 mogućnosti. OmniRoute: 383 pružalac, više od 150 besplatnih nivoa ugrađenih, 20 strategija rutiranja, kompresija tokena pomoću 12 mašina, ugrađeni MCP server sa 110 alata, A2A protokol agenata, trajna memorija, zaštitne mjere, agenti u oblaku, prikrivanje TLS otiska, Desktop/Termux/PWA i 67 i18n lokalizacija korisničkog interfejsa. OmniRoute je licenciran pod MIT licencom i može se samostalno hostovati. Mogućnosti i brojke konkurenata mogu se promijeniti; pogledajte povezanu metodologiju. +Po čemu se OmniRoute izdvaja — datirani pregled funkcionalnosti u odnosu na 9router, OpenRouter, CLIProxyAPI i LiteLLM kroz 13 mogućnosti. OmniRoute: 384 pružalac, više od 150 besplatnih nivoa ugrađenih, 20 strategija rutiranja, kompresija tokena pomoću 12 mašina, ugrađeni MCP server sa 110 alata, A2A protokol agenata, trajna memorija, zaštitne mjere, agenti u oblaku, prikrivanje TLS otiska, Desktop/Termux/PWA i 67 i18n lokalizacija korisničkog interfejsa. OmniRoute je licenciran pod MIT licencom i može se samostalno hostovati. Mogućnosti i brojke konkurenata mogu se promijeniti; pogledajte povezanu metodologiju. 📊 Kompletna metodologija i detalji po funkcionalnostima u odnosu na 9router, OpenRouter, CLIProxyAPI i LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -613,7 +613,7 @@ trenutnom katalogu na **[radar.omniroute.online/planos](https://radar.omniroute. - **🖼️ Nove krajnje tačke** — `/v1/ocr` (Mistral OCR) i `/v1/audio/translations` (u stilu Whispera) upotpunjuju medijski interfejs. → [API referenca](docs/reference/API_REFERENCE.md) - **🎨 Generisanje slika / videa / zvuka** — jedan API za medije: xAI Grok Imagine i Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind i pružaoci govora poput ElevenLabs. → [API referenca](docs/reference/API_REFERENCE.md) - **🌍 Implementacija i operacije** — `basePath` obrnutog proxyja, automatsko prepoznavanje jezika preglednika, praćenje uređaja po ključu, MITM povjerenje bez root ovlasti, lokalizacija zh-TW. → [Okruženje](docs/reference/ENVIRONMENT.md) -- **🤝 Više pružalaca i agenata** — agenti u oblaku (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) s prijavom putem preglednika + OAuth, prvoklasna kartica za Ollama, Claude Opus 5 i Sonnet 5, službeno partnerstvo s Kimijem (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… i osvježeni **katalog s 383 pružaocem**. → [Pružaoci](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Više pružalaca i agenata** — agenti u oblaku (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) s prijavom putem preglednika + OAuth, prvoklasna kartica za Ollama, Claude Opus 5 i Sonnet 5, službeno partnerstvo s Kimijem (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… i osvježeni **katalog s 384 pružaocem**. → [Pružaoci](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparentnost usmjeravanja** — svaki odgovor sadrži zaglavlje `X-OmniRoute-Decision` koje navodi strategiju/pružaoca/latenciju koji su ga poslužili, nova strategija kombinovanja `cache-optimized` + faktor `cacheAffinity` automatskog kombinovanja usmjerava ponovljene zahtjeve nazad na konekciju koja sadrži keširani prefiks, a krajnja tačka samo za čitanje `/v1/auto-combo/{channel}/candidates` izlaže aktivni skup kandidata `auto/*` kanala. → [Automatsko kombinovanje](docs/routing/AUTO-COMBO.md) - **⚡ Lokalne performanse i infrastruktura** — lokalni Redis jednim klikom, deployeri za Cloudflare Workers / Deno Deploy relay, Bifrost i Mux kao nadzirane ugrađene usluge. → [Ugrađene usluge](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Također uključeno** — okvir za dodatke + marketplace, Omni/Agent/GitHub okviri vještina, integracija s Obsidian trezorom (22 MCP alata), OpenAI-kompatibilni Batch i Files API-ji, semantička predmemorija odgovora, gamifikacija s rang-listama, ACP otkrivanje agenata (15 ugrađenih agenata), zakazani izvoz logova u BigQuery, `auto/chaos` paralelno prosljeđivanje prema više modela, most za Telegram bot, upravljač verzijama unutar aplikacije i rang-liste besplatnih pružalaca usluga LMArena-ELO. → [Dokumentacija](docs/README.md) @@ -697,11 +697,11 @@ izvan vaše shell historije. → [CLI Integracije](docs/guides/CLI-INTEGRATIONS.
-## 🌐 383 AI provajder — 153 označena kao besplatna u katalogu +## 🌐 384 AI provajder — 153 označena kao besplatna u katalogu
-> **383 registrovani provajder** u kanonskim kolekcijama za chat, medije, pretragu, lokalne i cloud agente te sistemske kolekcije, uključujući **153 provajdera s metapodacima za otkrivanje `hasFree: true`**. Registar chat modela obuhvata **239 provajdera / 3,059 različitih parova provajder-model / 1,475 sirovih ID-jeva modela**; zasebni katalog besplatnog budžeta sadrži **471 red po modelu**, **34 periodična fonda** i **52 periodična/beskonačno besplatna provajdera bez ključa**. Ovo su različiti nazivnici po dizajnu; definicije i izračuni s uklonjenim duplikatima fondova dostupni su u dokumentima [Referenca provajdera](docs/reference/PROVIDER_REFERENCE.md) i [Besplatni nivoi](docs/reference/FREE_TIERS.md). +> **384 registrovani provajder** u kanonskim kolekcijama za chat, medije, pretragu, lokalne i cloud agente te sistemske kolekcije, uključujući **153 provajdera s metapodacima za otkrivanje `hasFree: true`**. Registar chat modela obuhvata **239 provajdera / 3,059 različitih parova provajder-model / 1,475 sirovih ID-jeva modela**; zasebni katalog besplatnog budžeta sadrži **471 red po modelu**, **34 periodična fonda** i **52 periodična/beskonačno besplatna provajdera bez ključa**. Ovo su različiti nazivnici po dizajnu; definicije i izračuni s uklonjenim duplikatima fondova dostupni su u dokumentima [Referenca provajdera](docs/reference/PROVIDER_REFERENCE.md) i [Besplatni nivoi](docs/reference/FREE_TIERS.md).
@@ -1305,7 +1305,7 @@ Kanonske metrike na dan 2026-10-18: **1.976 jedinstvenih videozapisa** · **20.2 Izvršno okruženjeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 JezikTypeScript 6.0 — 100% TypeScript kroz src/ i open-sse/ (nula any u jezgri od v2.0) OkvirNext.js 16 + React 19 + Tailwind CSS 4 - Baza podatakabetter-sqlite3 (SQLite, WAL evidentiranje) + LowDB (naslijeđeni JSON) — 147 domenskih modula, 230 migracija + Baza podatakabetter-sqlite3 (SQLite, WAL evidentiranje) + LowDB (naslijeđeni JSON) — 147 domenskih modula, 232 migracija MemorijaSQLite FTS5 pretraga punog teksta + vektorska ugrađivanja kvantizovana na int8, tipizirano opadanje ŠemeZod 4 — validacija I/O MCP alata + API ugovori ProtokoliMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/bs/llm.txt b/docs/i18n/bs/llm.txt index 9950d02acf7..514c1e80517 100644 --- a/docs/i18n/bs/llm.txt +++ b/docs/i18n/bs/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/cs/README.md b/docs/i18n/cs/README.md index 3b10ac9d65a..7c17ee98d36 100644 --- a/docs/i18n/cs/README.md +++ b/docs/i18n/cs/README.md @@ -15,7 +15,7 @@ **Vytvořil [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Nikdy nepřestávejte programovat. Každý nástroj AI → 383 poskytovatelů — 150+ bezplatných — prostřednictvím jediného endpointu. Claude Code, Codex, Cursor, Cline, Copilot a Antigravity do BEZPLATNÝCH služeb Claude / GPT / Gemini s automatickým přepnutím. Komprese RTK + Caveman ve vrstvách šetří 15–95 % tokenů (průměrně ~89 %) — nikdy nenarazíte na limity. 383 poskytovatelů AI · 150+ bezplatných tarifů · ~1,62 miliardy bezplatných tokenů/měsíc · 20 strategií směrování · začněte za 0 $. +OmniRoute — Nikdy nepřestávejte programovat. Každý nástroj AI → 384 poskytovatelů — 150+ bezplatných — prostřednictvím jediného endpointu. Claude Code, Codex, Cursor, Cline, Copilot a Antigravity do BEZPLATNÝCH služeb Claude / GPT / Gemini s automatickým přepnutím. Komprese RTK + Caveman ve vrstvách šetří 15–95 % tokenů (průměrně ~89 %) — nikdy nenarazíte na limity. 384 poskytovatelů AI · 150+ bezplatných tarifů · ~1,62 miliardy bezplatných tokenů/měsíc · 20 strategií směrování · začněte za 0 $.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ---------------------------------- | :----------------------: | :---------: | :-------------------: | -| 🌐 Poskytovatelé | 352 | **358** | 383 | +| 🌐 Poskytovatelé | 352 | **358** | 384 | | 🧠 Jedinečná ID chatovacích modelů | 1320 | **1374** | 1491 | | 🖼️ Most modalit | 🆕 vidění + zvuk + video | ✓ | ✓ | | 📡 Bezplatný katalog Radar | 🆕 opt-in | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funkce 🎯 Kombinace - 🌐 Poskytovatelé + 🌐 Poskytovatelé 🔌 CLI a MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Příslib — jeden endpoint a 383 poskytovatelů. Automatický fallback udržuje směrování, dokud je k dispozici jiný zdravý cíl. Šest pilířů: odolný fallback napříč 383 poskytovateli · až 95% úspora tokenů u způsobilých úloh · $0 na začátek se 150+ bezplatnými tarify a 52 opakovaně bezplatnými poskytovateli bez klíče · 39 integrací CLI/agentů prostřednictvím jediné konfigurace · kompatibilita s OpenAI, Claude, Gemini a Responses API na /v1 · produkční ovládací prvky včetně jističů, TLS stealth, MCP se 110 nástroji, A2A, paměti, ochranných mechanismů, evaluací a více než 56 000 statických deklarací testů napříč více než 7 800 sledovanými testovacími soubory. +Příslib — jeden endpoint a 384 poskytovatelů. Automatický fallback udržuje směrování, dokud je k dispozici jiný zdravý cíl. Šest pilířů: odolný fallback napříč 384 poskytovateli · až 95% úspora tokenů u způsobilých úloh · $0 na začátek se 150+ bezplatnými tarify a 52 opakovaně bezplatnými poskytovateli bez klíče · 39 integrací CLI/agentů prostřednictvím jediné konfigurace · kompatibilita s OpenAI, Claude, Gemini a Responses API na /v1 · produkční ovládací prvky včetně jističů, TLS stealth, MCP se 110 nástroji, A2A, paměti, ochranných mechanismů, evaluací a více než 56 000 statických deklarací testů napříč více než 7 800 sledovanými testovacími soubory.

@@ -510,7 +510,7 @@ Všech **20** strategií — libovolně je kombinujte v jednotlivých krocích k -V čem OmniRoute vyniká — aktuální snímek funkcí oproti 9router, OpenRouter, CLIProxyAPI a LiteLLM v rámci 13 schopností. OmniRoute: 383 poskytovatelů, více než 150 integrovaných bezplatných tarifů, 20 strategií směrování, komprese tokenů pomocí 12 enginů, integrovaný server MCP se 110 nástroji, protokol agentů A2A, trvalá paměť, ochranná pravidla, cloudoví agenti, utajení pomocí otisku TLS, Desktop/Termux/PWA a 67 lokalizací uživatelského rozhraní i18n. OmniRoute je licencován pod MIT a lze jej hostovat samostatně. Funkce a počty konkurenčních řešení se mohou měnit; viz odkazovaná metodika. +V čem OmniRoute vyniká — aktuální snímek funkcí oproti 9router, OpenRouter, CLIProxyAPI a LiteLLM v rámci 13 schopností. OmniRoute: 384 poskytovatelů, více než 150 integrovaných bezplatných tarifů, 20 strategií směrování, komprese tokenů pomocí 12 enginů, integrovaný server MCP se 110 nástroji, protokol agentů A2A, trvalá paměť, ochranná pravidla, cloudoví agenti, utajení pomocí otisku TLS, Desktop/Termux/PWA a 67 lokalizací uživatelského rozhraní i18n. OmniRoute je licencován pod MIT a lze jej hostovat samostatně. Funkce a počty konkurenčních řešení se mohou měnit; viz odkazovaná metodika. 📊 Kompletní metodika a podrobnosti o jednotlivých funkcích oproti 9router, OpenRouter, CLIProxyAPI a LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -613,7 +613,7 @@ aktuálním katalogu najdete na **[radar.omniroute.online/planos](https://radar. - **🖼️ Nové endpointy** — `/v1/ocr` (Mistral OCR) a `/v1/audio/translations` (ve stylu Whisper) doplňují rozhraní pro média. → [Referenční dokumentace API](docs/reference/API_REFERENCE.md) - **🎨 Generování obrázků / videí / zvuku** — jedno API pro média: xAI Grok Imagine a video od Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind a poskytovatelé řeči, jako je ElevenLabs. → [Referenční dokumentace API](docs/reference/API_REFERENCE.md) - **🌍 Nasazení a provoz** — `basePath` reverse proxy, automatická detekce jazyka prohlížeče, sledování zařízení podle klíče, důvěryhodnost MITM bez práv root a lokalizace zh-TW. → [Prostředí](docs/reference/ENVIRONMENT.md) -- **🤝 Další poskytovatelé a agenti** — cloudoví agenti (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) s prohlížečem a přihlášením OAuth, plnohodnotná karta Ollama, Claude Opus 5 a Sonnet 5, oficiální partnerství s Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… a aktualizovaný **katalog 383 poskytovatelů**. → [Poskytovatelé](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Další poskytovatelé a agenti** — cloudoví agenti (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) s prohlížečem a přihlášením OAuth, plnohodnotná karta Ollama, Claude Opus 5 a Sonnet 5, oficiální partnerství s Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… a aktualizovaný **katalog 384 poskytovatelů**. → [Poskytovatelé](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparentnost směrování** — každá odpověď obsahuje hlavičku `X-OmniRoute-Decision` s názvem strategie, poskytovatele a latence, které ji obsloužily; nová kombinovaná strategie `cache-optimized` a faktor `cacheAffinity` v Auto-Combo směrují opakované požadavky zpět na připojení, které obsahuje předponu v mezipaměti; endpoint pouze pro čtení `/v1/auto-combo/{channel}/candidates` zpřístupňuje živý fond kandidátů kanálu `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Lokální výkon a infrastruktura** — lokální Redis na jedno kliknutí, nasazovače relay pro Cloudflare Workers / Deno Deploy, Bifrost a Mux jako spravované vestavěné služby. → [Vestavěné služby](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Také součástí balení** — framework pro pluginy + marketplace, frameworky dovedností Omni/Agent/GitHub, integrace trezoru Obsidian (22 nástrojů MCP), rozhraní Batch & Files API kompatibilní s OpenAI, sémantická mezipaměť odpovědí, gamifikace s žebříčky, vyhledávání agentů ACP (15 integrovaných agentů), plánovaný export protokolů do BigQuery, paralelní fan-out více modelů `auto/chaos`, most pro bota Telegram, správce verzí v aplikaci a žebříčky bezplatných poskytovatelů LMArena-ELO. → [Dokumentace](docs/README.md) @@ -697,11 +697,11 @@ mimo historii vašeho shellu. → [Integrace CLI](docs/guides/CLI-INTEGRATIONS.m
-## 🌐 383 poskytovatelů AI — 153 označených katalogem jako bezplatní +## 🌐 384 poskytovatelů AI — 153 označených katalogem jako bezplatní
-> **383 registrovaných poskytovatelů** napříč kanonickými kolekcemi chatu, médií, vyhledávání, lokálních, cloudových agentů a systémů, včetně **153 poskytovatelů s metadaty pro objevování `hasFree: true`**. Registr modelů chatu pokrývá **239 poskytovatelů / 3 059 jedinečných dvojic poskytovatel–model / 1 475 nezpracovaných ID modelů**; samostatný katalog bezplatných rozpočtů obsahuje **471 řádků pro jednotlivé modely**, **34 opakujících se fondů** a **52 poskytovatelů s opakovanou či bezklíčovou bezplatnou dostupností navždy**. Jedná se záměrně o různé jmenovatele; definice a výpočty deduplikované podle fondů jsou uvedeny v dokumentech [Reference poskytovatelů](docs/reference/PROVIDER_REFERENCE.md) a [Bezplatné tarify](docs/reference/FREE_TIERS.md). +> **384 registrovaných poskytovatelů** napříč kanonickými kolekcemi chatu, médií, vyhledávání, lokálních, cloudových agentů a systémů, včetně **153 poskytovatelů s metadaty pro objevování `hasFree: true`**. Registr modelů chatu pokrývá **239 poskytovatelů / 3 059 jedinečných dvojic poskytovatel–model / 1 475 nezpracovaných ID modelů**; samostatný katalog bezplatných rozpočtů obsahuje **471 řádků pro jednotlivé modely**, **34 opakujících se fondů** a **52 poskytovatelů s opakovanou či bezklíčovou bezplatnou dostupností navždy**. Jedná se záměrně o různé jmenovatele; definice a výpočty deduplikované podle fondů jsou uvedeny v dokumentech [Reference poskytovatelů](docs/reference/PROVIDER_REFERENCE.md) a [Bezplatné tarify](docs/reference/FREE_TIERS.md).
@@ -1302,7 +1302,7 @@ Kanonické metriky k 2026-10-18: **1.976 unikátních videí** · **20.220.289 z RuntimeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 JazykTypeScript 6.0 — 100% TypeScript napříč src/ a open-sse/ (v jádře od v2.0 žádné any) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Databázebetter-sqlite3 (SQLite, žurnálování WAL) + LowDB (starší JSON) — 147 doménových modulů, 230 migrací + Databázebetter-sqlite3 (SQLite, žurnálování WAL) + LowDB (starší JSON) — 147 doménových modulů, 232 migrací PaměťFulltextové vyhledávání SQLite FTS5 + vektorové embeddingy kvantizované na int8, typované stárnutí SchémataZod 4 — validace I/O nástrojů MCP + kontrakty API ProtokolyMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/cs/llm.txt b/docs/i18n/cs/llm.txt index 15fa26a6c27..a85133161ac 100644 --- a/docs/i18n/cs/llm.txt +++ b/docs/i18n/cs/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/da/README.md b/docs/i18n/da/README.md index f91642b720d..551cd9e12ff 100644 --- a/docs/i18n/da/README.md +++ b/docs/i18n/da/README.md @@ -15,7 +15,7 @@ **Oprettet af [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Stop aldrig med at kode. Alle AI-værktøjer → 383 udbydere — 150+ gratis — gennem ét endpoint. Claude Code, Codex, Cursor, Cline, Copilot og Antigravity til GRATIS Claude / GPT / Gemini med automatisk fallback. RTK + Caveman-stablet komprimering sparer 15–95 % tokens (~89 % i gennemsnit) — ram aldrig grænserne. 383 AI-udbydere · 150+ gratis niveauer · ~1,62 mia. gratis tokens/md. · 20 routingstrategier · $0 for at komme i gang. +OmniRoute — Stop aldrig med at kode. Alle AI-værktøjer → 384 udbydere — 150+ gratis — gennem ét endpoint. Claude Code, Codex, Cursor, Cline, Copilot og Antigravity til GRATIS Claude / GPT / Gemini med automatisk fallback. RTK + Caveman-stablet komprimering sparer 15–95 % tokens (~89 % i gennemsnit) — ram aldrig grænserne. 384 AI-udbydere · 150+ gratis niveauer · ~1,62 mia. gratis tokens/md. · 20 routingstrategier · $0 for at komme i gang.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------- | :---------------------: | :---------: | :-------------------: | -| 🌐 Udbydere | 352 | **358** | 383 | +| 🌐 Udbydere | 352 | **358** | 384 | | 🧠 Unikke chatmodel-id'er | 1320 | **1374** | 1491 | | 🖼️ Modalitetsbro | 🆕 vision + lyd + video | ✓ | ✓ | | 📡 Radars gratis katalog | 🆕 tilvalg | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funktioner 🎯 Kombinationer - 🌐 Udbydere + 🌐 Udbydere 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Løftet — Ét endpoint og 383 udbydere. Automatisk fallback holder routingen i gang, så længe der er et andet sundt mål tilgængeligt. Seks søjler: robust fallback på tværs af 383 udbydere · op til 95 % tokenbesparelse på kvalificerede arbejdsbelastninger · $0 for at komme i gang med over 150 gratis niveauer og 52 tilbagevendende/nøglefri udbydere, der er gratis for altid · 39 CLI-/agentintegrationer via én konfiguration · kompatibilitet med OpenAI-, Claude-, Gemini- og Responses API på /v1 · produktionskontroller, herunder circuit breakers, TLS-stealth, MCP med 110 værktøjer, A2A, hukommelse, guardrails, evalueringer og over 56.000 statiske testdeklarationer på tværs af over 7.800 registrerede testfiler. +Løftet — Ét endpoint og 384 udbydere. Automatisk fallback holder routingen i gang, så længe der er et andet sundt mål tilgængeligt. Seks søjler: robust fallback på tværs af 384 udbydere · op til 95 % tokenbesparelse på kvalificerede arbejdsbelastninger · $0 for at komme i gang med over 150 gratis niveauer og 52 tilbagevendende/nøglefri udbydere, der er gratis for altid · 39 CLI-/agentintegrationer via én konfiguration · kompatibilitet med OpenAI-, Claude-, Gemini- og Responses API på /v1 · produktionskontroller, herunder circuit breakers, TLS-stealth, MCP med 110 værktøjer, A2A, hukommelse, guardrails, evalueringer og over 56.000 statiske testdeklarationer på tværs af over 7.800 registrerede testfiler.

@@ -510,7 +510,7 @@ Alle **20** strategier — kombinér frit for hvert trin i en combo: -Hvad adskiller OmniRoute – et dateret øjebliksbillede af funktioner sammenlignet med 9router, OpenRouter, CLIProxyAPI og LiteLLM på tværs af 13 funktioner. OmniRoute: 383 udbydere, over 150 indbyggede gratis niveauer, 20 routingstrategier, tokenkomprimering med 12 motorer, indbygget MCP-server med 110 værktøjer, A2A-agentprotokol, persistent hukommelse, sikkerhedsregler, cloud-agenter, TLS-fingeraftryks-stealth, Desktop/Termux/PWA og 67 i18n-brugerfladelokaliteter. OmniRoute er MIT-licenseret og kan hostes selv. Konkurrenternes funktioner og antal kan ændre sig; se den linkede metodologi. +Hvad adskiller OmniRoute – et dateret øjebliksbillede af funktioner sammenlignet med 9router, OpenRouter, CLIProxyAPI og LiteLLM på tværs af 13 funktioner. OmniRoute: 384 udbydere, over 150 indbyggede gratis niveauer, 20 routingstrategier, tokenkomprimering med 12 motorer, indbygget MCP-server med 110 værktøjer, A2A-agentprotokol, persistent hukommelse, sikkerhedsregler, cloud-agenter, TLS-fingeraftryks-stealth, Desktop/Termux/PWA og 67 i18n-brugerfladelokaliteter. OmniRoute er MIT-licenseret og kan hostes selv. Konkurrenternes funktioner og antal kan ændre sig; se den linkede metodologi. 📊 Fuld metodologi og detaljer pr. funktion sammenlignet med 9router, OpenRouter, CLIProxyAPI og LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -614,7 +614,7 @@ adgangskrav og det aktuelle katalog på **[radar.omniroute.online/planos](https: - **🖼️ Nye endpoints** — `/v1/ocr` (Mistral OCR) og `/v1/audio/translations` (Whisper-lignende) fuldender medieoverfladen. → [API-reference](docs/reference/API_REFERENCE.md) - **🎨 Generering af billeder / video / lyd** — én API til medier: xAI Grok Imagine og Novita AI-video, ComfyUI, Magnific, Adobe Firefly, Segmind samt taleprovidere såsom ElevenLabs. → [API-reference](docs/reference/API_REFERENCE.md) - **🌍 Udrulning og drift** — reverse-proxy-`basePath`, automatisk registrering af browsersprog, enhedssporing pr. nøgle, rootless MITM-tillid og zh-TW-lokalisering. → [Miljø](docs/reference/ENVIRONMENT.md) -- **🤝 Flere providere og agenter** — cloud-agenter (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) med browser- og OAuth-login, førsteklasses Ollama-kort, Claude Opus 5 og Sonnet 5, officielt Kimi-partnerskab (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI … samt et opdateret **383-provider-katalog**. → [Providere](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Flere providere og agenter** — cloud-agenter (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) med browser- og OAuth-login, førsteklasses Ollama-kort, Claude Opus 5 og Sonnet 5, officielt Kimi-partnerskab (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI … samt et opdateret **384-provider-katalog**. → [Providere](docs/reference/PROVIDER_REFERENCE.md) - **📡 Gennemsigtig routing** — hvert svar indeholder en `X-OmniRoute-Decision`-header, der angiver strategien, provideren og latenstiden, som leverede svaret; en ny `cache-optimized`-kombinationsstrategi samt Auto-Combo-`cacheAffinity`-faktoren dirigerer gentagne anmodninger tilbage til forbindelsen, der holder det cachelagrede præfiks; og et skrivebeskyttet `/v1/auto-combo/{channel}/candidates`-endpoint viser en `auto/*`-kanals aktuelle kandidatsamling. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Lokal ydeevne og infrastruktur** — lokal Redis med ét klik, relay-deployere til Cloudflare Workers / Deno Deploy, Bifrost og Mux som overvågede indlejrede tjenester. → [Indlejrede tjenester](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Også inkluderet** — plugin-framework og markedsplads, Omni/Agent/GitHub-skills-frameworks, Obsidian vault-integration (22 MCP-værktøjer), OpenAI-kompatible Batch- og Files-API'er, semantisk responscache, gamification med ranglister, ACP-agentopdagelse (15 indbyggede agenter), planlagt eksport af logfiler til BigQuery, parallel multi-model-fan-out med `auto/chaos`, en Telegram-botbro, en versionsstyring i appen og LMArena-ELO-rangeringer af gratisudbydere. → [Dokumentation](docs/README.md) @@ -698,11 +698,11 @@ af din shell-historik. → [CLI-integrationer](docs/guides/CLI-INTEGRATIONS.md)
-## 🌐 383 AI-udbydere — 153 katalogmarkerede gratis +## 🌐 384 AI-udbydere — 153 katalogmarkerede gratis
-> **383 registrerede udbydere** på tværs af de kanoniske samlinger for chat, medier, søgning, lokale tjenester, cloud-agenter og systemer, herunder **153 med `hasFree: true`-opdagelsesmetadata**. Chatmodellens register dækker **239 udbydere / 3.059 distinkte udbyder-model-par / 1.475 rå model-ID'er**; det separate katalog over gratis budgetter har **471 rækker pr. model**, **34 tilbagevendende puljer** og **52 tilbagevendende/keyless gratis-for-altid-udbydere**. Dette er forskellige nævnere med vilje; definitioner og pool-deduplicerede beregninger findes i [Udbyderreference](docs/reference/PROVIDER_REFERENCE.md) og [Gratis niveauer](docs/reference/FREE_TIERS.md). +> **384 registrerede udbydere** på tværs af de kanoniske samlinger for chat, medier, søgning, lokale tjenester, cloud-agenter og systemer, herunder **153 med `hasFree: true`-opdagelsesmetadata**. Chatmodellens register dækker **239 udbydere / 3.059 distinkte udbyder-model-par / 1.475 rå model-ID'er**; det separate katalog over gratis budgetter har **471 rækker pr. model**, **34 tilbagevendende puljer** og **52 tilbagevendende/keyless gratis-for-altid-udbydere**. Dette er forskellige nævnere med vilje; definitioner og pool-deduplicerede beregninger findes i [Udbyderreference](docs/reference/PROVIDER_REFERENCE.md) og [Gratis niveauer](docs/reference/FREE_TIERS.md).
@@ -1304,7 +1304,7 @@ Kanoniske målinger pr. 2026-10-18: **1.976 unikke videoer** · **20.220.289 ken KørselstidNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 SprogTypeScript 6.0 — 100 % TypeScript på tværs af src/ og open-sse/ (ingen any i kernen siden v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Databasebetter-sqlite3 (SQLite, WAL-journalføring) + LowDB (JSON-legacy) — 147 domænemoduler, 230 migreringer + Databasebetter-sqlite3 (SQLite, WAL-journalføring) + LowDB (JSON-legacy) — 147 domænemoduler, 232 migreringer HukommelseSQLite FTS5-fuldtekst + int8-kvantiserede vektorembeddings, typet henfald SkemaerZod 4 — MCP-værktøjs-I/O-validering + API-kontrakter ProtokollerMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/da/llm.txt b/docs/i18n/da/llm.txt index 5cec1d247e5..c87aedf309e 100644 --- a/docs/i18n/da/llm.txt +++ b/docs/i18n/da/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/de/README.md b/docs/i18n/de/README.md index a6ab9fe69d4..bba6c600006 100644 --- a/docs/i18n/de/README.md +++ b/docs/i18n/de/README.md @@ -15,7 +15,7 @@ **Erstellt von [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Höre niemals auf zu programmieren. Jedes KI-Tool → 383 Anbieter — 150+ kostenlos — über einen einzigen Endpunkt. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity mit automatischem Fallback in KOSTENLOSES Claude / GPT / Gemini. Die gestapelte RTK- + Caveman-Komprimierung spart 15–95 % an Tokens (~89 % im Durchschnitt) — erreiche niemals Limits. 383 KI-Anbieter · 150+ kostenlose Tarife · ~1,62 Mrd. kostenlose Tokens/Monat · 20 Routing-Strategien · Start bei 0 $. +OmniRoute — Höre niemals auf zu programmieren. Jedes KI-Tool → 384 Anbieter — 150+ kostenlos — über einen einzigen Endpunkt. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity mit automatischem Fallback in KOSTENLOSES Claude / GPT / Gemini. Die gestapelte RTK- + Caveman-Komprimierung spart 15–95 % an Tokens (~89 % im Durchschnitt) — erreiche niemals Limits. 384 KI-Anbieter · 150+ kostenlose Tarife · ~1,62 Mrd. kostenlose Tokens/Monat · 20 Routing-Strategien · Start bei 0 $.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ----------------------------- | :-----------------------: | :---------: | :------------------------: | -| 🌐 Anbieter | 352 | **358** | 383 | +| 🌐 Anbieter | 352 | **358** | 384 | | 🧠 Eindeutige Chat-Modell-IDs | 1320 | **1374** | 1491 | | 🖼️ Modalitätsbrücke | 🆕 Vision + Audio + Video | ✓ | ✓ | | 📡 Kostenloser Radar-Katalog | 🆕 optional | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funktionen 🎯 Kombinationen - 🌐 Anbieter + 🌐 Anbieter 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Das Versprechen — Ein Endpunkt und 383 Anbieter. Automatisches Fallback hält das Routing aufrecht, solange ein anderes funktionsfähiges Ziel verfügbar ist. Sechs Säulen: robustes Fallback über 383 Anbieter hinweg · bis zu 95 % Token-Einsparungen bei geeigneten Workloads · $0 zum Start mit mehr als 150 kostenlosen Tarifen und 52 wiederkehrenden/schlüssellosen dauerhaft kostenlosen Anbietern · 39 CLI-/Agent-Integrationen über eine Konfiguration · Kompatibilität mit OpenAI-, Claude-, Gemini- und Responses-APIs unter /v1 · Produktivitätskontrollen einschließlich Circuit Breakern, TLS-Stealth, MCP mit 110 Tools, A2A, Memory, Guardrails, Evals und mehr als 56.000 statischen Testdeklarationen in über 7.800 erfassten Testdateien. +Das Versprechen — Ein Endpunkt und 384 Anbieter. Automatisches Fallback hält das Routing aufrecht, solange ein anderes funktionsfähiges Ziel verfügbar ist. Sechs Säulen: robustes Fallback über 384 Anbieter hinweg · bis zu 95 % Token-Einsparungen bei geeigneten Workloads · $0 zum Start mit mehr als 150 kostenlosen Tarifen und 52 wiederkehrenden/schlüssellosen dauerhaft kostenlosen Anbietern · 39 CLI-/Agent-Integrationen über eine Konfiguration · Kompatibilität mit OpenAI-, Claude-, Gemini- und Responses-APIs unter /v1 · Produktivitätskontrollen einschließlich Circuit Breakern, TLS-Stealth, MCP mit 110 Tools, A2A, Memory, Guardrails, Evals und mehr als 56.000 statischen Testdeklarationen in über 7.800 erfassten Testdateien.

@@ -510,7 +510,7 @@ Alle **20** Strategien — pro Combo-Schritt frei kombinierbar: -Was OmniRoute auszeichnet – eine datierte Funktionsübersicht im Vergleich zu 9router, OpenRouter, CLIProxyAPI und LiteLLM über 13 Fähigkeiten hinweg. OmniRoute: 383 Anbieter, über 150 integrierte kostenlose Kontingente, 20 Routing-Strategien, Token-Komprimierung mit 12 Engines, integrierter MCP-Server mit 110 Tools, A2A-Agentenprotokoll, dauerhafter Speicher, Schutzmechanismen, Cloud-Agenten, TLS-Fingerprinting-Tarnung, Desktop/Termux/PWA und 67 i18n-UI-Lokalisierungen. OmniRoute ist MIT-lizenziert und kann selbst gehostet werden. Die Funktionen und Anzahlen der Wettbewerber können sich ändern; siehe die verlinkte Methodik. +Was OmniRoute auszeichnet – eine datierte Funktionsübersicht im Vergleich zu 9router, OpenRouter, CLIProxyAPI und LiteLLM über 13 Fähigkeiten hinweg. OmniRoute: 384 Anbieter, über 150 integrierte kostenlose Kontingente, 20 Routing-Strategien, Token-Komprimierung mit 12 Engines, integrierter MCP-Server mit 110 Tools, A2A-Agentenprotokoll, dauerhafter Speicher, Schutzmechanismen, Cloud-Agenten, TLS-Fingerprinting-Tarnung, Desktop/Termux/PWA und 67 i18n-UI-Lokalisierungen. OmniRoute ist MIT-lizenziert und kann selbst gehostet werden. Die Funktionen und Anzahlen der Wettbewerber können sich ändern; siehe die verlinkte Methodik. 📊 Vollständige Methodik und Details zu den einzelnen Funktionen im Vergleich zu 9router, OpenRouter, CLIProxyAPI und LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -614,7 +614,7 @@ zum aktuellen Katalog finden Sie unter **[radar.omniroute.online/planos](https:/ - **🖼️ Neue Endpunkte** — `/v1/ocr` (Mistral OCR) und `/v1/audio/translations` (im Whisper-Stil) vervollständigen die Medienfunktionen. → [API-Referenz](docs/reference/API_REFERENCE.md) - **🎨 Bild-, Video- und Audiogenerierung** — eine API für Medien: xAI Grok Imagine und Novita AI Video, ComfyUI, Magnific, Adobe Firefly, Segmind sowie Sprachanbieter wie ElevenLabs. → [API-Referenz](docs/reference/API_REFERENCE.md) - **🌍 Bereitstellung und Betrieb** — Reverse-Proxy-`basePath`, automatische Erkennung der Browsersprache, Gerätemitverfolgung pro Schlüssel, MITM-Trust ohne Root-Rechte und zh-TW-Lokalisierung. → [Umgebung](docs/reference/ENVIRONMENT.md) -- **🤝 Mehr Anbieter und Agents** — Cloud-Agents (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) mit Browser- und OAuth-Anmeldung, eine erstklassige Ollama-Karte, Claude Opus 5 und Sonnet 5, offizielle Kimi-Partnerschaft (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI … sowie ein überarbeiteter **383-Anbieter-Katalog**. → [Anbieter](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Mehr Anbieter und Agents** — Cloud-Agents (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) mit Browser- und OAuth-Anmeldung, eine erstklassige Ollama-Karte, Claude Opus 5 und Sonnet 5, offizielle Kimi-Partnerschaft (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI … sowie ein überarbeiteter **384-Anbieter-Katalog**. → [Anbieter](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparenz beim Routing** — jede Antwort enthält einen `X-OmniRoute-Decision`-Header mit Strategie, Anbieter und Latenz, die sie bereitgestellt haben; eine neue `cache-optimized`-Kombi-Strategie sowie der Auto-Combo-Faktor `cacheAffinity` leiten wiederholte Anfragen an die Verbindung zurück, die das zwischengespeicherte Präfix enthält, und ein schreibgeschützter Endpunkt `/v1/auto-combo/{channel}/candidates` stellt den aktuellen Kandidatenpool eines `auto/*`-Kanals bereit. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Lokale Performance & Infrastruktur** — lokales Redis mit einem Klick, Relay-Deployer für Cloudflare Workers / Deno Deploy sowie Bifrost & Mux als überwachte eingebettete Dienste. → [Eingebettete Dienste](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Ebenfalls enthalten** — Plugin-Framework + Marketplace, Omni/Agent/GitHub-Skills-Frameworks, Obsidian-Vault-Integration (22 MCP-Tools), mit OpenAI kompatible Batch- & Files-APIs, semantischer Antwort-Cache, Gamification mit Bestenlisten, ACP-Agent-Erkennung (15 integrierte Agents), geplante Log-Exporte nach BigQuery, parallele Multi-Modell-Fan-outs mit `auto/chaos`, eine Telegram-Bot-Bridge, ein integrierter Versionsmanager und kostenlose LMArena-ELO-Ranglisten für Anbieter. → [Dokumentation](docs/README.md) @@ -698,11 +698,11 @@ aus deiner Shell-Historie heraus. → [CLI-Integrationen](docs/guides/CLI-INTEGR
-## 🌐 383 KI-Anbieter — 153 im Katalog als kostenlos markiert +## 🌐 384 KI-Anbieter — 153 im Katalog als kostenlos markiert
-> **383 registrierte Anbieter** in den kanonischen Sammlungen für Chat, Medien, Suche, lokal, Cloud-Agenten und Systeme, darunter **153 mit `hasFree: true`-Erkennungsmetadaten**. Das Chat-Modellregister umfasst **239 Anbieter / 3.059 eindeutige Anbieter-Modell-Paare / 1.475 rohe Modell-IDs**; der separate Katalog für kostenlose Kontingente enthält **471 Zeilen pro Modell**, **34 wiederkehrende Kontingente** und **52 dauerhaft kostenlose Anbieter ohne wiederkehrende Kosten bzw. ohne erforderlichen Schlüssel**. Diese Nenner unterscheiden sich absichtlich; Definitionen und poolbereinigte Berechnungen finden Sie in der [Anbieterreferenz](docs/reference/PROVIDER_REFERENCE.md) und unter [Kostenlose Tarife](docs/reference/FREE_TIERS.md). +> **384 registrierte Anbieter** in den kanonischen Sammlungen für Chat, Medien, Suche, lokal, Cloud-Agenten und Systeme, darunter **153 mit `hasFree: true`-Erkennungsmetadaten**. Das Chat-Modellregister umfasst **239 Anbieter / 3.059 eindeutige Anbieter-Modell-Paare / 1.475 rohe Modell-IDs**; der separate Katalog für kostenlose Kontingente enthält **471 Zeilen pro Modell**, **34 wiederkehrende Kontingente** und **52 dauerhaft kostenlose Anbieter ohne wiederkehrende Kosten bzw. ohne erforderlichen Schlüssel**. Diese Nenner unterscheiden sich absichtlich; Definitionen und poolbereinigte Berechnungen finden Sie in der [Anbieterreferenz](docs/reference/PROVIDER_REFERENCE.md) und unter [Kostenlose Tarife](docs/reference/FREE_TIERS.md).
@@ -1304,7 +1304,7 @@ Kanonische Metriken vom 2026-10-18: **1.976 einzigartige Videos** · **20.220.28 LaufzeitNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 SpracheTypeScript 6.0 — 100 % TypeScript in src/ und open-sse/ (seit v2.0 kein any im Kern) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Datenbankbetter-sqlite3 (SQLite, WAL-Journaling) + LowDB (veraltetes JSON) — 147 Domänenmodule, 230 Migrationen + Datenbankbetter-sqlite3 (SQLite, WAL-Journaling) + LowDB (veraltetes JSON) — 147 Domänenmodule, 232 Migrationen SpeicherSQLite-FTS5-Volltextsuche + int8-quantisierte Vektoreinbettungen, typisierte Abklingfunktion SchemasZod 4 — MCP-Tool-E/A-Validierung + API-Verträge ProtokolleMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/de/llm.txt b/docs/i18n/de/llm.txt index 7ba62f3196c..c045839dbee 100644 --- a/docs/i18n/de/llm.txt +++ b/docs/i18n/de/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/el/README.md b/docs/i18n/el/README.md index c38ea35188f..4b59eae2386 100644 --- a/docs/i18n/el/README.md +++ b/docs/i18n/el/README.md @@ -15,7 +15,7 @@ **Δημιουργήθηκε από τον [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Μην σταματάς ποτέ να γράφεις κώδικα. Κάθε εργαλείο AI → 383 πάροχοι — 150+ δωρεάν — μέσω ενός endpoint. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity σε δωρεάν Claude / GPT / Gemini με αυτόματη εναλλακτική δρομολόγηση. Η συμπιεσμένη διαδοχική συμπίεση RTK + Caveman εξοικονομεί 15–95% των tokens (~89% κατά μέσο όρο) — δεν θα φτάσεις ποτέ τα όρια. 383 πάροχοι AI · 150+ δωρεάν επίπεδα · ~1,62 δισ. δωρεάν tokens/μήνα · 20 στρατηγικές δρομολόγησης · $0 για να ξεκινήσεις. +OmniRoute — Μην σταματάς ποτέ να γράφεις κώδικα. Κάθε εργαλείο AI → 384 πάροχοι — 150+ δωρεάν — μέσω ενός endpoint. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity σε δωρεάν Claude / GPT / Gemini με αυτόματη εναλλακτική δρομολόγηση. Η συμπιεσμένη διαδοχική συμπίεση RTK + Caveman εξοικονομεί 15–95% των tokens (~89% κατά μέσο όρο) — δεν θα φτάσεις ποτέ τα όρια. 384 πάροχοι AI · 150+ δωρεάν επίπεδα · ~1,62 δισ. δωρεάν tokens/μήνα · 20 στρατηγικές δρομολόγησης · $0 για να ξεκινήσεις.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------------------------- | :----------------------: | :---------: | :-------------------: | -| 🌐 Πάροχοι | 352 | **358** | 383 | +| 🌐 Πάροχοι | 352 | **358** | 384 | | 🧠 Μοναδικά αναγνωριστικά μοντέλων συνομιλίας | 1320 | **1374** | 1491 | | 🖼️ Γέφυρα τροπικοτήτων | 🆕 όραση + ήχος + βίντεο | ✓ | ✓ | | 📡 Δωρεάν κατάλογος Radar | 🆕 προαιρετική συμμετοχή | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Δυνατότητες 🎯 Συνδυασμοί - 🌐 Πάροχοι + 🌐 Πάροχοι 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Η υπόσχεση — Ένα endpoint και 383 πάροχοι. Η αυτόματη εφεδρική μετάβαση διατηρεί τη δρομολόγηση όσο υπάρχει διαθέσιμος άλλος υγιής στόχος. Έξι πυλώνες: ανθεκτική εφεδρική μετάβαση μεταξύ 383 παρόχων · έως και 95% εξοικονόμηση token σε κατάλληλους φόρτους εργασίας · $0 για εκκίνηση με περισσότερα από 150 δωρεάν προγράμματα και 52 επαναλαμβανόμενους/χωρίς κλειδί παρόχους που είναι δωρεάν για πάντα · 39 ενσωματώσεις CLI/agent μέσω μίας ρύθμισης · συμβατότητα με τα OpenAI, Claude, Gemini και Responses API στο /v1 · έλεγχοι παραγωγής, όπως circuit breakers, TLS stealth, MCP με 110 εργαλεία, A2A, μνήμη, guardrails, evals και περισσότερες από 56.000 δηλώσεις στατικών δοκιμών σε περισσότερα από 7.800 καταγεγραμμένα αρχεία δοκιμών. +Η υπόσχεση — Ένα endpoint και 384 πάροχοι. Η αυτόματη εφεδρική μετάβαση διατηρεί τη δρομολόγηση όσο υπάρχει διαθέσιμος άλλος υγιής στόχος. Έξι πυλώνες: ανθεκτική εφεδρική μετάβαση μεταξύ 384 παρόχων · έως και 95% εξοικονόμηση token σε κατάλληλους φόρτους εργασίας · $0 για εκκίνηση με περισσότερα από 150 δωρεάν προγράμματα και 52 επαναλαμβανόμενους/χωρίς κλειδί παρόχους που είναι δωρεάν για πάντα · 39 ενσωματώσεις CLI/agent μέσω μίας ρύθμισης · συμβατότητα με τα OpenAI, Claude, Gemini και Responses API στο /v1 · έλεγχοι παραγωγής, όπως circuit breakers, TLS stealth, MCP με 110 εργαλεία, A2A, μνήμη, guardrails, evals και περισσότερες από 56.000 δηλώσεις στατικών δοκιμών σε περισσότερα από 7.800 καταγεγραμμένα αρχεία δοκιμών.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -Τι ξεχωρίζει το OmniRoute — ένα στιγμιότυπο δυνατοτήτων με ημερομηνία σε σύγκριση με τα 9router, OpenRouter, CLIProxyAPI και LiteLLM σε 13 δυνατότητες. Το OmniRoute: 383 πάροχοι, ενσωματωμένα 150+ δωρεάν επίπεδα, 20 στρατηγικές δρομολόγησης, συμπίεση token με 12 μηχανές, ενσωματωμένος διακομιστής MCP με 110 εργαλεία, πρωτόκολλο πρακτόρων A2A, μόνιμη μνήμη, guardrails, cloud agents, stealth αποτύπωμα TLS, Desktop/Termux/PWA και 67 τοπικές ρυθμίσεις διεπαφής χρήστη i18n. Το OmniRoute διαθέτει άδεια MIT και μπορεί να φιλοξενηθεί από τον ίδιο τον χρήστη. Οι δυνατότητες και οι μετρήσεις των ανταγωνιστών ενδέχεται να αλλάξουν· ανατρέξτε στη συνδεδεμένη μεθοδολογία. +Τι ξεχωρίζει το OmniRoute — ένα στιγμιότυπο δυνατοτήτων με ημερομηνία σε σύγκριση με τα 9router, OpenRouter, CLIProxyAPI και LiteLLM σε 13 δυνατότητες. Το OmniRoute: 384 πάροχοι, ενσωματωμένα 150+ δωρεάν επίπεδα, 20 στρατηγικές δρομολόγησης, συμπίεση token με 12 μηχανές, ενσωματωμένος διακομιστής MCP με 110 εργαλεία, πρωτόκολλο πρακτόρων A2A, μόνιμη μνήμη, guardrails, cloud agents, stealth αποτύπωμα TLS, Desktop/Termux/PWA και 67 τοπικές ρυθμίσεις διεπαφής χρήστη i18n. Το OmniRoute διαθέτει άδεια MIT και μπορεί να φιλοξενηθεί από τον ίδιο τον χρήστη. Οι δυνατότητες και οι μετρήσεις των ανταγωνιστών ενδέχεται να αλλάξουν· ανατρέξτε στη συνδεδεμένη μεθοδολογία. 📊 Πλήρης μεθοδολογία και λεπτομέρειες ανά δυνατότητα σε σύγκριση με τα 9router, OpenRouter, CLIProxyAPI και LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -614,7 +614,7 @@ Pix copia-e-cola: - **🖼️ Νέα endpoints** — τα `/v1/ocr` (Mistral OCR) και `/v1/audio/translations` (τύπου Whisper) ολοκληρώνουν την επιφάνεια πολυμέσων. → [Αναφορά API](docs/reference/API_REFERENCE.md) - **🎨 Δημιουργία εικόνας / βίντεο / ήχου** — ένα API για πολυμέσα: xAI Grok Imagine και βίντεο Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind και πάροχοι ομιλίας όπως το ElevenLabs. → [Αναφορά API](docs/reference/API_REFERENCE.md) - **🌍 Ανάπτυξη και λειτουργίες** — `basePath` reverse-proxy, αυτόματος εντοπισμός γλώσσας προγράμματος περιήγησης, παρακολούθηση συσκευών ανά κλειδί, trust MITM χωρίς δικαιώματα root, τοπικοποίηση zh-TW. → [Περιβάλλον](docs/reference/ENVIRONMENT.md) -- **🤝 Περισσότεροι πάροχοι και agent** — cloud agent (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) με σύνδεση μέσω browser + OAuth, κάρτα πρώτης κατηγορίας για Ollama, Claude Opus 5 και Sonnet 5, επίσημη συνεργασία με την Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… και ανανεωμένος **κατάλογος 383 παρόχων**. → [Πάροχοι](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Περισσότεροι πάροχοι και agent** — cloud agent (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) με σύνδεση μέσω browser + OAuth, κάρτα πρώτης κατηγορίας για Ollama, Claude Opus 5 και Sonnet 5, επίσημη συνεργασία με την Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… και ανανεωμένος **κατάλογος 384 παρόχων**. → [Πάροχοι](docs/reference/PROVIDER_REFERENCE.md) - **📡 Διαφάνεια δρομολόγησης** — κάθε απόκριση περιλαμβάνει μια κεφαλίδα `X-OmniRoute-Decision` που αναφέρει τη στρατηγική/πάροχο/καθυστέρηση που την εξυπηρέτησε, μια νέα στρατηγική συνδυασμού `cache-optimized` και ο παράγοντας `cacheAffinity` του Auto-Combo δρομολογούν τα επαναλαμβανόμενα αιτήματα πίσω στη σύνδεση που διατηρεί το αποθηκευμένο πρόθεμα, ενώ ένα μόνο για ανάγνωση endpoint `/v1/auto-combo/{channel}/candidates` εκθέτει τη ζωντανή ομάδα υποψηφίων ενός καναλιού `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Τοπική απόδοση και υποδομή** — τοπικό Redis με ένα κλικ, deployers relay για Cloudflare Workers / Deno Deploy, Bifrost και Mux ως εποπτευόμενες ενσωματωμένες υπηρεσίες. → [Ενσωματωμένες υπηρεσίες](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Περιλαμβάνονται επίσης** — πλαίσιο προσθηκών και marketplace, πλαίσια δεξιοτήτων Omni/Agent/GitHub, ενσωμάτωση με θησαυροφυλάκιο Obsidian (22 εργαλεία MCP), API Batch και Files συμβατά με OpenAI, σημασιολογική cache αποκρίσεων, gamification με βαθμολογικούς πίνακες, ανακάλυψη πρακτόρων ACP (15 ενσωματωμένοι πράκτορες), προγραμματισμένη εξαγωγή αρχείων καταγραφής στο BigQuery, παράλληλο multi-model fan-out `auto/chaos`, γέφυρα bot Telegram, διαχειριστής εκδόσεων εντός της εφαρμογής και δωρεάν κατατάξεις παρόχων LMArena-ELO. → [Τεκμηρίωση](docs/README.md) @@ -698,11 +698,11 @@ omniroute configure codex # επίσης: claude opencode qwen aider go
-## 🌐 383 πάροχοι AI — 153 σημειωμένοι στον κατάλογο ως δωρεάν +## 🌐 384 πάροχοι AI — 153 σημειωμένοι στον κατάλογο ως δωρεάν
-> **383 εγγεγραμμένοι πάροχοι** στις κανονικές συλλογές συνομιλίας, πολυμέσων, αναζήτησης, τοπικών, cloud-agent και συστήματος, συμπεριλαμβανομένων **153 που διαθέτουν μεταδεδομένα ανακάλυψης `hasFree: true`**. Το μητρώο μοντέλων συνομιλίας καλύπτει **239 παρόχους / 3,059 διακριτά ζεύγη παρόχου-μοντέλου / 1,475 ακατέργαστα αναγνωριστικά μοντέλων**· ο ξεχωριστός κατάλογος δωρεάν προϋπολογισμού διαθέτει **471 γραμμές ανά μοντέλο**, **34 επαναλαμβανόμενες δεξαμενές** και **52 επαναλαμβανόμενους/χωρίς κλειδί δωρεάν παρόχους για πάντα**. Αυτοί είναι διαφορετικοί παρονομαστές εκ σχεδιασμού· οι ορισμοί και οι υπολογισμοί με αφαίρεση διπλοτύπων δεξαμενών βρίσκονται στην [Αναφορά παρόχων](docs/reference/PROVIDER_REFERENCE.md) και στις [Δωρεάν βαθμίδες](docs/reference/FREE_TIERS.md). +> **384 εγγεγραμμένοι πάροχοι** στις κανονικές συλλογές συνομιλίας, πολυμέσων, αναζήτησης, τοπικών, cloud-agent και συστήματος, συμπεριλαμβανομένων **153 που διαθέτουν μεταδεδομένα ανακάλυψης `hasFree: true`**. Το μητρώο μοντέλων συνομιλίας καλύπτει **239 παρόχους / 3,059 διακριτά ζεύγη παρόχου-μοντέλου / 1,475 ακατέργαστα αναγνωριστικά μοντέλων**· ο ξεχωριστός κατάλογος δωρεάν προϋπολογισμού διαθέτει **471 γραμμές ανά μοντέλο**, **34 επαναλαμβανόμενες δεξαμενές** και **52 επαναλαμβανόμενους/χωρίς κλειδί δωρεάν παρόχους για πάντα**. Αυτοί είναι διαφορετικοί παρονομαστές εκ σχεδιασμού· οι ορισμοί και οι υπολογισμοί με αφαίρεση διπλοτύπων δεξαμενών βρίσκονται στην [Αναφορά παρόχων](docs/reference/PROVIDER_REFERENCE.md) και στις [Δωρεάν βαθμίδες](docs/reference/FREE_TIERS.md).
@@ -1304,7 +1304,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 επίσης τη RuntimeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ΓλώσσαTypeScript 6.0 — 100% TypeScript στα src/ και open-sse/ (μηδενικά any στον πυρήνα από την v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Βάση δεδομένωνbetter-sqlite3 (SQLite, καταγραφή WAL) + LowDB (παλαιού τύπου JSON) — 147 αρθρώματα τομέα, 230 μεταναστεύσεις + Βάση δεδομένωνbetter-sqlite3 (SQLite, καταγραφή WAL) + LowDB (παλαιού τύπου JSON) — 147 αρθρώματα τομέα, 232 μεταναστεύσεις ΜνήμηΠλήρες κείμενο SQLite FTS5 + διανυσματικά embeddings κβαντισμένα σε int8, τυποποιημένη εξασθένηση ΣχήματαZod 4 — επικύρωση I/O εργαλείων MCP + συμβάσεις API ΠρωτόκολλαMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/el/llm.txt b/docs/i18n/el/llm.txt index 3c9c49cd05a..1562d766439 100644 --- a/docs/i18n/el/llm.txt +++ b/docs/i18n/el/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/es/README.md b/docs/i18n/es/README.md index 8bb14c7a68a..406214ba5af 100644 --- a/docs/i18n/es/README.md +++ b/docs/i18n/es/README.md @@ -15,7 +15,7 @@ **Creado por [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Nunca dejes de programar. Cada herramienta de IA → 383 proveedores — 150+ gratuitos — a través de un único endpoint. Claude Code, Codex, Cursor, Cline, Copilot y Antigravity en Claude / GPT / Gemini GRATUITOS con conmutación automática por error. La compresión apilada de RTK + Caveman ahorra entre un 15 % y un 95 % de tokens (~89 % de media): nunca alcances los límites. 383 proveedores de IA · más de 150 niveles gratuitos · ~1,62 mil millones de tokens gratuitos/mes · 20 estrategias de enrutamiento · $0 para comenzar. +OmniRoute — Nunca dejes de programar. Cada herramienta de IA → 384 proveedores — 150+ gratuitos — a través de un único endpoint. Claude Code, Codex, Cursor, Cline, Copilot y Antigravity en Claude / GPT / Gemini GRATUITOS con conmutación automática por error. La compresión apilada de RTK + Caveman ahorra entre un 15 % y un 95 % de tokens (~89 % de media): nunca alcances los límites. 384 proveedores de IA · más de 150 niveles gratuitos · ~1,62 mil millones de tokens gratuitos/mes · 20 estrategias de enrutamiento · $0 para comenzar.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | -------------------------------------- | :-----------------------: | :---------: | :-----------------------: | -| 🌐 Proveedores | 352 | **358** | 383 | +| 🌐 Proveedores | 352 | **358** | 384 | | 🧠 IDs de modelos de chat únicos | 1320 | **1374** | 1491 | | 🖼️ Puente de modalidades | 🆕 visión + audio + vídeo | ✓ | ✓ | | 📡 Catálogo gratuito de Radar | 🆕 adhesión voluntaria | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funciones 🎯 Combos - 🌐 Proveedores + 🌐 Proveedores 🔌 CLI y MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -La promesa — Un endpoint y 383 proveedores. El fallback automático mantiene el enrutamiento mientras haya otro destino saludable disponible. Seis pilares: fallback resiliente entre 383 proveedores · hasta un 95 % de ahorro de tokens en cargas de trabajo elegibles · $0 para empezar, con más de 150 niveles gratuitos y 52 proveedores gratuitos recurrentes o sin clave para siempre · 39 integraciones de CLI/agentes mediante una única configuración · compatibilidad con las API de OpenAI, Claude, Gemini y Responses en /v1 · controles de producción que incluyen disyuntores, sigilo TLS, 110 herramientas MCP, A2A, memoria, barreras de protección, evaluaciones y más de 56.000 declaraciones de pruebas estáticas en más de 7.800 archivos de pruebas registrados. +La promesa — Un endpoint y 384 proveedores. El fallback automático mantiene el enrutamiento mientras haya otro destino saludable disponible. Seis pilares: fallback resiliente entre 384 proveedores · hasta un 95 % de ahorro de tokens en cargas de trabajo elegibles · $0 para empezar, con más de 150 niveles gratuitos y 52 proveedores gratuitos recurrentes o sin clave para siempre · 39 integraciones de CLI/agentes mediante una única configuración · compatibilidad con las API de OpenAI, Claude, Gemini y Responses en /v1 · controles de producción que incluyen disyuntores, sigilo TLS, 110 herramientas MCP, A2A, memoria, barreras de protección, evaluaciones y más de 56.000 declaraciones de pruebas estáticas en más de 7.800 archivos de pruebas registrados.

@@ -510,7 +510,7 @@ Las **20** estrategias: combínalas como quieras en cada paso del combo: -Qué distingue a OmniRoute: una instantánea fechada de las funcionalidades frente a 9router, OpenRouter, CLIProxyAPI y LiteLLM en 13 capacidades. OmniRoute: 383 proveedores, más de 150 niveles gratuitos integrados, 20 estrategias de enrutamiento, compresión de tokens con 12 motores, servidor MCP integrado con 110 herramientas, protocolo de agentes A2A, memoria persistente, protecciones, agentes en la nube, sigilo de huella TLS, Desktop/Termux/PWA y 67 idiomas de interfaz de usuario i18n. OmniRoute cuenta con licencia MIT y puede autoalojarse. Las capacidades y cantidades de los competidores pueden cambiar; consulta la metodología enlazada. +Qué distingue a OmniRoute: una instantánea fechada de las funcionalidades frente a 9router, OpenRouter, CLIProxyAPI y LiteLLM en 13 capacidades. OmniRoute: 384 proveedores, más de 150 niveles gratuitos integrados, 20 estrategias de enrutamiento, compresión de tokens con 12 motores, servidor MCP integrado con 110 herramientas, protocolo de agentes A2A, memoria persistente, protecciones, agentes en la nube, sigilo de huella TLS, Desktop/Termux/PWA y 67 idiomas de interfaz de usuario i18n. OmniRoute cuenta con licencia MIT y puede autoalojarse. Las capacidades y cantidades de los competidores pueden cambiar; consulta la metodología enlazada. 📊 Metodología completa y detalles por funcionalidad frente a 9router, OpenRouter, CLIProxyAPI y LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar es opcional y solo utiliza GET. El cliente de OmniRoute no carga prompts, - **🖼️ Nuevos endpoints** — `/v1/ocr` (Mistral OCR) y `/v1/audio/translations` (estilo Whisper) completan la superficie multimedia. → [Referencia de la API](docs/reference/API_REFERENCE.md) - **🎨 Generación de imágenes, vídeo y audio** — una API para medios: vídeo de xAI Grok Imagine y Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind y proveedores de voz como ElevenLabs. → [Referencia de la API](docs/reference/API_REFERENCE.md) - **🌍 Implementación y operaciones** — `basePath` de proxy inverso, detección automática del idioma del navegador, seguimiento de dispositivos por clave, confianza MITM sin privilegios de root y localización zh-TW. → [Entorno](docs/reference/ENVIRONMENT.md) -- **🤝 Más proveedores y agentes** — agentes en la nube (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) con navegador e inicio de sesión OAuth, tarjeta de primera clase para Ollama, Claude Opus 5 y Sonnet 5, colaboración oficial con Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… y un catálogo actualizado de **383 proveedores**. → [Proveedores](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Más proveedores y agentes** — agentes en la nube (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) con navegador e inicio de sesión OAuth, tarjeta de primera clase para Ollama, Claude Opus 5 y Sonnet 5, colaboración oficial con Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… y un catálogo actualizado de **384 proveedores**. → [Proveedores](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparencia del enrutamiento** — cada respuesta incluye un encabezado `X-OmniRoute-Decision` que indica la estrategia, el proveedor y la latencia que la atendieron; una nueva estrategia de combinación `cache-optimized` y el factor `cacheAffinity` de Auto-Combo redirigen las solicitudes repetidas a la conexión que contiene el prefijo almacenado en caché; además, un endpoint de solo lectura `/v1/auto-combo/{channel}/candidates` expone el conjunto de candidatos activo del canal `auto/*`. → [Combinación automática](docs/routing/AUTO-COMBO.md) - **⚡ Rendimiento e infraestructura locales** — Redis local con un clic, implementadores de relays para Cloudflare Workers / Deno Deploy, Bifrost y Mux como servicios integrados supervisados. → [Servicios integrados](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 También incluido** — marco de plugins + marketplace, marcos de habilidades de Omni/Agent/GitHub, integración con bóvedas de Obsidian (22 herramientas MCP), APIs de Batch y Files compatibles con OpenAI, caché semántica de respuestas, gamificación con clasificaciones, descubrimiento de agentes ACP (15 agentes integrados), exportación programada de registros a BigQuery, distribución paralela de múltiples modelos `auto/chaos`, un puente para bots de Telegram, un administrador de versiones integrado en la aplicación y clasificaciones de proveedores gratuitos de LMArena-ELO. → [Documentación](docs/README.md) @@ -689,11 +689,11 @@ del historial de tu shell. → [Integraciones CLI](docs/guides/CLI-INTEGRATIONS.
-## 🌐 383 proveedores de IA — 153 marcados como gratuitos en el catálogo +## 🌐 384 proveedores de IA — 153 marcados como gratuitos en el catálogo
-> **383 proveedores registrados** en las colecciones canónicas de chat, multimedia, búsqueda, local, agentes en la nube y sistema, incluidos **153 con metadatos de descubrimiento `hasFree: true`**. El registro de modelos de chat cubre **239 proveedores / 3,059 pares distintos de proveedor-modelo / 1,475 ID de modelos sin procesar**; el catálogo independiente de presupuesto gratuito contiene **471 filas por modelo**, **34 grupos recurrentes** y **52 proveedores gratuitos recurrentes/sin clave para siempre**. Estos son denominadores diferentes por diseño; las definiciones y los cálculos sin duplicar grupos se encuentran en la [Referencia de proveedores](docs/reference/PROVIDER_REFERENCE.md) y [Niveles gratuitos](docs/reference/FREE_TIERS.md). +> **384 proveedores registrados** en las colecciones canónicas de chat, multimedia, búsqueda, local, agentes en la nube y sistema, incluidos **153 con metadatos de descubrimiento `hasFree: true`**. El registro de modelos de chat cubre **239 proveedores / 3,059 pares distintos de proveedor-modelo / 1,475 ID de modelos sin procesar**; el catálogo independiente de presupuesto gratuito contiene **471 filas por modelo**, **34 grupos recurrentes** y **52 proveedores gratuitos recurrentes/sin clave para siempre**. Estos son denominadores diferentes por diseño; las definiciones y los cálculos sin duplicar grupos se encuentran en la [Referencia de proveedores](docs/reference/PROVIDER_REFERENCE.md) y [Niveles gratuitos](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ Métricas canónicas al 2026-10-18: **1.976 vídeos únicos** · **20.220.289 vi Entorno de ejecuciónNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 LenguajeTypeScript 6.0 — 100% TypeScript en src/ y open-sse/ (cero any en el núcleo desde v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Base de datosbetter-sqlite3 (SQLite, registro WAL) + LowDB (JSON heredado) — 147 módulos de dominio, 230 migraciones + Base de datosbetter-sqlite3 (SQLite, registro WAL) + LowDB (JSON heredado) — 147 módulos de dominio, 232 migraciones MemoriaBúsqueda de texto completo SQLite FTS5 + embeddings vectoriales cuantizados a int8, decaimiento tipado EsquemasZod 4 — validación de E/S de herramientas MCP + contratos de API ProtocolosMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/es/llm.txt b/docs/i18n/es/llm.txt index 523aa763a21..d2b6acc8697 100644 --- a/docs/i18n/es/llm.txt +++ b/docs/i18n/es/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/et/README.md b/docs/i18n/et/README.md index 384f3884121..c178e5f552d 100644 --- a/docs/i18n/et/README.md +++ b/docs/i18n/et/README.md @@ -15,7 +15,7 @@ **Loonud [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Ära kunagi lõpeta kodeerimist. Iga AI-tööriist → 383 pakkujat — neist 150+ tasuta — ühe lõpp-punkti kaudu. Claude Code, Codex, Cursor, Cline, Copilot ja Antigravity tasuta Claude'i / GPT / Gemini teenustesse koos automaatse varulahendusega. RTK + Cavemani virnastatud tihendus säästab 15–95% žetoone (keskmiselt ~89%) — ära kunagi ületa limiite. 383 AI-pakkujat · 150+ tasuta taset · ~1,62 miljardit tasuta žetooni kuus · 20 marsruutimisstrateegiat · alustamine hinnaga 0 $. +OmniRoute — Ära kunagi lõpeta kodeerimist. Iga AI-tööriist → 384 pakkujat — neist 150+ tasuta — ühe lõpp-punkti kaudu. Claude Code, Codex, Cursor, Cline, Copilot ja Antigravity tasuta Claude'i / GPT / Gemini teenustesse koos automaatse varulahendusega. RTK + Cavemani virnastatud tihendus säästab 15–95% žetoone (keskmiselt ~89%) — ära kunagi ületa limiite. 384 AI-pakkujat · 150+ tasuta taset · ~1,62 miljardit tasuta žetooni kuus · 20 marsruutimisstrateegiat · alustamine hinnaga 0 $.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ---------------------------------- | :------------------------: | :---------: | :-------------------: | -| 🌐 Pakkujad | 352 | **358** | 383 | +| 🌐 Pakkujad | 352 | **358** | 384 | | 🧠 Unikaalsed vestlusmudelite ID-d | 1320 | **1374** | 1491 | | 🖼️ Modaalsuse sild | 🆕 nägemine + heli + video | ✓ | ✓ | | 📡 Radari tasuta kataloog | 🆕 lubamispõhine | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funktsioonid 🎯 Kombod - 🌐 Teenusepakkujad + 🌐 Teenusepakkujad 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Lubadus — üks lõpp-punkt ja 383 teenusepakkujat. Automaatne varumarsruutimine hoiab marsruutimise toimivana seni, kuni saadaval on mõni teine terve sihtmärk. Kuus alustala: vastupidav varumarsruutimine 383 teenusepakkuja vahel · kuni 95% tokenite kokkuhoidu sobilike töökoormuste korral · alustamiseks $0, üle 150 tasuta taseme ning 52 korduva või võtmeta piiramatult tasuta teenusepakkujat · 39 CLI-/agendintegratsiooni ühe seadistuse kaudu · OpenAI, Claude'i, Gemini ja Responses API ühilduvus aadressil /v1 · tootmiskeskkonna juhtfunktsioonid, sealhulgas kaitselülitid, TLS-i varjamine, MCP 110 tööriistaga, A2A, mälu, kaitsepiirded, hindamised ning üle 56 000 staatilise testideklaratsiooni enam kui 7800 jälgitavas testifailis. +Lubadus — üks lõpp-punkt ja 384 teenusepakkujat. Automaatne varumarsruutimine hoiab marsruutimise toimivana seni, kuni saadaval on mõni teine terve sihtmärk. Kuus alustala: vastupidav varumarsruutimine 384 teenusepakkuja vahel · kuni 95% tokenite kokkuhoidu sobilike töökoormuste korral · alustamiseks $0, üle 150 tasuta taseme ning 52 korduva või võtmeta piiramatult tasuta teenusepakkujat · 39 CLI-/agendintegratsiooni ühe seadistuse kaudu · OpenAI, Claude'i, Gemini ja Responses API ühilduvus aadressil /v1 · tootmiskeskkonna juhtfunktsioonid, sealhulgas kaitselülitid, TLS-i varjamine, MCP 110 tööriistaga, A2A, mälu, kaitsepiirded, hindamised ning üle 56 000 staatilise testideklaratsiooni enam kui 7800 jälgitavas testifailis.

@@ -510,7 +510,7 @@ Kõik **20** strateegiat — kombineeri neid kombo igas etapis: -Mille poolest OmniRoute silma paistab — kuupäevaga funktsioonide ülevaade võrreldes 9routeri, OpenRouteri, CLIProxyAPI ja LiteLLM-iga 13 võimekuse lõikes. OmniRoute: 383 pakkujat, üle 150 sisseehitatud tasuta taset, 20 marsruutimisstrateegiat, 12-mootoriline tokenite tihendamine, sisseehitatud MCP-server 110 tööriistaga, A2A agendiprotokoll, püsiv mälu, kaitsepiirded, pilveagendid, TLS-i sõrmejälje varjamine, Desktop/Termux/PWA ja 67 i18n-i kasutajaliidese lokaati. OmniRoute on MIT-litsentsiga ja ise hostitav. Konkurentide võimalused ja arvud võivad muutuda; vaadake lingitud metoodikat. +Mille poolest OmniRoute silma paistab — kuupäevaga funktsioonide ülevaade võrreldes 9routeri, OpenRouteri, CLIProxyAPI ja LiteLLM-iga 13 võimekuse lõikes. OmniRoute: 384 pakkujat, üle 150 sisseehitatud tasuta taset, 20 marsruutimisstrateegiat, 12-mootoriline tokenite tihendamine, sisseehitatud MCP-server 110 tööriistaga, A2A agendiprotokoll, püsiv mälu, kaitsepiirded, pilveagendid, TLS-i sõrmejälje varjamine, Desktop/Termux/PWA ja 67 i18n-i kasutajaliidese lokaati. OmniRoute on MIT-litsentsiga ja ise hostitav. Konkurentide võimalused ja arvud võivad muutuda; vaadake lingitud metoodikat. 📊 Täielik metoodika ja funktsioonide üksikasjad võrreldes 9routeri, OpenRouteri, CLIProxyAPI ja LiteLLM-iga → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -614,7 +614,7 @@ praeguse kataloogi kohta leiate teavet aadressilt **[radar.omniroute.online/plan - **🖼️ Uued lõpp-punktid** — `/v1/ocr` (Mistral OCR) ja `/v1/audio/translations` (Whisperi-laadne) täiendavad meedialiidest. → [API viide](docs/reference/API_REFERENCE.md) - **🎨 Piltide / video / heli genereerimine** — üks meediateenuste API: xAI Grok Imagine ja Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind ning kõnepakkujad nagu ElevenLabs. → [API viide](docs/reference/API_REFERENCE.md) - **🌍 Juurutamine ja käitamine** — pöördpuhverserveri `basePath`, brauseri keele automaattuvastus, võtme-põhine seadmete jälgimine, juurkasutajata MITM-i usaldus ja zh-TW lokaliseerimine. → [Keskkond](docs/reference/ENVIRONMENT.md) -- **🤝 Rohkem pakkujaid ja agente** — pilveagendid (Codex Cloud, Cursor, Devin, Jules), brauseri ja OAuthi sisselogimisega Grok Build (xAI), esmaklassiline Ollama kaart, Claude Opus 5 ja Sonnet 5, Kimi ametlik partnerlus (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… ning värskendatud **383 pakkuja kataloog**. → [Pakkujad](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Rohkem pakkujaid ja agente** — pilveagendid (Codex Cloud, Cursor, Devin, Jules), brauseri ja OAuthi sisselogimisega Grok Build (xAI), esmaklassiline Ollama kaart, Claude Opus 5 ja Sonnet 5, Kimi ametlik partnerlus (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… ning värskendatud **384 pakkuja kataloog**. → [Pakkujad](docs/reference/PROVIDER_REFERENCE.md) - **📡 Marsruutimise läbipaistvus** — iga vastus sisaldab `X-OmniRoute-Decision` päist, mis nimetab seda teenindanud strateegia/pakkuja/latentsuse; uus `cache-optimized` kombineeritud strateegia ja Auto-Combo `cacheAffinity` tegur suunavad korduspäringud tagasi ühendusele, millel on vahemällu salvestatud prefiks; kirjutuskaitstud `/v1/auto-combo/{channel}/candidates` lõpp-punkt avaldab `auto/*` kanali aktiivse kandidaatide kogumi. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Kohalik jõudlus ja taristu** — ühe klõpsuga kohalik Redis, Cloudflare Workersi / Deno Deployi relee juurutajad, Bifrost ja Mux järelevalve all töötavate manustatud teenustena. → [Manustatud teenused](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Samuti sisaldub komplektis** — pluginate raamistik ja turg, Omni/Agenti/GitHubi oskuste raamistikud, Obsidiani varamu integratsioon (22 MCP-tööriista), OpenAI-ga ühilduvad Batchi ja Filesi API-d, semantiline vastuste vahemälu, edetabelitega mängustamine, ACP-agentide tuvastus (15 sisseehitatud agenti), ajastatud logide eksport BigQuerysse, `auto/chaos` paralleelne mitme mudeli fänn-out, Telegrami roboti sild, rakendusesisene versioonihaldur ja LMArena-ELO tasuta pakkujate edetabelid. → [Dokumentatsioon](docs/README.md) @@ -698,11 +698,11 @@ eemal sinu shelli ajaloost. → [CLI integratsioonid](docs/guides/CLI-INTEGRATIO
-## 🌐 383 tehisintellekti pakkujat — 153 kataloogis märgitud tasuta +## 🌐 384 tehisintellekti pakkujat — 153 kataloogis märgitud tasuta
-> **383 registreeritud pakkujat** kanoonilistes vestluse, meedia, otsingu, kohalikes, pilveagentide ja süsteemikogudes, sealhulgas **153 pakkujat, mille avastamise metaandmed sisaldavad `hasFree: true`**. Vestlusmudelite register hõlmab **239 pakkujat / 3 059 erinevat pakkuja-mudeli paari / 1 475 algset mudeli ID-d**; eraldi tasuta eelarve kataloogis on **471 mudelipõhist rida**, **34 korduvat kogumit** ja **52 korduvat või võtmeta igavesti tasuta pakkujat**. Need ongi kavandatult erinevad nimetajad; määratlused ja kogumite duplikaatide eemaldamisega arvutused on toodud [pakkujate teatmikus](docs/reference/PROVIDER_REFERENCE.md) ja [tasuta pakettide](docs/reference/FREE_TIERS.md) ülevaates. +> **384 registreeritud pakkujat** kanoonilistes vestluse, meedia, otsingu, kohalikes, pilveagentide ja süsteemikogudes, sealhulgas **153 pakkujat, mille avastamise metaandmed sisaldavad `hasFree: true`**. Vestlusmudelite register hõlmab **239 pakkujat / 3 059 erinevat pakkuja-mudeli paari / 1 475 algset mudeli ID-d**; eraldi tasuta eelarve kataloogis on **471 mudelipõhist rida**, **34 korduvat kogumit** ja **52 korduvat või võtmeta igavesti tasuta pakkujat**. Need ongi kavandatult erinevad nimetajad; määratlused ja kogumite duplikaatide eemaldamisega arvutused on toodud [pakkujate teatmikus](docs/reference/PROVIDER_REFERENCE.md) ja [tasuta pakettide](docs/reference/FREE_TIERS.md) ülevaates.
@@ -1307,7 +1307,7 @@ Kanoonilised mõõdikud seisuga 2026-10-18: **1.976 unikaalset videot** · **20. KäituskeskkondNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 KeelTypeScript 6.0 — 100% TypeScript kataloogides src/ ja open-sse/ (tuumas pole alates versioonist v2.0 ühtki any-t) RaamistikNext.js 16 + React 19 + Tailwind CSS 4 - Andmebaasbetter-sqlite3 (SQLite, WAL-logimine) + LowDB (JSON-i pärand) — 147 domeenimoodulit, 230 migratsiooni + Andmebaasbetter-sqlite3 (SQLite, WAL-logimine) + LowDB (JSON-i pärand) — 147 domeenimoodulit, 232 migratsiooni MäluSQLite FTS5 täistekstiotsing + int8-kvantiseeritud vektormanused, tüübistatud kahanemine SkeemidZod 4 — MCP-tööriistade sisendi/väljundi valideerimine + API lepingud ProtokollidMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/et/llm.txt b/docs/i18n/et/llm.txt index d5e75f97ef5..a5e7dbee8bb 100644 --- a/docs/i18n/et/llm.txt +++ b/docs/i18n/et/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/fa/README.md b/docs/i18n/fa/README.md index b5bad5c2f7e..cf1b5c2ae99 100644 --- a/docs/i18n/fa/README.md +++ b/docs/i18n/fa/README.md @@ -15,7 +15,7 @@ **ساختهشده توسط [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — هرگز کدنویسی را متوقف نکنید. هر ابزار هوش مصنوعی → 383 ارائهدهنده — 150+ مورد رایگان — از طریق یک نقطه پایانی. Claude Code، Codex، Cursor، Cline، Copilot و Antigravity به Claude / GPT / Gemini رایگان با بازگشت خودکار. فشردهسازی ترکیبی RTK + Caveman بین 15 تا 95٪ از توکنها (~89٪ میانگین) صرفهجویی میکند — هرگز به محدودیتها نرسید. 383 ارائهدهنده هوش مصنوعی · 150+ سطح رایگان · ~1.62B توکن رایگان در ماه · 20 راهبرد مسیریابی · شروع با 0 دلار. +OmniRoute — هرگز کدنویسی را متوقف نکنید. هر ابزار هوش مصنوعی → 384 ارائهدهنده — 150+ مورد رایگان — از طریق یک نقطه پایانی. Claude Code، Codex، Cursor، Cline، Copilot و Antigravity به Claude / GPT / Gemini رایگان با بازگشت خودکار. فشردهسازی ترکیبی RTK + Caveman بین 15 تا 95٪ از توکنها (~89٪ میانگین) صرفهجویی میکند — هرگز به محدودیتها نرسید. 384 ارائهدهنده هوش مصنوعی · 150+ سطح رایگان · ~1.62B توکن رایگان در ماه · 20 راهبرد مسیریابی · شروع با 0 دلار.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ---------------------------- | :---------------------: | :---------: | :-------------------: | -| 🌐 ارائهدهندگان | 352 | **358** | 383 | +| 🌐 ارائهدهندگان | 352 | **358** | 384 | | 🧠 شناسههای یکتای مدل گفتوگو | 1320 | **1374** | 1491 | | 🖼️ پل چندوجهی | 🆕 بینایی + صوت + ویدئو | ✓ | ✓ | | 📡 کاتالوگ رایگان رادار | 🆕 اختیاری | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ قابلیتها 🎯 ترکیبها - 🌐 ارائهدهندگان + 🌐 ارائهدهندگان 🔌 CLI و MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -وعده — یک نقطه پایانی و 383 ارائهدهنده. زمانی که یک هدف سالم دیگر در دسترس باشد، بازگشت خودکار مسیریابی را ادامه میدهد. شش ستون: بازگشتپذیری پایدار در میان 383 ارائهدهنده · صرفهجویی حداکثر 95٪ در توکن برای بارهای کاری واجد شرایط · شروع با هزینه 0 دلار، با بیش از 150 سطح رایگان و 52 ارائهدهنده رایگان دائمیِ تکرارشونده/بدون کلید · 39 یکپارچهسازی CLI/عامل از طریق یک پیکربندی · سازگاری با OpenAI، Claude، Gemini و Responses API در /v1 · کنترلهای تولیدی شامل قطعکنندههای مدار، پنهانسازی TLS، 110 ابزار MCP، A2A، حافظه، محافظها، ارزیابیها و بیش از 56,000 اعلان تست ایستا در بیش از 7,800 فایل تست تحت رهگیری. +وعده — یک نقطه پایانی و 384 ارائهدهنده. زمانی که یک هدف سالم دیگر در دسترس باشد، بازگشت خودکار مسیریابی را ادامه میدهد. شش ستون: بازگشتپذیری پایدار در میان 384 ارائهدهنده · صرفهجویی حداکثر 95٪ در توکن برای بارهای کاری واجد شرایط · شروع با هزینه 0 دلار، با بیش از 150 سطح رایگان و 52 ارائهدهنده رایگان دائمیِ تکرارشونده/بدون کلید · 39 یکپارچهسازی CLI/عامل از طریق یک پیکربندی · سازگاری با OpenAI، Claude، Gemini و Responses API در /v1 · کنترلهای تولیدی شامل قطعکنندههای مدار، پنهانسازی TLS، 110 ابزار MCP، A2A، حافظه، محافظها، ارزیابیها و بیش از 56,000 اعلان تست ایستا در بیش از 7,800 فایل تست تحت رهگیری.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -چه چیزی OmniRoute را متمایز میکند — تصویری بهروز از قابلیتها در مقایسه با 9router، OpenRouter، CLIProxyAPI و LiteLLM در ۱۳ قابلیت. OmniRoute: 383 ارائهدهنده، بیش از 150 سطح رایگان داخلی، 20 راهبرد مسیریابی، فشردهسازی توکن با 12 موتور، سرور MCP داخلی با 110 ابزار، پروتکل عامل A2A، حافظه پایدار، محافظها، عاملهای ابری، پنهانسازی اثرانگشت TLS، Desktop/Termux/PWA و 67 زبان محلی رابط کاربری i18n. OmniRoute تحت مجوز MIT است و میتوان آن را روی زیرساخت شخصی میزبانی کرد. قابلیتها و تعدادهای مربوط به رقبا ممکن است تغییر کنند؛ روششناسی پیوندشده را ببینید. +چه چیزی OmniRoute را متمایز میکند — تصویری بهروز از قابلیتها در مقایسه با 9router، OpenRouter، CLIProxyAPI و LiteLLM در ۱۳ قابلیت. OmniRoute: 384 ارائهدهنده، بیش از 150 سطح رایگان داخلی، 20 راهبرد مسیریابی، فشردهسازی توکن با 12 موتور، سرور MCP داخلی با 110 ابزار، پروتکل عامل A2A، حافظه پایدار، محافظها، عاملهای ابری، پنهانسازی اثرانگشت TLS، Desktop/Termux/PWA و 67 زبان محلی رابط کاربری i18n. OmniRoute تحت مجوز MIT است و میتوان آن را روی زیرساخت شخصی میزبانی کرد. قابلیتها و تعدادهای مربوط به رقبا ممکن است تغییر کنند؛ روششناسی پیوندشده را ببینید. 📊 روششناسی کامل و جزئیات هر قابلیت در مقایسه با 9router، OpenRouter، CLIProxyAPI و LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ OmniRoute دارای مجوز MIT است و بهصورت عمومی نگهدار - **🖼️ endpointهای جدید** — `/v1/ocr` (Mistral OCR) و `/v1/audio/translations` (سبک Whisper) سطح قابلیتهای رسانهای را کامل میکنند. → [مرجع API](docs/reference/API_REFERENCE.md) - **🎨 تولید تصویر / ویدئو / صدا** — یک API برای رسانه: xAI Grok Imagine و ویدئوی Novita AI، ComfyUI، Magnific، Adobe Firefly، Segmind و providerهای گفتار مانند ElevenLabs. → [مرجع API](docs/reference/API_REFERENCE.md) - **🌍 استقرار و عملیات** — `basePath` برای reverse-proxy، تشخیص خودکار زبان مرورگر، ردیابی دستگاه برای هر کلید، اعتماد MITM بدون root و بومیسازی zh-TW. → [محیط](docs/reference/ENVIRONMENT.md) -- **🤝 providerها و agentهای بیشتر** — agentهای ابری (Codex Cloud، Cursor، Devin، Jules)، Grok Build (xAI) با مرورگر و ورود OAuth، کارت اختصاصی Ollama، Claude Opus 5 و Sonnet 5، همکاری رسمی Kimi (Code/Web/Moonshot)، Zed، Requesty، SenseNova، Yuanbao، Agnes AI… و کاتالوگ بهروزشده **۳۸۳ provider**. → [Providerها](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 providerها و agentهای بیشتر** — agentهای ابری (Codex Cloud، Cursor، Devin، Jules)، Grok Build (xAI) با مرورگر و ورود OAuth، کارت اختصاصی Ollama، Claude Opus 5 و Sonnet 5، همکاری رسمی Kimi (Code/Web/Moonshot)، Zed، Requesty، SenseNova، Yuanbao، Agnes AI… و کاتالوگ بهروزشده **۳۸۴ provider**. → [Providerها](docs/reference/PROVIDER_REFERENCE.md) - **📡 شفافیت مسیریابی** — هر پاسخ یک هدر `X-OmniRoute-Decision` دارد که strategy/provider/latency ارائهدهنده آن را مشخص میکند؛ یک strategy ترکیبی جدید با نام `cache-optimized` بههمراه عامل `cacheAffinity` در Auto-Combo، درخواستهای تکراری را به اتصالی که prefix کششده را در اختیار دارد بازمیگرداند؛ و endpoint فقطخواندنی `/v1/auto-combo/{channel}/candidates` مجموعه کاندیداهای زنده یک channel از نوع `auto/*` را نمایش میدهد. → [ترکیب خودکار](docs/routing/AUTO-COMBO.md) - **⚡ عملکرد محلی و زیرساخت** — راهاندازی Redis محلی با یک کلیک، استقراردهندههای رله برای Cloudflare Workers / Deno Deploy، و Bifrost و Mux بهعنوان سرویسهای تعبیهشده تحت نظارت. → [سرویسهای تعبیهشده](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 همچنین در بسته** — چارچوب افزونه + بازارچه، چارچوبهای مهارت Omni/Agent/GitHub، یکپارچهسازی با مخزن Obsidian (۲۲ ابزار MCP)، APIهای Batch و Files سازگار با OpenAI، حافظهٔ نهان معنایی پاسخ، بازیوارسازی با جدولهای رتبهبندی، کشف عامل ACP (۱۵ عامل داخلی)، صدور زمانبندیشدهٔ گزارشها به BigQuery، توزیع موازی چندمدلی `auto/chaos`، پل ربات Telegram، مدیر نسخهٔ درونبرنامهای و رتبهبندی ارائهدهندگان رایگان LMArena-ELO. → [مستندات](docs/README.md) @@ -689,11 +689,11 @@ omniroute configure codex # همچنین: claude opencode qwen aider go
-## 🌐 383 ارائهدهندهٔ هوش مصنوعی — 153 مورد علامتگذاریشده بهعنوان رایگان در کاتالوگ +## 🌐 384 ارائهدهندهٔ هوش مصنوعی — 153 مورد علامتگذاریشده بهعنوان رایگان در کاتالوگ
-> **383 ارائهدهندهٔ ثبتشده** در مجموعههای مرجع گفتوگو، رسانه، جستوجو، محلی، عامل ابری و سیستم؛ از جمله **153 ارائهدهنده که متادیتای شناسایی `hasFree: true` را دارند**. رجیستری مدلهای گفتوگو شامل **239 ارائهدهنده / 3,059 جفت متمایز ارائهدهنده-مدل / 1,475 شناسهٔ خام مدل** است؛ کاتالوگ جداگانهٔ بودجهٔ رایگان نیز دارای **471 ردیف برای هر مدل**، **34 استخر تکرارشونده** و **52 ارائهدهندهٔ رایگان همیشگیِ تکرارشونده/بدون کلید** است. اینها عمداً مخرجهای متفاوتی هستند؛ تعریفها و محاسبات حذف تکرار بر اساس استخر در [مرجع ارائهدهندگان](docs/reference/PROVIDER_REFERENCE.md) و [سطوح رایگان](docs/reference/FREE_TIERS.md) قرار دارند. +> **384 ارائهدهندهٔ ثبتشده** در مجموعههای مرجع گفتوگو، رسانه، جستوجو، محلی، عامل ابری و سیستم؛ از جمله **153 ارائهدهنده که متادیتای شناسایی `hasFree: true` را دارند**. رجیستری مدلهای گفتوگو شامل **239 ارائهدهنده / 3,059 جفت متمایز ارائهدهنده-مدل / 1,475 شناسهٔ خام مدل** است؛ کاتالوگ جداگانهٔ بودجهٔ رایگان نیز دارای **471 ردیف برای هر مدل**، **34 استخر تکرارشونده** و **52 ارائهدهندهٔ رایگان همیشگیِ تکرارشونده/بدون کلید** است. اینها عمداً مخرجهای متفاوتی هستند؛ تعریفها و محاسبات حذف تکرار بر اساس استخر در [مرجع ارائهدهندگان](docs/reference/PROVIDER_REFERENCE.md) و [سطوح رایگان](docs/reference/FREE_TIERS.md) قرار دارند.
@@ -1295,7 +1295,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 نیز از آن زمان اجراNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 زبانTypeScript 6.0 — ۱۰۰٪ TypeScript در سراسر src/ و open-sse/ (بدون any در هسته از نسخهٔ v2.0) چارچوبNext.js 16 + React 19 + Tailwind CSS 4 - پایگاه دادهbetter-sqlite3 (SQLite، ثبت رویداد WAL) + LowDB (JSON قدیمی) — ۱۴۷ ماژول دامنه، ۲۳۰ مهاجرت + پایگاه دادهbetter-sqlite3 (SQLite، ثبت رویداد WAL) + LowDB (JSON قدیمی) — ۱۴۷ ماژول دامنه، ۲۳۲ مهاجرت حافظهجستوجوی تماممتن SQLite FTS5 + embeddingهای برداری کوانتیزهشده با int8، زوال نوعدار طرحوارههاZod 4 — اعتبارسنجی ورودی/خروجی ابزار MCP + قراردادهای API پروتکلهاMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/fa/llm.txt b/docs/i18n/fa/llm.txt index abd779b8dec..95d5d1baa25 100644 --- a/docs/i18n/fa/llm.txt +++ b/docs/i18n/fa/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/fi/README.md b/docs/i18n/fi/README.md index c8bf0cace31..ff53b4f5a9a 100644 --- a/docs/i18n/fi/README.md +++ b/docs/i18n/fi/README.md @@ -15,7 +15,7 @@ **Tekijä: [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Älä lopeta koodaamista. Jokainen tekoälytyökalu → 383 palveluntarjoajaa — yli 150 ilmaista — yhden päätepisteen kautta. Claude Code, Codex, Cursor, Cline, Copilot ja Antigravity → ILMAINEN Claude / GPT / Gemini automaattisella varajärjestelmällä. RTK + Caveman -pinottu pakkaus säästää 15–95 % tokeneista (keskimäärin noin 89 %) — rajat eivät tule koskaan vastaan. 383 tekoälypalveluntarjoajaa · yli 150 ilmaista tasoa · noin 1,62 miljardia ilmaista tokenia kuukaudessa · 20 reititysstrategiaa · aloitus 0 dollarilla. +OmniRoute — Älä lopeta koodaamista. Jokainen tekoälytyökalu → 384 palveluntarjoajaa — yli 150 ilmaista — yhden päätepisteen kautta. Claude Code, Codex, Cursor, Cline, Copilot ja Antigravity → ILMAINEN Claude / GPT / Gemini automaattisella varajärjestelmällä. RTK + Caveman -pinottu pakkaus säästää 15–95 % tokeneista (keskimäärin noin 89 %) — rajat eivät tule koskaan vastaan. 384 tekoälypalveluntarjoajaa · yli 150 ilmaista tasoa · noin 1,62 miljardia ilmaista tokenia kuukaudessa · 20 reititysstrategiaa · aloitus 0 dollarilla.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------------------- | :--------------------: | :---------: | :-------------------: | -| 🌐 Palveluntarjoajat | 352 | **358** | 383 | +| 🌐 Palveluntarjoajat | 352 | **358** | 384 | | 🧠 Yksilölliset keskustelumallien tunnukset | 1320 | **1374** | 1491 | | 🖼️ Modaliteettisilta | 🆕 näkö + ääni + video | ✓ | ✓ | | 📡 Radarin maksuton luettelo | 🆕 opt-in | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Ominaisuudet 🎯 Yhdistelmät - 🌐 Palveluntarjoajat + 🌐 Palveluntarjoajat 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Lupaus — yksi päätepiste ja 383 palveluntarjoajaa. Automaattinen varareititys pitää reitityksen toiminnassa, kun toinen terve kohde on käytettävissä. Kuusi pilaria: vikasietoinen varareititys 383 palveluntarjoajan välillä · jopa 95 %:n token-säästöt soveltuvissa työkuormissa · 0 $:sta alkaen, yli 150 ilmaistasoa ja 52 toistuvaa tai avaimetonta ikuisesti ilmaista palveluntarjoajaa · 39 CLI-/agentti-integraatiota yhden määrityksen kautta · OpenAI-, Claude-, Gemini- ja Responses API -yhteensopivuus osoitteessa /v1 · tuotantotason hallintaominaisuudet, kuten virtakatkaisimet, TLS-häivytys, MCP:n 110 työkalua, A2A, muisti, suojakaiteet, evaluoinnit sekä yli 56 000 staattista testimääritystä yli 7 800 seurattavassa testitiedostossa. +Lupaus — yksi päätepiste ja 384 palveluntarjoajaa. Automaattinen varareititys pitää reitityksen toiminnassa, kun toinen terve kohde on käytettävissä. Kuusi pilaria: vikasietoinen varareititys 384 palveluntarjoajan välillä · jopa 95 %:n token-säästöt soveltuvissa työkuormissa · 0 $:sta alkaen, yli 150 ilmaistasoa ja 52 toistuvaa tai avaimetonta ikuisesti ilmaista palveluntarjoajaa · 39 CLI-/agentti-integraatiota yhden määrityksen kautta · OpenAI-, Claude-, Gemini- ja Responses API -yhteensopivuus osoitteessa /v1 · tuotantotason hallintaominaisuudet, kuten virtakatkaisimet, TLS-häivytys, MCP:n 110 työkalua, A2A, muisti, suojakaiteet, evaluoinnit sekä yli 56 000 staattista testimääritystä yli 7 800 seurattavassa testitiedostossa.

@@ -510,7 +510,7 @@ Kaikki **20** strategiaa — yhdistele vapaasti yhdistelmän jokaisessa vaiheess -Mikä erottaa OmniRouten muista — päivätty ominaisuusvertailu 9routeriin, OpenRouteriin, CLIProxyAPI:in ja LiteLLM:ään nähden 13 ominaisuudessa. OmniRoute: 383 palveluntarjoajaa, yli 150 ilmaista käyttöoikeustasoa sisäänrakennettuna, 20 reititysstrategiaa, tunnusten pakkaus 12 moottorilla, sisäänrakennettu MCP-palvelin, jossa on 110 työkalua, A2A-agenttiprotokolla, pysyvä muisti, suojaukset, pilviagentit, TLS-sormenjäljen häivytys, Desktop/Termux/PWA ja 67 käyttöliittymän i18n-kielialuetta. OmniRoute on MIT-lisensoitu ja itse isännöitävissä. Kilpailijoiden ominaisuudet ja määrät voivat muuttua; katso linkitetty menetelmäkuvaus. +Mikä erottaa OmniRouten muista — päivätty ominaisuusvertailu 9routeriin, OpenRouteriin, CLIProxyAPI:in ja LiteLLM:ään nähden 13 ominaisuudessa. OmniRoute: 384 palveluntarjoajaa, yli 150 ilmaista käyttöoikeustasoa sisäänrakennettuna, 20 reititysstrategiaa, tunnusten pakkaus 12 moottorilla, sisäänrakennettu MCP-palvelin, jossa on 110 työkalua, A2A-agenttiprotokolla, pysyvä muisti, suojaukset, pilviagentit, TLS-sormenjäljen häivytys, Desktop/Termux/PWA ja 67 käyttöliittymän i18n-kielialuetta. OmniRoute on MIT-lisensoitu ja itse isännöitävissä. Kilpailijoiden ominaisuudet ja määrät voivat muuttua; katso linkitetty menetelmäkuvaus. 📊 Täydellinen menetelmäkuvaus ja ominaisuuskohtaiset tiedot verrattuna 9routeriin, OpenRouteriin, CLIProxyAPI:in ja LiteLLM:ään → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -614,7 +614,7 @@ nykyisestä luettelosta on osoitteessa **[radar.omniroute.online/planos](https:/ - **🖼️ Uudet päätepisteet** — `/v1/ocr` (Mistral OCR) ja `/v1/audio/translations` (Whisper-tyylinen) täydentävät median käsittelyn rajapintaa. → [API-viite](docs/reference/API_REFERENCE.md) - **🎨 Kuvien, videoiden ja äänen luonti** — yksi media-API: xAI Grok Imagine ja Novita AI -video, ComfyUI, Magnific, Adobe Firefly, Segmind sekä puhepalveluntarjoajat, kuten ElevenLabs. → [API-viite](docs/reference/API_REFERENCE.md) - **🌍 Käyttöönotto ja operointi** — käänteisen välityspalvelimen `basePath`, selaimen kielen automaattinen tunnistus, avainkohtainen laitteiden seuranta, pääkäyttäjättömän MITM-luottamus ja zh-TW-lokalisointi. → [Ympäristö](docs/reference/ENVIRONMENT.md) -- **🤝 Lisää palveluntarjoajia ja agentteja** — pilviagentit (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) selain- ja OAuth-kirjautumisella, Ollaman ensiluokkainen kortti, Claude Opus 5 ja Sonnet 5, Kimin virallinen kumppanuus (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… sekä uudistettu **383 palveluntarjoajan luettelo**. → [Palveluntarjoajat](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Lisää palveluntarjoajia ja agentteja** — pilviagentit (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) selain- ja OAuth-kirjautumisella, Ollaman ensiluokkainen kortti, Claude Opus 5 ja Sonnet 5, Kimin virallinen kumppanuus (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… sekä uudistettu **384 palveluntarjoajan luettelo**. → [Palveluntarjoajat](docs/reference/PROVIDER_REFERENCE.md) - **📡 Reitityksen läpinäkyvyys** — jokainen vastaus sisältää `X-OmniRoute-Decision`-otsakkeen, joka ilmoittaa sitä palvelleen strategian, palveluntarjoajan ja viiveen; uusi `cache-optimized`-yhdistelmästrategia sekä Auto-Combossa oleva `cacheAffinity`-tekijän reitti toistuvat pyynnöt takaisin välimuistissa olevaa etuliitettä hallussaan pitävään yhteyteen, ja vain lukuun tarkoitettu `/v1/auto-combo/{channel}/candidates`-päätepiste paljastaa `auto/*`-kanavan reaaliaikaisen ehdokasjoukon. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Paikallinen suorituskyky ja infrastruktuuri** — paikallinen Redis yhdellä napsautuksella, Cloudflare Workers- / Deno Deploy -välityspalvelinten käyttöönottajat, Bifrost ja Mux valvottuina upotettuina palveluina. → [Upotetut palvelut](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Myös mukana** — lisäosakehys ja markkinapaikka, Omni/Agent/GitHub-taitokehykset, Obsidian-holvi-integraatio (22 MCP-työkalua), OpenAI-yhteensopivat Batch- ja Files-rajapinnat, semanttinen vastausvälimuisti, pelillistäminen tulostaulukoilla, ACP-agenttien haku (15 sisäänrakennettua agenttia), ajoitettu lokien vienti BigQueryyn, `auto/chaos`-rinnakkainen usean mallin fan-out, Telegram-bottisilta, sovelluksen sisäinen versionhallinta ja LMArena-ELO:n ilmaisten palveluntarjoajien sijoitukset. → [Dokumentaatio](docs/README.md) @@ -698,11 +698,11 @@ komentorivihistoriastasi. → [CLI-integraatiot](docs/guides/CLI-INTEGRATIONS.md
-## 🌐 383 tekoälypalveluntarjoajaa — 153 katalogiin merkittyä maksutonta +## 🌐 384 tekoälypalveluntarjoajaa — 153 katalogiin merkittyä maksutonta
-> **383 rekisteröityä palveluntarjoajaa** kanonisissa chat-, media-, haku-, paikallis-, pilviagentti- ja järjestelmäkokoelmissa, joista **153:lla on `hasFree: true` -hakumetatiedot**. Chat-mallirekisteri kattaa **239 palveluntarjoajaa / 3 059 erillistä palveluntarjoaja–malli-paria / 1 475 raakaa mallitunnistetta**; erillisessä ilmaisen budjetin luettelossa on **471 mallikohtaista riviä**, **34 toistuvaa allasta** ja **52 toistuvasti tai ilman avainta maksutta käytettävää palveluntarjoajaa**. Nämä ovat tarkoituksella eri nimittäjiä; määritelmät ja allasduplikaatit poistavat laskelmat ovat saatavilla [palveluntarjoajien viiteoppaassa](docs/reference/PROVIDER_REFERENCE.md) ja [ilmaisissa käyttöoikeustasoissa](docs/reference/FREE_TIERS.md). +> **384 rekisteröityä palveluntarjoajaa** kanonisissa chat-, media-, haku-, paikallis-, pilviagentti- ja järjestelmäkokoelmissa, joista **153:lla on `hasFree: true` -hakumetatiedot**. Chat-mallirekisteri kattaa **239 palveluntarjoajaa / 3 059 erillistä palveluntarjoaja–malli-paria / 1 475 raakaa mallitunnistetta**; erillisessä ilmaisen budjetin luettelossa on **471 mallikohtaista riviä**, **34 toistuvaa allasta** ja **52 toistuvasti tai ilman avainta maksutta käytettävää palveluntarjoajaa**. Nämä ovat tarkoituksella eri nimittäjiä; määritelmät ja allasduplikaatit poistavat laskelmat ovat saatavilla [palveluntarjoajien viiteoppaassa](docs/reference/PROVIDER_REFERENCE.md) ja [ilmaisissa käyttöoikeustasoissa](docs/reference/FREE_TIERS.md).
@@ -1304,7 +1304,7 @@ Kanoniset mittarit 2026-10-18: **1.976 yksilöllistä videota** · **20.220.289 AjoympäristöNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 KieliTypeScript 6.0 — 100 % TypeScriptiä hakemistoissa src/ ja open-sse/ (ytimessä ei yhtään any-tyyppiä versiosta v2.0 alkaen) KehysNext.js 16 + React 19 + Tailwind CSS 4 - Tietokantabetter-sqlite3 (SQLite, WAL-kirjaus) + LowDB (JSON-legacy) — 147 toimialamoduulia, 230 migraatiota + Tietokantabetter-sqlite3 (SQLite, WAL-kirjaus) + LowDB (JSON-legacy) — 147 toimialamoduulia, 232 migraatiota MuistiSQLite FTS5 -kokotekstihaku + int8-kvantisointia käyttävät vektoriupotukset, tyypitetty vanheneminen SkeematZod 4 — MCP-työkalujen I/O-validointi + API-sopimukset ProtokollatMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/fi/llm.txt b/docs/i18n/fi/llm.txt index 468fea7238c..ec9e91f301c 100644 --- a/docs/i18n/fi/llm.txt +++ b/docs/i18n/fi/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/fr/README.md b/docs/i18n/fr/README.md index cea8558ae3a..3f455b6441c 100644 --- a/docs/i18n/fr/README.md +++ b/docs/i18n/fr/README.md @@ -15,7 +15,7 @@ **Créé par [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Ne cessez jamais de coder. Chaque outil d’IA → 383 fournisseurs — dont 150+ gratuits — via un seul endpoint. Claude Code, Codex, Cursor, Cline, Copilot et Antigravity vers Claude / GPT / Gemini GRATUITS avec basculement automatique. La compression empilée RTK + Caveman économise 15–95 % de tokens (~89 % en moyenne) — ne rencontrez jamais de limites. 383 fournisseurs d’IA · 150+ offres gratuites · ~1,62 milliard de tokens gratuits/mois · 20 stratégies de routage · 0 $ pour commencer. +OmniRoute — Ne cessez jamais de coder. Chaque outil d’IA → 384 fournisseurs — dont 150+ gratuits — via un seul endpoint. Claude Code, Codex, Cursor, Cline, Copilot et Antigravity vers Claude / GPT / Gemini GRATUITS avec basculement automatique. La compression empilée RTK + Caveman économise 15–95 % de tokens (~89 % en moyenne) — ne rencontrez jamais de limites. 384 fournisseurs d’IA · 150+ offres gratuites · ~1,62 milliard de tokens gratuits/mois · 20 stratégies de routage · 0 $ pour commencer.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------------------ | :-----------------------: | :---------: | :----------------------: | -| 🌐 Fournisseurs | 352 | **358** | 383 | +| 🌐 Fournisseurs | 352 | **358** | 384 | | 🧠 Identifiants de modèles de chat uniques | 1320 | **1374** | 1491 | | 🖼️ Passerelle de modalités | 🆕 vision + audio + vidéo | ✓ | ✓ | | 📡 Catalogue gratuit du radar | 🆕 inscription volontaire | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Fonctionnalités 🎯 Combos - 🌐 Fournisseurs + 🌐 Fournisseurs 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -La promesse — Un endpoint et 383 fournisseurs. Le basculement automatique maintient le routage tant qu’une autre cible saine est disponible. Six piliers : basculement résilient entre 383 fournisseurs · jusqu’à 95 % d’économies de tokens sur les charges de travail éligibles · 0 $ pour commencer avec plus de 150 offres gratuites et 52 fournisseurs gratuits récurrents ou sans clé, pour toujours · 39 intégrations CLI/agent via une seule configuration · compatibilité avec les API OpenAI, Claude, Gemini et Responses à /v1 · contrôles de production, notamment disjoncteurs, furtivité TLS, 110 outils MCP, A2A, mémoire, garde-fous, évaluations et plus de 56 000 déclarations de tests statiques réparties dans plus de 7 800 fichiers de test suivis. +La promesse — Un endpoint et 384 fournisseurs. Le basculement automatique maintient le routage tant qu’une autre cible saine est disponible. Six piliers : basculement résilient entre 384 fournisseurs · jusqu’à 95 % d’économies de tokens sur les charges de travail éligibles · 0 $ pour commencer avec plus de 150 offres gratuites et 52 fournisseurs gratuits récurrents ou sans clé, pour toujours · 39 intégrations CLI/agent via une seule configuration · compatibilité avec les API OpenAI, Claude, Gemini et Responses à /v1 · contrôles de production, notamment disjoncteurs, furtivité TLS, 110 outils MCP, A2A, mémoire, garde-fous, évaluations et plus de 56 000 déclarations de tests statiques réparties dans plus de 7 800 fichiers de test suivis.

@@ -510,7 +510,7 @@ Les **20** stratégies — à combiner librement à chaque étape du combo : -Ce qui distingue OmniRoute — instantané daté des fonctionnalités comparant 9router, OpenRouter, CLIProxyAPI et LiteLLM sur 13 capacités. OmniRoute : 383 fournisseurs, plus de 150 niveaux gratuits intégrés, 20 stratégies de routage, compression des tokens avec 12 moteurs, serveur MCP intégré avec 110 outils, protocole d’agents A2A, mémoire persistante, garde-fous, agents cloud, furtivité de l’empreinte TLS, Desktop/Termux/PWA et 67 paramètres régionaux d’interface utilisateur i18nisés. OmniRoute est sous licence MIT et peut être auto-hébergé. Les capacités et les chiffres des concurrents peuvent évoluer ; consultez la méthodologie liée. +Ce qui distingue OmniRoute — instantané daté des fonctionnalités comparant 9router, OpenRouter, CLIProxyAPI et LiteLLM sur 13 capacités. OmniRoute : 384 fournisseurs, plus de 150 niveaux gratuits intégrés, 20 stratégies de routage, compression des tokens avec 12 moteurs, serveur MCP intégré avec 110 outils, protocole d’agents A2A, mémoire persistante, garde-fous, agents cloud, furtivité de l’empreinte TLS, Desktop/Termux/PWA et 67 paramètres régionaux d’interface utilisateur i18nisés. OmniRoute est sous licence MIT et peut être auto-hébergé. Les capacités et les chiffres des concurrents peuvent évoluer ; consultez la méthodologie liée. 📊 Méthodologie complète et détails par fonctionnalité comparant 9router, OpenRouter, CLIProxyAPI et LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar est facultatif et utilise uniquement des requêtes GET. Le client OmniRout - **🖼️ Nouveaux endpoints** — `/v1/ocr` (Mistral OCR) et `/v1/audio/translations` (de type Whisper) complètent l’offre multimédia. → [Référence de l’API](docs/reference/API_REFERENCE.md) - **🎨 Génération d’images, de vidéos et d’audio** — une API pour les médias : vidéo xAI Grok Imagine et Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind et fournisseurs vocaux tels qu’ElevenLabs. → [Référence de l’API](docs/reference/API_REFERENCE.md) - **🌍 Déploiement et exploitation** — `basePath` pour les reverse proxies, détection automatique de la langue du navigateur, suivi des appareils par clé, confiance MITM sans privilèges root et localisation zh-TW. → [Environnement](docs/reference/ENVIRONMENT.md) -- **🤝 Davantage de fournisseurs et d’agents** — agents cloud (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) avec navigateur et connexion OAuth, fiche Ollama de première classe, Claude Opus 5 et Sonnet 5, partenariat officiel avec Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… et un catalogue renouvelé de **383 fournisseurs**. → [Fournisseurs](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Davantage de fournisseurs et d’agents** — agents cloud (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) avec navigateur et connexion OAuth, fiche Ollama de première classe, Claude Opus 5 et Sonnet 5, partenariat officiel avec Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… et un catalogue renouvelé de **384 fournisseurs**. → [Fournisseurs](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparence du routage** — chaque réponse contient un en-tête `X-OmniRoute-Decision` indiquant la stratégie, le fournisseur et la latence qui l’ont servie ; une nouvelle stratégie combinée `cache-optimized` et le facteur `cacheAffinity` d’Auto-Combo réacheminent les requêtes répétées vers la connexion qui détient le préfixe mis en cache ; enfin, un endpoint en lecture seule `/v1/auto-combo/{channel}/candidates` expose le pool de candidats actif d’un canal `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Performances locales et infrastructure** — Redis local en un clic, déployeurs de relais Cloudflare Workers / Deno Deploy, Bifrost et Mux en tant que services intégrés supervisés. → [Services intégrés](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Également inclus** — framework de plugins + marketplace, frameworks de skills Omni/Agent/GitHub, intégration avec un coffre Obsidian (22 outils MCP), API Batch & Files compatibles avec OpenAI, cache sémantique des réponses, gamification avec classements, découverte d’agents ACP (15 agents intégrés), export planifié des journaux vers BigQuery, fan-out parallèle multi-modèles `auto/chaos`, pont avec un bot Telegram, gestionnaire de versions intégré à l’application et classements des fournisseurs gratuits LMArena-ELO. → [Documentation](docs/README.md) @@ -689,11 +689,11 @@ hors de l'historique de votre shell. → [Intégrations CLI](docs/guides/CLI-INT
-## 🌐 383 fournisseurs d’IA — 153 signalés comme gratuits dans le catalogue +## 🌐 384 fournisseurs d’IA — 153 signalés comme gratuits dans le catalogue
-> **383 fournisseurs enregistrés** dans les collections canoniques de chat, de médias, de recherche, locales, d’agents cloud et système, dont **153 avec les métadonnées de découverte `hasFree: true`**. Le registre des modèles de chat couvre **239 fournisseurs / 3 059 paires fournisseur-modèle distinctes / 1 475 identifiants de modèles bruts** ; le catalogue séparé des budgets gratuits contient **471 lignes par modèle**, **34 pools récurrents** et **52 fournisseurs récurrents/gratuits sans clé, pour toujours**. Ces dénominateurs diffèrent intentionnellement ; les définitions et les calculs dédupliqués par pool sont disponibles dans la [Référence des fournisseurs](docs/reference/PROVIDER_REFERENCE.md) et les [Offres gratuites](docs/reference/FREE_TIERS.md). +> **384 fournisseurs enregistrés** dans les collections canoniques de chat, de médias, de recherche, locales, d’agents cloud et système, dont **153 avec les métadonnées de découverte `hasFree: true`**. Le registre des modèles de chat couvre **239 fournisseurs / 3 059 paires fournisseur-modèle distinctes / 1 475 identifiants de modèles bruts** ; le catalogue séparé des budgets gratuits contient **471 lignes par modèle**, **34 pools récurrents** et **52 fournisseurs récurrents/gratuits sans clé, pour toujours**. Ces dénominateurs diffèrent intentionnellement ; les définitions et les calculs dédupliqués par pool sont disponibles dans la [Référence des fournisseurs](docs/reference/PROVIDER_REFERENCE.md) et les [Offres gratuites](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ Métriques canoniques au 2026-10-18 : **1.976 vidéos uniques** · **20.220.289 ExécutionNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 LangageTypeScript 6.0 — 100 % TypeScript dans src/ et open-sse/ (any absent du cœur depuis la v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Base de donnéesbetter-sqlite3 (SQLite, journalisation WAL) + LowDB (JSON legacy) — 147 modules métier, 230 migrations + Base de donnéesbetter-sqlite3 (SQLite, journalisation WAL) + LowDB (JSON legacy) — 147 modules métier, 232 migrations MémoireRecherche en texte intégral SQLite FTS5 + embeddings vectoriels quantifiés en int8, décroissance typée SchémasZod 4 — validation des entrées/sorties des outils MCP + contrats d’API ProtocolesMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/fr/llm.txt b/docs/i18n/fr/llm.txt index 81526e30e01..b5371b397ed 100644 --- a/docs/i18n/fr/llm.txt +++ b/docs/i18n/fr/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ga/README.md b/docs/i18n/ga/README.md index 5650f1d3ae9..6fb9d779ca4 100644 --- a/docs/i18n/ga/README.md +++ b/docs/i18n/ga/README.md @@ -15,7 +15,7 @@ **Cruthaithe ag [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Ná stop den chódú riamh. Gach uirlis AI → 383 soláthraí — 150+ saor in aisce — trí chríochphointe amháin. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity isteach i Claude / GPT / Gemini SAOR IN AISCE le cúltaca uathoibríoch. Sábhálann comhbhrú cruachta RTK + Caveman 15–95% de na comharthaí (~89% ar an meán) — ná sroich teorainneacha riamh. 383 soláthraí AI · 150+ sraith saor in aisce · ~1.62B comhartha saor in aisce/mí · 20 straitéis ródaithe · $0 chun tosú. +OmniRoute — Ná stop den chódú riamh. Gach uirlis AI → 384 soláthraí — 150+ saor in aisce — trí chríochphointe amháin. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity isteach i Claude / GPT / Gemini SAOR IN AISCE le cúltaca uathoibríoch. Sábhálann comhbhrú cruachta RTK + Caveman 15–95% de na comharthaí (~89% ar an meán) — ná sroich teorainneacha riamh. 384 soláthraí AI · 150+ sraith saor in aisce · ~1.62B comhartha saor in aisce/mí · 20 straitéis ródaithe · $0 chun tosú.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ---------------------------------------- | :---------------------: | :---------: | :-------------------: | -| 🌐 Soláthraithe | 352 | **358** | 383 | +| 🌐 Soláthraithe | 352 | **358** | 384 | | 🧠 Aitheantóirí uathúla samhlacha comhrá | 1320 | **1374** | 1491 | | 🖼️ Droichead ilmhódúlachta | 🆕 fís + fuaim + físeán | ✓ | ✓ | | 📡 Catalóg shaor Radar | 🆕 rogha isteach | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Gnéithe 🎯 Teaglamaí - 🌐 Soláthraithe + 🌐 Soláthraithe 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -An Gealltanas — Críochphointe amháin agus 383 soláthraí. Coinníonn cúltaca uathoibríoch an ródáil ar siúl fad is atá sprioc shláintiúil eile ar fáil. Sé cholún: cúltaca athléimneach trasna 383 soláthraí · suas le 95% coigilteas comharthaí ar ualaí oibre incháilithe · $0 le tosú le 150+ sraith saor agus 52 soláthraí saor in aisce go deo ar bhonn athfhillteach/gan eochair · 39 comhtháthú CLI/gníomhaire trí chumraíocht amháin · comhoiriúnacht le OpenAI, Claude, Gemini agus Responses API ag /v1 · rialuithe táirgeachta lena n-áirítear scoradáin chiorcaid, stealth TLS, MCP 110 uirlis, A2A, cuimhne, ráillí cosanta, meastóireachtaí agus 56,000+ dearbhú tástála statach trasna 7,800+ comhad tástála rianaithe. +An Gealltanas — Críochphointe amháin agus 384 soláthraí. Coinníonn cúltaca uathoibríoch an ródáil ar siúl fad is atá sprioc shláintiúil eile ar fáil. Sé cholún: cúltaca athléimneach trasna 384 soláthraí · suas le 95% coigilteas comharthaí ar ualaí oibre incháilithe · $0 le tosú le 150+ sraith saor agus 52 soláthraí saor in aisce go deo ar bhonn athfhillteach/gan eochair · 39 comhtháthú CLI/gníomhaire trí chumraíocht amháin · comhoiriúnacht le OpenAI, Claude, Gemini agus Responses API ag /v1 · rialuithe táirgeachta lena n-áirítear scoradáin chiorcaid, stealth TLS, MCP 110 uirlis, A2A, cuimhne, ráillí cosanta, meastóireachtaí agus 56,000+ dearbhú tástála statach trasna 7,800+ comhad tástála rianaithe.

@@ -510,7 +510,7 @@ Gach ceann de na **20** straitéis — measc agus meaitseáil iad ag gach céim -Cad a leagann OmniRoute amach — léargas gnéithe dátaithe i gcomparáid le 9router, OpenRouter, CLIProxyAPI agus LiteLLM thar 13 chumas. OmniRoute: 383 soláthraí, breis agus 150 sraith saor in aisce ionsuite, 20 straitéis ródaithe, comhbhrú comharthaí le 12 inneall, freastalaí MCP ionsuite le 110 uirlis, prótacal gníomhairí A2A, cuimhne bhuan, rialtáin, gníomhairí néil, duaithníocht méarlorg TLS, Deasc/Termux/PWA agus 67 logán UI i18n. Tá OmniRoute ceadúnaithe faoin MIT agus is féidir é a óstáil féin. D’fhéadfadh cumais agus comhaireamh na n-iomaitheoirí athrú; féach ar an modheolaíocht nasctha. +Cad a leagann OmniRoute amach — léargas gnéithe dátaithe i gcomparáid le 9router, OpenRouter, CLIProxyAPI agus LiteLLM thar 13 chumas. OmniRoute: 384 soláthraí, breis agus 150 sraith saor in aisce ionsuite, 20 straitéis ródaithe, comhbhrú comharthaí le 12 inneall, freastalaí MCP ionsuite le 110 uirlis, prótacal gníomhairí A2A, cuimhne bhuan, rialtáin, gníomhairí néil, duaithníocht méarlorg TLS, Deasc/Termux/PWA agus 67 logán UI i18n. Tá OmniRoute ceadúnaithe faoin MIT agus is féidir é a óstáil féin. D’fhéadfadh cumais agus comhaireamh na n-iomaitheoirí athrú; féach ar an modheolaíocht nasctha. 📊 Modheolaíocht iomlán agus sonraí in aghaidh na gné i gcomparáid le 9router, OpenRouter, CLIProxyAPI agus LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -614,7 +614,7 @@ faoin gcatalóg reatha ag **[radar.omniroute.online/planos](https://radar.omniro - **🖼️ Críochphointí nua** — comhlánaíonn `/v1/ocr` (Mistral OCR) agus `/v1/audio/translations` (stíl Whisper) an dromchla meán. → [Tagairt API](docs/reference/API_REFERENCE.md) - **🎨 Giniúint íomhánna / físeáin / fuaime** — API amháin do mheáin: físeán xAI Grok Imagine agus Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind, agus soláthraithe cainte ar nós ElevenLabs. → [Tagairt API](docs/reference/API_REFERENCE.md) - **🌍 Imscaradh agus oibríochtaí** — `basePath` seachfhreastalaí droim ar ais, uathbhraith teanga an bhrabhsálaí, rianú gléis in aghaidh na heochrach, muinín MITM gan fhréamh, logánú zh-TW. → [Timpeallacht](docs/reference/ENVIRONMENT.md) -- **🤝 Tuilleadh soláthraithe agus gníomhairí** — gníomhairí néil (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) le brabhsálaí + logáil isteach OAuth, cárta Ollama den chéad scoth, Claude Opus 5 agus Sonnet 5, comhpháirtíocht oifigiúil Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… agus catalóg athnuaite de **383 soláthraí**. → [Soláthraithe](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Tuilleadh soláthraithe agus gníomhairí** — gníomhairí néil (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) le brabhsálaí + logáil isteach OAuth, cárta Ollama den chéad scoth, Claude Opus 5 agus Sonnet 5, comhpháirtíocht oifigiúil Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… agus catalóg athnuaite de **384 soláthraí**. → [Soláthraithe](docs/reference/PROVIDER_REFERENCE.md) - **📡 Trédhearcacht ródaithe** — iompraíonn gach freagra ceanntásc `X-OmniRoute-Decision` a ainmníonn an straitéis/soláthraí/moille a sheirbheáil é; cuireann straitéis teaglaim nua `cache-optimized` + fachtóir Auto-Combo `cacheAffinity` iarratais athdhéanta ar ais chuig an nasc a bhfuil an réimír taiscthe aige; agus nochtann críochphointe inléite amháin `/v1/auto-combo/{channel}/candidates` linn bheo iarrthóirí cainéil `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md). - **⚡ Feidhmíocht áitiúil agus bonneagar** — Redis áitiúil le cliceáil amháin, imlonnóirí athsheachadáin Cloudflare Workers / Deno Deploy, agus Bifrost & Mux mar sheirbhísí leabaithe faoi mhaoirseacht. → [Seirbhísí Leabaithe](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Sa bhosca freisin** — creat breiseán + margadh, creataí scileanna Omni/Agent/GitHub, comhtháthú cruinneacháin Obsidian (22 uirlis MCP), APIs Batch & Files atá comhoiriúnach le OpenAI, taisce freagartha shéimeantach, cluichíocht le cláir cheannródaíochta, aimsiú gníomhairí ACP (15 ghníomhaire ionsuite), easpórtáil sceidealaithe logaí chuig BigQuery, luchtú amach comhthreomhar ilmhúnlaí `auto/chaos`, droichead bot Telegram, bainisteoir leaganacha in-aip agus ranguithe soláthraithe saor in aisce LMArena-ELO. → [Doiciméadú](docs/README.md) @@ -697,11 +697,11 @@ an timpeallacht/args cruinn gan é a rith, agus coimeádann `--api-key-env NAME`
-## 🌐 383 Soláthraí AI — 153 Marcáilte sa Chatalóg mar Shaor +## 🌐 384 Soláthraí AI — 153 Marcáilte sa Chatalóg mar Shaor
-> **383 soláthraí cláraithe** sna bailiúcháin chanónacha comhrá, meán, cuardaigh, áitiúla, gníomhairí néalbhunaithe agus chórais, lena n-áirítear **153 a bhfuil meiteashonraí fionnachtana `hasFree: true` acu**. Clúdaíonn clárlann na samhlacha comhrá **239 soláthraí / 3,059 péire soláthraí-samhail ar leith / 1,475 aitheantóir amh samhla**; tá **471 ró in aghaidh na samhla**, **34 linn athfhillteacha** agus **52 soláthraí athfhillteacha/saor go deo gan eochair** sa chatalóg bhuiséid shaor ar leith. Is ainmneoirí éagsúla iad seo de réir dearaidh; tá na sainmhínithe agus na ríomhanna dí-dúbláilte de réir linne sa [Tagairt Soláthraithe](docs/reference/PROVIDER_REFERENCE.md) agus sna [Sraitheanna Saora](docs/reference/FREE_TIERS.md). +> **384 soláthraí cláraithe** sna bailiúcháin chanónacha comhrá, meán, cuardaigh, áitiúla, gníomhairí néalbhunaithe agus chórais, lena n-áirítear **153 a bhfuil meiteashonraí fionnachtana `hasFree: true` acu**. Clúdaíonn clárlann na samhlacha comhrá **239 soláthraí / 3,059 péire soláthraí-samhail ar leith / 1,475 aitheantóir amh samhla**; tá **471 ró in aghaidh na samhla**, **34 linn athfhillteacha** agus **52 soláthraí athfhillteacha/saor go deo gan eochair** sa chatalóg bhuiséid shaor ar leith. Is ainmneoirí éagsúla iad seo de réir dearaidh; tá na sainmhínithe agus na ríomhanna dí-dúbláilte de réir linne sa [Tagairt Soláthraithe](docs/reference/PROVIDER_REFERENCE.md) agus sna [Sraitheanna Saora](docs/reference/FREE_TIERS.md).
@@ -1305,7 +1305,7 @@ Méadrachtaí canónacha ar 2026-10-18: **1.976 físeán uathúil** · **20.220. Am riteNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 TeangaTypeScript 6.0 — TypeScript 100% ar fud src/ agus open-sse/ (nialas any sa chroílár ó v2.0) CreatNext.js 16 + React 19 + Tailwind CSS 4 - Bunachar sonraíbetter-sqlite3 (SQLite, dialannú WAL) + LowDB (JSON oidhreachta) — 147 modúl fearainn, 230 imirce + Bunachar sonraíbetter-sqlite3 (SQLite, dialannú WAL) + LowDB (JSON oidhreachta) — 147 modúl fearainn, 232 imirce CuimhneTéacs iomlán SQLite FTS5 + leabuithe veicteora cainníochtaithe int8, meath clóscríofa ScéimeannaZod 4 — bailíochtú I/A uirlisí MCP + conarthaí API PrótacailMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ga/llm.txt b/docs/i18n/ga/llm.txt index 78535e0fc67..f0de544e021 100644 --- a/docs/i18n/ga/llm.txt +++ b/docs/i18n/ga/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/gu/README.md b/docs/i18n/gu/README.md index b942ef7fa91..24ce4da1836 100644 --- a/docs/i18n/gu/README.md +++ b/docs/i18n/gu/README.md @@ -15,7 +15,7 @@ **[Diego Souza](https://github.com/diegosouzapw) દ્વારા બનાવાયેલ** -OmniRoute — ક્યારેય કોડિંગ બંધ ન કરો. દરેક AI ટૂલ → 383 પ્રદાતાઓ — જેમાંથી 150+ મફત — એક જ એન્ડપૉઇન્ટ દ્વારા. Claude Code, Codex, Cursor, Cline, Copilot અને Antigravity ને ઓટો-ફૉલબૅક સાથે મફત Claude / GPT / Gemini માં જોડો. RTK + Caveman સ્ટૅક્ડ કમ્પ્રેશન 15–95% ટોકન (~89% સરેરાશ) બચાવે છે — ક્યારેય મર્યાદા સુધી ન પહોંચો. 383 AI પ્રદાતાઓ · 150+ મફત ટિયર · ~1.62B મફત ટોકન/મહિનો · 20 રાઉટિંગ વ્યૂહરચનાઓ · શરૂ કરવા માટે $0. +OmniRoute — ક્યારેય કોડિંગ બંધ ન કરો. દરેક AI ટૂલ → 384 પ્રદાતાઓ — જેમાંથી 150+ મફત — એક જ એન્ડપૉઇન્ટ દ્વારા. Claude Code, Codex, Cursor, Cline, Copilot અને Antigravity ને ઓટો-ફૉલબૅક સાથે મફત Claude / GPT / Gemini માં જોડો. RTK + Caveman સ્ટૅક્ડ કમ્પ્રેશન 15–95% ટોકન (~89% સરેરાશ) બચાવે છે — ક્યારેય મર્યાદા સુધી ન પહોંચો. 384 AI પ્રદાતાઓ · 150+ મફત ટિયર · ~1.62B મફત ટોકન/મહિનો · 20 રાઉટિંગ વ્યૂહરચનાઓ · શરૂ કરવા માટે $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | -------------------------- | :----------------------: | :---------: | :-------------------: | -| 🌐 પ્રદાતાઓ | 352 | **358** | 383 | +| 🌐 પ્રદાતાઓ | 352 | **358** | 384 | | 🧠 અનન્ય ચેટ મોડેલ IDs | 1320 | **1374** | 1491 | | 🖼️ મોડેલિટી બ્રિજ | 🆕 વિઝન + ઑડિયો + વિડિયો | ✓ | ✓ | | 📡 રડાર મફત કેટલોગ | 🆕 વૈકલ્પિક | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ સુવિધાઓ 🎯 કોમ્બોઝ - 🌐 પ્રદાતાઓ + 🌐 પ્રદાતાઓ 🔌 CLI અને MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -વચન — એક એન્ડપૉઇન્ટ અને 383 પ્રોવાઇડર્સ. બીજું સ્વસ્થ લક્ષ્ય ઉપલબ્ધ હોય ત્યાં સુધી સ્વચાલિત ફૉલબૅક રૂટિંગ ચાલુ રાખે છે. છ આધારસ્તંભો: 383 પ્રોવાઇડર્સમાં સ્થિતિસ્થાપક ફૉલબૅક · યોગ્ય વર્કલોડ પર 95% સુધી ટોકન બચત · 150+ મફત ટિયર્સ અને 52 પુનરાવર્તિત/કીવિહોણા હંમેશાં-મફત પ્રોવાઇડર્સ સાથે શરૂઆત માટે $0 · એક કૉન્ફિગરેશન દ્વારા 39 CLI/એજન્ટ ઇન્ટિગ્રેશન્સ · /v1 પર OpenAI, Claude, Gemini અને Responses API સુસંગતતા · સર્કિટ બ્રેકર્સ, TLS stealth, MCP 110 ટૂલ્સ, A2A, મેમરી, ગાર્ડરેલ્સ, evals અને 7,800+ ટ્રૅક કરેલી ટેસ્ટ ફાઇલો પર 56,000+ સ્ટૅટિક ટેસ્ટ ડિક્લેરેશન્સ સહિતના પ્રોડક્શન નિયંત્રણો. +વચન — એક એન્ડપૉઇન્ટ અને 384 પ્રોવાઇડર્સ. બીજું સ્વસ્થ લક્ષ્ય ઉપલબ્ધ હોય ત્યાં સુધી સ્વચાલિત ફૉલબૅક રૂટિંગ ચાલુ રાખે છે. છ આધારસ્તંભો: 384 પ્રોવાઇડર્સમાં સ્થિતિસ્થાપક ફૉલબૅક · યોગ્ય વર્કલોડ પર 95% સુધી ટોકન બચત · 150+ મફત ટિયર્સ અને 52 પુનરાવર્તિત/કીવિહોણા હંમેશાં-મફત પ્રોવાઇડર્સ સાથે શરૂઆત માટે $0 · એક કૉન્ફિગરેશન દ્વારા 39 CLI/એજન્ટ ઇન્ટિગ્રેશન્સ · /v1 પર OpenAI, Claude, Gemini અને Responses API સુસંગતતા · સર્કિટ બ્રેકર્સ, TLS stealth, MCP 110 ટૂલ્સ, A2A, મેમરી, ગાર્ડરેલ્સ, evals અને 7,800+ ટ્રૅક કરેલી ટેસ્ટ ફાઇલો પર 56,000+ સ્ટૅટિક ટેસ્ટ ડિક્લેરેશન્સ સહિતના પ્રોડક્શન નિયંત્રણો.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRoute ને શું અલગ બનાવે છે — 13 ક્ષમતાઓમાં 9router, OpenRouter, CLIProxyAPI અને LiteLLM સામેની તારીખવાળી સુવિધા ઝલક. OmniRoute: 383 પ્રદાતાઓ, બિલ્ટ-ઇન 150+ મફત ટિયર્સ, 20 રૂટિંગ વ્યૂહરચનાઓ, 12-એન્જિન ટોકન કમ્પ્રેશન, 110 ટૂલ્સ સાથેનું બિલ્ટ-ઇન MCP સર્વર, A2A એજન્ટ પ્રોટોકોલ, કાયમી મેમરી, ગાર્ડરેલ્સ, ક્લાઉડ એજન્ટ્સ, TLS ફિંગરપ્રિન્ટ સ્ટેલ્થ, Desktop/Termux/PWA અને 67 i18n UI લોકેલ્સ. OmniRoute MIT-લાઇસન્સ ધરાવે છે અને તેને સ્વ-હોસ્ટ કરી શકાય છે. સ્પર્ધકોની ક્ષમતાઓ અને ગણતરીઓ બદલાઈ શકે છે; લિંક કરેલી પદ્ધતિ જુઓ. +OmniRoute ને શું અલગ બનાવે છે — 13 ક્ષમતાઓમાં 9router, OpenRouter, CLIProxyAPI અને LiteLLM સામેની તારીખવાળી સુવિધા ઝલક. OmniRoute: 384 પ્રદાતાઓ, બિલ્ટ-ઇન 150+ મફત ટિયર્સ, 20 રૂટિંગ વ્યૂહરચનાઓ, 12-એન્જિન ટોકન કમ્પ્રેશન, 110 ટૂલ્સ સાથેનું બિલ્ટ-ઇન MCP સર્વર, A2A એજન્ટ પ્રોટોકોલ, કાયમી મેમરી, ગાર્ડરેલ્સ, ક્લાઉડ એજન્ટ્સ, TLS ફિંગરપ્રિન્ટ સ્ટેલ્થ, Desktop/Termux/PWA અને 67 i18n UI લોકેલ્સ. OmniRoute MIT-લાઇસન્સ ધરાવે છે અને તેને સ્વ-હોસ્ટ કરી શકાય છે. સ્પર્ધકોની ક્ષમતાઓ અને ગણતરીઓ બદલાઈ શકે છે; લિંક કરેલી પદ્ધતિ જુઓ. 📊 9router, OpenRouter, CLIProxyAPI અને LiteLLM સામેની સંપૂર્ણ પદ્ધતિ અને દરેક સુવિધાની વિગતો → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar વૈકલ્પિક છે અને માત્ર GET નો ઉ - **🖼️ નવા endpoints** — `/v1/ocr` (Mistral OCR) અને `/v1/audio/translations` (Whisper-style) મીડિયા સપાટી પૂર્ણ કરે છે. → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 છબી / વિડિયો / ઑડિયો જનરેશન** — મીડિયા માટે એક API: xAI Grok Imagine અને Novita AI વિડિયો, ComfyUI, Magnific, Adobe Firefly, Segmind અને ElevenLabs જેવા speech providers. → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 ડિપ્લોયમેન્ટ અને ઑપ્સ** — reverse-proxy `basePath`, બ્રાઉઝર-ભાષા ઑટો-ડિટેક્ટ, દરેક કી માટે device tracking, root-less MITM trust, zh-TW સ્થાનિકીકરણ. → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 વધુ પ્રોવાઇડર્સ અને એજન્ટ્સ** — cloud agents (Codex Cloud, Cursor, Devin, Jules), બ્રાઉઝર + OAuth login સાથે Grok Build (xAI), first-class card તરીકે Ollama, Claude Opus 5 અને Sonnet 5, Kimiની સત્તાવાર ભાગીદારી (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… અને નવેસરથી તૈયાર કરેલો **383-provider catalog**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 વધુ પ્રોવાઇડર્સ અને એજન્ટ્સ** — cloud agents (Codex Cloud, Cursor, Devin, Jules), બ્રાઉઝર + OAuth login સાથે Grok Build (xAI), first-class card તરીકે Ollama, Claude Opus 5 અને Sonnet 5, Kimiની સત્તાવાર ભાગીદારી (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… અને નવેસરથી તૈયાર કરેલો **384-provider catalog**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 રાઉટિંગ પારદર્શિતા** — દરેક પ્રતિભાવમાં તેને સેવા આપનાર strategy/provider/latencyનું નામ આપતું `X-OmniRoute-Decision` હેડર હોય છે; નવી `cache-optimized` કોમ્બો વ્યૂહરચના અને Auto-Combo `cacheAffinity` factor વારંવાર આવતી વિનંતીઓને cached prefix ધરાવતા કનેક્શન તરફ પાછી રાઉટ કરે છે; અને read-only `/v1/auto-combo/{channel}/candidates` endpoint `auto/*` ચેનલનો live candidate pool જાહેર કરે છે. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ સ્થાનિક કામગીરી અને ઇન્ફ્રા** — એક-ક્લિક સ્થાનિક Redis, Cloudflare Workers / Deno Deploy રિલે ડિપ્લોયર્સ, દેખરેખ હેઠળની એમ્બેડેડ સેવાઓ તરીકે Bifrost અને Mux. → [એમ્બેડેડ સેવાઓ](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 બૉક્સમાં આ પણ સામેલ છે** — પ્લગિન ફ્રેમવર્ક + માર્કેટપ્લેસ, Omni/Agent/GitHub સ્કિલ્સ ફ્રેમવર્ક્સ, Obsidian વૉલ્ટ ઇન્ટિગ્રેશન (22 MCP ટૂલ્સ), OpenAI-સુસંગત Batch અને Files APIs, સેમેન્ટિક રિસ્પોન્સ કૅશ, લીડરબોર્ડ્સ સાથે ગેમિફિકેશન, ACP એજન્ટ ડિસ્કવરી (15 બિલ્ટ-ઇન એજન્ટ્સ), BigQuery પર શેડ્યૂલ કરેલું લૉગ એક્સપોર્ટ, `auto/chaos` પેરલલ મલ્ટિ-મોડલ ફૅન-આઉટ, Telegram બોટ બ્રિજ, ઇન-ઍપ વર્ઝન મેનેજર અને LMArena-ELO મફત-પ્રદાતા રેન્કિંગ્સ. → [દસ્તાવેજો](docs/README.md) @@ -688,11 +688,11 @@ omniroute configure codex # આ માટે પણ: claude opencode qw
-## 🌐 383 AI પ્રોવાઇડર્સ — 153 કેટલોગ-ચિહ્નિત મફત +## 🌐 384 AI પ્રોવાઇડર્સ — 153 કેટલોગ-ચિહ્નિત મફત
-> **383 નોંધાયેલા પ્રોવાઇડર્સ** કેનોનિકલ ચેટ, મીડિયા, શોધ, સ્થાનિક, ક્લાઉડ-એજન્ટ અને સિસ્ટમ સંગ્રહોમાં, જેમાં **153 પાસે `hasFree: true` ડિસ્કવરી મેટાડેટા છે**. ચેટ મોડેલ રજિસ્ટ્રીમાં **239 પ્રોવાઇડર્સ / 3,059 અલગ પ્રોવાઇડર-મોડેલ જોડીઓ / 1,475 કાચા મોડેલ IDs** આવરી લેવાય છે; અલગ મફત-બજેટ કેટલોગમાં **471 પ્રતિ-મોડેલ પંક્તિઓ**, **34 પુનરાવર્તિત પૂલ્સ** અને **52 પુનરાવર્તિત/કીલેસ હંમેશાં-મફત પ્રોવાઇડર્સ** છે. આ અલગ છેદકો જાણપૂર્વક રાખવામાં આવ્યા છે; વ્યાખ્યાઓ અને પૂલ-ડિડુપ્લિકેટેડ ગણતરીઓ [પ્રોવાઇડર સંદર્ભ](docs/reference/PROVIDER_REFERENCE.md) અને [મફત ટિયર્સ](docs/reference/FREE_TIERS.md) માં ઉપલબ્ધ છે. +> **384 નોંધાયેલા પ્રોવાઇડર્સ** કેનોનિકલ ચેટ, મીડિયા, શોધ, સ્થાનિક, ક્લાઉડ-એજન્ટ અને સિસ્ટમ સંગ્રહોમાં, જેમાં **153 પાસે `hasFree: true` ડિસ્કવરી મેટાડેટા છે**. ચેટ મોડેલ રજિસ્ટ્રીમાં **239 પ્રોવાઇડર્સ / 3,059 અલગ પ્રોવાઇડર-મોડેલ જોડીઓ / 1,475 કાચા મોડેલ IDs** આવરી લેવાય છે; અલગ મફત-બજેટ કેટલોગમાં **471 પ્રતિ-મોડેલ પંક્તિઓ**, **34 પુનરાવર્તિત પૂલ્સ** અને **52 પુનરાવર્તિત/કીલેસ હંમેશાં-મફત પ્રોવાઇડર્સ** છે. આ અલગ છેદકો જાણપૂર્વક રાખવામાં આવ્યા છે; વ્યાખ્યાઓ અને પૂલ-ડિડુપ્લિકેટેડ ગણતરીઓ [પ્રોવાઇડર સંદર્ભ](docs/reference/PROVIDER_REFERENCE.md) અને [મફત ટિયર્સ](docs/reference/FREE_TIERS.md) માં ઉપલબ્ધ છે.
@@ -1294,7 +1294,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 પણ તેન રનટાઇમNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ભાષાTypeScript 6.0 — 100% TypeScript સમગ્ર src/ અને open-sse/માં (v2.0થી કોરમાં શૂન્ય any) ફ્રેમવર્કNext.js 16 + React 19 + Tailwind CSS 4 - ડેટાબેઝbetter-sqlite3 (SQLite, WAL જર્નલિંગ) + LowDB (JSON લેગસી) — 147 ડોમેન મોડ્યુલ્સ, 230 માઇગ્રેશન્સ + ડેટાબેઝbetter-sqlite3 (SQLite, WAL જર્નલિંગ) + LowDB (JSON લેગસી) — 147 ડોમેન મોડ્યુલ્સ, 232 માઇગ્રેશન્સ મેમરીSQLite FTS5 પૂર્ણ-ટેક્સ્ટ + int8-quantized વેક્ટર એમ્બેડિંગ્સ, ટાઇપ્ડ ડિકે સ્કીમાZod 4 — MCP ટૂલ I/O માન્યકરણ + API કૉન્ટ્રેક્ટ્સ પ્રોટોકોલ્સMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/gu/llm.txt b/docs/i18n/gu/llm.txt index df65db3875b..653dd1e6148 100644 --- a/docs/i18n/gu/llm.txt +++ b/docs/i18n/gu/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ha/README.md b/docs/i18n/ha/README.md index 746435d36f7..b05bf585a11 100644 --- a/docs/i18n/ha/README.md +++ b/docs/i18n/ha/README.md @@ -15,7 +15,7 @@ **Diego Souza ne ya ƙirƙira shi ([Diego Souza](https://github.com/diegosouzapw))** -OmniRoute — Kada ka taɓa daina rubuta lamba. Kowane kayan aikin AI → masu samarwa 383 — 150+ kyauta — ta wurin endpoint guda ɗaya. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity zuwa Claude / GPT / Gemini KYauta tare da komawar atomatik. Matsewar RTK + Caveman da aka haɗa tana adana alamomi 15–95% (~89% a matsakaici) — kada ka taɓa kaiwa iyaka. Masu samar da AI 383 · matakan kyauta 150+ · ~alamomi biliyan 1.62 kyauta a wata · dabarun turawa 20 · $0 don farawa. +OmniRoute — Kada ka taɓa daina rubuta lamba. Kowane kayan aikin AI → masu samarwa 384 — 150+ kyauta — ta wurin endpoint guda ɗaya. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity zuwa Claude / GPT / Gemini KYauta tare da komawar atomatik. Matsewar RTK + Caveman da aka haɗa tana adana alamomi 15–95% (~89% a matsakaici) — kada ka taɓa kaiwa iyaka. Masu samar da AI 384 · matakan kyauta 150+ · ~alamomi biliyan 1.62 kyauta a wata · dabarun turawa 20 · $0 don farawa.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------------------- | :-----------------------------: | :---------: | :-------------------: | -| 🌐 Masu samarwa | 352 | **358** | 383 | +| 🌐 Masu samarwa | 352 | **358** | 384 | | 🧠 ID na musamman na samfuran hira | 1320 | **1374** | 1491 | | 🖼️ Gadar nau'ikan bayanai | 🆕 hangen nesa + sauti + bidiyo | ✓ | ✓ | | 📡 Kundin kyauta na Radar | 🆕 zaɓin shiga | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Fasaloli 🎯 Haɗe-haɗe - 🌐 Masu samarwa + 🌐 Masu samarwa 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Alkawarin — Wurin ƙarshen hanya guda da masu bayarwa 383. Komawar atomatik tana ci gaba da turawa muddin akwai wata makoma mai lafiya. Ginshiƙai shida: ingantacciyar komawa a tsakanin masu bayarwa 383 · har zuwa 95% rage ajiyar tokens a ayyukan da suka cancanta · $0 don farawa tare da matakan kyauta sama da 150 da masu bayarwa kyauta har abada 52 masu maimaituwa/marasa maɓalli · haɗin CLI/agent 39 ta hanyar tsari guda · daidaituwar OpenAI, Claude, Gemini da Responses API a /v1 · sarrafawar samarwa da ta haɗa da masu katse da’ira, ɓoye TLS, kayan aikin MCP 110, A2A, ƙwaƙwalwa, matakan kariya, gwaje-gwajen tantancewa da bayanan gwaji na tsaye sama da 56,000 a cikin fayilolin gwaji sama da 7,800 da ake bibiyarsu. +Alkawarin — Wurin ƙarshen hanya guda da masu bayarwa 384. Komawar atomatik tana ci gaba da turawa muddin akwai wata makoma mai lafiya. Ginshiƙai shida: ingantacciyar komawa a tsakanin masu bayarwa 384 · har zuwa 95% rage ajiyar tokens a ayyukan da suka cancanta · $0 don farawa tare da matakan kyauta sama da 150 da masu bayarwa kyauta har abada 52 masu maimaituwa/marasa maɓalli · haɗin CLI/agent 39 ta hanyar tsari guda · daidaituwar OpenAI, Claude, Gemini da Responses API a /v1 · sarrafawar samarwa da ta haɗa da masu katse da’ira, ɓoye TLS, kayan aikin MCP 110, A2A, ƙwaƙwalwa, matakan kariya, gwaje-gwajen tantancewa da bayanan gwaji na tsaye sama da 56,000 a cikin fayilolin gwaji sama da 7,800 da ake bibiyarsu.

@@ -510,7 +510,7 @@ Dukkan dabaru **20** — haɗa su yadda kake so a kowane mataki na haɗi: -Abin da ya bambanta OmniRoute — wani hoton kwatancen fasaloli na zamani da 9router, OpenRouter, CLIProxyAPI da LiteLLM a fannoni 13. OmniRoute: masu samarwa 383, matakan kyauta sama da 150 da aka gina a ciki, dabarun routing 20, matse alamomin token ta injuna 12, uwar garken MCP da aka gina a ciki mai kayan aiki 110, ƙa'idar wakilan A2A, ma'ajiya mai ɗorewa, matakan kariya, wakilan girgije, ɓoyayyen sawun TLS, Desktop/Termux/PWA da harsunan UI na i18n 67. OmniRoute yana da lasisin MIT kuma ana iya karɓar bakuncinsa da kansa. Ƙwarewa da ƙididdigar masu fafatawa na iya sauyawa; duba hanyoyin da aka haɗa. +Abin da ya bambanta OmniRoute — wani hoton kwatancen fasaloli na zamani da 9router, OpenRouter, CLIProxyAPI da LiteLLM a fannoni 13. OmniRoute: masu samarwa 384, matakan kyauta sama da 150 da aka gina a ciki, dabarun routing 20, matse alamomin token ta injuna 12, uwar garken MCP da aka gina a ciki mai kayan aiki 110, ƙa'idar wakilan A2A, ma'ajiya mai ɗorewa, matakan kariya, wakilan girgije, ɓoyayyen sawun TLS, Desktop/Termux/PWA da harsunan UI na i18n 67. OmniRoute yana da lasisin MIT kuma ana iya karɓar bakuncinsa da kansa. Ƙwarewa da ƙididdigar masu fafatawa na iya sauyawa; duba hanyoyin da aka haɗa. 📊 Cikakkiyar hanya & cikakkun bayanai kan kowane fasali idan aka kwatanta da 9router, OpenRouter, CLIProxyAPI & LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar na zaɓi ne kuma GET-only. Abokin hulɗar OmniRoute ba ya loda prompts, zi - **🖼️ Sabbin ƙarshen API** — `/v1/ocr` (Mistral OCR) da `/v1/audio/translations` (mai salo irin na Whisper) sun cika ɓangaren kafofin watsa labarai. → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 Ƙirƙirar hoto / bidiyo / sauti** — API ɗaya don kafofin watsa labarai: xAI Grok Imagine & Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind, da masu bayar da magana kamar ElevenLabs. → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 Girke-girke da gudanarwa** — `basePath` na reverse-proxy, gano harshen mai bincike ta atomatik, bin diddigin na'ura bisa kowane maɓalli, amincewar MITM ba tare da root ba, da fassarar zh-TW. → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 Ƙarin masu bayarwa da wakilai** — wakilan gajimare (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) tare da mai bincike + shiga OAuth, katin Ollama na matakin farko, Claude Opus 5 & Sonnet 5, haɗin gwiwar hukuma da Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… da sabunta **katalojin masu bayarwa 383**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Ƙarin masu bayarwa da wakilai** — wakilan gajimare (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) tare da mai bincike + shiga OAuth, katin Ollama na matakin farko, Claude Opus 5 & Sonnet 5, haɗin gwiwar hukuma da Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… da sabunta **katalojin masu bayarwa 384**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 Bayyananniyar kewayawa** — kowace amsa tana ɗauke da taken `X-OmniRoute-Decision` mai bayyana dabarar/mai bayarwa/jinkirin da ya ba da ita, sabuwar dabarar haɗi ta `cache-optimized` + abin `cacheAffinity` na Auto-Combo suna mayar da maimaitattun buƙatu zuwa haɗin da ke riƙe da prefix da aka adana, kuma ƙarshen API mai karatu-kawai `/v1/auto-combo/{channel}/candidates` yana bayyana rukunin zaɓaɓɓun kai tsaye na tashar `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md). - **⚡ Ayyukan aiki na gida da kayan more rayuwa** — Redis na gida da ake kunna da dannawa ɗaya, masu tura relay na Cloudflare Workers / Deno Deploy, Bifrost da Mux a matsayin ayyukan da aka saka kuma ake sa ido a kansu. → [Ayyukan da Aka Saka](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Haka kuma yana cikin kunshin** — tsarin plugins + kasuwa, tsarin ƙwarewar Omni/Agent/GitHub, haɗin Obsidian vault (kayan aikin MCP 22), APIs na Batch & Files masu dacewa da OpenAI, ma'ajiyar amsoshin semantic, tsarin gamification mai allunan jagora, gano wakilan ACP (wakilai 15 da aka riga aka gina), fitar da log zuwa BigQuery bisa jadawali, `auto/chaos` don rarraba ayyuka ga samfuran AI da yawa a layi ɗaya, gadar bot ta Telegram, mai sarrafa sigar manhaja a ciki, da martabobin masu samarwa kyauta na LMArena-ELO. → [Takardu](docs/README.md) @@ -688,11 +688,11 @@ yana nuna ainihin muhalli/huɗɗa ba tare da aiwatarwa ba, kuma `--api-key-env N
-## 🌐 Masu Samar da AI 383 — 153 Kundin Ya Yi Wa Alama a Matsayin Kyauta +## 🌐 Masu Samar da AI 384 — 153 Kundin Ya Yi Wa Alama a Matsayin Kyauta
-> **Masu samarwa 383 da aka yi rajista** a cikin tarin hirarraki, kafofin watsa labarai, bincike, na gida, wakilan girgije da tsarin da aka amince da su, ciki har da **153 masu ɗauke da bayanan gano `hasFree: true`**. Rijistar samfuran hira ta ƙunshi **masu samarwa 239 / nau’ikan haɗin mai samarwa da samfuri 3,059 daban-daban / ID ɗin samfura 1,475 na asali**; yayin da kundin kasafin kyauta na daban ya ƙunshi **layukan kowane samfuri 471**, **rukunai masu maimaituwa 34** da **masu samarwa 52 masu maimaituwa/ba sa buƙatar maɓalli kuma kyauta har abada**. Waɗannan ma’aunai ne daban bisa ƙira; ma’anoni da ƙididdigar da aka cire maimaitawar rukunai suna cikin [Marejen Mai Samarwa](docs/reference/PROVIDER_REFERENCE.md) da [Matakan Kyauta](docs/reference/FREE_TIERS.md). +> **Masu samarwa 384 da aka yi rajista** a cikin tarin hirarraki, kafofin watsa labarai, bincike, na gida, wakilan girgije da tsarin da aka amince da su, ciki har da **153 masu ɗauke da bayanan gano `hasFree: true`**. Rijistar samfuran hira ta ƙunshi **masu samarwa 239 / nau’ikan haɗin mai samarwa da samfuri 3,059 daban-daban / ID ɗin samfura 1,475 na asali**; yayin da kundin kasafin kyauta na daban ya ƙunshi **layukan kowane samfuri 471**, **rukunai masu maimaituwa 34** da **masu samarwa 52 masu maimaituwa/ba sa buƙatar maɓalli kuma kyauta har abada**. Waɗannan ma’aunai ne daban bisa ƙira; ma’anoni da ƙididdigar da aka cire maimaitawar rukunai suna cikin [Marejen Mai Samarwa](docs/reference/PROVIDER_REFERENCE.md) da [Matakan Kyauta](docs/reference/FREE_TIERS.md).
@@ -1294,7 +1294,7 @@ Ma'aunai na asali a 2026-10-18: **bidiyoyi na musamman 1.976** · **sanannun kal Lokacin guduNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 Harshen shirye-shiryeTypeScript 6.0 — TypeScript 100% a cikin src/ da open-sse/ (any sifili a cikin core tun v2.0) Tsarin aikiNext.js 16 + React 19 + Tailwind CSS 4 - Ma’adanar bayanaibetter-sqlite3 (SQLite, rubutun WAL) + LowDB (JSON na gado) — modules 147 na fannin aiki, migrations 230 + Ma’adanar bayanaibetter-sqlite3 (SQLite, rubutun WAL) + LowDB (JSON na gado) — modules 147 na fannin aiki, migrations 232 ƘwaƙwalwaBinciken cikakken rubutu na SQLite FTS5 + embeddings na vector masu auna int8, tare da raguwar ƙima mai nau’i Tsare-tsareZod 4 — ingantaccen bayanan shigarwa/fitarwa na kayan aikin MCP + yarjejeniyoyin API Ka’idojiMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ha/llm.txt b/docs/i18n/ha/llm.txt index 31d540613a1..fefe9d0ff67 100644 --- a/docs/i18n/ha/llm.txt +++ b/docs/i18n/ha/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/he/README.md b/docs/i18n/he/README.md index 608c45ceb87..f0ad88f5440 100644 --- a/docs/i18n/he/README.md +++ b/docs/i18n/he/README.md @@ -15,7 +15,7 @@ **נוצר על ידי [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — לעולם אל תפסיקו לתכנת. כל כלי AI → 383 ספקים — 150+ בחינם — דרך נקודת קצה אחת. Claude Code, Codex, Cursor, Cline, Copilot ו-Antigravity אל Claude / GPT / Gemini בחינם עם מעבר אוטומטי במקרה כשל. דחיסה משולבת של RTK + Caveman חוסכת 15–95% מהטוקנים (~89% בממוצע) — לעולם לא תגיעו למגבלות. 383 ספקי AI · 150+ מסלולים בחינם · ~1.62B טוקנים בחינם בחודש · 20 אסטרטגיות ניתוב · $0 כדי להתחיל. +OmniRoute — לעולם אל תפסיקו לתכנת. כל כלי AI → 384 ספקים — 150+ בחינם — דרך נקודת קצה אחת. Claude Code, Codex, Cursor, Cline, Copilot ו-Antigravity אל Claude / GPT / Gemini בחינם עם מעבר אוטומטי במקרה כשל. דחיסה משולבת של RTK + Caveman חוסכת 15–95% מהטוקנים (~89% בממוצע) — לעולם לא תגיעו למגבלות. 384 ספקי AI · 150+ מסלולים בחינם · ~1.62B טוקנים בחינם בחודש · 20 אסטרטגיות ניתוב · $0 כדי להתחיל.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------- | :--------------------: | :---------: | :-------------------: | -| 🌐 ספקים | 352 | **358** | 383 | +| 🌐 ספקים | 352 | **358** | 384 | | 🧠 מזהי מודלי צ'אט ייחודיים | 1320 | **1374** | 1491 | | 🖼️ גשר מודאליות | 🆕 ראייה + שמע + וידאו | ✓ | ✓ | | 📡 קטלוג חינמי של Radar | 🆕 opt-in | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ תכונות 🎯 שילובים - 🌐 ספקים + 🌐 ספקים 🔌 CLI ו-MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -ההבטחה — נקודת קצה אחת ו־383 ספקים. מנגנון נסיגה אוטומטי ממשיך את הניתוב כל עוד יעד בריא אחר זמין. שישה עקרונות: נסיגה עמידה בין 383 ספקים · חיסכון של עד 95% באסימונים בעומסי עבודה מתאימים · התחלה בעלות של $0 עם יותר מ־150 שכבות חינמיות ו־52 ספקים חינמיים לתמיד, חוזרים או ללא מפתח · 39 אינטגרציות CLI/agent באמצעות תצורה אחת · תאימות ל־OpenAI, Claude, Gemini ול־Responses API ב־/v1 · בקרות ייצור הכוללות מפסקי זרם, הסוואת TLS, ‏MCP עם 110 כלים, A2A, זיכרון, מעקות בטיחות, הערכות ויותר מ־56,000 הצהרות בדיקה סטטיות על פני יותר מ־7,800 קובצי בדיקה במעקב. +ההבטחה — נקודת קצה אחת ו־384 ספקים. מנגנון נסיגה אוטומטי ממשיך את הניתוב כל עוד יעד בריא אחר זמין. שישה עקרונות: נסיגה עמידה בין 384 ספקים · חיסכון של עד 95% באסימונים בעומסי עבודה מתאימים · התחלה בעלות של $0 עם יותר מ־150 שכבות חינמיות ו־52 ספקים חינמיים לתמיד, חוזרים או ללא מפתח · 39 אינטגרציות CLI/agent באמצעות תצורה אחת · תאימות ל־OpenAI, Claude, Gemini ול־Responses API ב־/v1 · בקרות ייצור הכוללות מפסקי זרם, הסוואת TLS, ‏MCP עם 110 כלים, A2A, זיכרון, מעקות בטיחות, הערכות ויותר מ־56,000 הצהרות בדיקה סטטיות על פני יותר מ־7,800 קובצי בדיקה במעקב.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -מה מבדיל את OmniRoute — תמונת מצב מתוארכת של תכונות בהשוואה ל־9router, OpenRouter, CLIProxyAPI ו־LiteLLM על פני 13 יכולות. OmniRoute: 383 ספקים, יותר מ־150 שכבות חינמיות מובנות, 20 אסטרטגיות ניתוב, דחיסת טוקנים באמצעות 12 מנועים, שרת MCP מובנה עם 110 כלים, פרוטוקול סוכנים A2A, זיכרון מתמשך, אמצעי הגנה, סוכני ענן, הסוואת טביעת אצבע של TLS, Desktop/Termux/PWA ו־67 אזורי שפה בממשק. OmniRoute מופץ ברישיון MIT וניתן לאירוח עצמי. היכולות והספירות של המתחרים עשויות להשתנות; למידע נוסף ראו את המתודולוגיה המקושרת. +מה מבדיל את OmniRoute — תמונת מצב מתוארכת של תכונות בהשוואה ל־9router, OpenRouter, CLIProxyAPI ו־LiteLLM על פני 13 יכולות. OmniRoute: 384 ספקים, יותר מ־150 שכבות חינמיות מובנות, 20 אסטרטגיות ניתוב, דחיסת טוקנים באמצעות 12 מנועים, שרת MCP מובנה עם 110 כלים, פרוטוקול סוכנים A2A, זיכרון מתמשך, אמצעי הגנה, סוכני ענן, הסוואת טביעת אצבע של TLS, Desktop/Termux/PWA ו־67 אזורי שפה בממשק. OmniRoute מופץ ברישיון MIT וניתן לאירוח עצמי. היכולות והספירות של המתחרים עשויות להשתנות; למידע נוסף ראו את המתודולוגיה המקושרת. 📊 המתודולוגיה המלאה ופירוט לפי תכונה בהשוואה ל־9router, OpenRouter, CLIProxyAPI ו־LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar הוא שירות אופציונלי המבוסס על GET בלבד. לק - **🖼️ נקודות קצה חדשות** — `/v1/ocr` (‏Mistral OCR) ו־`/v1/audio/translations` (בסגנון Whisper) משלימות את מעטפת המדיה. → [מדריך ה־API](docs/reference/API_REFERENCE.md) - **🎨 יצירת תמונות / וידאו / אודיו** — API אחד למדיה: xAI Grok Imagine ו־Novita AI video, ‏ComfyUI, ‏Magnific, ‏Adobe Firefly, ‏Segmind, וספקי דיבור כגון ElevenLabs. → [מדריך ה־API](docs/reference/API_REFERENCE.md) - **🌍 פריסה ותפעול** — `basePath` עבור reverse-proxy, זיהוי אוטומטי של שפת הדפדפן, מעקב אחר מכשירים לפי מפתח, אמון MITM ללא הרשאות root, ולוקליזציה ל־zh-TW. → [סביבה](docs/reference/ENVIRONMENT.md) -- **🤝 ספקים וסוכנים נוספים** — סוכני ענן (Codex Cloud, Cursor, Devin, Jules), ‏Grok Build ‏(xAI) עם דפדפן וכניסה באמצעות OAuth, כרטיס רשמי ל־Ollama, ‏Claude Opus 5 ו־Sonnet 5, שותפות רשמית עם Kimi ‏(Code/Web/Moonshot), ‏Zed, ‏Requesty, ‏SenseNova, ‏Yuanbao, ‏Agnes AI… וקטלוג מעודכן של **383 ספקים**. → [ספקים](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 ספקים וסוכנים נוספים** — סוכני ענן (Codex Cloud, Cursor, Devin, Jules), ‏Grok Build ‏(xAI) עם דפדפן וכניסה באמצעות OAuth, כרטיס רשמי ל־Ollama, ‏Claude Opus 5 ו־Sonnet 5, שותפות רשמית עם Kimi ‏(Code/Web/Moonshot), ‏Zed, ‏Requesty, ‏SenseNova, ‏Yuanbao, ‏Agnes AI… וקטלוג מעודכן של **384 ספקים**. → [ספקים](docs/reference/PROVIDER_REFERENCE.md) - **📡 שקיפות בניתוב** — כל תגובה כוללת כותרת `X-OmniRoute-Decision` המציינת את האסטרטגיה/הספק/השהיה ששירתו אותה, אסטרטגיית קומבו חדשה בשם `cache-optimized` וגורם `cacheAffinity` של Auto-Combo מנתבים בקשות חוזרות בחזרה לחיבור שמחזיק בקידומת שבמטמון, ונקודת קצה לקריאה בלבד `/v1/auto-combo/{channel}/candidates` חושפת את מאגר המועמדים החי של ערוץ `auto/*`. → [קומבו אוטומטי](docs/routing/AUTO-COMBO.md) - **⚡ ביצועים ותשתיות מקומיים** — Redis מקומי בלחיצה אחת, מפעילי פריסה לממסרים של Cloudflare Workers / Deno Deploy, ו־Bifrost ו־Mux כשירותים משובצים ומנוהלים. → [שירותים משובצים](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 גם כלול בחבילה** — מסגרת תוספים ושוק תוספים, מסגרות מיומנויות של Omni/Agent/GitHub, אינטגרציה עם כספת Obsidian (22 כלי MCP), ממשקי Batch ו־Files תואמי OpenAI, מטמון תגובות סמנטי, מנגנון משחק עם טבלאות מובילים, גילוי סוכני ACP (15 סוכנים מובנים), ייצוא מתוזמן של יומנים ל־BigQuery, הפצת עומסים מקבילית של מודלים מרובים באמצעות `auto/chaos`, גשר לבוט Telegram, מנהל גרסאות בתוך האפליקציה ודירוגי ספקים חינמיים של LMArena-ELO. → [תיעוד](docs/README.md) @@ -688,11 +688,11 @@ omniroute configure codex # גם: claude opencode qwen aider goose gemi
-## 🌐 383 ספקי AI — 153 מסומנים בקטלוג כחינמיים +## 🌐 384 ספקי AI — 153 מסומנים בקטלוג כחינמיים
-> **383 ספקים רשומים** באוספים הקנוניים של צ'אט, מדיה, חיפוש, מקומי, סוכני ענן ומערכת, כולל **153 ספקים עם מטא-נתוני גילוי `hasFree: true`**. רישום מודלי הצ'אט כולל **239 ספקים / 3,059 צמדי ספק-מודל ייחודיים / 1,475 מזהי מודלים גולמיים**; קטלוג התקציב החינמי הנפרד כולל **471 שורות לכל מודל**, **34 מאגרים חוזרים** ו-**52 ספקים חינמיים לתמיד, חוזרים/ללא מפתח**. אלה מכנים שונים בכוונה; ההגדרות והחישובים לאחר הסרת כפילויות ממאגרים נמצאים ב-[Reference לספקים](docs/reference/PROVIDER_REFERENCE.md) וב-[רמות חינמיות](docs/reference/FREE_TIERS.md). +> **384 ספקים רשומים** באוספים הקנוניים של צ'אט, מדיה, חיפוש, מקומי, סוכני ענן ומערכת, כולל **153 ספקים עם מטא-נתוני גילוי `hasFree: true`**. רישום מודלי הצ'אט כולל **239 ספקים / 3,059 צמדי ספק-מודל ייחודיים / 1,475 מזהי מודלים גולמיים**; קטלוג התקציב החינמי הנפרד כולל **471 שורות לכל מודל**, **34 מאגרים חוזרים** ו-**52 ספקים חינמיים לתמיד, חוזרים/ללא מפתח**. אלה מכנים שונים בכוונה; ההגדרות והחישובים לאחר הסרת כפילויות ממאגרים נמצאים ב-[Reference לספקים](docs/reference/PROVIDER_REFERENCE.md) וב-[רמות חינמיות](docs/reference/FREE_TIERS.md).
@@ -1294,7 +1294,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # גם CI=1 מדלג ע סביבת ריצהNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 שפהTypeScript 6.0 — 100% TypeScript בכל src/ ו-open-sse/ (אפס מופעים של any בליבה מאז v2.0) תשתיתNext.js 16 + React 19 + Tailwind CSS 4 - מסד נתוניםbetter-sqlite3 (SQLite, רישום WAL) + LowDB (JSON מדור קודם) — 147 מודולי תחום, 230 הגירות + מסד נתוניםbetter-sqlite3 (SQLite, רישום WAL) + LowDB (JSON מדור קודם) — 147 מודולי תחום, 232 הגירות זיכרוןחיפוש טקסט מלא באמצעות SQLite FTS5 + הטמעות וקטוריות שעברו קוונטיזציה ל-int8, דעיכה טיפוסית סכמותZod 4 — אימות קלט/פלט של כלי MCP + חוזי API פרוטוקוליםMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/he/llm.txt b/docs/i18n/he/llm.txt index 3642eddeaa8..5fc39fba7e6 100644 --- a/docs/i18n/he/llm.txt +++ b/docs/i18n/he/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/hi/README.md b/docs/i18n/hi/README.md index 5e4e455b805..5e03c110499 100644 --- a/docs/i18n/hi/README.md +++ b/docs/i18n/hi/README.md @@ -15,7 +15,7 @@ **[Diego Souza](https://github.com/diegosouzapw) द्वारा निर्मित** -OmniRoute — कोडिंग कभी न रोकें। हर AI टूल → 383 प्रदाता — 150+ निःशुल्क — एक ही एंडपॉइंट के माध्यम से। Claude Code, Codex, Cursor, Cline, Copilot और Antigravity को ऑटो-फ़ॉलबैक के साथ निःशुल्क Claude / GPT / Gemini में बदलें। RTK + Caveman की संयुक्त कंप्रेशन 15–95% टोकन (~89% औसत) बचाती है — सीमाओं का सामना कभी न करें। 383 AI प्रदाता · 150+ निःशुल्क टियर · ~1.62B निःशुल्क टोकन/माह · 20 रूटिंग रणनीतियाँ · शुरुआत की लागत $0. +OmniRoute — कोडिंग कभी न रोकें। हर AI टूल → 384 प्रदाता — 150+ निःशुल्क — एक ही एंडपॉइंट के माध्यम से। Claude Code, Codex, Cursor, Cline, Copilot और Antigravity को ऑटो-फ़ॉलबैक के साथ निःशुल्क Claude / GPT / Gemini में बदलें। RTK + Caveman की संयुक्त कंप्रेशन 15–95% टोकन (~89% औसत) बचाती है — सीमाओं का सामना कभी न करें। 384 AI प्रदाता · 150+ निःशुल्क टियर · ~1.62B निःशुल्क टोकन/माह · 20 रूटिंग रणनीतियाँ · शुरुआत की लागत $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------- | :-----------------------: | :---------: | :-------------------: | -| 🌐 प्रदाता | 352 | **358** | 383 | +| 🌐 प्रदाता | 352 | **358** | 384 | | 🧠 अद्वितीय चैट मॉडल IDs | 1320 | **1374** | 1491 | | 🖼️ मोडैलिटी ब्रिज | 🆕 विज़न + ऑडियो + वीडियो | ✓ | ✓ | | 📡 रडार मुफ़्त कैटलॉग | 🆕 ऑप्ट-इन | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ सुविधाएँ 🎯 कॉम्बो - 🌐 प्रदाता + 🌐 प्रदाता 🔌 CLI और MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -वादा — एक एंडपॉइंट और 383 प्रोवाइडर। जब कोई अन्य स्वस्थ लक्ष्य उपलब्ध रहता है, तो स्वचालित फ़ॉलबैक रूटिंग जारी रखता है। छह स्तंभ: 383 प्रोवाइडर के बीच लचीला फ़ॉलबैक · योग्य वर्कलोड पर टोकन में 95% तक की बचत · 150+ मुफ़्त टियर और 52 आवर्ती/ keyless हमेशा-मुफ़्त प्रोवाइडर के साथ शुरू करने के लिए $0 · एक कॉन्फ़िगरेशन के माध्यम से 39 CLI/एजेंट इंटीग्रेशन · /v1 पर OpenAI, Claude, Gemini और Responses API संगतता · सर्किट ब्रेकर, TLS stealth, MCP 110 टूल, A2A, मेमोरी, guardrails, evals और 7,800+ ट्रैक किए गए टेस्ट फ़ाइलों में 56,000+ स्थिर टेस्ट घोषणाओं सहित प्रोडक्शन नियंत्रण। +वादा — एक एंडपॉइंट और 384 प्रोवाइडर। जब कोई अन्य स्वस्थ लक्ष्य उपलब्ध रहता है, तो स्वचालित फ़ॉलबैक रूटिंग जारी रखता है। छह स्तंभ: 384 प्रोवाइडर के बीच लचीला फ़ॉलबैक · योग्य वर्कलोड पर टोकन में 95% तक की बचत · 150+ मुफ़्त टियर और 52 आवर्ती/ keyless हमेशा-मुफ़्त प्रोवाइडर के साथ शुरू करने के लिए $0 · एक कॉन्फ़िगरेशन के माध्यम से 39 CLI/एजेंट इंटीग्रेशन · /v1 पर OpenAI, Claude, Gemini और Responses API संगतता · सर्किट ब्रेकर, TLS stealth, MCP 110 टूल, A2A, मेमोरी, guardrails, evals और 7,800+ ट्रैक किए गए टेस्ट फ़ाइलों में 56,000+ स्थिर टेस्ट घोषणाओं सहित प्रोडक्शन नियंत्रण।

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRoute को अलग पहचान देने वाली बातें — 13 क्षमताओं में 9router, OpenRouter, CLIProxyAPI और LiteLLM की तुलना वाला दिनांकित फीचर स्नैपशॉट। OmniRoute: 383 प्रदाता, पहले से निर्मित 150+ निःशुल्क टियर, 20 रूटिंग रणनीतियाँ, 12-इंजन टोकन संपीड़न, 110 टूल वाला अंतर्निर्मित MCP सर्वर, A2A एजेंट प्रोटोकॉल, स्थायी मेमोरी, सुरक्षा नियंत्रण, क्लाउड एजेंट, TLS फ़िंगरप्रिंट स्टील्थ, Desktop/Termux/PWA और 67 i18n UI लोकेल। OmniRoute MIT-लाइसेंस प्राप्त और स्वयं-होस्ट करने योग्य है। प्रतिस्पर्धी क्षमताएँ और संख्याएँ बदल सकती हैं; लिंक की गई कार्यप्रणाली देखें। +OmniRoute को अलग पहचान देने वाली बातें — 13 क्षमताओं में 9router, OpenRouter, CLIProxyAPI और LiteLLM की तुलना वाला दिनांकित फीचर स्नैपशॉट। OmniRoute: 384 प्रदाता, पहले से निर्मित 150+ निःशुल्क टियर, 20 रूटिंग रणनीतियाँ, 12-इंजन टोकन संपीड़न, 110 टूल वाला अंतर्निर्मित MCP सर्वर, A2A एजेंट प्रोटोकॉल, स्थायी मेमोरी, सुरक्षा नियंत्रण, क्लाउड एजेंट, TLS फ़िंगरप्रिंट स्टील्थ, Desktop/Termux/PWA और 67 i18n UI लोकेल। OmniRoute MIT-लाइसेंस प्राप्त और स्वयं-होस्ट करने योग्य है। प्रतिस्पर्धी क्षमताएँ और संख्याएँ बदल सकती हैं; लिंक की गई कार्यप्रणाली देखें। 📊 9router, OpenRouter, CLIProxyAPI और LiteLLM की तुलना की पूरी कार्यप्रणाली और प्रति-फीचर विवरण → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -614,7 +614,7 @@ Radar ऑप्ट-इन और केवल-GET है। OmniRoute क्ल - **🖼️ नए endpoints** — `/v1/ocr` (Mistral OCR) और `/v1/audio/translations` (Whisper-style) मीडिया सतह को पूरा करते हैं। → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 Image / video / audio generation** — मीडिया के लिए एक API: xAI Grok Imagine और Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind तथा ElevenLabs जैसे speech प्रोवाइडर। → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 Deployment & ops** — reverse-proxy `basePath`, ब्राउज़र-भाषा का ऑटो-डिटेक्शन, प्रति-कुंजी डिवाइस ट्रैकिंग, root-less MITM trust और zh-TW स्थानीयकरण। → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 अधिक प्रोवाइडर और एजेंट** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login वाला Grok Build (xAI), first-class card वाला Ollama, Claude Opus 5 और Sonnet 5, Kimi की आधिकारिक साझेदारी (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… और नया **383-provider catalog**। → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 अधिक प्रोवाइडर और एजेंट** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login वाला Grok Build (xAI), first-class card वाला Ollama, Claude Opus 5 और Sonnet 5, Kimi की आधिकारिक साझेदारी (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… और नया **384-provider catalog**। → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 रूटिंग पारदर्शिता** — हर response में उसे सर्व करने वाली strategy/provider/latency का नाम देने वाला `X-OmniRoute-Decision` हेडर होता है; नई `cache-optimized` कॉम्बो रणनीति और Auto-Combo `cacheAffinity` factor दोहराए गए अनुरोधों को उस कनेक्शन पर वापस भेजते हैं जिसके पास cached prefix है; और केवल-पठन `/v1/auto-combo/{channel}/candidates` endpoint किसी `auto/*` channel के लाइव candidate pool को दिखाता है। → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ स्थानीय प्रदर्शन और इन्फ्रा** — एक-क्लिक स्थानीय Redis, Cloudflare Workers / Deno Deploy रिले डिप्लॉयर, तथा सुपरवाइज़ किए गए एम्बेडेड सर्विस के रूप में Bifrost और Mux। → [एम्बेडेड सर्विस](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 बॉक्स में और भी** — प्लगइन फ़्रेमवर्क + मार्केटप्लेस, Omni/Agent/GitHub स्किल्स फ़्रेमवर्क, Obsidian वॉल्ट इंटीग्रेशन (22 MCP टूल), OpenAI-संगत Batch और Files APIs, सिमेंटिक रिस्पॉन्स कैश, लीडरबोर्ड सहित गेमिफ़िकेशन, ACP एजेंट डिस्कवरी (15 बिल्ट-इन एजेंट), BigQuery में शेड्यूल किया गया लॉग एक्सपोर्ट, `auto/chaos` पैरेलल मल्टी-मॉडल फ़ैन-आउट, Telegram बॉट ब्रिज, इन-ऐप वर्शन मैनेजर और LMArena-ELO मुफ़्त-प्रदाता रैंकिंग। → [दस्तावेज़](docs/README.md) @@ -697,11 +697,11 @@ omniroute configure codex # साथ ही: claude opencode qwen aide
-## 🌐 383 AI प्रदाता — 153 कैटलॉग-चिह्नित निःशुल्क +## 🌐 384 AI प्रदाता — 153 कैटलॉग-चिह्नित निःशुल्क
-> **383 पंजीकृत प्रदाता** canonical chat, media, search, local, cloud-agent और system collections में फैले हुए हैं, जिनमें **153 के पास `hasFree: true` discovery metadata है**। Chat model registry में **239 प्रदाता / 3,059 विशिष्ट provider-model pairs / 1,475 raw model IDs** शामिल हैं; अलग free-budget catalog में **471 per-model rows**, **34 recurring pools** और **52 recurring/keyless free-forever providers** हैं। ये जानबूझकर अलग-अलग denominators हैं; परिभाषाएँ और pool-deduped calculations [प्रदाता संदर्भ](docs/reference/PROVIDER_REFERENCE.md) और [निःशुल्क स्तर](docs/reference/FREE_TIERS.md) में उपलब्ध हैं। +> **384 पंजीकृत प्रदाता** canonical chat, media, search, local, cloud-agent और system collections में फैले हुए हैं, जिनमें **153 के पास `hasFree: true` discovery metadata है**। Chat model registry में **239 प्रदाता / 3,059 विशिष्ट provider-model pairs / 1,475 raw model IDs** शामिल हैं; अलग free-budget catalog में **471 per-model rows**, **34 recurring pools** और **52 recurring/keyless free-forever providers** हैं। ये जानबूझकर अलग-अलग denominators हैं; परिभाषाएँ और pool-deduped calculations [प्रदाता संदर्भ](docs/reference/PROVIDER_REFERENCE.md) और [निःशुल्क स्तर](docs/reference/FREE_TIERS.md) में उपलब्ध हैं।
@@ -1303,7 +1303,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 भी इसे रनटाइमNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 भाषाTypeScript 6.0 — src/ और open-sse/ में 100% TypeScript (v2.0 से कोर में शून्य any) फ़्रेमवर्कNext.js 16 + React 19 + Tailwind CSS 4 - डेटाबेसbetter-sqlite3 (SQLite, WAL जर्नलिंग) + LowDB (JSON लेगेसी) — 147 डोमेन मॉड्यूल, 230 माइग्रेशन + डेटाबेसbetter-sqlite3 (SQLite, WAL जर्नलिंग) + LowDB (JSON लेगेसी) — 147 डोमेन मॉड्यूल, 232 माइग्रेशन मेमोरीSQLite FTS5 पूर्ण-पाठ + int8-क्वांटाइज़्ड वेक्टर एम्बेडिंग, टाइपित क्षय स्कीमाZod 4 — MCP टूल I/O सत्यापन + API अनुबंध प्रोटोकॉलMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/hi/llm.txt b/docs/i18n/hi/llm.txt index cc164b02038..72438f9296a 100644 --- a/docs/i18n/hi/llm.txt +++ b/docs/i18n/hi/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/hr/README.md b/docs/i18n/hr/README.md index e887f68d46a..afe2980b143 100644 --- a/docs/i18n/hr/README.md +++ b/docs/i18n/hr/README.md @@ -15,7 +15,7 @@ **Izradio [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Nikad ne prestajte programirati. Svaki AI alat → 383 pružatelj — više od 150 besplatnih — putem jedne krajnje točke. Claude Code, Codex, Cursor, Cline, Copilot i Antigravity u BESPLATNI Claude / GPT / Gemini s automatskim prebacivanjem. Kombinirana kompresija RTK + Caveman štedi 15–95 % tokena (prosječno ~89 %) — nikad ne dosegnite ograničenja. 383 AI pružatelj · više od 150 besplatnih razina · ~1,62 mlrd. besplatnih tokena mjesečno · 20 strategija usmjeravanja · početak po cijeni od 0 USD. +OmniRoute — Nikad ne prestajte programirati. Svaki AI alat → 384 pružatelj — više od 150 besplatnih — putem jedne krajnje točke. Claude Code, Codex, Cursor, Cline, Copilot i Antigravity u BESPLATNI Claude / GPT / Gemini s automatskim prebacivanjem. Kombinirana kompresija RTK + Caveman štedi 15–95 % tokena (prosječno ~89 %) — nikad ne dosegnite ograničenja. 384 AI pružatelj · više od 150 besplatnih razina · ~1,62 mlrd. besplatnih tokena mjesečno · 20 strategija usmjeravanja · početak po cijeni od 0 USD.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------------- | :-------------------: | :---------: | :-------------------: | -| 🌐 Pružatelji | 352 | **358** | 383 | +| 🌐 Pružatelji | 352 | **358** | 384 | | 🧠 Jedinstveni ID-ovi chat modela | 1320 | **1374** | 1491 | | 🖼️ Most modaliteta | 🆕 vid + zvuk + video | ✓ | ✓ | | 📡 Besplatni katalog Radara | 🆕 uz prijavu | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Značajke 🎯 Kombinacije - 🌐 Pružatelji + 🌐 Pružatelji 🔌 CLI i MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Obećanje — Jedna krajnja točka i 383 pružatelj. Automatsko prebacivanje održava usmjeravanje dok je dostupno drugo ispravno odredište. Šest stupova: otporno prebacivanje između 383 pružatelja · do 95 % uštede tokena za prihvatljiva radna opterećenja · $0 za početak uz više od 150 besplatnih paketa i 52 besplatna pružatelja koji se obnavljaju ili su trajno besplatni bez ključa · 39 CLI/agentskih integracija kroz jednu konfiguraciju · kompatibilnost s OpenAI, Claude, Gemini i Responses API na /v1 · produkcijske kontrole, uključujući prekidače strujnog kruga, TLS stealth, MCP sa 110 alata, A2A, memoriju, zaštitne mehanizme, evaluacije i više od 56 000 statičkih deklaracija testova u više od 7 800 praćenih datoteka testova. +Obećanje — Jedna krajnja točka i 384 pružatelj. Automatsko prebacivanje održava usmjeravanje dok je dostupno drugo ispravno odredište. Šest stupova: otporno prebacivanje između 384 pružatelja · do 95 % uštede tokena za prihvatljiva radna opterećenja · $0 za početak uz više od 150 besplatnih paketa i 52 besplatna pružatelja koji se obnavljaju ili su trajno besplatni bez ključa · 39 CLI/agentskih integracija kroz jednu konfiguraciju · kompatibilnost s OpenAI, Claude, Gemini i Responses API na /v1 · produkcijske kontrole, uključujući prekidače strujnog kruga, TLS stealth, MCP sa 110 alata, A2A, memoriju, zaštitne mehanizme, evaluacije i više od 56 000 statičkih deklaracija testova u više od 7 800 praćenih datoteka testova.

@@ -510,7 +510,7 @@ Svih **20** strategija — kombinirajte ih po koraku kombinacije: -Po čemu se OmniRoute izdvaja — aktualni snimak značajki u usporedbi s 9router, OpenRouter, CLIProxyAPI i LiteLLM kroz 13 mogućnosti. OmniRoute: 383 pružatelj, više od 150 ugrađenih besplatnih tarifa, 20 strategija usmjeravanja, kompresija tokena uz 12 mehanizama, ugrađeni MCP poslužitelj sa 110 alata, A2A protokol agenata, trajna memorija, zaštitne mjere, agenti u oblaku, prikrivanje TLS otiska, Desktop/Termux/PWA i 67 lokalizacija korisničkog sučelja. OmniRoute je licenciran pod MIT licencom i može se samostalno ugostiti. Mogućnosti i brojke konkurenata mogu se promijeniti; pogledajte povezanu metodologiju. +Po čemu se OmniRoute izdvaja — aktualni snimak značajki u usporedbi s 9router, OpenRouter, CLIProxyAPI i LiteLLM kroz 13 mogućnosti. OmniRoute: 384 pružatelj, više od 150 ugrađenih besplatnih tarifa, 20 strategija usmjeravanja, kompresija tokena uz 12 mehanizama, ugrađeni MCP poslužitelj sa 110 alata, A2A protokol agenata, trajna memorija, zaštitne mjere, agenti u oblaku, prikrivanje TLS otiska, Desktop/Termux/PWA i 67 lokalizacija korisničkog sučelja. OmniRoute je licenciran pod MIT licencom i može se samostalno ugostiti. Mogućnosti i brojke konkurenata mogu se promijeniti; pogledajte povezanu metodologiju. 📊 Cjelovita metodologija & pojedinosti o svakoj značajci u usporedbi s 9router, OpenRouter, CLIProxyAPI & LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -613,7 +613,7 @@ trenutačnom katalogu na **[radar.omniroute.online/planos](https://radar.omnirou - **🖼️ Nove krajnje točke** — `/v1/ocr` (Mistral OCR) i `/v1/audio/translations` (u stilu Whispera) zaokružuju medijski sloj. → [Referenca API-ja](docs/reference/API_REFERENCE.md) - **🎨 Generiranje slika / videozapisa / zvuka** — jedan API za medije: xAI Grok Imagine i Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind te pružatelji govora poput ElevenLabsa. → [Referenca API-ja](docs/reference/API_REFERENCE.md) - **🌍 Implementacija i operacije** — `basePath` obrnutog proxyja, automatsko otkrivanje jezika preglednika, praćenje uređaja po ključu, MITM povjerenje bez root ovlasti, lokalizacija zh-TW. → [Okruženje](docs/reference/ENVIRONMENT.md) -- **🤝 Više pružatelja i agenata** — agenti u oblaku (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) s prijavom putem preglednika + OAuth, prvoklasna kartica za Ollamu, Claude Opus 5 i Sonnet 5, službeno partnerstvo s Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… i osvježeni **katalog s 383 pružateljem**. → [Pružatelji](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Više pružatelja i agenata** — agenti u oblaku (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) s prijavom putem preglednika + OAuth, prvoklasna kartica za Ollamu, Claude Opus 5 i Sonnet 5, službeno partnerstvo s Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… i osvježeni **katalog s 384 pružateljem**. → [Pružatelji](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparentnost usmjeravanja** — svaki odgovor sadrži zaglavlje `X-OmniRoute-Decision` koje navodi strategiju/pružatelja/latenciju koji su ga poslužili, nova strategija kombinacije `cache-optimized` + faktor usmjeravanja `cacheAffinity` u Automatskim kombinacijama vraća ponovljene zahtjeve vezi koja sadrži predmemorirani prefiks, a krajnja točka samo za čitanje `/v1/auto-combo/{channel}/candidates` izlaže aktivni skup kandidata kanala `auto/*`. → [Automatske kombinacije](docs/routing/AUTO-COMBO.md). - **⚡ Lokalne performanse i infrastruktura** — lokalni Redis jednim klikom, implementatori releja za Cloudflare Workers / Deno Deploy, Bifrost i Mux kao nadzirane ugrađene usluge. → [Ugrađene usluge](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Također uključeno** — okvir za dodatke i tržnica, okviri vještina Omni/Agent/GitHub, integracija s trezorom Obsidiana (22 MCP alata), API-ji Batch i Files kompatibilni s OpenAI-jem, semantička predmemorija odgovora, gamifikacija s ljestvicama, otkrivanje ACP agenata (15 ugrađenih agenata), zakazani izvoz zapisnika u BigQuery, paralelno grananje prema više modela `auto/chaos`, most za Telegram bota, upravitelj verzija unutar aplikacije i poredak besplatnih pružatelja usluga LMArena-ELO. → [Dokumentacija](docs/README.md) @@ -697,11 +697,11 @@ povijesti vaše ljuske. → [CLI Integracije](docs/guides/CLI-INTEGRATIONS.md)
-## 🌐 383 pružatelj AI usluga — 153 označena u katalogu kao besplatna +## 🌐 384 pružatelj AI usluga — 153 označena u katalogu kao besplatna
-> **383 registrirani pružatelj** u kanonskim zbirkama za chat, medije, pretraživanje, lokalne usluge, agente u oblaku i sustav, uključujući **153 pružatelja s metapodacima za otkrivanje `hasFree: true`**. Registar chat modela obuhvaća **239 pružatelja / 3,059 različitih parova pružatelj-model / 1,475 neobrađenih ID-ova modela**; zasebni katalog besplatnih kvota ima **471 redak po modelu**, **34 ponavljajuća skupa** i **52 pružatelja s ponavljajućom/besplatnom uslugom bez ključa, trajno besplatna**. Ti su nazivnici namjerno različiti; definicije i izračuni s deduplikacijom skupova dostupni su u dokumentima [Referenca pružatelja](docs/reference/PROVIDER_REFERENCE.md) i [Besplatne razine](docs/reference/FREE_TIERS.md). +> **384 registrirani pružatelj** u kanonskim zbirkama za chat, medije, pretraživanje, lokalne usluge, agente u oblaku i sustav, uključujući **153 pružatelja s metapodacima za otkrivanje `hasFree: true`**. Registar chat modela obuhvaća **239 pružatelja / 3,059 različitih parova pružatelj-model / 1,475 neobrađenih ID-ova modela**; zasebni katalog besplatnih kvota ima **471 redak po modelu**, **34 ponavljajuća skupa** i **52 pružatelja s ponavljajućom/besplatnom uslugom bez ključa, trajno besplatna**. Ti su nazivnici namjerno različiti; definicije i izračuni s deduplikacijom skupova dostupni su u dokumentima [Referenca pružatelja](docs/reference/PROVIDER_REFERENCE.md) i [Besplatne razine](docs/reference/FREE_TIERS.md).
@@ -1305,7 +1305,7 @@ Kanonske metrike na dan 2026-10-18: **1.976 jedinstvenih videozapisa** · **20.2 Izvršno okruženjeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 JezikTypeScript 6.0 — 100% TypeScript u src/ i open-sse/ (nula any u jezgri od verzije v2.0) OkvirNext.js 16 + React 19 + Tailwind CSS 4 - Baza podatakabetter-sqlite3 (SQLite, WAL zapisivanje) + LowDB (naslijeđeni JSON) — 147 domenskih modula, 230 migracija + Baza podatakabetter-sqlite3 (SQLite, WAL zapisivanje) + LowDB (naslijeđeni JSON) — 147 domenskih modula, 232 migracija MemorijaSQLite FTS5 pretraživanje cijelog teksta + vektorski ugrađeni zapisi kvantizirani na int8, tipizirano opadanje ShemeZod 4 — provjera I/O-a MCP alata + API ugovori ProtokoliMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/hr/llm.txt b/docs/i18n/hr/llm.txt index 4c1285b9761..cf10775b6a4 100644 --- a/docs/i18n/hr/llm.txt +++ b/docs/i18n/hr/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/hu/README.md b/docs/i18n/hu/README.md index ac5dae6afd5..16981960399 100644 --- a/docs/i18n/hu/README.md +++ b/docs/i18n/hu/README.md @@ -15,7 +15,7 @@ **Készítette: [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Soha ne hagyd abba a programozást. Minden AI-eszköz → 383 szolgáltató — ebből 150+ ingyenes — egyetlen végponton keresztül. Claude Code, Codex, Cursor, Cline, Copilot és Antigravity → INGYENES Claude / GPT / Gemini automatikus tartalékútválasztással. Az RTK + Caveman rétegezett tömörítés 15–95%-kal kevesebb tokent használ (átlagosan ~89%) — soha többé nem éred el a korlátokat. 383 AI-szolgáltató · 150+ ingyenes csomag · ~1,62B ingyenes token/hó · 20 útválasztási stratégia · 0 dollárról indulhatsz. +OmniRoute — Soha ne hagyd abba a programozást. Minden AI-eszköz → 384 szolgáltató — ebből 150+ ingyenes — egyetlen végponton keresztül. Claude Code, Codex, Cursor, Cline, Copilot és Antigravity → INGYENES Claude / GPT / Gemini automatikus tartalékútválasztással. Az RTK + Caveman rétegezett tömörítés 15–95%-kal kevesebb tokent használ (átlagosan ~89%) — soha többé nem éred el a korlátokat. 384 AI-szolgáltató · 150+ ingyenes csomag · ~1,62B ingyenes token/hó · 20 útválasztási stratégia · 0 dollárról indulhatsz.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------- | :-----------------------: | :---------: | :-------------------: | -| 🌐 Szolgáltatók | 352 | **358** | 383 | +| 🌐 Szolgáltatók | 352 | **358** | 384 | | 🧠 Egyedi chatmodell-azonosítók | 1320 | **1374** | 1491 | | 🖼️ Modalitás-híd | 🆕 vision + audio + video | ✓ | ✓ | | 📡 Radar ingyenes katalógusa | 🆕 opt-in | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funkciók 🎯 Kombók - 🌐 Szolgáltatók + 🌐 Szolgáltatók 🔌 CLI és MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Az ígéret — Egyetlen végpont és 383 szolgáltató. Az automatikus tartalékútválasztás folytatja az útválasztást, amíg elérhető egy másik egészséges célpont. Hat pillér: rugalmas tartalékútválasztás 383 szolgáltatón keresztül · akár 95%-os tokenmegtakarítás a jogosult munkaterheléseknél · 0 dollárról indulás 150+ ingyenes csomaggal és 52 rendszeresen megújuló vagy kulcs nélküli, örökre ingyenes szolgáltatóval · 39 CLI-/ügynökintegráció egyetlen konfiguráción keresztül · OpenAI-, Claude-, Gemini- és Responses API-kompatibilitás a /v1 címen · éles környezethez használható vezérlők, többek között megszakítók, TLS-rejtés, 110 MCP-eszköz, A2A, memória, védelmi korlátok, kiértékelések, valamint több mint 56 000 statikus tesztdeklaráció több mint 7 800 nyomon követett tesztfájlban. +Az ígéret — Egyetlen végpont és 384 szolgáltató. Az automatikus tartalékútválasztás folytatja az útválasztást, amíg elérhető egy másik egészséges célpont. Hat pillér: rugalmas tartalékútválasztás 384 szolgáltatón keresztül · akár 95%-os tokenmegtakarítás a jogosult munkaterheléseknél · 0 dollárról indulás 150+ ingyenes csomaggal és 52 rendszeresen megújuló vagy kulcs nélküli, örökre ingyenes szolgáltatóval · 39 CLI-/ügynökintegráció egyetlen konfiguráción keresztül · OpenAI-, Claude-, Gemini- és Responses API-kompatibilitás a /v1 címen · éles környezethez használható vezérlők, többek között megszakítók, TLS-rejtés, 110 MCP-eszköz, A2A, memória, védelmi korlátok, kiértékelések, valamint több mint 56 000 statikus tesztdeklaráció több mint 7 800 nyomon követett tesztfájlban.

@@ -510,7 +510,7 @@ Mind a **20** stratégia — szabadon kombinálhatók a kombó egyes lépéseibe -Mi különbözteti meg az OmniRoute-ot — egy dátummal ellátott funkció-összehasonlítás a 9router, az OpenRouter, a CLIProxyAPI és a LiteLLM között 13 képesség tekintetében. OmniRoute: 383 szolgáltató, több mint 150 beépített ingyenes csomag, 20 útválasztási stratégia, 12 motoros tokenkompresszió, beépített MCP-szerver 110 eszközzel, A2A-ügynökprotokoll, tartós memória, védelmi korlátozások, felhőügynökök, TLS-ujjlenyomat-alapú rejtőzködés, Desktop/Termux/PWA és 67 i18n felhasználói felületi nyelvi terület. Az OmniRoute MIT-licencelt, és saját infrastruktúrán üzemeltethető. A versenytársak képességei és számai változhatnak; lásd a hivatkozott módszertant. +Mi különbözteti meg az OmniRoute-ot — egy dátummal ellátott funkció-összehasonlítás a 9router, az OpenRouter, a CLIProxyAPI és a LiteLLM között 13 képesség tekintetében. OmniRoute: 384 szolgáltató, több mint 150 beépített ingyenes csomag, 20 útválasztási stratégia, 12 motoros tokenkompresszió, beépített MCP-szerver 110 eszközzel, A2A-ügynökprotokoll, tartós memória, védelmi korlátozások, felhőügynökök, TLS-ujjlenyomat-alapú rejtőzködés, Desktop/Termux/PWA és 67 i18n felhasználói felületi nyelvi terület. Az OmniRoute MIT-licencelt, és saját infrastruktúrán üzemeltethető. A versenytársak képességei és számai változhatnak; lásd a hivatkozott módszertant. 📊 Teljes módszertan és funkciónkénti részletek a 9routerrel, az OpenRouterrel, a CLIProxyAPI-val és a LiteLLM-mel való összehasonlításról → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ A Radar használata opcionális, és kizárólag GET-kéréseket alkalmaz. Az Om - **🖼️ Új végpontok** — a `/v1/ocr` (Mistral OCR) és a `/v1/audio/translations` (Whisper-stílusú) teljessé teszi a médiavégpontokat. → [API-referencia](docs/reference/API_REFERENCE.md) - **🎨 Kép-, videó- és hanggenerálás** — egyetlen API a médiához: xAI Grok Imagine és Novita AI videó, ComfyUI, Magnific, Adobe Firefly, Segmind, valamint olyan beszédszolgáltatók, mint az ElevenLabs. → [API-referencia](docs/reference/API_REFERENCE.md) - **🌍 Telepítés és üzemeltetés** — reverse proxy `basePath`, böngészőnyelv automatikus felismerése, kulcsonkénti eszközkövetés, root nélküli MITM-megbízhatóság, zh-TW-lokalizáció. → [Környezet](docs/reference/ENVIRONMENT.md) -- **🤝 További szolgáltatók és ügynökök** — felhőügynökök (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) böngésző- és OAuth-bejelentkezéssel, első osztályú Ollama-kártya, Claude Opus 5 és Sonnet 5, hivatalos Kimi-partnerség (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… és egy frissített **383 szolgáltatós katalógus**. → [Szolgáltatók](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 További szolgáltatók és ügynökök** — felhőügynökök (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) böngésző- és OAuth-bejelentkezéssel, első osztályú Ollama-kártya, Claude Opus 5 és Sonnet 5, hivatalos Kimi-partnerség (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… és egy frissített **384 szolgáltatós katalógus**. → [Szolgáltatók](docs/reference/PROVIDER_REFERENCE.md) - **📡 Routingátláthatóság** — minden válasz tartalmaz egy `X-OmniRoute-Decision` fejlécet, amely megnevezi a kiszolgálásáért felelős stratégiát/szolgáltatót/késleltetést; az új `cache-optimized` kombinációs stratégia és az Auto-Combo `cacheAffinity` faktora a megismételt kéréseket visszairányítja a gyorsítótárazott előtagot birtokló kapcsolathoz; a csak olvasható `/v1/auto-combo/{channel}/candidates` végpont pedig közzéteszi egy `auto/*` csatorna élő jelöltkészletét. → [Auto-Combo](docs/routing/AUTO-COMBO.md). - **⚡ Helyi teljesítmény és infrastruktúra** — egykattintásos helyi Redis, Cloudflare Workers / Deno Deploy relay-telepítők, valamint felügyelt beágyazott szolgáltatásként futó Bifrost és Mux. → [Beágyazott szolgáltatások](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 A csomag része továbbá** — bővítménykeretrendszer és piactér, Omni/Agent/GitHub skills-keretrendszerek, Obsidian vault-integráció (22 MCP-eszköz), OpenAI-kompatibilis Batch- és Files API-k, szemantikus válaszgyorsítótár, ranglistákkal rendelkező gamifikáció, ACP-ügynökfelderítés (15 beépített ügynök), ütemezett naplóexport a BigQuery-be, `auto/chaos` párhuzamos, többmodelles fan-out, Telegram-bot-híd, alkalmazáson belüli verziókezelő, valamint az LMArena-ELO ingyenes szolgáltatói rangsora. → [Dokumentáció](docs/README.md) @@ -689,11 +689,11 @@ megóvja a titkos adatokat a shell előzményétől. → [CLI Integrációk](doc
-## 🌐 383 AI-szolgáltató — közülük 153 katalógus szerint ingyenes +## 🌐 384 AI-szolgáltató — közülük 153 katalógus szerint ingyenes
-> **383 regisztrált szolgáltató** a kanonikus csevegési, média-, keresési, helyi, felhőügynök- és rendszerkollekciókban, köztük **153 olyan szolgáltató, amely `hasFree: true` felfedezési metaadatot tartalmaz**. A csevegési modellregiszter **239 szolgáltatót / 3 059 különálló szolgáltató-modell párt / 1 475 nyers modellazonosítót** tartalmaz; a különálló ingyenes keretkatalógusban **471 modellenkénti sor**, **34 ismétlődő keret** és **52 ismétlődő vagy kulcs nélküli, örökre ingyenes szolgáltató** található. Ezek szándékosan eltérő nevezők; a definíciók és a keretduplikációktól megtisztított számítások a [Szolgáltatói referenciában](docs/reference/PROVIDER_REFERENCE.md) és az [Ingyenes csomagok](docs/reference/FREE_TIERS.md) oldalon találhatók. +> **384 regisztrált szolgáltató** a kanonikus csevegési, média-, keresési, helyi, felhőügynök- és rendszerkollekciókban, köztük **153 olyan szolgáltató, amely `hasFree: true` felfedezési metaadatot tartalmaz**. A csevegési modellregiszter **239 szolgáltatót / 3 059 különálló szolgáltató-modell párt / 1 475 nyers modellazonosítót** tartalmaz; a különálló ingyenes keretkatalógusban **471 modellenkénti sor**, **34 ismétlődő keret** és **52 ismétlődő vagy kulcs nélküli, örökre ingyenes szolgáltató** található. Ezek szándékosan eltérő nevezők; a definíciók és a keretduplikációktól megtisztított számítások a [Szolgáltatói referenciában](docs/reference/PROVIDER_REFERENCE.md) és az [Ingyenes csomagok](docs/reference/FREE_TIERS.md) oldalon találhatók.
@@ -1292,7 +1292,7 @@ Kanonikus mérőszámok 2026-10-18-án: **1.976 egyedi videó** · **20.220.289 Futási környezetNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 NyelvTypeScript 6.0 — 100%-ban TypeScript a src/ és open-sse/ könyvtárakban (a magban a v2.0 óta nincs any) KeretrendszerNext.js 16 + React 19 + Tailwind CSS 4 - Adatbázisbetter-sqlite3 (SQLite, WAL-naplózás) + LowDB (örökölt JSON) — 147 tartományi modul, 230 migráció + Adatbázisbetter-sqlite3 (SQLite, WAL-naplózás) + LowDB (örökölt JSON) — 147 tartományi modul, 232 migráció MemóriaSQLite FTS5 teljes szöveges keresés + int8-kvantált vektorbeágyazások, típusos lecsengés SémákZod 4 — MCP-eszközök I/O-validációja + API-szerződések ProtokollokMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/hu/llm.txt b/docs/i18n/hu/llm.txt index 784513ef39b..47e95730321 100644 --- a/docs/i18n/hu/llm.txt +++ b/docs/i18n/hu/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/hy/README.md b/docs/i18n/hy/README.md index ab43784495b..f509a054dba 100644 --- a/docs/i18n/hy/README.md +++ b/docs/i18n/hy/README.md @@ -15,7 +15,7 @@ **Ստեղծողը՝ [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Երբեք մի՛ դադարեք կոդ գրել։ Յուրաքանչյուր AI գործիք → 383 մատակարար — 150+ անվճար — մեկ վերջնակետի միջոցով։ Claude Code, Codex, Cursor, Cline, Copilot և Antigravity՝ FREE Claude / GPT / Gemini համակարգերում՝ ավտոմատ պահուստային անցմամբ։ RTK + Caveman-ի համակցված սեղմումը խնայում է նշանների 15–95%-ը (միջինը՝ ~89%)՝ երբեք չհասնելով սահմանաչափերին։ 383 AI մատակարար · 150+ անվճար մակարդակ · ~1.62B անվճար նշան/ամիս · երթուղավորման 20 ռազմավարություն · մեկնարկը՝ $0։ +OmniRoute — Երբեք մի՛ դադարեք կոդ գրել։ Յուրաքանչյուր AI գործիք → 384 մատակարար — 150+ անվճար — մեկ վերջնակետի միջոցով։ Claude Code, Codex, Cursor, Cline, Copilot և Antigravity՝ FREE Claude / GPT / Gemini համակարգերում՝ ավտոմատ պահուստային անցմամբ։ RTK + Caveman-ի համակցված սեղմումը խնայում է նշանների 15–95%-ը (միջինը՝ ~89%)՝ երբեք չհասնելով սահմանաչափերին։ 384 AI մատակարար · 150+ անվճար մակարդակ · ~1.62B անվճար նշան/ամիս · երթուղավորման 20 ռազմավարություն · մեկնարկը՝ $0։
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------------- | :----------------------------------: | :---------: | :-------------------: | -| 🌐 Մատակարարներ | 352 | **358** | 383 | +| 🌐 Մատակարարներ | 352 | **358** | 384 | | 🧠 Զրույցի եզակի մոդելների ID-ներ | 1320 | **1374** | 1491 | | 🖼️ Մոդալությունների կամուրջ | 🆕 տեսողություն + աուդիո + տեսանյութ | ✓ | ✓ | | 📡 Radar-ի անվճար կատալոգ | 🆕 ընտրովի միացում | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Հնարավորություններ 🎯 Կոմբինացիաներ - 🌐 Մատակարարներ + 🌐 Մատակարարներ 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Խոստումը՝ մեկ վերջնակետ և 383 մատակարար։ Ավտոմատ պահուստային անցումը պահպանում է ուղղորդումը, քանի դեռ հասանելի է մեկ այլ առողջ թիրախ։ Վեց հիմնասյուներ՝ դիմացկուն պահուստային անցում 383 մատակարարների միջև · համապատասխան աշխատանքային ծանրաբեռնվածությունների դեպքում մինչև 95% խնայողություն թոքենների վրա · մեկնարկը՝ $0-ից՝ 150+ անվճար մակարդակներով և 52 պարբերական/առանց բանալիի՝ ընդմիշտ անվճար մատակարարներով · 39 CLI/գործակալի ինտեգրում մեկ կարգավորման միջոցով · OpenAI, Claude, Gemini և Responses API-ի համատեղելիություն /v1-ում · արտադրական կառավարման միջոցներ, ներառյալ շղթայի անջատիչներ, TLS stealth, MCP-ի 110 գործիք, A2A, հիշողություն, պաշտպանիչ սահմանափակումներ, գնահատումներ և 56,000+ ստատիկ թեստային հայտարարություններ՝ 7,800+ վերահսկվող թեստային ֆայլերում։ +Խոստումը՝ մեկ վերջնակետ և 384 մատակարար։ Ավտոմատ պահուստային անցումը պահպանում է ուղղորդումը, քանի դեռ հասանելի է մեկ այլ առողջ թիրախ։ Վեց հիմնասյուներ՝ դիմացկուն պահուստային անցում 384 մատակարարների միջև · համապատասխան աշխատանքային ծանրաբեռնվածությունների դեպքում մինչև 95% խնայողություն թոքենների վրա · մեկնարկը՝ $0-ից՝ 150+ անվճար մակարդակներով և 52 պարբերական/առանց բանալիի՝ ընդմիշտ անվճար մատակարարներով · 39 CLI/գործակալի ինտեգրում մեկ կարգավորման միջոցով · OpenAI, Claude, Gemini և Responses API-ի համատեղելիություն /v1-ում · արտադրական կառավարման միջոցներ, ներառյալ շղթայի անջատիչներ, TLS stealth, MCP-ի 110 գործիք, A2A, հիշողություն, պաշտպանիչ սահմանափակումներ, գնահատումներ և 56,000+ ստատիկ թեստային հայտարարություններ՝ 7,800+ վերահսկվող թեստային ֆայլերում։

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -Ինչով է OmniRoute-ն առանձնանում — հնարավորությունների ժամանակային պատկեր՝ 9router-ի, OpenRouter-ի, CLIProxyAPI-ի և LiteLLM-ի համեմատությամբ՝ 13 հնարավորությունների շրջանակում։ OmniRoute՝ 383 մատակարար, ավելի քան 150 ներկառուցված անվճար մակարդակ, երթուղավորման 20 ռազմավարություն, թոքենների սեղմում 12 շարժիչով, ներկառուցված MCP սերվեր՝ 110 գործիքով, A2A գործակալների արձանագրություն, մշտական հիշողություն, պաշտպանիչ սահմանափակումներ, ամպային գործակալներ, TLS մատնահետքի քողարկում, Desktop/Termux/PWA և i18n UI-ի 67 տեղայնացման լեզու։ OmniRoute-ը լիցենզավորված է MIT լիցենզիայով և կարող է հոսթվել ինքնուրույն։ Մրցակիցների հնարավորություններն ու քանակները կարող են փոխվել․ տես կից մեթոդաբանությունը։ +Ինչով է OmniRoute-ն առանձնանում — հնարավորությունների ժամանակային պատկեր՝ 9router-ի, OpenRouter-ի, CLIProxyAPI-ի և LiteLLM-ի համեմատությամբ՝ 13 հնարավորությունների շրջանակում։ OmniRoute՝ 384 մատակարար, ավելի քան 150 ներկառուցված անվճար մակարդակ, երթուղավորման 20 ռազմավարություն, թոքենների սեղմում 12 շարժիչով, ներկառուցված MCP սերվեր՝ 110 գործիքով, A2A գործակալների արձանագրություն, մշտական հիշողություն, պաշտպանիչ սահմանափակումներ, ամպային գործակալներ, TLS մատնահետքի քողարկում, Desktop/Termux/PWA և i18n UI-ի 67 տեղայնացման լեզու։ OmniRoute-ը լիցենզավորված է MIT լիցենզիայով և կարող է հոսթվել ինքնուրույն։ Մրցակիցների հնարավորություններն ու քանակները կարող են փոխվել․ տես կից մեթոդաբանությունը։ 📊 Ամբողջական մեթոդաբանություն և յուրաքանչյուր հնարավորության մանրամասներ՝ 9router-ի, OpenRouter-ի, CLIProxyAPI-ի և LiteLLM-ի համեմատությամբ → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar-ի օգտագործումը կամընտիր է, և այն աշխատու - **🖼️ Նոր endpoint-ներ** — `/v1/ocr` (Mistral OCR) և `/v1/audio/translations` (Whisper-ի ոճով) լրացնում են մեդիայի հնարավորությունները։ → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 Պատկերների / տեսանյութերի / աուդիոյի գեներացում** — մեդիայի համար մեկ API՝ xAI Grok Imagine և Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind և ձայնային մատակարարներ, օրինակ՝ ElevenLabs։ → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 Տեղակայում և գործարկում** — հակադարձ պրոքսիի `basePath`, դիտարկիչի լեզվի ավտոմատ հայտնաբերում, յուրաքանչյուր բանալու սարքի հետագծում, root-less MITM վստահություն, zh-TW տեղայնացում։ → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 Ավելի շատ մատակարարներ և գործակալներ** — ամպային գործակալներ (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI)՝ դիտարկիչով + OAuth մուտքով, Ollama-ի առաջին կարգի քարտ, Claude Opus 5 և Sonnet 5, Kimi-ի պաշտոնական գործընկերություն (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… և թարմացված **383 մատակարարի կատալոգ**։ → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Ավելի շատ մատակարարներ և գործակալներ** — ամպային գործակալներ (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI)՝ դիտարկիչով + OAuth մուտքով, Ollama-ի առաջին կարգի քարտ, Claude Opus 5 և Sonnet 5, Kimi-ի պաշտոնական գործընկերություն (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… և թարմացված **384 մատակարարի կատալոգ**։ → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 Երթուղավորման թափանցիկություն** — յուրաքանչյուր պատասխան պարունակում է `X-OmniRoute-Decision` վերնագիր, որը նշում է այն սպասարկած ռազմավարությունը/մատակարարին/ուշացումը, նոր `cache-optimized` կոմբո ռազմավարություն + Auto-Combo-ի `cacheAffinity` գործոնը կրկնվող հարցումները վերադարձնում է քեշավորված նախածանցը պահող կապին, իսկ միայն ընթերցման համար նախատեսված `/v1/auto-combo/{channel}/candidates` endpoint-ը հրապարակում է `auto/*` ալիքի թեկնածուների ընթացիկ խումբը։ → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Տեղական արտադրողականություն և ենթակառուցվածք** — մեկ սեղմումով տեղական Redis, Cloudflare Workers / Deno Deploy relay-ի տեղակայման գործիքներ, Bifrost և Mux՝ վերահսկվող ներկառուցված ծառայությունների տեսքով։ → [Ներկառուցված ծառայություններ](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Ներառված են նաև** — պլագինների ֆրեյմվորք և շուկա, Omni/Agent/GitHub հմտությունների ֆրեյմվորքներ, Obsidian vault-ի ինտեգրում (22 MCP գործիք), OpenAI-ին համատեղելի Batch & Files API-ներ, սեմանտիկ պատասխանների քեշ, առաջատարների աղյուսակներով գեյմիֆիկացում, ACP գործակալների հայտնաբերում (15 ներկառուցված գործակալ), պլանավորված լոգերի արտահանում BigQuery, `auto/chaos` զուգահեռ բազմամոդելային fan-out, Telegram բոտի կամուրջ, հավելվածում ներկառուցված տարբերակների կառավարիչ և LMArena-ELO-ի անվճար մատակարարների վարկանիշներ։ → [Փաստաթղթեր](docs/README.md) @@ -689,11 +689,11 @@ omniroute configure codex # նաև՝ claude opencode qwen aider goose g
-## 🌐 383 AI մատակարար — 153-ը կատալոգում նշված են որպես անվճար +## 🌐 384 AI մատակարար — 153-ը կատալոգում նշված են որպես անվճար
-> **383 գրանցված մատակարար** կան կանոնական chat, media, search, local, cloud-agent և system հավաքածուներում, որոնցից **153-ն ունեն `hasFree: true` հայտնաբերման մետատվյալներ**։ Chat մոդելների ռեեստրը ներառում է **239 մատակարար / 3,059 տարբեր մատակարար-մոդել զույգ / 1,475 չմշակված մոդելի ID**․ առանձին անվճար բյուջեի կատալոգն ունի **471՝ ըստ մոդելի տող**, **34 պարբերական փունջ** և **52 պարբերական/բանալազուրկ՝ ընդմիշտ անվճար մատակարար**։ Սրանք դիտավորյալ տարբեր հայտարարիչներ են․ սահմանումները և փունջերի կրկնօրինակների հեռացմամբ հաշվարկները ներկայացված են [Մատակարարների տեղեկատուում](docs/reference/PROVIDER_REFERENCE.md) և [Անվճար մակարդակներ](docs/reference/FREE_TIERS.md) բաժնում։ +> **384 գրանցված մատակարար** կան կանոնական chat, media, search, local, cloud-agent և system հավաքածուներում, որոնցից **153-ն ունեն `hasFree: true` հայտնաբերման մետատվյալներ**։ Chat մոդելների ռեեստրը ներառում է **239 մատակարար / 3,059 տարբեր մատակարար-մոդել զույգ / 1,475 չմշակված մոդելի ID**․ առանձին անվճար բյուջեի կատալոգն ունի **471՝ ըստ մոդելի տող**, **34 պարբերական փունջ** և **52 պարբերական/բանալազուրկ՝ ընդմիշտ անվճար մատակարար**։ Սրանք դիտավորյալ տարբեր հայտարարիչներ են․ սահմանումները և փունջերի կրկնօրինակների հեռացմամբ հաշվարկները ներկայացված են [Մատակարարների տեղեկատուում](docs/reference/PROVIDER_REFERENCE.md) և [Անվճար մակարդակներ](docs/reference/FREE_TIERS.md) բաժնում։
@@ -1295,7 +1295,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1-ը նույնպե Կատարման միջավայրNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ԼեզուTypeScript 6.0 — 100% TypeScript src/ և open-sse/ պանակներում (v2.0-ից ի վեր core-ում զրո any) ՖրեյմվորքNext.js 16 + React 19 + Tailwind CSS 4 - Տվյալների բազաbetter-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — տիրույթի 147 մոդուլ, 230 միգրացիա + Տվյալների բազաbetter-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — տիրույթի 147 մոդուլ, 232 միգրացիա ՀիշողությունSQLite FTS5 ամբողջական տեքստի որոնում + int8-քվանտավորված վեկտորային ներդրումներ, տիպավորված մարում ՍխեմաներZod 4 — MCP գործիքների I/O վավերացում + API պայմանագրեր ԱրձանագրություններMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/hy/llm.txt b/docs/i18n/hy/llm.txt index 00420461978..6634995743d 100644 --- a/docs/i18n/hy/llm.txt +++ b/docs/i18n/hy/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/id/README.md b/docs/i18n/id/README.md index d25675393e3..95646cab55d 100644 --- a/docs/i18n/id/README.md +++ b/docs/i18n/id/README.md @@ -15,7 +15,7 @@ **Dibuat oleh [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Jangan pernah berhenti membuat kode. Setiap alat AI → 383 penyedia — 150+ gratis — melalui satu endpoint. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity ke Claude / GPT / Gemini GRATIS dengan fallback otomatis. Kompresi bertingkat RTK + Caveman menghemat 15–95% token (~89% rata-rata) — jangan pernah mencapai batas. 383 penyedia AI · 150+ tingkat gratis · ~1,62 miliar token gratis/bln · 20 strategi perutean · mulai dengan $0. +OmniRoute — Jangan pernah berhenti membuat kode. Setiap alat AI → 384 penyedia — 150+ gratis — melalui satu endpoint. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity ke Claude / GPT / Gemini GRATIS dengan fallback otomatis. Kompresi bertingkat RTK + Caveman menghemat 15–95% token (~89% rata-rata) — jangan pernah mencapai batas. 384 penyedia AI · 150+ tingkat gratis · ~1,62 miliar token gratis/bln · 20 strategi perutean · mulai dengan $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | -------------------------- | :---------------------: | :---------: | :-------------------: | -| 🌐 Penyedia | 352 | **358** | 383 | +| 🌐 Penyedia | 352 | **358** | 384 | | 🧠 ID model chat unik | 1320 | **1374** | 1491 | | 🖼️ Jembatan modalitas | 🆕 visi + audio + video | ✓ | ✓ | | 📡 Katalog gratis Radar | 🆕 ikut serta | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Fitur 🎯 Kombinasi - 🌐 Penyedia + 🌐 Penyedia 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Janji — Satu endpoint dan 383 penyedia. Fallback otomatis menjaga perutean tetap berjalan selama masih tersedia target sehat lainnya. Enam pilar: fallback tangguh di 383 penyedia · penghematan token hingga 95% pada beban kerja yang memenuhi syarat · $0 untuk memulai dengan 150+ tingkat gratis dan 52 penyedia gratis selamanya yang berulang/tanpa kunci · 39 integrasi CLI/agen melalui satu konfigurasi · kompatibilitas dengan API OpenAI, Claude, Gemini, dan Responses di /v1 · kontrol produksi termasuk pemutus sirkuit, penyamaran TLS, 110 alat MCP, A2A, memori, guardrail, evaluasi, dan 56.000+ deklarasi pengujian statis di lebih dari 7.800 file pengujian yang dilacak. +Janji — Satu endpoint dan 384 penyedia. Fallback otomatis menjaga perutean tetap berjalan selama masih tersedia target sehat lainnya. Enam pilar: fallback tangguh di 384 penyedia · penghematan token hingga 95% pada beban kerja yang memenuhi syarat · $0 untuk memulai dengan 150+ tingkat gratis dan 52 penyedia gratis selamanya yang berulang/tanpa kunci · 39 integrasi CLI/agen melalui satu konfigurasi · kompatibilitas dengan API OpenAI, Claude, Gemini, dan Responses di /v1 · kontrol produksi termasuk pemutus sirkuit, penyamaran TLS, 110 alat MCP, A2A, memori, guardrail, evaluasi, dan 56.000+ deklarasi pengujian statis di lebih dari 7.800 file pengujian yang dilacak.

@@ -510,7 +510,7 @@ Seluruh **20** strategi — padukan sesuai kebutuhan pada setiap langkah combo: -Keunggulan OmniRoute — snapshot fitur bertanggal vs 9router, OpenRouter, CLIProxyAPI, dan LiteLLM dalam 13 kapabilitas. OmniRoute: 383 provider, 150+ tingkat gratis bawaan, 20 strategi perutean, kompresi token dengan 12 engine, server MCP bawaan dengan 110 alat, protokol agen A2A, memori persisten, guardrails, agen cloud, stealth sidik jari TLS, Desktop/Termux/PWA, dan 67 lokal UI i18n. OmniRoute berlisensi MIT dan dapat di-host sendiri. Kapabilitas dan jumlah kompetitor dapat berubah; lihat metodologi yang ditautkan. +Keunggulan OmniRoute — snapshot fitur bertanggal vs 9router, OpenRouter, CLIProxyAPI, dan LiteLLM dalam 13 kapabilitas. OmniRoute: 384 provider, 150+ tingkat gratis bawaan, 20 strategi perutean, kompresi token dengan 12 engine, server MCP bawaan dengan 110 alat, protokol agen A2A, memori persisten, guardrails, agen cloud, stealth sidik jari TLS, Desktop/Termux/PWA, dan 67 lokal UI i18n. OmniRoute berlisensi MIT dan dapat di-host sendiri. Kapabilitas dan jumlah kompetitor dapat berubah; lihat metodologi yang ditautkan. 📊 Metodologi lengkap & detail per fitur vs 9router, OpenRouter, CLIProxyAPI & LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar bersifat pilihan dan hanya menggunakan GET. Klien OmniRoute tidak mengungg - **🖼️ Endpoint baru** — `/v1/ocr` (Mistral OCR) dan `/v1/audio/translations` (bergaya Whisper) melengkapi cakupan media. → [Referensi API](docs/reference/API_REFERENCE.md) - **🎨 Pembuatan gambar / video / audio** — satu API untuk media: xAI Grok Imagine & video Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind, dan penyedia ucapan seperti ElevenLabs. → [Referensi API](docs/reference/API_REFERENCE.md) - **🌍 Deployment & operasi** — `basePath` reverse-proxy, deteksi otomatis bahasa browser, pelacakan perangkat per kunci, trust MITM tanpa hak root, serta lokalisasi zh-TW. → [Lingkungan](docs/reference/ENVIRONMENT.md) -- **🤝 Lebih banyak penyedia & agen** — agen cloud (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) dengan browser + login OAuth, kartu Ollama kelas utama, Claude Opus 5 & Sonnet 5, kemitraan resmi Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… dan **katalog 383 penyedia** yang diperbarui. → [Penyedia](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Lebih banyak penyedia & agen** — agen cloud (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) dengan browser + login OAuth, kartu Ollama kelas utama, Claude Opus 5 & Sonnet 5, kemitraan resmi Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… dan **katalog 384 penyedia** yang diperbarui. → [Penyedia](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparansi perutean** — setiap respons membawa header `X-OmniRoute-Decision` yang menyebutkan strategi/penyedia/latensi yang melayaninya, strategi combo `cache-optimized` baru + faktor rute `cacheAffinity` Auto-Combo yang mengarahkan permintaan berulang kembali ke koneksi yang menyimpan prefiks dalam cache, serta endpoint hanya-baca `/v1/auto-combo/{channel}/candidates` yang menampilkan kumpulan kandidat aktif dari channel `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Performa & infrastruktur lokal** — Redis lokal sekali klik, deployer relay Cloudflare Workers / Deno Deploy, Bifrost & Mux sebagai layanan tersemat yang diawasi. → [Layanan Tersemat](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Juga tersedia** — framework plugin + marketplace, framework skill Omni/Agent/GitHub, integrasi vault Obsidian (22 alat MCP), API Batch & Files yang kompatibel dengan OpenAI, cache respons semantik, gamifikasi dengan papan peringkat, penemuan agen ACP (15 agen bawaan), ekspor log terjadwal ke BigQuery, fan-out paralel multi-model `auto/chaos`, bridge bot Telegram, pengelola versi dalam aplikasi, dan peringkat penyedia gratis LMArena-ELO. → [Dokumentasi](docs/README.md) @@ -689,11 +689,11 @@ tetap aman dari riwayat shell Anda. → [Integrasi CLI](docs/guides/CLI-INTEGRAT
-## 🌐 383 Penyedia AI — 153 Ditandai Katalog sebagai Gratis +## 🌐 384 Penyedia AI — 153 Ditandai Katalog sebagai Gratis
-> **383 penyedia terdaftar** di seluruh koleksi kanonis chat, media, pencarian, lokal, agen cloud, dan sistem, termasuk **153 yang memiliki metadata penemuan `hasFree: true`**. Registri model chat mencakup **239 penyedia / 3.059 pasangan penyedia-model berbeda / 1.475 ID model mentah**; katalog anggaran gratis terpisah memiliki **471 baris per-model**, **34 kumpulan berulang**, dan **52 penyedia gratis selamanya yang berulang/tanpa kunci**. Penyebut ini memang berbeda berdasarkan desain; definisi dan perhitungan yang dideduplikasi berdasarkan kumpulan tersedia di [Referensi Penyedia](docs/reference/PROVIDER_REFERENCE.md) dan [Tingkat Gratis](docs/reference/FREE_TIERS.md). +> **384 penyedia terdaftar** di seluruh koleksi kanonis chat, media, pencarian, lokal, agen cloud, dan sistem, termasuk **153 yang memiliki metadata penemuan `hasFree: true`**. Registri model chat mencakup **239 penyedia / 3.059 pasangan penyedia-model berbeda / 1.475 ID model mentah**; katalog anggaran gratis terpisah memiliki **471 baris per-model**, **34 kumpulan berulang**, dan **52 penyedia gratis selamanya yang berulang/tanpa kunci**. Penyebut ini memang berbeda berdasarkan desain; definisi dan perhitungan yang dideduplikasi berdasarkan kumpulan tersedia di [Referensi Penyedia](docs/reference/PROVIDER_REFERENCE.md) dan [Tingkat Gratis](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ Metrik kanonis pada 2026-10-18: **1.976 video unik** · **20.220.289 tayangan ya RuntimeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 BahasaTypeScript 6.0 — 100% TypeScript di seluruh src/ dan open-sse/ (nol any di inti sejak v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Basis databetter-sqlite3 (SQLite, pencatatan WAL) + LowDB (JSON lama) — 147 modul domain, 230 migrasi + Basis databetter-sqlite3 (SQLite, pencatatan WAL) + LowDB (JSON lama) — 147 modul domain, 232 migrasi MemoriFTS5 SQLite untuk teks lengkap + embedding vektor terkuantisasi int8, peluruhan bertipe SkemaZod 4 — validasi I/O alat MCP + kontrak API ProtokolMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/id/llm.txt b/docs/i18n/id/llm.txt index 1764cd24c68..03b9d639266 100644 --- a/docs/i18n/id/llm.txt +++ b/docs/i18n/id/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ig/README.md b/docs/i18n/ig/README.md index d65086c0da5..2f7baeb7895 100644 --- a/docs/i18n/ig/README.md +++ b/docs/i18n/ig/README.md @@ -15,7 +15,7 @@ **Ọ bụ [Diego Souza](https://github.com/diegosouzapw) kere ya** -OmniRoute — Akwụsịla ide koodu. Ngwaọrụ AI ọ bụla → ndị na-eweta ọrụ 383 — 150+ n’efu — site n’otu njedebe. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity banye na Claude / GPT / Gemini N’efu site na ndaghachi akpaghị aka. Mkpakọ RTK + Caveman jikọtara ọnụ na-echekwa 15–95% token (~89% nkezi) — agaghị eru oke. Ndị na-eweta AI 383 · ọkwa efu 150+ · ~1.62B token efu kwa ọnwa · usoro ntụgharị ụzọ 20 · $0 iji malite. +OmniRoute — Akwụsịla ide koodu. Ngwaọrụ AI ọ bụla → ndị na-eweta ọrụ 384 — 150+ n’efu — site n’otu njedebe. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity banye na Claude / GPT / Gemini N’efu site na ndaghachi akpaghị aka. Mkpakọ RTK + Caveman jikọtara ọnụ na-echekwa 15–95% token (~89% nkezi) — agaghị eru oke. Ndị na-eweta AI 384 · ọkwa efu 150+ · ~1.62B token efu kwa ọnwa · usoro ntụgharị ụzọ 20 · $0 iji malite.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | -------------------------------- | :--------------------: | :---------: | :-------------------: | -| 🌐 Ndị na-eweta ọrụ | 352 | **358** | 383 | +| 🌐 Ndị na-eweta ọrụ | 352 | **358** | 384 | | 🧠 NJirimara ụdị nkata pụrụ iche | 1320 | **1374** | 1491 | | 🖼️ Àkwà njikọ ụdị mgbasa ozi | 🆕 ọhụụ + ụda + vidiyo | ✓ | ✓ | | 📡 Katalọgụ efu Radar | 🆕 nhọrọ ịbanye | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Njirimara 🎯 Ngwakọta - 🌐 Ndị na-eweta ọrụ + 🌐 Ndị na-eweta ọrụ 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Nkwa ahụ — Otu endpoint na ndị na-enye ọrụ 383. Ndabere akpaghị aka na-eme ka nhazi okporo ụzọ gaa n’ihu mgbe e nwere ebe ọzọ dị mma. Ogidi isii: ndabere siri ike n’ofe ndị na-enye ọrụ 383 · ruo 95% nchekwa token n’ọrụ tozuru oke · $0 iji malite na ọkwa ọrụ n’efu 150+ na ndị na-enye ọrụ 52 na-enye ọrụ n’efu ruo mgbe ebighị ebi, ma ọ bụ na-enweghị igodo · njikọ CLI/agent 39 site n’otu nhazi · ndakọrịta OpenAI, Claude, Gemini na Responses API na /v1 · njikwa mmepụta gụnyere circuit breakers, TLS stealth, MCP 110 tools, A2A, ebe nchekwa, guardrails, evals na nkwupụta ule static karịrị 56,000 n’ofe faịlụ ule a na-enyocha karịrị 7,800. +Nkwa ahụ — Otu endpoint na ndị na-enye ọrụ 384. Ndabere akpaghị aka na-eme ka nhazi okporo ụzọ gaa n’ihu mgbe e nwere ebe ọzọ dị mma. Ogidi isii: ndabere siri ike n’ofe ndị na-enye ọrụ 384 · ruo 95% nchekwa token n’ọrụ tozuru oke · $0 iji malite na ọkwa ọrụ n’efu 150+ na ndị na-enye ọrụ 52 na-enye ọrụ n’efu ruo mgbe ebighị ebi, ma ọ bụ na-enweghị igodo · njikọ CLI/agent 39 site n’otu nhazi · ndakọrịta OpenAI, Claude, Gemini na Responses API na /v1 · njikwa mmepụta gụnyere circuit breakers, TLS stealth, MCP 110 tools, A2A, ebe nchekwa, guardrails, evals na nkwupụta ule static karịrị 56,000 n’ofe faịlụ ule a na-enyocha karịrị 7,800.

@@ -510,7 +510,7 @@ Usoro **20** niile — gwakọta ma dakọtara ha na nzọụkwụ combo ọ b -Ihe na-eme OmniRoute ka ọ dị iche — nseta atụmatụ nke nwere ụbọchị ma e jiri ya tụnyere 9router, OpenRouter, CLIProxyAPI na LiteLLM n'ofe ikike 13. OmniRoute: ndị na-enye ọrụ 383, ọkwa efu 150+ arụnyere, atụmatụ nhazi ụzọ 20, mkpakọ token nke injin 12, sava MCP arụnyere nke nwere ngwaọrụ 110, usoro onye nnọchi anya A2A, ebe nchekwa na-adịgide adịgide, guardrails, ndị nnọchi anya igwe ojii, nzuzo akara mkpịsị aka TLS, Desktop/Termux/PWA na mpaghara UI i18n 67. OmniRoute nwere ikikere MIT ma ị nwere ike ịnabata ya n'onwe gị. Ikike na ọnụọgụ ndị asọmpi nwere ike ịgbanwe; lee usoro ndị ejikọtara. +Ihe na-eme OmniRoute ka ọ dị iche — nseta atụmatụ nke nwere ụbọchị ma e jiri ya tụnyere 9router, OpenRouter, CLIProxyAPI na LiteLLM n'ofe ikike 13. OmniRoute: ndị na-enye ọrụ 384, ọkwa efu 150+ arụnyere, atụmatụ nhazi ụzọ 20, mkpakọ token nke injin 12, sava MCP arụnyere nke nwere ngwaọrụ 110, usoro onye nnọchi anya A2A, ebe nchekwa na-adịgide adịgide, guardrails, ndị nnọchi anya igwe ojii, nzuzo akara mkpịsị aka TLS, Desktop/Termux/PWA na mpaghara UI i18n 67. OmniRoute nwere ikikere MIT ma ị nwere ike ịnabata ya n'onwe gị. Ikike na ọnụọgụ ndị asọmpi nwere ike ịgbanwe; lee usoro ndị ejikọtara. 📊 Usoro zuru ezu & nkọwa atụmatụ n'otu n'otu ma e jiri ya tụnyere 9router, OpenRouter, CLIProxyAPI & LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ A na-ahọrọ iji Radar n'onwe gị, ọ na-ejikwa naanị GET. Klayentị Omni - **🖼️ Endpoints ọhụrụ** — `/v1/ocr` (Mistral OCR) na `/v1/audio/translations` (ụdị Whisper) na-eme ka media surface zuo oke. → [Ntụaka API](docs/reference/API_REFERENCE.md) - **🎨 Ịmepụta image / video / audio** — otu API maka media: xAI Grok Imagine na Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind, na providers okwu dị ka ElevenLabs. → [Ntụaka API](docs/reference/API_REFERENCE.md) - **🌍 Nnyefe na ops** — `basePath` reverse-proxy, nchọpụta asụsụ browser na-akpaghị aka, nsuso ngwaọrụ kwa-key, MITM trust na-enweghị root, na localization zh-TW. → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 Providers na agents ndị ọzọ** — cloud agents (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) nwere browser + OAuth login, kaadị Ollama nke mbụ, Claude Opus 5 na Sonnet 5, mmekọrịta gọọmentị Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… yana catalog **383-provider** emezigharịrị. → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Providers na agents ndị ọzọ** — cloud agents (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) nwere browser + OAuth login, kaadị Ollama nke mbụ, Claude Opus 5 na Sonnet 5, mmekọrịta gọọmentị Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… yana catalog **384-provider** emezigharịrị. → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 Nghọta ntụgharị** — nzaghachi ọ bụla na-ebu header `X-OmniRoute-Decision` nke na-akpọ aha strategy/provider/latency nke nyere ya ọrụ, atụmatụ combo `cache-optimized` ọhụrụ + ihe `cacheAffinity` nke Auto-Combo na-ezighachi arịrịọ ndị a na-eme ugboro ugboro na connection nwere cached prefix, ebe endpoint `/v1/auto-combo/{channel}/candidates` nke naanị ịgụ na-egosi pool candidates dị ndụ nke channel `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Ọrụ ngwa ngwa mpaghara & akụrụngwa** — Redis mpaghara eji otu pịa, ndị na-ebubata relay Cloudflare Workers / Deno Deploy, yana Bifrost & Mux dị ka ọrụ agbakwunyere a na-elekọta. → [Ọrụ Agbakwunyere](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Ọ dịkwa n’ime ngwugwu** — usoro plugin + ahịa, usoro nka Omni/Agent/GitHub, njikọta Obsidian vault (ngwaọrụ MCP 22), API Batch & Files dakọtara na OpenAI, cache nzaghachi semantic, gamification nwere tebụl ọkwa, nchọpụta ndị nnọchi anya ACP (ndị nnọchi anya 15 arụnyere n’ime), mbupụ log ahaziri ahazi gaa BigQuery, `auto/chaos` nkesa ụdị dị iche iche n’otu oge, njikọ bot Telegram, onye njikwa ụdị n’ime ngwa, yana ọkwa ndị na-enye ọrụ n’efu LMArena-ELO. → [Akwụkwọ Nkọwa](docs/README.md) @@ -689,11 +689,11 @@ n’akụkọ ihe omume shell gị. → [CLI Integrations](docs/guides/CLI-INTEG
-## 🌐 Ndị na-eweta AI 383 — 153 akara katalọgụ n’efu +## 🌐 Ndị na-eweta AI 384 — 153 akara katalọgụ n’efu
-> **Ndị na-eweta ọrụ 383 edebanyere aha** n’ofe nchịkọta chat, media, ọchụchọ, mpaghara, cloud-agent na sistemụ nke canonical, gụnyere **153 nwere metadata nchọpụta `hasFree: true`**. Ndebanye aha ụdị chat ahụ kpuchiri **ndị na-eweta ọrụ 239 / njikọ ndị na-eweta ọrụ na ụdị 3,059 pụrụ iche / NJ ụdị raw 1,475**; katalọgụ mmefu n’efu dị iche nwere **ahịrị kwa ụdị 471**, **ọdọ ego na-emegharị 34** na **ndị na-eweta ọrụ 52 na-enye n’efu ruo mgbe ebighị ebi na-emegharị ma ọ bụ na-achọghị igodo**. Ndị a bụ denomineta dị iche iche n’ụzọ e mere ha; nkọwa na ngụkọta ndị e wepụrụ mmegharị ọdọ ego dị na [Ntụaka Ndị Na-eweta Ọrụ](docs/reference/PROVIDER_REFERENCE.md) na [Ọkwa N’efu](docs/reference/FREE_TIERS.md). +> **Ndị na-eweta ọrụ 384 edebanyere aha** n’ofe nchịkọta chat, media, ọchụchọ, mpaghara, cloud-agent na sistemụ nke canonical, gụnyere **153 nwere metadata nchọpụta `hasFree: true`**. Ndebanye aha ụdị chat ahụ kpuchiri **ndị na-eweta ọrụ 239 / njikọ ndị na-eweta ọrụ na ụdị 3,059 pụrụ iche / NJ ụdị raw 1,475**; katalọgụ mmefu n’efu dị iche nwere **ahịrị kwa ụdị 471**, **ọdọ ego na-emegharị 34** na **ndị na-eweta ọrụ 52 na-enye n’efu ruo mgbe ebighị ebi na-emegharị ma ọ bụ na-achọghị igodo**. Ndị a bụ denomineta dị iche iche n’ụzọ e mere ha; nkọwa na ngụkọta ndị e wepụrụ mmegharị ọdọ ego dị na [Ntụaka Ndị Na-eweta Ọrụ](docs/reference/PROVIDER_REFERENCE.md) na [Ọkwa N’efu](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ Metrik iwu na 2026-10-18: **vidiyo pụrụ iche 1.976** · **nlele amaara 20.22 Oge ịrụ ọrụNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 AsụsụTypeScript 6.0 — 100% TypeScript n'ofe src/ na open-sse/ (any efu na isi sistemụ kemgbe v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Ebe nchekwa databetter-sqlite3 (SQLite, ndekọ WAL) + LowDB (JSON ochie) — modul ngalaba 147, mbugharị 230 + Ebe nchekwa databetter-sqlite3 (SQLite, ndekọ WAL) + LowDB (JSON ochie) — modul ngalaba 147, mbugharị 232 Ụdị ebe nchekwaSQLite FTS5 ọchụchọ ederede zuru ezu + ntinye vector int8-quantized, mbelata nwere ụdị SchemasZod 4 — nkwado MCP tool I/O + nkwekọrịta API Usoro nkwukọrịtaMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ig/llm.txt b/docs/i18n/ig/llm.txt index b01e4632e75..da92efd70df 100644 --- a/docs/i18n/ig/llm.txt +++ b/docs/i18n/ig/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/it/README.md b/docs/i18n/it/README.md index 958dabf67c5..1eb6f0a246d 100644 --- a/docs/i18n/it/README.md +++ b/docs/i18n/it/README.md @@ -15,7 +15,7 @@ **Creato da [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Non smettere mai di programmare. Ogni strumento AI → 383 provider — 150+ gratuiti — attraverso un unico endpoint. Claude Code, Codex, Cursor, Cline, Copilot e Antigravity verso Claude / GPT / Gemini GRATUITI con fallback automatico. La compressione combinata RTK + Caveman consente di risparmiare il 15–95% dei token (~89% in media) — non raggiungere mai i limiti. 383 provider AI · oltre 150 piani gratuiti · ~1,62 miliardi di token gratuiti/mese · 20 strategie di routing · 0 $ per iniziare. +OmniRoute — Non smettere mai di programmare. Ogni strumento AI → 384 provider — 150+ gratuiti — attraverso un unico endpoint. Claude Code, Codex, Cursor, Cline, Copilot e Antigravity verso Claude / GPT / Gemini GRATUITI con fallback automatico. La compressione combinata RTK + Caveman consente di risparmiare il 15–95% dei token (~89% in media) — non raggiungere mai i limiti. 384 provider AI · oltre 150 piani gratuiti · ~1,62 miliardi di token gratuiti/mese · 20 strategie di routing · 0 $ per iniziare.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ----------------------------------------- | :------------------------: | :---------: | :-------------------: | -| 🌐 Provider | 352 | **358** | 383 | +| 🌐 Provider | 352 | **358** | 384 | | 🧠 ID univoci dei modelli di chat | 1320 | **1374** | 1491 | | 🖼️ Bridge delle modalità | 🆕 visione + audio + video | ✓ | ✓ | | 📡 Catalogo gratuito Radar | 🆕 adesione volontaria | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funzionalità 🎯 Combo - 🌐 Provider + 🌐 Provider 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -La promessa — Un endpoint e 383 provider. Il fallback automatico mantiene l'instradamento attivo finché è disponibile un'altra destinazione operativa. Sei pilastri: fallback resiliente tra 383 provider · fino al 95% di risparmio sui token per i carichi di lavoro idonei · $0 per iniziare con oltre 150 piani gratuiti e 52 provider gratuiti ricorrenti/senza chiave per sempre · 39 integrazioni CLI/agent tramite un'unica configurazione · compatibilità con le API OpenAI, Claude, Gemini e Responses su /v1 · controlli per la produzione, inclusi circuit breaker, TLS stealth, MCP con 110 strumenti, A2A, memoria, guardrail, valutazioni e oltre 56.000 dichiarazioni di test statiche in oltre 7.800 file di test monitorati. +La promessa — Un endpoint e 384 provider. Il fallback automatico mantiene l'instradamento attivo finché è disponibile un'altra destinazione operativa. Sei pilastri: fallback resiliente tra 384 provider · fino al 95% di risparmio sui token per i carichi di lavoro idonei · $0 per iniziare con oltre 150 piani gratuiti e 52 provider gratuiti ricorrenti/senza chiave per sempre · 39 integrazioni CLI/agent tramite un'unica configurazione · compatibilità con le API OpenAI, Claude, Gemini e Responses su /v1 · controlli per la produzione, inclusi circuit breaker, TLS stealth, MCP con 110 strumenti, A2A, memoria, guardrail, valutazioni e oltre 56.000 dichiarazioni di test statiche in oltre 7.800 file di test monitorati.

@@ -510,7 +510,7 @@ Tutte e **20** le strategie — combinabili liberamente per ogni passaggio della -Cosa distingue OmniRoute — un'istantanea datata delle funzionalità rispetto a 9router, OpenRouter, CLIProxyAPI e LiteLLM per 13 capacità. OmniRoute: 383 provider, oltre 150 livelli gratuiti integrati, 20 strategie di routing, compressione dei token tramite 12 motori, server MCP integrato con 110 strumenti, protocollo per agenti A2A, memoria persistente, guardrail, agenti cloud, stealth dell'impronta TLS, Desktop/Termux/PWA e 67 locali dell'interfaccia utente per l'internazionalizzazione. OmniRoute è distribuito con licenza MIT e può essere ospitato autonomamente. Le funzionalità e i conteggi dei concorrenti possono cambiare; consulta la metodologia collegata. +Cosa distingue OmniRoute — un'istantanea datata delle funzionalità rispetto a 9router, OpenRouter, CLIProxyAPI e LiteLLM per 13 capacità. OmniRoute: 384 provider, oltre 150 livelli gratuiti integrati, 20 strategie di routing, compressione dei token tramite 12 motori, server MCP integrato con 110 strumenti, protocollo per agenti A2A, memoria persistente, guardrail, agenti cloud, stealth dell'impronta TLS, Desktop/Termux/PWA e 67 locali dell'interfaccia utente per l'internazionalizzazione. OmniRoute è distribuito con licenza MIT e può essere ospitato autonomamente. Le funzionalità e i conteggi dei concorrenti possono cambiare; consulta la metodologia collegata. 📊 Metodologia completa e dettagli per funzionalità rispetto a 9router, OpenRouter, CLIProxyAPI e LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -614,7 +614,7 @@ il catalogo attuale su **[radar.omniroute.online/planos](https://radar.omniroute - **🖼️ Nuovi endpoint** — `/v1/ocr` (Mistral OCR) e `/v1/audio/translations` (in stile Whisper) completano la superficie multimediale. → [Riferimento API](docs/reference/API_REFERENCE.md) - **🎨 Generazione di immagini / video / audio** — un'unica API per i contenuti multimediali: xAI Grok Imagine e video Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind e provider vocali come ElevenLabs. → [Riferimento API](docs/reference/API_REFERENCE.md) - **🌍 Deployment e operazioni** — `basePath` per il reverse proxy, rilevamento automatico della lingua del browser, tracciamento dei dispositivi per chiave, trust MITM senza privilegi root, localizzazione zh-TW. → [Ambiente](docs/reference/ENVIRONMENT.md) -- **🤝 Più provider e agenti** — agenti cloud (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) con browser + accesso OAuth, card nativa per Ollama, Claude Opus 5 e Sonnet 5, partnership ufficiale con Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… e un catalogo aggiornato di **383 provider**. → [Provider](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Più provider e agenti** — agenti cloud (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) con browser + accesso OAuth, card nativa per Ollama, Claude Opus 5 e Sonnet 5, partnership ufficiale con Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… e un catalogo aggiornato di **384 provider**. → [Provider](docs/reference/PROVIDER_REFERENCE.md) - **📡 Trasparenza del routing** — ogni risposta include un header `X-OmniRoute-Decision` che indica la strategia/provider/latenza che l'ha gestita, una nuova strategia combo `cache-optimized` + il fattore `cacheAffinity` di Auto-Combo reindirizza le richieste ripetute alla connessione che contiene il prefisso nella cache e un endpoint in sola lettura `/v1/auto-combo/{channel}/candidates` espone il pool di candidati attivo di un canale `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Prestazioni e infrastruttura locali** — Redis locale con un clic, strumenti di distribuzione relay per Cloudflare Workers / Deno Deploy, Bifrost e Mux come servizi incorporati supervisionati. → [Servizi incorporati](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Incluso anche** — framework per plugin e marketplace, framework di competenze Omni/Agent/GitHub, integrazione con i vault di Obsidian (22 strumenti MCP), API Batch e Files compatibili con OpenAI, cache semantica delle risposte, gamification con classifiche, rilevamento degli agenti ACP (15 agenti integrati), esportazione pianificata dei log in BigQuery, fan-out parallelo multi-modello `auto/chaos`, un bridge per bot Telegram, un gestore delle versioni integrato nell'app e classifiche dei provider gratuiti LMArena-ELO. → [Documentazione](docs/README.md) @@ -698,11 +698,11 @@ dalla cronologia della tua shell. → [Integrazioni CLI](docs/guides/CLI-INTEGRA
-## 🌐 383 provider AI — 153 contrassegnati come gratuiti nel catalogo +## 🌐 384 provider AI — 153 contrassegnati come gratuiti nel catalogo
-> **383 provider registrati** nelle raccolte canoniche di chat, media, ricerca, locali, agenti cloud e sistema, inclusi **153 con metadati di rilevamento `hasFree: true`**. Il registro dei modelli di chat comprende **239 provider / 3.059 coppie provider-modello distinte / 1.475 ID modello grezzi**; il catalogo separato dei budget gratuiti conta **471 righe per modello**, **34 pool ricorrenti** e **52 provider gratuiti ricorrenti/senza chiave per sempre**. Questi denominatori sono diversi per progettazione; le definizioni e i calcoli deduplicati per pool sono disponibili in [Riferimento provider](docs/reference/PROVIDER_REFERENCE.md) e [Livelli gratuiti](docs/reference/FREE_TIERS.md). +> **384 provider registrati** nelle raccolte canoniche di chat, media, ricerca, locali, agenti cloud e sistema, inclusi **153 con metadati di rilevamento `hasFree: true`**. Il registro dei modelli di chat comprende **239 provider / 3.059 coppie provider-modello distinte / 1.475 ID modello grezzi**; il catalogo separato dei budget gratuiti conta **471 righe per modello**, **34 pool ricorrenti** e **52 provider gratuiti ricorrenti/senza chiave per sempre**. Questi denominatori sono diversi per progettazione; le definizioni e i calcoli deduplicati per pool sono disponibili in [Riferimento provider](docs/reference/PROVIDER_REFERENCE.md) e [Livelli gratuiti](docs/reference/FREE_TIERS.md).
@@ -1306,7 +1306,7 @@ Metriche canoniche al 2026-10-18: **1.976 video unici** · **20.220.289 visualiz RuntimeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 LinguaggioTypeScript 6.0 — 100% TypeScript in src/ e open-sse/ (zero any nel core dalla v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Databasebetter-sqlite3 (SQLite, journaling WAL) + LowDB (JSON legacy) — 147 moduli di dominio, 230 migrazioni + Databasebetter-sqlite3 (SQLite, journaling WAL) + LowDB (JSON legacy) — 147 moduli di dominio, 232 migrazioni MemoriaRicerca full-text SQLite FTS5 + embedding vettoriali quantizzati int8, decadimento tipizzato SchemiZod 4 — validazione I/O degli strumenti MCP + contratti API ProtocolliMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/it/llm.txt b/docs/i18n/it/llm.txt index 2eddd0b8569..3a66828a904 100644 --- a/docs/i18n/it/llm.txt +++ b/docs/i18n/it/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ja/README.md b/docs/i18n/ja/README.md index e762ac8fb62..57c11202961 100644 --- a/docs/i18n/ja/README.md +++ b/docs/i18n/ja/README.md @@ -15,7 +15,7 @@ **[Diego Souza](https://github.com/diegosouzapw) によって作成** -OmniRoute — コーディングを止めない。すべての AI ツールを、1 つのエンドポイント経由で 383 のプロバイダー(無料のものは 150 以上)へ。Claude Code、Codex、Cursor、Cline、Copilot、Antigravity を、オートフォールバック対応の無料の Claude / GPT / Gemini へ接続。RTK + Caveman の積層圧縮により、トークンを 15~95% 節約(平均約 89%)し、制限に達することはありません。AI プロバイダー 383 社 · 無料ティア 150 以上 · 月間無料トークン約 16.2 億 · ルーティング戦略 20 種類 · $0 で開始可能。 +OmniRoute — コーディングを止めない。すべての AI ツールを、1 つのエンドポイント経由で 384 のプロバイダー(無料のものは 150 以上)へ。Claude Code、Codex、Cursor、Cline、Copilot、Antigravity を、オートフォールバック対応の無料の Claude / GPT / Gemini へ接続。RTK + Caveman の積層圧縮により、トークンを 15~95% 節約(平均約 89%)し、制限に達することはありません。AI プロバイダー 384 社 · 無料ティア 150 以上 · 月間無料トークン約 16.2 億 · ルーティング戦略 20 種類 · $0 で開始可能。
@@ -73,7 +73,7 @@ OMNIROUTEが費用の節約や作業の効率化に役立ったなら、リポ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------- | :-------------------------------: | :---------: | :-------------------: | -| 🌐 プロバイダー | 352 | **358** | 383 | +| 🌐 プロバイダー | 352 | **358** | 384 | | 🧠 一意のチャットモデル ID | 1320 | **1374** | 1491 | | 🖼️ モダリティブリッジ | 🆕 ビジョン + オーディオ + ビデオ | ✓ | ✓ | | 📡 Radar 無料カタログ | 🆕 オプトイン | ✓ | ✓ | @@ -114,7 +114,7 @@ OMNIROUTEが費用の節約や作業の効率化に役立ったなら、リポ ⚙️ 機能 🎯 コンボ - 🌐 プロバイダー + 🌐 プロバイダー 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRouteが約束するもの — 1つのエンドポイントと383のプロバイダー。利用可能で健全な別の対象がある限り、自動フォールバックによってルーティングを継続します。6つの柱:383のプロバイダーにまたがる回復力の高いフォールバック · 対象となるワークロードで最大95%のトークン削減 · 150以上の無料ティアと、継続利用可能またはキー不要で永久無料の52プロバイダーにより、0ドルから開始 · 1つの設定で39のCLI/エージェント連携 · /v1でのOpenAI、Claude、Gemini、Responses API互換性 · サーキットブレーカー、TLSステルス、MCP 110ツール、A2A、メモリ、ガードレール、評価、7,800以上の追跡テストファイルにわたる56,000以上の静的テスト宣言など、本番環境向けの制御機能。 +OmniRouteが約束するもの — 1つのエンドポイントと384のプロバイダー。利用可能で健全な別の対象がある限り、自動フォールバックによってルーティングを継続します。6つの柱:384のプロバイダーにまたがる回復力の高いフォールバック · 対象となるワークロードで最大95%のトークン削減 · 150以上の無料ティアと、継続利用可能またはキー不要で永久無料の52プロバイダーにより、0ドルから開始 · 1つの設定で39のCLI/エージェント連携 · /v1でのOpenAI、Claude、Gemini、Responses API互換性 · サーキットブレーカー、TLSステルス、MCP 110ツール、A2A、メモリ、ガードレール、評価、7,800以上の追跡テストファイルにわたる56,000以上の静的テスト宣言など、本番環境向けの制御機能。

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRouteを際立たせる特徴 — 9router、OpenRouter、CLIProxyAPI、LiteLLMと13の機能を比較した日付入りのスナップショット。OmniRouteには、383のプロバイダー、組み込みの150以上の無料ティア、20のルーティング戦略、12エンジンによるトークン圧縮、110ツールを備えた組み込みMCPサーバー、A2Aエージェントプロトコル、永続メモリ、ガードレール、クラウドエージェント、TLSフィンガープリントによるステルス、Desktop/Termux/PWA、67のi18n UIロケールが含まれます。OmniRouteはMITライセンスで提供され、セルフホストできます。競合製品の機能や数は変更される場合があります。詳細はリンク先の方法論をご覧ください。 +OmniRouteを際立たせる特徴 — 9router、OpenRouter、CLIProxyAPI、LiteLLMと13の機能を比較した日付入りのスナップショット。OmniRouteには、384のプロバイダー、組み込みの150以上の無料ティア、20のルーティング戦略、12エンジンによるトークン圧縮、110ツールを備えた組み込みMCPサーバー、A2Aエージェントプロトコル、永続メモリ、ガードレール、クラウドエージェント、TLSフィンガープリントによるステルス、Desktop/Termux/PWA、67のi18n UIロケールが含まれます。OmniRouteはMITライセンスで提供され、セルフホストできます。競合製品の機能や数は変更される場合があります。詳細はリンク先の方法論をご覧ください。 📊 9router、OpenRouter、CLIProxyAPI、LiteLLMとの完全な方法論と機能別詳細 → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radarはオプトイン方式で、GETリクエストのみを使用します。 - **🖼️ 新しいエンドポイント** — `/v1/ocr`(Mistral OCR)と `/v1/audio/translations`(Whisper形式)を追加し、メディアAPIを拡充しました。→ [APIリファレンス](docs/reference/API_REFERENCE.md) - **🎨 画像 / 動画 / 音声生成** — メディア向けの単一APIで、xAI Grok ImagineとNovita AIの動画、ComfyUI、Magnific、Adobe Firefly、Segmind、ElevenLabsなどの音声プロバイダーを利用できます。→ [APIリファレンス](docs/reference/API_REFERENCE.md) - **🌍 デプロイと運用** — リバースプロキシの `basePath`、ブラウザー言語の自動検出、キーごとのデバイス追跡、root権限不要のMITM信頼、zh-TWローカライゼーションに対応しました。→ [環境変数](docs/reference/ENVIRONMENT.md) -- **🤝 さらに多くのプロバイダーとエージェント** — クラウドエージェント(Codex Cloud、Cursor、Devin、Jules)、ブラウザー + OAuthログイン対応のGrok Build(xAI)、正式対応のOllamaカード、Claude Opus 5とSonnet 5、Kimi公式パートナーシップ(Code/Web/Moonshot)、Zed、Requesty、SenseNova、Yuanbao、Agnes AIなどに対応し、**383プロバイダーのカタログ**を刷新しました。→ [プロバイダー](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 さらに多くのプロバイダーとエージェント** — クラウドエージェント(Codex Cloud、Cursor、Devin、Jules)、ブラウザー + OAuthログイン対応のGrok Build(xAI)、正式対応のOllamaカード、Claude Opus 5とSonnet 5、Kimi公式パートナーシップ(Code/Web/Moonshot)、Zed、Requesty、SenseNova、Yuanbao、Agnes AIなどに対応し、**384プロバイダーのカタログ**を刷新しました。→ [プロバイダー](docs/reference/PROVIDER_REFERENCE.md) - **📡 ルーティングの透明性** — すべてのレスポンスに、処理した戦略/プロバイダー/レイテンシーを示す `X-OmniRoute-Decision` ヘッダーが付与されます。新しい `cache-optimized` コンボ戦略とAuto-Comboの `cacheAffinity` 要素により、繰り返しリクエストをキャッシュ済みプレフィックスを保持する接続へ戻します。また、読み取り専用の `/v1/auto-combo/{channel}/candidates` エンドポイントで、`auto/*` チャネルの現在の候補プールを公開します。→ [Auto-Combo](docs/routing/AUTO-COMBO.md)。 - **⚡ ローカルパフォーマンスとインフラ** — ワンクリックで起動できるローカル Redis、Cloudflare Workers / Deno Deploy リレーデプロイヤー、監視対象の組み込みサービスとしての Bifrost と Mux。→ [組み込みサービス](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 これらも標準搭載** — プラグインフレームワーク + マーケットプレイス、Omni/Agent/GitHub スキルフレームワーク、Obsidian vault 統合(22 個の MCP ツール)、OpenAI 互換の Batch & Files API、セマンティックレスポンスキャッシュ、リーダーボード付きのゲーミフィケーション、ACP エージェント検出(15 個の組み込みエージェント)、BigQuery へのスケジュール済みログエクスポート、`auto/chaos` による並列マルチモデルファンアウト、Telegram ボットブリッジ、アプリ内バージョンマネージャー、LMArena-ELO による無料プロバイダーランキング。→ [ドキュメント](docs/README.md) @@ -689,11 +689,11 @@ omniroute configure codex # 他にも: claude opencode qwen aider goose
-## 🌐 383 AI プロバイダー — カタログ登録済み無料 153 +## 🌐 384 AI プロバイダー — カタログ登録済み無料 153
-> **383 の登録済みプロバイダー**が、正規のチャット、メディア、検索、ローカル、クラウドエージェント、システムの各コレクションにまたがって存在し、そのうち **153 は `hasFree: true` の検出メタデータを付与**されています。チャットモデルレジストリには **239 プロバイダー / 3,059 個の個別プロバイダー・モデルペア / 1,475 個の生モデル ID**が含まれます。別個に管理される無料予算カタログには、**モデルごとの行が 471 件**、**継続的なプールが 34 件**、**継続利用可能・キー不要で永久無料のプロバイダーが 52 件**あります。これらは意図的に異なる分母を使用しています。定義およびプール重複排除後の計算については、[プロバイダーリファレンス](docs/reference/PROVIDER_REFERENCE.md)と[無料ティア](docs/reference/FREE_TIERS.md)をご覧ください。 +> **384 の登録済みプロバイダー**が、正規のチャット、メディア、検索、ローカル、クラウドエージェント、システムの各コレクションにまたがって存在し、そのうち **153 は `hasFree: true` の検出メタデータを付与**されています。チャットモデルレジストリには **239 プロバイダー / 3,059 個の個別プロバイダー・モデルペア / 1,475 個の生モデル ID**が含まれます。別個に管理される無料予算カタログには、**モデルごとの行が 471 件**、**継続的なプールが 34 件**、**継続利用可能・キー不要で永久無料のプロバイダーが 52 件**あります。これらは意図的に異なる分母を使用しています。定義およびプール重複排除後の計算については、[プロバイダーリファレンス](docs/reference/PROVIDER_REFERENCE.md)と[無料ティア](docs/reference/FREE_TIERS.md)をご覧ください。
@@ -1295,7 +1295,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 でも省略さ ランタイムNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 言語TypeScript 6.0 — src/ と open-sse/ 全体で 100% TypeScript(v2.0 以降のコアに any はゼロ) フレームワークNext.js 16 + React 19 + Tailwind CSS 4 - データベースbetter-sqlite3(SQLite、WAL ジャーナリング)+ LowDB(JSON レガシー)— 147 のドメインモジュール、230 のマイグレーション + データベースbetter-sqlite3(SQLite、WAL ジャーナリング)+ LowDB(JSON レガシー)— 147 のドメインモジュール、232 のマイグレーション メモリSQLite FTS5 全文検索 + int8 量子化ベクトル埋め込み、型付き減衰 スキーマZod 4 — MCP ツールの I/O 検証 + API コントラクト プロトコルMCP(stdio / HTTP / SSE)+ A2A v0.3(JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ja/llm.txt b/docs/i18n/ja/llm.txt index c6508c6e23e..bac705d91dc 100644 --- a/docs/i18n/ja/llm.txt +++ b/docs/i18n/ja/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ka/README.md b/docs/i18n/ka/README.md index a9cfb4b099b..f2c550a402b 100644 --- a/docs/i18n/ka/README.md +++ b/docs/i18n/ka/README.md @@ -15,7 +15,7 @@ **შექმნა [Diego Souza](https://github.com/diegosouzapw)-მ** -OmniRoute — არასოდეს შეწყვიტო კოდირება. ყველა AI ინსტრუმენტი → 383 პროვაიდერი — 150-ზე მეტი უფასო — ერთი endpoint-ის გავლით. Claude Code, Codex, Cursor, Cline, Copilot და Antigravity — უფასო Claude / GPT / Gemini-ზე ავტომატური fallback-ით. RTK + Caveman-ის კომბინირებული შეკუმშვა ზოგავს ტოკენების 15–95%-ს (საშუალოდ ~89%) — ლიმიტებს არასოდეს გადააჭარბო. 383 AI პროვაიდერი · 150-ზე მეტი უფასო ტარიფი · ~1.62B უფასო ტოკენი/თვეში · მარშრუტიზაციის 20 სტრატეგია · დასაწყებად $0. +OmniRoute — არასოდეს შეწყვიტო კოდირება. ყველა AI ინსტრუმენტი → 384 პროვაიდერი — 150-ზე მეტი უფასო — ერთი endpoint-ის გავლით. Claude Code, Codex, Cursor, Cline, Copilot და Antigravity — უფასო Claude / GPT / Gemini-ზე ავტომატური fallback-ით. RTK + Caveman-ის კომბინირებული შეკუმშვა ზოგავს ტოკენების 15–95%-ს (საშუალოდ ~89%) — ლიმიტებს არასოდეს გადააჭარბო. 384 AI პროვაიდერი · 150-ზე მეტი უფასო ტარიფი · ~1.62B უფასო ტოკენი/თვეში · მარშრუტიზაციის 20 სტრატეგია · დასაწყებად $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ----------------------------------- | :----------------------: | :---------: | :-------------------: | -| 🌐 პროვაიდერები | 352 | **358** | 383 | +| 🌐 პროვაიდერები | 352 | **358** | 384 | | 🧠 უნიკალური ჩატის მოდელების ID-ები | 1320 | **1374** | 1491 | | 🖼️ მოდალობების ხიდი | 🆕 ხედვა + აუდიო + ვიდეო | ✓ | ✓ | | 📡 Radar-ის უფასო კატალოგი | 🆕 ჩართვის არჩევანი | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ ფუნქციები 🎯 კომბინაციები - 🌐 პროვაიდერები + 🌐 პროვაიდერები 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -დაპირება — ერთი endpoint და 383 პროვაიდერი. ავტომატური fallback ინარჩუნებს მარშრუტიზაციას, სანამ სხვა ჯანსაღი სამიზნე ხელმისაწვდომია. ექვსი საყრდენი: მდგრადი fallback 383 პროვაიდერს შორის · ტოკენების 95%-მდე დაზოგვა შესაბამის დატვირთვებზე · დასაწყებად $0, 150-ზე მეტი უფასო ტარიფითა და 52 განმეორებადი/გასაღების გარეშე, სამუდამოდ უფასო პროვაიდერით · 39 CLI/აგენტის ინტეგრაცია ერთი კონფიგურაციით · OpenAI-ს, Claude-ის, Gemini-სა და Responses API-ს თავსებადობა /v1-ზე · საწარმოო კონტროლები, მათ შორის circuit breaker-ები, TLS stealth, MCP-ის 110 ხელსაწყო, A2A, მეხსიერება, guardrail-ები, eval-ები და 56,000-ზე მეტი სტატიკური ტესტის დეკლარაცია 7,800-ზე მეტ თვალთვალის ქვეშ მყოფ სატესტო ფაილში. +დაპირება — ერთი endpoint და 384 პროვაიდერი. ავტომატური fallback ინარჩუნებს მარშრუტიზაციას, სანამ სხვა ჯანსაღი სამიზნე ხელმისაწვდომია. ექვსი საყრდენი: მდგრადი fallback 384 პროვაიდერს შორის · ტოკენების 95%-მდე დაზოგვა შესაბამის დატვირთვებზე · დასაწყებად $0, 150-ზე მეტი უფასო ტარიფითა და 52 განმეორებადი/გასაღების გარეშე, სამუდამოდ უფასო პროვაიდერით · 39 CLI/აგენტის ინტეგრაცია ერთი კონფიგურაციით · OpenAI-ს, Claude-ის, Gemini-სა და Responses API-ს თავსებადობა /v1-ზე · საწარმოო კონტროლები, მათ შორის circuit breaker-ები, TLS stealth, MCP-ის 110 ხელსაწყო, A2A, მეხსიერება, guardrail-ები, eval-ები და 56,000-ზე მეტი სტატიკური ტესტის დეკლარაცია 7,800-ზე მეტ თვალთვალის ქვეშ მყოფ სატესტო ფაილში.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -რა განასხვავებს OmniRoute-ს — ფუნქციების მოძველებული სურათი 9router-თან, OpenRouter-თან, CLIProxyAPI-სა და LiteLLM-თან შედარებით 13 შესაძლებლობის მიხედვით. OmniRoute: 383 პროვაიდერი, 150+-ზე მეტი უფასო ტარიფი ჩაშენებული, მარშრუტიზაციის 20 სტრატეგია, ტოკენების შეკუმშვის 12 ძრავა, ჩაშენებული MCP სერვერი 110 ხელსაწყოთი, A2A აგენტების პროტოკოლი, მუდმივი მეხსიერება, დამცავი შეზღუდვები, ღრუბლოვანი აგენტები, TLS fingerprint-ის ფარული რეჟიმი, Desktop/Termux/PWA და i18n UI-ის 67 ლოკალი. OmniRoute ლიცენზირებულია MIT-ით და შესაძლებელია მისი თვითჰოსტინგი. კონკურენტების შესაძლებლობები და რაოდენობები შეიძლება შეიცვალოს; იხილეთ ბმულზე მოცემული მეთოდოლოგია. +რა განასხვავებს OmniRoute-ს — ფუნქციების მოძველებული სურათი 9router-თან, OpenRouter-თან, CLIProxyAPI-სა და LiteLLM-თან შედარებით 13 შესაძლებლობის მიხედვით. OmniRoute: 384 პროვაიდერი, 150+-ზე მეტი უფასო ტარიფი ჩაშენებული, მარშრუტიზაციის 20 სტრატეგია, ტოკენების შეკუმშვის 12 ძრავა, ჩაშენებული MCP სერვერი 110 ხელსაწყოთი, A2A აგენტების პროტოკოლი, მუდმივი მეხსიერება, დამცავი შეზღუდვები, ღრუბლოვანი აგენტები, TLS fingerprint-ის ფარული რეჟიმი, Desktop/Termux/PWA და i18n UI-ის 67 ლოკალი. OmniRoute ლიცენზირებულია MIT-ით და შესაძლებელია მისი თვითჰოსტინგი. კონკურენტების შესაძლებლობები და რაოდენობები შეიძლება შეიცვალოს; იხილეთ ბმულზე მოცემული მეთოდოლოგია. 📊 სრული მეთოდოლოგია და თითოეული ფუნქციის დეტალები 9router-თან, OpenRouter-თან, CLIProxyAPI-სა და LiteLLM-თან შედარებით → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar-ში მონაწილეობა ნებაყოფლობი - **🖼️ ახალი endpoint-ები** — `/v1/ocr` (Mistral OCR) და `/v1/audio/translations` (Whisper-ის სტილის) ავსებს მედიის შესაძლებლობებს. → [API-ის ცნობარი](docs/reference/API_REFERENCE.md) - **🎨 სურათების / ვიდეოს / აუდიოს გენერაცია** — მედიისთვის ერთი API: xAI Grok Imagine და Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind და მეტყველების პროვაიდერები, მათ შორის ElevenLabs. → [API-ის ცნობარი](docs/reference/API_REFERENCE.md) - **🌍 განთავსება და ოპერაციები** — reverse-proxy `basePath`, ბრაუზერის ენის ავტომატური აღმოჩენა, თითოეული გასაღების მოწყობილობის თვალყურის დევნება, root-ის გარეშე MITM ნდობა და zh-TW ლოკალიზაცია. → [გარემო](docs/reference/ENVIRONMENT.md) -- **🤝 მეტი პროვაიდერი და აგენტი** — ღრუბლოვანი აგენტები (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) ბრაუზერითა და OAuth-ით შესვლით, Ollama-ს პირველი კლასის ბარათი, Claude Opus 5 და Sonnet 5, Kimi-ს ოფიციალური პარტნიორობა (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… და განახლებული **383-პროვაიდერიანი კატალოგი**. → [პროვაიდერები](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 მეტი პროვაიდერი და აგენტი** — ღრუბლოვანი აგენტები (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) ბრაუზერითა და OAuth-ით შესვლით, Ollama-ს პირველი კლასის ბარათი, Claude Opus 5 და Sonnet 5, Kimi-ს ოფიციალური პარტნიორობა (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… და განახლებული **384-პროვაიდერიანი კატალოგი**. → [პროვაიდერები](docs/reference/PROVIDER_REFERENCE.md) - **📡 მარშრუტიზაციის გამჭვირვალობა** — თითოეული პასუხი შეიცავს `X-OmniRoute-Decision` ჰედერს, რომელშიც მითითებულია მისი მომსახურე სტრატეგია/პროვაიდერი/დაყოვნება; ახალი `cache-optimized` კომბინირებული სტრატეგია და Auto-Combo-ის `cacheAffinity` ფაქტორი განმეორებით მოთხოვნებს აბრუნებს იმ კავშირზე, რომელსაც ქეშირებული პრეფიქსი აქვს; მხოლოდ წაკითხვისთვის განკუთვნილი `/v1/auto-combo/{channel}/candidates` endpoint-ი კი ავლენს `auto/*` არხის კანდიდატთა ცოცხალ ნაკრებს. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ ლოკალური წარმადობა და ინფრასტრუქტურა** — ლოკალური Redis-ის გაშვება ერთი დაწკაპუნებით, Cloudflare Workers / Deno Deploy relay-ის დიპლოიერები, Bifrost და Mux ზედამხედველობის ქვეშ გაშვებული ჩაშენებული სერვისების სახით. → [ჩაშენებული სერვისები](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 ასევე მოყვება** — პლაგინების ფრეიმვორკი და მარკეტფლეისი, Omni/Agent/GitHub-ის უნარების ფრეიმვორკები, Obsidian vault-ის ინტეგრაცია (22 MCP ხელსაწყო), OpenAI-თან თავსებადი Batch და Files API-ები, სემანტიკური პასუხების ქეში, გეიმიფიკაცია ლიდერბორდებით, ACP აგენტების აღმოჩენა (15 ჩაშენებული აგენტი), ჟურნალების დაგეგმილი ექსპორტი BigQuery-ში, `auto/chaos` პარალელური მრავალმოდელიანი fan-out, Telegram ბოტის ხიდი, აპში ჩაშენებული ვერსიების მენეჯერი და LMArena-ELO-ს უფასო პროვაიდერების რეიტინგები. → [დოკუმენტაცია](docs/README.md) @@ -688,11 +688,11 @@ omniroute configure codex # ასევე: claude opencode qwen aider
-## 🌐 383 AI პროვაიდერი — 153 კატალოგში უფასოდ მონიშნული +## 🌐 384 AI პროვაიდერი — 153 კატალოგში უფასოდ მონიშნული
-> **383 რეგისტრირებული პროვაიდერი** კანონიკური ჩატის, მედიის, ძიების, ლოკალური, cloud-agent და სისტემის კოლექციებში, მათ შორის **153 პროვაიდერი, რომელსაც აღმოჩენის მეტამონაცემებში `hasFree: true` აქვს მითითებული**. ჩატის მოდელების რეესტრი მოიცავს **239 პროვაიდერს / 3,059 უნიკალურ provider-model წყვილს / 1,475 დაუმუშავებელ model ID-ს**; ცალკე უფასო ბიუჯეტის კატალოგში არის **471 ჩანაწერი თითოეულ მოდელზე**, **34 განმეორებადი პული** და **52 განმეორებადი/გასაღების გარეშე სამუდამოდ უფასო პროვაიდერი**. ეს სხვადასხვა მნიშვნელი განზრახ არის გამოყენებული; განსაზღვრებები და პულების დუბლიკატების მოცილებით გამოთვლები მოცემულია [პროვაიდერების ცნობარში](docs/reference/PROVIDER_REFERENCE.md) და [უფასო ტარიფებში](docs/reference/FREE_TIERS.md). +> **384 რეგისტრირებული პროვაიდერი** კანონიკური ჩატის, მედიის, ძიების, ლოკალური, cloud-agent და სისტემის კოლექციებში, მათ შორის **153 პროვაიდერი, რომელსაც აღმოჩენის მეტამონაცემებში `hasFree: true` აქვს მითითებული**. ჩატის მოდელების რეესტრი მოიცავს **239 პროვაიდერს / 3,059 უნიკალურ provider-model წყვილს / 1,475 დაუმუშავებელ model ID-ს**; ცალკე უფასო ბიუჯეტის კატალოგში არის **471 ჩანაწერი თითოეულ მოდელზე**, **34 განმეორებადი პული** და **52 განმეორებადი/გასაღების გარეშე სამუდამოდ უფასო პროვაიდერი**. ეს სხვადასხვა მნიშვნელი განზრახ არის გამოყენებული; განსაზღვრებები და პულების დუბლიკატების მოცილებით გამოთვლები მოცემულია [პროვაიდერების ცნობარში](docs/reference/PROVIDER_REFERENCE.md) და [უფასო ტარიფებში](docs/reference/FREE_TIERS.md).
@@ -1294,7 +1294,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1-იც გამ გაშვების გარემოNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ენაTypeScript 6.0 — 100% TypeScript src/-სა და open-sse/-ში (any-ის ნულოვანი გამოყენება core-ში v2.0-დან) ფრეიმვორკიNext.js 16 + React 19 + Tailwind CSS 4 - მონაცემთა ბაზაbetter-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — 147 დომენური მოდული, 230 მიგრაცია + მონაცემთა ბაზაbetter-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — 147 დომენური მოდული, 232 მიგრაცია მეხსიერებაSQLite FTS5 სრული ტექსტის ძიება + int8-quantized ვექტორული ემბედინგები, ტიპიზებული decay სქემებიZod 4 — MCP ხელსაწყოების I/O ვალიდაცია + API კონტრაქტები პროტოკოლებიMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ka/llm.txt b/docs/i18n/ka/llm.txt index ea6cdd9975b..2d3be6537a8 100644 --- a/docs/i18n/ka/llm.txt +++ b/docs/i18n/ka/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/km/README.md b/docs/i18n/km/README.md index 24751dfd551..a9bc9bc5344 100644 --- a/docs/i18n/km/README.md +++ b/docs/i18n/km/README.md @@ -15,7 +15,7 @@ **បង្កើតដោយ [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — កុំឈប់សរសេរកូដ។ ឧបករណ៍ AI ទាំងអស់ → អ្នកផ្តល់សេវា 383 — ឥតគិតថ្លៃជាង 150 — តាមរយៈ endpoint តែមួយ។ Claude Code, Codex, Cursor, Cline, Copilot និង Antigravity ទៅកាន់ Claude / GPT / Gemini ឥតគិតថ្លៃ ជាមួយការប្ដូរបម្រុងស្វ័យប្រវត្តិ។ ការបង្ហាប់ជាជង់ RTK + Caveman សន្សំ tokens បាន 15–95% (ជាមធ្យម ~89%) — មិនជួបដែនកំណត់ឡើយ។ អ្នកផ្តល់សេវា AI 383 · កម្រិតឥតគិតថ្លៃជាង 150 · tokens ឥតគិតថ្លៃប្រហែល 1.62B ក្នុងមួយខែ · យុទ្ធសាស្ត្រកំណត់ផ្លូវ 20 · ចាប់ផ្តើមដោយ $0។ +OmniRoute — កុំឈប់សរសេរកូដ។ ឧបករណ៍ AI ទាំងអស់ → អ្នកផ្តល់សេវា 384 — ឥតគិតថ្លៃជាង 150 — តាមរយៈ endpoint តែមួយ។ Claude Code, Codex, Cursor, Cline, Copilot និង Antigravity ទៅកាន់ Claude / GPT / Gemini ឥតគិតថ្លៃ ជាមួយការប្ដូរបម្រុងស្វ័យប្រវត្តិ។ ការបង្ហាប់ជាជង់ RTK + Caveman សន្សំ tokens បាន 15–95% (ជាមធ្យម ~89%) — មិនជួបដែនកំណត់ឡើយ។ អ្នកផ្តល់សេវា AI 384 · កម្រិតឥតគិតថ្លៃជាង 150 · tokens ឥតគិតថ្លៃប្រហែល 1.62B ក្នុងមួយខែ · យុទ្ធសាស្ត្រកំណត់ផ្លូវ 20 · ចាប់ផ្តើមដោយ $0។
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | -------------------------------- | :--------------------------: | :---------: | :-------------------: | -| 🌐 អ្នកផ្តល់សេវា | 352 | **358** | 383 | +| 🌐 អ្នកផ្តល់សេវា | 352 | **358** | 384 | | 🧠 លេខសម្គាល់ម៉ូដែលជជែកតែមួយគត់ | 1320 | **1374** | 1491 | | 🖼️ ស្ពានម៉ូឌាល់លីតេ | 🆕 រូបភាព + អូឌីយ៉ូ + វីដេអូ | ✓ | ✓ | | 📡 កាតាឡុកឥតគិតថ្លៃរបស់ Radar | 🆕 ជ្រើសរើសចូលរួម | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ មុខងារ 🎯 បន្សំ - 🌐 អ្នកផ្ដល់សេវា + 🌐 អ្នកផ្ដល់សេវា 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -ការសន្យា — endpoint តែមួយ និងអ្នកផ្តល់សេវា 383។ ការបម្រុងទុកដោយស្វ័យប្រវត្តិធានាឱ្យការបញ្ជូនបន្ត នៅពេលគោលដៅដែលមានសុខភាពល្អផ្សេងទៀតនៅតែមាន។ សសរស្តម្ភប្រាំមួយ៖ ការបម្រុងទុកដែលមានភាពធន់ទ្រាំឆ្លងកាត់អ្នកផ្តល់សេវា 383 · សន្សំ token បានដល់ 95% សម្រាប់ workload ដែលមានសិទ្ធិ · ចាប់ផ្ដើមដោយ $0 ជាមួយ free tier ជាង 150 និងអ្នកផ្តល់សេវាឥតគិតថ្លៃជារៀងរហូតចំនួន 52 ដែលកើតឡើងជាប្រចាំ/មិនត្រូវការឃី · ការរួមបញ្ចូល CLI/agent ចំនួន 39 តាមរយៈ config តែមួយ · ភាពឆបគ្នាជាមួយ OpenAI, Claude, Gemini និង Responses API នៅ /v1 · ការគ្រប់គ្រងសម្រាប់ production រួមមាន circuit breaker, TLS stealth, MCP 110 tools, A2A, memory, guardrails, evals និង static test declaration ជាង 56,000 នៅទូទាំង test file ដែលបានតាមដានជាង 7,800។ +ការសន្យា — endpoint តែមួយ និងអ្នកផ្តល់សេវា 384។ ការបម្រុងទុកដោយស្វ័យប្រវត្តិធានាឱ្យការបញ្ជូនបន្ត នៅពេលគោលដៅដែលមានសុខភាពល្អផ្សេងទៀតនៅតែមាន។ សសរស្តម្ភប្រាំមួយ៖ ការបម្រុងទុកដែលមានភាពធន់ទ្រាំឆ្លងកាត់អ្នកផ្តល់សេវា 384 · សន្សំ token បានដល់ 95% សម្រាប់ workload ដែលមានសិទ្ធិ · ចាប់ផ្ដើមដោយ $0 ជាមួយ free tier ជាង 150 និងអ្នកផ្តល់សេវាឥតគិតថ្លៃជារៀងរហូតចំនួន 52 ដែលកើតឡើងជាប្រចាំ/មិនត្រូវការឃី · ការរួមបញ្ចូល CLI/agent ចំនួន 39 តាមរយៈ config តែមួយ · ភាពឆបគ្នាជាមួយ OpenAI, Claude, Gemini និង Responses API នៅ /v1 · ការគ្រប់គ្រងសម្រាប់ production រួមមាន circuit breaker, TLS stealth, MCP 110 tools, A2A, memory, guardrails, evals និង static test declaration ជាង 56,000 នៅទូទាំង test file ដែលបានតាមដានជាង 7,800។

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -អ្វីដែលធ្វើឱ្យ OmniRoute ខុសប្លែក — សេចក្ដីសង្ខេបមុខងារតាមកាលបរិច្ឆេទ ប្រៀបធៀបជាមួយ 9router, OpenRouter, CLIProxyAPI និង LiteLLM លើសមត្ថភាព 13 ប្រភេទ។ OmniRoute៖ អ្នកផ្តល់សេវា 383, កម្រិតប្រើប្រាស់ឥតគិតថ្លៃជាង 150 ដែលបានបង្កប់, យុទ្ធសាស្ត្រកំណត់ផ្លូវ 20, ការបង្រួម token ដោយប្រើម៉ាស៊ីន 12, ម៉ាស៊ីនមេ MCP ដែលបានបង្កប់ជាមួយឧបករណ៍ 110, ពិធីការ A2A agent, អង្គចងចាំអចិន្ត្រៃយ៍, guardrails, cloud agents, ភាពលាក់បាំងស្នាមម្រាមដៃ TLS, Desktop/Termux/PWA និងមូលដ្ឋានភាសា UI i18n ចំនួន 67។ OmniRoute មានអាជ្ញាប័ណ្ណ MIT និងអាចបង្ហោះដោយខ្លួនឯងបាន។ សមត្ថភាព និងចំនួនរបស់គូប្រកួតអាចផ្លាស់ប្តូរ។ សូមមើលវិធីសាស្ត្រដែលបានភ្ជាប់។ +អ្វីដែលធ្វើឱ្យ OmniRoute ខុសប្លែក — សេចក្ដីសង្ខេបមុខងារតាមកាលបរិច្ឆេទ ប្រៀបធៀបជាមួយ 9router, OpenRouter, CLIProxyAPI និង LiteLLM លើសមត្ថភាព 13 ប្រភេទ។ OmniRoute៖ អ្នកផ្តល់សេវា 384, កម្រិតប្រើប្រាស់ឥតគិតថ្លៃជាង 150 ដែលបានបង្កប់, យុទ្ធសាស្ត្រកំណត់ផ្លូវ 20, ការបង្រួម token ដោយប្រើម៉ាស៊ីន 12, ម៉ាស៊ីនមេ MCP ដែលបានបង្កប់ជាមួយឧបករណ៍ 110, ពិធីការ A2A agent, អង្គចងចាំអចិន្ត្រៃយ៍, guardrails, cloud agents, ភាពលាក់បាំងស្នាមម្រាមដៃ TLS, Desktop/Termux/PWA និងមូលដ្ឋានភាសា UI i18n ចំនួន 67។ OmniRoute មានអាជ្ញាប័ណ្ណ MIT និងអាចបង្ហោះដោយខ្លួនឯងបាន។ សមត្ថភាព និងចំនួនរបស់គូប្រកួតអាចផ្លាស់ប្តូរ។ សូមមើលវិធីសាស្ត្រដែលបានភ្ជាប់។ 📊 វិធីសាស្ត្រពេញលេញ និងព័ត៌មានលម្អិតតាមមុខងារ ប្រៀបធៀបជាមួយ 9router, OpenRouter, CLIProxyAPI និង LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar គឺជាមុខងារដែលត្រូវជ្រើសរ - **🖼️ endpoint ថ្មី** — `/v1/ocr` (Mistral OCR) និង `/v1/audio/translations` (រចនាប័ទ្ម Whisper) បំពេញផ្ទៃប្រព័ន្ធផ្សព្វផ្សាយ។ → [ឯកសារយោង API](docs/reference/API_REFERENCE.md) - **🎨 ការបង្កើតរូបភាព / វីដេអូ / សំឡេង** — API តែមួយសម្រាប់ប្រព័ន្ធផ្សព្វផ្សាយ៖ xAI Grok Imagine & Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind និង provider សំឡេងដូចជា ElevenLabs។ → [ឯកសារយោង API](docs/reference/API_REFERENCE.md) - **🌍 ការដាក់ឱ្យប្រើប្រាស់ និងប្រតិបត្តិការ** — reverse-proxy `basePath`, ការរកឃើញភាសាកម្មវិធីរុករកដោយស្វ័យប្រវត្តិ, ការតាមដានឧបករណ៍តាម key, MITM trust ដោយគ្មាន root និងការធ្វើមូលដ្ឋានីយកម្ម zh-TW។ → [បរិស្ថាន](docs/reference/ENVIRONMENT.md) -- **🤝 provider និង agent បន្ថែមទៀត** — cloud agent (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) ជាមួយ browser + OAuth login, កាត Ollama ថ្នាក់ដំបូង, Claude Opus 5 & Sonnet 5, ភាពជាដៃគូផ្លូវការជាមួយ Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… និង catalog provider ដែលបានធ្វើបច្ចុប្បន្នភាពមាន **383 provider**។ → [Provider](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 provider និង agent បន្ថែមទៀត** — cloud agent (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) ជាមួយ browser + OAuth login, កាត Ollama ថ្នាក់ដំបូង, Claude Opus 5 & Sonnet 5, ភាពជាដៃគូផ្លូវការជាមួយ Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… និង catalog provider ដែលបានធ្វើបច្ចុប្បន្នភាពមាន **384 provider**។ → [Provider](docs/reference/PROVIDER_REFERENCE.md) - **📡 តម្លាភាពនៃការកំណត់ផ្លូវ** — រាល់ការឆ្លើយតបមានបឋមកថា `X-OmniRoute-Decision` ដែលបញ្ជាក់ឈ្មោះយុទ្ធសាស្ត្រ/provider/latency ដែលបានបម្រើវា, យុទ្ធសាស្ត្រ combo `cache-optimized` ថ្មី + កត្តា `cacheAffinity` របស់ Auto-Combo បញ្ជូនសំណើដដែលៗត្រឡប់ទៅ connection ដែលកាន់ cached prefix ហើយ endpoint `/v1/auto-combo/{channel}/candidates` ដែលអាចអានបានតែប៉ុណ្ណោះ បង្ហាញ candidate pool បច្ចុប្បន្នរបស់ channel `auto/*`។ → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ ប្រតិបត្តិការមូលដ្ឋាន និងហេដ្ឋារចនាសម្ព័ន្ធដែលមានប្រសិទ្ធភាព** — Redis មូលដ្ឋានដោយចុចតែម្តង, ឧបករណ៍ដាក់ឱ្យដំណើរការ relay សម្រាប់ Cloudflare Workers / Deno Deploy, និង Bifrost & Mux ជាសេវាកម្មបង្កប់ដែលមានការត្រួតពិនិត្យ។ → [សេវាកម្មបង្កប់](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 មានភ្ជាប់មកជាមួយផងដែរ** — ក្របខណ្ឌកម្មវិធីជំនួយ + ទីផ្សារ, ក្របខណ្ឌជំនាញ Omni/Agent/GitHub, ការរួមបញ្ចូល Obsidian vault (ឧបករណ៍ MCP ចំនួន 22), Batch & Files APIs ដែលអាចប្រើជាមួយ OpenAI, ឃ្លាំងសម្ងាត់ការឆ្លើយតបតាមន័យ, ប្រព័ន្ធលេងហ្គេមជាមួយតារាងចំណាត់ថ្នាក់, ការស្វែងរកភ្នាក់ងារ ACP (ភ្នាក់ងារដែលមានស្រាប់ចំនួន 15), ការនាំចេញកំណត់ហេតុតាមកាលវិភាគទៅ BigQuery, ការបញ្ជូនបន្តម៉ូដែលច្រើនស្របគ្នាតាមរយៈ `auto/chaos`, ស្ពាន bot Telegram, កម្មវិធីគ្រប់គ្រងកំណែក្នុងកម្មវិធី និងចំណាត់ថ្នាក់អ្នកផ្តល់សេវាឥតគិតថ្លៃ LMArena-ELO។ → [ឯកសារ](docs/README.md) @@ -688,11 +688,11 @@ omniroute configure codex # ក៏ដូចជា: claude opencode qwen
-## 🌐 អ្នកផ្តល់សេវា AI ចំនួន 383 — 153 ត្រូវបានសម្គាល់ថាឥតគិតថ្លៃក្នុងកាតាឡុក +## 🌐 អ្នកផ្តល់សេវា AI ចំនួន 384 — 153 ត្រូវបានសម្គាល់ថាឥតគិតថ្លៃក្នុងកាតាឡុក
-> **អ្នកផ្តល់សេវាដែលបានចុះបញ្ជីចំនួន 383** នៅទូទាំងបណ្តុំ canonical chat, media, search, local, cloud-agent និង system ដែលក្នុងនោះមាន **153 មានទិន្នន័យមេតាសម្រាប់ការស្វែងរក `hasFree: true`**។ បញ្ជីឈ្មោះម៉ូដែលជជែកគ្របដណ្តប់ **អ្នកផ្តល់សេវា 239 / គូអ្នកផ្តល់សេវា-ម៉ូដែលខុសគ្នា 3,059 / លេខសម្គាល់ម៉ូដែលដើម 1,475**; កាតាឡុកថវិកាឥតគិតថ្លៃដាច់ដោយឡែកមាន **ជួរតាមម៉ូដែល 471**, **pool ដែលកើតឡើងជាប្រចាំ 34** និង **អ្នកផ្តល់សេវាឥតគិតថ្លៃជារៀងរហូត ដែលកើតឡើងជាប្រចាំ/មិនត្រូវការសោ 52**។ ទាំងនេះជាភាគបែងខុសគ្នាតាមការរចនា; និយមន័យ និងការគណនាដែលបានដកស្ទួនតាម pool មាននៅក្នុង [ឯកសារយោងអ្នកផ្តល់សេវា](docs/reference/PROVIDER_REFERENCE.md) និង [កម្រិតឥតគិតថ្លៃ](docs/reference/FREE_TIERS.md)។ +> **អ្នកផ្តល់សេវាដែលបានចុះបញ្ជីចំនួន 384** នៅទូទាំងបណ្តុំ canonical chat, media, search, local, cloud-agent និង system ដែលក្នុងនោះមាន **153 មានទិន្នន័យមេតាសម្រាប់ការស្វែងរក `hasFree: true`**។ បញ្ជីឈ្មោះម៉ូដែលជជែកគ្របដណ្តប់ **អ្នកផ្តល់សេវា 239 / គូអ្នកផ្តល់សេវា-ម៉ូដែលខុសគ្នា 3,059 / លេខសម្គាល់ម៉ូដែលដើម 1,475**; កាតាឡុកថវិកាឥតគិតថ្លៃដាច់ដោយឡែកមាន **ជួរតាមម៉ូដែល 471**, **pool ដែលកើតឡើងជាប្រចាំ 34** និង **អ្នកផ្តល់សេវាឥតគិតថ្លៃជារៀងរហូត ដែលកើតឡើងជាប្រចាំ/មិនត្រូវការសោ 52**។ ទាំងនេះជាភាគបែងខុសគ្នាតាមការរចនា; និយមន័យ និងការគណនាដែលបានដកស្ទួនតាម pool មាននៅក្នុង [ឯកសារយោងអ្នកផ្តល់សេវា](docs/reference/PROVIDER_REFERENCE.md) និង [កម្រិតឥតគិតថ្លៃ](docs/reference/FREE_TIERS.md)។
@@ -1294,7 +1294,7 @@ process ដូចគ្នានៅលើ port តែមួយ ដូច្ន បរិស្ថានដំណើរការNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ភាសាTypeScript 6.0 — TypeScript 100% នៅទូទាំង src/ និង open-sse/ (គ្មាន any ក្នុងស្នូល ចាប់ពី v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - មូលដ្ឋានទិន្នន័យbetter-sqlite3 (SQLite, កំណត់ហេតុ WAL) + LowDB (JSON ចាស់) — ម៉ូឌុលដែន 147, migrations 230 + មូលដ្ឋានទិន្នន័យbetter-sqlite3 (SQLite, កំណត់ហេតុ WAL) + LowDB (JSON ចាស់) — ម៉ូឌុលដែន 147, migrations 232 អង្គចងចាំអត្ថបទពេញលេញ SQLite FTS5 + vector embeddings ដែលបាន quantize ទៅជា int8, ការថយចុះតាមប្រភេទ SchemaZod 4 — ការផ្ទៀងផ្ទាត់ I/O របស់ឧបករណ៍ MCP + កិច្ចសន្យា API ពិធីការMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/km/llm.txt b/docs/i18n/km/llm.txt index 9d9ada2fce5..9e762b3fa21 100644 --- a/docs/i18n/km/llm.txt +++ b/docs/i18n/km/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/kn/README.md b/docs/i18n/kn/README.md index ab326609640..a6e9db5eb6e 100644 --- a/docs/i18n/kn/README.md +++ b/docs/i18n/kn/README.md @@ -15,7 +15,7 @@ **ರಚನೆ: [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — ಎಂದಿಗೂ ಕೋಡಿಂಗ್ ನಿಲ್ಲಿಸಬೇಡಿ. ಪ್ರತಿಯೊಂದು AI ಟೂಲ್ → 383 ಪೂರೈಕೆದಾರರು — 150+ ಉಚಿತ — ಒಂದೇ ಎಂಡ್ಪಾಯಿಂಟ್ ಮೂಲಕ. Claude Code, Codex, Cursor, Cline, Copilot ಮತ್ತು Antigravity ಅನ್ನು ಸ್ವಯಂ-ಫಾಲ್ಬ್ಯಾಕ್ನೊಂದಿಗೆ ಉಚಿತ Claude / GPT / Gemini ಗೆ ಸಂಪರ್ಕಿಸಿ. RTK + Caveman ಸ್ಟ್ಯಾಕ್ ಮಾಡಲಾದ ಸಂಕುಚನವು 15–95% ಟೋಕನ್ಗಳನ್ನು (~89% ಸರಾಸರಿ) ಉಳಿಸುತ್ತದೆ — ಮಿತಿಗಳನ್ನು ಎಂದಿಗೂ ತಲುಪಬೇಡಿ. 383 AI ಪೂರೈಕೆದಾರರು · 150+ ಉಚಿತ ಶ್ರೇಣಿಗಳು · ತಿಂಗಳಿಗೆ ~1.62B ಉಚಿತ ಟೋಕನ್ಗಳು · 20 ರೂಟಿಂಗ್ ತಂತ್ರಗಳು · ಪ್ರಾರಂಭಿಸಲು $0. +OmniRoute — ಎಂದಿಗೂ ಕೋಡಿಂಗ್ ನಿಲ್ಲಿಸಬೇಡಿ. ಪ್ರತಿಯೊಂದು AI ಟೂಲ್ → 384 ಪೂರೈಕೆದಾರರು — 150+ ಉಚಿತ — ಒಂದೇ ಎಂಡ್ಪಾಯಿಂಟ್ ಮೂಲಕ. Claude Code, Codex, Cursor, Cline, Copilot ಮತ್ತು Antigravity ಅನ್ನು ಸ್ವಯಂ-ಫಾಲ್ಬ್ಯಾಕ್ನೊಂದಿಗೆ ಉಚಿತ Claude / GPT / Gemini ಗೆ ಸಂಪರ್ಕಿಸಿ. RTK + Caveman ಸ್ಟ್ಯಾಕ್ ಮಾಡಲಾದ ಸಂಕುಚನವು 15–95% ಟೋಕನ್ಗಳನ್ನು (~89% ಸರಾಸರಿ) ಉಳಿಸುತ್ತದೆ — ಮಿತಿಗಳನ್ನು ಎಂದಿಗೂ ತಲುಪಬೇಡಿ. 384 AI ಪೂರೈಕೆದಾರರು · 150+ ಉಚಿತ ಶ್ರೇಣಿಗಳು · ತಿಂಗಳಿಗೆ ~1.62B ಉಚಿತ ಟೋಕನ್ಗಳು · 20 ರೂಟಿಂಗ್ ತಂತ್ರಗಳು · ಪ್ರಾರಂಭಿಸಲು $0.
@@ -73,7 +73,7 @@ OMNIROUTE ನಿಮ್ಮ ಹಣ ಉಳಿಸಲು ಮತ್ತು ಕೆಲಸ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------- | :-----------------------: | :---------: | :-------------------: | -| 🌐 ಪೂರೈಕೆದಾರರು | 352 | **358** | 383 | +| 🌐 ಪೂರೈಕೆದಾರರು | 352 | **358** | 384 | | 🧠 ವಿಶಿಷ್ಟ ಚಾಟ್ ಮಾದರಿ IDಗಳು | 1320 | **1374** | 1491 | | 🖼️ ಮೋಡಾಲಿಟಿ ಸೇತುವೆ | 🆕 ದೃಶ್ಯ + ಆಡಿಯೋ + ವೀಡಿಯೋ | ✓ | ✓ | | 📡 Radar ಉಚಿತ ಕ್ಯಾಟಲಾಗ್ | 🆕 ಆಯ್ಕೆ-ಆಧಾರಿತ | ✓ | ✓ | @@ -113,7 +113,7 @@ OMNIROUTE ನಿಮ್ಮ ಹಣ ಉಳಿಸಲು ಮತ್ತು ಕೆಲಸ ⚙️ ವೈಶಿಷ್ಟ್ಯಗಳು 🎯 ಕಾಂಬೊಗಳು - 🌐 ಪೂರೈಕೆದಾರರು + 🌐 ಪೂರೈಕೆದಾರರು 🔌 CLI & MCP @@ -249,7 +249,7 @@ curl http://localhost:20128/v1/chat/completions \ -ಭರವಸೆ — ಒಂದೇ ಎಂಡ್ಪಾಯಿಂಟ್ ಮತ್ತು 383 ಒದಗಿಸುವವರು. ಮತ್ತೊಂದು ಆರೋಗ್ಯಕರ ಗುರಿ ಲಭ್ಯವಿರುವವರೆಗೆ ಸ್ವಯಂಚಾಲಿತ ಫಾಲ್ಬ್ಯಾಕ್ ಮಾರ್ಗೀಕರಣವನ್ನು ಮುಂದುವರಿಸುತ್ತದೆ. ಆರು ಸ್ತಂಭಗಳು: 383 ಒದಗಿಸುವವರಾದ್ಯಂತ ಸ್ಥಿರವಾದ ಫಾಲ್ಬ್ಯಾಕ್ · ಅರ್ಹ ವರ್ಕ್ಲೋಡ್ಗಳಲ್ಲಿ 95% ವರೆಗೆ ಟೋಕನ್ ಉಳಿತಾಯ · 150+ ಉಚಿತ ಹಂತಗಳು ಮತ್ತು ಮರುಕಳಿಸುವ/ಕೀಲಿರಹಿತವಾಗಿ ಶಾಶ್ವತವಾಗಿ ಉಚಿತವಾಗಿರುವ 52 ಒದಗಿಸುವವರೊಂದಿಗೆ ಪ್ರಾರಂಭಿಸಲು $0 · ಒಂದೇ ಕಾನ್ಫಿಗರೇಶನ್ ಮೂಲಕ 39 CLI/ಏಜೆಂಟ್ ಏಕೀಕರಣಗಳು · /v1 ನಲ್ಲಿ OpenAI, Claude, Gemini ಮತ್ತು Responses API ಹೊಂದಾಣಿಕೆ · ಸರ್ಕ್ಯೂಟ್ ಬ್ರೇಕರ್ಗಳು, TLS ಸ್ಟೆಲ್ತ್, 110 MCP ಟೂಲ್ಗಳು, A2A, ಮೆಮೊರಿ, ಗಾರ್ಡ್ರೇಲ್ಗಳು, ಮೌಲ್ಯಮಾಪನಗಳು ಮತ್ತು 7,800+ ಟ್ರ್ಯಾಕ್ ಮಾಡಲಾದ ಟೆಸ್ಟ್ ಫೈಲ್ಗಳಲ್ಲಿ 56,000+ ಸ್ಥಿರ ಟೆಸ್ಟ್ ಘೋಷಣೆಗಳನ್ನು ಒಳಗೊಂಡ ಉತ್ಪಾದನಾ ನಿಯಂತ್ರಣಗಳು. +ಭರವಸೆ — ಒಂದೇ ಎಂಡ್ಪಾಯಿಂಟ್ ಮತ್ತು 384 ಒದಗಿಸುವವರು. ಮತ್ತೊಂದು ಆರೋಗ್ಯಕರ ಗುರಿ ಲಭ್ಯವಿರುವವರೆಗೆ ಸ್ವಯಂಚಾಲಿತ ಫಾಲ್ಬ್ಯಾಕ್ ಮಾರ್ಗೀಕರಣವನ್ನು ಮುಂದುವರಿಸುತ್ತದೆ. ಆರು ಸ್ತಂಭಗಳು: 384 ಒದಗಿಸುವವರಾದ್ಯಂತ ಸ್ಥಿರವಾದ ಫಾಲ್ಬ್ಯಾಕ್ · ಅರ್ಹ ವರ್ಕ್ಲೋಡ್ಗಳಲ್ಲಿ 95% ವರೆಗೆ ಟೋಕನ್ ಉಳಿತಾಯ · 150+ ಉಚಿತ ಹಂತಗಳು ಮತ್ತು ಮರುಕಳಿಸುವ/ಕೀಲಿರಹಿತವಾಗಿ ಶಾಶ್ವತವಾಗಿ ಉಚಿತವಾಗಿರುವ 52 ಒದಗಿಸುವವರೊಂದಿಗೆ ಪ್ರಾರಂಭಿಸಲು $0 · ಒಂದೇ ಕಾನ್ಫಿಗರೇಶನ್ ಮೂಲಕ 39 CLI/ಏಜೆಂಟ್ ಏಕೀಕರಣಗಳು · /v1 ನಲ್ಲಿ OpenAI, Claude, Gemini ಮತ್ತು Responses API ಹೊಂದಾಣಿಕೆ · ಸರ್ಕ್ಯೂಟ್ ಬ್ರೇಕರ್ಗಳು, TLS ಸ್ಟೆಲ್ತ್, 110 MCP ಟೂಲ್ಗಳು, A2A, ಮೆಮೊರಿ, ಗಾರ್ಡ್ರೇಲ್ಗಳು, ಮೌಲ್ಯಮಾಪನಗಳು ಮತ್ತು 7,800+ ಟ್ರ್ಯಾಕ್ ಮಾಡಲಾದ ಟೆಸ್ಟ್ ಫೈಲ್ಗಳಲ್ಲಿ 56,000+ ಸ್ಥಿರ ಟೆಸ್ಟ್ ಘೋಷಣೆಗಳನ್ನು ಒಳಗೊಂಡ ಉತ್ಪಾದನಾ ನಿಯಂತ್ರಣಗಳು.

@@ -509,7 +509,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRoute ಅನ್ನು ವಿಭಿನ್ನಗೊಳಿಸುವ ಅಂಶಗಳು — 13 ಸಾಮರ್ಥ್ಯಗಳಲ್ಲಿ 9router, OpenRouter, CLIProxyAPI ಮತ್ತು LiteLLM ಗಳೊಂದಿಗೆ ದಿನಾಂಕಸಹಿತ ವೈಶಿಷ್ಟ್ಯ ಸ್ನ್ಯಾಪ್ಶಾಟ್ ಹೋಲಿಕೆ. OmniRoute: 383 ಪೂರೈಕೆದಾರರು, ಅಂತರ್ನಿರ್ಮಿತ 150+ ಉಚಿತ ಶ್ರೇಣಿಗಳು, 20 ರೌಟಿಂಗ್ ತಂತ್ರಗಳು, 12-ಎಂಜಿನ್ ಟೋಕನ್ ಸಂಕುಚನ, 110 ಪರಿಕರಗಳೊಂದಿಗೆ ಅಂತರ್ನಿರ್ಮಿತ MCP ಸರ್ವರ್, A2A ಏಜೆಂಟ್ ಪ್ರೋಟೋಕಾಲ್, ನಿರಂತರ ಮೆಮೊರಿ, ಗಾರ್ಡ್ರೇಲ್ಗಳು, ಕ್ಲೌಡ್ ಏಜೆಂಟ್ಗಳು, TLS ಫಿಂಗರ್ಪ್ರಿಂಟ್ ಸ್ಟೆಲ್ತ್, Desktop/Termux/PWA ಮತ್ತು 67 i18n UI ಲೊಕೇಲ್ಗಳು. OmniRoute MIT-ಲೈಸೆನ್ಸ್ ಹೊಂದಿದ್ದು, ಸ್ವಯಂ-ಹೋಸ್ಟ್ ಮಾಡಬಹುದು. ಸ್ಪರ್ಧಿ ಸಾಮರ್ಥ್ಯಗಳು ಮತ್ತು ಎಣಿಕೆಗಳು ಬದಲಾಗಬಹುದು; ಲಿಂಕ್ ಮಾಡಲಾದ ವಿಧಾನಶಾಸ್ತ್ರವನ್ನು ನೋಡಿ. +OmniRoute ಅನ್ನು ವಿಭಿನ್ನಗೊಳಿಸುವ ಅಂಶಗಳು — 13 ಸಾಮರ್ಥ್ಯಗಳಲ್ಲಿ 9router, OpenRouter, CLIProxyAPI ಮತ್ತು LiteLLM ಗಳೊಂದಿಗೆ ದಿನಾಂಕಸಹಿತ ವೈಶಿಷ್ಟ್ಯ ಸ್ನ್ಯಾಪ್ಶಾಟ್ ಹೋಲಿಕೆ. OmniRoute: 384 ಪೂರೈಕೆದಾರರು, ಅಂತರ್ನಿರ್ಮಿತ 150+ ಉಚಿತ ಶ್ರೇಣಿಗಳು, 20 ರೌಟಿಂಗ್ ತಂತ್ರಗಳು, 12-ಎಂಜಿನ್ ಟೋಕನ್ ಸಂಕುಚನ, 110 ಪರಿಕರಗಳೊಂದಿಗೆ ಅಂತರ್ನಿರ್ಮಿತ MCP ಸರ್ವರ್, A2A ಏಜೆಂಟ್ ಪ್ರೋಟೋಕಾಲ್, ನಿರಂತರ ಮೆಮೊರಿ, ಗಾರ್ಡ್ರೇಲ್ಗಳು, ಕ್ಲೌಡ್ ಏಜೆಂಟ್ಗಳು, TLS ಫಿಂಗರ್ಪ್ರಿಂಟ್ ಸ್ಟೆಲ್ತ್, Desktop/Termux/PWA ಮತ್ತು 67 i18n UI ಲೊಕೇಲ್ಗಳು. OmniRoute MIT-ಲೈಸೆನ್ಸ್ ಹೊಂದಿದ್ದು, ಸ್ವಯಂ-ಹೋಸ್ಟ್ ಮಾಡಬಹುದು. ಸ್ಪರ್ಧಿ ಸಾಮರ್ಥ್ಯಗಳು ಮತ್ತು ಎಣಿಕೆಗಳು ಬದಲಾಗಬಹುದು; ಲಿಂಕ್ ಮಾಡಲಾದ ವಿಧಾನಶಾಸ್ತ್ರವನ್ನು ನೋಡಿ. 📊 9router, OpenRouter, CLIProxyAPI ಮತ್ತು LiteLLM ಗಳೊಂದಿಗೆ ಸಂಪೂರ್ಣ ವಿಧಾನಶಾಸ್ತ್ರ ಮತ್ತು ಪ್ರತಿ ವೈಶಿಷ್ಟ್ಯದ ವಿವರ → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -604,7 +604,7 @@ Radar ಬಳಸುವುದು ಐಚ್ಛಿಕವಾಗಿದ್ದು, ಅ - **🖼️ ಹೊಸ endpoints** — `/v1/ocr` (Mistral OCR) ಮತ್ತು `/v1/audio/translations` (Whisper-style) media surface ಅನ್ನು ಸಂಪೂರ್ಣಗೊಳಿಸುತ್ತವೆ. → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 Image / video / audio generation** — media ಗಾಗಿ ಒಂದೇ API: xAI Grok Imagine ಮತ್ತು Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind ಹಾಗೂ ElevenLabsನಂತಹ speech providers. → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 Deployment ಮತ್ತು ops** — reverse-proxy `basePath`, browser-language auto-detect, ಪ್ರತಿ ಕೀಗೆ device tracking, root-less MITM trust, zh-TW localization. → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 ಇನ್ನಷ್ಟು providers ಮತ್ತು agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login ಹೊಂದಿರುವ Grok Build (xAI), first-class card ಹೊಂದಿರುವ Ollama, Claude Opus 5 ಮತ್ತು Sonnet 5, Kimi ಅಧಿಕೃತ partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… ಮತ್ತು ನವೀಕರಿಸಲಾದ **383-provider catalog**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 ಇನ್ನಷ್ಟು providers ಮತ್ತು agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login ಹೊಂದಿರುವ Grok Build (xAI), first-class card ಹೊಂದಿರುವ Ollama, Claude Opus 5 ಮತ್ತು Sonnet 5, Kimi ಅಧಿಕೃತ partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… ಮತ್ತು ನವೀಕರಿಸಲಾದ **384-provider catalog**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 Routing transparency** — ಪ್ರತಿಯೊಂದು response ಅದನ್ನು ಪೂರೈಸಿದ strategy/provider/latency ಅನ್ನು ಹೆಸರಿಸುವ `X-OmniRoute-Decision` header ಅನ್ನು ಹೊಂದಿರುತ್ತದೆ; ಹೊಸ `cache-optimized` combo strategy ಮತ್ತು Auto-Combo `cacheAffinity` factor, cached prefix ಹೊಂದಿರುವ connectionಗೆ ಪುನರಾವರ್ತಿತ requests ಅನ್ನು ಹಿಂದಿರುಗಿಸುತ್ತವೆ; read-only `/v1/auto-combo/{channel}/candidates` endpoint `auto/*` channelನ live candidate pool ಅನ್ನು ಬಹಿರಂಗಪಡಿಸುತ್ತದೆ. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ ಸ್ಥಳೀಯ ಕಾರ್ಯಕ್ಷಮತೆ ಮತ್ತು ಮೂಲಸೌಕರ್ಯ** — ಒಂದೇ ಕ್ಲಿಕ್ನಲ್ಲಿ ಸ್ಥಳೀಯ Redis, Cloudflare Workers / Deno Deploy ರಿಲೇ ಡಿಪ್ಲಾಯರ್ಗಳು, ಮೇಲ್ವಿಚಾರಿತ ಎಂಬೆಡೆಡ್ ಸೇವೆಗಳಾಗಿ Bifrost ಮತ್ತು Mux. → [ಎಂಬೆಡೆಡ್ ಸೇವೆಗಳು](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 ಇದೂ ಸಹ ಒಳಗೊಂಡಿದೆ** — ಪ್ಲಗಿನ್ ಫ್ರೇಮ್ವರ್ಕ್ ಮತ್ತು ಮಾರ್ಕೆಟ್ಪ್ಲೇಸ್, Omni/Agent/GitHub ಸ್ಕಿಲ್ಸ್ ಫ್ರೇಮ್ವರ್ಕ್ಗಳು, Obsidian ವಾಲ್ಟ್ ಇಂಟಿಗ್ರೇಶನ್ (22 MCP ಟೂಲ್ಗಳು), OpenAI-ಹೊಂದಾಣಿಕೆಯ Batch ಮತ್ತು Files APIಗಳು, ಸಿಮೆಂಟಿಕ್ ಪ್ರತಿಕ್ರಿಯೆ ಕ್ಯಾಶ್, ಲೀಡರ್ಬೋರ್ಡ್ಗಳೊಂದಿಗೆ ಗ್ಯಾಮಿಫಿಕೇಶನ್, ACP ಏಜೆಂಟ್ ಡಿಸ್ಕವರಿ (15 ಅಂತರ್ನಿರ್ಮಿತ ಏಜೆಂಟ್ಗಳು), BigQueryಗೆ ನಿಗದಿತ ಲಾಗ್ ರಫ್ತು, `auto/chaos` ಸಮಾಂತರ ಬಹು-ಮಾದರಿ ಫ್ಯಾನ್-ಔಟ್, Telegram ಬಾಟ್ ಬ್ರಿಡ್ಜ್, ಅಪ್ಲಿಕೇಶನ್ನಲ್ಲೇ ಇರುವ ಆವೃತ್ತಿ ನಿರ್ವಾಹಕ ಮತ್ತು LMArena-ELO ಉಚಿತ-ಪ್ರೊವೈಡರ್ ಶ್ರೇಯಾಂಕಗಳು. → [ದಾಖಲೆಗಳು](docs/README.md) @@ -687,11 +687,11 @@ omniroute configure codex # ಇದಕ್ಕೂ: claude opencode qwen ai
-## 🌐 383 AI ಪೂರೈಕೆದಾರರು — ಕ್ಯಾಟಲಾಗ್ನಲ್ಲಿ ಗುರುತಿಸಲಾದ 153 ಉಚಿತ +## 🌐 384 AI ಪೂರೈಕೆದಾರರು — ಕ್ಯಾಟಲಾಗ್ನಲ್ಲಿ ಗುರುತಿಸಲಾದ 153 ಉಚಿತ
-> **383 ನೋಂದಾಯಿತ ಪೂರೈಕೆದಾರರು** canonical chat, media, search, local, cloud-agent ಮತ್ತು system ಸಂಗ್ರಹಗಳಾದ್ಯಂತ ಇದ್ದಾರೆ; ಇವರಲ್ಲಿ **153 ಪೂರೈಕೆದಾರರು `hasFree: true` ಅನ್ವೇಷಣಾ ಮೆಟಾಡೇಟಾವನ್ನು ಹೊಂದಿದ್ದಾರೆ**. chat model registryಯಲ್ಲಿ **239 ಪೂರೈಕೆದಾರರು / 3,059 ವಿಭಿನ್ನ provider-model ಜೋಡಿಗಳು / 1,475 ಕಚ್ಚಾ model IDಗಳು** ಒಳಗೊಂಡಿವೆ; ಪ್ರತ್ಯೇಕ free-budget catalogನಲ್ಲಿ **471 ಪ್ರತಿ-model ಸಾಲುಗಳು**, **34 ಮರುಕಳಿಸುವ poolಗಳು** ಮತ್ತು **52 ಮರುಕಳಿಸುವ/keyless free-forever ಪೂರೈಕೆದಾರರು** ಇದ್ದಾರೆ. ಇವು ವಿನ್ಯಾಸಪೂರ್ವಕವಾಗಿ ವಿಭಿನ್ನ denominatorಗಳಾಗಿವೆ; ವ್ಯಾಖ್ಯಾನಗಳು ಮತ್ತು pool-deduped ಲೆಕ್ಕಾಚಾರಗಳು [ಪೂರೈಕೆದಾರರ ಉಲ್ಲೇಖ](docs/reference/PROVIDER_REFERENCE.md) ಮತ್ತು [ಉಚಿತ ಹಂತಗಳು](docs/reference/FREE_TIERS.md)ಗಳಲ್ಲಿ ಲಭ್ಯವಿವೆ. +> **384 ನೋಂದಾಯಿತ ಪೂರೈಕೆದಾರರು** canonical chat, media, search, local, cloud-agent ಮತ್ತು system ಸಂಗ್ರಹಗಳಾದ್ಯಂತ ಇದ್ದಾರೆ; ಇವರಲ್ಲಿ **153 ಪೂರೈಕೆದಾರರು `hasFree: true` ಅನ್ವೇಷಣಾ ಮೆಟಾಡೇಟಾವನ್ನು ಹೊಂದಿದ್ದಾರೆ**. chat model registryಯಲ್ಲಿ **239 ಪೂರೈಕೆದಾರರು / 3,059 ವಿಭಿನ್ನ provider-model ಜೋಡಿಗಳು / 1,475 ಕಚ್ಚಾ model IDಗಳು** ಒಳಗೊಂಡಿವೆ; ಪ್ರತ್ಯೇಕ free-budget catalogನಲ್ಲಿ **471 ಪ್ರತಿ-model ಸಾಲುಗಳು**, **34 ಮರುಕಳಿಸುವ poolಗಳು** ಮತ್ತು **52 ಮರುಕಳಿಸುವ/keyless free-forever ಪೂರೈಕೆದಾರರು** ಇದ್ದಾರೆ. ಇವು ವಿನ್ಯಾಸಪೂರ್ವಕವಾಗಿ ವಿಭಿನ್ನ denominatorಗಳಾಗಿವೆ; ವ್ಯಾಖ್ಯಾನಗಳು ಮತ್ತು pool-deduped ಲೆಕ್ಕಾಚಾರಗಳು [ಪೂರೈಕೆದಾರರ ಉಲ್ಲೇಖ](docs/reference/PROVIDER_REFERENCE.md) ಮತ್ತು [ಉಚಿತ ಹಂತಗಳು](docs/reference/FREE_TIERS.md)ಗಳಲ್ಲಿ ಲಭ್ಯವಿವೆ.
@@ -1293,7 +1293,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 ಕೂಡ ಇದ ರನ್ಟೈಮ್Node.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ಭಾಷೆTypeScript 6.0 — src/ ಮತ್ತು open-sse/ ಗಳಾದ್ಯಂತ 100% TypeScript (v2.0ರಿಂದ ಕೋರ್ನಲ್ಲಿ ಶೂನ್ಯ any) ಫ್ರೇಮ್ವರ್ಕ್Next.js 16 + React 19 + Tailwind CSS 4 - ಡೇಟಾಬೇಸ್better-sqlite3 (SQLite, WAL ಜರ್ನಲಿಂಗ್) + LowDB (JSON ಪರಂಪರাগত) — 147 ಡೊಮೇನ್ ಮಾಡ್ಯೂಲ್ಗಳು, 230 ಮೈಗ್ರೇಶನ್ಗಳು + ಡೇಟಾಬೇಸ್better-sqlite3 (SQLite, WAL ಜರ್ನಲಿಂಗ್) + LowDB (JSON ಪರಂಪರাগত) — 147 ಡೊಮೇನ್ ಮಾಡ್ಯೂಲ್ಗಳು, 232 ಮೈಗ್ರೇಶನ್ಗಳು ಮೆಮೊರಿSQLite FTS5 ಪೂರ್ಣ-ಪಠ್ಯ + int8-ಕ್ವಾಂಟೈಸ್ ಮಾಡಲಾದ ವೆಕ್ಟರ್ ಎಂಬೆಡಿಂಗ್ಗಳು, ಟೈಪ್ ಮಾಡಲಾದ ಕ್ಷಯ ಸ್ಕೀಮಾಗಳುZod 4 — MCP ಟೂಲ್ I/O ಪರಿಶೀಲನೆ + API ಒಪ್ಪಂದಗಳು ಪ್ರೋಟೋಕಾಲ್ಗಳುMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/kn/llm.txt b/docs/i18n/kn/llm.txt index 2c2c9619fe9..fe54033c0f1 100644 --- a/docs/i18n/kn/llm.txt +++ b/docs/i18n/kn/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ko/README.md b/docs/i18n/ko/README.md index 7916da1a0b9..9b16dc898f5 100644 --- a/docs/i18n/ko/README.md +++ b/docs/i18n/ko/README.md @@ -15,7 +15,7 @@ **[Diego Souza](https://github.com/diegosouzapw)가 제작** -OmniRoute — 코딩을 절대 멈추지 마세요. 모든 AI 도구 → 하나의 엔드포인트를 통해 383개 제공업체 — 150개 이상 무료 — Claude Code, Codex, Cursor, Cline, Copilot 및 Antigravity를 무료 Claude / GPT / Gemini로 자동 대체 기능과 함께 연결합니다. RTK + Caveman 스택 압축으로 토큰을 15–95% 절약(평균 약 89%)하여 한도에 도달하지 않습니다. AI 제공업체 383개 · 무료 티어 150개 이상 · 월 약 16.2억 개의 무료 토큰 · 라우팅 전략 20개 · $0로 시작. +OmniRoute — 코딩을 절대 멈추지 마세요. 모든 AI 도구 → 하나의 엔드포인트를 통해 384개 제공업체 — 150개 이상 무료 — Claude Code, Codex, Cursor, Cline, Copilot 및 Antigravity를 무료 Claude / GPT / Gemini로 자동 대체 기능과 함께 연결합니다. RTK + Caveman 스택 압축으로 토큰을 15–95% 절약(평균 약 89%)하여 한도에 도달하지 않습니다. AI 제공업체 384개 · 무료 티어 150개 이상 · 월 약 16.2억 개의 무료 토큰 · 라우팅 전략 20개 · $0로 시작.
@@ -73,7 +73,7 @@ OMNIROUTE가 비용을 절약하고 작업을 더 쉽게 만드는 데 도움이 | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ----------------------- | :-----------------------: | :---------: | :-------------------: | -| 🌐 제공업체 | 352 | **358** | 383 | +| 🌐 제공업체 | 352 | **358** | 384 | | 🧠 고유 채팅 모델 ID | 1320 | **1374** | 1491 | | 🖼️ 모달리티 브리지 | 🆕 비전 + 오디오 + 비디오 | ✓ | ✓ | | 📡 Radar 무료 카탈로그 | 🆕 옵트인 | ✓ | ✓ | @@ -114,7 +114,7 @@ OMNIROUTE가 비용을 절약하고 작업을 더 쉽게 만드는 데 도움이 ⚙️ 기능 🎯 콤보 - 🌐 제공업체 + 🌐 제공업체 🔌 CLI 및 MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -약속 — 하나의 엔드포인트와 383개 공급자. 다른 정상 대상이 사용 가능한 동안 자동 폴백이 라우팅을 유지합니다. 6가지 핵심 요소: 383개 공급자에 걸친 복원력 있는 폴백 · 적격 워크로드에서 최대 95% 토큰 절약 · 150개 이상의 무료 티어와 52개의 반복 제공/키 없는 영구 무료 공급자를 통해 $0로 시작 · 하나의 구성으로 39개의 CLI/에이전트 통합 · /v1에서 OpenAI, Claude, Gemini 및 Responses API 호환 · 회로 차단기, TLS 스텔스, MCP 110개 도구, A2A, 메모리, 가드레일, 평가 및 7,800개 이상의 추적 테스트 파일에 걸친 56,000개 이상의 정적 테스트 선언을 포함한 프로덕션 제어 기능. +약속 — 하나의 엔드포인트와 384개 공급자. 다른 정상 대상이 사용 가능한 동안 자동 폴백이 라우팅을 유지합니다. 6가지 핵심 요소: 384개 공급자에 걸친 복원력 있는 폴백 · 적격 워크로드에서 최대 95% 토큰 절약 · 150개 이상의 무료 티어와 52개의 반복 제공/키 없는 영구 무료 공급자를 통해 $0로 시작 · 하나의 구성으로 39개의 CLI/에이전트 통합 · /v1에서 OpenAI, Claude, Gemini 및 Responses API 호환 · 회로 차단기, TLS 스텔스, MCP 110개 도구, A2A, 메모리, 가드레일, 평가 및 7,800개 이상의 추적 테스트 파일에 걸친 56,000개 이상의 정적 테스트 선언을 포함한 프로덕션 제어 기능.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRoute의 차별점 — 13가지 기능에 걸쳐 9router, OpenRouter, CLIProxyAPI 및 LiteLLM과 비교한 날짜 기준 기능 스냅샷. OmniRoute: 383개 프로바이더, 기본 제공되는 150개 이상의 무료 티어, 20가지 라우팅 전략, 12개 엔진 토큰 압축, 110개 도구가 포함된 기본 제공 MCP 서버, A2A 에이전트 프로토콜, 영구 메모리, 가드레일, 클라우드 에이전트, TLS 지문 스텔스, Desktop/Termux/PWA 및 67개 i18n UI 로캘. OmniRoute는 MIT 라이선스로 제공되며 자체 호스팅할 수 있습니다. 경쟁 제품의 기능 및 수치는 변경될 수 있습니다. 자세한 내용은 링크된 방법론을 참조하세요. +OmniRoute의 차별점 — 13가지 기능에 걸쳐 9router, OpenRouter, CLIProxyAPI 및 LiteLLM과 비교한 날짜 기준 기능 스냅샷. OmniRoute: 384개 프로바이더, 기본 제공되는 150개 이상의 무료 티어, 20가지 라우팅 전략, 12개 엔진 토큰 압축, 110개 도구가 포함된 기본 제공 MCP 서버, A2A 에이전트 프로토콜, 영구 메모리, 가드레일, 클라우드 에이전트, TLS 지문 스텔스, Desktop/Termux/PWA 및 67개 i18n UI 로캘. OmniRoute는 MIT 라이선스로 제공되며 자체 호스팅할 수 있습니다. 경쟁 제품의 기능 및 수치는 변경될 수 있습니다. 자세한 내용은 링크된 방법론을 참조하세요. 📊 9router, OpenRouter, CLIProxyAPI 및 LiteLLM과의 전체 방법론 및 기능별 세부 정보 → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar는 선택적으로 사용할 수 있으며 GET 전용입니다. OmniRoute - **🖼️ 새로운 엔드포인트** — `/v1/ocr`(Mistral OCR) 및 `/v1/audio/translations`(Whisper 스타일)가 미디어 기능을 확장합니다. → [API 레퍼런스](docs/reference/API_REFERENCE.md) - **🎨 이미지 / 동영상 / 오디오 생성** — 미디어를 위한 단일 API를 제공하며, xAI Grok Imagine 및 Novita AI 동영상, ComfyUI, Magnific, Adobe Firefly, Segmind, ElevenLabs와 같은 음성 제공자를 지원합니다. → [API 레퍼런스](docs/reference/API_REFERENCE.md) - **🌍 배포 및 운영** — 리버스 프록시 `basePath`, 브라우저 언어 자동 감지, 키별 장치 추적, 루트 권한 없는 MITM trust, zh-TW 현지화를 지원합니다. → [환경](docs/reference/ENVIRONMENT.md) -- **🤝 더 많은 제공자 및 에이전트** — 클라우드 에이전트(Codex Cloud, Cursor, Devin, Jules), 브라우저 + OAuth 로그인을 지원하는 Grok Build(xAI), Ollama 퍼스트 클래스 카드, Claude Opus 5 및 Sonnet 5, Kimi 공식 파트너십(Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… 그리고 새롭게 정비된 **383개 제공자 카탈로그**를 제공합니다. → [제공자](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 더 많은 제공자 및 에이전트** — 클라우드 에이전트(Codex Cloud, Cursor, Devin, Jules), 브라우저 + OAuth 로그인을 지원하는 Grok Build(xAI), Ollama 퍼스트 클래스 카드, Claude Opus 5 및 Sonnet 5, Kimi 공식 파트너십(Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… 그리고 새롭게 정비된 **384개 제공자 카탈로그**를 제공합니다. → [제공자](docs/reference/PROVIDER_REFERENCE.md) - **📡 라우팅 투명성** — 모든 응답에 해당 응답을 처리한 전략/제공자/지연 시간을 명시하는 `X-OmniRoute-Decision` 헤더가 포함됩니다. 새로운 `cache-optimized` 콤보 전략과 Auto-Combo `cacheAffinity` 요소는 반복 요청을 캐시된 접두사를 보유한 연결로 다시 라우팅하며, 읽기 전용 `/v1/auto-combo/{channel}/candidates` 엔드포인트는 `auto/*` 채널의 실시간 후보 풀을 노출합니다. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ 로컬 성능 및 인프라** — 원클릭 로컬 Redis, Cloudflare Workers / Deno Deploy 릴레이 배포 도구, 감독형 임베디드 서비스로 제공되는 Bifrost 및 Mux. → [임베디드 서비스](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 기본 포함 기능** — 플러그인 프레임워크 및 마켓플레이스, Omni/Agent/GitHub 스킬 프레임워크, Obsidian 볼트 통합(22개의 MCP 도구), OpenAI 호환 Batch 및 Files API, 시맨틱 응답 캐시, 리더보드가 포함된 게이미피케이션, ACP 에이전트 검색(기본 제공 에이전트 15개), BigQuery로의 예약된 로그 내보내기, `auto/chaos` 병렬 다중 모델 팬아웃, Telegram 봇 브리지, 앱 내 버전 관리자, LMArena-ELO 무료 제공자 순위. → [문서](docs/README.md) @@ -689,11 +689,11 @@ omniroute configure codex # 또한: claude opencode qwen aider goose ge
-## 🌐 383개 AI 제공업체 — 카탈로그에 표시된 무료 제공업체 153개 +## 🌐 384개 AI 제공업체 — 카탈로그에 표시된 무료 제공업체 153개
-> **등록된 제공업체 383개**가 표준 채팅, 미디어, 검색, 로컬, 클라우드 에이전트 및 시스템 컬렉션에 걸쳐 포함되어 있으며, 이 중 **153개는 `hasFree: true` 검색 메타데이터를 보유**합니다. 채팅 모델 레지스트리에는 **제공업체 239개 / 서로 다른 제공업체-모델 쌍 3,059개 / 원시 모델 ID 1,475개**가 포함되어 있으며, 별도의 무료 예산 카탈로그에는 **모델별 행 471개**, **반복 풀 34개**, **반복/키리스 영구 무료 제공업체 52개**가 있습니다. 이는 설계상 서로 다른 분모입니다. 정의와 풀 중복 제거 계산은 [제공업체 참조](docs/reference/PROVIDER_REFERENCE.md) 및 [무료 요금제](docs/reference/FREE_TIERS.md)에 정리되어 있습니다. +> **등록된 제공업체 384개**가 표준 채팅, 미디어, 검색, 로컬, 클라우드 에이전트 및 시스템 컬렉션에 걸쳐 포함되어 있으며, 이 중 **153개는 `hasFree: true` 검색 메타데이터를 보유**합니다. 채팅 모델 레지스트리에는 **제공업체 239개 / 서로 다른 제공업체-모델 쌍 3,059개 / 원시 모델 ID 1,475개**가 포함되어 있으며, 별도의 무료 예산 카탈로그에는 **모델별 행 471개**, **반복 풀 34개**, **반복/키리스 영구 무료 제공업체 52개**가 있습니다. 이는 설계상 서로 다른 분모입니다. 정의와 풀 중복 제거 계산은 [제공업체 참조](docs/reference/PROVIDER_REFERENCE.md) 및 [무료 요금제](docs/reference/FREE_TIERS.md)에 정리되어 있습니다.
@@ -1295,7 +1295,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1도 이를 건너 런타임Node.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 언어TypeScript 6.0 — src/ 및 open-sse/ 전체에서 100% TypeScript 사용 (v2.0 이후 코어에 any 없음) 프레임워크Next.js 16 + React 19 + Tailwind CSS 4 - 데이터베이스better-sqlite3 (SQLite, WAL 저널링) + LowDB (JSON 레거시) — 도메인 모듈 147개, 마이그레이션 230개 + 데이터베이스better-sqlite3 (SQLite, WAL 저널링) + LowDB (JSON 레거시) — 도메인 모듈 147개, 마이그레이션 232개 메모리SQLite FTS5 전문 검색 + int8 양자화 벡터 임베딩, 타입 지정 감쇠 스키마Zod 4 — MCP 도구 I/O 검증 + API 계약 프로토콜MCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ko/llm.txt b/docs/i18n/ko/llm.txt index 867f1b981de..cad5ad200b8 100644 --- a/docs/i18n/ko/llm.txt +++ b/docs/i18n/ko/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/lt/README.md b/docs/i18n/lt/README.md index b8f72b7524c..f228662dd9a 100644 --- a/docs/i18n/lt/README.md +++ b/docs/i18n/lt/README.md @@ -15,7 +15,7 @@ **Sukūrė [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Niekada nenustokite programuoti. Kiekvienas AI įrankis → 383 teikėjas — daugiau nei 150 nemokamų — per vieną galinį tašką. Claude Code, Codex, Cursor, Cline, Copilot ir Antigravity į NEMOKAMUS Claude / GPT / Gemini su automatiniu atsarginiu perjungimu. RTK + Caveman pakopinis glaudinimas sutaupo 15–95 % žetonų (~89 % vidutiniškai) — niekada nepasiekite limitų. 383 AI teikėjas · daugiau nei 150 nemokamų planų · ~1,62 mlrd. nemokamų žetonų per mėnesį · 20 maršrutizavimo strategijų · pradėkite už 0 USD. +OmniRoute — Niekada nenustokite programuoti. Kiekvienas AI įrankis → 384 teikėjas — daugiau nei 150 nemokamų — per vieną galinį tašką. Claude Code, Codex, Cursor, Cline, Copilot ir Antigravity į NEMOKAMUS Claude / GPT / Gemini su automatiniu atsarginiu perjungimu. RTK + Caveman pakopinis glaudinimas sutaupo 15–95 % žetonų (~89 % vidutiniškai) — niekada nepasiekite limitų. 384 AI teikėjas · daugiau nei 150 nemokamų planų · ~1,62 mlrd. nemokamų žetonų per mėnesį · 20 maršrutizavimo strategijų · pradėkite už 0 USD.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------------- | :---------------------------------: | :---------: | :-------------------: | -| 🌐 Teikėjai | 352 | **358** | 383 | +| 🌐 Teikėjai | 352 | **358** | 384 | | 🧠 Unikalūs pokalbių modelių ID | 1320 | **1374** | 1491 | | 🖼️ Modalumo tiltas | 🆕 vaizdas + garsas + vaizdo įrašas | ✓ | ✓ | | 📡 Nemokamas „Radar“ katalogas | 🆕 pasirenkamas | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funkcijos 🎯 Deriniai - 🌐 Teikėjai + 🌐 Teikėjai 🔌 CLI ir MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Pažadas — vienas galinis punktas ir 383 teikėjas. Automatinis atsarginis perjungimas užtikrina maršrutų parinkimą, kol pasiekiamas kitas veikiantis taikinys. Šeši pagrindiniai privalumai: atsparus atsarginis perjungimas tarp 383 teikėjo · iki 95 % žetonų sutaupymas tinkamiems darbo krūviams · pradėti galima už 0 USD, turint daugiau nei 150 nemokamų planų ir 52 pasikartojančius arba rakto nereikalaujančius nemokamus planus visam laikui · 39 CLI / agentų integracijos naudojant vieną konfigūraciją · OpenAI, Claude, Gemini ir Responses API suderinamumas adresu /v1 · gamybinės aplinkos valdikliai, įskaitant grandinės pertraukiklius, TLS slėpimą, MCP su 110 įrankių, A2A, atmintį, apsaugos priemones, vertinimus ir daugiau nei 56 000 statinių testų deklaracijų daugiau nei 7 800 sekamų testų failų. +Pažadas — vienas galinis punktas ir 384 teikėjas. Automatinis atsarginis perjungimas užtikrina maršrutų parinkimą, kol pasiekiamas kitas veikiantis taikinys. Šeši pagrindiniai privalumai: atsparus atsarginis perjungimas tarp 384 teikėjo · iki 95 % žetonų sutaupymas tinkamiems darbo krūviams · pradėti galima už 0 USD, turint daugiau nei 150 nemokamų planų ir 52 pasikartojančius arba rakto nereikalaujančius nemokamus planus visam laikui · 39 CLI / agentų integracijos naudojant vieną konfigūraciją · OpenAI, Claude, Gemini ir Responses API suderinamumas adresu /v1 · gamybinės aplinkos valdikliai, įskaitant grandinės pertraukiklius, TLS slėpimą, MCP su 110 įrankių, A2A, atmintį, apsaugos priemones, vertinimus ir daugiau nei 56 000 statinių testų deklaracijų daugiau nei 7 800 sekamų testų failų.

@@ -510,7 +510,7 @@ Visos **20** strategijų – derinkite jas kiekviename kombinacijos žingsnyje: -Kuo išsiskiria „OmniRoute“ — atnaujinta funkcijų momentinė apžvalga, kurioje pagal 13 galimybių lyginami 9router, OpenRouter, CLIProxyAPI ir LiteLLM. „OmniRoute“: 383 tiekėjas, daugiau nei 150 integruotų nemokamų planų, 20 maršrutizavimo strategijų, žetonų glaudinimas naudojant 12 variklių, integruotas MCP serveris su 110 įrankių, A2A agentų protokolas, nuolatinė atmintis, apsaugos taisyklės, debesijos agentai, TLS pirštų atspaudų slėpimas, Desktop/Termux/PWA ir 67 i18n naudotojo sąsajos lokalės. „OmniRoute“ licencijuota pagal MIT licenciją ir gali būti talpinama savarankiškai. Konkurentų galimybės ir skaičiai gali keistis; žr. susietą metodiką. +Kuo išsiskiria „OmniRoute“ — atnaujinta funkcijų momentinė apžvalga, kurioje pagal 13 galimybių lyginami 9router, OpenRouter, CLIProxyAPI ir LiteLLM. „OmniRoute“: 384 tiekėjas, daugiau nei 150 integruotų nemokamų planų, 20 maršrutizavimo strategijų, žetonų glaudinimas naudojant 12 variklių, integruotas MCP serveris su 110 įrankių, A2A agentų protokolas, nuolatinė atmintis, apsaugos taisyklės, debesijos agentai, TLS pirštų atspaudų slėpimas, Desktop/Termux/PWA ir 67 i18n naudotojo sąsajos lokalės. „OmniRoute“ licencijuota pagal MIT licenciją ir gali būti talpinama savarankiškai. Konkurentų galimybės ir skaičiai gali keistis; žr. susietą metodiką. 📊 Visa metodika ir išsami kiekvienos funkcijos informacija, palyginti su 9router, OpenRouter, CLIProxyAPI ir LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -613,7 +613,7 @@ dabartinį katalogą rasite **[radar.omniroute.online/planos](https://radar.omni - **🖼️ Nauji galiniai taškai** — `/v1/ocr` (Mistral OCR) ir `/v1/audio/translations` (Whisper tipo) užbaigia medijos funkcijų rinkinį. → [API nuoroda](docs/reference/API_REFERENCE.md) - **🎨 Vaizdų / vaizdo įrašų / garso generavimas** — viena medijos API: xAI Grok Imagine ir Novita AI vaizdo įrašai, ComfyUI, Magnific, Adobe Firefly, Segmind bei kalbos teikėjai, tokie kaip ElevenLabs. → [API nuoroda](docs/reference/API_REFERENCE.md) - **🌍 Diegimas ir operacijos** — atvirkštinio tarpinio serverio `basePath`, automatinis naršyklės kalbos aptikimas, įrenginių sekimas pagal raktą, MITM pasitikėjimas be root teisių, zh-TW lokalizacija. → [Aplinka](docs/reference/ENVIRONMENT.md) -- **🤝 Daugiau teikėjų ir agentų** — debesijos agentai (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) su naršyklės ir OAuth prisijungimu, visavertė Ollama kortelė, Claude Opus 5 ir Sonnet 5, oficiali Kimi partnerystė (Code / Web / Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… ir atnaujintas **383 teikėjo katalogas**. → [Teikėjai](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Daugiau teikėjų ir agentų** — debesijos agentai (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) su naršyklės ir OAuth prisijungimu, visavertė Ollama kortelė, Claude Opus 5 ir Sonnet 5, oficiali Kimi partnerystė (Code / Web / Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… ir atnaujintas **384 teikėjo katalogas**. → [Teikėjai](docs/reference/PROVIDER_REFERENCE.md) - **📡 Maršrutizavimo skaidrumas** — kiekviename atsakyme pateikiama `X-OmniRoute-Decision` antraštė, nurodanti strategiją, teikėją ir delsą, aptarnavusius užklausą; nauja `cache-optimized` derinio strategija ir Auto-Combo `cacheAffinity` faktorius nukreipia pasikartojančias užklausas atgal į ryšį, kuriame laikoma podėlyje esanti priešdėlio dalis; tik skaitymui skirtas `/v1/auto-combo/{channel}/candidates` galinis taškas atskleidžia tiesioginį `auto/*` kanalo kandidatų telkinį. → [Automatiniai deriniai](docs/routing/AUTO-COMBO.md). - **⚡ Vietinis našumas ir infrastruktūra** — vienu spustelėjimu paleidžiamas vietinis Redis, „Cloudflare Workers“ / „Deno Deploy“ perdavimo serverių diegikliai, „Bifrost“ ir „Mux“ kaip prižiūrimosios įterptosios paslaugos. → [Įterptosios paslaugos](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Taip pat įtraukta** — įskiepių sistema ir prekyvietė, „Omni“ / „Agent“ / „GitHub“ įgūdžių sistemos, „Obsidian“ saugyklos integracija (22 MCP įrankiai), su „OpenAI“ suderinamos „Batch“ ir „Files“ API, semantinė atsakymų podėlio sistema, žaidybinimas su lyderių lentelėmis, ACP agentų aptikimas (15 integruotų agentų), suplanuotas žurnalų eksportavimas į „BigQuery“, `auto/chaos` lygiagretus kelių modelių paskirstymas, „Telegram“ roboto sąsaja, programoje integruota versijų tvarkyklė ir nemokamų teikėjų reitingai pagal LMArena-ELO. → [Dokumentacija](docs/README.md) @@ -697,11 +697,11 @@ slaptus duomenis nuo patekimo į jūsų shell istoriją. → [CLI integracijos](
-## 🌐 383 AI teikėjai — 153 kataloge pažymėti kaip nemokami +## 🌐 384 AI teikėjai — 153 kataloge pažymėti kaip nemokami
-> **383 registruotas teikėjas** kanoninėse pokalbių, medijos, paieškos, vietinėse, debesijos agentų ir sistemos kolekcijose, įskaitant **153 teikėjus, turinčius `hasFree: true` aptikimo metaduomenis**. Pokalbių modelių registre yra **239 teikėjai / 3 059 skirtingos teikėjo ir modelio poros / 1 475 neapdoroti modelių ID**; atskirame nemokamų planų kataloge yra **471 eilutė pagal modelį**, **34 pasikartojantys fondai** ir **52 pasikartojantys arba be rakto visam laikui nemokami teikėjai**. Tai skirtingi vardikliai pagal paskirtį; apibrėžimai ir pagal fondus deduplikuoti skaičiavimai pateikiami [Teikėjų žinyne](docs/reference/PROVIDER_REFERENCE.md) ir [Nemokamuose planuose](docs/reference/FREE_TIERS.md). +> **384 registruotas teikėjas** kanoninėse pokalbių, medijos, paieškos, vietinėse, debesijos agentų ir sistemos kolekcijose, įskaitant **153 teikėjus, turinčius `hasFree: true` aptikimo metaduomenis**. Pokalbių modelių registre yra **239 teikėjai / 3 059 skirtingos teikėjo ir modelio poros / 1 475 neapdoroti modelių ID**; atskirame nemokamų planų kataloge yra **471 eilutė pagal modelį**, **34 pasikartojantys fondai** ir **52 pasikartojantys arba be rakto visam laikui nemokami teikėjai**. Tai skirtingi vardikliai pagal paskirtį; apibrėžimai ir pagal fondus deduplikuoti skaičiavimai pateikiami [Teikėjų žinyne](docs/reference/PROVIDER_REFERENCE.md) ir [Nemokamuose planuose](docs/reference/FREE_TIERS.md).
@@ -1306,7 +1306,7 @@ Kanoninės metrikos 2026-10-18: **1.976 unikalūs vaizdo įrašai** · **20.220. Vykdymo aplinkaNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 KalbaTypeScript 6.0 — 100% TypeScript visame src/ ir open-sse/ (nuo v2.0 pagrindiniame kode nėra any) KarkasasNext.js 16 + React 19 + Tailwind CSS 4 - Duomenų bazėbetter-sqlite3 (SQLite, WAL žurnalinimas) + LowDB (JSON senoji sistema) — 147 domeno modulių, 230 migracijos + Duomenų bazėbetter-sqlite3 (SQLite, WAL žurnalinimas) + LowDB (JSON senoji sistema) — 147 domeno modulių, 232 migracijos AtmintisSQLite FTS5 viso teksto paieška + int8 kvantuoti vektoriniai įterpiniai, tipizuotas nykimas SchemosZod 4 — MCP įrankių įvesties / išvesties tikrinimas + API sutartys ProtokolaiMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/lt/llm.txt b/docs/i18n/lt/llm.txt index df186a205eb..8313135b2a3 100644 --- a/docs/i18n/lt/llm.txt +++ b/docs/i18n/lt/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/lv/README.md b/docs/i18n/lv/README.md index ad8150da021..4857f9b4a30 100644 --- a/docs/i18n/lv/README.md +++ b/docs/i18n/lv/README.md @@ -15,7 +15,7 @@ **Radījis [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Nekad nepārtrauciet programmēt. Katrs MI rīks → 383 nodrošinātājs — 150+ bezmaksas — izmantojot vienu galapunktu. Claude Code, Codex, Cursor, Cline, Copilot un Antigravity uz BEZMAKSAS Claude / GPT / Gemini ar automātisku atkāpšanos. RTK + Caveman kombinētā saspiešana ietaupa 15–95% marķieru (~89% vidēji) — nekad nesasniedziet limitus. 383 MI nodrošinātājs · 150+ bezmaksas līmeņi · ~1,62 miljardi bezmaksas marķieru mēnesī · 20 maršrutēšanas stratēģijas · Sāciet par $0. +OmniRoute — Nekad nepārtrauciet programmēt. Katrs MI rīks → 384 nodrošinātājs — 150+ bezmaksas — izmantojot vienu galapunktu. Claude Code, Codex, Cursor, Cline, Copilot un Antigravity uz BEZMAKSAS Claude / GPT / Gemini ar automātisku atkāpšanos. RTK + Caveman kombinētā saspiešana ietaupa 15–95% marķieru (~89% vidēji) — nekad nesasniedziet limitus. 384 MI nodrošinātājs · 150+ bezmaksas līmeņi · ~1,62 miljardi bezmaksas marķieru mēnesī · 20 maršrutēšanas stratēģijas · Sāciet par $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------- | :----------------------: | :---------: | :-------------------: | -| 🌐 Nodrošinātāji | 352 | **358** | 383 | +| 🌐 Nodrošinātāji | 352 | **358** | 384 | | 🧠 Unikāli tērzēšanas modeļu ID | 1320 | **1374** | 1491 | | 🖼️ Modalitāšu tilts | 🆕 redze + audio + video | ✓ | ✓ | | 📡 Radar bezmaksas katalogs | 🆕 pēc izvēles | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funkcijas 🎯 Kombinācijas - 🌐 Nodrošinātāji + 🌐 Nodrošinātāji 🔌 CLI un MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Solījums — viens galapunkts un 383 pakalpojumu sniedzējs. Automātiskā atkāpšanās nodrošina maršrutēšanas turpināšanu, kamēr ir pieejams cits darbspējīgs mērķis. Seši pīlāri: noturīga atkāpšanās starp 383 pakalpojumu sniedzēju · līdz 95% žetonu ietaupījums piemērotām darba slodzēm · $0, lai sāktu, ar vairāk nekā 150 bezmaksas līmeņiem un 52 atkārtoti pieejamiem vai bezatslēgas bezmaksas pakalpojumu sniedzējiem uz visiem laikiem · 39 CLI/aģentu integrācijas, izmantojot vienu konfigurāciju · OpenAI, Claude, Gemini un Responses API saderība ar /v1 · ražošanas vadīklas, tostarp ķēdes pārtraucēji, TLS slepenais režīms, MCP 110 rīki, A2A, atmiņa, aizsargmehānismi, novērtējumi un vairāk nekā 56 000 statisku testu deklarāciju vairāk nekā 7800 izsekotos testu failos. +Solījums — viens galapunkts un 384 pakalpojumu sniedzējs. Automātiskā atkāpšanās nodrošina maršrutēšanas turpināšanu, kamēr ir pieejams cits darbspējīgs mērķis. Seši pīlāri: noturīga atkāpšanās starp 384 pakalpojumu sniedzēju · līdz 95% žetonu ietaupījums piemērotām darba slodzēm · $0, lai sāktu, ar vairāk nekā 150 bezmaksas līmeņiem un 52 atkārtoti pieejamiem vai bezatslēgas bezmaksas pakalpojumu sniedzējiem uz visiem laikiem · 39 CLI/aģentu integrācijas, izmantojot vienu konfigurāciju · OpenAI, Claude, Gemini un Responses API saderība ar /v1 · ražošanas vadīklas, tostarp ķēdes pārtraucēji, TLS slepenais režīms, MCP 110 rīki, A2A, atmiņa, aizsargmehānismi, novērtējumi un vairāk nekā 56 000 statisku testu deklarāciju vairāk nekā 7800 izsekotos testu failos.

@@ -510,7 +510,7 @@ Visas **20** stratēģijas — kombinējiet tās katrā kombinācijas solī: -Ar ko OmniRoute atšķiras — aktuāls funkciju salīdzinājums ar 9router, OpenRouter, CLIProxyAPI un LiteLLM 13 iespēju kategorijās. OmniRoute: 383 nodrošinātājs, vairāk nekā 150 iebūvēti bezmaksas līmeņi, 20 maršrutēšanas stratēģijas, marķieru saspiešana ar 12 dzinējiem, iebūvēts MCP serveris ar 110 rīkiem, A2A aģentu protokols, pastāvīgā atmiņa, drošības ierobežojumi, mākoņa aģenti, TLS pirkstu nospiedumu slēpšana, Desktop/Termux/PWA un 67 i18n saskarnes lokalizācijas. OmniRoute ir licencēts saskaņā ar MIT licenci, un to var mitināt patstāvīgi. Konkurentu iespējas un skaits var mainīties; skatiet saiti uz metodoloģiju. +Ar ko OmniRoute atšķiras — aktuāls funkciju salīdzinājums ar 9router, OpenRouter, CLIProxyAPI un LiteLLM 13 iespēju kategorijās. OmniRoute: 384 nodrošinātājs, vairāk nekā 150 iebūvēti bezmaksas līmeņi, 20 maršrutēšanas stratēģijas, marķieru saspiešana ar 12 dzinējiem, iebūvēts MCP serveris ar 110 rīkiem, A2A aģentu protokols, pastāvīgā atmiņa, drošības ierobežojumi, mākoņa aģenti, TLS pirkstu nospiedumu slēpšana, Desktop/Termux/PWA un 67 i18n saskarnes lokalizācijas. OmniRoute ir licencēts saskaņā ar MIT licenci, un to var mitināt patstāvīgi. Konkurentu iespējas un skaits var mainīties; skatiet saiti uz metodoloģiju. 📊 Pilna metodoloģija un detalizēts salīdzinājums par katru funkciju ar 9router, OpenRouter, CLIProxyAPI un LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar ir jāiespējo brīvprātīgi, un tas izmanto tikai GET pieprasījumus. Om - **🖼️ Jauni galapunkti** — `/v1/ocr` (Mistral OCR) un `/v1/audio/translations` (Whisper stila) papildina multivides iespēju klāstu. → [API atsauce](docs/reference/API_REFERENCE.md) - **🎨 Attēlu / video / audio ģenerēšana** — viena API multividei: xAI Grok Imagine un Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind un runas pakalpojumu sniedzēji, piemēram, ElevenLabs. → [API atsauce](docs/reference/API_REFERENCE.md) - **🌍 Izvietošana un darbība** — reversā starpniekservera `basePath`, automātiska pārlūkprogrammas valodas noteikšana, ierīču uzskaite katrai atslēgai, MITM uzticamība bez root tiesībām, zh-TW lokalizācija. → [Vide](docs/reference/ENVIRONMENT.md) -- **🤝 Vairāk pakalpojumu sniedzēju un aģentu** — mākoņa aģenti (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) ar pārlūkprogrammas un OAuth pieteikšanos, pirmās klases Ollama karte, Claude Opus 5 un Sonnet 5, oficiāla Kimi partnerība (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… un atjaunināts **383 pakalpojumu sniedzēja katalogs**. → [Pakalpojumu sniedzēji](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Vairāk pakalpojumu sniedzēju un aģentu** — mākoņa aģenti (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) ar pārlūkprogrammas un OAuth pieteikšanos, pirmās klases Ollama karte, Claude Opus 5 un Sonnet 5, oficiāla Kimi partnerība (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… un atjaunināts **384 pakalpojumu sniedzēja katalogs**. → [Pakalpojumu sniedzēji](docs/reference/PROVIDER_REFERENCE.md) - **📡 Maršrutēšanas pārredzamība** — katra atbilde ietver `X-OmniRoute-Decision` galveni, kas norāda to apkalpojušo stratēģiju/pakalpojumu sniedzēju/aizturi; jaunā `cache-optimized` kombinētā stratēģija un Auto-Combo `cacheAffinity` faktora maršruts atkārtotus pieprasījumus novirza atpakaļ uz savienojumu, kurā atrodas kešotais prefikss; tikai lasāms `/v1/auto-combo/{channel}/candidates` galapunkts atklāj `auto/*` kanāla aktuālo kandidātu kopu. → [Automātiskās kombinācijas](docs/routing/AUTO-COMBO.md) - **⚡ Vietējā veiktspēja un infrastruktūra** — lokālais Redis ar vienu klikšķi, Cloudflare Workers / Deno Deploy releju izvietotāji, Bifrost un Mux kā uzraudzīti iegulti pakalpojumi. → [Iegultie pakalpojumi](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Arī komplektā** — spraudņu ietvars un tirgus, Omni/Agent/GitHub prasmju ietvari, Obsidian glabātuves integrācija (22 MCP rīki), ar OpenAI saderīgas Batch un Files API, semantiskā atbilžu kešatmiņa, spēļošana ar līderu tabulām, ACP aģentu atklāšana (15 iebūvēti aģenti), plānota žurnālu eksportēšana uz BigQuery, `auto/chaos` paralēla vairāku modeļu izkliedēšana, Telegram robota tilts, lietotnē iebūvēts versiju pārvaldnieks un LMArena-ELO bezmaksas pakalpojumu sniedzēju reitingi. → [Dokumentācija](docs/README.md) @@ -688,11 +688,11 @@ parāda precīzu vidi/argumentus pirms izpildes, un `--api-key-env NAME` pasarg
-## 🌐 383 AI nodrošinātāji — 153 katalogā atzīmēti kā bezmaksas +## 🌐 384 AI nodrošinātāji — 153 katalogā atzīmēti kā bezmaksas
-> **383 reģistrēts nodrošinātājs** kanoniskajās tērzēšanas, multivides, meklēšanas, lokālajās, mākoņa aģentu un sistēmu kolekcijās, tostarp **153 ar `hasFree: true` atklāšanas metadatiem**. Tērzēšanas modeļu reģistrs aptver **239 nodrošinātājus / 3 059 atšķirīgus nodrošinātāja–modeļa pārus / 1 475 neapstrādātus modeļu ID**; atsevišķajā bezmaksas budžeta katalogā ir **471 rinda katram modelim**, **34 periodiski resursu pūli** un **52 periodiski / bez atslēgas pieejami nodrošinātāji, kas vienmēr ir bezmaksas**. Šie saucēji pēc būtības atšķiras; definīcijas un aprēķini, kuros pūli tiek uzskaitīti tikai vienreiz, ir pieejami sadaļās [Nodrošinātāju atsauce](docs/reference/PROVIDER_REFERENCE.md) un [Bezmaksas līmeņi](docs/reference/FREE_TIERS.md). +> **384 reģistrēts nodrošinātājs** kanoniskajās tērzēšanas, multivides, meklēšanas, lokālajās, mākoņa aģentu un sistēmu kolekcijās, tostarp **153 ar `hasFree: true` atklāšanas metadatiem**. Tērzēšanas modeļu reģistrs aptver **239 nodrošinātājus / 3 059 atšķirīgus nodrošinātāja–modeļa pārus / 1 475 neapstrādātus modeļu ID**; atsevišķajā bezmaksas budžeta katalogā ir **471 rinda katram modelim**, **34 periodiski resursu pūli** un **52 periodiski / bez atslēgas pieejami nodrošinātāji, kas vienmēr ir bezmaksas**. Šie saucēji pēc būtības atšķiras; definīcijas un aprēķini, kuros pūli tiek uzskaitīti tikai vienreiz, ir pieejami sadaļās [Nodrošinātāju atsauce](docs/reference/PROVIDER_REFERENCE.md) un [Bezmaksas līmeņi](docs/reference/FREE_TIERS.md).
@@ -1294,7 +1294,7 @@ Kanoniskie rādītāji 2026-10-18: **1.976 unikāli videoklipi** · **20.220.289 IzpildlaiksNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ValodaTypeScript 6.0 — 100% TypeScript visā src/ un open-sse/ (kopš v2.0 kodolā nav neviena any) ietvarsNext.js 16 + React 19 + Tailwind CSS 4 - Datu bāzebetter-sqlite3 (SQLite, WAL žurnalēšana) + LowDB (mantotais JSON) — 147 domēna moduļi, 230 migrācijas + Datu bāzebetter-sqlite3 (SQLite, WAL žurnalēšana) + LowDB (mantotais JSON) — 147 domēna moduļi, 232 migrācijas AtmiņaSQLite FTS5 pilnteksta meklēšana + int8 kvantizēti vektoru iegulumi, tipizēta novecošana ShēmasZod 4 — MCP rīku I/O validācija + API līgumi ProtokoliMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/lv/llm.txt b/docs/i18n/lv/llm.txt index 253be98e1b2..c5a86ecdc50 100644 --- a/docs/i18n/lv/llm.txt +++ b/docs/i18n/lv/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ml/README.md b/docs/i18n/ml/README.md index 9942a8ca623..30f5927db83 100644 --- a/docs/i18n/ml/README.md +++ b/docs/i18n/ml/README.md @@ -15,7 +15,7 @@ **രൂപകൽപ്പന ചെയ്തത് [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — ഒരിക്കലും കോഡിംഗ് നിർത്തരുത്. ഓരോ AI ടൂളും → 383 പ്രൊവൈഡർമാർ — 150+ സൗജന്യമായി — ഒരു എൻഡ്പോയിന്റിലൂടെ. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity എന്നിവയെ FREE Claude / GPT / Gemini-യിലേക്ക് ഓട്ടോ-ഫോൾബാക്കോടെ ബന്ധിപ്പിക്കുന്നു. RTK + Caveman സ്റ്റാക്ക് ചെയ്ത കംപ്രഷൻ 15–95% ടോക്കണുകൾ (~89% ശരാശരി) ലാഭിക്കുന്നു — പരിധികളിൽ ഒരിക്കലും എത്തില്ല. 383 AI പ്രൊവൈഡർമാർ · 150+ സൗജന്യ ടയറുകൾ · പ്രതിമാസം ~1.62B സൗജന്യ ടോക്കണുകൾ · 20 റൂട്ടിംഗ് തന്ത്രങ്ങൾ · ആരംഭിക്കാൻ $0. +OmniRoute — ഒരിക്കലും കോഡിംഗ് നിർത്തരുത്. ഓരോ AI ടൂളും → 384 പ്രൊവൈഡർമാർ — 150+ സൗജന്യമായി — ഒരു എൻഡ്പോയിന്റിലൂടെ. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity എന്നിവയെ FREE Claude / GPT / Gemini-യിലേക്ക് ഓട്ടോ-ഫോൾബാക്കോടെ ബന്ധിപ്പിക്കുന്നു. RTK + Caveman സ്റ്റാക്ക് ചെയ്ത കംപ്രഷൻ 15–95% ടോക്കണുകൾ (~89% ശരാശരി) ലാഭിക്കുന്നു — പരിധികളിൽ ഒരിക്കലും എത്തില്ല. 384 AI പ്രൊവൈഡർമാർ · 150+ സൗജന്യ ടയറുകൾ · പ്രതിമാസം ~1.62B സൗജന്യ ടോക്കണുകൾ · 20 റൂട്ടിംഗ് തന്ത്രങ്ങൾ · ആരംഭിക്കാൻ $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------------ | :----------------------: | :---------: | :-------------------: | -| 🌐 ദാതാക്കൾ | 352 | **358** | 383 | +| 🌐 ദാതാക്കൾ | 352 | **358** | 384 | | 🧠 തനതായ ചാറ്റ് മോഡൽ ID-കൾ | 1320 | **1374** | 1491 | | 🖼️ മോഡാലിറ്റി ബ്രിഡ്ജ് | 🆕 വിഷൻ + ഓഡിയോ + വീഡിയോ | ✓ | ✓ | | 📡 റഡാർ സൗജന്യ കാറ്റലോഗ് | 🆕 opt-in | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ സവിശേഷതകൾ 🎯 കോംബോകൾ - 🌐 പ്രൊവൈഡറുകൾ + 🌐 പ്രൊവൈഡറുകൾ 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -വാഗ്ദാനം — ഒരു എൻഡ്പോയിന്റും 383 പ്രൊവൈഡർമാരും. മറ്റൊരു ആരോഗ്യകരമായ ടാർഗെറ്റ് ലഭ്യമായിരിക്കുമ്പോൾ ഓട്ടോമാറ്റിക് ഫാൾബാക്ക് റൂട്ടിംഗ് തുടരാൻ സഹായിക്കുന്നു. ആറ് തൂണുകൾ: 383 പ്രൊവൈഡർമാരിലുടനീളം പ്രതിരോധശേഷിയുള്ള ഫാൾബാക്ക് · യോഗ്യമായ വർക്ക്ലോഡുകളിൽ 95% വരെ ടോക്കൺ ലാഭം · 150-ലധികം സൗജന്യ ടിയറുകളും ആവർത്തിച്ചുള്ള/കീ ഇല്ലാത്ത, എന്നേക്കുമായി സൗജന്യമായ 52 പ്രൊവൈഡർമാരുമായി $0 ചെലവിൽ ആരംഭിക്കാം · ഒരു കോൺഫിഗറേഷനിലൂടെ 39 CLI/ഏജന്റ് ഇന്റഗ്രേഷനുകൾ · /v1-ൽ OpenAI, Claude, Gemini, Responses API അനുയോജ്യത · സർക്യൂട്ട് ബ്രേക്കറുകൾ, TLS സ്റ്റെൽത്ത്, 110 MCP ടൂളുകൾ, A2A, മെമ്മറി, ഗാർഡ്റെയിലുകൾ, ഇവാലുകൾ, കൂടാതെ 7,800-ലധികം ട്രാക്ക് ചെയ്ത ടെസ്റ്റ് ഫയലുകളിലായി 56,000-ലധികം സ്റ്റാറ്റിക് ടെസ്റ്റ് ഡിക്ലറേഷനുകൾ ഉൾപ്പെടെയുള്ള പ്രൊഡക്ഷൻ നിയന്ത്രണങ്ങൾ. +വാഗ്ദാനം — ഒരു എൻഡ്പോയിന്റും 384 പ്രൊവൈഡർമാരും. മറ്റൊരു ആരോഗ്യകരമായ ടാർഗെറ്റ് ലഭ്യമായിരിക്കുമ്പോൾ ഓട്ടോമാറ്റിക് ഫാൾബാക്ക് റൂട്ടിംഗ് തുടരാൻ സഹായിക്കുന്നു. ആറ് തൂണുകൾ: 384 പ്രൊവൈഡർമാരിലുടനീളം പ്രതിരോധശേഷിയുള്ള ഫാൾബാക്ക് · യോഗ്യമായ വർക്ക്ലോഡുകളിൽ 95% വരെ ടോക്കൺ ലാഭം · 150-ലധികം സൗജന്യ ടിയറുകളും ആവർത്തിച്ചുള്ള/കീ ഇല്ലാത്ത, എന്നേക്കുമായി സൗജന്യമായ 52 പ്രൊവൈഡർമാരുമായി $0 ചെലവിൽ ആരംഭിക്കാം · ഒരു കോൺഫിഗറേഷനിലൂടെ 39 CLI/ഏജന്റ് ഇന്റഗ്രേഷനുകൾ · /v1-ൽ OpenAI, Claude, Gemini, Responses API അനുയോജ്യത · സർക്യൂട്ട് ബ്രേക്കറുകൾ, TLS സ്റ്റെൽത്ത്, 110 MCP ടൂളുകൾ, A2A, മെമ്മറി, ഗാർഡ്റെയിലുകൾ, ഇവാലുകൾ, കൂടാതെ 7,800-ലധികം ട്രാക്ക് ചെയ്ത ടെസ്റ്റ് ഫയലുകളിലായി 56,000-ലധികം സ്റ്റാറ്റിക് ടെസ്റ്റ് ഡിക്ലറേഷനുകൾ ഉൾപ്പെടെയുള്ള പ്രൊഡക്ഷൻ നിയന്ത്രണങ്ങൾ.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRoute-നെ വേറിട്ടുനിർത്തുന്ന കാര്യങ്ങൾ — 13 കഴിവുകളിലായി 9router, OpenRouter, CLIProxyAPI, LiteLLM എന്നിവയുമായുള്ള തീയതിയോടുകൂടിയ ഫീച്ചർ സ്നാപ്ഷോട്ട്. OmniRoute: 383 പ്രൊവൈഡർമാർ, ഇതിനകം ഉൾപ്പെടുത്തിയ 150-ലധികം സൗജന്യ ടിയറുകൾ, 20 റൂട്ടിംഗ് തന്ത്രങ്ങൾ, 12-എഞ്ചിൻ ടോക്കൺ കംപ്രഷൻ, 110 ടൂളുകളുള്ള ബിൽറ്റ്-ഇൻ MCP സെർവർ, A2A ഏജന്റ് പ്രോട്ടോക്കോൾ, സ്ഥിരതയുള്ള മെമ്മറി, ഗാർഡ്റെയിലുകൾ, ക്ലൗഡ് ഏജന്റുകൾ, TLS ഫിംഗർപ്രിന്റ് സ്റ്റെൽത്ത്, Desktop/Termux/PWA, 67 i18n UI ലോക്കേലുകൾ. OmniRoute MIT-ലൈസൻസുള്ളതും സ്വയം ഹോസ്റ്റ് ചെയ്യാനാകുന്നതുമാണ്. മത്സരികളിലെ കഴിവുകളും എണ്ണങ്ങളും മാറിയേക്കാം; ലിങ്ക് ചെയ്ത മെത്തഡോളജി കാണുക. +OmniRoute-നെ വേറിട്ടുനിർത്തുന്ന കാര്യങ്ങൾ — 13 കഴിവുകളിലായി 9router, OpenRouter, CLIProxyAPI, LiteLLM എന്നിവയുമായുള്ള തീയതിയോടുകൂടിയ ഫീച്ചർ സ്നാപ്ഷോട്ട്. OmniRoute: 384 പ്രൊവൈഡർമാർ, ഇതിനകം ഉൾപ്പെടുത്തിയ 150-ലധികം സൗജന്യ ടിയറുകൾ, 20 റൂട്ടിംഗ് തന്ത്രങ്ങൾ, 12-എഞ്ചിൻ ടോക്കൺ കംപ്രഷൻ, 110 ടൂളുകളുള്ള ബിൽറ്റ്-ഇൻ MCP സെർവർ, A2A ഏജന്റ് പ്രോട്ടോക്കോൾ, സ്ഥിരതയുള്ള മെമ്മറി, ഗാർഡ്റെയിലുകൾ, ക്ലൗഡ് ഏജന്റുകൾ, TLS ഫിംഗർപ്രിന്റ് സ്റ്റെൽത്ത്, Desktop/Termux/PWA, 67 i18n UI ലോക്കേലുകൾ. OmniRoute MIT-ലൈസൻസുള്ളതും സ്വയം ഹോസ്റ്റ് ചെയ്യാനാകുന്നതുമാണ്. മത്സരികളിലെ കഴിവുകളും എണ്ണങ്ങളും മാറിയേക്കാം; ലിങ്ക് ചെയ്ത മെത്തഡോളജി കാണുക. 📊 9router, OpenRouter, CLIProxyAPI, LiteLLM എന്നിവയുമായുള്ള പൂർണ്ണ മെത്തഡോളജിയും ഓരോ ഫീച്ചറിന്റെയും വിശദാംശങ്ങളും → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar തിരഞ്ഞെടുക്കുന്നവർക്ക് മാ - **🖼️ പുതിയ endpoints** — `/v1/ocr` (Mistral OCR), `/v1/audio/translations` (Whisper-style) എന്നിവ media surface പൂർണ്ണമാക്കുന്നു. → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 Image / video / audio generation** — media-ക്കായി ഒരൊറ്റ API: xAI Grok Imagine & Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind, ElevenLabs പോലുള്ള speech providers എന്നിവ. → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 Deployment & ops** — reverse-proxy `basePath`, browser-language auto-detect, ഓരോ key-ക്കും device tracking, root-less MITM trust, zh-TW localization. → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 കൂടുതൽ providers & agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login ഉള്ള Grok Build (xAI), first-class card ഉള്ള Ollama, Claude Opus 5 & Sonnet 5, Kimi official partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… കൂടാതെ പുതുക്കിയ **383-provider catalog**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 കൂടുതൽ providers & agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login ഉള്ള Grok Build (xAI), first-class card ഉള്ള Ollama, Claude Opus 5 & Sonnet 5, Kimi official partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… കൂടാതെ പുതുക്കിയ **384-provider catalog**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 Routing transparency** — ഓരോ response-ലും അത് നൽകിയ strategy/provider/latency വ്യക്തമാക്കുന്ന `X-OmniRoute-Decision` header ഉണ്ടായിരിക്കും; പുതിയ `cache-optimized` combo strategy-യും Auto-Combo `cacheAffinity` factor-ഉം ആവർത്തിച്ചുള്ള requests cached prefix കൈവശമുള്ള connection-ലേക്ക് തിരിച്ചുവിടുന്നു; read-only `/v1/auto-combo/{channel}/candidates` endpoint ഒരു `auto/*` channel-ന്റെ live candidate pool പുറത്തുവിടുന്നു. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ പ്രാദേശിക പ്രകടനവും ഇൻഫ്രയും** — ഒറ്റ ക്ലിക്കിൽ പ്രാദേശിക Redis, Cloudflare Workers / Deno Deploy റിലേ ഡിപ്ലോയർമാർ, മേൽനോട്ടത്തിലുള്ള എംബെഡഡ് സേവനങ്ങളായി Bifrost & Mux. → [എംബെഡഡ് സേവനങ്ങൾ](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 പാക്കേജിനുള്ളിൽ തന്നെ ലഭ്യമായവ** — പ്ലഗിൻ ഫ്രെയിംവർക്ക് + മാർക്കറ്റ്പ്ലേസ്, Omni/Agent/GitHub സ്കിൽസ് ഫ്രെയിംവർക്കുകൾ, Obsidian വോൾട്ട് ഇന്റഗ്രേഷൻ (22 MCP ടൂളുകൾ), OpenAI-യ്ക്ക് അനുയോജ്യമായ Batch & Files APIs, സെമാന്റിക് റെസ്പോൺസ് കാഷെ, ലീഡർബോർഡുകളോടെയുള്ള ഗെയിമിഫിക്കേഷൻ, ACP ഏജന്റ് കണ്ടെത്തൽ (15 ബിൽറ്റ്-ഇൻ ഏജന്റുകൾ), BigQuery-യിലേക്കുള്ള ഷെഡ്യൂൾ ചെയ്ത ലോഗ് എക്സ്പോർട്ട്, `auto/chaos` സമാന്തര മൾട്ടി-മോഡൽ ഫാൻ-ഔട്ട്, Telegram ബോട്ട് ബ്രിഡ്ജ്, ഇൻ-ആപ്പ് വേർഷൻ മാനേജർ, LMArena-ELO സൗജന്യ-പ്രൊവൈഡർ റാങ്കിങ്ങുകൾ. → [ഡോക്യുമെന്റേഷൻ](docs/README.md) @@ -688,11 +688,11 @@ omniroute configure codex # ഇതിലും: claude opencode qwen ai
-## 🌐 383 AI ദാതാക്കൾ — കാറ്റലോഗിൽ അടയാളപ്പെടുത്തിയ സൗജന്യമായവ 153 +## 🌐 384 AI ദാതാക്കൾ — കാറ്റലോഗിൽ അടയാളപ്പെടുത്തിയ സൗജന്യമായവ 153
-> **383 രജിസ്റ്റർ ചെയ്ത ദാതാക്കൾ** കാനോണിക്കൽ ചാറ്റ്, മീഡിയ, തിരയൽ, ലോക്കൽ, ക്ലൗഡ്-ഏജന്റ്, സിസ്റ്റം ശേഖരങ്ങളിലായി ഉണ്ട്. ഇതിൽ **153 എണ്ണത്തിൽ `hasFree: true` കണ്ടെത്തൽ മെറ്റാഡാറ്റ** ഉൾപ്പെടുന്നു. ചാറ്റ് മോഡൽ രജിസ്ട്രിയിൽ **239 ദാതാക്കൾ / 3,059 വ്യത്യസ്ത ദാതാവ്-മോഡൽ ജോഡികൾ / 1,475 റോ മോഡൽ ID-കൾ** ഉൾപ്പെടുന്നു; വേറിട്ട സൗജന്യ-ബജറ്റ് കാറ്റലോഗിൽ ഓരോ മോഡലിനുമുള്ള **471 വരികൾ**, **34 ആവർത്തിക്കുന്ന പൂളുകൾ**, **52 ആവർത്തിക്കുന്നതും കീ ആവശ്യമില്ലാത്തതുമായ എന്നേക്കും സൗജന്യ ദാതാക്കൾ** എന്നിവയുണ്ട്. ഇവ രൂപകൽപ്പനപ്രകാരം വ്യത്യസ്ത കണക്കാധാരങ്ങളാണ്; നിർവചനങ്ങളും പൂൾ-ഡ്യൂപ്ലിക്കേറ്റ് നീക്കം ചെയ്ത കണക്കുകൂട്ടലുകളും [ദാതാവ് റഫറൻസ്](docs/reference/PROVIDER_REFERENCE.md), [സൗജന്യ ശ്രേണികൾ](docs/reference/FREE_TIERS.md) എന്നിവയിൽ ലഭ്യമാണ്. +> **384 രജിസ്റ്റർ ചെയ്ത ദാതാക്കൾ** കാനോണിക്കൽ ചാറ്റ്, മീഡിയ, തിരയൽ, ലോക്കൽ, ക്ലൗഡ്-ഏജന്റ്, സിസ്റ്റം ശേഖരങ്ങളിലായി ഉണ്ട്. ഇതിൽ **153 എണ്ണത്തിൽ `hasFree: true` കണ്ടെത്തൽ മെറ്റാഡാറ്റ** ഉൾപ്പെടുന്നു. ചാറ്റ് മോഡൽ രജിസ്ട്രിയിൽ **239 ദാതാക്കൾ / 3,059 വ്യത്യസ്ത ദാതാവ്-മോഡൽ ജോഡികൾ / 1,475 റോ മോഡൽ ID-കൾ** ഉൾപ്പെടുന്നു; വേറിട്ട സൗജന്യ-ബജറ്റ് കാറ്റലോഗിൽ ഓരോ മോഡലിനുമുള്ള **471 വരികൾ**, **34 ആവർത്തിക്കുന്ന പൂളുകൾ**, **52 ആവർത്തിക്കുന്നതും കീ ആവശ്യമില്ലാത്തതുമായ എന്നേക്കും സൗജന്യ ദാതാക്കൾ** എന്നിവയുണ്ട്. ഇവ രൂപകൽപ്പനപ്രകാരം വ്യത്യസ്ത കണക്കാധാരങ്ങളാണ്; നിർവചനങ്ങളും പൂൾ-ഡ്യൂപ്ലിക്കേറ്റ് നീക്കം ചെയ്ത കണക്കുകൂട്ടലുകളും [ദാതാവ് റഫറൻസ്](docs/reference/PROVIDER_REFERENCE.md), [സൗജന്യ ശ്രേണികൾ](docs/reference/FREE_TIERS.md) എന്നിവയിൽ ലഭ്യമാണ്.
@@ -1295,7 +1295,7 @@ port-ൽ ഒരേ process വഴിയാണ് ലഭ്യമാക്കു റൺടൈംNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ഭാഷTypeScript 6.0 — src/, open-sse/ എന്നിവയിലുടനീളം 100% TypeScript (v2.0 മുതൽ core-ൽ any ഒന്നുമില്ല) ഫ്രെയിംവർക്ക്Next.js 16 + React 19 + Tailwind CSS 4 - ഡാറ്റാബേസ്better-sqlite3 (SQLite, WAL ജേണലിംഗ്) + LowDB (JSON ലെഗസി) — 147 ഡൊമെയ്ൻ മോഡ്യൂളുകൾ, 230 മൈഗ്രേഷനുകൾ + ഡാറ്റാബേസ്better-sqlite3 (SQLite, WAL ജേണലിംഗ്) + LowDB (JSON ലെഗസി) — 147 ഡൊമെയ്ൻ മോഡ്യൂളുകൾ, 232 മൈഗ്രേഷനുകൾ മെമ്മറിSQLite FTS5 പൂർണ്ണ-ടെക്സ്റ്റ് + int8-quantized വെക്ടർ എംബെഡ്ഡിംഗുകൾ, ടൈപ്പുചെയ്ത ഡിക്കേ സ്കീമകൾZod 4 — MCP ടൂൾ I/O സാധൂകരണം + API കരാറുകൾ പ്രോട്ടോക്കോളുകൾMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ml/llm.txt b/docs/i18n/ml/llm.txt index 425aed9a154..0158e00a28b 100644 --- a/docs/i18n/ml/llm.txt +++ b/docs/i18n/ml/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/mr/README.md b/docs/i18n/mr/README.md index 295b1d25cf9..d6136151a2b 100644 --- a/docs/i18n/mr/README.md +++ b/docs/i18n/mr/README.md @@ -15,7 +15,7 @@ **[Diego Souza](https://github.com/diegosouzapw) यांनी निर्मित** -OmniRoute — कोडिंग कधीही थांबवू नका. प्रत्येक AI टूल → 383 प्रदाते — 150+ मोफत — एका एंडपॉइंटद्वारे. Claude Code, Codex, Cursor, Cline, Copilot आणि Antigravity यांना ऑटो-फॉलबॅकसह FREE Claude / GPT / Gemini मध्ये जोडते. RTK + Caveman स्टॅक्ड कॉम्प्रेशनमुळे 15–95% टोकन्सची बचत (~89% सरासरी) — मर्यादा कधीही गाठू नका. 383 AI प्रदाते · 150+ मोफत टियर · ~1.62B मोफत टोकन्स/महिना · 20 राउटिंग स्ट्रॅटेजी · सुरू करण्यासाठी $0. +OmniRoute — कोडिंग कधीही थांबवू नका. प्रत्येक AI टूल → 384 प्रदाते — 150+ मोफत — एका एंडपॉइंटद्वारे. Claude Code, Codex, Cursor, Cline, Copilot आणि Antigravity यांना ऑटो-फॉलबॅकसह FREE Claude / GPT / Gemini मध्ये जोडते. RTK + Caveman स्टॅक्ड कॉम्प्रेशनमुळे 15–95% टोकन्सची बचत (~89% सरासरी) — मर्यादा कधीही गाठू नका. 384 AI प्रदाते · 150+ मोफत टियर · ~1.62B मोफत टोकन्स/महिना · 20 राउटिंग स्ट्रॅटेजी · सुरू करण्यासाठी $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------- | :------------------------: | :---------: | :-------------------: | -| 🌐 प्रदाते | 352 | **358** | 383 | +| 🌐 प्रदाते | 352 | **358** | 384 | | 🧠 अद्वितीय चॅट मॉडेल IDs | 1320 | **1374** | 1491 | | 🖼️ मोडॅलिटी ब्रिज | 🆕 दृष्टी + ऑडिओ + व्हिडिओ | ✓ | ✓ | | 📡 रडार विनामूल्य कॅटलॉग | 🆕 ऐच्छिक सहभाग | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ वैशिष्ट्ये 🎯 कॉम्बोज - 🌐 प्रदाते + 🌐 प्रदाते 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -वचन — एक एंडपॉइंट आणि 383 प्रदाते. दुसरे निरोगी लक्ष्य उपलब्ध असेपर्यंत स्वयंचलित फॉलबॅक रूटिंग सुरू ठेवते. सहा स्तंभ: 383 प्रदात्यांमध्ये लवचिक फॉलबॅक · पात्र वर्कलोडवर 95% पर्यंत टोकन बचत · 150+ विनामूल्य टियर आणि 52 आवर्ती/keyless कायमस्वरूपी विनामूल्य प्रदात्यांसह सुरू करण्यासाठी $0 · एका कॉन्फिगद्वारे 39 CLI/एजंट इंटिग्रेशन्स · /v1 वर OpenAI, Claude, Gemini आणि Responses API सुसंगतता · सर्किट ब्रेकर, TLS stealth, MCP 110 टूल्स, A2A, मेमरी, guardrails, evals आणि 7,800+ ट्रॅक केलेल्या टेस्ट फाइल्समधील 56,000+ स्थिर टेस्ट घोषणांसह उत्पादन नियंत्रण. +वचन — एक एंडपॉइंट आणि 384 प्रदाते. दुसरे निरोगी लक्ष्य उपलब्ध असेपर्यंत स्वयंचलित फॉलबॅक रूटिंग सुरू ठेवते. सहा स्तंभ: 384 प्रदात्यांमध्ये लवचिक फॉलबॅक · पात्र वर्कलोडवर 95% पर्यंत टोकन बचत · 150+ विनामूल्य टियर आणि 52 आवर्ती/keyless कायमस्वरूपी विनामूल्य प्रदात्यांसह सुरू करण्यासाठी $0 · एका कॉन्फिगद्वारे 39 CLI/एजंट इंटिग्रेशन्स · /v1 वर OpenAI, Claude, Gemini आणि Responses API सुसंगतता · सर्किट ब्रेकर, TLS stealth, MCP 110 टूल्स, A2A, मेमरी, guardrails, evals आणि 7,800+ ट्रॅक केलेल्या टेस्ट फाइल्समधील 56,000+ स्थिर टेस्ट घोषणांसह उत्पादन नियंत्रण.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRoute ला वेगळे ठरवणारी वैशिष्ट्ये — 13 क्षमतांमध्ये 9router, OpenRouter, CLIProxyAPI आणि LiteLLM यांच्या तुलनेतील दिनांकित वैशिष्ट्य-चित्र. OmniRoute: 383 प्रदाते, अंगभूत 150+ विनामूल्य स्तर, 20 रूटिंग धोरणे, 12-इंजिन टोकन कॉम्प्रेशन, 110 साधनांसह अंगभूत MCP सर्व्हर, A2A एजंट प्रोटोकॉल, पर्सिस्टंट मेमरी, गार्डरेल्स, क्लाउड एजंट्स, TLS फिंगरप्रिंट स्टेल्थ, Desktop/Termux/PWA आणि 67 i18n UI लोकेल्स. OmniRoute MIT-परवानाकृत असून स्वतः होस्ट करता येते. स्पर्धकांची क्षमता आणि संख्या बदलू शकतात; लिंक केलेली पद्धतशास्त्रीय माहिती पहा. +OmniRoute ला वेगळे ठरवणारी वैशिष्ट्ये — 13 क्षमतांमध्ये 9router, OpenRouter, CLIProxyAPI आणि LiteLLM यांच्या तुलनेतील दिनांकित वैशिष्ट्य-चित्र. OmniRoute: 384 प्रदाते, अंगभूत 150+ विनामूल्य स्तर, 20 रूटिंग धोरणे, 12-इंजिन टोकन कॉम्प्रेशन, 110 साधनांसह अंगभूत MCP सर्व्हर, A2A एजंट प्रोटोकॉल, पर्सिस्टंट मेमरी, गार्डरेल्स, क्लाउड एजंट्स, TLS फिंगरप्रिंट स्टेल्थ, Desktop/Termux/PWA आणि 67 i18n UI लोकेल्स. OmniRoute MIT-परवानाकृत असून स्वतः होस्ट करता येते. स्पर्धकांची क्षमता आणि संख्या बदलू शकतात; लिंक केलेली पद्धतशास्त्रीय माहिती पहा. 📊 9router, OpenRouter, CLIProxyAPI आणि LiteLLM यांच्या तुलनेतील संपूर्ण पद्धतशास्त्र व प्रत्येक वैशिष्ट्याचा तपशील → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar वापरणे ऐच्छिक असून तो केवळ GET - **🖼️ नवीन endpoints** — `/v1/ocr` (Mistral OCR) आणि `/v1/audio/translations` (Whisper-style) मुळे media surface अधिक पूर्ण होते. → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 Image / video / audio generation** — media साठी एक API: xAI Grok Imagine आणि Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind आणि ElevenLabs सारखे speech providers. → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 Deployment आणि ops** — reverse-proxy `basePath`, browser-language auto-detect, per-key device tracking, root-less MITM trust आणि zh-TW localization. → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 अधिक providers आणि agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login सह Grok Build (xAI), प्रथम-श्रेणीचे Ollama card, Claude Opus 5 आणि Sonnet 5, Kimi ची अधिकृत भागीदारी (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… आणि नव्याने अद्ययावत **383-provider catalog**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 अधिक providers आणि agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login सह Grok Build (xAI), प्रथम-श्रेणीचे Ollama card, Claude Opus 5 आणि Sonnet 5, Kimi ची अधिकृत भागीदारी (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… आणि नव्याने अद्ययावत **384-provider catalog**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 Routing transparency** — प्रत्येक response मध्ये सेवा देणारी strategy/provider/latency दर्शवणारा `X-OmniRoute-Decision` header असतो; नवीन `cache-optimized` combo strategy आणि Auto-Combo `cacheAffinity` factor repeat requests त्या cached prefix असलेल्या connection कडे परत पाठवतात; तसेच read-only `/v1/auto-combo/{channel}/candidates` endpoint `auto/*` channel चा live candidate pool उघड करते. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ स्थानिक कामगिरी आणि पायाभूत सुविधा** — एका क्लिकवर स्थानिक Redis, Cloudflare Workers / Deno Deploy रिले डिप्लॉयर, तसेच पर्यवेक्षित एम्बेडेड सेवा म्हणून Bifrost आणि Mux. → [एम्बेडेड सेवा](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 यामध्ये आणखी समाविष्ट** — प्लगइन फ्रेमवर्क + मार्केटप्लेस, Omni/Agent/GitHub स्किल्स फ्रेमवर्क, Obsidian व्हॉल्ट एकत्रीकरण (22 MCP टूल्स), OpenAI-सुसंगत Batch आणि Files API, सिमॅंटिक प्रतिसाद कॅशे, लीडरबोर्डसह गेमिफिकेशन, ACP एजंट शोध (15 अंगभूत एजंट), BigQuery मध्ये नियोजित लॉग निर्यात, `auto/chaos` समांतर मल्टी-मॉडेल फॅन-आउट, Telegram बॉट ब्रिज, अॅपमधील आवृत्ती व्यवस्थापक आणि LMArena-ELO मोफत-प्रदाता क्रमवारी. → [दस्तऐवज](docs/README.md) @@ -688,11 +688,11 @@ omniroute configure codex # तसेच: claude opencode qwen aider go
-## 🌐 383 AI प्रदाते — कॅटलॉगमध्ये नोंदवलेले 153 मोफत +## 🌐 384 AI प्रदाते — कॅटलॉगमध्ये नोंदवलेले 153 मोफत
-> **383 नोंदणीकृत प्रदाते** canonical chat, media, search, local, cloud-agent आणि system संग्रहांमध्ये उपलब्ध आहेत; यांपैकी **153 प्रदात्यांकडे `hasFree: true` शोध मेटाडेटा आहे**. chat model registry मध्ये **239 प्रदाते / 3,059 स्वतंत्र provider-model जोड्या / 1,475 कच्चे model IDs** समाविष्ट आहेत; स्वतंत्र free-budget catalog मध्ये **471 प्रति-मॉडेल नोंदी**, **34 आवर्ती पूल** आणि **52 आवर्ती/keyless कायमस्वरूपी मोफत प्रदाते** आहेत. हे वेगवेगळे denominators जाणीवपूर्वक ठेवले आहेत; व्याख्या आणि pool-deduped गणना [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) आणि [Free Tiers](docs/reference/FREE_TIERS.md) मध्ये उपलब्ध आहेत. +> **384 नोंदणीकृत प्रदाते** canonical chat, media, search, local, cloud-agent आणि system संग्रहांमध्ये उपलब्ध आहेत; यांपैकी **153 प्रदात्यांकडे `hasFree: true` शोध मेटाडेटा आहे**. chat model registry मध्ये **239 प्रदाते / 3,059 स्वतंत्र provider-model जोड्या / 1,475 कच्चे model IDs** समाविष्ट आहेत; स्वतंत्र free-budget catalog मध्ये **471 प्रति-मॉडेल नोंदी**, **34 आवर्ती पूल** आणि **52 आवर्ती/keyless कायमस्वरूपी मोफत प्रदाते** आहेत. हे वेगवेगळे denominators जाणीवपूर्वक ठेवले आहेत; व्याख्या आणि pool-deduped गणना [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) आणि [Free Tiers](docs/reference/FREE_TIERS.md) मध्ये उपलब्ध आहेत.
@@ -1294,7 +1294,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 देखील रनटाइमNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 भाषाTypeScript 6.0 — src/ आणि open-sse/ मध्ये 100% TypeScript (v2.0 पासून कोअरमध्ये शून्य any) फ्रेमवर्कNext.js 16 + React 19 + Tailwind CSS 4 - डेटाबेसbetter-sqlite3 (SQLite, WAL जर्नलिंग) + LowDB (JSON लेगसी) — 147 डोमेन मॉड्यूल्स, 230 माइग्रेशन्स + डेटाबेसbetter-sqlite3 (SQLite, WAL जर्नलिंग) + LowDB (JSON लेगसी) — 147 डोमेन मॉड्यूल्स, 232 माइग्रेशन्स मेमरीSQLite FTS5 पूर्ण-मजकूर + int8-क्वांटाइझ्ड व्हेक्टर एम्बेडिंग्ज, टाइपयुक्त क्षय स्कीमाZod 4 — MCP टूल I/O प्रमाणीकरण + API करार प्रोटोकॉलMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/mr/llm.txt b/docs/i18n/mr/llm.txt index b18fc5e0817..98c099a8178 100644 --- a/docs/i18n/mr/llm.txt +++ b/docs/i18n/mr/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ms/README.md b/docs/i18n/ms/README.md index c27e57ee677..78a3a12ea28 100644 --- a/docs/i18n/ms/README.md +++ b/docs/i18n/ms/README.md @@ -15,7 +15,7 @@ **Dicipta oleh [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Jangan berhenti menulis kod. Setiap alat AI → 383 pembekal — 150+ percuma — melalui satu titik akhir. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity kepada Claude / GPT / Gemini PERCUMA dengan sandaran automatik. Pemampatan bertindan RTK + Caveman menjimatkan 15–95% token (~89% secara purata) — jangan sekali-kali mencapai had. 383 pembekal AI · 150+ peringkat percuma · ~1.62B token percuma/bln · 20 strategi penghalaan · $0 untuk bermula. +OmniRoute — Jangan berhenti menulis kod. Setiap alat AI → 384 pembekal — 150+ percuma — melalui satu titik akhir. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity kepada Claude / GPT / Gemini PERCUMA dengan sandaran automatik. Pemampatan bertindan RTK + Caveman menjimatkan 15–95% token (~89% secara purata) — jangan sekali-kali mencapai had. 384 pembekal AI · 150+ peringkat percuma · ~1.62B token percuma/bln · 20 strategi penghalaan · $0 untuk bermula.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | -------------------------------- | :----------------------------: | :---------: | :-------------------: | -| 🌐 Penyedia | 352 | **358** | 383 | +| 🌐 Penyedia | 352 | **358** | 384 | | 🧠 ID model chat unik | 1320 | **1374** | 1491 | | 🖼️ Jambatan modaliti | 🆕 penglihatan + audio + video | ✓ | ✓ | | 📡 Katalog percuma Radar | 🆕 ikut serta | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Ciri 🎯 Kombinasi - 🌐 Pembekal + 🌐 Pembekal 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Janji — Satu titik akhir dan 383 penyedia. Fallback automatik memastikan penghalaan diteruskan selagi sasaran sihat lain masih tersedia. Enam tonggak: fallback berdaya tahan merentasi 383 penyedia · sehingga 95% penjimatan token untuk beban kerja yang layak · $0 untuk bermula dengan lebih 150 tingkatan percuma dan 52 penyedia percuma selama-lamanya yang berulang/tanpa kunci · 39 integrasi CLI/ejen melalui satu konfigurasi · keserasian OpenAI, Claude, Gemini dan Responses API di /v1 · kawalan pengeluaran termasuk pemutus litar, penyamaran TLS, 110 alat MCP, A2A, memori, pagar kawalan, penilaian serta 56,000+ pengisytiharan ujian statik merentasi 7,800+ fail ujian yang dijejaki. +Janji — Satu titik akhir dan 384 penyedia. Fallback automatik memastikan penghalaan diteruskan selagi sasaran sihat lain masih tersedia. Enam tonggak: fallback berdaya tahan merentasi 384 penyedia · sehingga 95% penjimatan token untuk beban kerja yang layak · $0 untuk bermula dengan lebih 150 tingkatan percuma dan 52 penyedia percuma selama-lamanya yang berulang/tanpa kunci · 39 integrasi CLI/ejen melalui satu konfigurasi · keserasian OpenAI, Claude, Gemini dan Responses API di /v1 · kawalan pengeluaran termasuk pemutus litar, penyamaran TLS, 110 alat MCP, A2A, memori, pagar kawalan, penilaian serta 56,000+ pengisytiharan ujian statik merentasi 7,800+ fail ujian yang dijejaki.

@@ -510,7 +510,7 @@ Kesemua **20** strategi — campur & padankan bagi setiap langkah kombo: -Perkara yang membezakan OmniRoute — gambaran ciri bertarikh berbanding 9router, OpenRouter, CLIProxyAPI dan LiteLLM merentasi 13 keupayaan. OmniRoute: 383 pembekal, lebih 150 peringkat percuma terbina dalam, 20 strategi penghalaan, pemampatan token 12 enjin, pelayan MCP terbina dalam dengan 110 alat, protokol ejen A2A, memori berterusan, perlindungan, ejen awan, penyamaran cap jari TLS, Desktop/Termux/PWA dan 67 lokal UI i18n. OmniRoute dilesenkan di bawah MIT dan boleh dihos sendiri. Keupayaan dan bilangan pesaing mungkin berubah; lihat metodologi yang dipautkan. +Perkara yang membezakan OmniRoute — gambaran ciri bertarikh berbanding 9router, OpenRouter, CLIProxyAPI dan LiteLLM merentasi 13 keupayaan. OmniRoute: 384 pembekal, lebih 150 peringkat percuma terbina dalam, 20 strategi penghalaan, pemampatan token 12 enjin, pelayan MCP terbina dalam dengan 110 alat, protokol ejen A2A, memori berterusan, perlindungan, ejen awan, penyamaran cap jari TLS, Desktop/Termux/PWA dan 67 lokal UI i18n. OmniRoute dilesenkan di bawah MIT dan boleh dihos sendiri. Keupayaan dan bilangan pesaing mungkin berubah; lihat metodologi yang dipautkan. 📊 Metodologi lengkap & perincian setiap ciri berbanding 9router, OpenRouter, CLIProxyAPI & LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar adalah secara ikut serta dan GET sahaja. Klien OmniRoute tidak memuat naik - **🖼️ Titik akhir baharu** — `/v1/ocr` (Mistral OCR) dan `/v1/audio/translations` (gaya Whisper) melengkapkan permukaan media. → [Rujukan API](docs/reference/API_REFERENCE.md) - **🎨 Penjanaan imej / video / audio** — satu API untuk media: xAI Grok Imagine & video Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind, serta pembekal pertuturan seperti ElevenLabs. → [Rujukan API](docs/reference/API_REFERENCE.md) - **🌍 Penerapan & operasi** — `basePath` proksi songsang, pengesanan automatik bahasa pelayar, penjejakan peranti mengikut kunci, kepercayaan MITM tanpa root, penyetempatan zh-TW. → [Persekitaran](docs/reference/ENVIRONMENT.md) -- **🤝 Lebih banyak pembekal & ejen** — ejen awan (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) dengan log masuk pelayar + OAuth, kad kelas pertama Ollama, Claude Opus 5 & Sonnet 5, kerjasama rasmi Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… serta **katalog 383 pembekal** yang diperbaharui. → [Pembekal](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Lebih banyak pembekal & ejen** — ejen awan (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) dengan log masuk pelayar + OAuth, kad kelas pertama Ollama, Claude Opus 5 & Sonnet 5, kerjasama rasmi Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… serta **katalog 384 pembekal** yang diperbaharui. → [Pembekal](docs/reference/PROVIDER_REFERENCE.md) - **📡 Ketelusan penghalaan** — setiap respons membawa pengepala `X-OmniRoute-Decision` yang menamakan strategi/pembekal/latensi yang menyampaikannya, strategi gabungan `cache-optimized` baharu + faktor laluan `cacheAffinity` Auto-Combo menghalakan permintaan berulang kembali kepada sambungan yang menyimpan awalan dalam cache, dan titik akhir baca sahaja `/v1/auto-combo/{channel}/candidates` mendedahkan kumpulan calon langsung saluran `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md)。 - **⚡ Prestasi & infrastruktur setempat** — Redis setempat satu klik, penyebar geganti Cloudflare Workers / Deno Deploy, Bifrost & Mux sebagai perkhidmatan terbenam yang diselia. → [Perkhidmatan Terbenam](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Turut disertakan** — rangka kerja pemalam + pasaran, rangka kerja kemahiran Omni/Agent/GitHub, integrasi peti besi Obsidian (22 alat MCP), API Batch & Files yang serasi dengan OpenAI, cache respons semantik, gamifikasi dengan papan pendahulu, penemuan ejen ACP (15 ejen terbina dalam), pengeksportan log berjadual ke BigQuery, fan-out selari berbilang model `auto/chaos`, jambatan bot Telegram, pengurus versi dalam aplikasi dan kedudukan penyedia percuma LMArena-ELO. → [Dokumentasi](docs/README.md) @@ -689,11 +689,11 @@ masuk dalam sejarah shell anda. → [Integrasi CLI](docs/guides/CLI-INTEGRATIONS
-## 🌐 383 Penyedia AI — 153 Ditandai sebagai Percuma dalam Katalog +## 🌐 384 Penyedia AI — 153 Ditandai sebagai Percuma dalam Katalog
-> **383 penyedia berdaftar** merentasi koleksi kanonik chat, media, carian, setempat, ejen awan dan sistem, termasuk **153 yang membawa metadata penemuan `hasFree: true`**. Daftar model chat merangkumi **239 penyedia / 3,059 pasangan penyedia-model berbeza / 1,475 ID model mentah**; katalog bajet percuma yang berasingan mempunyai **471 baris setiap model**, **34 kumpulan berulang** dan **52 penyedia percuma selama-lamanya yang berulang/tanpa kunci**. Ini ialah penyebut yang berbeza mengikut reka bentuk; definisi dan pengiraan yang dinyahgandakan mengikut kumpulan tersedia dalam [Rujukan Penyedia](docs/reference/PROVIDER_REFERENCE.md) dan [Peringkat Percuma](docs/reference/FREE_TIERS.md). +> **384 penyedia berdaftar** merentasi koleksi kanonik chat, media, carian, setempat, ejen awan dan sistem, termasuk **153 yang membawa metadata penemuan `hasFree: true`**. Daftar model chat merangkumi **239 penyedia / 3,059 pasangan penyedia-model berbeza / 1,475 ID model mentah**; katalog bajet percuma yang berasingan mempunyai **471 baris setiap model**, **34 kumpulan berulang** dan **52 penyedia percuma selama-lamanya yang berulang/tanpa kunci**. Ini ialah penyebut yang berbeza mengikut reka bentuk; definisi dan pengiraan yang dinyahgandakan mengikut kumpulan tersedia dalam [Rujukan Penyedia](docs/reference/PROVIDER_REFERENCE.md) dan [Peringkat Percuma](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ Metrik kanonik pada 2026-10-18: **1.976 video unik** · **20.220.289 tontonan ya Masa jalanNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 BahasaTypeScript 6.0 — 100% TypeScript merentasi src/ dan open-sse/ (sifar any dalam teras sejak v2.0) Rangka kerjaNext.js 16 + React 19 + Tailwind CSS 4 - Pangkalan databetter-sqlite3 (SQLite, pengjurnalan WAL) + LowDB (JSON legasi) — 147 modul domain, 230 migrasi + Pangkalan databetter-sqlite3 (SQLite, pengjurnalan WAL) + LowDB (JSON legasi) — 147 modul domain, 232 migrasi MemoriTeks penuh SQLite FTS5 + pembenaman vektor terkuantum int8, pereputan bertip SkemaZod 4 — pengesahan I/O alat MCP + kontrak API ProtokolMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ms/llm.txt b/docs/i18n/ms/llm.txt index c4f71fdfd4c..ec4160bb652 100644 --- a/docs/i18n/ms/llm.txt +++ b/docs/i18n/ms/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/mt/README.md b/docs/i18n/mt/README.md index e4fb8f9dc32..bf101f682a8 100644 --- a/docs/i18n/mt/README.md +++ b/docs/i18n/mt/README.md @@ -15,7 +15,7 @@ **Maħluq minn [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Qatt tieqaf tipprogramma. Kull għodda tal-IA → 383 fornitur — 150+ b’xejn — permezz ta’ endpoint wieħed. Claude Code, Codex, Cursor, Cline, Copilot u Antigravity f’Claude / GPT / Gemini B’XEJN b’fallback awtomatiku. Il-kompressjoni f’saffi ta’ RTK + Caveman tiffranka 15–95% tat-tokens (~89% bħala medja) — qatt ma tilħaq il-limiti. 383 fornitur tal-IA · 150+ livelli b’xejn · ~1.62B tokens b’xejn fix-xahar · 20 strateġija ta’ routing · $0 biex tibda. +OmniRoute — Qatt tieqaf tipprogramma. Kull għodda tal-IA → 384 fornitur — 150+ b’xejn — permezz ta’ endpoint wieħed. Claude Code, Codex, Cursor, Cline, Copilot u Antigravity f’Claude / GPT / Gemini B’XEJN b’fallback awtomatiku. Il-kompressjoni f’saffi ta’ RTK + Caveman tiffranka 15–95% tat-tokens (~89% bħala medja) — qatt ma tilħaq il-limiti. 384 fornitur tal-IA · 150+ livelli b’xejn · ~1.62B tokens b’xejn fix-xahar · 20 strateġija ta’ routing · $0 biex tibda.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ---------------------------------- | :------------------------: | :---------: | :-------------------: | -| 🌐 Fornituri | 352 | **358** | 383 | +| 🌐 Fornituri | 352 | **358** | 384 | | 🧠 IDs uniċi tal-mudelli taċ-chat | 1320 | **1374** | 1491 | | 🖼️ Pont tal-modalitajiet | 🆕 viżjoni + awdjo + vidjo | ✓ | ✓ | | 📡 Katalgu b’xejn tar-Radar | 🆕 b’għażla | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Karatteristiċi 🎯 Kombinazzjonijiet - 🌐 Fornituri + 🌐 Fornituri 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Il-Wegħda — Endpoint wieħed u 383 fornitur. Ir-ripplijar awtomatiku jżomm ir-routing għaddej sakemm tkun disponibbli destinazzjoni oħra b’saħħitha. Sitt pilastri: ripplijar reżiljenti bejn 383 fornitur · sa 95% iffrankar fit-tokens fuq xogħolijiet eliġibbli · $0 biex tibda b’aktar minn 150 livell bla ħlas u 52 fornitur rikorrenti/bla ċavetta bla ħlas għal dejjem · 39 integrazzjoni CLI/aġent permezz ta’ konfigurazzjoni waħda · kompatibbiltà ma’ OpenAI, Claude, Gemini u Responses API f’/v1 · kontrolli għall-produzzjoni li jinkludu circuit breakers, stealth ta’ TLS, 110 għodda MCP, A2A, memorja, guardrails, evals u aktar minn 56,000 dikjarazzjoni ta’ test statiku f’aktar minn 7,800 fajl tat-test traċċat. +Il-Wegħda — Endpoint wieħed u 384 fornitur. Ir-ripplijar awtomatiku jżomm ir-routing għaddej sakemm tkun disponibbli destinazzjoni oħra b’saħħitha. Sitt pilastri: ripplijar reżiljenti bejn 384 fornitur · sa 95% iffrankar fit-tokens fuq xogħolijiet eliġibbli · $0 biex tibda b’aktar minn 150 livell bla ħlas u 52 fornitur rikorrenti/bla ċavetta bla ħlas għal dejjem · 39 integrazzjoni CLI/aġent permezz ta’ konfigurazzjoni waħda · kompatibbiltà ma’ OpenAI, Claude, Gemini u Responses API f’/v1 · kontrolli għall-produzzjoni li jinkludu circuit breakers, stealth ta’ TLS, 110 għodda MCP, A2A, memorja, guardrails, evals u aktar minn 56,000 dikjarazzjoni ta’ test statiku f’aktar minn 7,800 fajl tat-test traċċat.

@@ -510,7 +510,7 @@ L-**20** strateġija kollha — ħallat u qabbel f’kull pass tal-combo: -X’jiddistingwi lil OmniRoute — ħarsa ġenerali datata tal-karatteristiċi mqabbla ma’ 9router, OpenRouter, CLIProxyAPI u LiteLLM fuq 13-il kapaċità. OmniRoute: 383 fornitur, aktar minn 150 livell b’xejn integrati, 20 strateġija ta’ routing, kompressjoni tat-tokens b’12-il magna, server MCP integrat b’110 għodda, protokoll tal-aġenti A2A, memorja persistenti, guardrails, aġenti cloud, stealth tal-marki tas-swaba’ TLS, Desktop/Termux/PWA u 67 locale tal-UI għall-i18n. OmniRoute huwa liċenzjat taħt MIT u jista’ jiġi ospitat lokalment. Il-kapaċitajiet u l-għadd tal-kompetituri jistgħu jinbidlu; ara l-metodoloġija marbuta. +X’jiddistingwi lil OmniRoute — ħarsa ġenerali datata tal-karatteristiċi mqabbla ma’ 9router, OpenRouter, CLIProxyAPI u LiteLLM fuq 13-il kapaċità. OmniRoute: 384 fornitur, aktar minn 150 livell b’xejn integrati, 20 strateġija ta’ routing, kompressjoni tat-tokens b’12-il magna, server MCP integrat b’110 għodda, protokoll tal-aġenti A2A, memorja persistenti, guardrails, aġenti cloud, stealth tal-marki tas-swaba’ TLS, Desktop/Termux/PWA u 67 locale tal-UI għall-i18n. OmniRoute huwa liċenzjat taħt MIT u jista’ jiġi ospitat lokalment. Il-kapaċitajiet u l-għadd tal-kompetituri jistgħu jinbidlu; ara l-metodoloġija marbuta. 📊 Metodoloġija sħiħa u dettalji għal kull karatteristika mqabbla ma’ 9router, OpenRouter, CLIProxyAPI u LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar huwa fakultattiv u juża GET biss. Il-klijent OmniRoute ma jtellax prompts - **🖼️ Endpoints ġodda** — `/v1/ocr` (Mistral OCR) u `/v1/audio/translations` (stil Whisper) jikkompletaw il-wiċċ tal-midja. → [Referenza tal-API](docs/reference/API_REFERENCE.md) - **🎨 Ġenerazzjoni ta’ stampi / vidjows / awdjo** — API waħda għall-midja: xAI Grok Imagine u Novita AI għall-vidjows, ComfyUI, Magnific, Adobe Firefly, Segmind, u fornituri tad-diskors bħal ElevenLabs. → [Referenza tal-API](docs/reference/API_REFERENCE.md) - **🌍 Skjerament u operazzjonijiet** — `basePath` ta’ reverse-proxy, identifikazzjoni awtomatika tal-lingwa tal-browser, traċċar tal-apparati għal kull ċavetta, fiduċja MITM mingħajr root, u lokalizzazzjoni zh-TW. → [Ambjent](docs/reference/ENVIRONMENT.md) -- **🤝 Aktar fornituri u aġenti** — aġenti cloud (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) b’login permezz tal-browser u OAuth, karta uffiċjali għal Ollama, Claude Opus 5 u Sonnet 5, sħubija uffiċjali ma’ Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… u katalgu aġġornat ta’ **383 fornitur**. → [Fornituri](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Aktar fornituri u aġenti** — aġenti cloud (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) b’login permezz tal-browser u OAuth, karta uffiċjali għal Ollama, Claude Opus 5 u Sonnet 5, sħubija uffiċjali ma’ Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… u katalgu aġġornat ta’ **384 fornitur**. → [Fornituri](docs/reference/PROVIDER_REFERENCE.md) - **📡 Trasparenza tar-rotot** — kull tweġiba tinkludi header `X-OmniRoute-Decision` li jindika l-istrateġija/fornitur/latency li serva t-talba, strateġija combo ġdida `cache-optimized` + fattur `cacheAffinity` ta’ Auto-Combo li jirritorna talbiet ripetuti lejn il-konnessjoni li żżomm il-prefiss fil-cache, u endpoint ġdid li jinqara biss `/v1/auto-combo/{channel}/candidates` jesponi l-pool attiv ta’ kandidati tal-kanal `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md). - **⚡ Prestazzjoni u infrastruttura lokali** — Redis lokali b’klikk waħda, deployers tar-relay ta’ Cloudflare Workers / Deno Deploy, Bifrost u Mux bħala servizzi inkorporati ssorveljati. → [Servizzi Inkorporati](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Inkluż ukoll** — qafas tal-plugins + suq, frameworks tal-ħiliet ta’ Omni/Agent/GitHub, integrazzjoni ma’ Obsidian vault (22 għodda MCP), APIs ta’ Batch & Files kompatibbli ma’ OpenAI, cache semantiku għar-risposti, gamification b’klassifiki, skoperta ta’ aġenti ACP (15-il aġent inkorporat), esportazzjoni skedata tar-reġistri lejn BigQuery, fan-out parallel ta’ diversi mudelli `auto/chaos`, bridge għal bot ta’ Telegram, maniġer tal-verżjonijiet fl-applikazzjoni u klassifiki LMArena-ELO ta’ fornituri b’xejn. → [Dokumentazzjoni](docs/README.md) @@ -689,11 +689,11 @@ mill-istorja tas-shell tiegħek. → [CLI Integrations](docs/guides/CLI-INTEGRAT
-## 🌐 383 Fornitur tal-AI — 153 Immarkati fil-Katalgu bħala Bla Ħlas +## 🌐 384 Fornitur tal-AI — 153 Immarkati fil-Katalgu bħala Bla Ħlas
-> **383 fornitur irreġistrat** fil-kollezzjonijiet kanoniċi taċ-chat, tal-midja, tat-tfittxija, lokali, tal-aġenti fil-cloud u tas-sistema, inklużi **153 li għandhom il-metadata ta’ skoperta `hasFree: true`**. Ir-reġistru tal-mudelli taċ-chat ikopri **239 fornitur / 3,059 par distint ta’ fornitur-mudell / 1,475 ID mhux ipproċessat tal-mudelli**; il-katalgu separat tal-baġit bla ħlas għandu **471 ringiela għal kull mudell**, **34 pool rikorrenti** u **52 fornitur rikorrenti/bla ċavetta u bla ħlas għal dejjem**. Dawn huma denominaturi differenti skont id-disinn; id-definizzjonijiet u l-kalkoli deduplikati skont il-pool jinsabu fir-[Referenza tal-Fornituri](docs/reference/PROVIDER_REFERENCE.md) u fil-[Livelli Bla Ħlas](docs/reference/FREE_TIERS.md). +> **384 fornitur irreġistrat** fil-kollezzjonijiet kanoniċi taċ-chat, tal-midja, tat-tfittxija, lokali, tal-aġenti fil-cloud u tas-sistema, inklużi **153 li għandhom il-metadata ta’ skoperta `hasFree: true`**. Ir-reġistru tal-mudelli taċ-chat ikopri **239 fornitur / 3,059 par distint ta’ fornitur-mudell / 1,475 ID mhux ipproċessat tal-mudelli**; il-katalgu separat tal-baġit bla ħlas għandu **471 ringiela għal kull mudell**, **34 pool rikorrenti** u **52 fornitur rikorrenti/bla ċavetta u bla ħlas għal dejjem**. Dawn huma denominaturi differenti skont id-disinn; id-definizzjonijiet u l-kalkoli deduplikati skont il-pool jinsabu fir-[Referenza tal-Fornituri](docs/reference/PROVIDER_REFERENCE.md) u fil-[Livelli Bla Ħlas](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ Metriċi kanoniċi fl-2026-10-18: **1.976 vidjo uniku** · **20.220.289 darba ma Ambjent ta' eżekuzzjoniNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 LingwaTypeScript 6.0 — 100% TypeScript f’src/ u open-sse/ (żero any fil-qalba minn v2.0) QafasNext.js 16 + React 19 + Tailwind CSS 4 - Bażi tad-dejtabetter-sqlite3 (SQLite, ġurnalar WAL) + LowDB (JSON ta’ wirt) — 147-il modulu tad-dominju, 230-il migrazzjoni + Bażi tad-dejtabetter-sqlite3 (SQLite, ġurnalar WAL) + LowDB (JSON ta’ wirt) — 147-il modulu tad-dominju, 232-il migrazzjoni MemorjaTest sħiħ b’SQLite FTS5 + embeddings vettorjali kwantizzati b’int8, tħassir ittajpjat SkemiZod 4 — validazzjoni tal-I/O tal-għodod MCP + kuntratti tal-API ProtokolliMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/mt/llm.txt b/docs/i18n/mt/llm.txt index 4180ee4d30b..1966eb4ff15 100644 --- a/docs/i18n/mt/llm.txt +++ b/docs/i18n/mt/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/my/README.md b/docs/i18n/my/README.md index 676191aabbb..154b39b5c8e 100644 --- a/docs/i18n/my/README.md +++ b/docs/i18n/my/README.md @@ -15,7 +15,7 @@ **ဖန်တီးသူ [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — ကုဒ်ရေးသားခြင်းကို မရပ်တန့်ပါနှင့်။ AI tool တိုင်း → provider 383 ခု — အခမဲ့ 150+ ခု — endpoint တစ်ခုတည်းမှတစ်ဆင့်။ Claude Code၊ Codex၊ Cursor၊ Cline၊ Copilot နှင့် Antigravity ကို အလိုအလျောက် fallback ဖြင့် အခမဲ့ Claude / GPT / Gemini သို့ ချိတ်ဆက်အသုံးပြုနိုင်သည်။ RTK + Caveman ပေါင်းစပ် compression သည် token 15–95% (~89% ပျမ်းမျှ) ချွေတာပေးသောကြောင့် ကန့်သတ်ချက်များကို မည်သည့်အခါမျှ မထိတွေ့ရပါ။ AI provider 383 ခု · အခမဲ့ tier 150+ ခု · ~1.62B အခမဲ့ token/mo · routing strategy 20 ခု · စတင်ရန် $0. +OmniRoute — ကုဒ်ရေးသားခြင်းကို မရပ်တန့်ပါနှင့်။ AI tool တိုင်း → provider 384 ခု — အခမဲ့ 150+ ခု — endpoint တစ်ခုတည်းမှတစ်ဆင့်။ Claude Code၊ Codex၊ Cursor၊ Cline၊ Copilot နှင့် Antigravity ကို အလိုအလျောက် fallback ဖြင့် အခမဲ့ Claude / GPT / Gemini သို့ ချိတ်ဆက်အသုံးပြုနိုင်သည်။ RTK + Caveman ပေါင်းစပ် compression သည် token 15–95% (~89% ပျမ်းမျှ) ချွေတာပေးသောကြောင့် ကန့်သတ်ချက်များကို မည်သည့်အခါမျှ မထိတွေ့ရပါ။ AI provider 384 ခု · အခမဲ့ tier 150+ ခု · ~1.62B အခမဲ့ token/mo · routing strategy 20 ခု · စတင်ရန် $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | -------------------------------------------------- | :-------------------------: | :---------: | :------------------------: | -| 🌐 ပံ့ပိုးသူများ | 352 | **358** | 383 | +| 🌐 ပံ့ပိုးသူများ | 352 | **358** | 384 | | 🧠 ထူးခြားသော စကားပြောမော်ဒယ် ID များ | 1320 | **1374** | 1491 | | 🖼️ မီဒီယာအမျိုးအစား ချိတ်ဆက်တံတား | 🆕 အမြင် + အသံ + ဗီဒီယို | ✓ | ✓ | | 📡 Radar အခမဲ့ကတ်တလောက် | 🆕 ရွေးချယ်အသုံးပြုနိုင်သည် | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ အင်္ဂါရပ်များ 🎯 ပေါင်းစပ်မှုများ - 🌐 ပံ့ပိုးပေးသူများ + 🌐 ပံ့ပိုးပေးသူများ 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -ကတိပြုချက် — endpoint တစ်ခုနှင့် ပံ့ပိုးသူ 383 ခု။ ကျန်းမာနေသော အခြား target တစ်ခု ရရှိနိုင်သရွေ့ အလိုအလျောက် fallback သည် လမ်းကြောင်းပို့ခြင်းကို ဆက်လက်ထိန်းသိမ်းပေးသည်။ အဓိကတိုင် ၆ ခု — ပံ့ပိုးသူ 383 ခုအကြား ခံနိုင်ရည်ရှိသော fallback · သတ်မှတ်ချက်နှင့်ကိုက်ညီသော workload များတွင် token ကုန်ကျမှု 95% အထိ ချွေတာနိုင်ခြင်း · အခမဲ့အဆင့် 150+ ခုနှင့် ထပ်တလဲလဲရရှိနိုင်သော/keyless အမြဲတမ်းအခမဲ့ ပံ့ပိုးသူ 52 ခုဖြင့် $0 ဖြင့် စတင်နိုင်ခြင်း · config တစ်ခုတည်းမှတစ်ဆင့် CLI/agent ပေါင်းစည်းမှု 39 ခု · /v1 တွင် OpenAI၊ Claude၊ Gemini နှင့် Responses API ကိုက်ညီမှု · circuit breaker များ၊ TLS stealth၊ MCP tools 110 ခု၊ A2A၊ memory၊ guardrail များ၊ eval များနှင့် စောင့်ကြည့်ထားသော test file 7,800+ ခုအတွင်း static test declaration 56,000+ ခု အပါအဝင် production ထိန်းချုပ်မှုများ။ +ကတိပြုချက် — endpoint တစ်ခုနှင့် ပံ့ပိုးသူ 384 ခု။ ကျန်းမာနေသော အခြား target တစ်ခု ရရှိနိုင်သရွေ့ အလိုအလျောက် fallback သည် လမ်းကြောင်းပို့ခြင်းကို ဆက်လက်ထိန်းသိမ်းပေးသည်။ အဓိကတိုင် ၆ ခု — ပံ့ပိုးသူ 384 ခုအကြား ခံနိုင်ရည်ရှိသော fallback · သတ်မှတ်ချက်နှင့်ကိုက်ညီသော workload များတွင် token ကုန်ကျမှု 95% အထိ ချွေတာနိုင်ခြင်း · အခမဲ့အဆင့် 150+ ခုနှင့် ထပ်တလဲလဲရရှိနိုင်သော/keyless အမြဲတမ်းအခမဲ့ ပံ့ပိုးသူ 52 ခုဖြင့် $0 ဖြင့် စတင်နိုင်ခြင်း · config တစ်ခုတည်းမှတစ်ဆင့် CLI/agent ပေါင်းစည်းမှု 39 ခု · /v1 တွင် OpenAI၊ Claude၊ Gemini နှင့် Responses API ကိုက်ညီမှု · circuit breaker များ၊ TLS stealth၊ MCP tools 110 ခု၊ A2A၊ memory၊ guardrail များ၊ eval များနှင့် စောင့်ကြည့်ထားသော test file 7,800+ ခုအတွင်း static test declaration 56,000+ ခု အပါအဝင် production ထိန်းချုပ်မှုများ။

@@ -510,7 +510,7 @@ strategy **20** ခုလုံးကို ကွန်ဘိုအဆင့် -OmniRoute ကို ထူးခြားစေသော အချက်များ — စွမ်းဆောင်ရည် ၁၃ မျိုးအတွက် 9router၊ OpenRouter၊ CLIProxyAPI နှင့် LiteLLM တို့နှင့် နှိုင်းယှဉ်ထားသော ရက်စွဲပါ အင်္ဂါရပ်အနှစ်ချုပ်။ OmniRoute တွင် provider 383 ခု၊ ထည့်သွင်းတည်ဆောက်ထားသော အခမဲ့ tier 150+ ခု၊ routing strategy 20 ခု၊ engine 12 ခုကို အသုံးပြုသည့် token compression၊ tool 110 ခုပါဝင်သော ထည့်သွင်းတည်ဆောက်ထားသည့် MCP server၊ A2A agent protocol၊ အမြဲတမ်းမှတ်ဉာဏ်၊ guardrails၊ cloud agents၊ TLS fingerprint stealth၊ Desktop/Termux/PWA နှင့် i18n UI locale 67 ခု ပါဝင်သည်။ OmniRoute သည် MIT လိုင်စင်ရထားပြီး မိမိကိုယ်တိုင် host ပြုလုပ်နိုင်သည်။ ပြိုင်ဘက်များ၏ စွမ်းဆောင်ရည်များနှင့် အရေအတွက်များသည် ပြောင်းလဲနိုင်သောကြောင့် လင့်ခ်ချိတ်ထားသည့် နည်းစနစ်ကို ကြည့်ပါ။ +OmniRoute ကို ထူးခြားစေသော အချက်များ — စွမ်းဆောင်ရည် ၁၃ မျိုးအတွက် 9router၊ OpenRouter၊ CLIProxyAPI နှင့် LiteLLM တို့နှင့် နှိုင်းယှဉ်ထားသော ရက်စွဲပါ အင်္ဂါရပ်အနှစ်ချုပ်။ OmniRoute တွင် provider 384 ခု၊ ထည့်သွင်းတည်ဆောက်ထားသော အခမဲ့ tier 150+ ခု၊ routing strategy 20 ခု၊ engine 12 ခုကို အသုံးပြုသည့် token compression၊ tool 110 ခုပါဝင်သော ထည့်သွင်းတည်ဆောက်ထားသည့် MCP server၊ A2A agent protocol၊ အမြဲတမ်းမှတ်ဉာဏ်၊ guardrails၊ cloud agents၊ TLS fingerprint stealth၊ Desktop/Termux/PWA နှင့် i18n UI locale 67 ခု ပါဝင်သည်။ OmniRoute သည် MIT လိုင်စင်ရထားပြီး မိမိကိုယ်တိုင် host ပြုလုပ်နိုင်သည်။ ပြိုင်ဘက်များ၏ စွမ်းဆောင်ရည်များနှင့် အရေအတွက်များသည် ပြောင်းလဲနိုင်သောကြောင့် လင့်ခ်ချိတ်ထားသည့် နည်းစနစ်ကို ကြည့်ပါ။ 📊 9router၊ OpenRouter၊ CLIProxyAPI နှင့် LiteLLM တို့နှင့် အပြည့်အစုံသော နည်းစနစ်နှင့် အင်္ဂါရပ်တစ်ခုချင်းစီ၏ အသေးစိတ်နှိုင်းယှဉ်ချက် → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -691,11 +691,11 @@ omniroute configure codex # ထို့အပြင်: claude opencod
-## 🌐 AI ပံ့ပိုးသူ 383 ခု — Catalog တွင် အခမဲ့အဖြစ် အမှတ်အသားပြုထားသော 153 ခု +## 🌐 AI ပံ့ပိုးသူ 384 ခု — Catalog တွင် အခမဲ့အဖြစ် အမှတ်အသားပြုထားသော 153 ခု
-> **မှတ်ပုံတင်ထားသော ပံ့ပိုးသူ 383 ခု** ကို canonical chat၊ media၊ search၊ local၊ cloud-agent နှင့် system collection များအနှံ့ တွေ့ရပြီး၊ ၎င်းတို့အနက် **`hasFree: true` ရှာဖွေတွေ့ရှိမှု metadata ပါဝင်သော 153 ခု** ရှိသည်။ Chat model registry တွင် **ပံ့ပိုးသူ 239 ခု / ကွဲပြားသော provider-model pair 3,059 ခု / မူရင်း model ID 1,475 ခု** ပါဝင်သည်။ သီးခြား free-budget catalog တွင် **per-model row 471 ခု**၊ **ထပ်တလဲလဲအသုံးပြုနိုင်သော pool 34 ခု** နှင့် **ထပ်တလဲလဲအသုံးပြုနိုင်ပြီး key မလိုသော အမြဲတမ်းအခမဲ့ ပံ့ပိုးသူ 52 ခု** ပါဝင်သည်။ ဤကိန်းဂဏန်းများသည် ရည်ရွယ်ချက်ရှိရှိ denominator မတူညီကြခြင်းဖြစ်ပြီး၊ အဓိပ္ပာယ်ဖွင့်ဆိုချက်များနှင့် pool များကို ထပ်မတွက်ဘဲ တွက်ချက်ထားသော အချက်အလက်များကို [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) နှင့် [Free Tiers](docs/reference/FREE_TIERS.md) တွင် ကြည့်ရှုနိုင်သည်။ +> **မှတ်ပုံတင်ထားသော ပံ့ပိုးသူ 384 ခု** ကို canonical chat၊ media၊ search၊ local၊ cloud-agent နှင့် system collection များအနှံ့ တွေ့ရပြီး၊ ၎င်းတို့အနက် **`hasFree: true` ရှာဖွေတွေ့ရှိမှု metadata ပါဝင်သော 153 ခု** ရှိသည်။ Chat model registry တွင် **ပံ့ပိုးသူ 239 ခု / ကွဲပြားသော provider-model pair 3,059 ခု / မူရင်း model ID 1,475 ခု** ပါဝင်သည်။ သီးခြား free-budget catalog တွင် **per-model row 471 ခု**၊ **ထပ်တလဲလဲအသုံးပြုနိုင်သော pool 34 ခု** နှင့် **ထပ်တလဲလဲအသုံးပြုနိုင်ပြီး key မလိုသော အမြဲတမ်းအခမဲ့ ပံ့ပိုးသူ 52 ခု** ပါဝင်သည်။ ဤကိန်းဂဏန်းများသည် ရည်ရွယ်ချက်ရှိရှိ denominator မတူညီကြခြင်းဖြစ်ပြီး၊ အဓိပ္ပာယ်ဖွင့်ဆိုချက်များနှင့် pool များကို ထပ်မတွက်ဘဲ တွက်ချက်ထားသော အချက်အလက်များကို [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) နှင့် [Free Tiers](docs/reference/FREE_TIERS.md) တွင် ကြည့်ရှုနိုင်သည်။
@@ -1297,7 +1297,7 @@ port တစ်ခုတည်းရှိ process တစ်ခုတည်းက လုပ်ဆောင်မှုပတ်ဝန်းကျင်Node.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ဘာသာစကားTypeScript 6.0 — src/ နှင့် open-sse/ တစ်လျှောက် TypeScript 100% (v2.0 မှစ၍ core တွင် any လုံးဝမရှိ) ဖရိမ်းဝေါ့ခ်Next.js 16 + React 19 + Tailwind CSS 4 - ဒေတာဘေ့စ်better-sqlite3 (SQLite၊ WAL မှတ်တမ်းရေးသားခြင်း) + LowDB (JSON အမွေအနှစ်စနစ်) — ဒိုမိန်းမော်ဂျူး 147 ခု၊ မိုင်ဂရေးရှင်း 230 ခု + ဒေတာဘေ့စ်better-sqlite3 (SQLite၊ WAL မှတ်တမ်းရေးသားခြင်း) + LowDB (JSON အမွေအနှစ်စနစ်) — ဒိုမိန်းမော်ဂျူး 147 ခု၊ မိုင်ဂရေးရှင်း 232 ခု မှတ်ဉာဏ်SQLite FTS5 စာသားအပြည့်အစုံရှာဖွေမှု + int8-quantized vector embeddings၊ အမျိုးအစားသတ်မှတ်ထားသော decay စကီးမားများZod 4 — MCP tool I/O validation + API contracts ပရိုတိုကောများMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/my/llm.txt b/docs/i18n/my/llm.txt index a411ff410a6..6fdf8d5706d 100644 --- a/docs/i18n/my/llm.txt +++ b/docs/i18n/my/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ne/README.md b/docs/i18n/ne/README.md index 62839cea531..8e33f014e1e 100644 --- a/docs/i18n/ne/README.md +++ b/docs/i18n/ne/README.md @@ -15,7 +15,7 @@ **[Diego Souza](https://github.com/diegosouzapw) द्वारा निर्मित** -OmniRoute — कोडिङ कहिल्यै नरोक्नुहोस्। प्रत्येक AI उपकरण → एउटै endpoint मार्फत 383 प्रदायक — 150+ निःशुल्क —। Claude Code, Codex, Cursor, Cline, Copilot र Antigravity लाई स्वतः fallback सहित निःशुल्क Claude / GPT / Gemini मा। RTK + Caveman को संयुक्त कम्प्रेसनले 15–95% टोकन (~89% औसत) बचत गर्छ — सीमामा कहिल्यै नठोक्किनुहोस्। 383 AI प्रदायक · 150+ निःशुल्क तहहरू · ~1.62B निःशुल्क टोकन/महिना · 20 राउटिङ रणनीतिहरू · सुरु गर्न $0। +OmniRoute — कोडिङ कहिल्यै नरोक्नुहोस्। प्रत्येक AI उपकरण → एउटै endpoint मार्फत 384 प्रदायक — 150+ निःशुल्क —। Claude Code, Codex, Cursor, Cline, Copilot र Antigravity लाई स्वतः fallback सहित निःशुल्क Claude / GPT / Gemini मा। RTK + Caveman को संयुक्त कम्प्रेसनले 15–95% टोकन (~89% औसत) बचत गर्छ — सीमामा कहिल्यै नठोक्किनुहोस्। 384 AI प्रदायक · 150+ निःशुल्क तहहरू · ~1.62B निःशुल्क टोकन/महिना · 20 राउटिङ रणनीतिहरू · सुरु गर्न $0।
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------ | :------------------------: | :---------: | :-------------------: | -| 🌐 प्रदायकहरू | 352 | **358** | 383 | +| 🌐 प्रदायकहरू | 352 | **358** | 384 | | 🧠 अद्वितीय च्याट मोडेल ID हरू | 1320 | **1374** | 1491 | | 🖼️ मोडालिटी ब्रिज | 🆕 दृष्टि + अडियो + भिडियो | ✓ | ✓ | | 📡 राडार निःशुल्क क्याटलग | 🆕 अप्ट-इन | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ सुविधाहरू 🎯 कम्बोहरू - 🌐 प्रदायकहरू + 🌐 प्रदायकहरू 🔌 CLI र MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -प्रतिज्ञा — एउटा endpoint र 383 प्रदायकहरू। अर्को स्वस्थ target उपलब्ध रहँदासम्म स्वचालित fallback ले रुटिङ जारी राख्छ। छ वटा आधारहरू: 383 प्रदायकहरूमा resilient fallback · योग्य workload मा token बचत 95% सम्म · 150+ free tiers र 52 recurring/keyless free-forever प्रदायकहरूसहित सुरु गर्न $0 · एउटा config मार्फत 39 CLI/agent integrations · /v1 मा OpenAI, Claude, Gemini र Responses API compatibility · circuit breakers, TLS stealth, MCP 110 tools, A2A, memory, guardrails, evals, र 7,800+ tracked test files मा 56,000+ static test declarations सहितका production controls। +प्रतिज्ञा — एउटा endpoint र 384 प्रदायकहरू। अर्को स्वस्थ target उपलब्ध रहँदासम्म स्वचालित fallback ले रुटिङ जारी राख्छ। छ वटा आधारहरू: 384 प्रदायकहरूमा resilient fallback · योग्य workload मा token बचत 95% सम्म · 150+ free tiers र 52 recurring/keyless free-forever प्रदायकहरूसहित सुरु गर्न $0 · एउटा config मार्फत 39 CLI/agent integrations · /v1 मा OpenAI, Claude, Gemini र Responses API compatibility · circuit breakers, TLS stealth, MCP 110 tools, A2A, memory, guardrails, evals, र 7,800+ tracked test files मा 56,000+ static test declarations सहितका production controls।

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRoute लाई के कुराले अलग बनाउँछ — 13 वटा क्षमतामा 9router, OpenRouter, CLIProxyAPI र LiteLLM सँगको मिति-आधारित सुविधाको तुलनात्मक स्न्यापसट। OmniRoute: 383 प्रदायक, 150+ निःशुल्क तहहरू समावेश, 20 राउटिङ रणनीतिहरू, 12-इन्जिन टोकन कम्प्रेसन, 110 उपकरणसहितको अन्तर्निर्मित MCP सर्भर, A2A एजेन्ट प्रोटोकल, स्थायी मेमोरी, सुरक्षा नियन्त्रणहरू, क्लाउड एजेन्टहरू, TLS फिङ्गरप्रिन्ट स्टेल्थ, Desktop/Termux/PWA र 67 वटा i18n UI स्थानीय भाषाहरू। OmniRoute MIT-लाइसेन्स प्राप्त र आफैं होस्ट गर्न सकिने छ। प्रतिस्पर्धीहरूको क्षमता र सङ्ख्या परिवर्तन हुन सक्छ; लिङ्क गरिएको विधिशास्त्र हेर्नुहोस्। +OmniRoute लाई के कुराले अलग बनाउँछ — 13 वटा क्षमतामा 9router, OpenRouter, CLIProxyAPI र LiteLLM सँगको मिति-आधारित सुविधाको तुलनात्मक स्न्यापसट। OmniRoute: 384 प्रदायक, 150+ निःशुल्क तहहरू समावेश, 20 राउटिङ रणनीतिहरू, 12-इन्जिन टोकन कम्प्रेसन, 110 उपकरणसहितको अन्तर्निर्मित MCP सर्भर, A2A एजेन्ट प्रोटोकल, स्थायी मेमोरी, सुरक्षा नियन्त्रणहरू, क्लाउड एजेन्टहरू, TLS फिङ्गरप्रिन्ट स्टेल्थ, Desktop/Termux/PWA र 67 वटा i18n UI स्थानीय भाषाहरू। OmniRoute MIT-लाइसेन्स प्राप्त र आफैं होस्ट गर्न सकिने छ। प्रतिस्पर्धीहरूको क्षमता र सङ्ख्या परिवर्तन हुन सक्छ; लिङ्क गरिएको विधिशास्त्र हेर्नुहोस्। 📊 9router, OpenRouter, CLIProxyAPI र LiteLLM सँगको पूर्ण विधिशास्त्र तथा प्रत्येक सुविधाको विस्तृत विवरण → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -613,7 +613,7 @@ Radar वैकल्पिक रूपमा सक्रिय गर्न - **🖼️ नयाँ endpoints** — `/v1/ocr` (Mistral OCR) र `/v1/audio/translations` (Whisper-style) ले media surface पूरा गर्छन्। → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 Image / video / audio generation** — media का लागि एउटै API: xAI Grok Imagine र Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind, तथा ElevenLabs जस्ता speech providers। → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 Deployment र ops** — reverse-proxy `basePath`, browser-language auto-detect, प्रत्येक key का लागि device tracking, root-less MITM trust, zh-TW localization। → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 थप providers र agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login सहित Grok Build (xAI), first-class card सहित Ollama, Claude Opus 5 र Sonnet 5, Kimi official partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… र पुनः तयार गरिएको **383-provider catalog**। → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 थप providers र agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login सहित Grok Build (xAI), first-class card सहित Ollama, Claude Opus 5 र Sonnet 5, Kimi official partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… र पुनः तयार गरिएको **384-provider catalog**। → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 राउटिङ पारदर्शिता** — प्रत्येक response ले त्यसलाई सेवा दिने strategy/provider/latency को नामसहित `X-OmniRoute-Decision` header समावेश गर्छ; नयाँ `cache-optimized` combo strategy र Auto-Combo `cacheAffinity` factor ले repeat requests लाई cached prefix भएको connection मा फर्काउँछ; साथै read-only `/v1/auto-combo/{channel}/candidates` endpoint ले `auto/*` channel को live candidate pool देखाउँछ। → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ स्थानीय कार्यसम्पादन र पूर्वाधार** — एक-क्लिक स्थानीय Redis, Cloudflare Workers / Deno Deploy रिले डिप्लोयरहरू, सुपरभाइज गरिएका एम्बेडेड सेवाका रूपमा Bifrost र Mux। → [एम्बेडेड सेवाहरू](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 प्याकेजमै समावेश** — प्लगइन फ्रेमवर्क + मार्केटप्लेस, Omni/Agent/GitHub स्किल्स फ्रेमवर्कहरू, Obsidian भल्ट एकीकरण (22 MCP उपकरणहरू), OpenAI-compatible Batch र Files APIs, सिमान्टिक प्रतिक्रिया क्यास, लिडरबोर्डसहितको गेमिफिकेसन, ACP एजेन्ट डिस्कभरी (15 निर्मित एजेन्टहरू), BigQuery मा निर्धारित लग निर्यात, `auto/chaos` समानान्तर बहु-मोडेल फ्यान-आउट, Telegram बट ब्रिज, इन-एप संस्करण व्यवस्थापक र LMArena-ELO निःशुल्क-प्रदायक र्याङ्किङहरू। → [दस्तावेज](docs/README.md) @@ -697,11 +697,11 @@ omniroute configure codex # साथै: claude opencode qwen aider go
-## 🌐 383 AI प्रदायक — क्याटलगमा चिह्नित 153 निःशुल्क +## 🌐 384 AI प्रदायक — क्याटलगमा चिह्नित 153 निःशुल्क
-> **383 दर्ता गरिएका प्रदायकहरू** आधिकारिक chat, media, search, local, cloud-agent र system सङ्ग्रहहरूमा समेटिएका छन्, जसमा **`hasFree: true` डिस्कभरी मेटाडेटा भएका 153 प्रदायकहरू** पनि छन्। chat model registry मा **239 प्रदायक / 3,059 विशिष्ट provider-model जोडी / 1,475 कच्चा model ID** समेटिएका छन्; छुट्टै free-budget catalog मा **471 प्रति-मोडेल पङ्क्तिहरू**, **34 पुनरावर्ती पूलहरू** र **52 पुनरावर्ती/कुञ्जीविहीन सधैँ-निःशुल्क प्रदायकहरू** छन्। यी फरक denominator हरू डिजाइनकै अंश हुन्; परिभाषा र pool-deduped गणनाहरू [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) र [Free Tiers](docs/reference/FREE_TIERS.md) मा उपलब्ध छन्। +> **384 दर्ता गरिएका प्रदायकहरू** आधिकारिक chat, media, search, local, cloud-agent र system सङ्ग्रहहरूमा समेटिएका छन्, जसमा **`hasFree: true` डिस्कभरी मेटाडेटा भएका 153 प्रदायकहरू** पनि छन्। chat model registry मा **239 प्रदायक / 3,059 विशिष्ट provider-model जोडी / 1,475 कच्चा model ID** समेटिएका छन्; छुट्टै free-budget catalog मा **471 प्रति-मोडेल पङ्क्तिहरू**, **34 पुनरावर्ती पूलहरू** र **52 पुनरावर्ती/कुञ्जीविहीन सधैँ-निःशुल्क प्रदायकहरू** छन्। यी फरक denominator हरू डिजाइनकै अंश हुन्; परिभाषा र pool-deduped गणनाहरू [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) र [Free Tiers](docs/reference/FREE_TIERS.md) मा उपलब्ध छन्।
@@ -1303,7 +1303,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 ले पनि रनटाइमNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 भाषाTypeScript 6.0 — 100% TypeScript src/ र open-sse/ मा (v2.0 देखि कोरमा शून्य any) फ्रेमवर्कNext.js 16 + React 19 + Tailwind CSS 4 - डाटाबेसbetter-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — 147 डोमेन मोड्युल, 230 माइग्रेसन + डाटाबेसbetter-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — 147 डोमेन मोड्युल, 232 माइग्रेसन मेमोरीSQLite FTS5 पूर्ण-पाठ + int8-quantized vector embeddings, typed decay स्किमाZod 4 — MCP टुल I/O प्रमाणीकरण + API contracts प्रोटोकलMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ne/llm.txt b/docs/i18n/ne/llm.txt index 9392f42a070..77efd1e2bcd 100644 --- a/docs/i18n/ne/llm.txt +++ b/docs/i18n/ne/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/nl/README.md b/docs/i18n/nl/README.md index 15457f5e03c..042e5658cd7 100644 --- a/docs/i18n/nl/README.md +++ b/docs/i18n/nl/README.md @@ -15,7 +15,7 @@ **Gemaakt door [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Stop nooit met coderen. Elke AI-tool → 383 providers — 150+ gratis — via één endpoint. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity naar GRATIS Claude / GPT / Gemini met automatische fallback. RTK + Caveman gestapelde compressie bespaart 15–95% tokens (~89% gemiddeld) — bereik nooit de limieten. 383 AI-providers · 150+ gratis niveaus · ~1,62 miljard gratis tokens/maand · 20 routeringsstrategieën · $0 om te beginnen. +OmniRoute — Stop nooit met coderen. Elke AI-tool → 384 providers — 150+ gratis — via één endpoint. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity naar GRATIS Claude / GPT / Gemini met automatische fallback. RTK + Caveman gestapelde compressie bespaart 15–95% tokens (~89% gemiddeld) — bereik nooit de limieten. 384 AI-providers · 150+ gratis niveaus · ~1,62 miljard gratis tokens/maand · 20 routeringsstrategieën · $0 om te beginnen.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------- | :----------------------: | :---------: | :-------------------: | -| 🌐 Aanbieders | 352 | **358** | 383 | +| 🌐 Aanbieders | 352 | **358** | 384 | | 🧠 Unieke chatmodel-ID's | 1320 | **1374** | 1491 | | 🖼️ Modaliteitsbrug | 🆕 visie + audio + video | ✓ | ✓ | | 📡 Gratis radarcatalogus | 🆕 opt-in | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Functies 🎯 Combinaties - 🌐 Providers + 🌐 Providers 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -De belofte — Eén endpoint en 383 providers. Automatische fallback houdt de routing actief zolang er een andere gezonde target beschikbaar is. Zes pijlers: veerkrachtige fallback over 383 providers · tot 95% tokenbesparing voor geschikte workloads · $0 om te starten met 150+ gratis tiers en 52 terugkerende/keyless providers die voor altijd gratis zijn · 39 CLI/agent-integraties via één configuratie · compatibiliteit met OpenAI, Claude, Gemini en de Responses API op /v1 · productiecontroles, waaronder circuit breakers, TLS stealth, MCP met 110 tools, A2A, geheugen, guardrails, evals en meer dan 56.000 statische testdeclaraties verspreid over meer dan 7.800 bijgehouden testbestanden. +De belofte — Eén endpoint en 384 providers. Automatische fallback houdt de routing actief zolang er een andere gezonde target beschikbaar is. Zes pijlers: veerkrachtige fallback over 384 providers · tot 95% tokenbesparing voor geschikte workloads · $0 om te starten met 150+ gratis tiers en 52 terugkerende/keyless providers die voor altijd gratis zijn · 39 CLI/agent-integraties via één configuratie · compatibiliteit met OpenAI, Claude, Gemini en de Responses API op /v1 · productiecontroles, waaronder circuit breakers, TLS stealth, MCP met 110 tools, A2A, geheugen, guardrails, evals en meer dan 56.000 statische testdeclaraties verspreid over meer dan 7.800 bijgehouden testbestanden.

@@ -510,7 +510,7 @@ Alle **20** strategieën — naar wens te combineren per combostap: -Wat OmniRoute onderscheidt — een gedateerde functiesnapshot versus 9router, OpenRouter, CLIProxyAPI en LiteLLM voor 13 mogelijkheden. OmniRoute: 383 providers, meer dan 150 ingebouwde gratis tiers, 20 routeringsstrategieën, tokencompressie met 12 engines, een ingebouwde MCP-server met 110 tools, het A2A-agentprotocol, persistent geheugen, guardrails, cloudagents, TLS-fingerprint-stealth, Desktop/Termux/PWA en 67 i18n-UI-lokalisaties. OmniRoute heeft een MIT-licentie en kan zelf worden gehost. De mogelijkheden en aantallen van concurrenten kunnen veranderen; zie de gekoppelde methodologie. +Wat OmniRoute onderscheidt — een gedateerde functiesnapshot versus 9router, OpenRouter, CLIProxyAPI en LiteLLM voor 13 mogelijkheden. OmniRoute: 384 providers, meer dan 150 ingebouwde gratis tiers, 20 routeringsstrategieën, tokencompressie met 12 engines, een ingebouwde MCP-server met 110 tools, het A2A-agentprotocol, persistent geheugen, guardrails, cloudagents, TLS-fingerprint-stealth, Desktop/Termux/PWA en 67 i18n-UI-lokalisaties. OmniRoute heeft een MIT-licentie en kan zelf worden gehost. De mogelijkheden en aantallen van concurrenten kunnen veranderen; zie de gekoppelde methodologie. 📊 Volledige methodologie & details per functie versus 9router, OpenRouter, CLIProxyAPI & LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar is opt-in en gebruikt alleen GET. De OmniRoute-client uploadt geen prompts - **🖼️ Nieuwe endpoints** — `/v1/ocr` (Mistral OCR) en `/v1/audio/translations` (Whisper-achtig) completeren het media-aanbod. → [API-referentie](docs/reference/API_REFERENCE.md) - **🎨 Generatie van afbeeldingen, video en audio** — één API voor media: xAI Grok Imagine en Novita AI-video, ComfyUI, Magnific, Adobe Firefly, Segmind en spraakproviders zoals ElevenLabs. → [API-referentie](docs/reference/API_REFERENCE.md) - **🌍 Implementatie en beheer** — reverse-proxy-`basePath`, automatische detectie van de browsertaal, apparaatregistratie per sleutel, MITM-vertrouwen zonder root en zh-TW-lokalisatie. → [Omgeving](docs/reference/ENVIRONMENT.md) -- **🤝 Meer providers en agents** — cloudagents (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) met browser- en OAuth-aanmelding, een volwaardige Ollama-kaart, Claude Opus 5 en Sonnet 5, officiële samenwerking met Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… en een vernieuwde **383-providercatalogus**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Meer providers en agents** — cloudagents (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) met browser- en OAuth-aanmelding, een volwaardige Ollama-kaart, Claude Opus 5 en Sonnet 5, officiële samenwerking met Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… en een vernieuwde **384-providercatalogus**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparantie van routering** — elke respons bevat een `X-OmniRoute-Decision`-header met de strategie, provider en latentie die de respons hebben geleverd; een nieuwe `cache-optimized`-combinatiestrategie en de `cacheAffinity`-factor van Auto-Combo sturen herhaalde verzoeken terug naar de verbinding met het gecachte prefix; en een alleen-lezen endpoint `/v1/auto-combo/{channel}/candidates` maakt de live kandidatenpool van een `auto/*`-kanaal zichtbaar. → [Auto-Combo](docs/routing/AUTO-COMBO.md). - **⚡ Lokale prestaties & infrastructuur** — lokale Redis met één klik, relay-deployers voor Cloudflare Workers / Deno Deploy, Bifrost & Mux als bewaakte ingesloten services. → [Embedded Services](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Ook inbegrepen** — pluginframework + marketplace, Omni/Agent/GitHub-skillsframeworks, Obsidian-kluisintegratie (22 MCP-tools), OpenAI-compatibele Batch- & Files-API's, semantische responscache, gamification met ranglijsten, ACP-agentdetectie (15 ingebouwde agents), geplande logexport naar BigQuery, parallelle multi-model-fan-out via `auto/chaos`, een Telegram-botbridge, een ingebouwde versiebeheerder en gratis-provider-ranglijsten op basis van LMArena-ELO. → [Documentatie](docs/README.md) @@ -689,11 +689,11 @@ je shellgeschiedenis. → [CLI-integraties](docs/guides/CLI-INTEGRATIONS.md)
-## 🌐 383 AI-providers — 153 in de catalogus als gratis gemarkeerd +## 🌐 384 AI-providers — 153 in de catalogus als gratis gemarkeerd
-> **383 geregistreerde providers** verspreid over de canonieke collecties voor chat, media, zoeken, lokaal, cloud-agents en systemen, waaronder **153 met `hasFree: true`-metadata voor ontdekking**. Het chatmodelregister omvat **239 providers / 3.059 unieke provider-modelparen / 1.475 onbewerkte model-ID's**; de afzonderlijke catalogus voor gratis budgetten bevat **471 rijen per model**, **34 terugkerende pools** en **52 terugkerende/keyless providers die voor altijd gratis zijn**. Dit zijn opzettelijk verschillende noemers; definities en pool-ontdubbeld berekeningen staan in de [Providerreferentie](docs/reference/PROVIDER_REFERENCE.md) en [Gratis niveaus](docs/reference/FREE_TIERS.md). +> **384 geregistreerde providers** verspreid over de canonieke collecties voor chat, media, zoeken, lokaal, cloud-agents en systemen, waaronder **153 met `hasFree: true`-metadata voor ontdekking**. Het chatmodelregister omvat **239 providers / 3.059 unieke provider-modelparen / 1.475 onbewerkte model-ID's**; de afzonderlijke catalogus voor gratis budgetten bevat **471 rijen per model**, **34 terugkerende pools** en **52 terugkerende/keyless providers die voor altijd gratis zijn**. Dit zijn opzettelijk verschillende noemers; definities en pool-ontdubbeld berekeningen staan in de [Providerreferentie](docs/reference/PROVIDER_REFERENCE.md) en [Gratis niveaus](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ Canonieke statistieken op 2026-10-18: **1.976 unieke video's** · **20.220.289 b RuntimeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 TaalTypeScript 6.0 — 100% TypeScript in src/ en open-sse/ (geen any in de core sinds v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Databasebetter-sqlite3 (SQLite, WAL-journaling) + LowDB (JSON-legacy) — 147 domeinmodules, 230 migraties + Databasebetter-sqlite3 (SQLite, WAL-journaling) + LowDB (JSON-legacy) — 147 domeinmodules, 232 migraties GeheugenSQLite FTS5-volledige-tekstzoekopdracht + int8-gekwantiseerde vector-embeddings, getypeerde vervaging SchemataZod 4 — validatie van MCP-tool-I/O + API-contracten ProtocollenMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/nl/llm.txt b/docs/i18n/nl/llm.txt index 48a2a8d466f..e92d87e0e2c 100644 --- a/docs/i18n/nl/llm.txt +++ b/docs/i18n/nl/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/no/README.md b/docs/i18n/no/README.md index cb6ef698562..7bc09a8e8e4 100644 --- a/docs/i18n/no/README.md +++ b/docs/i18n/no/README.md @@ -15,7 +15,7 @@ **Laget av [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Slutt aldri å programmere. Alle AI-verktøy → 383 leverandører — 150+ kostnadsfrie — gjennom ett endepunkt. Claude Code, Codex, Cursor, Cline, Copilot og Antigravity til GRATIS Claude / GPT / Gemini med automatisk reservevalg. RTK + Caveman-komprimert lagring sparer 15–95 % tokens (~89 % i gjennomsnitt) — nå aldri grensene. 383 AI-leverandører · 150+ kostnadsfrie nivåer · ~1,62 mrd. kostnadsfrie tokens/mnd. · 20 rutingsstrategier · $0 for å starte. +OmniRoute — Slutt aldri å programmere. Alle AI-verktøy → 384 leverandører — 150+ kostnadsfrie — gjennom ett endepunkt. Claude Code, Codex, Cursor, Cline, Copilot og Antigravity til GRATIS Claude / GPT / Gemini med automatisk reservevalg. RTK + Caveman-komprimert lagring sparer 15–95 % tokens (~89 % i gjennomsnitt) — nå aldri grensene. 384 AI-leverandører · 150+ kostnadsfrie nivåer · ~1,62 mrd. kostnadsfrie tokens/mnd. · 20 rutingsstrategier · $0 for å starte.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------- | :------------------: | :---------: | :-------------------: | -| 🌐 Leverandører | 352 | **358** | 383 | +| 🌐 Leverandører | 352 | **358** | 384 | | 🧠 Unike chatmodell-ID-er | 1320 | **1374** | 1491 | | 🖼️ Modalitetsbro | 🆕 syn + lyd + video | ✓ | ✓ | | 📡 Radars gratiskatalog | 🆕 valgfri | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funksjoner 🎯 Kombinasjoner - 🌐 Leverandører + 🌐 Leverandører 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Løftet — Ett endepunkt og 383 leverandører. Automatisk reserveomkobling opprettholder rutingen så lenge et annet tilgjengelig og friskt mål finnes. Seks søyler: robust reserveomkobling på tvers av 383 leverandører · opptil 95 % tokenbesparelse på kvalifiserte arbeidsbelastninger · 0 USD for å komme i gang med over 150 gratistilbud og 52 tilbakevendende eller nøkkelløse gratistilbud for alltid · 39 CLI-/agentintegrasjoner gjennom én konfigurasjon · kompatibilitet med OpenAI-, Claude-, Gemini- og Responses API på /v1 · produksjonskontroller, inkludert kretsbrytere, TLS-stealth, MCP med 110 verktøy, A2A, minne, sikkerhetsregler, evalueringer og over 56 000 statiske testdeklarasjoner på tvers av over 7 800 sporede testfiler. +Løftet — Ett endepunkt og 384 leverandører. Automatisk reserveomkobling opprettholder rutingen så lenge et annet tilgjengelig og friskt mål finnes. Seks søyler: robust reserveomkobling på tvers av 384 leverandører · opptil 95 % tokenbesparelse på kvalifiserte arbeidsbelastninger · 0 USD for å komme i gang med over 150 gratistilbud og 52 tilbakevendende eller nøkkelløse gratistilbud for alltid · 39 CLI-/agentintegrasjoner gjennom én konfigurasjon · kompatibilitet med OpenAI-, Claude-, Gemini- og Responses API på /v1 · produksjonskontroller, inkludert kretsbrytere, TLS-stealth, MCP med 110 verktøy, A2A, minne, sikkerhetsregler, evalueringer og over 56 000 statiske testdeklarasjoner på tvers av over 7 800 sporede testfiler.

@@ -510,7 +510,7 @@ Alle **20** strategiene — kombiner fritt per combo-trinn: -Hva som skiller OmniRoute fra andre — et datert funksjonsøyeblikksbilde sammenlignet med 9router, OpenRouter, CLIProxyAPI og LiteLLM på tvers av 13 funksjoner. OmniRoute: 383 leverandører, over 150 innebygde gratiskvoter, 20 rutingsstrategier, tokenkomprimering med 12 motorer, innebygd MCP-server med 110 verktøy, A2A-agentprotokoll, persistent minne, sikkerhetsregler, skyagenter, TLS-fingeravtrykksskjuling, Desktop/Termux/PWA og 67 i18n-brukergrensesnittlokaler. OmniRoute er lisensiert under MIT og kan driftes selv. Konkurrentenes funksjoner og antall kan endres; se den lenkede metoden. +Hva som skiller OmniRoute fra andre — et datert funksjonsøyeblikksbilde sammenlignet med 9router, OpenRouter, CLIProxyAPI og LiteLLM på tvers av 13 funksjoner. OmniRoute: 384 leverandører, over 150 innebygde gratiskvoter, 20 rutingsstrategier, tokenkomprimering med 12 motorer, innebygd MCP-server med 110 verktøy, A2A-agentprotokoll, persistent minne, sikkerhetsregler, skyagenter, TLS-fingeravtrykksskjuling, Desktop/Termux/PWA og 67 i18n-brukergrensesnittlokaler. OmniRoute er lisensiert under MIT og kan driftes selv. Konkurrentenes funksjoner og antall kan endres; se den lenkede metoden. 📊 Full metodebeskrivelse og detaljer per funksjon sammenlignet med 9router, OpenRouter, CLIProxyAPI og LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -614,7 +614,7 @@ den gjeldende katalogen på **[radar.omniroute.online/planos](https://radar.omni - **🖼️ Nye endepunkter** — `/v1/ocr` (Mistral OCR) og `/v1/audio/translations` (Whisper-stil) kompletterer medieoverflaten. → [API-referanse](docs/reference/API_REFERENCE.md) - **🎨 Generering av bilder, video og lyd** — ett API for medier: xAI Grok Imagine og Novita AI-video, ComfyUI, Magnific, Adobe Firefly, Segmind og taleleverandører som ElevenLabs. → [API-referanse](docs/reference/API_REFERENCE.md) - **🌍 Distribusjon og drift** — reverse-proxy-`basePath`, automatisk gjenkjenning av nettleserspråk, enhetssporing per nøkkel, MITM-tillit uten root og zh-TW-lokalisering. → [Miljø](docs/reference/ENVIRONMENT.md) -- **🤝 Flere leverandører og agenter** — skyagenter (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) med nettleser- og OAuth-innlogging, førsteklasses Ollama-kort, Claude Opus 5 og Sonnet 5, offisielt Kimi-partnerskap (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI … og en oppdatert **383-leverandørskatalog**. → [Leverandører](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Flere leverandører og agenter** — skyagenter (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) med nettleser- og OAuth-innlogging, førsteklasses Ollama-kort, Claude Opus 5 og Sonnet 5, offisielt Kimi-partnerskap (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI … og en oppdatert **384-leverandørskatalog**. → [Leverandører](docs/reference/PROVIDER_REFERENCE.md) - **📡 Åpenhet i ruting** — alle svar inneholder en `X-OmniRoute-Decision`-header som angir strategien, leverandøren og forsinkelsen som leverte svaret, en ny `cache-optimized`-kombinasjonsstrategi og faktoren `cacheAffinity` i Auto-Combo sender gjentatte forespørsler tilbake til tilkoblingen som har det bufrede prefikset, og et skrivebeskyttet `/v1/auto-combo/{channel}/candidates`-endepunkt viser en `auto/*`-kanals aktive kandidatpool. → [Auto-Combo](docs/routing/AUTO-COMBO.md). - **⚡ Lokal ytelse og infrastruktur** — Redis lokalt med ett klikk, reléimplementeringer for Cloudflare Workers / Deno Deploy, Bifrost og Mux som overvåkede innebygde tjenester. → [Innebygde tjenester](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Også inkludert** — plugin-rammeverk og markedsplass, Omni/Agent/GitHub-ferdighetsrammeverk, Obsidian-hvelvintegrasjon (22 MCP-verktøy), OpenAI-kompatible Batch- og Files-API-er, semantisk responsbuffer, spillifisering med ledertavler, ACP-agentoppdagelse (15 innebygde agenter), planlagt loggeksport til BigQuery, `auto/chaos` parallell fan-out til flere modeller, en Telegram-botbro, en versjonsbehandler i appen og LMArena-ELO-rangeringer for gratisleverandører. → [Dokumentasjon](docs/README.md) @@ -698,11 +698,11 @@ av shell-historikken din. → [CLI-integrasjoner](docs/guides/CLI-INTEGRATIONS.m
-## 🌐 383 AI-leverandører — 153 katalogmerkede gratis +## 🌐 384 AI-leverandører — 153 katalogmerkede gratis
-> **383 registrerte leverandører** på tvers av de kanoniske samlingene for chat, medier, søk, lokale tjenester, skyagenter og systemer, inkludert **153 med `hasFree: true`-metadata for oppdagelse**. Chatmodellregisteret dekker **239 leverandører / 3 059 unike leverandør-modell-par / 1 475 rå modell-ID-er**; den separate katalogen for gratisbudsjetter har **471 rader per modell**, **34 tilbakevendende kvoter** og **52 tilbakevendende/gratis-for-alltid-leverandører uten nøkkel**. Dette er ulike nevnerverdier med hensikt; definisjoner og beregninger deduplisert etter kvote finnes i [Leverandørreferansen](docs/reference/PROVIDER_REFERENCE.md) og [Gratisnivåer](docs/reference/FREE_TIERS.md). +> **384 registrerte leverandører** på tvers av de kanoniske samlingene for chat, medier, søk, lokale tjenester, skyagenter og systemer, inkludert **153 med `hasFree: true`-metadata for oppdagelse**. Chatmodellregisteret dekker **239 leverandører / 3 059 unike leverandør-modell-par / 1 475 rå modell-ID-er**; den separate katalogen for gratisbudsjetter har **471 rader per modell**, **34 tilbakevendende kvoter** og **52 tilbakevendende/gratis-for-alltid-leverandører uten nøkkel**. Dette er ulike nevnerverdier med hensikt; definisjoner og beregninger deduplisert etter kvote finnes i [Leverandørreferansen](docs/reference/PROVIDER_REFERENCE.md) og [Gratisnivåer](docs/reference/FREE_TIERS.md).
@@ -1304,7 +1304,7 @@ Kanoniske måledata per 2026-10-18: **1.976 unike videoer** · **20.220.289 kjen KjøretidNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 SpråkTypeScript 6.0 — 100 % TypeScript i src/ og open-sse/ (ingen any i kjernen siden v2.0) RammeverkNext.js 16 + React 19 + Tailwind CSS 4 - Databasebetter-sqlite3 (SQLite, WAL-journalføring) + LowDB (JSON-legacy) — 147 domenemoduler, 230 migreringer + Databasebetter-sqlite3 (SQLite, WAL-journalføring) + LowDB (JSON-legacy) — 147 domenemoduler, 232 migreringer MinneSQLite FTS5-fulltekstsøk + int8-kvantisert vektorinbedding, typet henfall SkjemaerZod 4 — validering av MCP-verktøy-I/O + API-kontrakter ProtokollerMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/no/llm.txt b/docs/i18n/no/llm.txt index 18794417884..10df3b9c3dc 100644 --- a/docs/i18n/no/llm.txt +++ b/docs/i18n/no/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/or/README.md b/docs/i18n/or/README.md index 6b3a6942dae..5352d54b5e7 100644 --- a/docs/i18n/or/README.md +++ b/docs/i18n/or/README.md @@ -15,7 +15,7 @@ **[Diego Souza](https://github.com/diegosouzapw)ଙ୍କ ଦ୍ୱାରା ସୃଷ୍ଟି** -OmniRoute — କେବେ କୋଡିଂ ବନ୍ଦ କରନ୍ତୁ ନାହିଁ। ପ୍ରତ୍ୟେକ AI ଟୁଲ୍ → 383 ପ୍ରଦାନକାରୀ — 150+ ମାଗଣା — ଗୋଟିଏ ଏଣ୍ଡପଏଣ୍ଟ ମାଧ୍ୟମରେ। Claude Code, Codex, Cursor, Cline, Copilot ଏବଂ Antigravityକୁ ସ୍ୱୟଂଚାଳିତ ଫଲ୍ବ୍ୟାକ୍ ସହିତ ମାଗଣା Claude / GPT / Geminiରେ ପରିଣତ କରନ୍ତୁ। RTK + Caveman ଷ୍ଟ୍ୟାକ୍ଡ କମ୍ପ୍ରେସନ୍ 15–95% ଟୋକେନ୍ (~89% ହାରାହାରି) ସଞ୍ଚୟ କରେ — କେବେ ସୀମାରେ ପହଞ୍ଚିବେ ନାହିଁ। 383 AI ପ୍ରଦାନକାରୀ · 150+ ମାଗଣା ଟିଅର୍ · ପ୍ରତି ମାସରେ ~1.62B ମାଗଣା ଟୋକେନ୍ · 20ଟି ରାଉଟିଂ କୌଶଳ · ଆରମ୍ଭ ପାଇଁ $0। +OmniRoute — କେବେ କୋଡିଂ ବନ୍ଦ କରନ୍ତୁ ନାହିଁ। ପ୍ରତ୍ୟେକ AI ଟୁଲ୍ → 384 ପ୍ରଦାନକାରୀ — 150+ ମାଗଣା — ଗୋଟିଏ ଏଣ୍ଡପଏଣ୍ଟ ମାଧ୍ୟମରେ। Claude Code, Codex, Cursor, Cline, Copilot ଏବଂ Antigravityକୁ ସ୍ୱୟଂଚାଳିତ ଫଲ୍ବ୍ୟାକ୍ ସହିତ ମାଗଣା Claude / GPT / Geminiରେ ପରିଣତ କରନ୍ତୁ। RTK + Caveman ଷ୍ଟ୍ୟାକ୍ଡ କମ୍ପ୍ରେସନ୍ 15–95% ଟୋକେନ୍ (~89% ହାରାହାରି) ସଞ୍ଚୟ କରେ — କେବେ ସୀମାରେ ପହଞ୍ଚିବେ ନାହିଁ। 384 AI ପ୍ରଦାନକାରୀ · 150+ ମାଗଣା ଟିଅର୍ · ପ୍ରତି ମାସରେ ~1.62B ମାଗଣା ଟୋକେନ୍ · 20ଟି ରାଉଟିଂ କୌଶଳ · ଆରମ୍ଭ ପାଇଁ $0।
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | -------------------------------- | :---------------------: | :---------: | :-------------------: | -| 🌐 ପ୍ରୋଭାଇଡର୍ଗୁଡ଼ିକ | 352 | **358** | 383 | +| 🌐 ପ୍ରୋଭାଇଡର୍ଗୁଡ଼ିକ | 352 | **358** | 384 | | 🧠 ସ୍ୱତନ୍ତ୍ର ଚାଟ୍ ମଡେଲ୍ IDଗୁଡ଼ିକ | 1320 | **1374** | 1491 | | 🖼️ ମୋଡାଲିଟି ବ୍ରିଜ୍ | 🆕 ଭିଜନ୍ + ଅଡିଓ + ଭିଡିଓ | ✓ | ✓ | | 📡 ରାଡାର୍ ମାଗଣା କ୍ୟାଟାଲଗ୍ | 🆕 ଅପ୍ଟ-ଇନ୍ | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ ବୈଶିଷ୍ଟ୍ୟଗୁଡ଼ିକ 🎯 କମ୍ବୋଗୁଡ଼ିକ - 🌐 ପ୍ରଦାତାମାନେ + 🌐 ପ୍ରଦାତାମାନେ 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -ପ୍ରତିଶ୍ରୁତି — ଗୋଟିଏ ଏଣ୍ଡପଏଣ୍ଟ ଏବଂ 383 ପ୍ରଦାତା। ଅନ୍ୟ ଏକ ସୁସ୍ଥ ଲକ୍ଷ୍ୟ ଉପଲବ୍ଧ ଥିବା ପର୍ଯ୍ୟନ୍ତ ସ୍ୱୟଂଚାଳିତ ଫଲ୍ବ୍ୟାକ୍ ରାଉଟିଂ ଜାରି ରଖେ। ଛଅଟି ସ୍ତମ୍ଭ: 383 ପ୍ରଦାତା ମଧ୍ୟରେ ସୁଦୃଢ଼ ଫଲ୍ବ୍ୟାକ୍ · ଯୋଗ୍ୟ ୱର୍କଲୋଡ୍ରେ 95% ପର୍ଯ୍ୟନ୍ତ ଟୋକେନ୍ ସଞ୍ଚୟ · 150+ ମାଗଣା ଟିଅର୍ ଏବଂ 52ଟି ପୁନରାବୃତ୍ତିଶୀଳ/କୀ-ବିହୀନ ସର୍ବଦା-ମାଗଣା ପ୍ରଦାତା ସହିତ ଆରମ୍ଭ କରିବାକୁ $0 · ଗୋଟିଏ କନ୍ଫିଗ୍ ମାଧ୍ୟମରେ 39ଟି CLI/ଏଜେଣ୍ଟ ଇଣ୍ଟିଗ୍ରେସନ୍ · /v1 ରେ OpenAI, Claude, Gemini ଏବଂ Responses API ସହିତ ସୁସଙ୍ଗତତା · ସର୍କିଟ୍ ବ୍ରେକର୍, TLS ଷ୍ଟେଲ୍ଥ୍, MCP 110 ଟୁଲ୍, A2A, ମେମୋରୀ, ଗାର୍ଡରେଲ୍, ଇଭାଲ୍ସ ଏବଂ 7,800+ ଟ୍ରାକ୍ ହୋଇଥିବା ଟେଷ୍ଟ ଫାଇଲ୍ରେ 56,000+ ଷ୍ଟାଟିକ୍ ଟେଷ୍ଟ ଘୋଷଣା ସହିତ ପ୍ରଡକ୍ସନ୍ ନିୟନ୍ତ୍ରଣ। +ପ୍ରତିଶ୍ରୁତି — ଗୋଟିଏ ଏଣ୍ଡପଏଣ୍ଟ ଏବଂ 384 ପ୍ରଦାତା। ଅନ୍ୟ ଏକ ସୁସ୍ଥ ଲକ୍ଷ୍ୟ ଉପଲବ୍ଧ ଥିବା ପର୍ଯ୍ୟନ୍ତ ସ୍ୱୟଂଚାଳିତ ଫଲ୍ବ୍ୟାକ୍ ରାଉଟିଂ ଜାରି ରଖେ। ଛଅଟି ସ୍ତମ୍ଭ: 384 ପ୍ରଦାତା ମଧ୍ୟରେ ସୁଦୃଢ଼ ଫଲ୍ବ୍ୟାକ୍ · ଯୋଗ୍ୟ ୱର୍କଲୋଡ୍ରେ 95% ପର୍ଯ୍ୟନ୍ତ ଟୋକେନ୍ ସଞ୍ଚୟ · 150+ ମାଗଣା ଟିଅର୍ ଏବଂ 52ଟି ପୁନରାବୃତ୍ତିଶୀଳ/କୀ-ବିହୀନ ସର୍ବଦା-ମାଗଣା ପ୍ରଦାତା ସହିତ ଆରମ୍ଭ କରିବାକୁ $0 · ଗୋଟିଏ କନ୍ଫିଗ୍ ମାଧ୍ୟମରେ 39ଟି CLI/ଏଜେଣ୍ଟ ଇଣ୍ଟିଗ୍ରେସନ୍ · /v1 ରେ OpenAI, Claude, Gemini ଏବଂ Responses API ସହିତ ସୁସଙ୍ଗତତା · ସର୍କିଟ୍ ବ୍ରେକର୍, TLS ଷ୍ଟେଲ୍ଥ୍, MCP 110 ଟୁଲ୍, A2A, ମେମୋରୀ, ଗାର୍ଡରେଲ୍, ଇଭାଲ୍ସ ଏବଂ 7,800+ ଟ୍ରାକ୍ ହୋଇଥିବା ଟେଷ୍ଟ ଫାଇଲ୍ରେ 56,000+ ଷ୍ଟାଟିକ୍ ଟେଷ୍ଟ ଘୋଷଣା ସହିତ ପ୍ରଡକ୍ସନ୍ ନିୟନ୍ତ୍ରଣ।

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRouteକୁ କ’ଣ ସ୍ୱତନ୍ତ୍ର କରେ — 13ଟି କ୍ଷମତାରେ 9router, OpenRouter, CLIProxyAPI ଏବଂ LiteLLM ସହ ତାରିଖ-ସହିତ ବୈଶିଷ୍ଟ୍ୟ ତୁଳନା। OmniRoute: 383ଟି ପ୍ରଦାତା, ଅନ୍ତର୍ନିର୍ମିତ 150+ ମାଗଣା ଟିଅର୍, 20ଟି ରାଉଟିଂ କୌଶଳ, 12-ଇଞ୍ଜିନ୍ ଟୋକେନ୍ ସଙ୍କୋଚନ, 110ଟି ଟୁଲ୍ ସହ ଅନ୍ତର୍ନିର୍ମିତ MCP ସର୍ଭର, A2A ଏଜେଣ୍ଟ ପ୍ରୋଟୋକଲ୍, ସ୍ଥାୟୀ ମେମୋରି, ଗାର୍ଡରେଲ୍, କ୍ଲାଉଡ୍ ଏଜେଣ୍ଟ, TLS ଫିଙ୍ଗରପ୍ରିଣ୍ଟ ଷ୍ଟିଲ୍ଥ୍, Desktop/Termux/PWA ଏବଂ 67ଟି i18n UI ଲୋକେଲ୍। OmniRoute MIT-ଲାଇସେନ୍ସପ୍ରାପ୍ତ ଏବଂ ନିଜେ ହୋଷ୍ଟ କରାଯାଇପାରେ। ପ୍ରତିଦ୍ୱନ୍ଦ୍ୱୀମାନଙ୍କର କ୍ଷମତା ଏବଂ ସଂଖ୍ୟା ବଦଳିପାରେ; ଲିଙ୍କ୍ କରାଯାଇଥିବା ପଦ୍ଧତି ଦେଖନ୍ତୁ। +OmniRouteକୁ କ’ଣ ସ୍ୱତନ୍ତ୍ର କରେ — 13ଟି କ୍ଷମତାରେ 9router, OpenRouter, CLIProxyAPI ଏବଂ LiteLLM ସହ ତାରିଖ-ସହିତ ବୈଶିଷ୍ଟ୍ୟ ତୁଳନା। OmniRoute: 384ଟି ପ୍ରଦାତା, ଅନ୍ତର୍ନିର୍ମିତ 150+ ମାଗଣା ଟିଅର୍, 20ଟି ରାଉଟିଂ କୌଶଳ, 12-ଇଞ୍ଜିନ୍ ଟୋକେନ୍ ସଙ୍କୋଚନ, 110ଟି ଟୁଲ୍ ସହ ଅନ୍ତର୍ନିର୍ମିତ MCP ସର୍ଭର, A2A ଏଜେଣ୍ଟ ପ୍ରୋଟୋକଲ୍, ସ୍ଥାୟୀ ମେମୋରି, ଗାର୍ଡରେଲ୍, କ୍ଲାଉଡ୍ ଏଜେଣ୍ଟ, TLS ଫିଙ୍ଗରପ୍ରିଣ୍ଟ ଷ୍ଟିଲ୍ଥ୍, Desktop/Termux/PWA ଏବଂ 67ଟି i18n UI ଲୋକେଲ୍। OmniRoute MIT-ଲାଇସେନ୍ସପ୍ରାପ୍ତ ଏବଂ ନିଜେ ହୋଷ୍ଟ କରାଯାଇପାରେ। ପ୍ରତିଦ୍ୱନ୍ଦ୍ୱୀମାନଙ୍କର କ୍ଷମତା ଏବଂ ସଂଖ୍ୟା ବଦଳିପାରେ; ଲିଙ୍କ୍ କରାଯାଇଥିବା ପଦ୍ଧତି ଦେଖନ୍ତୁ। 📊 9router, OpenRouter, CLIProxyAPI ଏବଂ LiteLLM ସହ ସମ୍ପୂର୍ଣ୍ଣ ପଦ୍ଧତି ଏବଂ ପ୍ରତ୍ୟେକ ବୈଶିଷ୍ଟ୍ୟର ବିବରଣୀ → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -689,11 +689,11 @@ omniroute configure codex # ଏହିଠାରେ ମଧ୍ୟ: claude
-## 🌐 383 AI ପ୍ରଦାତା — 153 କ୍ୟାଟାଲଗ୍-ଚିହ୍ନିତ ମାଗଣା +## 🌐 384 AI ପ୍ରଦାତା — 153 କ୍ୟାଟାଲଗ୍-ଚିହ୍ନିତ ମାଗଣା
-> **383 ପଞ୍ଜୀକୃତ ପ୍ରଦାତା** canonical chat, media, search, local, cloud-agent ଏବଂ system ସଂଗ୍ରହଗୁଡ଼ିକରେ ରହିଛନ୍ତି, ଯାହାମଧ୍ୟରୁ **153 ଜଣଙ୍କ ପାଖରେ `hasFree: true` ଆବିଷ୍କାର ମେଟାଡାଟା** ରହିଛି। Chat model registry ରେ **239 ପ୍ରଦାତା / 3,059 ସ୍ୱତନ୍ତ୍ର provider-model ଯୁଗଳ / 1,475 କାଚା model ID** ରହିଛି; ଅଲଗା free-budget catalog ରେ **471 ପ୍ରତି-model ଧାଡ଼ି**, **34 ପୁନରାବୃତ୍ତିଶୀଳ pool** ଏବଂ **52 ପୁନରାବୃତ୍ତିଶୀଳ/keyless ସଦା-ମାଗଣା ପ୍ରଦାତା** ଅଛନ୍ତି। ଏଗୁଡ଼ିକ ଡିଜାଇନ୍ ଅନୁସାରେ ଭିନ୍ନ denominator; ସଂଜ୍ଞା ଏବଂ pool-deduped ଗଣନା [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) ଏବଂ [Free Tiers](docs/reference/FREE_TIERS.md) ରେ ରହିଛି। +> **384 ପଞ୍ଜୀକୃତ ପ୍ରଦାତା** canonical chat, media, search, local, cloud-agent ଏବଂ system ସଂଗ୍ରହଗୁଡ଼ିକରେ ରହିଛନ୍ତି, ଯାହାମଧ୍ୟରୁ **153 ଜଣଙ୍କ ପାଖରେ `hasFree: true` ଆବିଷ୍କାର ମେଟାଡାଟା** ରହିଛି। Chat model registry ରେ **239 ପ୍ରଦାତା / 3,059 ସ୍ୱତନ୍ତ୍ର provider-model ଯୁଗଳ / 1,475 କାଚା model ID** ରହିଛି; ଅଲଗା free-budget catalog ରେ **471 ପ୍ରତି-model ଧାଡ଼ି**, **34 ପୁନରାବୃତ୍ତିଶୀଳ pool** ଏବଂ **52 ପୁନରାବୃତ୍ତିଶୀଳ/keyless ସଦା-ମାଗଣା ପ୍ରଦାତା** ଅଛନ୍ତି। ଏଗୁଡ଼ିକ ଡିଜାଇନ୍ ଅନୁସାରେ ଭିନ୍ନ denominator; ସଂଜ୍ଞା ଏବଂ pool-deduped ଗଣନା [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) ଏବଂ [Free Tiers](docs/reference/FREE_TIERS.md) ରେ ରହିଛି।
@@ -1295,7 +1295,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 ମଧ୍ୟ ଏ ରନ୍ଟାଇମ୍Node.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ଭାଷାTypeScript 6.0 — src/ ଏବଂ open-sse/ ରେ 100% TypeScript (v2.0 ଠାରୁ କୋର୍ରେ ଶୂନ୍ୟ any) ଫ୍ରେମୱର୍କNext.js 16 + React 19 + Tailwind CSS 4 - ଡାଟାବେସ୍better-sqlite3 (SQLite, WAL ଜର୍ନାଲିଂ) + LowDB (JSON ପୁରୁଣା) — 147 ଡୋମେନ୍ ମଡ୍ୟୁଲ୍, 230 ମାଇଗ୍ରେସନ୍ + ଡାଟାବେସ୍better-sqlite3 (SQLite, WAL ଜର୍ନାଲିଂ) + LowDB (JSON ପୁରୁଣା) — 147 ଡୋମେନ୍ ମଡ୍ୟୁଲ୍, 232 ମାଇଗ୍ରେସନ୍ ମେମୋରିSQLite FTS5 ପୂର୍ଣ୍ଣ-ଟେକ୍ସଟ୍ + int8-କ୍ୱାଣ୍ଟାଇଜ୍ଡ୍ ଭେକ୍ଟର୍ ଏମ୍ବେଡିଂ, ଟାଇପ୍ଡ୍ ଡିକେ ସ୍କିମାZod 4 — MCP ଟୁଲ୍ I/O ଭ୍ୟାଲିଡେସନ୍ + API କନ୍ଟ୍ରାକ୍ଟ ପ୍ରୋଟୋକଲ୍MCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/or/llm.txt b/docs/i18n/or/llm.txt index 50cfc740d4e..9c5a99e4bfb 100644 --- a/docs/i18n/or/llm.txt +++ b/docs/i18n/or/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/pa/README.md b/docs/i18n/pa/README.md index 2f4d3dd4b0d..ce9dea246d2 100644 --- a/docs/i18n/pa/README.md +++ b/docs/i18n/pa/README.md @@ -15,7 +15,7 @@ **[Diego Souza](https://github.com/diegosouzapw) ਦੁਆਰਾ ਬਣਾਇਆ ਗਿਆ** -OmniRoute — ਕਦੇ ਵੀ ਕੋਡਿੰਗ ਨਾ ਰੋਕੋ। ਹਰ AI ਟੂਲ → 383 ਪ੍ਰਦਾਤਾ — 150+ ਮੁਫ਼ਤ — ਇੱਕ ਐਂਡਪੌਇੰਟ ਰਾਹੀਂ। Claude Code, Codex, Cursor, Cline, Copilot ਅਤੇ Antigravity ਨੂੰ ਮੁਫ਼ਤ Claude / GPT / Gemini ਨਾਲ ਆਟੋ-ਫਾਲਬੈਕ ਸਮੇਤ ਜੋੜੋ। RTK + Caveman ਦੀ ਸਟੈਕ ਕੀਤੀ ਕੰਪ੍ਰੈਸ਼ਨ 15–95% ਟੋਕਨ (~89% ਔਸਤ) ਬਚਾਉਂਦੀ ਹੈ — ਕਦੇ ਵੀ ਸੀਮਾਵਾਂ ਨਾਲ ਨਾ ਟਕਰਾਓ। 383 AI ਪ੍ਰਦਾਤਾ · 150+ ਮੁਫ਼ਤ ਟੀਅਰ · ~1.62B ਮੁਫ਼ਤ ਟੋਕਨ/ਮਹੀਨਾ · 20 ਰੂਟਿੰਗ ਰਣਨੀਤੀਆਂ · ਸ਼ੁਰੂਆਤ ਲਈ $0। +OmniRoute — ਕਦੇ ਵੀ ਕੋਡਿੰਗ ਨਾ ਰੋਕੋ। ਹਰ AI ਟੂਲ → 384 ਪ੍ਰਦਾਤਾ — 150+ ਮੁਫ਼ਤ — ਇੱਕ ਐਂਡਪੌਇੰਟ ਰਾਹੀਂ। Claude Code, Codex, Cursor, Cline, Copilot ਅਤੇ Antigravity ਨੂੰ ਮੁਫ਼ਤ Claude / GPT / Gemini ਨਾਲ ਆਟੋ-ਫਾਲਬੈਕ ਸਮੇਤ ਜੋੜੋ। RTK + Caveman ਦੀ ਸਟੈਕ ਕੀਤੀ ਕੰਪ੍ਰੈਸ਼ਨ 15–95% ਟੋਕਨ (~89% ਔਸਤ) ਬਚਾਉਂਦੀ ਹੈ — ਕਦੇ ਵੀ ਸੀਮਾਵਾਂ ਨਾਲ ਨਾ ਟਕਰਾਓ। 384 AI ਪ੍ਰਦਾਤਾ · 150+ ਮੁਫ਼ਤ ਟੀਅਰ · ~1.62B ਮੁਫ਼ਤ ਟੋਕਨ/ਮਹੀਨਾ · 20 ਰੂਟਿੰਗ ਰਣਨੀਤੀਆਂ · ਸ਼ੁਰੂਆਤ ਲਈ $0।
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------ | :---------------------: | :---------: | :-------------------: | -| 🌐 ਪ੍ਰਦਾਤਾ | 352 | **358** | 383 | +| 🌐 ਪ੍ਰਦਾਤਾ | 352 | **358** | 384 | | 🧠 ਵਿਲੱਖਣ ਚੈਟ ਮਾਡਲ ID | 1320 | **1374** | 1491 | | 🖼️ ਮੋਡੈਲਿਟੀ ਬ੍ਰਿਜ | 🆕 ਵਿਜ਼ਨ + ਆਡੀਓ + ਵੀਡੀਓ | ✓ | ✓ | | 📡 ਰਡਾਰ ਮੁਫ਼ਤ ਕੈਟਾਲਾਗ | 🆕 ਆਪਟ-ਇਨ | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ ਵਿਸ਼ੇਸ਼ਤਾਵਾਂ 🎯 ਕੰਬੋਜ਼ - 🌐 ਪ੍ਰਦਾਤਾ + 🌐 ਪ੍ਰਦਾਤਾ 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -ਵਾਅਦਾ — ਇੱਕ ਐਂਡਪੌਇੰਟ ਅਤੇ 383 ਪ੍ਰੋਵਾਈਡਰ। ਜਦੋਂ ਹੋਰ ਇੱਕ ਸਿਹਤਮੰਦ ਟਾਰਗੇਟ ਉਪਲਬਧ ਹੋਵੇ, ਤਾਂ ਆਟੋਮੈਟਿਕ fallback ਰੂਟਿੰਗ ਜਾਰੀ ਰੱਖਦਾ ਹੈ। ਛੇ ਥੰਮ੍ਹ: 383 ਪ੍ਰੋਵਾਈਡਰਾਂ ਵਿੱਚ resilient fallback · ਯੋਗ workloads 'ਤੇ 95% ਤੱਕ token ਬਚਤ · 150+ ਮੁਫ਼ਤ tiers ਅਤੇ 52 recurring/keyless free-forever ਪ੍ਰੋਵਾਈਡਰਾਂ ਨਾਲ ਸ਼ੁਰੂਆਤ ਲਈ $0 · ਇੱਕ config ਰਾਹੀਂ 39 CLI/agent integrations · /v1 'ਤੇ OpenAI, Claude, Gemini ਅਤੇ Responses API compatibility · circuit breakers, TLS stealth, MCP 110 tools, A2A, memory, guardrails, evals ਅਤੇ 7,800+ tracked test files ਵਿੱਚ 56,000+ static test declarations ਸਮੇਤ production controls। +ਵਾਅਦਾ — ਇੱਕ ਐਂਡਪੌਇੰਟ ਅਤੇ 384 ਪ੍ਰੋਵਾਈਡਰ। ਜਦੋਂ ਹੋਰ ਇੱਕ ਸਿਹਤਮੰਦ ਟਾਰਗੇਟ ਉਪਲਬਧ ਹੋਵੇ, ਤਾਂ ਆਟੋਮੈਟਿਕ fallback ਰੂਟਿੰਗ ਜਾਰੀ ਰੱਖਦਾ ਹੈ। ਛੇ ਥੰਮ੍ਹ: 384 ਪ੍ਰੋਵਾਈਡਰਾਂ ਵਿੱਚ resilient fallback · ਯੋਗ workloads 'ਤੇ 95% ਤੱਕ token ਬਚਤ · 150+ ਮੁਫ਼ਤ tiers ਅਤੇ 52 recurring/keyless free-forever ਪ੍ਰੋਵਾਈਡਰਾਂ ਨਾਲ ਸ਼ੁਰੂਆਤ ਲਈ $0 · ਇੱਕ config ਰਾਹੀਂ 39 CLI/agent integrations · /v1 'ਤੇ OpenAI, Claude, Gemini ਅਤੇ Responses API compatibility · circuit breakers, TLS stealth, MCP 110 tools, A2A, memory, guardrails, evals ਅਤੇ 7,800+ tracked test files ਵਿੱਚ 56,000+ static test declarations ਸਮੇਤ production controls।

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRoute ਨੂੰ ਕੀ ਵੱਖਰਾ ਬਣਾਉਂਦਾ ਹੈ — 13 ਸਮਰੱਥਾਵਾਂ ਵਿੱਚ 9router, OpenRouter, CLIProxyAPI ਅਤੇ LiteLLM ਦੇ ਮੁਕਾਬਲੇ ਦੀ ਮਿਤੀ-ਸਮੇਤ ਵਿਸ਼ੇਸ਼ਤਾ ਝਲਕ। OmniRoute: 383 ਪ੍ਰਦਾਤਾ, ਪਹਿਲਾਂ ਤੋਂ ਸ਼ਾਮਲ 150+ ਮੁਫ਼ਤ ਟੀਅਰ, 20 ਰੂਟਿੰਗ ਰਣਨੀਤੀਆਂ, 12-ਇੰਜਣ ਟੋਕਨ ਕੰਪ੍ਰੈਸ਼ਨ, 110 ਟੂਲਾਂ ਵਾਲਾ ਬਿਲਟ-ਇਨ MCP ਸਰਵਰ, A2A ਏਜੰਟ ਪ੍ਰੋਟੋਕੋਲ, ਸਥਾਈ ਮੈਮੋਰੀ, ਗਾਰਡਰੇਲ, ਕਲਾਊਡ ਏਜੰਟ, TLS ਫਿੰਗਰਪ੍ਰਿੰਟ ਸਟੀਲਥ, Desktop/Termux/PWA ਅਤੇ 67 i18n UI ਲੋਕੇਲ। OmniRoute MIT-ਲਾਇਸੰਸਸ਼ੁਦਾ ਅਤੇ ਖੁਦ ਹੋਸਟ ਕਰਨਯੋਗ ਹੈ। ਮੁਕਾਬਲੇਦਾਰਾਂ ਦੀਆਂ ਸਮਰੱਥਾਵਾਂ ਅਤੇ ਗਿਣਤੀਆਂ ਬਦਲ ਸਕਦੀਆਂ ਹਨ; ਲਿੰਕ ਕੀਤੀ ਵਿਧੀ ਵੇਖੋ। +OmniRoute ਨੂੰ ਕੀ ਵੱਖਰਾ ਬਣਾਉਂਦਾ ਹੈ — 13 ਸਮਰੱਥਾਵਾਂ ਵਿੱਚ 9router, OpenRouter, CLIProxyAPI ਅਤੇ LiteLLM ਦੇ ਮੁਕਾਬਲੇ ਦੀ ਮਿਤੀ-ਸਮੇਤ ਵਿਸ਼ੇਸ਼ਤਾ ਝਲਕ। OmniRoute: 384 ਪ੍ਰਦਾਤਾ, ਪਹਿਲਾਂ ਤੋਂ ਸ਼ਾਮਲ 150+ ਮੁਫ਼ਤ ਟੀਅਰ, 20 ਰੂਟਿੰਗ ਰਣਨੀਤੀਆਂ, 12-ਇੰਜਣ ਟੋਕਨ ਕੰਪ੍ਰੈਸ਼ਨ, 110 ਟੂਲਾਂ ਵਾਲਾ ਬਿਲਟ-ਇਨ MCP ਸਰਵਰ, A2A ਏਜੰਟ ਪ੍ਰੋਟੋਕੋਲ, ਸਥਾਈ ਮੈਮੋਰੀ, ਗਾਰਡਰੇਲ, ਕਲਾਊਡ ਏਜੰਟ, TLS ਫਿੰਗਰਪ੍ਰਿੰਟ ਸਟੀਲਥ, Desktop/Termux/PWA ਅਤੇ 67 i18n UI ਲੋਕੇਲ। OmniRoute MIT-ਲਾਇਸੰਸਸ਼ੁਦਾ ਅਤੇ ਖੁਦ ਹੋਸਟ ਕਰਨਯੋਗ ਹੈ। ਮੁਕਾਬਲੇਦਾਰਾਂ ਦੀਆਂ ਸਮਰੱਥਾਵਾਂ ਅਤੇ ਗਿਣਤੀਆਂ ਬਦਲ ਸਕਦੀਆਂ ਹਨ; ਲਿੰਕ ਕੀਤੀ ਵਿਧੀ ਵੇਖੋ। 📊 9router, OpenRouter, CLIProxyAPI ਅਤੇ LiteLLM ਦੇ ਮੁਕਾਬਲੇ ਦੀ ਪੂਰੀ ਵਿਧੀ ਅਤੇ ਪ੍ਰਤੀ-ਵਿਸ਼ੇਸ਼ਤਾ ਵੇਰਵਾ → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar ਦੀ ਵਰਤੋਂ ਇੱਛਿਕ ਹੈ ਅਤੇ ਇਹ ਸਿਰ - **🖼️ ਨਵੇਂ ਐਂਡਪੌਇੰਟ** — `/v1/ocr` (Mistral OCR) ਅਤੇ `/v1/audio/translations` (Whisper-ਸ਼ੈਲੀ) ਮੀਡੀਆ ਸਤਹ ਨੂੰ ਪੂਰਾ ਕਰਦੇ ਹਨ। → [API ਹਵਾਲਾ](docs/reference/API_REFERENCE.md) - **🎨 ਚਿੱਤਰ / ਵੀਡੀਓ / ਆਡੀਓ ਜਨਰੇਸ਼ਨ** — ਮੀਡੀਆ ਲਈ ਇੱਕ API: xAI Grok Imagine ਅਤੇ Novita AI ਵੀਡੀਓ, ComfyUI, Magnific, Adobe Firefly, Segmind ਅਤੇ ElevenLabs ਵਰਗੇ ਸਪੀਚ ਪ੍ਰੋਵਾਈਡਰ। → [API ਹਵਾਲਾ](docs/reference/API_REFERENCE.md) - **🌍 ਡਿਪਲੌਇਮੈਂਟ ਅਤੇ ਓਪਰੇਸ਼ਨ** — ਰਿਵਰਸ-ਪ੍ਰਾਕਸੀ `basePath`, ਬ੍ਰਾਊਜ਼ਰ-ਭਾਸ਼ਾ ਆਟੋ-ਡਿਟੈਕਟ, ਪ੍ਰਤੀ-ਕੁੰਜੀ ਡਿਵਾਈਸ ਟ੍ਰੈਕਿੰਗ, root-less MITM trust ਅਤੇ zh-TW ਲੋਕਲਾਈਜ਼ੇਸ਼ਨ। → [ਵਾਤਾਵਰਣ](docs/reference/ENVIRONMENT.md) -- **🤝 ਹੋਰ ਪ੍ਰੋਵਾਈਡਰ ਅਤੇ ਏਜੰਟ** — ਕਲਾਊਡ ਏਜੰਟ (Codex Cloud, Cursor, Devin, Jules), ਬ੍ਰਾਊਜ਼ਰ + OAuth ਲੌਗਇਨ ਵਾਲਾ Grok Build (xAI), ਪਹਿਲੀ ਸ਼੍ਰੇਣੀ ਵਾਲਾ Ollama ਕਾਰਡ, Claude Opus 5 ਅਤੇ Sonnet 5, Kimi ਦੀ ਅਧਿਕਾਰਤ ਭਾਈਵਾਲੀ (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… ਅਤੇ ਤਾਜ਼ਾ ਕੀਤਾ ਹੋਇਆ **383-ਪ੍ਰੋਵਾਈਡਰ ਕੈਟਾਲਾਗ**। → [ਪ੍ਰੋਵਾਈਡਰ](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 ਹੋਰ ਪ੍ਰੋਵਾਈਡਰ ਅਤੇ ਏਜੰਟ** — ਕਲਾਊਡ ਏਜੰਟ (Codex Cloud, Cursor, Devin, Jules), ਬ੍ਰਾਊਜ਼ਰ + OAuth ਲੌਗਇਨ ਵਾਲਾ Grok Build (xAI), ਪਹਿਲੀ ਸ਼੍ਰੇਣੀ ਵਾਲਾ Ollama ਕਾਰਡ, Claude Opus 5 ਅਤੇ Sonnet 5, Kimi ਦੀ ਅਧਿਕਾਰਤ ਭਾਈਵਾਲੀ (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… ਅਤੇ ਤਾਜ਼ਾ ਕੀਤਾ ਹੋਇਆ **384-ਪ੍ਰੋਵਾਈਡਰ ਕੈਟਾਲਾਗ**। → [ਪ੍ਰੋਵਾਈਡਰ](docs/reference/PROVIDER_REFERENCE.md) - **📡 ਰੂਟਿੰਗ ਪਾਰਦਰਸ਼ਤਾ** — ਹਰ ਜਵਾਬ ਵਿੱਚ ਇੱਕ `X-OmniRoute-Decision` ਹੈਡਰ ਹੁੰਦਾ ਹੈ, ਜੋ ਸੇਵਾ ਦੇਣ ਵਾਲੀ ਰਣਨੀਤੀ/ਪ੍ਰੋਵਾਈਡਰ/ਲੇਟੈਂਸੀ ਦਾ ਨਾਮ ਦਿੰਦਾ ਹੈ; ਨਵੀਂ `cache-optimized` ਕੰਬੋ ਰਣਨੀਤੀ + Auto-Combo `cacheAffinity` ਫੈਕਟਰ ਦੁਹਰਾਈਆਂ ਬੇਨਤੀਆਂ ਨੂੰ ਉਸ ਕਨੈਕਸ਼ਨ ਵੱਲ ਵਾਪਸ ਭੇਜਦੇ ਹਨ ਜਿਸ ਕੋਲ ਕੈਸ਼ ਕੀਤਾ ਪ੍ਰੀਫਿਕਸ ਹੈ; ਅਤੇ ਸਿਰਫ਼-ਪੜ੍ਹਨਯੋਗ `/v1/auto-combo/{channel}/candidates` ਐਂਡਪੌਇੰਟ ਕਿਸੇ `auto/*` ਚੈਨਲ ਦਾ ਲਾਈਵ ਉਮੀਦਵਾਰ ਪੂਲ ਦਿਖਾਉਂਦਾ ਹੈ। → [ਆਟੋ-ਕੰਬੋ](docs/routing/AUTO-COMBO.md) - **⚡ ਸਥਾਨਕ ਕਾਰਗੁਜ਼ਾਰੀ ਅਤੇ ਇਨਫਰਾ** — ਇੱਕ-ਕਲਿੱਕ ਸਥਾਨਕ Redis, Cloudflare Workers / Deno Deploy ਰੀਲੇ ਡਿਪਲੋਅਰ, ਅਤੇ ਨਿਗਰਾਨੀ ਹੇਠ ਏਮਬੈਡ ਕੀਤੀਆਂ ਸੇਵਾਵਾਂ ਵਜੋਂ Bifrost ਅਤੇ Mux। → [ਏਮਬੈਡ ਕੀਤੀਆਂ ਸੇਵਾਵਾਂ](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 ਬਾਕਸ ਵਿੱਚ ਹੋਰ ਵੀ** — ਪਲੱਗਇਨ ਫ੍ਰੇਮਵਰਕ + ਮਾਰਕੀਟਪਲੇਸ, Omni/Agent/GitHub ਸਕਿਲਜ਼ ਫ੍ਰੇਮਵਰਕ, Obsidian ਵੌਲਟ ਇੰਟੀਗ੍ਰੇਸ਼ਨ (22 MCP ਟੂਲ), OpenAI-ਅਨੁਕੂਲ Batch ਅਤੇ Files APIs, ਸੈਮੈਂਟਿਕ ਰਿਸਪਾਂਸ ਕੈਸ਼, ਲੀਡਰਬੋਰਡਾਂ ਸਮੇਤ ਗੇਮੀਫਿਕੇਸ਼ਨ, ACP ਏਜੰਟ ਖੋਜ (15 ਬਿਲਟ-ਇਨ ਏਜੰਟ), BigQuery ਵਿੱਚ ਨਿਯਤ ਲੌਗ ਐਕਸਪੋਰਟ, `auto/chaos` ਪੈਰਲਲ ਮਲਟੀ-ਮਾਡਲ ਫੈਨ-ਆਊਟ, Telegram ਬੋਟ ਬ੍ਰਿਜ, ਇਨ-ਐਪ ਵਰਜ਼ਨ ਮੈਨੇਜਰ ਅਤੇ LMArena-ELO ਮੁਫ਼ਤ-ਪ੍ਰਦਾਤਾ ਰੈਂਕਿੰਗਾਂ। → [ਦਸਤਾਵੇਜ਼](docs/README.md) @@ -688,11 +688,11 @@ omniroute configure codex # ਇਹ ਵੀ: claude opencode qwen aider g
-## 🌐 383 AI ਪ੍ਰਦਾਤਾ — 153 ਕੈਟਾਲਾਗ-ਚਿੰਨ੍ਹਿਤ ਮੁਫ਼ਤ +## 🌐 384 AI ਪ੍ਰਦਾਤਾ — 153 ਕੈਟਾਲਾਗ-ਚਿੰਨ੍ਹਿਤ ਮੁਫ਼ਤ
-> **383 ਰਜਿਸਟਰ ਕੀਤੇ ਪ੍ਰਦਾਤਾ** ਕੈਨੋਨਿਕਲ ਚੈਟ, ਮੀਡੀਆ, ਖੋਜ, ਲੋਕਲ, ਕਲਾਉਡ-ਏਜੰਟ ਅਤੇ ਸਿਸਟਮ ਕਲੈਕਸ਼ਨਾਂ ਵਿੱਚ ਫੈਲੇ ਹੋਏ ਹਨ, ਜਿਨ੍ਹਾਂ ਵਿੱਚੋਂ **153 ਕੋਲ `hasFree: true` ਡਿਸਕਵਰੀ ਮੈਟਾਡਾਟਾ ਹੈ**। ਚੈਟ ਮਾਡਲ ਰਜਿਸਟਰੀ ਵਿੱਚ **239 ਪ੍ਰਦਾਤਾ / 3,059 ਵਿਲੱਖਣ provider-model ਜੋੜੇ / 1,475 ਕੱਚੇ ਮਾਡਲ ID** ਸ਼ਾਮਲ ਹਨ; ਵੱਖਰੇ ਮੁਫ਼ਤ-ਬਜਟ ਕੈਟਾਲਾਗ ਵਿੱਚ **471 ਪ੍ਰਤੀ-ਮਾਡਲ ਕਤਾਰਾਂ**, **34 ਆਵਰਤੀ ਪੂਲ** ਅਤੇ **52 ਆਵਰਤੀ/ਕੀ-ਰਹਿਤ ਹਮੇਸ਼ਾ-ਮੁਫ਼ਤ ਪ੍ਰਦਾਤਾ** ਹਨ। ਇਹ ਜਾਣ-ਬੁੱਝ ਕੇ ਵੱਖਰੇ ਹਰ ਹਨ; ਪਰਿਭਾਸ਼ਾਵਾਂ ਅਤੇ ਪੂਲ-ਡਿਡਿਊਪ ਕੀਤੀਆਂ ਗਣਨਾਵਾਂ [ਪ੍ਰਦਾਤਾ ਹਵਾਲਾ](docs/reference/PROVIDER_REFERENCE.md) ਅਤੇ [ਮੁਫ਼ਤ ਟੀਅਰ](docs/reference/FREE_TIERS.md) ਵਿੱਚ ਉਪਲਬਧ ਹਨ। +> **384 ਰਜਿਸਟਰ ਕੀਤੇ ਪ੍ਰਦਾਤਾ** ਕੈਨੋਨਿਕਲ ਚੈਟ, ਮੀਡੀਆ, ਖੋਜ, ਲੋਕਲ, ਕਲਾਉਡ-ਏਜੰਟ ਅਤੇ ਸਿਸਟਮ ਕਲੈਕਸ਼ਨਾਂ ਵਿੱਚ ਫੈਲੇ ਹੋਏ ਹਨ, ਜਿਨ੍ਹਾਂ ਵਿੱਚੋਂ **153 ਕੋਲ `hasFree: true` ਡਿਸਕਵਰੀ ਮੈਟਾਡਾਟਾ ਹੈ**। ਚੈਟ ਮਾਡਲ ਰਜਿਸਟਰੀ ਵਿੱਚ **239 ਪ੍ਰਦਾਤਾ / 3,059 ਵਿਲੱਖਣ provider-model ਜੋੜੇ / 1,475 ਕੱਚੇ ਮਾਡਲ ID** ਸ਼ਾਮਲ ਹਨ; ਵੱਖਰੇ ਮੁਫ਼ਤ-ਬਜਟ ਕੈਟਾਲਾਗ ਵਿੱਚ **471 ਪ੍ਰਤੀ-ਮਾਡਲ ਕਤਾਰਾਂ**, **34 ਆਵਰਤੀ ਪੂਲ** ਅਤੇ **52 ਆਵਰਤੀ/ਕੀ-ਰਹਿਤ ਹਮੇਸ਼ਾ-ਮੁਫ਼ਤ ਪ੍ਰਦਾਤਾ** ਹਨ। ਇਹ ਜਾਣ-ਬੁੱਝ ਕੇ ਵੱਖਰੇ ਹਰ ਹਨ; ਪਰਿਭਾਸ਼ਾਵਾਂ ਅਤੇ ਪੂਲ-ਡਿਡਿਊਪ ਕੀਤੀਆਂ ਗਣਨਾਵਾਂ [ਪ੍ਰਦਾਤਾ ਹਵਾਲਾ](docs/reference/PROVIDER_REFERENCE.md) ਅਤੇ [ਮੁਫ਼ਤ ਟੀਅਰ](docs/reference/FREE_TIERS.md) ਵਿੱਚ ਉਪਲਬਧ ਹਨ।
@@ -1294,7 +1294,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 ਵੀ ਇਸਨ ਰਨਟਾਈਮNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ਭਾਸ਼ਾTypeScript 6.0 — src/ ਅਤੇ open-sse/ ਵਿੱਚ 100% TypeScript (v2.0 ਤੋਂ ਬਾਅਦ ਕੋਰ ਵਿੱਚ ਜ਼ੀਰੋ any) ਫਰੇਮਵਰਕNext.js 16 + React 19 + Tailwind CSS 4 - ਡੇਟਾਬੇਸbetter-sqlite3 (SQLite, WAL ਜਰਨਲਿੰਗ) + LowDB (JSON ਲੈਗੇਸੀ) — 147 ਡੋਮੇਨ ਮੋਡੀਊਲ, 230 ਮਾਈਗ੍ਰੇਸ਼ਨ + ਡੇਟਾਬੇਸbetter-sqlite3 (SQLite, WAL ਜਰਨਲਿੰਗ) + LowDB (JSON ਲੈਗੇਸੀ) — 147 ਡੋਮੇਨ ਮੋਡੀਊਲ, 232 ਮਾਈਗ੍ਰੇਸ਼ਨ ਮੈਮੋਰੀSQLite FTS5 ਫੁੱਲ-ਟੈਕਸਟ + int8-ਕਵਾਂਟਾਈਜ਼ਡ ਵੈਕਟਰ ਐਂਬੈਡਿੰਗ, ਟਾਈਪਡ ਡਿਕੇ ਸਕੀਮਾਵਾਂZod 4 — MCP ਟੂਲ I/O ਵੈਲੀਡੇਸ਼ਨ + API ਕਾਂਟ੍ਰੈਕਟ ਪ੍ਰੋਟੋਕੋਲMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/pa/llm.txt b/docs/i18n/pa/llm.txt index ff710144eaa..0a444116f96 100644 --- a/docs/i18n/pa/llm.txt +++ b/docs/i18n/pa/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/phi/README.md b/docs/i18n/phi/README.md index 61f755aca52..c484e5cbb86 100644 --- a/docs/i18n/phi/README.md +++ b/docs/i18n/phi/README.md @@ -15,7 +15,7 @@ **Ginawa ni [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Huwag tumigil sa pagko-code. Bawat AI tool → 383 provider — 150+ libre — sa pamamagitan ng isang endpoint. Claude Code, Codex, Cursor, Cline, Copilot at Antigravity patungo sa LIBRENG Claude / GPT / Gemini na may awtomatikong fallback. Ang stacked compression ng RTK + Caveman ay nakakatipid ng 15–95% token (~89% ang average) — hindi kailanman tatama sa mga limitasyon. 383 AI provider · 150+ libreng tier · ~1.62B libreng token/buwan · 20 estratehiya sa pagruruta · $0 para magsimula. +OmniRoute — Huwag tumigil sa pagko-code. Bawat AI tool → 384 provider — 150+ libre — sa pamamagitan ng isang endpoint. Claude Code, Codex, Cursor, Cline, Copilot at Antigravity patungo sa LIBRENG Claude / GPT / Gemini na may awtomatikong fallback. Ang stacked compression ng RTK + Caveman ay nakakatipid ng 15–95% token (~89% ang average) — hindi kailanman tatama sa mga limitasyon. 384 AI provider · 150+ libreng tier · ~1.62B libreng token/buwan · 20 estratehiya sa pagruruta · $0 para magsimula.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ----------------------------------- | :-----------------------: | :---------: | :--------------------: | -| 🌐 Mga provider | 352 | **358** | 383 | +| 🌐 Mga provider | 352 | **358** | 384 | | 🧠 Mga natatanging ID ng chat model | 1320 | **1374** | 1491 | | 🖼️ Tulay ng modality | 🆕 bisyon + audio + video | ✓ | ✓ | | 📡 Libreng katalogo ng Radar | 🆕 opt-in | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Mga Tampok 🎯 Mga Combo - 🌐 Mga Provider + 🌐 Mga Provider 🔌 CLI at MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Ang Pangako — Isang endpoint at 383 provider. Pinananatili ng awtomatikong fallback ang routing habang may available na isa pang maayos na target. Anim na haligi: matatag na fallback sa 383 provider · hanggang 95% pagtitipid sa token para sa mga kwalipikadong workload · $0 para magsimula, na may 150+ libreng tier at 52 paulit-ulit/keyless na provider na libre magpakailanman · 39 CLI/agent integration sa pamamagitan ng isang config · compatibility sa OpenAI, Claude, Gemini at Responses API sa /v1 · mga production control kabilang ang circuit breaker, TLS stealth, MCP na may 110 tool, A2A, memory, guardrail, eval at 56,000+ static test declaration sa 7,800+ sinusubaybayang test file. +Ang Pangako — Isang endpoint at 384 provider. Pinananatili ng awtomatikong fallback ang routing habang may available na isa pang maayos na target. Anim na haligi: matatag na fallback sa 384 provider · hanggang 95% pagtitipid sa token para sa mga kwalipikadong workload · $0 para magsimula, na may 150+ libreng tier at 52 paulit-ulit/keyless na provider na libre magpakailanman · 39 CLI/agent integration sa pamamagitan ng isang config · compatibility sa OpenAI, Claude, Gemini at Responses API sa /v1 · mga production control kabilang ang circuit breaker, TLS stealth, MCP na may 110 tool, A2A, memory, guardrail, eval at 56,000+ static test declaration sa 7,800+ sinusubaybayang test file.

@@ -510,7 +510,7 @@ Lahat ng **20** diskarte — paghalu-haluin at pagtugmain sa bawat hakbang ng co -Mga tampok na nagpapaiba sa OmniRoute — isang snapshot ng mga tampok ayon sa petsa kumpara sa 9router, OpenRouter, CLIProxyAPI at LiteLLM sa 13 kakayahan. OmniRoute: 383 provider, 150+ libreng tier na nakapaloob, 20 estratehiya sa pagruruta, compression ng token gamit ang 12 engine, nakapaloob na MCP server na may 110 tool, A2A agent protocol, persistent na memorya, mga guardrail, cloud agent, TLS fingerprint stealth, Desktop/Termux/PWA at 67 locale ng UI para sa i18n. Lisensyado ang OmniRoute sa ilalim ng MIT at maaaring i-self-host. Maaaring magbago ang mga kakayahan at bilang ng mga kakompetensya; tingnan ang naka-link na metodolohiya. +Mga tampok na nagpapaiba sa OmniRoute — isang snapshot ng mga tampok ayon sa petsa kumpara sa 9router, OpenRouter, CLIProxyAPI at LiteLLM sa 13 kakayahan. OmniRoute: 384 provider, 150+ libreng tier na nakapaloob, 20 estratehiya sa pagruruta, compression ng token gamit ang 12 engine, nakapaloob na MCP server na may 110 tool, A2A agent protocol, persistent na memorya, mga guardrail, cloud agent, TLS fingerprint stealth, Desktop/Termux/PWA at 67 locale ng UI para sa i18n. Lisensyado ang OmniRoute sa ilalim ng MIT at maaaring i-self-host. Maaaring magbago ang mga kakayahan at bilang ng mga kakompetensya; tingnan ang naka-link na metodolohiya. 📊 Buong metodolohiya at detalye sa bawat tampok kumpara sa 9router, OpenRouter, CLIProxyAPI & LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Ang Radar ay opt-in at GET-only. Hindi nag-a-upload ang OmniRoute client ng mga - **🖼️ Mga bagong endpoint** — kinukumpleto ng `/v1/ocr` (Mistral OCR) at `/v1/audio/translations` (istilong Whisper) ang media surface. → [Sanggunian ng API](docs/reference/API_REFERENCE.md) - **🎨 Pagbuo ng image / video / audio** — isang API para sa media: xAI Grok Imagine at Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind, at mga speech provider gaya ng ElevenLabs. → [Sanggunian ng API](docs/reference/API_REFERENCE.md) - **🌍 Deployment at mga operasyon** — reverse-proxy `basePath`, awtomatikong pagtukoy ng wika ng browser, pagsubaybay sa device bawat key, root-less na MITM trust, at zh-TW localization. → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 Mas maraming provider at agent** — mga cloud agent (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) na may browser + OAuth login, first-class card para sa Ollama, Claude Opus 5 at Sonnet 5, opisyal na pakikipagtulungan sa Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… at isang bagong **383-provider catalog**. → [Mga Provider](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Mas maraming provider at agent** — mga cloud agent (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) na may browser + OAuth login, first-class card para sa Ollama, Claude Opus 5 at Sonnet 5, opisyal na pakikipagtulungan sa Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… at isang bagong **384-provider catalog**. → [Mga Provider](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparency sa routing** — may `X-OmniRoute-Decision` header ang bawat response na nagsasaad ng strategy/provider/latency na naghatid nito; isang bagong `cache-optimized` combo strategy + Auto-Combo `cacheAffinity` factor ang nagbabalik ng mga paulit-ulit na request sa connection na may hawak ng naka-cache na prefix; at inilalantad ng read-only na `/v1/auto-combo/{channel}/candidates` endpoint ang live candidate pool ng isang `auto/*` channel. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Lokal na performance at imprastraktura** — mga lokal na Redis na isang-click na sinisimulan, mga deployer ng relay para sa Cloudflare Workers / Deno Deploy, at Bifrost at Mux bilang mga naka-supervise na naka-embed na serbisyo. → [Mga Naka-embed na Serbisyo](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Kasama rin sa package** — framework ng plugin at marketplace, mga framework ng kasanayan para sa Omni/Agent/GitHub, integrasyon sa Obsidian vault (22 MCP tool), mga API na tugma sa OpenAI para sa Batch at Files, semantic response cache, gamification na may mga leaderboard, pagtuklas ng ACP agent (15 built-in na agent), nakaiskedyul na pag-export ng log sa BigQuery, parallel multi-model fan-out na `auto/chaos`, bridge ng Telegram bot, in-app na version manager, at mga ranking ng libreng provider ng LMArena-ELO. → [Dokumentasyon](docs/README.md) @@ -689,11 +689,11 @@ mula sa iyong shell history. → [CLI Integrations](docs/guides/CLI-INTEGRATIONS
-## 🌐 383 AI Provider — 153 Minarkahan ng Catalogo bilang Libre +## 🌐 384 AI Provider — 153 Minarkahan ng Catalogo bilang Libre
-> **383 nakarehistrong provider** sa mga canonical na koleksyon ng chat, media, search, local, cloud-agent at system, kabilang ang **153 na may `hasFree: true` na metadata para sa discovery**. Saklaw ng chat model registry ang **239 provider / 3,059 natatanging pares ng provider-model / 1,475 raw model ID**; ang hiwalay na catalog ng libreng budget ay may **471 row bawat modelo**, **34 umuulit na pool** at **52 provider na paulit-ulit/keyless at libreng-libre**. Magkakaiba ang mga denominator na ito ayon sa disenyo; makikita ang mga depinisyon at kalkulasyong inalisan ng duplicate sa pool sa [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) at [Free Tiers](docs/reference/FREE_TIERS.md). +> **384 nakarehistrong provider** sa mga canonical na koleksyon ng chat, media, search, local, cloud-agent at system, kabilang ang **153 na may `hasFree: true` na metadata para sa discovery**. Saklaw ng chat model registry ang **239 provider / 3,059 natatanging pares ng provider-model / 1,475 raw model ID**; ang hiwalay na catalog ng libreng budget ay may **471 row bawat modelo**, **34 umuulit na pool** at **52 provider na paulit-ulit/keyless at libreng-libre**. Magkakaiba ang mga denominator na ito ayon sa disenyo; makikita ang mga depinisyon at kalkulasyong inalisan ng duplicate sa pool sa [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) at [Free Tiers](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ Mga canonical na sukatan noong 2026-10-18: **1.976 natatanging video** · **20.2 RuntimeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 WikaTypeScript 6.0 — 100% TypeScript sa src/ at open-sse/ (walang any sa core mula noong v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Databasebetter-sqlite3 (SQLite, WAL journaling) + LowDB (legacy na JSON) — 147 domain module, 230 migration + Databasebetter-sqlite3 (SQLite, WAL journaling) + LowDB (legacy na JSON) — 147 domain module, 232 migration MemoryaFull-text ng SQLite FTS5 + mga vector embedding na naka-quantize sa int8, may type na pagkabulok Mga SchemaZod 4 — pagpapatunay ng I/O ng MCP tool + mga kontrata ng API Mga ProtocolMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/phi/llm.txt b/docs/i18n/phi/llm.txt index d7d2e7dc8f8..79ec7833e53 100644 --- a/docs/i18n/phi/llm.txt +++ b/docs/i18n/phi/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/pl/README.md b/docs/i18n/pl/README.md index 04a17015089..1f1507b7105 100644 --- a/docs/i18n/pl/README.md +++ b/docs/i18n/pl/README.md @@ -15,7 +15,7 @@ **Autor: [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Nigdy nie przestawaj programować. Każde narzędzie AI → 383 dostawców — 150+ bezpłatnych — za pośrednictwem jednego punktu końcowego. Claude Code, Codex, Cursor, Cline, Copilot i Antigravity z dostępem do BEZPŁATNYCH modeli Claude / GPT / Gemini z automatycznym przełączaniem awaryjnym. Kompresja RTK + Caveman pozwala oszczędzić 15–95% tokenów (średnio ~89%) — nigdy nie napotykaj limitów. 383 dostawców AI · 150+ bezpłatnych poziomów · ~1,62 mld bezpłatnych tokenów/mies. · 20 strategii routingu · $0 na start. +OmniRoute — Nigdy nie przestawaj programować. Każde narzędzie AI → 384 dostawców — 150+ bezpłatnych — za pośrednictwem jednego punktu końcowego. Claude Code, Codex, Cursor, Cline, Copilot i Antigravity z dostępem do BEZPŁATNYCH modeli Claude / GPT / Gemini z automatycznym przełączaniem awaryjnym. Kompresja RTK + Caveman pozwala oszczędzić 15–95% tokenów (średnio ~89%) — nigdy nie napotykaj limitów. 384 dostawców AI · 150+ bezpłatnych poziomów · ~1,62 mld bezpłatnych tokenów/mies. · 20 strategii routingu · $0 na start.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------------------- | :----------------------: | :---------: | :-------------------: | -| 🌐 Dostawcy | 352 | **358** | 383 | +| 🌐 Dostawcy | 352 | **358** | 384 | | 🧠 Unikalne identyfikatory modeli czatu | 1320 | **1374** | 1491 | | 🖼️ Mostek modalności | 🆕 obraz + audio + wideo | ✓ | ✓ | | 📡 Bezpłatny katalog Radar | 🆕 opt-in | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funkcje 🎯 Kombinacje - 🌐 Dostawcy + 🌐 Dostawcy 🔌 CLI i MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Obietnica — jeden endpoint i 383 dostawców. Automatyczny fallback utrzymuje routing, dopóki dostępny jest inny sprawny cel. Sześć filarów: odporny fallback obejmujący 383 dostawców · do 95% oszczędności tokenów w kwalifikujących się zadaniach · 0 USD na start dzięki ponad 150 bezpłatnym planom oraz 52 cyklicznym lub niewymagającym klucza dostawcom oferującym bezpłatny dostęp bezterminowo · 39 integracji CLI/agentów za pomocą jednej konfiguracji · zgodność z interfejsami API OpenAI, Claude, Gemini i Responses pod adresem /v1 · mechanizmy kontroli produkcyjnej, w tym wyłączniki obwodu, TLS stealth, 110 narzędzi MCP, A2A, pamięć, zabezpieczenia, ewaluacje oraz ponad 56 000 statycznych deklaracji testów w ponad 7800 śledzonych plikach testów. +Obietnica — jeden endpoint i 384 dostawców. Automatyczny fallback utrzymuje routing, dopóki dostępny jest inny sprawny cel. Sześć filarów: odporny fallback obejmujący 384 dostawców · do 95% oszczędności tokenów w kwalifikujących się zadaniach · 0 USD na start dzięki ponad 150 bezpłatnym planom oraz 52 cyklicznym lub niewymagającym klucza dostawcom oferującym bezpłatny dostęp bezterminowo · 39 integracji CLI/agentów za pomocą jednej konfiguracji · zgodność z interfejsami API OpenAI, Claude, Gemini i Responses pod adresem /v1 · mechanizmy kontroli produkcyjnej, w tym wyłączniki obwodu, TLS stealth, 110 narzędzi MCP, A2A, pamięć, zabezpieczenia, ewaluacje oraz ponad 56 000 statycznych deklaracji testów w ponad 7800 śledzonych plikach testów.

@@ -510,7 +510,7 @@ Wszystkie **20** strategii — łącz je dowolnie na poszczególnych etapach kom -Co wyróżnia OmniRoute — aktualne zestawienie funkcji w porównaniu z 9router, OpenRouter, CLIProxyAPI i LiteLLM w 13 obszarach. OmniRoute: 383 dostawców, ponad 150 wbudowanych bezpłatnych progów, 20 strategii routingu, kompresja tokenów z wykorzystaniem 12 silników, wbudowany serwer MCP ze 110 narzędziami, protokół agentów A2A, trwała pamięć, mechanizmy ochronne, agenci chmurowi, ukrywanie odcisku TLS, Desktop/Termux/PWA oraz 67 lokalizacji interfejsu i18n. OmniRoute jest objęty licencją MIT i można go samodzielnie hostować. Możliwości i liczby dotyczące konkurencyjnych rozwiązań mogą się zmieniać; zobacz powiązaną metodologię. +Co wyróżnia OmniRoute — aktualne zestawienie funkcji w porównaniu z 9router, OpenRouter, CLIProxyAPI i LiteLLM w 13 obszarach. OmniRoute: 384 dostawców, ponad 150 wbudowanych bezpłatnych progów, 20 strategii routingu, kompresja tokenów z wykorzystaniem 12 silników, wbudowany serwer MCP ze 110 narzędziami, protokół agentów A2A, trwała pamięć, mechanizmy ochronne, agenci chmurowi, ukrywanie odcisku TLS, Desktop/Termux/PWA oraz 67 lokalizacji interfejsu i18n. OmniRoute jest objęty licencją MIT i można go samodzielnie hostować. Możliwości i liczby dotyczące konkurencyjnych rozwiązań mogą się zmieniać; zobacz powiązaną metodologię. 📊 Pełna metodologia i szczegóły dotyczące poszczególnych funkcji w porównaniu z 9router, OpenRouter, CLIProxyAPI i LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -614,7 +614,7 @@ aktualnym katalogu znajdziesz na stronie **[radar.omniroute.online/planos](https - **🖼️ Nowe endpointy** — `/v1/ocr` (Mistral OCR) i `/v1/audio/translations` (w stylu Whisper) uzupełniają obsługę multimediów. → [Dokumentacja API](docs/reference/API_REFERENCE.md) - **🎨 Generowanie obrazów / wideo / dźwięku** — jedno API do multimediów: xAI Grok Imagine i wideo Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind oraz dostawcy mowy, tacy jak ElevenLabs. → [Dokumentacja API](docs/reference/API_REFERENCE.md) - **🌍 Wdrażanie i operacje** — `basePath` odwrotnego proxy, automatyczne wykrywanie języka przeglądarki, śledzenie urządzeń dla poszczególnych kluczy, zaufanie MITM bez uprawnień roota, lokalizacja zh-TW. → [Środowisko](docs/reference/ENVIRONMENT.md) -- **🤝 Więcej dostawców i agentów** — agenci chmurowi (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) z przeglądarką i logowaniem OAuth, karta Ollama jako pełnoprawny element, Claude Opus 5 i Sonnet 5, oficjalne partnerstwo z Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… oraz odświeżony **katalog 383 dostawców**. → [Dostawcy](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Więcej dostawców i agentów** — agenci chmurowi (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) z przeglądarką i logowaniem OAuth, karta Ollama jako pełnoprawny element, Claude Opus 5 i Sonnet 5, oficjalne partnerstwo z Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… oraz odświeżony **katalog 384 dostawców**. → [Dostawcy](docs/reference/PROVIDER_REFERENCE.md) - **📡 Przejrzystość routingu** — każda odpowiedź zawiera nagłówek `X-OmniRoute-Decision` z nazwą strategii/dostawcy i opóźnieniem, które obsłużyły żądanie; nowa strategia combo `cache-optimized` oraz współczynnik `cacheAffinity` Auto-Combo kierują powtarzane żądania z powrotem do połączenia przechowującego buforowany prefiks, a tylko do odczytu endpoint `/v1/auto-combo/{channel}/candidates` udostępnia bieżącą pulę kandydatów kanału `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Lokalna wydajność i infrastruktura** — lokalny Redis uruchamiany jednym kliknięciem, wdrażanie przekaźników Cloudflare Workers / Deno Deploy, Bifrost i Mux jako nadzorowane usługi osadzone. → [Usługi osadzone](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Także w zestawie** — framework wtyczek i marketplace, frameworki umiejętności Omni/Agent/GitHub, integracja z sejfem Obsidian (22 narzędzia MCP), zgodne z OpenAI interfejsy API Batch i Files, semantyczna pamięć podręczna odpowiedzi, grywalizacja z tabelami wyników, wykrywanie agentów ACP (15 wbudowanych agentów), zaplanowany eksport logów do BigQuery, równoległe rozgałęzianie na wiele modeli `auto/chaos`, mostek bota Telegram, menedżer wersji w aplikacji oraz rankingi bezpłatnych dostawców LMArena-ELO. → [Dokumentacja](docs/README.md) @@ -697,11 +697,11 @@ pokazuje dokładnie jakie zmienne środowiskowe/argumenty zostaną użyte bez wy
-## 🌐 383 dostawców AI — 153 oznaczonych w katalogu jako bezpłatni +## 🌐 384 dostawców AI — 153 oznaczonych w katalogu jako bezpłatni
-> **383 zarejestrowanych dostawców** w kanonicznych kolekcjach czatu, multimediów, wyszukiwania, lokalnych, agentów chmurowych i systemowych, w tym **153 z metadanymi wykrywania `hasFree: true`**. Rejestr modeli czatu obejmuje **239 dostawców / 3 059 unikalnych par dostawca–model / 1 475 surowych identyfikatorów modeli**; oddzielny katalog bezpłatnych limitów obejmuje **471 wierszy dotyczących poszczególnych modeli**, **34 cykliczne pule** oraz **52 cyklicznych lub bezkluczowych dostawców bezpłatnych na zawsze**. Są to celowo różne mianowniki; definicje i obliczenia z deduplikacją pul znajdują się w [Informatorze dostawców](docs/reference/PROVIDER_REFERENCE.md) i [Bezpłatnych poziomach](docs/reference/FREE_TIERS.md). +> **384 zarejestrowanych dostawców** w kanonicznych kolekcjach czatu, multimediów, wyszukiwania, lokalnych, agentów chmurowych i systemowych, w tym **153 z metadanymi wykrywania `hasFree: true`**. Rejestr modeli czatu obejmuje **239 dostawców / 3 059 unikalnych par dostawca–model / 1 475 surowych identyfikatorów modeli**; oddzielny katalog bezpłatnych limitów obejmuje **471 wierszy dotyczących poszczególnych modeli**, **34 cykliczne pule** oraz **52 cyklicznych lub bezkluczowych dostawców bezpłatnych na zawsze**. Są to celowo różne mianowniki; definicje i obliczenia z deduplikacją pul znajdują się w [Informatorze dostawców](docs/reference/PROVIDER_REFERENCE.md) i [Bezpłatnych poziomach](docs/reference/FREE_TIERS.md).
@@ -1302,7 +1302,7 @@ Dane kanoniczne na 2026-10-18: **1.976 unikalnych filmów** · **20.220.289 znan Środowisko uruchomienioweNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 JęzykTypeScript 6.0 — 100% TypeScript w całych katalogach src/ i open-sse/ (zero any w rdzeniu od wersji v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Baza danychbetter-sqlite3 (SQLite, rejestrowanie WAL) + LowDB (starszy format JSON) — 147 modułów domenowych, 230 migracji + Baza danychbetter-sqlite3 (SQLite, rejestrowanie WAL) + LowDB (starszy format JSON) — 147 modułów domenowych, 232 migracji PamięćPełnotekstowe wyszukiwanie SQLite FTS5 + osadzenia wektorowe z kwantyzacją int8, typowane wygaszanie SchematyZod 4 — walidacja wejścia/wyjścia narzędzi MCP + kontrakty API ProtokołyMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/pl/llm.txt b/docs/i18n/pl/llm.txt index c5db8aedc7a..fcc1251d31b 100644 --- a/docs/i18n/pl/llm.txt +++ b/docs/i18n/pl/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/pt-BR/README.md b/docs/i18n/pt-BR/README.md index 66920754a0d..1fb48396c38 100644 --- a/docs/i18n/pt-BR/README.md +++ b/docs/i18n/pt-BR/README.md @@ -15,7 +15,7 @@ **Criado por [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Nunca pare de programar. Cada ferramenta de IA → 383 provedores — 150+ gratuitos — por meio de um único endpoint. Claude Code, Codex, Cursor, Cline, Copilot e Antigravity em Claude / GPT / Gemini GRATUITOS com fallback automático. A compressão combinada RTK + Caveman economiza de 15 a 95% dos tokens (~89% em média) — nunca atinja os limites. 383 provedores de IA · 150+ níveis gratuitos · ~1,62 bilhão de tokens gratuitos/mês · 20 estratégias de roteamento · $0 para começar. +OmniRoute — Nunca pare de programar. Cada ferramenta de IA → 384 provedores — 150+ gratuitos — por meio de um único endpoint. Claude Code, Codex, Cursor, Cline, Copilot e Antigravity em Claude / GPT / Gemini GRATUITOS com fallback automático. A compressão combinada RTK + Caveman economiza de 15 a 95% dos tokens (~89% em média) — nunca atinja os limites. 384 provedores de IA · 150+ níveis gratuitos · ~1,62 bilhão de tokens gratuitos/mês · 20 estratégias de roteamento · $0 para começar.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ----------------------------------------- | :----------------------: | :---------: | :---------------------: | -| 🌐 Provedores | 352 | **358** | 383 | +| 🌐 Provedores | 352 | **358** | 384 | | 🧠 IDs exclusivos de modelos de chat | 1320 | **1374** | 1491 | | 🖼️ Ponte de modalidades | 🆕 visão + áudio + vídeo | ✓ | ✓ | | 📡 Catálogo gratuito do Radar | 🆕 adesão opcional | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Recursos 🎯 Combos - 🌐 Provedores + 🌐 Provedores 🔌 CLI e MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -A promessa — Um endpoint e 383 provedores. O fallback automático mantém o roteamento enquanto outro destino saudável estiver disponível. Seis pilares: fallback resiliente entre 383 provedores · economia de até 95% em tokens em cargas de trabalho elegíveis · $0 para começar, com mais de 150 níveis gratuitos e 52 provedores gratuitos recorrentes/sem chave para sempre · 39 integrações de CLI/agentes por meio de uma única configuração · compatibilidade com as APIs OpenAI, Claude, Gemini e Responses em /v1 · controles de produção, incluindo disjuntores, furtividade TLS, 110 ferramentas MCP, A2A, memória, proteções, avaliações e mais de 56.000 declarações de testes estáticos em mais de 7.800 arquivos de teste acompanhados. +A promessa — Um endpoint e 384 provedores. O fallback automático mantém o roteamento enquanto outro destino saudável estiver disponível. Seis pilares: fallback resiliente entre 384 provedores · economia de até 95% em tokens em cargas de trabalho elegíveis · $0 para começar, com mais de 150 níveis gratuitos e 52 provedores gratuitos recorrentes/sem chave para sempre · 39 integrações de CLI/agentes por meio de uma única configuração · compatibilidade com as APIs OpenAI, Claude, Gemini e Responses em /v1 · controles de produção, incluindo disjuntores, furtividade TLS, 110 ferramentas MCP, A2A, memória, proteções, avaliações e mais de 56.000 declarações de testes estáticos em mais de 7.800 arquivos de teste acompanhados.

@@ -510,7 +510,7 @@ Todas as **20** estratégias — combine-as livremente em cada etapa do combo: -O que diferencia o OmniRoute — um panorama datado de recursos em comparação com 9router, OpenRouter, CLIProxyAPI e LiteLLM em 13 capacidades. OmniRoute: 383 provedores, mais de 150 níveis gratuitos integrados, 20 estratégias de roteamento, compressão de tokens com 12 mecanismos, servidor MCP integrado com 110 ferramentas, protocolo de agentes A2A, memória persistente, proteções, agentes na nuvem, furtividade de impressão digital TLS, Desktop/Termux/PWA e 67 localidades de interface do usuário para i18n. O OmniRoute é licenciado sob MIT e pode ser hospedado por você. Os recursos e as contagens dos concorrentes podem mudar; consulte a metodologia vinculada. +O que diferencia o OmniRoute — um panorama datado de recursos em comparação com 9router, OpenRouter, CLIProxyAPI e LiteLLM em 13 capacidades. OmniRoute: 384 provedores, mais de 150 níveis gratuitos integrados, 20 estratégias de roteamento, compressão de tokens com 12 mecanismos, servidor MCP integrado com 110 ferramentas, protocolo de agentes A2A, memória persistente, proteções, agentes na nuvem, furtividade de impressão digital TLS, Desktop/Termux/PWA e 67 localidades de interface do usuário para i18n. O OmniRoute é licenciado sob MIT e pode ser hospedado por você. Os recursos e as contagens dos concorrentes podem mudar; consulte a metodologia vinculada. 📊 Metodologia completa e detalhes por recurso em comparação com 9router, OpenRouter, CLIProxyAPI e LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ O Radar é opcional e usa apenas GET. O cliente OmniRoute não envia prompts, tr - **🖼️ Novos endpoints** — `/v1/ocr` (Mistral OCR) e `/v1/audio/translations` (no estilo Whisper) completam a superfície de mídia. → [Referência da API](docs/reference/API_REFERENCE.md) - **🎨 Geração de imagens / vídeos / áudio** — uma API para mídia: xAI Grok Imagine e vídeos da Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind e provedores de fala como ElevenLabs. → [Referência da API](docs/reference/API_REFERENCE.md) - **🌍 Implantação e operações** — `basePath` de proxy reverso, detecção automática do idioma do navegador, rastreamento de dispositivos por chave, confiança MITM sem root e localização zh-TW. → [Ambiente](docs/reference/ENVIRONMENT.md) -- **🤝 Mais provedores e agentes** — agentes na nuvem (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) com navegador + login OAuth, cartão de primeira classe do Ollama, Claude Opus 5 e Sonnet 5, parceria oficial com a Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… e um catálogo de **383 provedores** atualizado. → [Provedores](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Mais provedores e agentes** — agentes na nuvem (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) com navegador + login OAuth, cartão de primeira classe do Ollama, Claude Opus 5 e Sonnet 5, parceria oficial com a Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… e um catálogo de **384 provedores** atualizado. → [Provedores](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparência do roteamento** — cada resposta traz um cabeçalho `X-OmniRoute-Decision` que identifica a estratégia/provedor/latência que a processou; uma nova estratégia de combinação `cache-optimized` + o fator `cacheAffinity` do Auto-Combo redirecionam solicitações repetidas para a conexão que contém o prefixo em cache; e um endpoint somente leitura `/v1/auto-combo/{channel}/candidates` expõe o conjunto de candidatos ativo de um canal `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Desempenho e infraestrutura locais** — Redis local com um clique, implementadores de relay para Cloudflare Workers / Deno Deploy, Bifrost e Mux como serviços incorporados supervisionados. → [Serviços incorporados](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Também incluído** — framework de plugins + marketplace, frameworks de skills Omni/Agent/GitHub, integração com cofres do Obsidian (22 ferramentas MCP), APIs Batch e Files compatíveis com OpenAI, cache semântico de respostas, gamificação com tabelas de classificação, descoberta de agentes ACP (15 agentes integrados), exportação agendada de logs para o BigQuery, distribuição paralela de vários modelos com `auto/chaos`, uma ponte para bots do Telegram, um gerenciador de versões integrado ao aplicativo e rankings de provedores gratuitos LMArena-ELO. → [Documentação](docs/README.md) @@ -689,11 +689,11 @@ do histórico do seu shell. → [Integrações CLI](docs/guides/CLI-INTEGRATIONS
-## 🌐 383 provedores de IA — 153 marcados como gratuitos no catálogo +## 🌐 384 provedores de IA — 153 marcados como gratuitos no catálogo
-> **383 provedores registrados** nas coleções canônicas de chat, mídia, pesquisa, locais, agentes de nuvem e sistema, incluindo **153 com metadados de descoberta `hasFree: true`**. O registro de modelos de chat abrange **239 provedores / 3.059 pares distintos de provedor-modelo / 1.475 IDs de modelo brutos**; o catálogo separado de orçamento gratuito tem **471 linhas por modelo**, **34 pools recorrentes** e **52 provedores gratuitos para sempre, recorrentes e sem chave**. Esses são denominadores diferentes por definição; as definições e os cálculos deduplicados por pool estão disponíveis em [Referência de provedores](docs/reference/PROVIDER_REFERENCE.md) e [Planos gratuitos](docs/reference/FREE_TIERS.md). +> **384 provedores registrados** nas coleções canônicas de chat, mídia, pesquisa, locais, agentes de nuvem e sistema, incluindo **153 com metadados de descoberta `hasFree: true`**. O registro de modelos de chat abrange **239 provedores / 3.059 pares distintos de provedor-modelo / 1.475 IDs de modelo brutos**; o catálogo separado de orçamento gratuito tem **471 linhas por modelo**, **34 pools recorrentes** e **52 provedores gratuitos para sempre, recorrentes e sem chave**. Esses são denominadores diferentes por definição; as definições e os cálculos deduplicados por pool estão disponíveis em [Referência de provedores](docs/reference/PROVIDER_REFERENCE.md) e [Planos gratuitos](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ Métricas canônicas em 2026-10-18: **1.976 vídeos únicos** · **20.220.289 vi RuntimeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 LinguagemTypeScript 6.0 — 100% TypeScript em src/ e open-sse/ (any zero no núcleo desde a v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Banco de dadosbetter-sqlite3 (SQLite, journaling WAL) + LowDB (JSON legado) — 147 módulos de domínio, 230 migrações + Banco de dadosbetter-sqlite3 (SQLite, journaling WAL) + LowDB (JSON legado) — 147 módulos de domínio, 232 migrações MemóriaTexto completo com SQLite FTS5 + embeddings vetoriais quantizados em int8, decaimento tipado EsquemasZod 4 — validação de I/O de ferramentas MCP + contratos de API ProtocolosMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/pt-BR/llm.txt b/docs/i18n/pt-BR/llm.txt index c93f96aab4b..e1755d1b57f 100644 --- a/docs/i18n/pt-BR/llm.txt +++ b/docs/i18n/pt-BR/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/pt/README.md b/docs/i18n/pt/README.md index 84ce356dcb6..83a97ebc755 100644 --- a/docs/i18n/pt/README.md +++ b/docs/i18n/pt/README.md @@ -15,7 +15,7 @@ **Criado por [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Nunca pare de programar. Todas as ferramentas de IA → 383 fornecedores — 150+ gratuitos — através de um único endpoint. Claude Code, Codex, Cursor, Cline, Copilot e Antigravity para Claude / GPT / Gemini GRATUITOS com fallback automático. A compressão combinada RTK + Caveman poupa 15–95% dos tokens (~89% em média) — nunca atinja os limites. 383 fornecedores de IA · 150+ níveis gratuitos · ~1,62 mil milhões de tokens gratuitos/mês · 20 estratégias de encaminhamento · $0 para começar. +OmniRoute — Nunca pare de programar. Todas as ferramentas de IA → 384 fornecedores — 150+ gratuitos — através de um único endpoint. Claude Code, Codex, Cursor, Cline, Copilot e Antigravity para Claude / GPT / Gemini GRATUITOS com fallback automático. A compressão combinada RTK + Caveman poupa 15–95% dos tokens (~89% em média) — nunca atinja os limites. 384 fornecedores de IA · 150+ níveis gratuitos · ~1,62 mil milhões de tokens gratuitos/mês · 20 estratégias de encaminhamento · $0 para começar.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------------------ | :----------------------: | :---------: | :---------------------: | -| 🌐 Fornecedores | 352 | **358** | 383 | +| 🌐 Fornecedores | 352 | **358** | 384 | | 🧠 IDs únicos de modelos de chat | 1320 | **1374** | 1491 | | 🖼️ Ponte de modalidades | 🆕 visão + áudio + vídeo | ✓ | ✓ | | 📡 Catálogo gratuito do Radar | 🆕 opt-in | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funcionalidades 🎯 Combinações - 🌐 Fornecedores + 🌐 Fornecedores 🔌 CLI e MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -A promessa — Um endpoint e 383 fornecedores. O fallback automático mantém o encaminhamento enquanto estiver disponível outro destino em bom estado. Seis pilares: fallback resiliente entre 383 fornecedores · até 95% de poupança de tokens em cargas de trabalho elegíveis · $0 para começar, com mais de 150 escalões gratuitos e 52 fornecedores gratuitos recorrentes/sem chave para sempre · 39 integrações de CLI/agentes através de uma única configuração · compatibilidade com as APIs OpenAI, Claude, Gemini e Responses em /v1 · controlos de produção, incluindo disjuntores, furtividade TLS, 110 ferramentas MCP, A2A, memória, guardrails, avaliações e mais de 56 000 declarações de testes estáticos em mais de 7 800 ficheiros de testes monitorizados. +A promessa — Um endpoint e 384 fornecedores. O fallback automático mantém o encaminhamento enquanto estiver disponível outro destino em bom estado. Seis pilares: fallback resiliente entre 384 fornecedores · até 95% de poupança de tokens em cargas de trabalho elegíveis · $0 para começar, com mais de 150 escalões gratuitos e 52 fornecedores gratuitos recorrentes/sem chave para sempre · 39 integrações de CLI/agentes através de uma única configuração · compatibilidade com as APIs OpenAI, Claude, Gemini e Responses em /v1 · controlos de produção, incluindo disjuntores, furtividade TLS, 110 ferramentas MCP, A2A, memória, guardrails, avaliações e mais de 56 000 declarações de testes estáticos em mais de 7 800 ficheiros de testes monitorizados.

@@ -510,7 +510,7 @@ Todas as **20** estratégias — combine-as livremente em cada etapa do combo: -O que distingue o OmniRoute — uma comparação datada de funcionalidades com o 9router, OpenRouter, CLIProxyAPI e LiteLLM em 13 capacidades. OmniRoute: 383 fornecedores, mais de 150 níveis gratuitos integrados, 20 estratégias de encaminhamento, compressão de tokens com 12 motores, servidor MCP integrado com 110 ferramentas, protocolo de agentes A2A, memória persistente, salvaguardas, agentes na cloud, furtividade de impressão digital TLS, Desktop/Termux/PWA e 67 localidades de interface i18n. O OmniRoute é licenciado ao abrigo da MIT e pode ser autoalojado. As capacidades e contagens dos concorrentes podem mudar; consulte a metodologia associada. +O que distingue o OmniRoute — uma comparação datada de funcionalidades com o 9router, OpenRouter, CLIProxyAPI e LiteLLM em 13 capacidades. OmniRoute: 384 fornecedores, mais de 150 níveis gratuitos integrados, 20 estratégias de encaminhamento, compressão de tokens com 12 motores, servidor MCP integrado com 110 ferramentas, protocolo de agentes A2A, memória persistente, salvaguardas, agentes na cloud, furtividade de impressão digital TLS, Desktop/Termux/PWA e 67 localidades de interface i18n. O OmniRoute é licenciado ao abrigo da MIT e pode ser autoalojado. As capacidades e contagens dos concorrentes podem mudar; consulte a metodologia associada. 📊 Metodologia completa e detalhes por funcionalidade face ao 9router, OpenRouter, CLIProxyAPI e LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ O Radar é opcional e utiliza apenas pedidos GET. O cliente OmniRoute não envia - **🖼️ Novos endpoints** — `/v1/ocr` (Mistral OCR) e `/v1/audio/translations` (ao estilo do Whisper) completam a superfície multimédia. → [Referência da API](docs/reference/API_REFERENCE.md) - **🎨 Geração de imagens / vídeo / áudio** — uma API para multimédia: vídeo xAI Grok Imagine e Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind e fornecedores de voz como ElevenLabs. → [Referência da API](docs/reference/API_REFERENCE.md) - **🌍 Implementação e operações** — `basePath` de proxy inverso, deteção automática do idioma do navegador, monitorização de dispositivos por chave, confiança MITM sem root e localização zh-TW. → [Ambiente](docs/reference/ENVIRONMENT.md) -- **🤝 Mais fornecedores e agentes** — agentes na cloud (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) com início de sessão por navegador + OAuth, cartão de primeira classe do Ollama, Claude Opus 5 e Sonnet 5, parceria oficial com a Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… e um catálogo de **383 fornecedores** renovado. → [Fornecedores](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Mais fornecedores e agentes** — agentes na cloud (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) com início de sessão por navegador + OAuth, cartão de primeira classe do Ollama, Claude Opus 5 e Sonnet 5, parceria oficial com a Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… e um catálogo de **384 fornecedores** renovado. → [Fornecedores](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparência do roteamento** — cada resposta inclui um cabeçalho `X-OmniRoute-Decision` que indica a estratégia/fornecedor/latência que a serviu; uma nova estratégia de combinação `cache-optimized` + o fator `cacheAffinity` das combinações automáticas encaminha os pedidos repetidos de volta para a ligação que contém o prefixo em cache; e um endpoint apenas de leitura `/v1/auto-combo/{channel}/candidates` expõe o conjunto de candidatos ativos de um canal `auto/*`. → [Combinações automáticas](docs/routing/AUTO-COMBO.md) - **⚡ Desempenho e infraestrutura locais** — Redis local com um clique, implementadores de relay para Cloudflare Workers / Deno Deploy, Bifrost e Mux como serviços incorporados supervisionados. → [Serviços incorporados](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Também incluído** — framework de plugins + marketplace, frameworks de competências Omni/Agent/GitHub, integração com cofres do Obsidian (22 ferramentas MCP), APIs Batch e Files compatíveis com OpenAI, cache semântico de respostas, gamificação com tabelas de classificação, descoberta de agentes ACP (15 agentes integrados), exportação agendada de registos para o BigQuery, distribuição paralela de múltiplos modelos `auto/chaos`, uma ponte para bots do Telegram, um gestor de versões integrado na aplicação e classificações de fornecedores gratuitos LMArena-ELO. → [Documentação](docs/README.md) @@ -689,11 +689,11 @@ do histórico do seu shell. → [Integrações CLI](docs/guides/CLI-INTEGRATIONS
-## 🌐 383 fornecedores de IA — 153 marcados como gratuitos no catálogo +## 🌐 384 fornecedores de IA — 153 marcados como gratuitos no catálogo
-> **383 fornecedores registados** nas coleções canónicas de chat, multimédia, pesquisa, locais, agentes na cloud e sistema, incluindo **153 com metadados de descoberta `hasFree: true`**. O registo de modelos de chat abrange **239 fornecedores / 3 059 pares distintos fornecedor-modelo / 1 475 IDs de modelos brutos**; o catálogo separado de orçamentos gratuitos tem **471 linhas por modelo**, **34 grupos recorrentes** e **52 fornecedores gratuitos para sempre, recorrentes e sem chave**. Estes são denominadores diferentes por conceção; as definições e os cálculos deduplicados por grupo estão disponíveis na [Referência de fornecedores](docs/reference/PROVIDER_REFERENCE.md) e nos [Níveis gratuitos](docs/reference/FREE_TIERS.md). +> **384 fornecedores registados** nas coleções canónicas de chat, multimédia, pesquisa, locais, agentes na cloud e sistema, incluindo **153 com metadados de descoberta `hasFree: true`**. O registo de modelos de chat abrange **239 fornecedores / 3 059 pares distintos fornecedor-modelo / 1 475 IDs de modelos brutos**; o catálogo separado de orçamentos gratuitos tem **471 linhas por modelo**, **34 grupos recorrentes** e **52 fornecedores gratuitos para sempre, recorrentes e sem chave**. Estes são denominadores diferentes por conceção; as definições e os cálculos deduplicados por grupo estão disponíveis na [Referência de fornecedores](docs/reference/PROVIDER_REFERENCE.md) e nos [Níveis gratuitos](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ Métricas canónicas em 2026-10-18: **1.976 vídeos únicos** · **20.220.289 vi RuntimeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 LinguagemTypeScript 6.0 — 100% TypeScript em src/ e open-sse/ (zero any no núcleo desde a v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Base de dadosbetter-sqlite3 (SQLite, journaling WAL) + LowDB (JSON legado) — 147 módulos de domínio, 230 migrações + Base de dadosbetter-sqlite3 (SQLite, journaling WAL) + LowDB (JSON legado) — 147 módulos de domínio, 232 migrações MemóriaPesquisa de texto integral SQLite FTS5 + embeddings vetoriais quantizados em int8, decaimento tipado EsquemasZod 4 — validação de I/O de ferramentas MCP + contratos de API ProtocolosMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/pt/llm.txt b/docs/i18n/pt/llm.txt index ced8dab1c57..0cb502c552d 100644 --- a/docs/i18n/pt/llm.txt +++ b/docs/i18n/pt/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ro/README.md b/docs/i18n/ro/README.md index 012eb8274ee..8a010a34c98 100644 --- a/docs/i18n/ro/README.md +++ b/docs/i18n/ro/README.md @@ -15,7 +15,7 @@ **Creat de [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Nu te opri niciodată din programat. Fiecare instrument AI → 383 de furnizori — peste 150 gratuit — printr-un singur endpoint. Claude Code, Codex, Cursor, Cline, Copilot și Antigravity către Claude / GPT / Gemini GRATUITE, cu fallback automat. Compresia combinată RTK + Caveman economisește 15–95% din tokenuri (aprox. 89% în medie) — nu atinge niciodată limitele. 383 de furnizori AI · peste 150 de niveluri gratuite · aprox. 1,62 miliarde de tokenuri gratuite/lună · 20 de strategii de rutare · începe de la 0 $. +OmniRoute — Nu te opri niciodată din programat. Fiecare instrument AI → 384 de furnizori — peste 150 gratuit — printr-un singur endpoint. Claude Code, Codex, Cursor, Cline, Copilot și Antigravity către Claude / GPT / Gemini GRATUITE, cu fallback automat. Compresia combinată RTK + Caveman economisește 15–95% din tokenuri (aprox. 89% în medie) — nu atinge niciodată limitele. 384 de furnizori AI · peste 150 de niveluri gratuite · aprox. 1,62 miliarde de tokenuri gratuite/lună · 20 de strategii de rutare · începe de la 0 $.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------------- | :------------------------: | :---------: | :-------------------: | -| 🌐 Furnizori | 352 | **358** | 383 | +| 🌐 Furnizori | 352 | **358** | 384 | | 🧠 ID-uri unice de modele de chat | 1320 | **1374** | 1491 | | 🖼️ Punte de modalități | 🆕 viziune + audio + video | ✓ | ✓ | | 📡 Catalog gratuit Radar | 🆕 opt-in | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funcționalități 🎯 Combinații - 🌐 Furnizori + 🌐 Furnizori 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Promisiunea — Un singur endpoint și 383 de furnizori. Fallback-ul automat menține direcționarea atât timp cât este disponibilă o altă țintă funcțională. Șase piloni: fallback rezilient între 383 de furnizori · economii de până la 95% la tokenuri pentru sarcinile eligibile · 0 $ pentru început, cu peste 150 de niveluri gratuite și 52 de furnizori recurenți/fără cheie, gratuiți pentru totdeauna · 39 de integrări CLI/agent printr-o singură configurație · compatibilitate cu API-urile OpenAI, Claude, Gemini și Responses la /v1 · controale pentru producție, inclusiv întrerupătoare de circuit, discreție TLS, 110 instrumente MCP, A2A, memorie, guardrails, evaluări și peste 56.000 de declarații statice de testare în peste 7.800 de fișiere de testare urmărite. +Promisiunea — Un singur endpoint și 384 de furnizori. Fallback-ul automat menține direcționarea atât timp cât este disponibilă o altă țintă funcțională. Șase piloni: fallback rezilient între 384 de furnizori · economii de până la 95% la tokenuri pentru sarcinile eligibile · 0 $ pentru început, cu peste 150 de niveluri gratuite și 52 de furnizori recurenți/fără cheie, gratuiți pentru totdeauna · 39 de integrări CLI/agent printr-o singură configurație · compatibilitate cu API-urile OpenAI, Claude, Gemini și Responses la /v1 · controale pentru producție, inclusiv întrerupătoare de circuit, discreție TLS, 110 instrumente MCP, A2A, memorie, guardrails, evaluări și peste 56.000 de declarații statice de testare în peste 7.800 de fișiere de testare urmărite.

@@ -510,7 +510,7 @@ Toate cele **20** de strategii — combină-le după preferință pentru fiecare -Ce diferențiază OmniRoute — o prezentare comparativă actualizată a funcționalităților față de 9router, OpenRouter, CLIProxyAPI și LiteLLM pentru 13 capabilități. OmniRoute: 383 de furnizori, peste 150 de niveluri gratuite integrate, 20 de strategii de rutare, compresia tokenurilor prin 12 motoare, server MCP integrat cu 110 instrumente, protocol pentru agenți A2A, memorie persistentă, mecanisme de protecție, agenți cloud, metode de evitare a detectării prin amprenta TLS, Desktop/Termux/PWA și 67 de localizări i18n ale interfeței. OmniRoute este licențiat sub MIT și poate fi găzduit autonom. Capabilitățile și numărul acestora pentru concurenți se pot schimba; consultați metodologia din link. +Ce diferențiază OmniRoute — o prezentare comparativă actualizată a funcționalităților față de 9router, OpenRouter, CLIProxyAPI și LiteLLM pentru 13 capabilități. OmniRoute: 384 de furnizori, peste 150 de niveluri gratuite integrate, 20 de strategii de rutare, compresia tokenurilor prin 12 motoare, server MCP integrat cu 110 instrumente, protocol pentru agenți A2A, memorie persistentă, mecanisme de protecție, agenți cloud, metode de evitare a detectării prin amprenta TLS, Desktop/Termux/PWA și 67 de localizări i18n ale interfeței. OmniRoute este licențiat sub MIT și poate fi găzduit autonom. Capabilitățile și numărul acestora pentru concurenți se pot schimba; consultați metodologia din link. 📊 Metodologia completă & detalii pentru fiecare funcționalitate față de 9router, OpenRouter, CLIProxyAPI & LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -613,7 +613,7 @@ catalogul actual la **[radar.omniroute.online/planos](https://radar.omniroute.on - **🖼️ Endpoint-uri noi** — `/v1/ocr` (Mistral OCR) și `/v1/audio/translations` (de tip Whisper) completează suprafața media. → [Referință API](docs/reference/API_REFERENCE.md) - **🎨 Generare de imagini / videoclipuri / audio** — un singur API pentru media: xAI Grok Imagine și videoclipuri Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind și furnizori de vorbire precum ElevenLabs. → [Referință API](docs/reference/API_REFERENCE.md) - **🌍 Implementare și operațiuni** — `basePath` pentru proxy invers, detectarea automată a limbii browserului, urmărirea dispozitivelor per cheie, încredere MITM fără privilegii root, localizare zh-TW. → [Mediu](docs/reference/ENVIRONMENT.md) -- **🤝 Mai mulți furnizori și agenți** — agenți cloud (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) cu autentificare în browser + OAuth, card de primă clasă pentru Ollama, Claude Opus 5 și Sonnet 5, parteneriat oficial Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… și un **catalog cu 383 de furnizori** actualizat. → [Furnizori](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Mai mulți furnizori și agenți** — agenți cloud (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) cu autentificare în browser + OAuth, card de primă clasă pentru Ollama, Claude Opus 5 și Sonnet 5, parteneriat oficial Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… și un **catalog cu 384 de furnizori** actualizat. → [Furnizori](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparența rutării** — fiecare răspuns include un antet `X-OmniRoute-Decision` care indică strategia/furnizorul/latența care l-au deservit, o nouă strategie combo `cache-optimized` + factorul de rutare Auto-Combo `cacheAffinity` redirecționează solicitările repetate către conexiunea care deține prefixul memorat în cache, iar un endpoint numai pentru citire `/v1/auto-combo/{channel}/candidates` expune grupul de candidați activi al unui canal `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Performanță locală și infrastructură** — Redis local cu un singur clic, deployere pentru relee Cloudflare Workers / Deno Deploy, Bifrost și Mux ca servicii încorporate supervizate. → [Servicii încorporate](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Inclus în pachet** — framework pentru pluginuri + marketplace, frameworkuri de skill-uri Omni/Agent/GitHub, integrare cu vault-uri Obsidian (22 de instrumente MCP), API-uri Batch și Files compatibile cu OpenAI, cache semantic pentru răspunsuri, gamificare cu clasamente, descoperirea agenților ACP (15 agenți încorporați), export programat al jurnalelor în BigQuery, distribuire paralelă multi-model `auto/chaos`, o punte pentru boți Telegram, un manager de versiuni în aplicație și clasamente LMArena-ELO ale furnizorilor gratuiți. → [Documentație](docs/README.md) @@ -697,11 +697,11 @@ previzualizează exact env/args fără a executa, iar `--api-key-env NAME` păst
-## 🌐 383 de furnizori AI — 153 marcați ca gratuiți în catalog +## 🌐 384 de furnizori AI — 153 marcați ca gratuiți în catalog
-> **383 de furnizori înregistrați** în colecțiile canonice de chat, media, căutare, locale, agenți cloud și sistem, inclusiv **153 care au metadatele de descoperire `hasFree: true`**. Registrul modelelor de chat acoperă **239 de furnizori / 3.059 de perechi distincte furnizor-model / 1.475 de ID-uri brute de modele**; catalogul separat cu buget gratuit are **471 de rânduri per model**, **34 de pool-uri recurente** și **52 de furnizori recurenți/gratuiți permanent, fără cheie**. Aceștia sunt numitori diferiți în mod intenționat; definițiile și calculele deduplicate la nivel de pool se găsesc în [Referința furnizorilor](docs/reference/PROVIDER_REFERENCE.md) și [Niveluri gratuite](docs/reference/FREE_TIERS.md). +> **384 de furnizori înregistrați** în colecțiile canonice de chat, media, căutare, locale, agenți cloud și sistem, inclusiv **153 care au metadatele de descoperire `hasFree: true`**. Registrul modelelor de chat acoperă **239 de furnizori / 3.059 de perechi distincte furnizor-model / 1.475 de ID-uri brute de modele**; catalogul separat cu buget gratuit are **471 de rânduri per model**, **34 de pool-uri recurente** și **52 de furnizori recurenți/gratuiți permanent, fără cheie**. Aceștia sunt numitori diferiți în mod intenționat; definițiile și calculele deduplicate la nivel de pool se găsesc în [Referința furnizorilor](docs/reference/PROVIDER_REFERENCE.md) și [Niveluri gratuite](docs/reference/FREE_TIERS.md).
@@ -1303,7 +1303,7 @@ Metrici canonice la 2026-10-18: **1.976 de videoclipuri unice** · **20.220.289 RuntimeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 LimbajTypeScript 6.0 — 100% TypeScript în src/ și open-sse/ (zero any în nucleu începând cu v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Bază de datebetter-sqlite3 (SQLite, jurnalizare WAL) + LowDB (JSON legacy) — 147 de module de domeniu, 230 de migrări + Bază de datebetter-sqlite3 (SQLite, jurnalizare WAL) + LowDB (JSON legacy) — 147 de module de domeniu, 232 de migrări MemorieCăutare full-text SQLite FTS5 + înglobări vectoriale cuantizate int8, cu degradare tipizată SchemeZod 4 — validarea I/O a instrumentelor MCP + contracte API ProtocoaleMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ro/llm.txt b/docs/i18n/ro/llm.txt index 8d144fa742c..18abb592073 100644 --- a/docs/i18n/ro/llm.txt +++ b/docs/i18n/ro/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ru/README.md b/docs/i18n/ru/README.md index 39f8e076742..9c98cc2c040 100644 --- a/docs/i18n/ru/README.md +++ b/docs/i18n/ru/README.md @@ -15,7 +15,7 @@ **Создано [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Никогда не прекращайте программировать. Каждый инструмент ИИ → 383 провайдер — 150+ бесплатных — через одну конечную точку. Claude Code, Codex, Cursor, Cline, Copilot и Antigravity для БЕСПЛАТНЫХ Claude / GPT / Gemini с автоматическим резервным переключением. Компрессия RTK + Caveman в несколько этапов экономит 15–95% токенов (~89% в среднем) — лимиты больше не помеха. 383 провайдер ИИ · 150+ бесплатных тарифов · ~1,62 млрд бесплатных токенов в месяц · 20 стратегий маршрутизации · старт за $0. +OmniRoute — Никогда не прекращайте программировать. Каждый инструмент ИИ → 384 провайдер — 150+ бесплатных — через одну конечную точку. Claude Code, Codex, Cursor, Cline, Copilot и Antigravity для БЕСПЛАТНЫХ Claude / GPT / Gemini с автоматическим резервным переключением. Компрессия RTK + Caveman в несколько этапов экономит 15–95% токенов (~89% в среднем) — лимиты больше не помеха. 384 провайдер ИИ · 150+ бесплатных тарифов · ~1,62 млрд бесплатных токенов в месяц · 20 стратегий маршрутизации · старт за $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ---------------------------------------- | :-----------------------: | :---------: | :-------------------: | -| 🌐 Провайдеры | 352 | **358** | 383 | +| 🌐 Провайдеры | 352 | **358** | 384 | | 🧠 Уникальные идентификаторы чат-моделей | 1320 | **1374** | 1491 | | 🖼️ Мост модальностей | 🆕 зрение + аудио + видео | ✓ | ✓ | | 📡 Бесплатный каталог Radar | 🆕 по желанию | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Возможности 🎯 Комбинации - 🌐 Провайдеры + 🌐 Провайдеры 🔌 CLI и MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Обещание — одна конечная точка и 383 провайдер. Автоматический откат сохраняет маршрутизацию, пока доступна другая работоспособная цель. Шесть основных преимуществ: устойчивый откат между 383 провайдером · экономия до 95% токенов для подходящих рабочих нагрузок · $0 для начала, 150+ бесплатных тарифов и 52 регулярно предоставляемых/бесплатных навсегда провайдера без ключа · 39 интеграций CLI/агентов через одну конфигурацию · совместимость с API OpenAI, Claude, Gemini и Responses по адресу /v1 · средства управления для продакшена, включая автоматические выключатели, скрытие TLS, MCP со 110 инструментами, A2A, память, защитные ограничения, оценивание и более 56 000 статических тестовых деклараций в 7 800+ отслеживаемых тестовых файлах. +Обещание — одна конечная точка и 384 провайдер. Автоматический откат сохраняет маршрутизацию, пока доступна другая работоспособная цель. Шесть основных преимуществ: устойчивый откат между 384 провайдером · экономия до 95% токенов для подходящих рабочих нагрузок · $0 для начала, 150+ бесплатных тарифов и 52 регулярно предоставляемых/бесплатных навсегда провайдера без ключа · 39 интеграций CLI/агентов через одну конфигурацию · совместимость с API OpenAI, Claude, Gemini и Responses по адресу /v1 · средства управления для продакшена, включая автоматические выключатели, скрытие TLS, MCP со 110 инструментами, A2A, память, защитные ограничения, оценивание и более 56 000 статических тестовых деклараций в 7 800+ отслеживаемых тестовых файлах.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -Чем OmniRoute отличается от других — актуальный на определённую дату снимок возможностей по сравнению с 9router, OpenRouter, CLIProxyAPI и LiteLLM по 13 направлениям. OmniRoute: 383 провайдер, более 150 встроенных бесплатных тарифов, 20 стратегий маршрутизации, сжатие токенов на базе 12 движков, встроенный MCP-сервер со 110 инструментами, протокол агентов A2A, постоянная память, защитные ограничения, облачные агенты, скрытность TLS-отпечатка, Desktop/Termux/PWA и 67 языковых локалей интерфейса i18n. OmniRoute распространяется по лицензии MIT и может размещаться на собственных серверах. Возможности и количество функций у конкурентов могут измениться; см. методологию по ссылке. +Чем OmniRoute отличается от других — актуальный на определённую дату снимок возможностей по сравнению с 9router, OpenRouter, CLIProxyAPI и LiteLLM по 13 направлениям. OmniRoute: 384 провайдер, более 150 встроенных бесплатных тарифов, 20 стратегий маршрутизации, сжатие токенов на базе 12 движков, встроенный MCP-сервер со 110 инструментами, протокол агентов A2A, постоянная память, защитные ограничения, облачные агенты, скрытность TLS-отпечатка, Desktop/Termux/PWA и 67 языковых локалей интерфейса i18n. OmniRoute распространяется по лицензии MIT и может размещаться на собственных серверах. Возможности и количество функций у конкурентов могут измениться; см. методологию по ссылке. 📊 Полная методология и подробные сведения по каждой функции в сравнении с 9router, OpenRouter, CLIProxyAPI и LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar используется по желанию и выполняет тол - **🖼️ Новые конечные точки** — `/v1/ocr` (Mistral OCR) и `/v1/audio/translations` (в стиле Whisper) дополняют набор возможностей для работы с медиа. → [Справочник API](docs/reference/API_REFERENCE.md) - **🎨 Генерация изображений, видео и аудио** — единый API для медиа: xAI Grok Imagine и видео Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind и провайдеры речи, такие как ElevenLabs. → [Справочник API](docs/reference/API_REFERENCE.md) - **🌍 Развёртывание и эксплуатация** — `basePath` для обратного прокси, автоматическое определение языка браузера, отслеживание устройств для каждого ключа, доверие к MITM без root-доступа, локализация zh-TW. → [Переменные окружения](docs/reference/ENVIRONMENT.md) -- **🤝 Больше провайдеров и агентов** — облачные агенты (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) с входом через браузер и OAuth, полноценная карточка Ollama, Claude Opus 5 и Sonnet 5, официальное партнёрство с Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… и обновлённый **каталог из 383 провайдера**. → [Провайдеры](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Больше провайдеров и агентов** — облачные агенты (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) с входом через браузер и OAuth, полноценная карточка Ollama, Claude Opus 5 и Sonnet 5, официальное партнёрство с Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… и обновлённый **каталог из 384 провайдера**. → [Провайдеры](docs/reference/PROVIDER_REFERENCE.md) - **📡 Прозрачность маршрутизации** — каждый ответ содержит заголовок `X-OmniRoute-Decision` с указанием стратегии, провайдера и задержки, обслуживших запрос; новая стратегия комбинирования `cache-optimized` и фактор маршрута `cacheAffinity` в Auto-Combo направляют повторные запросы обратно к подключению, содержащему кэшированный префикс; доступная только для чтения конечная точка `/v1/auto-combo/{channel}/candidates` показывает текущий пул кандидатов канала `auto/*`. → [Автоматическое комбинирование](docs/routing/AUTO-COMBO.md) - **⚡ Локальная производительность и инфраструктура** — локальный Redis в один клик, средства развертывания ретрансляторов Cloudflare Workers / Deno Deploy, а также Bifrost и Mux в качестве контролируемых встроенных сервисов. → [Встроенные сервисы](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Также в комплекте** — фреймворк плагинов и маркетплейс, фреймворки навыков Omni/Agent/GitHub, интеграция с хранилищем Obsidian (22 инструмента MCP), совместимые с OpenAI API Batch и Files API, семантический кэш ответов, игровая система с таблицами лидеров, обнаружение агентов ACP (15 встроенных агентов), запланированный экспорт журналов в BigQuery, параллельное распределение запросов между несколькими моделями `auto/chaos`, мост для бота Telegram, встроенный менеджер версий и рейтинги бесплатных провайдеров LMArena-ELO. → [Документация](docs/README.md) @@ -689,11 +689,11 @@ omniroute configure codex # также: claude opencode qwen aider goos
-## 🌐 383 провайдер ИИ — 153 отмечены в каталоге как бесплатные +## 🌐 384 провайдер ИИ — 153 отмечены в каталоге как бесплатные
-> **383 зарегистрированный провайдер** в канонических коллекциях чатов, медиа, поиска, локальных, облачных агентов и систем, включая **153 провайдера с метаданными обнаружения `hasFree: true`**. Реестр моделей чата охватывает **239 провайдеров / 3,059 уникальных пар провайдер–модель / 1,475 исходных идентификаторов моделей**; отдельный каталог бесплатных тарифов содержит **471 строку для отдельных моделей**, **34 регулярно пополняемых пула** и **52 провайдера с регулярно возобновляемым или бессрочно бесплатным доступом без ключа**. Это разные знаменатели по замыслу; определения и расчёты с дедупликацией пулов приведены в [справочнике провайдеров](docs/reference/PROVIDER_REFERENCE.md) и разделе [Бесплатные тарифы](docs/reference/FREE_TIERS.md). +> **384 зарегистрированный провайдер** в канонических коллекциях чатов, медиа, поиска, локальных, облачных агентов и систем, включая **153 провайдера с метаданными обнаружения `hasFree: true`**. Реестр моделей чата охватывает **239 провайдеров / 3,059 уникальных пар провайдер–модель / 1,475 исходных идентификаторов моделей**; отдельный каталог бесплатных тарифов содержит **471 строку для отдельных моделей**, **34 регулярно пополняемых пула** и **52 провайдера с регулярно возобновляемым или бессрочно бесплатным доступом без ключа**. Это разные знаменатели по замыслу; определения и расчёты с дедупликацией пулов приведены в [справочнике провайдеров](docs/reference/PROVIDER_REFERENCE.md) и разделе [Бесплатные тарифы](docs/reference/FREE_TIERS.md).
@@ -1294,7 +1294,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 также про Среда выполненияNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ЯзыкTypeScript 6.0 — 100% TypeScript в src/ и open-sse/ (ноль any в ядре начиная с v2.0) ФреймворкNext.js 16 + React 19 + Tailwind CSS 4 - База данныхbetter-sqlite3 (SQLite, журналирование WAL) + LowDB (устаревший JSON) — 147 предметных модулей, 230 миграций + База данныхbetter-sqlite3 (SQLite, журналирование WAL) + LowDB (устаревший JSON) — 147 предметных модулей, 232 миграций ПамятьПолнотекстовый поиск SQLite FTS5 + векторные эмбеддинги с квантованием int8, типизированное затухание СхемыZod 4 — проверка ввода-вывода инструментов MCP + контракты API ПротоколыMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ru/llm.txt b/docs/i18n/ru/llm.txt index 5be25c74920..4d3d121bb51 100644 --- a/docs/i18n/ru/llm.txt +++ b/docs/i18n/ru/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/si/README.md b/docs/i18n/si/README.md index d3f6f977a4d..d74e6a31634 100644 --- a/docs/i18n/si/README.md +++ b/docs/i18n/si/README.md @@ -15,7 +15,7 @@ **නිර්මාණය කළේ [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — කේතකරණය කිසිදා නවත්වන්න එපා. සෑම AI මෙවලමක්ම → එක් endpoint එකක් හරහා providers 383ක් — නොමිලේ 150කට වැඩි ප්රමාණයක්. Claude Code, Codex, Cursor, Cline, Copilot සහ Antigravity, ස්වයංක්රීය fallback සමඟ FREE Claude / GPT / Gemini වෙත. RTK + Caveman එකට ගොඩගැසූ සම්පීඩනය මඟින් tokens 15–95%ක් (~89% සාමාන්යයෙන්) ඉතිරි කරයි — සීමාවන්ට කිසිදා ළඟා නොවන්න. AI providers 383ක් · නොමිලේ tiers 150කට වැඩි ප්රමාණයක් · මසකට ~1.62B නොමිලේ tokens · routing strategies 20ක් · ආරම්භ කිරීමට $0. +OmniRoute — කේතකරණය කිසිදා නවත්වන්න එපා. සෑම AI මෙවලමක්ම → එක් endpoint එකක් හරහා providers 384ක් — නොමිලේ 150කට වැඩි ප්රමාණයක්. Claude Code, Codex, Cursor, Cline, Copilot සහ Antigravity, ස්වයංක්රීය fallback සමඟ FREE Claude / GPT / Gemini වෙත. RTK + Caveman එකට ගොඩගැසූ සම්පීඩනය මඟින් tokens 15–95%ක් (~89% සාමාන්යයෙන්) ඉතිරි කරයි — සීමාවන්ට කිසිදා ළඟා නොවන්න. AI providers 384ක් · නොමිලේ tiers 150කට වැඩි ප්රමාණයක් · මසකට ~1.62B නොමිලේ tokens · routing strategies 20ක් · ආරම්භ කිරීමට $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------------- | :------------------------: | :---------: | :-------------------: | -| 🌐 සැපයුම්කරුවන් | 352 | **358** | 383 | +| 🌐 සැපයුම්කරුවන් | 352 | **358** | 384 | | 🧠 අනන්ය chat model IDs | 1320 | **1374** | 1491 | | 🖼️ මාදිලි සම්බන්ධකය | 🆕 දෘශ්ය + ශ්රව්ය + වීඩියෝ | ✓ | ✓ | | 📡 Radar නොමිලේ නාමාවලිය | 🆕 තෝරාගැනීමේ විකල්පය | ✓ | ✓ | @@ -113,7 +113,7 @@ ⚙️ විශේෂාංග 🎯 සංයෝජන - 🌐 සැපයුම්කරුවන් + 🌐 සැපයුම්කරුවන් 🔌 CLI & MCP @@ -249,7 +249,7 @@ curl http://localhost:20128/v1/chat/completions \ -පොරොන්දුව — එක් endpoint එකක් සහ providers 383ක්. තවත් සෞඛ්ය සම්පන්න target එකක් පවතින තාක් automatic fallback මඟින් routing අඛණ්ඩව පවත්වා ගනී. ප්රධාන කුලුනු හය: providers 383ක් හරහා resilient fallback · සුදුසු workloads සඳහා token ඉතිරිකිරීම 95% දක්වා · free tiers 150කට වැඩි ප්රමාණයක් සහ නැවත නැවත ලැබෙන/keyless free-forever providers 52ක් සමඟ ආරම්භ කිරීමට $0 · එක් config එකක් හරහා CLI/agent integrations 39ක් · /v1 හි OpenAI, Claude, Gemini සහ Responses API compatibility · circuit breakers, TLS stealth, MCP tools 110ක්, A2A, memory, guardrails, evals සහ tracked test files 7,800කට වැඩි ප්රමාණයක් හරහා static test declarations 56,000කට වැඩි ප්රමාණයක් ඇතුළත් production controls. +පොරොන්දුව — එක් endpoint එකක් සහ providers 384ක්. තවත් සෞඛ්ය සම්පන්න target එකක් පවතින තාක් automatic fallback මඟින් routing අඛණ්ඩව පවත්වා ගනී. ප්රධාන කුලුනු හය: providers 384ක් හරහා resilient fallback · සුදුසු workloads සඳහා token ඉතිරිකිරීම 95% දක්වා · free tiers 150කට වැඩි ප්රමාණයක් සහ නැවත නැවත ලැබෙන/keyless free-forever providers 52ක් සමඟ ආරම්භ කිරීමට $0 · එක් config එකක් හරහා CLI/agent integrations 39ක් · /v1 හි OpenAI, Claude, Gemini සහ Responses API compatibility · circuit breakers, TLS stealth, MCP tools 110ක්, A2A, memory, guardrails, evals සහ tracked test files 7,800කට වැඩි ප්රමාණයක් හරහා static test declarations 56,000කට වැඩි ප්රමාණයක් ඇතුළත් production controls.

@@ -509,7 +509,7 @@ combo එකක් සෑදීමට අවශ්ය නැත. ඔබේ ම -OmniRoute වෙනස් කරන්නේ කුමක්ද — 13 හැකියාවන් හරහා 9router, OpenRouter, CLIProxyAPI සහ LiteLLM සමඟ සසඳන දිනಾಂಕගත විශේෂාංග සාරාංශයක්. OmniRoute: providers 383ක්, අන්තර්ගතව ඇති නොමිලේ භාවිත ස්ථර 150කට වැඩි ප්රමාණයක්, routing strategies 20ක්, token compression engines 12ක්, tools 110ක් සහිත built-in MCP server එකක්, A2A agent protocol, persistent memory, guardrails, cloud agents, TLS fingerprint stealth, Desktop/Termux/PWA සහ i18n UI locales 67ක්. OmniRoute MIT-licensed වන අතර self-host කළ හැක. තරඟකරුවන්ගේ හැකියාවන් සහ ගණන් වෙනස් විය හැක; සබැඳි methodology බලන්න. +OmniRoute වෙනස් කරන්නේ කුමක්ද — 13 හැකියාවන් හරහා 9router, OpenRouter, CLIProxyAPI සහ LiteLLM සමඟ සසඳන දිනಾಂಕගත විශේෂාංග සාරාංශයක්. OmniRoute: providers 384ක්, අන්තර්ගතව ඇති නොමිලේ භාවිත ස්ථර 150කට වැඩි ප්රමාණයක්, routing strategies 20ක්, token compression engines 12ක්, tools 110ක් සහිත built-in MCP server එකක්, A2A agent protocol, persistent memory, guardrails, cloud agents, TLS fingerprint stealth, Desktop/Termux/PWA සහ i18n UI locales 67ක්. OmniRoute MIT-licensed වන අතර self-host කළ හැක. තරඟකරුවන්ගේ හැකියාවන් සහ ගණන් වෙනස් විය හැක; සබැඳි methodology බලන්න. 📊 9router, OpenRouter, CLIProxyAPI සහ LiteLLM සමඟ සසඳන සම්පූර්ණ methodology සහ එක් එක් විශේෂාංගයේ විස්තර → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -604,7 +604,7 @@ Radar භාවිතය තෝරාගත හැකි අතර එය GET - **🖼️ නව endpoints** — `/v1/ocr` (Mistral OCR) සහ `/v1/audio/translations` (Whisper-style) මඟින් media surface එක සම්පූර්ණ කරයි. → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 Image / video / audio generation** — media සඳහා එක් API එකක්: xAI Grok Imagine සහ Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind සහ ElevenLabs වැනි speech providers. → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 Deployment & ops** — reverse-proxy `basePath`, browser-language auto-detect, එක් එක් key සඳහා device tracking, root-less MITM trust, zh-TW localization. → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 තවත් providers සහ agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login සහිත Grok Build (xAI), පළමු පන්තියේ card එකක් ලෙස Ollama, Claude Opus 5 සහ Sonnet 5, Kimi official partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… සහ යාවත්කාලීන කළ **383-provider catalog** එකක්. → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 තවත් providers සහ agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login සහිත Grok Build (xAI), පළමු පන්තියේ card එකක් ලෙස Ollama, Claude Opus 5 සහ Sonnet 5, Kimi official partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… සහ යාවත්කාලීන කළ **384-provider catalog** එකක්. → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 Routing transparency** — සෑම response එකක්ම එය සේවය කළ strategy/provider/latency නම් කරන `X-OmniRoute-Decision` header එකක් රැගෙන එයි; නව `cache-optimized` combo strategy එක සහ Auto-Combo `cacheAffinity` factor එක මඟින් නැවත නැවත එන requests cached prefix එක රඳවාගෙන සිටින connection වෙත යොමු කරයි; read-only `/v1/auto-combo/{channel}/candidates` endpoint එකක් `auto/*` channel එකක සජීවී candidate pool එක හෙළි කරයි. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ දේශීය කාර්යසාධනය සහ යටිතල පහසුකම්** — එක් ක්ලික් කිරීමකින් දේශීය Redis, Cloudflare Workers / Deno Deploy relay deployers, අධීක්ෂණය යටතේ ක්රියාත්මක වන embedded services ලෙස Bifrost සහ Mux. → [Embedded Services](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 මෙහිම ඇතුළත්** — ප්ලගින framework එකක් සහ marketplace එකක්, Omni/Agent/GitHub skills frameworks, Obsidian vault ඒකාබද්ධකරණය (MCP tools 22ක්), OpenAI-compatible Batch & Files APIs, semantic response cache, leaderboards සහිත gamification, ACP agent discovery (built-in agents 15ක්), BigQuery වෙත නියමිත වේලාවට log export කිරීම, `auto/chaos` සමාන්තර multi-model fan-out, Telegram bot bridge එකක්, යෙදුම තුළම version manager එකක් සහ LMArena-ELO නොමිලේ සැපයුම්කරුවන්ගේ rankings. → [ලේඛන](docs/README.md) @@ -688,11 +688,11 @@ omniroute configure codex # මෙයද: claude opencode qwen aider go
-## 🌐 AI සපයන්නන් 383ක් — නාමාවලි-සලකුණු කළ නොමිලේ සපයන්නන් 153ක් +## 🌐 AI සපයන්නන් 384ක් — නාමාවලි-සලකුණු කළ නොමිලේ සපයන්නන් 153ක්
-> **ලියාපදිංචි කළ සපයන්නන් 383ක්** canonical chat, media, search, local, cloud-agent සහ system එකතු හරහා, **`hasFree: true` සොයාගැනීමේ metadata සහිත සපයන්නන් 153ක්** ඇතුළුව. Chat model registry එකේ **සපයන්නන් 239ක් / වෙනස් provider-model යුගල 3,059ක් / raw model IDs 1,475ක්** අඩංගු වේ; වෙනම free-budget catalog එකේ **එක් model එකකට පේළි 471ක්**, **නැවත නැවත ලැබෙන pools 34ක්** සහ **නැවත නැවත ලැබෙන/keyless සදාකාලික නොමිලේ සපයන්නන් 52ක්** ඇත. මේවා සැලසුම් සහගතව වෙනස් denominator වේ; අර්ථ දැක්වීම් සහ pool-deduped ගණනය කිරීම් [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) සහ [Free Tiers](docs/reference/FREE_TIERS.md) තුළ ඇත. +> **ලියාපදිංචි කළ සපයන්නන් 384ක්** canonical chat, media, search, local, cloud-agent සහ system එකතු හරහා, **`hasFree: true` සොයාගැනීමේ metadata සහිත සපයන්නන් 153ක්** ඇතුළුව. Chat model registry එකේ **සපයන්නන් 239ක් / වෙනස් provider-model යුගල 3,059ක් / raw model IDs 1,475ක්** අඩංගු වේ; වෙනම free-budget catalog එකේ **එක් model එකකට පේළි 471ක්**, **නැවත නැවත ලැබෙන pools 34ක්** සහ **නැවත නැවත ලැබෙන/keyless සදාකාලික නොමිලේ සපයන්නන් 52ක්** ඇත. මේවා සැලසුම් සහගතව වෙනස් denominator වේ; අර්ථ දැක්වීම් සහ pool-deduped ගණනය කිරීම් [Provider Reference](docs/reference/PROVIDER_REFERENCE.md) සහ [Free Tiers](docs/reference/FREE_TIERS.md) තුළ ඇත.
@@ -1294,7 +1294,7 @@ Dashboard රහිත, headless runtime එකක් සඳහා Docker `base` ධාවන පරිසරයNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 භාෂාවTypeScript 6.0 — src/ සහ open-sse/ පුරා 100% TypeScript (v2.0 සිට core තුළ any ශුන්යයි) රාමුවNext.js 16 + React 19 + Tailwind CSS 4 - දත්ත සමුදායbetter-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — domain මොඩියුල 147ක්, migrations 230ක් + දත්ත සමුදායbetter-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — domain මොඩියුල 147ක්, migrations 232ක් මතකයSQLite FTS5 පූර්ණ-පාඨ + int8-quantized vector embeddings, වර්ගීකරණය කළ ක්ෂය වීම ස්කීමාZod 4 — MCP මෙවලම් I/O වලංගුකරණය + API ගිවිසුම් ප්රොටෝකෝලMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/si/llm.txt b/docs/i18n/si/llm.txt index c40cecd25c9..0dabd4f23ca 100644 --- a/docs/i18n/si/llm.txt +++ b/docs/i18n/si/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/sk/README.md b/docs/i18n/sk/README.md index 6ec4ab0fa5b..5ccf8e6859f 100644 --- a/docs/i18n/sk/README.md +++ b/docs/i18n/sk/README.md @@ -15,7 +15,7 @@ **Vytvoril [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Nikdy neprestaňte programovať. Každý nástroj AI → 383 poskytovateľov — 150+ bezplatných — cez jeden endpoint. Claude Code, Codex, Cursor, Cline, Copilot a Antigravity do BEZPLATNÉHO Claude / GPT / Gemini s automatickým záložným riešením. Kompresia RTK + Caveman v niekoľkých vrstvách šetrí 15–95 % tokenov (~89 % v priemere) — nikdy nenarazíte na limity. 383 poskytovateľov AI · 150+ bezplatných úrovní · ~1,62 miliardy bezplatných tokenov/mes. · 20 stratégií smerovania · začnite za $0. +OmniRoute — Nikdy neprestaňte programovať. Každý nástroj AI → 384 poskytovateľov — 150+ bezplatných — cez jeden endpoint. Claude Code, Codex, Cursor, Cline, Copilot a Antigravity do BEZPLATNÉHO Claude / GPT / Gemini s automatickým záložným riešením. Kompresia RTK + Caveman v niekoľkých vrstvách šetrí 15–95 % tokenov (~89 % v priemere) — nikdy nenarazíte na limity. 384 poskytovateľov AI · 150+ bezplatných úrovní · ~1,62 miliardy bezplatných tokenov/mes. · 20 stratégií smerovania · začnite za $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ----------------------------------- | :------------------------: | :---------: | :-------------------: | -| 🌐 Poskytovatelia | 352 | **358** | 383 | +| 🌐 Poskytovatelia | 352 | **358** | 384 | | 🧠 Jedinečné ID chatovacích modelov | 1320 | **1374** | 1491 | | 🖼️ Most modalít | 🆕 obraz + zvuk + video | ✓ | ✓ | | 📡 Bezplatný katalóg Radar | 🆕 s možnosťou prihlásenia | ✓ | ✓ | @@ -113,7 +113,7 @@ ⚙️ Funkcie 🎯 Kombá - 🌐 Poskytovatelia + 🌐 Poskytovatelia 🔌 CLI a MCP @@ -249,7 +249,7 @@ curl http://localhost:20128/v1/chat/completions \ -Prísľub — jeden endpoint a 383 poskytovateľov. Automatický fallback udržiava smerovanie, kým je k dispozícii iný funkčný cieľ. Šesť pilierov: odolný fallback naprieč 383 poskytovateľmi · až 95 % úspora tokenov pri vhodných pracovných zaťaženiach · $0 na začiatok so 150+ bezplatnými úrovňami a 52 opakujúcimi sa/bezplatnými poskytovateľmi navždy bez kľúča · 39 integrácií CLI/agentov prostredníctvom jednej konfigurácie · kompatibilita s OpenAI, Claude, Gemini a Responses API na /v1 · produkčné ovládacie prvky vrátane ističov, TLS stealth, MCP so 110 nástrojmi, A2A, pamäte, ochranných mechanizmov, evaluácií a viac než 56 000 statických deklarácií testov v rámci viac než 7 800 sledovaných súborov s testami. +Prísľub — jeden endpoint a 384 poskytovateľov. Automatický fallback udržiava smerovanie, kým je k dispozícii iný funkčný cieľ. Šesť pilierov: odolný fallback naprieč 384 poskytovateľmi · až 95 % úspora tokenov pri vhodných pracovných zaťaženiach · $0 na začiatok so 150+ bezplatnými úrovňami a 52 opakujúcimi sa/bezplatnými poskytovateľmi navždy bez kľúča · 39 integrácií CLI/agentov prostredníctvom jednej konfigurácie · kompatibilita s OpenAI, Claude, Gemini a Responses API na /v1 · produkčné ovládacie prvky vrátane ističov, TLS stealth, MCP so 110 nástrojmi, A2A, pamäte, ochranných mechanizmov, evaluácií a viac než 56 000 statických deklarácií testov v rámci viac než 7 800 sledovaných súborov s testami.

@@ -509,7 +509,7 @@ Všetkých **20** stratégií — kombinujte ich podľa potreby v jednotlivých -Čím sa OmniRoute odlišuje — aktuálny prehľad funkcií v porovnaní s nástrojmi 9router, OpenRouter, CLIProxyAPI a LiteLLM v 13 oblastiach. OmniRoute: 383 poskytovateľov, viac než 150 integrovaných bezplatných úrovní, 20 stratégií smerovania, kompresia tokenov pomocou 12 nástrojov, integrovaný server MCP so 110 nástrojmi, protokol agentov A2A, trvalá pamäť, ochranné mechanizmy, cloudoví agenti, utajenie pomocou odtlačku TLS, Desktop/Termux/PWA a 67 jazykových lokalizácií používateľského rozhrania i18n. OmniRoute je licencovaný pod licenciou MIT a možno ho hostovať vo vlastnej réžii. Funkcie a počty konkurenčných nástrojov sa môžu meniť; pozrite si prepojenú metodiku. +Čím sa OmniRoute odlišuje — aktuálny prehľad funkcií v porovnaní s nástrojmi 9router, OpenRouter, CLIProxyAPI a LiteLLM v 13 oblastiach. OmniRoute: 384 poskytovateľov, viac než 150 integrovaných bezplatných úrovní, 20 stratégií smerovania, kompresia tokenov pomocou 12 nástrojov, integrovaný server MCP so 110 nástrojmi, protokol agentov A2A, trvalá pamäť, ochranné mechanizmy, cloudoví agenti, utajenie pomocou odtlačku TLS, Desktop/Termux/PWA a 67 jazykových lokalizácií používateľského rozhrania i18n. OmniRoute je licencovaný pod licenciou MIT a možno ho hostovať vo vlastnej réžii. Funkcie a počty konkurenčných nástrojov sa môžu meniť; pozrite si prepojenú metodiku. 📊 Kompletná metodika a podrobnosti o jednotlivých funkciách v porovnaní s nástrojmi 9router, OpenRouter, CLIProxyAPI a LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -604,7 +604,7 @@ Radar je voliteľný a používa iba požiadavky GET. Klient OmniRoute neodosiel - **🖼️ Nové endpointy** — `/v1/ocr` (Mistral OCR) a `/v1/audio/translations` (v štýle Whisper) dopĺňajú mediálne rozhranie. → [Referencia API](docs/reference/API_REFERENCE.md) - **🎨 Generovanie obrázkov / videí / zvuku** — jedno API pre médiá: xAI Grok Imagine a video Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind a poskytovatelia reči, ako napríklad ElevenLabs. → [Referencia API](docs/reference/API_REFERENCE.md) - **🌍 Nasadenie a prevádzka** — `basePath` reverzného proxy, automatická detekcia jazyka prehliadača, sledovanie zariadení podľa kľúča, dôvera MITM bez oprávnení root, lokalizácia zh-TW. → [Prostredie](docs/reference/ENVIRONMENT.md) -- **🤝 Viac poskytovateľov a agentov** — cloudoví agenti (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) s prihlásením cez prehliadač + OAuth, plnohodnotná karta Ollama, Claude Opus 5 a Sonnet 5, oficiálne partnerstvo Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… a obnovený **katalóg 383 poskytovateľov**. → [Poskytovatelia](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Viac poskytovateľov a agentov** — cloudoví agenti (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) s prihlásením cez prehliadač + OAuth, plnohodnotná karta Ollama, Claude Opus 5 a Sonnet 5, oficiálne partnerstvo Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… a obnovený **katalóg 384 poskytovateľov**. → [Poskytovatelia](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparentnosť smerovania** — každá odpoveď obsahuje hlavičku `X-OmniRoute-Decision` s názvom stratégie/poskytovateľa a latenciou, ktoré ju obslúžili; nová kombinovaná stratégia `cache-optimized` + faktor Auto-Combo `cacheAffinity` smerujú opakované požiadavky späť na pripojenie, ktoré obsahuje uložený prefix, a koncový bod iba na čítanie `/v1/auto-combo/{channel}/candidates` sprístupňuje aktuálny fond kandidátov kanála `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Lokálny výkon a infraštruktúra** — lokálny Redis na jedno kliknutie, deployery relayov pre Cloudflare Workers / Deno Deploy, Bifrost a Mux ako spravované vstavané služby. → [Vstavané služby](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Aj toto je súčasťou balenia** — framework doplnkov a marketplace, frameworky zručností Omni/Agent/GitHub, integrácia trezoru Obsidian (22 nástrojov MCP), API Batch a Files kompatibilné s OpenAI, sémantická cache odpovedí, gamifikácia s rebríčkami, zisťovanie agentov ACP (15 vstavaných agentov), plánovaný export logov do BigQuery, paralelné rozosielanie medzi viacerými modelmi `auto/chaos`, mostík pre bota Telegram, správca verzií v aplikácii a rebríčky bezplatných poskytovateľov LMArena-ELO. → [Dokumentácia](docs/README.md) @@ -688,11 +688,11 @@ histórie vášho shellu. → [CLI Integrácie](docs/guides/CLI-INTEGRATIONS.md)
-## 🌐 383 poskytovateľov AI — 153 označených katalógom ako bezplatní +## 🌐 384 poskytovateľov AI — 153 označených katalógom ako bezplatní
-> **383 registrovaných poskytovateľov** v kanonických kolekciách chatu, médií, vyhľadávania, lokálnych, cloud-agentových a systémových kolekciách vrátane **153 s metadátami na objavovanie `hasFree: true`**. Register chatových modelov zahŕňa **239 poskytovateľov / 3 059 jedinečných dvojíc poskytovateľ–model / 1 475 nespracovaných ID modelov**; samostatný katalóg bezplatného rozpočtu obsahuje **471 riadkov pre jednotlivé modely**, **34 opakujúcich sa fondov** a **52 poskytovateľov s opakujúcim sa alebo bezkľúčovým bezplatným využívaním navždy**. Ide zámerne o rôzne menovatele; definície a výpočty deduplikované podľa fondov nájdete v [Referencii poskytovateľov](docs/reference/PROVIDER_REFERENCE.md) a v [Bezplatných úrovniach](docs/reference/FREE_TIERS.md). +> **384 registrovaných poskytovateľov** v kanonických kolekciách chatu, médií, vyhľadávania, lokálnych, cloud-agentových a systémových kolekciách vrátane **153 s metadátami na objavovanie `hasFree: true`**. Register chatových modelov zahŕňa **239 poskytovateľov / 3 059 jedinečných dvojíc poskytovateľ–model / 1 475 nespracovaných ID modelov**; samostatný katalóg bezplatného rozpočtu obsahuje **471 riadkov pre jednotlivé modely**, **34 opakujúcich sa fondov** a **52 poskytovateľov s opakujúcim sa alebo bezkľúčovým bezplatným využívaním navždy**. Ide zámerne o rôzne menovatele; definície a výpočty deduplikované podľa fondov nájdete v [Referencii poskytovateľov](docs/reference/PROVIDER_REFERENCE.md) a v [Bezplatných úrovniach](docs/reference/FREE_TIERS.md).
@@ -1293,7 +1293,7 @@ Kanonické metriky k 2026-10-18: **1.976 jedinečných videí** · **20.220.289 RuntimeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 JazykTypeScript 6.0 — 100 % TypeScript v src/ a open-sse/ (v jadre od v2.0 žiadne any) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Databázabetter-sqlite3 (SQLite, žurnálovanie WAL) + LowDB (starší JSON) — 147 doménových modulov, 230 migrácií + Databázabetter-sqlite3 (SQLite, žurnálovanie WAL) + LowDB (starší JSON) — 147 doménových modulov, 232 migrácií Pamäťfulltextové vyhľadávanie SQLite FTS5 + vektorové embeddingy kvantizované na int8, typovaný úpadok SchémyZod 4 — validácia I/O nástrojov MCP + kontrakty API ProtokolyMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/sk/llm.txt b/docs/i18n/sk/llm.txt index df4ded80478..5c74edd5362 100644 --- a/docs/i18n/sk/llm.txt +++ b/docs/i18n/sk/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/sl/README.md b/docs/i18n/sl/README.md index 19fbfb1a3dc..3e3c315da9b 100644 --- a/docs/i18n/sl/README.md +++ b/docs/i18n/sl/README.md @@ -15,7 +15,7 @@ **Ustvaril [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Nikoli ne prenehajte programirati. Vsako orodje UI → 383 ponudnikov — 150+ brezplačnih — prek ene končne točke. Claude Code, Codex, Cursor, Cline, Copilot in Antigravity v BREZPLAČNE Claude / GPT / Gemini z avtomatskim preklopom. Zloženo stiskanje RTK + Caveman prihrani 15–95 % žetonov (povprečno ~89 %) — nikoli ne dosezite omejitev. 383 ponudnikov UI · 150+ brezplačnih paketov · ~1,62 milijarde brezplačnih žetonov/mesec · 20 strategij usmerjanja · začnite za 0 $. +OmniRoute — Nikoli ne prenehajte programirati. Vsako orodje UI → 384 ponudnikov — 150+ brezplačnih — prek ene končne točke. Claude Code, Codex, Cursor, Cline, Copilot in Antigravity v BREZPLAČNE Claude / GPT / Gemini z avtomatskim preklopom. Zloženo stiskanje RTK + Caveman prihrani 15–95 % žetonov (povprečno ~89 %) — nikoli ne dosezite omejitev. 384 ponudnikov UI · 150+ brezplačnih paketov · ~1,62 milijarde brezplačnih žetonov/mesec · 20 strategij usmerjanja · začnite za 0 $.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ----------------------------------- | :---------------------: | :---------: | :-------------------: | -| 🌐 Ponudniki | 352 | **358** | 383 | +| 🌐 Ponudniki | 352 | **358** | 384 | | 🧠 Enolični ID-ji klepetnih modelov | 1320 | **1374** | 1491 | | 🖼️ Most modalnosti | 🆕 vid + zvok + video | ✓ | ✓ | | 📡 Brezplačni katalog Radar | 🆕 prostovoljna prijava | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funkcije 🎯 Kombinacije - 🌐 Ponudniki + 🌐 Ponudniki 🔌 CLI in MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Obljuba — ena končna točka in 383 ponudnikov. Samodejni nadomestni preklop ohranja usmerjanje, dokler je na voljo drug zdrav cilj. Šest stebrov: odporen nadomestni preklop med 383 ponudniki · do 95 % prihranka žetonov pri ustreznih delovnih obremenitvah · $0 za začetek s 150+ brezplačnimi paketi in 52 brezplačnimi ponudniki za vedno, ki se redno obnavljajo ali ne zahtevajo ključa · 39 integracij CLI/agentov prek ene konfiguracije · združljivost z API-ji OpenAI, Claude, Gemini in Responses na /v1 · produkcijski nadzor, vključno z odklopniki, prikritostjo TLS, 110 orodji MCP, A2A, pomnilnikom, zaščitnimi mehanizmi, evalvacijami ter več kot 56.000 statičnimi deklaracijami testov v več kot 7.800 sledenih testnih datotekah. +Obljuba — ena končna točka in 384 ponudnikov. Samodejni nadomestni preklop ohranja usmerjanje, dokler je na voljo drug zdrav cilj. Šest stebrov: odporen nadomestni preklop med 384 ponudniki · do 95 % prihranka žetonov pri ustreznih delovnih obremenitvah · $0 za začetek s 150+ brezplačnimi paketi in 52 brezplačnimi ponudniki za vedno, ki se redno obnavljajo ali ne zahtevajo ključa · 39 integracij CLI/agentov prek ene konfiguracije · združljivost z API-ji OpenAI, Claude, Gemini in Responses na /v1 · produkcijski nadzor, vključno z odklopniki, prikritostjo TLS, 110 orodji MCP, A2A, pomnilnikom, zaščitnimi mehanizmi, evalvacijami ter več kot 56.000 statičnimi deklaracijami testov v več kot 7.800 sledenih testnih datotekah.

@@ -510,7 +510,7 @@ Vseh **20** strategij — poljubno jih kombinirajte pri vsakem koraku kombinacij -Po čem se OmniRoute razlikuje — posnetek funkcij na določen datum v primerjavi z rešitvami 9router, OpenRouter, CLIProxyAPI in LiteLLM na področju 13 zmogljivosti. OmniRoute: 383 ponudnikov, več kot 150 vgrajenih brezplačnih paketov, 20 strategij usmerjanja, stiskanje žetonov z 12 mehanizmi, vgrajeni strežnik MCP s 110 orodji, protokol agentov A2A, trajni pomnilnik, varovalni mehanizmi, agenti v oblaku, prikrivanje prstnega odtisa TLS, Desktop/Termux/PWA in 67 jezikovnih nastavitev uporabniškega vmesnika i18n. OmniRoute je licenciran pod MIT in ga lahko gostite sami. Zmogljivosti in število funkcij konkurenčnih rešitev se lahko spremenijo; oglejte si povezano metodologijo. +Po čem se OmniRoute razlikuje — posnetek funkcij na določen datum v primerjavi z rešitvami 9router, OpenRouter, CLIProxyAPI in LiteLLM na področju 13 zmogljivosti. OmniRoute: 384 ponudnikov, več kot 150 vgrajenih brezplačnih paketov, 20 strategij usmerjanja, stiskanje žetonov z 12 mehanizmi, vgrajeni strežnik MCP s 110 orodji, protokol agentov A2A, trajni pomnilnik, varovalni mehanizmi, agenti v oblaku, prikrivanje prstnega odtisa TLS, Desktop/Termux/PWA in 67 jezikovnih nastavitev uporabniškega vmesnika i18n. OmniRoute je licenciran pod MIT in ga lahko gostite sami. Zmogljivosti in število funkcij konkurenčnih rešitev se lahko spremenijo; oglejte si povezano metodologijo. 📊 Celotna metodologija in podrobnosti po posameznih funkcijah v primerjavi z rešitvami 9router, OpenRouter, CLIProxyAPI in LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -613,7 +613,7 @@ trenutnem katalogu najdete na **[radar.omniroute.online/planos](https://radar.om - **🖼️ Nove končne točke** — `/v1/ocr` (Mistral OCR) in `/v1/audio/translations` (v slogu Whisper) dopolnjujeta medijski vmesnik. → [Referenca API-ja](docs/reference/API_REFERENCE.md) - **🎨 Ustvarjanje slik / videoposnetkov / zvoka** — en API za medije: xAI Grok Imagine in videoposnetki Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind ter ponudniki govora, kot je ElevenLabs. → [Referenca API-ja](docs/reference/API_REFERENCE.md) - **🌍 Namestitev in upravljanje** — `basePath` povratnega posredniškega strežnika, samodejno zaznavanje jezika brskalnika, sledenje napravam po ključih, zaupanje MITM brez korenskih pravic in lokalizacija zh-TW. → [Okolje](docs/reference/ENVIRONMENT.md) -- **🤝 Več ponudnikov in agentov** — agenti v oblaku (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) z brskalnikom in prijavo OAuth, prvovrstna kartica Ollama, Claude Opus 5 in Sonnet 5, uradno partnerstvo s Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI … in osvežen **katalog 383 ponudnikov**. → [Ponudniki](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Več ponudnikov in agentov** — agenti v oblaku (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) z brskalnikom in prijavo OAuth, prvovrstna kartica Ollama, Claude Opus 5 in Sonnet 5, uradno partnerstvo s Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI … in osvežen **katalog 384 ponudnikov**. → [Ponudniki](docs/reference/PROVIDER_REFERENCE.md) - **📡 Preglednost usmerjanja** — vsak odgovor vsebuje glavo `X-OmniRoute-Decision` z imenom strategije/ponudnika/zakasnitve, ki je odgovor zagotovil; nova strategija kombinacij `cache-optimized` in faktor usmerjanja `cacheAffinity` v funkciji Auto-Combo ponavljajoče se zahteve preusmerita nazaj na povezavo, ki vsebuje predpomnjeno predpono; končna točka samo za branje `/v1/auto-combo/{channel}/candidates` pa razkrije trenutni nabor kandidatov kanala `auto/*`. → [Samodejne kombinacije](docs/routing/AUTO-COMBO.md). - **⚡ Lokalna zmogljivost in infrastruktura** — lokalni Redis z enim klikom, posredniški uvajalniki za Cloudflare Workers / Deno Deploy, Bifrost in Mux kot nadzorovani vdelani storitvi. → [Vdelane storitve](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Tudi v paketu** — ogrodje vtičnikov in tržnica, ogrodja veščin Omni/Agent/GitHub, integracija z zbirko Obsidian (22 orodij MCP), API-ja Batch in Files, združljiva z OpenAI, semantični predpomnilnik odgovorov, igrifikacija z lestvicami, odkrivanje agentov ACP (15 vgrajenih agentov), načrtovan izvoz dnevnikov v BigQuery, vzporedno razpošiljanje med več modeli `auto/chaos`, most za bota Telegram, upravljalnik različic v aplikaciji in lestvice brezplačnih ponudnikov LMArena-ELO. → [Dokumentacija](docs/README.md) @@ -697,11 +697,11 @@ izven zgodovine vaše lupine. → [CLI integracije](docs/guides/CLI-INTEGRATIONS
-## 🌐 383 ponudnikov umetne inteligence — 153 označenih kot brezplačni v katalogu +## 🌐 384 ponudnikov umetne inteligence — 153 označenih kot brezplačni v katalogu
-> **383 registriranih ponudnikov** v kanoničnih zbirkah za klepet, medije, iskanje, lokalne ponudnike, agente v oblaku in sistemske ponudnike, vključno s **153 ponudniki z metapodatki za odkrivanje `hasFree: true`**. Register modelov za klepet zajema **239 ponudnikov / 3.059 različnih parov ponudnik–model / 1.475 neobdelanih ID-jev modelov**; ločeni katalog brezplačnih proračunov ima **471 vrstic po modelih**, **34 ponavljajočih se skladov** in **52 ponavljajočih se oziroma ponudnikov brez ključev, ki so brezplačni za vedno**. To so namerno različni imenovalci; definicije in izračuni z odstranitvijo podvojenih skladov so v dokumentih [Referenca ponudnikov](docs/reference/PROVIDER_REFERENCE.md) in [Brezplačne ravni](docs/reference/FREE_TIERS.md). +> **384 registriranih ponudnikov** v kanoničnih zbirkah za klepet, medije, iskanje, lokalne ponudnike, agente v oblaku in sistemske ponudnike, vključno s **153 ponudniki z metapodatki za odkrivanje `hasFree: true`**. Register modelov za klepet zajema **239 ponudnikov / 3.059 različnih parov ponudnik–model / 1.475 neobdelanih ID-jev modelov**; ločeni katalog brezplačnih proračunov ima **471 vrstic po modelih**, **34 ponavljajočih se skladov** in **52 ponavljajočih se oziroma ponudnikov brez ključev, ki so brezplačni za vedno**. To so namerno različni imenovalci; definicije in izračuni z odstranitvijo podvojenih skladov so v dokumentih [Referenca ponudnikov](docs/reference/PROVIDER_REFERENCE.md) in [Brezplačne ravni](docs/reference/FREE_TIERS.md).
@@ -1304,7 +1304,7 @@ Kanonične metrike na dan 2026-10-18: **1.976 edinstvenih videoposnetkov** · ** Izvajalno okoljeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 JezikTypeScript 6.0 — 100 % TypeScript v src/ in open-sse/ (v jedru od v2.0 ni nobenega any) OgrodjeNext.js 16 + React 19 + Tailwind CSS 4 - Zbirka podatkovbetter-sqlite3 (SQLite, beleženje WAL) + LowDB (stari JSON) — 147 domenskih modulov, 230 migracij + Zbirka podatkovbetter-sqlite3 (SQLite, beleženje WAL) + LowDB (stari JSON) — 147 domenskih modulov, 232 migracij PomnilnikIskanje po celotnem besedilu SQLite FTS5 + vektorske vdelave, kvantizirane na int8, s tipiziranim upadanjem ShemeZod 4 — preverjanje V/I orodij MCP + pogodbe API ProtokoliMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/sl/llm.txt b/docs/i18n/sl/llm.txt index 2d2d246d77c..32ac892cca3 100644 --- a/docs/i18n/sl/llm.txt +++ b/docs/i18n/sl/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/sr/README.md b/docs/i18n/sr/README.md index 123f4c0084d..1e418ae04c3 100644 --- a/docs/i18n/sr/README.md +++ b/docs/i18n/sr/README.md @@ -15,7 +15,7 @@ **Autor: [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Nikada ne prestajte da programirate. Svaki AI alat → 383 provajder — 150+ besplatnih — kroz jednu krajnju tačku. Claude Code, Codex, Cursor, Cline, Copilot i Antigravity ka BESPLATNOM Claude / GPT / Gemini uz automatski rezervni izbor. RTK + Caveman složena kompresija štedi 15–95% tokena (prosečno ~89%) — nikada ne dostižite ograničenja. 383 AI provajder · 150+ besplatnih nivoa · ~1,62 milijarde besplatnih tokena mesečno · 20 strategija rutiranja · $0 za početak. +OmniRoute — Nikada ne prestajte da programirate. Svaki AI alat → 384 provajder — 150+ besplatnih — kroz jednu krajnju tačku. Claude Code, Codex, Cursor, Cline, Copilot i Antigravity ka BESPLATNOM Claude / GPT / Gemini uz automatski rezervni izbor. RTK + Caveman složena kompresija štedi 15–95% tokena (prosečno ~89%) — nikada ne dostižite ograničenja. 384 AI provajder · 150+ besplatnih nivoa · ~1,62 milijarde besplatnih tokena mesečno · 20 strategija rutiranja · $0 za početak.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ---------------------------------- | :-----------------------: | :---------: | :-------------------: | -| 🌐 Provajderi | 352 | **358** | 383 | +| 🌐 Provajderi | 352 | **358** | 384 | | 🧠 Jedinstveni ID-jevi chat modela | 1320 | **1374** | 1491 | | 🖼️ Most za modalitete | 🆕 vid + audio + video | ✓ | ✓ | | 📡 Besplatni katalog Radara | 🆕 uključivanje po izboru | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Функције 🎯 Комбинације - 🌐 Добављачи + 🌐 Добављачи 🔌 CLI и MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Обећање — Једна крајња тачка и 383 добављач. Аутоматски резервни прелаз одржава усмеравање све док је доступно друго здраво одредиште. Шест стубова: поуздан резервни прелаз кроз 383 добављача · до 95% уштеде токена на одговарајућим радним оптерећењима · $0 за почетак уз више од 150 бесплатних тарифа и 52 бесплатна добављача са понављањем/без кључа заувек · 39 CLI/agent интеграција кроз једну конфигурацију · компатибилност са OpenAI, Claude, Gemini и Responses API на /v1 · продукционе контроле, укључујући прекидаче кола, TLS прикривање, MCP са 110 алата, A2A, меморију, заштитне механизме, евалуације и више од 56.000 статичких декларација тестова у више од 7.800 праћених датотека са тестовима. +Обећање — Једна крајња тачка и 384 добављач. Аутоматски резервни прелаз одржава усмеравање све док је доступно друго здраво одредиште. Шест стубова: поуздан резервни прелаз кроз 384 добављача · до 95% уштеде токена на одговарајућим радним оптерећењима · $0 за почетак уз више од 150 бесплатних тарифа и 52 бесплатна добављача са понављањем/без кључа заувек · 39 CLI/agent интеграција кроз једну конфигурацију · компатибилност са OpenAI, Claude, Gemini и Responses API на /v1 · продукционе контроле, укључујући прекидаче кола, TLS прикривање, MCP са 110 алата, A2A, меморију, заштитне механизме, евалуације и више од 56.000 статичких декларација тестова у више од 7.800 праћених датотека са тестовима.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -По чему се OmniRoute издваја — ажурирани преглед функција у поређењу са 9router, OpenRouter, CLIProxyAPI и LiteLLM кроз 13 могућности. OmniRoute: 383 добављач, више од 150 уграђених бесплатних нивоа, 20 стратегија рутирања, компресија токена помоћу 12 механизама, уграђени MCP сервер са 110 алатки, A2A протокол за агенте, трајна меморија, заштитне мере, агенти у облаку, прикривеност TLS отиска, Desktop/Termux/PWA и 67 локализација корисничког интерфејса. OmniRoute је лиценциран под MIT лиценцом и може се хостовати на сопственој инфраструктури. Могућности и бројеви конкурената могу се променити; погледајте методологију на наведеној вези. +По чему се OmniRoute издваја — ажурирани преглед функција у поређењу са 9router, OpenRouter, CLIProxyAPI и LiteLLM кроз 13 могућности. OmniRoute: 384 добављач, више од 150 уграђених бесплатних нивоа, 20 стратегија рутирања, компресија токена помоћу 12 механизама, уграђени MCP сервер са 110 алатки, A2A протокол за агенте, трајна меморија, заштитне мере, агенти у облаку, прикривеност TLS отиска, Desktop/Termux/PWA и 67 локализација корисничког интерфејса. OmniRoute је лиценциран под MIT лиценцом и може се хостовати на сопственој инфраструктури. Могућности и бројеви конкурената могу се променити; погледајте методологију на наведеној вези. 📊 Комплетна методологија и детаљи по функцијама у поређењу са 9router, OpenRouter, CLIProxyAPI и LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -613,7 +613,7 @@ Radar се користи по избору и подржава само GET. Om - **🖼️ Нове крајње тачке** — `/v1/ocr` (Mistral OCR) и `/v1/audio/translations` (у стилу Whisper-а) употпуњују медијску површину. → [API референца](docs/reference/API_REFERENCE.md) - **🎨 Генерисање слика / видео-записа / звука** — један API за медије: xAI Grok Imagine и Novita AI видео, ComfyUI, Magnific, Adobe Firefly, Segmind и добављачи говора као што је ElevenLabs. → [API референца](docs/reference/API_REFERENCE.md) - **🌍 Примена и оперативно управљање** — `basePath` обрнутог проксија, аутоматско откривање језика прегледача, праћење уређаја по кључу, MITM поверење без root привилегија, локализација zh-TW. → [Окружење](docs/reference/ENVIRONMENT.md) -- **🤝 Више добављача и агената** — агенти у облаку (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) са пријављивањем преко прегледача и OAuth-а, првокласна Ollama картица, Claude Opus 5 и Sonnet 5, званично партнерство са Kimi-јем (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… и освежен **каталог са 383 добављачем**. → [Добављачи](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Више добављача и агената** — агенти у облаку (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) са пријављивањем преко прегледача и OAuth-а, првокласна Ollama картица, Claude Opus 5 и Sonnet 5, званично партнерство са Kimi-јем (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… и освежен **каталог са 384 добављачем**. → [Добављачи](docs/reference/PROVIDER_REFERENCE.md) - **📡 Транспарентност рутирања** — сваки одговор садржи заглавље `X-OmniRoute-Decision` које наводи стратегију/добављача/кашњење који су га послужили, нова стратегија комбинације `cache-optimized` + фактор `cacheAffinity` аутоматских комбинација враћа поновљене захтеве вези која садржи кеширани префикс, а крајња тачка само за читање `/v1/auto-combo/{channel}/candidates` приказује живи скуп кандидата канала `auto/*`. → [Аутоматске комбинације](docs/routing/AUTO-COMBO.md)。 - **⚡ Локалне перформансе и инфраструктура** — локални Redis једним кликом, деплојери за Cloudflare Workers / Deno Deploy relay, Bifrost и Mux као надгледане уграђене услуге. → [Уграђене услуге](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Такође у пакету** — оквир за додатке + тржиште, Omni/Agent/GitHub оквири за вештине, интеграција са Obsidian трезором (22 MCP алатке), Batch и Files API-ји компатибилни са OpenAI-јем, семантички кеш одговора, гејмификација са ранг-листама, ACP откривање агената (15 уграђених агената), заказани извоз евиденција у BigQuery, `auto/chaos` паралелно прослеђивање ка више модела, мост за Telegram бота, менаџер верзија у апликацији и LMArena-ELO ранг-листе бесплатних провајдера. → [Документација](docs/README.md) @@ -697,11 +697,11 @@ omniroute configure codex # такође: claude opencode qwen aider go
-## 🌐 383 AI провајдер — 153 означена у каталогу као бесплатна +## 🌐 384 AI провајдер — 153 означена у каталогу као бесплатна
-> **383 регистрован провајдер** у канонским колекцијама за ћаскање, медије, претрагу, локалне системе, cloud-агенте и системске колекције, укључујући **153 провајдера са метаподацима за откривање `hasFree: true`**. Регистар модела за ћаскање обухвата **239 провајдера / 3,059 различитих парова провајдер-модел / 1,475 изворних ID-јева модела**; засебни каталог бесплатних буџета има **471 ред по моделу**, **34 периодична фонда** и **52 провајдера са бесплатним приступом заувек, без кључа и са периодичним фондом**. Ово су намерно различити имениоци; дефиниције и прорачуни са уклоњеним дупликатима фондова налазе се у документима [Референца за провајдере](docs/reference/PROVIDER_REFERENCE.md) и [Бесплатни нивои](docs/reference/FREE_TIERS.md). +> **384 регистрован провајдер** у канонским колекцијама за ћаскање, медије, претрагу, локалне системе, cloud-агенте и системске колекције, укључујући **153 провајдера са метаподацима за откривање `hasFree: true`**. Регистар модела за ћаскање обухвата **239 провајдера / 3,059 различитих парова провајдер-модел / 1,475 изворних ID-јева модела**; засебни каталог бесплатних буџета има **471 ред по моделу**, **34 периодична фонда** и **52 провајдера са бесплатним приступом заувек, без кључа и са периодичним фондом**. Ово су намерно различити имениоци; дефиниције и прорачуни са уклоњеним дупликатима фондова налазе се у документима [Референца за провајдере](docs/reference/PROVIDER_REFERENCE.md) и [Бесплатни нивои](docs/reference/FREE_TIERS.md).
@@ -1305,7 +1305,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 такође га Окружење за извршавањеNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ЈезикTypeScript 6.0 — 100% TypeScript у директоријумима src/ и open-sse/ (без any у језгру од v2.0) ОквирNext.js 16 + React 19 + Tailwind CSS 4 - База податакаbetter-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — 147 доменских модула, 230 миграција + База податакаbetter-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — 147 доменских модула, 232 миграција МеморијаSQLite FTS5 претрага целог текста + векторски embedding-и квантизовани на int8, типизовано опадање ШемеZod 4 — валидација I/O MCP алата + API уговори ПротоколиMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/sr/llm.txt b/docs/i18n/sr/llm.txt index c99a794d38a..0b9bf12b483 100644 --- a/docs/i18n/sr/llm.txt +++ b/docs/i18n/sr/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/sv/README.md b/docs/i18n/sv/README.md index f532e358858..ab16cf9815e 100644 --- a/docs/i18n/sv/README.md +++ b/docs/i18n/sv/README.md @@ -15,7 +15,7 @@ **Skapad av [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Sluta aldrig koda. Alla AI-verktyg → 383 leverantörer — 150+ kostnadsfria — via en enda endpoint. Claude Code, Codex, Cursor, Cline, Copilot och Antigravity till kostnadsfria Claude / GPT / Gemini med automatisk återställning. RTK + Caveman-staplad komprimering sparar 15–95 % tokens (~89 % i genomsnitt) — nå aldrig gränserna. 383 AI-leverantörer · 150+ kostnadsfria nivåer · ~1,62 miljarder kostnadsfria tokens/månad · 20 routningsstrategier · $0 för att komma igång. +OmniRoute — Sluta aldrig koda. Alla AI-verktyg → 384 leverantörer — 150+ kostnadsfria — via en enda endpoint. Claude Code, Codex, Cursor, Cline, Copilot och Antigravity till kostnadsfria Claude / GPT / Gemini med automatisk återställning. RTK + Caveman-staplad komprimering sparar 15–95 % tokens (~89 % i genomsnitt) — nå aldrig gränserna. 384 AI-leverantörer · 150+ kostnadsfria nivåer · ~1,62 miljarder kostnadsfria tokens/månad · 20 routningsstrategier · $0 för att komma igång.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------ | :----------------------: | :---------: | :-------------------: | -| 🌐 Leverantörer | 352 | **358** | 383 | +| 🌐 Leverantörer | 352 | **358** | 384 | | 🧠 Unika chattmodell-ID:n | 1320 | **1374** | 1491 | | 🖼️ Modalitetsbrygga | 🆕 vision + ljud + video | ✓ | ✓ | | 📡 Radars gratiskatalog | 🆕 valfritt deltagande | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Funktioner 🎯 Kombinationer - 🌐 Leverantörer + 🌐 Leverantörer 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Löftet — En slutpunkt och 383 leverantörer. Automatisk reservväxling fortsätter routningen så länge en annan fungerande målanslutning är tillgänglig. Sex pelare: motståndskraftig reservväxling över 383 leverantörer · upp till 95 % tokenbesparingar för kvalificerade arbetsbelastningar · $0 för att komma igång med över 150 kostnadsfria nivåer och 52 återkommande eller nyckellösa leverantörer som alltid är kostnadsfria · 39 CLI-/agentintegrationer via en enda konfiguration · kompatibilitet med OpenAI-, Claude-, Gemini- och Responses API på /v1 · produktionskontroller, inklusive kretsbrytare, TLS-stealth, MCP med 110 verktyg, A2A, minne, skyddsräcken, utvärderingar och över 56 000 statiska testdeklarationer i över 7 800 spårade testfiler. +Löftet — En slutpunkt och 384 leverantörer. Automatisk reservväxling fortsätter routningen så länge en annan fungerande målanslutning är tillgänglig. Sex pelare: motståndskraftig reservväxling över 384 leverantörer · upp till 95 % tokenbesparingar för kvalificerade arbetsbelastningar · $0 för att komma igång med över 150 kostnadsfria nivåer och 52 återkommande eller nyckellösa leverantörer som alltid är kostnadsfria · 39 CLI-/agentintegrationer via en enda konfiguration · kompatibilitet med OpenAI-, Claude-, Gemini- och Responses API på /v1 · produktionskontroller, inklusive kretsbrytare, TLS-stealth, MCP med 110 verktyg, A2A, minne, skyddsräcken, utvärderingar och över 56 000 statiska testdeklarationer i över 7 800 spårade testfiler.

@@ -510,7 +510,7 @@ Alla **20** strategier — blanda fritt per steg i kombinationen: -Vad som skiljer OmniRoute från mängden — en daterad funktionsöversikt jämfört med 9router, OpenRouter, CLIProxyAPI och LiteLLM inom 13 funktioner. OmniRoute: 383 leverantörer, över 150 inbyggda kostnadsfria nivåer, 20 routningsstrategier, tokenkomprimering med 12 motorer, inbyggd MCP-server med 110 verktyg, A2A-agentprotokoll, beständigt minne, skyddsräcken, molnagenter, stealth med TLS-fingeravtryck, Desktop/Termux/PWA och 67 i18n-gränssnittslokaler. OmniRoute är MIT-licensierat och kan egenhostas. Konkurrenternas funktioner och antal kan ändras; se den länkade metodbeskrivningen. +Vad som skiljer OmniRoute från mängden — en daterad funktionsöversikt jämfört med 9router, OpenRouter, CLIProxyAPI och LiteLLM inom 13 funktioner. OmniRoute: 384 leverantörer, över 150 inbyggda kostnadsfria nivåer, 20 routningsstrategier, tokenkomprimering med 12 motorer, inbyggd MCP-server med 110 verktyg, A2A-agentprotokoll, beständigt minne, skyddsräcken, molnagenter, stealth med TLS-fingeravtryck, Desktop/Termux/PWA och 67 i18n-gränssnittslokaler. OmniRoute är MIT-licensierat och kan egenhostas. Konkurrenternas funktioner och antal kan ändras; se den länkade metodbeskrivningen. 📊 Fullständig metodbeskrivning och detaljinformation per funktion jämfört med 9router, OpenRouter, CLIProxyAPI och LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -613,7 +613,7 @@ den aktuella katalogen på **[radar.omniroute.online/planos](https://radar.omnir - **🖼️ Nya endpoints** — `/v1/ocr` (Mistral OCR) och `/v1/audio/translations` (Whisper-liknande) kompletterar medieytan. → [API-referens](docs/reference/API_REFERENCE.md) - **🎨 Bild-, video- och ljudgenerering** — ett API för medier: xAI Grok Imagine och Novita AI-video, ComfyUI, Magnific, Adobe Firefly, Segmind samt talleverantörer som ElevenLabs. → [API-referens](docs/reference/API_REFERENCE.md) - **🌍 Driftsättning och drift** — reverse-proxy-`basePath`, automatisk identifiering av webbläsarspråk, enhetsspårning per nyckel, root-lös MITM-trust och zh-TW-lokalisering. → [Miljö](docs/reference/ENVIRONMENT.md) -- **🤝 Fler leverantörer och agenter** — molnagenter (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) med webbläsar- och OAuth-inloggning, förstaklasskort för Ollama, Claude Opus 5 och Sonnet 5, officiellt Kimi-samarbete (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI … samt en uppdaterad **383-leverantörskatalog**. → [Leverantörer](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Fler leverantörer och agenter** — molnagenter (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) med webbläsar- och OAuth-inloggning, förstaklasskort för Ollama, Claude Opus 5 och Sonnet 5, officiellt Kimi-samarbete (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI … samt en uppdaterad **384-leverantörskatalog**. → [Leverantörer](docs/reference/PROVIDER_REFERENCE.md) - **📡 Transparens i routningen** — varje svar innehåller en `X-OmniRoute-Decision`-header som anger strategin, leverantören och latensen som hanterade det, en ny `cache-optimized`-kombostrategi samt faktorn `cacheAffinity` i Auto-Combo som dirigerar upprepade förfrågningar tillbaka till anslutningen som innehåller det cachade prefixet, och en skrivskyddad endpoint på `/v1/auto-combo/{channel}/candidates` visar den aktuella kandidatpoolen för en `auto/*`-kanal. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Lokal prestanda och infrastruktur** — Redis lokalt med ett klick, relädistributörer för Cloudflare Workers / Deno Deploy, Bifrost och Mux som övervakade inbäddade tjänster. → [Inbäddade tjänster](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Ingår också** — pluginramverk och marknadsplats, ramverk för Omni-/Agent-/GitHub-färdigheter, Obsidian-valvintegrering (22 MCP-verktyg), OpenAI-kompatibla Batch- och Files-API:er, semantisk svarscache, spelifiering med topplistor, ACP-agentidentifiering (15 inbyggda agenter), schemalagd loggexport till BigQuery, parallell `auto/chaos`-fan-out med flera modeller, en Telegram-botbrygga, en versionshanterare i appen och LMArena-ELO-rankningar av kostnadsfria leverantörer. → [Dokumentation](docs/README.md) @@ -697,11 +697,11 @@ från din shellhistorik. → [CLI-integrationer](docs/guides/CLI-INTEGRATIONS.md
-## 🌐 383 AI-leverantörer — 153 katalogmarkerade som gratis +## 🌐 384 AI-leverantörer — 153 katalogmarkerade som gratis
-> **383 registrerade leverantörer** i de kanoniska samlingarna för chatt, media, sökning, lokalt, molnagenter och system, inklusive **153 med `hasFree: true`-metadata för upptäckt**. Chattmodellregistret omfattar **239 leverantörer / 3 059 unika leverantör-modell-par / 1 475 råa modell-ID:n**; den separata katalogen för gratisbudgetar har **471 rader per modell**, **34 återkommande pooler** och **52 återkommande eller nyckellösa leverantörer som är gratis för alltid**. Dessa är avsiktligt olika nämnare; definitioner och pooldeduplicerade beräkningar finns i [Leverantörsreferensen](docs/reference/PROVIDER_REFERENCE.md) och [Gratisnivåer](docs/reference/FREE_TIERS.md). +> **384 registrerade leverantörer** i de kanoniska samlingarna för chatt, media, sökning, lokalt, molnagenter och system, inklusive **153 med `hasFree: true`-metadata för upptäckt**. Chattmodellregistret omfattar **239 leverantörer / 3 059 unika leverantör-modell-par / 1 475 råa modell-ID:n**; den separata katalogen för gratisbudgetar har **471 rader per modell**, **34 återkommande pooler** och **52 återkommande eller nyckellösa leverantörer som är gratis för alltid**. Dessa är avsiktligt olika nämnare; definitioner och pooldeduplicerade beräkningar finns i [Leverantörsreferensen](docs/reference/PROVIDER_REFERENCE.md) och [Gratisnivåer](docs/reference/FREE_TIERS.md).
@@ -1303,7 +1303,7 @@ Kanoniska mätvärden den 2026-10-18: **1.976 unika videor** · **20.220.289 kä KörmiljöNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 SpråkTypeScript 6.0 — 100 % TypeScript i src/ och open-sse/ (ingen any i kärnan sedan v2.0) RamverkNext.js 16 + React 19 + Tailwind CSS 4 - Databasbetter-sqlite3 (SQLite, WAL-journalföring) + LowDB (äldre JSON) — 147 domänmoduler, 230 migreringar + Databasbetter-sqlite3 (SQLite, WAL-journalföring) + LowDB (äldre JSON) — 147 domänmoduler, 232 migreringar MinneSQLite FTS5-fulltext + int8-kvantiserade vektorinbäddningar, typad avklingning SchemanZod 4 — validering av MCP-verktygens indata/utdata + API-kontrakt ProtokollMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/sv/llm.txt b/docs/i18n/sv/llm.txt index a76b5422429..af8aae0974d 100644 --- a/docs/i18n/sv/llm.txt +++ b/docs/i18n/sv/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/sw/README.md b/docs/i18n/sw/README.md index 1b43eab718b..41c4aa9ffcc 100644 --- a/docs/i18n/sw/README.md +++ b/docs/i18n/sw/README.md @@ -15,7 +15,7 @@ **Imeundwa na [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Usikome kamwe kuandika msimbo. Kila zana ya AI → watoa huduma 383 — 150+ bila malipo — kupitia kituo kimoja. Claude Code, Codex, Cursor, Cline, Copilot na Antigravity kwenda Claude / GPT / Gemini BILA MALIPO kwa kutumia ubadilishaji wa kiotomatiki. Mfinyazo wa RTK + Caveman uliowekwa kwa pamoja huokoa tokeni 15–95% (~89% kwa wastani) — usiwahi kufikia vikomo. Watoa huduma 383 wa AI · viwango 150+ vya bila malipo · ~tokeni bilioni 1.62 za bila malipo kwa mwezi · mikakati 20 ya uelekezaji · $0 kuanza. +OmniRoute — Usikome kamwe kuandika msimbo. Kila zana ya AI → watoa huduma 384 — 150+ bila malipo — kupitia kituo kimoja. Claude Code, Codex, Cursor, Cline, Copilot na Antigravity kwenda Claude / GPT / Gemini BILA MALIPO kwa kutumia ubadilishaji wa kiotomatiki. Mfinyazo wa RTK + Caveman uliowekwa kwa pamoja huokoa tokeni 15–95% (~89% kwa wastani) — usiwahi kufikia vikomo. Watoa huduma 384 wa AI · viwango 150+ vya bila malipo · ~tokeni bilioni 1.62 za bila malipo kwa mwezi · mikakati 20 ya uelekezaji · $0 kuanza.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ---------------------------------------------------- | :----------------------: | :---------: | :-------------------: | -| 🌐 Watoa huduma | 352 | **358** | 383 | +| 🌐 Watoa huduma | 352 | **358** | 384 | | 🧠 Vitambulisho vya kipekee vya miundo ya mazungumzo | 1320 | **1374** | 1491 | | 🖼️ Daraja la modali | 🆕 kuona + sauti + video | ✓ | ✓ | | 📡 Katalogi ya bure ya Radar | 🆕 kujiunga kwa hiari | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Vipengele 🎯 Michanganyiko - 🌐 Watoa huduma + 🌐 Watoa huduma 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Ahadi — Endpoint moja na watoa huduma 383. Fallback ya kiotomatiki huendeleza uelekezaji mradi walengwa wengine wenye afya wanapatikana. Nguzo sita: fallback thabiti kati ya watoa huduma 383 · hadi 95% ya uokoaji wa tokeni kwenye mizigo ya kazi inayostahiki · $0 kuanza, ikiwa na viwango 150+ vya bure na watoa huduma 52 wa bure milele wenye kujirudia/wasiohitaji ufunguo · miunganisho 39 ya CLI/agent kupitia usanidi mmoja · uoanifu na OpenAI, Claude, Gemini na Responses API kwenye /v1 · vidhibiti vya uzalishaji vinavyojumuisha circuit breakers, TLS stealth, zana 110 za MCP, A2A, memory, guardrails, evals na matamko ya majaribio tuli 56,000+ katika faili za majaribio 7,800+ zinazofuatiliwa. +Ahadi — Endpoint moja na watoa huduma 384. Fallback ya kiotomatiki huendeleza uelekezaji mradi walengwa wengine wenye afya wanapatikana. Nguzo sita: fallback thabiti kati ya watoa huduma 384 · hadi 95% ya uokoaji wa tokeni kwenye mizigo ya kazi inayostahiki · $0 kuanza, ikiwa na viwango 150+ vya bure na watoa huduma 52 wa bure milele wenye kujirudia/wasiohitaji ufunguo · miunganisho 39 ya CLI/agent kupitia usanidi mmoja · uoanifu na OpenAI, Claude, Gemini na Responses API kwenye /v1 · vidhibiti vya uzalishaji vinavyojumuisha circuit breakers, TLS stealth, zana 110 za MCP, A2A, memory, guardrails, evals na matamko ya majaribio tuli 56,000+ katika faili za majaribio 7,800+ zinazofuatiliwa.

@@ -443,7 +443,7 @@ Mikakati yote **20** — changanya na kulinganisha kwa kila hatua ya combo: -Kinachoitofautisha OmniRoute — muhtasari wa vipengele wa tarehe fulani ukilinganisha na 9router, OpenRouter, CLIProxyAPI na LiteLLM katika uwezo 13. OmniRoute: watoa huduma 383, viwango vya bure zaidi ya 150 vilivyojengewa ndani, mikakati 20 ya uelekezaji, mbano wa tokeni wa injini 12, seva ya MCP iliyojengewa ndani yenye zana 110, itifaki ya mawakala ya A2A, kumbukumbu endelevu, vidhibiti, mawakala wa wingu, ujanja wa alama za vidole za TLS, Desktop/Termux/PWA na maeneo 67 ya lugha ya kiolesura cha i18n. OmniRoute imepewa leseni ya MIT na inaweza kupangishwa binafsi. Uwezo na idadi ya washindani inaweza kubadilika; tazama mbinu iliyounganishwa. +Kinachoitofautisha OmniRoute — muhtasari wa vipengele wa tarehe fulani ukilinganisha na 9router, OpenRouter, CLIProxyAPI na LiteLLM katika uwezo 13. OmniRoute: watoa huduma 384, viwango vya bure zaidi ya 150 vilivyojengewa ndani, mikakati 20 ya uelekezaji, mbano wa tokeni wa injini 12, seva ya MCP iliyojengewa ndani yenye zana 110, itifaki ya mawakala ya A2A, kumbukumbu endelevu, vidhibiti, mawakala wa wingu, ujanja wa alama za vidole za TLS, Desktop/Termux/PWA na maeneo 67 ya lugha ya kiolesura cha i18n. OmniRoute imepewa leseni ya MIT na inaweza kupangishwa binafsi. Uwezo na idadi ya washindani inaweza kubadilika; tazama mbinu iliyounganishwa. 📊 Mbinu kamili & maelezo ya kila kipengele dhidi ya 9router, OpenRouter, CLIProxyAPI & LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -538,7 +538,7 @@ Radar ni ya kujiunga kwa hiari na hutumia GET pekee. Kiteja cha OmniRoute hakipa - **🖼️ Endpoint mpya** — `/v1/ocr` (Mistral OCR) na `/v1/audio/translations` (ya mtindo wa Whisper) hukamilisha safu ya midia. → [Marejeo ya API](docs/reference/API_REFERENCE.md) - **🎨 Uzalishaji wa picha / video / sauti** — API moja ya midia: xAI Grok Imagine na video ya Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind, na watoa huduma wa sauti kama vile ElevenLabs. → [Marejeo ya API](docs/reference/API_REFERENCE.md) - **🌍 Usambazaji na uendeshaji** — `basePath` ya proksi kinyume, utambuzi otomatiki wa lugha ya kivinjari, ufuatiliaji wa kifaa kwa kila ufunguo, uaminifu wa MITM bila haki za mzizi, na ujanibishaji wa zh-TW. → [Mazingira](docs/reference/ENVIRONMENT.md) -- **🤝 Watoa huduma na mawakala zaidi** — mawakala wa wingu (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) yenye kivinjari + kuingia kwa OAuth, kadi ya kiwango cha kwanza ya Ollama, Claude Opus 5 na Sonnet 5, ushirikiano rasmi wa Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… na **katalogi ya watoa huduma 383** iliyoboreshwa. → [Watoa Huduma](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Watoa huduma na mawakala zaidi** — mawakala wa wingu (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) yenye kivinjari + kuingia kwa OAuth, kadi ya kiwango cha kwanza ya Ollama, Claude Opus 5 na Sonnet 5, ushirikiano rasmi wa Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… na **katalogi ya watoa huduma 384** iliyoboreshwa. → [Watoa Huduma](docs/reference/PROVIDER_REFERENCE.md) - **📡 Uwazi wa uelekezaji** — kila jibu hubeba kichwa cha `X-OmniRoute-Decision` kinachotaja mkakati/mtoa huduma/ucheleweshaji uliolihudumia, mkakati mpya wa mseto wa `cache-optimized` + kipengele cha `cacheAffinity` cha Auto-Combo huelekeza maombi yanayorudiwa kwenye muunganisho unaoshikilia kiambishi awali kilichowekwa kwenye akiba, na endpoint ya kusoma pekee ya `/v1/auto-combo/{channel}/candidates` hufichua kundi hai la wagombea la kituo cha `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md). - **⚡ Utendaji wa ndani & miundombinu** — Redis ya ndani kwa kubofya mara moja, zana za kupeleka relay za Cloudflare Workers / Deno Deploy, Bifrost & Mux kama huduma zilizopachikwa zinazosimamiwa. → [Huduma Zilizopachikwa](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Pia ndani ya kifurushi** — mfumo wa plugin + soko, mifumo ya ujuzi ya Omni/Agent/GitHub, ujumuishaji wa hazina ya Obsidian (zana 22 za MCP), API za Batch & Files zinazooana na OpenAI, akiba ya majibu ya kisemantiki, uchezeshaji wenye bao za nafasi, ugunduzi wa mawakala wa ACP (mawakala 15 waliojengewa ndani), uhamishaji uliopangwa wa kumbukumbu kwenda BigQuery, usambazaji sambamba wa miundo mingi kupitia `auto/chaos`, daraja la roboti ya Telegram, kidhibiti cha matoleo ndani ya programu na viwango vya watoa huduma bila malipo vya LMArena-ELO. → [Nyaraka](docs/README.md) @@ -622,11 +622,11 @@ ya historia ya shell yako. → [Muunganiko wa CLI](docs/guides/CLI-INTEGRATIONS.
-## 🌐 Watoa huduma wa AI 383 — 153 Waliowekwa Alama kuwa Bure kwenye Katalogi +## 🌐 Watoa huduma wa AI 384 — 153 Waliowekwa Alama kuwa Bure kwenye Katalogi
-> **Watoa huduma 383 waliosajiliwa** katika mikusanyo ya kikanoni ya gumzo, midia, utafutaji, ya ndani, mawakala wa wingu na mfumo, wakiwemo **153 wenye metadata ya ugunduzi ya `hasFree: true`**. Sajili ya miundo ya gumzo inashughulikia **watoa huduma 239 / jozi 3,059 tofauti za mtoa huduma-modeli / vitambulisho 1,475 ghafi vya miundo**; katalogi tofauti ya bajeti bila malipo ina **mistari 471 kwa kila modeli**, **madimbwi 34 yanayojirudia** na **watoa huduma 52 wa bila malipo milele wanaojirudia/wasiohitaji ufunguo**. Hivi ni vigezo tofauti kwa makusudi; fasili na hesabu zilizopunguzwa marudio ya madimbwi ziko katika [Marejeleo ya Watoa Huduma](docs/reference/PROVIDER_REFERENCE.md) na [Viwango vya Bure](docs/reference/FREE_TIERS.md). +> **Watoa huduma 384 waliosajiliwa** katika mikusanyo ya kikanoni ya gumzo, midia, utafutaji, ya ndani, mawakala wa wingu na mfumo, wakiwemo **153 wenye metadata ya ugunduzi ya `hasFree: true`**. Sajili ya miundo ya gumzo inashughulikia **watoa huduma 239 / jozi 3,059 tofauti za mtoa huduma-modeli / vitambulisho 1,475 ghafi vya miundo**; katalogi tofauti ya bajeti bila malipo ina **mistari 471 kwa kila modeli**, **madimbwi 34 yanayojirudia** na **watoa huduma 52 wa bila malipo milele wanaojirudia/wasiohitaji ufunguo**. Hivi ni vigezo tofauti kwa makusudi; fasili na hesabu zilizopunguzwa marudio ya madimbwi ziko katika [Marejeleo ya Watoa Huduma](docs/reference/PROVIDER_REFERENCE.md) na [Viwango vya Bure](docs/reference/FREE_TIERS.md).
@@ -1228,7 +1228,7 @@ Vipimo kanoniki mnamo 2026-10-18: **video 1.976 za kipekee** · **mara 20.220.28 Muda wa uendeshajiNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 LughaTypeScript 6.0 — 100% TypeScript katika src/ na open-sse/ (any sifuri katika msingi tangu v2.0) Mfumo wa uundajiNext.js 16 + React 19 + Tailwind CSS 4 - Hifadhidatabetter-sqlite3 (SQLite, uwekaji kumbukumbu wa WAL) + LowDB (JSON ya zamani) — moduli 147 za kikoa, uhamishaji 230 + Hifadhidatabetter-sqlite3 (SQLite, uwekaji kumbukumbu wa WAL) + LowDB (JSON ya zamani) — moduli 147 za kikoa, uhamishaji 232 KumbukumbuUtafutaji wa maandishi kamili wa SQLite FTS5 + vembeddings za vekta zilizopunguzwa kwa int8, uozo wenye aina MiundoZod 4 — uthibitishaji wa ingizo/tokeo la zana za MCP + mikataba ya API ItifakiMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/sw/llm.txt b/docs/i18n/sw/llm.txt index 20463bfb200..f127e1a54f0 100644 --- a/docs/i18n/sw/llm.txt +++ b/docs/i18n/sw/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ta/README.md b/docs/i18n/ta/README.md index 5af7e0683fd..e6079b6f0b4 100644 --- a/docs/i18n/ta/README.md +++ b/docs/i18n/ta/README.md @@ -15,7 +15,7 @@ **உருவாக்கியவர் [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — குறியீடு எழுதுவதை ஒருபோதும் நிறுத்தாதீர்கள். ஒவ்வொரு AI கருவியும் → 383 வழங்குநர்கள் — 150+ இலவசம் — ஒரே endpoint வழியாக. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity ஆகியவற்றை auto-fallback உடன் இலவச Claude / GPT / Gemini ஆக மாற்றுகிறது. RTK + Caveman அடுக்கப்பட்ட சுருக்கம் 15–95% டோக்கன்களை (~89% சராசரி) சேமிக்கிறது — வரம்புகளை ஒருபோதும் எட்டாதீர்கள். 383 AI வழங்குநர்கள் · 150+ இலவச அடுக்குகள் · மாதத்திற்கு ~1.62B இலவச டோக்கன்கள் · 20 routing உத்திகள் · தொடங்க $0. +OmniRoute — குறியீடு எழுதுவதை ஒருபோதும் நிறுத்தாதீர்கள். ஒவ்வொரு AI கருவியும் → 384 வழங்குநர்கள் — 150+ இலவசம் — ஒரே endpoint வழியாக. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity ஆகியவற்றை auto-fallback உடன் இலவச Claude / GPT / Gemini ஆக மாற்றுகிறது. RTK + Caveman அடுக்கப்பட்ட சுருக்கம் 15–95% டோக்கன்களை (~89% சராசரி) சேமிக்கிறது — வரம்புகளை ஒருபோதும் எட்டாதீர்கள். 384 AI வழங்குநர்கள் · 150+ இலவச அடுக்குகள் · மாதத்திற்கு ~1.62B இலவச டோக்கன்கள் · 20 routing உத்திகள் · தொடங்க $0.
@@ -73,7 +73,7 @@ OMNIROUTE உங்கள் பணத்தைச் சேமிக்கவ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------------- | :----------------------: | :---------: | :--------------------: | -| 🌐 வழங்குநர்கள் | 352 | **358** | 383 | +| 🌐 வழங்குநர்கள் | 352 | **358** | 384 | | 🧠 தனித்துவமான உரையாடல் மாதிரி ID-கள் | 1320 | **1374** | 1491 | | 🖼️ பல்முறைமைப் பாலம் | 🆕 பார்வை + ஒலி + காணொளி | ✓ | ✓ | | 📡 ரேடார் இலவசப் பட்டியல் | 🆕 விருப்பத்தேர்வு | ✓ | ✓ | @@ -114,7 +114,7 @@ OMNIROUTE உங்கள் பணத்தைச் சேமிக்கவ ⚙️ அம்சங்கள் 🎯 சேர்க்கைகள் - 🌐 வழங்குநர்கள் + 🌐 வழங்குநர்கள் 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -வாக்குறுதி — ஒரே endpoint மற்றும் 383 வழங்குநர்கள். மற்றொரு ஆரோக்கியமான இலக்கு கிடைக்கும் வரை தானியங்கி fallback வழிமாற்றத்தைத் தொடரச் செய்கிறது. ஆறு தூண்கள்: 383 வழங்குநர்கள் முழுவதும் நம்பகமான fallback · தகுதியான பணிச்சுமைகளில் token சேமிப்பு 95% வரை · 150+ இலவச tiers மற்றும் தொடர்ந்து கிடைக்கும் அல்லது keyless ஆக இலவசமாக இருக்கும் 52 வழங்குநர்களுடன் தொடங்க $0 · ஒரே config மூலம் 39 CLI/agent integrations · /v1-ல் OpenAI, Claude, Gemini மற்றும் Responses API compatibility · circuit breakers, TLS stealth, MCP 110 tools, A2A, memory, guardrails, evals மற்றும் 7,800+ கண்காணிக்கப்படும் test files முழுவதும் 56,000+ static test declarations உள்ளிட்ட production controls. +வாக்குறுதி — ஒரே endpoint மற்றும் 384 வழங்குநர்கள். மற்றொரு ஆரோக்கியமான இலக்கு கிடைக்கும் வரை தானியங்கி fallback வழிமாற்றத்தைத் தொடரச் செய்கிறது. ஆறு தூண்கள்: 384 வழங்குநர்கள் முழுவதும் நம்பகமான fallback · தகுதியான பணிச்சுமைகளில் token சேமிப்பு 95% வரை · 150+ இலவச tiers மற்றும் தொடர்ந்து கிடைக்கும் அல்லது keyless ஆக இலவசமாக இருக்கும் 52 வழங்குநர்களுடன் தொடங்க $0 · ஒரே config மூலம் 39 CLI/agent integrations · /v1-ல் OpenAI, Claude, Gemini மற்றும் Responses API compatibility · circuit breakers, TLS stealth, MCP 110 tools, A2A, memory, guardrails, evals மற்றும் 7,800+ கண்காணிக்கப்படும் test files முழுவதும் 56,000+ static test declarations உள்ளிட்ட production controls.

@@ -510,7 +510,7 @@ combo எதையும் உருவாக்கத் தேவையில -OmniRoute-ஐ தனித்துவமாக்குவது என்ன — 13 திறன்களில் 9router, OpenRouter, CLIProxyAPI மற்றும் LiteLLM உடனான தேதியிடப்பட்ட அம்ச ஒப்பீடு. OmniRoute: 383 வழங்குநர்கள், உள்ளமைக்கப்பட்ட 150+ இலவச அடுக்குகள், 20 வழிமுறை உத்திகள், 12-எஞ்சின் டோக்கன் சுருக்கம், 110 கருவிகளுடன் உள்ளமைக்கப்பட்ட MCP சேவையகம், A2A முகவர் நெறிமுறை, நிலையான நினைவகம், பாதுகாப்புக் கட்டுப்பாடுகள், கிளவுட் முகவர்கள், TLS கைரேகை மறைவு, Desktop/Termux/PWA மற்றும் 67 i18n UI மொழி வட்டாரங்கள். OmniRoute MIT உரிமம் பெற்றது மற்றும் உங்கள் சொந்த சேவையகத்தில் ஹோஸ்ட் செய்யலாம். போட்டியாளர்களின் திறன்களும் எண்ணிக்கைகளும் மாறக்கூடும்; இணைக்கப்பட்ட முறையியலைப் பார்க்கவும். +OmniRoute-ஐ தனித்துவமாக்குவது என்ன — 13 திறன்களில் 9router, OpenRouter, CLIProxyAPI மற்றும் LiteLLM உடனான தேதியிடப்பட்ட அம்ச ஒப்பீடு. OmniRoute: 384 வழங்குநர்கள், உள்ளமைக்கப்பட்ட 150+ இலவச அடுக்குகள், 20 வழிமுறை உத்திகள், 12-எஞ்சின் டோக்கன் சுருக்கம், 110 கருவிகளுடன் உள்ளமைக்கப்பட்ட MCP சேவையகம், A2A முகவர் நெறிமுறை, நிலையான நினைவகம், பாதுகாப்புக் கட்டுப்பாடுகள், கிளவுட் முகவர்கள், TLS கைரேகை மறைவு, Desktop/Termux/PWA மற்றும் 67 i18n UI மொழி வட்டாரங்கள். OmniRoute MIT உரிமம் பெற்றது மற்றும் உங்கள் சொந்த சேவையகத்தில் ஹோஸ்ட் செய்யலாம். போட்டியாளர்களின் திறன்களும் எண்ணிக்கைகளும் மாறக்கூடும்; இணைக்கப்பட்ட முறையியலைப் பார்க்கவும். 📊 9router, OpenRouter, CLIProxyAPI மற்றும் LiteLLM உடனான முழு முறையியல் & ஒவ்வொரு அம்சத்தின் விவரங்கள் → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -613,7 +613,7 @@ Radar-ஐப் பயன்படுத்தத் தனியாகச் - **🖼️ புதிய endpoints** — `/v1/ocr` (Mistral OCR) மற்றும் `/v1/audio/translations` (Whisper-style) ஆகியவை media surface-ஐ நிறைவு செய்கின்றன. → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 Image / video / audio generation** — media-க்கான ஒரே API: xAI Grok Imagine & Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind, மேலும் ElevenLabs போன்ற speech providers. → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 Deployment & ops** — reverse-proxy `basePath`, browser-language auto-detect, ஒவ்வொரு key-க்குமான device tracking, root-less MITM trust, zh-TW localization. → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 மேலும் providers & agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login உடன் Grok Build (xAI), first-class card கொண்ட Ollama, Claude Opus 5 & Sonnet 5, Kimi official partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… மேலும் புதுப்பிக்கப்பட்ட **383-provider catalog**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 மேலும் providers & agents** — cloud agents (Codex Cloud, Cursor, Devin, Jules), browser + OAuth login உடன் Grok Build (xAI), first-class card கொண்ட Ollama, Claude Opus 5 & Sonnet 5, Kimi official partnership (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… மேலும் புதுப்பிக்கப்பட்ட **384-provider catalog**. → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 Routing transparency** — ஒவ்வொரு response-உம் அதை வழங்கிய strategy/provider/latency-ஐ குறிப்பிடும் `X-OmniRoute-Decision` header-ஐ கொண்டுள்ளது; புதிய `cache-optimized` combo strategy மற்றும் Auto-Combo `cacheAffinity` factor ஆகியவை repeat requests-ஐ cached prefix-ஐ வைத்திருக்கும் connection-க்கு மீண்டும் வழிமாற்றுகின்றன; மேலும் read-only `/v1/auto-combo/{channel}/candidates` endpoint, `auto/*` channel-ன் live candidate pool-ஐ வெளிப்படுத்துகிறது. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ உள்ளமை செயல்திறன் & உள்கட்டமைப்பு** — ஒரே கிளிக்கில் உள்ளமை Redis, Cloudflare Workers / Deno Deploy ரிலே டிப்ளாயர்கள், கண்காணிக்கப்படும் உட்பொதிக்கப்பட்ட சேவைகளாக Bifrost & Mux. → [உட்பொதிக்கப்பட்ட சேவைகள்](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 பெட்டியிலேயே கிடைப்பவை** — செருகுநிரல் கட்டமைப்பு + சந்தை, Omni/Agent/GitHub திறன்கள் கட்டமைப்புகள், Obsidian vault ஒருங்கிணைப்பு (22 MCP கருவிகள்), OpenAI-இணக்கமான Batch & Files API-கள், பொருள் சார்ந்த பதில் கேச், லீடர்போர்டுகளுடன் கூடிய கேமிஃபிகேஷன், ACP ஏஜென்ட் கண்டறிதல் (15 உள்ளமை ஏஜென்ட்கள்), BigQuery-க்கு திட்டமிடப்பட்ட பதிவு ஏற்றுமதி, `auto/chaos` இணை பல-மாதிரி fan-out, Telegram bot பாலம், பயன்பாட்டிலேயே உள்ள பதிப்பு மேலாளர் மற்றும் LMArena-ELO இலவச வழங்குநர் தரவரிசைகள். → [ஆவணங்கள்](docs/README.md) @@ -696,11 +696,11 @@ omniroute configure codex # இதுவும்: claude opencode qwen
-## 🌐 383 AI வழங்குநர்கள் — 153 பட்டியலில் இலவசமாகக் குறிக்கப்பட்டவை +## 🌐 384 AI வழங்குநர்கள் — 153 பட்டியலில் இலவசமாகக் குறிக்கப்பட்டவை
-> **383 பதிவு செய்யப்பட்ட வழங்குநர்கள்**, canonical chat, media, search, local, cloud-agent மற்றும் system தொகுப்புகள் முழுவதும் உள்ளனர்; இதில் **153 வழங்குநர்கள் `hasFree: true` கண்டறிதல் metadata-வை கொண்டுள்ளனர்**. Chat model registry-யில் **239 வழங்குநர்கள் / 3,059 தனித்துவமான provider-model ஜோடிகள் / 1,475 மூல model IDs** உள்ளன; தனி இலவச-பட்ஜெட் பட்டியலில் **471 ஒவ்வொரு-model வரிசைகள்**, **34 தொடர்ச்சியான pool-கள்** மற்றும் **52 தொடர்ச்சியான/keyless எப்போதும்-இலவச வழங்குநர்கள்** உள்ளனர். இவை வடிவமைப்பின் அடிப்படையில் வெவ்வேறு denominator-கள்; வரையறைகளும் pool-deduped கணக்கீடுகளும் [வழங்குநர் குறிப்பு](docs/reference/PROVIDER_REFERENCE.md) மற்றும் [இலவச அடுக்குகள்](docs/reference/FREE_TIERS.md) ஆகியவற்றில் உள்ளன. +> **384 பதிவு செய்யப்பட்ட வழங்குநர்கள்**, canonical chat, media, search, local, cloud-agent மற்றும் system தொகுப்புகள் முழுவதும் உள்ளனர்; இதில் **153 வழங்குநர்கள் `hasFree: true` கண்டறிதல் metadata-வை கொண்டுள்ளனர்**. Chat model registry-யில் **239 வழங்குநர்கள் / 3,059 தனித்துவமான provider-model ஜோடிகள் / 1,475 மூல model IDs** உள்ளன; தனி இலவச-பட்ஜெட் பட்டியலில் **471 ஒவ்வொரு-model வரிசைகள்**, **34 தொடர்ச்சியான pool-கள்** மற்றும் **52 தொடர்ச்சியான/keyless எப்போதும்-இலவச வழங்குநர்கள்** உள்ளனர். இவை வடிவமைப்பின் அடிப்படையில் வெவ்வேறு denominator-கள்; வரையறைகளும் pool-deduped கணக்கீடுகளும் [வழங்குநர் குறிப்பு](docs/reference/PROVIDER_REFERENCE.md) மற்றும் [இலவச அடுக்குகள்](docs/reference/FREE_TIERS.md) ஆகியவற்றில் உள்ளன.
@@ -1302,7 +1302,7 @@ Dashboard இல்லாத headless runtime-க்கு Docker `base` profile இயக்கநேரம்Node.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 மொழிTypeScript 6.0 — src/ மற்றும் open-sse/ முழுவதும் 100% TypeScript (v2.0 முதல் core-ல் பூஜ்ய any) கட்டமைப்புNext.js 16 + React 19 + Tailwind CSS 4 - தரவுத்தளம்better-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — 147 domain modules, 230 migrations + தரவுத்தளம்better-sqlite3 (SQLite, WAL journaling) + LowDB (JSON legacy) — 147 domain modules, 232 migrations நினைவகம்SQLite FTS5 full-text + int8-quantized vector embeddings, typed decay ஸ்கீமாக்கள்Zod 4 — MCP tool I/O validation + API contracts நெறிமுறைகள்MCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ta/llm.txt b/docs/i18n/ta/llm.txt index 51ff99d7bef..283ac570fa7 100644 --- a/docs/i18n/ta/llm.txt +++ b/docs/i18n/ta/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/te/README.md b/docs/i18n/te/README.md index 0171aeae4b8..b5f0c2079f1 100644 --- a/docs/i18n/te/README.md +++ b/docs/i18n/te/README.md @@ -15,7 +15,7 @@ **[Diego Souza](https://github.com/diegosouzapw) రూపొందించారు** -OmniRoute — కోడింగ్ను ఎప్పటికీ ఆపవద్దు. ప్రతి AI సాధనం → 383 ప్రొవైడర్లు — 150+ ఉచితం — ఒకే ఎండ్పాయింట్ ద్వారా. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity → ఉచిత Claude / GPT / Geminiకి ఆటో-ఫాల్బ్యాక్తో. RTK + Caveman స్ట్యాక్డ్ కంప్రెషన్ 15–95% టోకెన్లను (~89% సగటు) ఆదా చేస్తుంది — పరిమితులను ఎప్పటికీ తాకవద్దు. 383 AI ప్రొవైడర్లు · 150+ ఉచిత శ్రేణులు · ~1.62B ఉచిత టోకెన్లు/నెల · 20 రూటింగ్ వ్యూహాలు · ప్రారంభించడానికి $0. +OmniRoute — కోడింగ్ను ఎప్పటికీ ఆపవద్దు. ప్రతి AI సాధనం → 384 ప్రొవైడర్లు — 150+ ఉచితం — ఒకే ఎండ్పాయింట్ ద్వారా. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity → ఉచిత Claude / GPT / Geminiకి ఆటో-ఫాల్బ్యాక్తో. RTK + Caveman స్ట్యాక్డ్ కంప్రెషన్ 15–95% టోకెన్లను (~89% సగటు) ఆదా చేస్తుంది — పరిమితులను ఎప్పటికీ తాకవద్దు. 384 AI ప్రొవైడర్లు · 150+ ఉచిత శ్రేణులు · ~1.62B ఉచిత టోకెన్లు/నెల · 20 రూటింగ్ వ్యూహాలు · ప్రారంభించడానికి $0.
@@ -73,7 +73,7 @@ OMNIROUTE మీకు డబ్బు ఆదా చేసి, మీ పని | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------ | :-----------------------: | :---------: | :-------------------: | -| 🌐 ప్రొవైడర్లు | 352 | **358** | 383 | +| 🌐 ప్రొవైడర్లు | 352 | **358** | 384 | | 🧠 ప్రత్యేకమైన చాట్ మోడల్ IDలు | 1320 | **1374** | 1491 | | 🖼️ మోడాలిటీ బ్రిడ్జ్ | 🆕 విజన్ + ఆడియో + వీడియో | ✓ | ✓ | | 📡 రాడార్ ఉచిత కేటలాగ్ | 🆕 ఆప్ట్-ఇన్ | ✓ | ✓ | @@ -114,7 +114,7 @@ OMNIROUTE మీకు డబ్బు ఆదా చేసి, మీ పని ⚙️ ఫీచర్లు 🎯 కాంబోలు - 🌐 ప్రొవైడర్లు + 🌐 ప్రొవైడర్లు 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -వాగ్దానం — ఒక ఎండ్పాయింట్ మరియు 383 ప్రొవైడర్లు. మరొక ఆరోగ్యకరమైన లక్ష్యం అందుబాటులో ఉన్నంతకాలం ఆటోమేటిక్ ఫాల్బ్యాక్ రూటింగ్ను కొనసాగిస్తుంది. ఆరు స్తంభాలు: 383 ప్రొవైడర్లలో స్థిరమైన ఫాల్బ్యాక్ · అర్హత కలిగిన వర్క్లోడ్లపై గరిష్ఠంగా 95% టోకెన్ ఆదా · 150+ ఉచిత టియర్లు మరియు పునరావృతంగా లభించే/కీ-లెస్ శాశ్వత ఉచిత ప్రొవైడర్లు 52తో $0తో ప్రారంభం · ఒకే కాన్ఫిగ్ ద్వారా 39 CLI/ఏజెంట్ ఇంటిగ్రేషన్లు · /v1 వద్ద OpenAI, Claude, Gemini మరియు Responses API అనుకూలత · సర్క్యూట్ బ్రేకర్లు, TLS స్టెల్త్, MCP 110 టూల్స్, A2A, మెమరీ, గార్డ్రెయిల్స్, ఎవాల్స్ మరియు 7,800+ ట్రాక్ చేయబడిన టెస్ట్ ఫైళ్లలో 56,000+ స్టాటిక్ టెస్ట్ డిక్లరేషన్లతో సహా ప్రొడక్షన్ నియంత్రణలు. +వాగ్దానం — ఒక ఎండ్పాయింట్ మరియు 384 ప్రొవైడర్లు. మరొక ఆరోగ్యకరమైన లక్ష్యం అందుబాటులో ఉన్నంతకాలం ఆటోమేటిక్ ఫాల్బ్యాక్ రూటింగ్ను కొనసాగిస్తుంది. ఆరు స్తంభాలు: 384 ప్రొవైడర్లలో స్థిరమైన ఫాల్బ్యాక్ · అర్హత కలిగిన వర్క్లోడ్లపై గరిష్ఠంగా 95% టోకెన్ ఆదా · 150+ ఉచిత టియర్లు మరియు పునరావృతంగా లభించే/కీ-లెస్ శాశ్వత ఉచిత ప్రొవైడర్లు 52తో $0తో ప్రారంభం · ఒకే కాన్ఫిగ్ ద్వారా 39 CLI/ఏజెంట్ ఇంటిగ్రేషన్లు · /v1 వద్ద OpenAI, Claude, Gemini మరియు Responses API అనుకూలత · సర్క్యూట్ బ్రేకర్లు, TLS స్టెల్త్, MCP 110 టూల్స్, A2A, మెమరీ, గార్డ్రెయిల్స్, ఎవాల్స్ మరియు 7,800+ ట్రాక్ చేయబడిన టెస్ట్ ఫైళ్లలో 56,000+ స్టాటిక్ టెస్ట్ డిక్లరేషన్లతో సహా ప్రొడక్షన్ నియంత్రణలు.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRouteను ప్రత్యేకంగా నిలబెట్టేవి — 13 సామర్థ్యాల్లో 9router, OpenRouter, CLIProxyAPI మరియు LiteLLMతో పోల్చిన తేదీతో కూడిన ఫీచర్ స్నాప్షాట్. OmniRoute: 383 ప్రొవైడర్లు, అంతర్నిర్మితంగా 150+ ఉచిత టియర్లు, 20 రూటింగ్ వ్యూహాలు, 12-ఇంజిన్ టోకెన్ కంప్రెషన్, 110 టూల్స్తో అంతర్నిర్మిత MCP సర్వర్, A2A ఏజెంట్ ప్రోటోకాల్, నిరంతర మెమరీ, గార్డ్రైల్స్, క్లౌడ్ ఏజెంట్లు, TLS ఫింగర్ప్రింట్ స్టెల్త్, Desktop/Termux/PWA మరియు 67 i18n UI లోకేల్స్. OmniRoute MIT లైసెన్స్తో అందుబాటులో ఉండి, స్వయంగా హోస్ట్ చేసుకోవచ్చు. ప్రత్యర్థుల సామర్థ్యాలు మరియు సంఖ్యలు మారవచ్చు; లింక్ చేసిన విధానాన్ని చూడండి. +OmniRouteను ప్రత్యేకంగా నిలబెట్టేవి — 13 సామర్థ్యాల్లో 9router, OpenRouter, CLIProxyAPI మరియు LiteLLMతో పోల్చిన తేదీతో కూడిన ఫీచర్ స్నాప్షాట్. OmniRoute: 384 ప్రొవైడర్లు, అంతర్నిర్మితంగా 150+ ఉచిత టియర్లు, 20 రూటింగ్ వ్యూహాలు, 12-ఇంజిన్ టోకెన్ కంప్రెషన్, 110 టూల్స్తో అంతర్నిర్మిత MCP సర్వర్, A2A ఏజెంట్ ప్రోటోకాల్, నిరంతర మెమరీ, గార్డ్రైల్స్, క్లౌడ్ ఏజెంట్లు, TLS ఫింగర్ప్రింట్ స్టెల్త్, Desktop/Termux/PWA మరియు 67 i18n UI లోకేల్స్. OmniRoute MIT లైసెన్స్తో అందుబాటులో ఉండి, స్వయంగా హోస్ట్ చేసుకోవచ్చు. ప్రత్యర్థుల సామర్థ్యాలు మరియు సంఖ్యలు మారవచ్చు; లింక్ చేసిన విధానాన్ని చూడండి. 📊 9router, OpenRouter, CLIProxyAPI మరియు LiteLLMతో పూర్తి విధానం & ప్రతి ఫీచర్ వివరాలు → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar ఐచ్ఛికం మరియు GET-మాత్రమే. OmniRoute - **🖼️ కొత్త ఎండ్పాయింట్లు** — `/v1/ocr` (Mistral OCR) మరియు `/v1/audio/translations` (Whisper-శైలి) మీడియా ఉపరితలాన్ని మరింత పూర్తి చేస్తాయి. → [API సూచిక](docs/reference/API_REFERENCE.md) - **🎨 చిత్రం / వీడియో / ఆడియో జనరేషన్** — మీడియా కోసం ఒకే API: xAI Grok Imagine మరియు Novita AI వీడియో, ComfyUI, Magnific, Adobe Firefly, Segmind, అలాగే ElevenLabs వంటి స్పీచ్ ప్రొవైడర్లు. → [API సూచిక](docs/reference/API_REFERENCE.md) - **🌍 డిప్లాయ్మెంట్ మరియు ఆప్స్** — రివర్స్-ప్రాక్సీ `basePath`, బ్రౌజర్-భాష ఆటో-డిటెక్షన్, ప్రతి కీకి డివైస్ ట్రాకింగ్, రూట్-లెస్ MITM ట్రస్ట్, zh-TW స్థానికీకరణ. → [ఎన్విరాన్మెంట్](docs/reference/ENVIRONMENT.md) -- **🤝 మరిన్ని ప్రొవైడర్లు మరియు ఏజెంట్లు** — క్లౌడ్ ఏజెంట్లు (Codex Cloud, Cursor, Devin, Jules), బ్రౌజర్ + OAuth లాగిన్తో Grok Build (xAI), ఫస్ట్-క్లాస్ కార్డ్గా Ollama, Claude Opus 5 మరియు Sonnet 5, Kimi అధికారిక భాగస్వామ్యం (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… అలాగే పునరుద్ధరించబడిన **383-ప్రొవైడర్ కాటలాగ్**. → [ప్రొవైడర్లు](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 మరిన్ని ప్రొవైడర్లు మరియు ఏజెంట్లు** — క్లౌడ్ ఏజెంట్లు (Codex Cloud, Cursor, Devin, Jules), బ్రౌజర్ + OAuth లాగిన్తో Grok Build (xAI), ఫస్ట్-క్లాస్ కార్డ్గా Ollama, Claude Opus 5 మరియు Sonnet 5, Kimi అధికారిక భాగస్వామ్యం (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… అలాగే పునరుద్ధరించబడిన **384-ప్రొవైడర్ కాటలాగ్**. → [ప్రొవైడర్లు](docs/reference/PROVIDER_REFERENCE.md) - **📡 రూటింగ్ పారదర్శకత** — ప్రతి ప్రతిస్పందనలో దాన్ని అందించిన వ్యూహం/ప్రొవైడర్/లేటెన్సీని సూచించే `X-OmniRoute-Decision` హెడర్ ఉంటుంది; కొత్త `cache-optimized` కాంబో వ్యూహం మరియు Auto-Combo `cacheAffinity` ఫ్యాక్టర్ పునరావృత అభ్యర్థనలను క్యాష్ చేసిన ప్రిఫిక్స్ను కలిగి ఉన్న కనెక్షన్కు తిరిగి పంపుతాయి; అలాగే కేవలం చదవగల `/v1/auto-combo/{channel}/candidates` ఎండ్పాయింట్ `auto/*` ఛానల్కి చెందిన ప్రత్యక్ష క్యాండిడేట్ పూల్ను బహిర్గతం చేస్తుంది. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ స్థానిక పనితీరు & మౌలిక సదుపాయాలు** — ఒక్క క్లిక్తో స్థానిక Redis, Cloudflare Workers / Deno Deploy రిలే డిప్లాయర్లు, పర్యవేక్షిత ఎంబెడెడ్ సేవలుగా Bifrost & Mux. → [ఎంబెడెడ్ సేవలు](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 ప్యాకేజీలో ఇవి కూడా ఉన్నాయి** — ప్లగిన్ ఫ్రేమ్వర్క్ + మార్కెట్ప్లేస్, Omni/Agent/GitHub స్కిల్స్ ఫ్రేమ్వర్క్లు, Obsidian వాల్ట్ ఇంటిగ్రేషన్ (22 MCP సాధనాలు), OpenAI-అనుకూల Batch & Files APIలు, సెమాంటిక్ ప్రతిస్పందన క్యాష్, లీడర్బోర్డ్లతో గేమిఫికేషన్, ACP ఏజెంట్ డిస్కవరీ (అంతర్నిర్మిత 15 ఏజెంట్లు), BigQueryకి షెడ్యూల్ చేసిన లాగ్ ఎగుమతి, `auto/chaos` సమాంతర బహుళ-మోడల్ ఫ్యాన్-అవుట్, Telegram బాట్ బ్రిడ్జ్, యాప్లోని వెర్షన్ మేనేజర్ మరియు LMArena-ELO ఉచిత ప్రొవైడర్ ర్యాంకింగ్లు. → [డాక్యుమెంటేషన్](docs/README.md) @@ -689,11 +689,11 @@ omniroute configure codex # అలాగే: claude opencode qwen aider
-## 🌐 383 AI ప్రొవైడర్లు — 153 క్యాటలాగ్లో ఉచితంగా గుర్తించబడినవి +## 🌐 384 AI ప్రొవైడర్లు — 153 క్యాటలాగ్లో ఉచితంగా గుర్తించబడినవి
-> **383 నమోదైన ప్రొవైడర్లు** canonical chat, media, search, local, cloud-agent మరియు system సేకరణలలో ఉన్నాయి. వీరిలో **153 ప్రొవైడర్లు `hasFree: true` డిస్కవరీ మెటాడేటాను కలిగి ఉన్నారు**. Chat model registryలో **239 ప్రొవైడర్లు / 3,059 ప్రత్యేక provider-model జతలు / 1,475 raw model IDs** ఉన్నాయి; వేరు free-budget catalogలో **471 ప్రతి-మోడల్ వరుసలు**, **34 పునరావృత poolలు** మరియు **52 పునరావృతం/keyless free-forever ప్రొవైడర్లు** ఉన్నాయి. ఇవి ఉద్దేశపూర్వకంగా వేర్వేరు denominators; నిర్వచనాలు మరియు pool-deduped లెక్కలు [ప్రొవైడర్ రిఫరెన్స్](docs/reference/PROVIDER_REFERENCE.md) మరియు [ఫ్రీ టియర్స్](docs/reference/FREE_TIERS.md)లో ఉన్నాయి. +> **384 నమోదైన ప్రొవైడర్లు** canonical chat, media, search, local, cloud-agent మరియు system సేకరణలలో ఉన్నాయి. వీరిలో **153 ప్రొవైడర్లు `hasFree: true` డిస్కవరీ మెటాడేటాను కలిగి ఉన్నారు**. Chat model registryలో **239 ప్రొవైడర్లు / 3,059 ప్రత్యేక provider-model జతలు / 1,475 raw model IDs** ఉన్నాయి; వేరు free-budget catalogలో **471 ప్రతి-మోడల్ వరుసలు**, **34 పునరావృత poolలు** మరియు **52 పునరావృతం/keyless free-forever ప్రొవైడర్లు** ఉన్నాయి. ఇవి ఉద్దేశపూర్వకంగా వేర్వేరు denominators; నిర్వచనాలు మరియు pool-deduped లెక్కలు [ప్రొవైడర్ రిఫరెన్స్](docs/reference/PROVIDER_REFERENCE.md) మరియు [ఫ్రీ టియర్స్](docs/reference/FREE_TIERS.md)లో ఉన్నాయి.
@@ -1295,7 +1295,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 కూడా ద రన్టైమ్Node.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 భాషTypeScript 6.0 — src/ మరియు open-sse/ అంతటా 100% TypeScript (v2.0 నుండి కోర్లో సున్నా any) ఫ్రేమ్వర్క్Next.js 16 + React 19 + Tailwind CSS 4 - డేటాబేస్better-sqlite3 (SQLite, WAL జర్నలింగ్) + LowDB (JSON లెగసీ) — 147 డొమైన్ మాడ్యూళ్లు, 230 మైగ్రేషన్లు + డేటాబేస్better-sqlite3 (SQLite, WAL జర్నలింగ్) + LowDB (JSON లెగసీ) — 147 డొమైన్ మాడ్యూళ్లు, 232 మైగ్రేషన్లు మెమరీSQLite FTS5 పూర్తి-పాఠ్య + int8-క్వాంటైజ్ చేసిన వెక్టర్ ఎంబెడ్డింగ్లు, టైప్ చేసిన డికే స్కీమాలుZod 4 — MCP టూల్ I/O ధృవీకరణ + API కాంట్రాక్ట్లు ప్రోటోకాల్లుMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/te/llm.txt b/docs/i18n/te/llm.txt index 14783335a7f..3ae3523c203 100644 --- a/docs/i18n/te/llm.txt +++ b/docs/i18n/te/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/th/README.md b/docs/i18n/th/README.md index 656c4cb4b48..7cc178aa9a8 100644 --- a/docs/i18n/th/README.md +++ b/docs/i18n/th/README.md @@ -15,7 +15,7 @@ **สร้างโดย [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — อย่าหยุดเขียนโค้ด เครื่องมือ AI ทุกตัว → ผู้ให้บริการ 383 ราย — ฟรีกว่า 150 ราย — ผ่านเอนด์พอยต์เดียว: Claude Code, Codex, Cursor, Cline, Copilot และ Antigravity สู่ Claude / GPT / Gemini ฟรี พร้อมการสลับเส้นทางอัตโนมัติ RTK + Caveman compression แบบซ้อนชั้นช่วยประหยัดโทเค็น 15–95% (~89% โดยเฉลี่ย) — ไม่จำกัดอีกต่อไป ผู้ให้บริการ AI 383 ราย · แพ็กเกจฟรีกว่า 150 ราย · โทเค็นฟรีประมาณ 1.62 พันล้านโทเค็น/เดือน · กลยุทธ์การกำหนดเส้นทาง 20 แบบ · เริ่มต้นที่ $0. +OmniRoute — อย่าหยุดเขียนโค้ด เครื่องมือ AI ทุกตัว → ผู้ให้บริการ 384 ราย — ฟรีกว่า 150 ราย — ผ่านเอนด์พอยต์เดียว: Claude Code, Codex, Cursor, Cline, Copilot และ Antigravity สู่ Claude / GPT / Gemini ฟรี พร้อมการสลับเส้นทางอัตโนมัติ RTK + Caveman compression แบบซ้อนชั้นช่วยประหยัดโทเค็น 15–95% (~89% โดยเฉลี่ย) — ไม่จำกัดอีกต่อไป ผู้ให้บริการ AI 384 ราย · แพ็กเกจฟรีกว่า 150 ราย · โทเค็นฟรีประมาณ 1.62 พันล้านโทเค็น/เดือน · กลยุทธ์การกำหนดเส้นทาง 20 แบบ · เริ่มต้นที่ $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | -------------------------- | :----------------------------: | :---------: | :-------------------: | -| 🌐 ผู้ให้บริการ | 352 | **358** | 383 | +| 🌐 ผู้ให้บริการ | 352 | **358** | 384 | | 🧠 ID โมเดลแชตที่ไม่ซ้ำกัน | 1320 | **1374** | 1491 | | 🖼️ บริดจ์ความสามารถ | 🆕 การมองเห็น + เสียง + วิดีโอ | ✓ | ✓ | | 📡 แคตตาล็อกฟรีของ Radar | 🆕 เลือกเข้าร่วม | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ ฟีเจอร์ 🎯 ชุดการใช้งาน - 🌐 ผู้ให้บริการ + 🌐 ผู้ให้บริการ 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -คำมั่นสัญญา — ปลายทางเดียวและผู้ให้บริการ 383 ราย การสำรองข้อมูลอัตโนมัติช่วยให้การกำหนดเส้นทางดำเนินต่อไปได้ตราบใดที่ยังมีเป้าหมายอื่นที่ทำงานปกติอยู่ เสาหลักหกประการ: การสำรองข้อมูลที่ยืดหยุ่นผ่านผู้ให้บริการ 383 ราย · ประหยัดโทเค็นได้สูงสุด 95% สำหรับเวิร์กโหลดที่เข้าเกณฑ์ · เริ่มต้นที่ $0 ด้วยระดับฟรีมากกว่า 150 รายการ และผู้ให้บริการฟรีตลอดไปแบบเกิดซ้ำ/ไม่ใช้คีย์ 52 ราย · การผสานรวม CLI/เอเจนต์ 39 รายการผ่านการตั้งค่าเดียว · ความเข้ากันได้กับ OpenAI, Claude, Gemini และ Responses API ที่ /v1 · การควบคุมสำหรับใช้งานจริง รวมถึง circuit breakers, TLS stealth, MCP 110 เครื่องมือ, A2A, memory, guardrails, evals และการประกาศการทดสอบแบบคงที่มากกว่า 56,000 รายการในไฟล์ทดสอบที่ติดตามมากกว่า 7,800 ไฟล์ +คำมั่นสัญญา — ปลายทางเดียวและผู้ให้บริการ 384 ราย การสำรองข้อมูลอัตโนมัติช่วยให้การกำหนดเส้นทางดำเนินต่อไปได้ตราบใดที่ยังมีเป้าหมายอื่นที่ทำงานปกติอยู่ เสาหลักหกประการ: การสำรองข้อมูลที่ยืดหยุ่นผ่านผู้ให้บริการ 384 ราย · ประหยัดโทเค็นได้สูงสุด 95% สำหรับเวิร์กโหลดที่เข้าเกณฑ์ · เริ่มต้นที่ $0 ด้วยระดับฟรีมากกว่า 150 รายการ และผู้ให้บริการฟรีตลอดไปแบบเกิดซ้ำ/ไม่ใช้คีย์ 52 ราย · การผสานรวม CLI/เอเจนต์ 39 รายการผ่านการตั้งค่าเดียว · ความเข้ากันได้กับ OpenAI, Claude, Gemini และ Responses API ที่ /v1 · การควบคุมสำหรับใช้งานจริง รวมถึง circuit breakers, TLS stealth, MCP 110 เครื่องมือ, A2A, memory, guardrails, evals และการประกาศการทดสอบแบบคงที่มากกว่า 56,000 รายการในไฟล์ทดสอบที่ติดตามมากกว่า 7,800 ไฟล์

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -สิ่งที่ทำให้ OmniRoute โดดเด่น — ภาพรวมฟีเจอร์ ณ วันที่จัดทำ เมื่อเปรียบเทียบกับ 9router, OpenRouter, CLIProxyAPI และ LiteLLM ในด้านความสามารถ 13 รายการ OmniRoute: ผู้ให้บริการ 383 รายการ, free tier ที่มีมาให้มากกว่า 150 รายการ, กลยุทธ์การกำหนดเส้นทาง 20 แบบ, การบีบอัดโทเค็นด้วยเอนจิน 12 ตัว, เซิร์ฟเวอร์ MCP ในตัวพร้อมเครื่องมือ 110 รายการ, โปรโตคอลตัวแทน A2A, หน่วยความจำถาวร, กลไกควบคุมความปลอดภัย, ตัวแทนบนคลาวด์, การพรางลายนิ้วมือ TLS, Desktop/Termux/PWA และภาษาสำหรับ UI ของ i18n 67 ภาษา OmniRoute ใช้สัญญาอนุญาต MIT และสามารถโฮสต์เองได้ ความสามารถและจำนวนของคู่แข่งอาจเปลี่ยนแปลงได้ โปรดดูวิธีการที่ลิงก์ไว้ +สิ่งที่ทำให้ OmniRoute โดดเด่น — ภาพรวมฟีเจอร์ ณ วันที่จัดทำ เมื่อเปรียบเทียบกับ 9router, OpenRouter, CLIProxyAPI และ LiteLLM ในด้านความสามารถ 13 รายการ OmniRoute: ผู้ให้บริการ 384 รายการ, free tier ที่มีมาให้มากกว่า 150 รายการ, กลยุทธ์การกำหนดเส้นทาง 20 แบบ, การบีบอัดโทเค็นด้วยเอนจิน 12 ตัว, เซิร์ฟเวอร์ MCP ในตัวพร้อมเครื่องมือ 110 รายการ, โปรโตคอลตัวแทน A2A, หน่วยความจำถาวร, กลไกควบคุมความปลอดภัย, ตัวแทนบนคลาวด์, การพรางลายนิ้วมือ TLS, Desktop/Termux/PWA และภาษาสำหรับ UI ของ i18n 67 ภาษา OmniRoute ใช้สัญญาอนุญาต MIT และสามารถโฮสต์เองได้ ความสามารถและจำนวนของคู่แข่งอาจเปลี่ยนแปลงได้ โปรดดูวิธีการที่ลิงก์ไว้ 📊 วิธีการฉบับเต็มและรายละเอียดแยกตามฟีเจอร์ เมื่อเปรียบเทียบกับ 9router, OpenRouter, CLIProxyAPI และ LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar เป็นฟีเจอร์ที่ต้องเลือกเ - **🖼️ endpoint ใหม่** — `/v1/ocr` (Mistral OCR) และ `/v1/audio/translations` (รูปแบบ Whisper) ช่วยเติมเต็มพื้นผิวการรองรับสื่อ → [API Reference](docs/reference/API_REFERENCE.md) - **🎨 การสร้างภาพ / วิดีโอ / เสียง** — API เดียวสำหรับสื่อ: วิดีโอจาก xAI Grok Imagine และ Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind และผู้ให้บริการเสียงพูด เช่น ElevenLabs → [API Reference](docs/reference/API_REFERENCE.md) - **🌍 การติดตั้งใช้งานและการปฏิบัติการ** — `basePath` ของ reverse proxy, การตรวจจับภาษาของเบราว์เซอร์อัตโนมัติ, การติดตามอุปกรณ์แยกตามคีย์, trust ของ MITM โดยไม่ต้องใช้ root และการแปลเป็น zh-TW → [Environment](docs/reference/ENVIRONMENT.md) -- **🤝 ผู้ให้บริการและเอเจนต์เพิ่มเติม** — เอเจนต์บนคลาวด์ (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) พร้อมเบราว์เซอร์และการเข้าสู่ระบบ OAuth, การ์ด Ollama แบบรองรับโดยตรง, Claude Opus 5 และ Sonnet 5, ความร่วมมืออย่างเป็นทางการกับ Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… และแค็ตตาล็อก **383 ผู้ให้บริการ** ที่ปรับปรุงใหม่ → [Providers](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 ผู้ให้บริการและเอเจนต์เพิ่มเติม** — เอเจนต์บนคลาวด์ (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) พร้อมเบราว์เซอร์และการเข้าสู่ระบบ OAuth, การ์ด Ollama แบบรองรับโดยตรง, Claude Opus 5 และ Sonnet 5, ความร่วมมืออย่างเป็นทางการกับ Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… และแค็ตตาล็อก **384 ผู้ให้บริการ** ที่ปรับปรุงใหม่ → [Providers](docs/reference/PROVIDER_REFERENCE.md) - **📡 ความโปร่งใสของการกำหนดเส้นทาง** — ทุกการตอบกลับมีเฮดเดอร์ `X-OmniRoute-Decision` ที่ระบุกลยุทธ์/ผู้ให้บริการ/เวลาแฝงที่ให้บริการคำขอนั้น, กลยุทธ์คอมโบ `cache-optimized` ใหม่ + ปัจจัย `cacheAffinity` ของ Auto-Combo จะกำหนดเส้นทางคำขอซ้ำกลับไปยังการเชื่อมต่อที่ถือคำนำหน้าซึ่งแคชไว้ และ endpoint แบบอ่านอย่างเดียว `/v1/auto-combo/{channel}/candidates` จะแสดงพูลผู้สมัครแบบสดของช่อง `auto/*` → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ ประสิทธิภาพและโครงสร้างพื้นฐานภายในเครื่อง** — Redis ภายในเครื่องแบบคลิกเดียว, ตัวติดตั้งรีเลย์ Cloudflare Workers / Deno Deploy, Bifrost และ Mux ในฐานะบริการแบบฝังตัวที่มีการกำกับดูแล → [บริการแบบฝังตัว](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 รวมอยู่ในชุดนี้ด้วย** — เฟรมเวิร์กปลั๊กอิน + มาร์เก็ตเพลส, เฟรมเวิร์กทักษะ Omni/Agent/GitHub, การผสานรวมคลัง Obsidian (เครื่องมือ MCP 22 รายการ), Batch และ Files APIs ที่เข้ากันได้กับ OpenAI, แคชการตอบสนองเชิงความหมาย, ระบบเกมมิฟิเคชันพร้อมลีดเดอร์บอร์ด, การค้นพบเอเจนต์ ACP (เอเจนต์ในตัว 15 รายการ), การส่งออกบันทึกตามกำหนดเวลาไปยัง BigQuery, การกระจายงานแบบขนานหลายโมเดลของ `auto/chaos`, บริดจ์บอต Telegram, ตัวจัดการเวอร์ชันภายในแอป และการจัดอันดับผู้ให้บริการฟรีของ LMArena-ELO → [เอกสาร](docs/README.md) @@ -689,11 +689,11 @@ omniroute configure codex # ใช้ได้กับ: claude opencod
-## 🌐 ผู้ให้บริการ AI 383 ราย — 153 รายถูกระบุว่าให้บริการฟรีในแคตตาล็อก +## 🌐 ผู้ให้บริการ AI 384 ราย — 153 รายถูกระบุว่าให้บริการฟรีในแคตตาล็อก
-> **ผู้ให้บริการที่ลงทะเบียนแล้ว 383 ราย** ครอบคลุมคอลเลกชันแชต สื่อ การค้นหา แบบโลคัล ตัวแทนบนคลาวด์ และระบบตามมาตรฐาน รวมถึง **153 รายที่มีเมทาดาทาการค้นพบ `hasFree: true`** รีจิสทรีโมเดลแชตครอบคลุม **ผู้ให้บริการ 239 ราย / คู่ผู้ให้บริการ-โมเดลที่แตกต่างกัน 3,059 คู่ / ID โมเดลดิบ 1,475 รายการ** ส่วนแคตตาล็อกงบประมาณฟรีแยกต่างหากมี **แถวต่อโมเดล 471 แถว**, **พูลแบบเกิดซ้ำ 34 พูล** และ **ผู้ให้บริการฟรีตลอดไปแบบเกิดซ้ำ/ไม่ต้องใช้คีย์ 52 ราย** ตัวเลขเหล่านี้ใช้ตัวหารที่แตกต่างกันโดยตั้งใจ ดูคำจำกัดความและการคำนวณที่หักรายการซ้ำของพูลได้ใน [ข้อมูลอ้างอิงผู้ให้บริการ](docs/reference/PROVIDER_REFERENCE.md) และ [ระดับฟรี](docs/reference/FREE_TIERS.md) +> **ผู้ให้บริการที่ลงทะเบียนแล้ว 384 ราย** ครอบคลุมคอลเลกชันแชต สื่อ การค้นหา แบบโลคัล ตัวแทนบนคลาวด์ และระบบตามมาตรฐาน รวมถึง **153 รายที่มีเมทาดาทาการค้นพบ `hasFree: true`** รีจิสทรีโมเดลแชตครอบคลุม **ผู้ให้บริการ 239 ราย / คู่ผู้ให้บริการ-โมเดลที่แตกต่างกัน 3,059 คู่ / ID โมเดลดิบ 1,475 รายการ** ส่วนแคตตาล็อกงบประมาณฟรีแยกต่างหากมี **แถวต่อโมเดล 471 แถว**, **พูลแบบเกิดซ้ำ 34 พูล** และ **ผู้ให้บริการฟรีตลอดไปแบบเกิดซ้ำ/ไม่ต้องใช้คีย์ 52 ราย** ตัวเลขเหล่านี้ใช้ตัวหารที่แตกต่างกันโดยตั้งใจ ดูคำจำกัดความและการคำนวณที่หักรายการซ้ำของพูลได้ใน [ข้อมูลอ้างอิงผู้ให้บริการ](docs/reference/PROVIDER_REFERENCE.md) และ [ระดับฟรี](docs/reference/FREE_TIERS.md)
@@ -1295,7 +1295,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 จะข้า รันไทม์Node.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ภาษาTypeScript 6.0 — TypeScript 100% ใน src/ และ open-sse/ (ไม่มี any ในคอร์นับตั้งแต่ v2.0) เฟรมเวิร์กNext.js 16 + React 19 + Tailwind CSS 4 - ฐานข้อมูลbetter-sqlite3 (SQLite, การทำเจอร์นัล WAL) + LowDB (JSON แบบเดิม) — โมดูลโดเมน 147 รายการ, migration 230 รายการ + ฐานข้อมูลbetter-sqlite3 (SQLite, การทำเจอร์นัล WAL) + LowDB (JSON แบบเดิม) — โมดูลโดเมน 147 รายการ, migration 232 รายการ หน่วยความจำฟูลเท็กซ์ SQLite FTS5 + vector embeddings แบบควอนไทซ์ int8 พร้อมการเสื่อมตามชนิดข้อมูล สคีมาZod 4 — การตรวจสอบ I/O ของเครื่องมือ MCP + สัญญา API โปรโตคอลMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/th/llm.txt b/docs/i18n/th/llm.txt index c25d7a580f5..e01cece0533 100644 --- a/docs/i18n/th/llm.txt +++ b/docs/i18n/th/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/tr/README.md b/docs/i18n/tr/README.md index bc96a472023..85fbd3e5ffa 100644 --- a/docs/i18n/tr/README.md +++ b/docs/i18n/tr/README.md @@ -15,7 +15,7 @@ **[Diego Souza](https://github.com/diegosouzapw) tarafından oluşturuldu** -OmniRoute — Kod yazmayı asla bırakmayın. Her yapay zekâ aracı → tek bir uç nokta üzerinden 383 sağlayıcı — 150+ ücretsiz —. Claude Code, Codex, Cursor, Cline, Copilot ve Antigravity ile ÜCRETSİZ Claude / GPT / Gemini; otomatik geri dönüş. RTK + Caveman yığılmış sıkıştırması %15–95 token tasarrufu sağlar (ortalama ~%89) — sınırlara asla takılmayın. 383 yapay zekâ sağlayıcısı · 150+ ücretsiz katman · ayda ~1,62 milyar ücretsiz token · 20 yönlendirme stratejisi · Başlamak için $0. +OmniRoute — Kod yazmayı asla bırakmayın. Her yapay zekâ aracı → tek bir uç nokta üzerinden 384 sağlayıcı — 150+ ücretsiz —. Claude Code, Codex, Cursor, Cline, Copilot ve Antigravity ile ÜCRETSİZ Claude / GPT / Gemini; otomatik geri dönüş. RTK + Caveman yığılmış sıkıştırması %15–95 token tasarrufu sağlar (ortalama ~%89) — sınırlara asla takılmayın. 384 yapay zekâ sağlayıcısı · 150+ ücretsiz katman · ayda ~1,62 milyar ücretsiz token · 20 yönlendirme stratejisi · Başlamak için $0.
@@ -73,7 +73,7 @@ OMNIROUTE paradan tasarruf etmenize ve işinizi kolaylaştırmanıza yardımcı | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------------ | :----------------------: | :---------: | :-------------------: | -| 🌐 Sağlayıcılar | 352 | **358** | 383 | +| 🌐 Sağlayıcılar | 352 | **358** | 384 | | 🧠 Benzersiz sohbet model kimlikleri | 1320 | **1374** | 1491 | | 🖼️ Modality Bridge | 🆕 görüntü + ses + video | ✓ | ✓ | | 📡 Radar ücretsiz kataloğu | 🆕 isteğe bağlı | ✓ | ✓ | @@ -114,7 +114,7 @@ OMNIROUTE paradan tasarruf etmenize ve işinizi kolaylaştırmanıza yardımcı ⚙️ Özellikler 🎯 Kombolar - 🌐 Sağlayıcılar + 🌐 Sağlayıcılar 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Vaat — Tek bir uç nokta ve 383 sağlayıcı. Otomatik geri dönüş, başka bir sağlıklı hedef kullanılabilir olduğu sürece yönlendirmeyi sürdürür. Altı temel özellik: 383 sağlayıcı genelinde dayanıklı geri dönüş · uygun iş yüklerinde %95'e varan token tasarrufu · 150'den fazla ücretsiz katman ve 52 sürekli/anahtarsız sonsuza kadar ücretsiz sağlayıcıyla başlamak için $0 · tek bir yapılandırma üzerinden 39 CLI/agent entegrasyonu · /v1 adresinde OpenAI, Claude, Gemini ve Responses API uyumluluğu · devre kesiciler, TLS gizliliği, 110 MCP aracı, A2A, bellek, guardrail'ler, değerlendirmeler ve 7.800'den fazla izlenen test dosyasında 56.000'den fazla statik test bildirimi dahil üretim kontrolleri. +Vaat — Tek bir uç nokta ve 384 sağlayıcı. Otomatik geri dönüş, başka bir sağlıklı hedef kullanılabilir olduğu sürece yönlendirmeyi sürdürür. Altı temel özellik: 384 sağlayıcı genelinde dayanıklı geri dönüş · uygun iş yüklerinde %95'e varan token tasarrufu · 150'den fazla ücretsiz katman ve 52 sürekli/anahtarsız sonsuza kadar ücretsiz sağlayıcıyla başlamak için $0 · tek bir yapılandırma üzerinden 39 CLI/agent entegrasyonu · /v1 adresinde OpenAI, Claude, Gemini ve Responses API uyumluluğu · devre kesiciler, TLS gizliliği, 110 MCP aracı, A2A, bellek, guardrail'ler, değerlendirmeler ve 7.800'den fazla izlenen test dosyasında 56.000'den fazla statik test bildirimi dahil üretim kontrolleri.

@@ -510,7 +510,7 @@ Kombo oluşturmanız gerekmez. Modelinizi `auto` (veya bir varyantı) olarak aya -OmniRoute'u farklı kılan özellikler — 13 yetenek genelinde 9router, OpenRouter, CLIProxyAPI ve LiteLLM ile güncel özellik karşılaştırması. OmniRoute: 383 sağlayıcı, yerleşik 150'den fazla ücretsiz katman, 20 yönlendirme stratejisi, 12 motorlu belirteç sıkıştırma, 110 araç içeren yerleşik MCP sunucusu, A2A aracı protokolü, kalıcı bellek, koruma kuralları, bulut aracıları, TLS parmak izi gizleme, Desktop/Termux/PWA ve 67 i18n kullanıcı arayüzü yerel ayarı. OmniRoute MIT lisanslıdır ve kendi altyapınızda barındırılabilir. Rakiplerin yetenekleri ve sayıları değişebilir; bağlantısı verilen metodolojiye bakın. +OmniRoute'u farklı kılan özellikler — 13 yetenek genelinde 9router, OpenRouter, CLIProxyAPI ve LiteLLM ile güncel özellik karşılaştırması. OmniRoute: 384 sağlayıcı, yerleşik 150'den fazla ücretsiz katman, 20 yönlendirme stratejisi, 12 motorlu belirteç sıkıştırma, 110 araç içeren yerleşik MCP sunucusu, A2A aracı protokolü, kalıcı bellek, koruma kuralları, bulut aracıları, TLS parmak izi gizleme, Desktop/Termux/PWA ve 67 i18n kullanıcı arayüzü yerel ayarı. OmniRoute MIT lisanslıdır ve kendi altyapınızda barındırılabilir. Rakiplerin yetenekleri ve sayıları değişebilir; bağlantısı verilen metodolojiye bakın. 📊 9router, OpenRouter, CLIProxyAPI ve LiteLLM ile tam metodoloji & özellik bazında ayrıntılar → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar isteğe bağlıdır ve yalnızca GET kullanır. OmniRoute istemcisi isteml - **🖼️ Yeni uç noktalar** — `/v1/ocr` (Mistral OCR) ve `/v1/audio/translations` (Whisper tarzı), medya yüzeyini tamamlar. → [API Referansı](docs/reference/API_REFERENCE.md) - **🎨 Görüntü / video / ses oluşturma** — medya için tek API: xAI Grok Imagine ve Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind ve ElevenLabs gibi konuşma sağlayıcıları. → [API Referansı](docs/reference/API_REFERENCE.md) - **🌍 Dağıtım ve operasyonlar** — ters proxy `basePath`, tarayıcı dili otomatik algılama, anahtar başına cihaz takibi, kök yetkisi gerektirmeyen MITM güveni ve zh-TW yerelleştirmesi. → [Ortam](docs/reference/ENVIRONMENT.md) -- **🤝 Daha fazla sağlayıcı ve aracı** — bulut aracıları (Codex Cloud, Cursor, Devin, Jules), tarayıcı + OAuth girişiyle Grok Build (xAI), birinci sınıf kart olarak Ollama, Claude Opus 5 ve Sonnet 5, Kimi ile resmî ortaklık (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… ve yenilenmiş **383 sağlayıcılı katalog**. → [Sağlayıcılar](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Daha fazla sağlayıcı ve aracı** — bulut aracıları (Codex Cloud, Cursor, Devin, Jules), tarayıcı + OAuth girişiyle Grok Build (xAI), birinci sınıf kart olarak Ollama, Claude Opus 5 ve Sonnet 5, Kimi ile resmî ortaklık (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… ve yenilenmiş **384 sağlayıcılı katalog**. → [Sağlayıcılar](docs/reference/PROVIDER_REFERENCE.md) - **📡 Yönlendirme şeffaflığı** — her yanıt, isteği karşılayan strateji/sağlayıcı/gecikmeyi belirten bir `X-OmniRoute-Decision` başlığı taşır; yeni bir `cache-optimized` birleşim stratejisi ile Auto-Combo `cacheAffinity` faktörü, yinelenen istekleri önbelleğe alınmış öneki tutan bağlantıya geri yönlendirir ve salt okunur `/v1/auto-combo/{channel}/candidates` uç noktası, bir `auto/*` kanalının canlı aday havuzunu kullanıma sunar. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Yerel performans ve altyapı** — tek tıkla yerel Redis, Cloudflare Workers / Deno Deploy aktarım dağıtıcıları, denetimli gömülü hizmetler olarak Bifrost ve Mux. → [Gömülü Hizmetler](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Kutuda ayrıca** — eklenti çerçevesi ve marketplace, Omni/Agent/GitHub beceri çerçeveleri, Obsidian kasa entegrasyonu (22 MCP aracı), OpenAI uyumlu Batch ve Files API'leri, anlamsal yanıt önbelleği, liderlik tablolarıyla oyunlaştırma, ACP aracı keşfi (yerleşik 15 aracı), BigQuery'ye zamanlanmış günlük dışa aktarımı, `auto/chaos` paralel çoklu model fan-out'u, Telegram bot köprüsü, uygulama içi sürüm yöneticisi ve LMArena-ELO ücretsiz sağlayıcı sıralamaları. → [Belgeler](docs/README.md) @@ -689,11 +689,11 @@ kabuk geçmişinizden uzak tutar. → [CLI Entegrasyonları](docs/guides/CLI-INT
-## 🌐 383 Yapay Zekâ Sağlayıcısı — Katalogda İşaretlenmiş 153 Ücretsiz +## 🌐 384 Yapay Zekâ Sağlayıcısı — Katalogda İşaretlenmiş 153 Ücretsiz
-> **Kanonik sohbet, medya, arama, yerel, bulut aracısı ve sistem koleksiyonlarında** toplam **383 kayıtlı sağlayıcı** bulunur; bunların **153’ü `hasFree: true` keşif meta verisini taşır**. Sohbet modeli kayıt defteri **239 sağlayıcıyı / 3.059 farklı sağlayıcı-model çiftini / 1.475 ham model kimliğini** kapsar; ayrı ücretsiz bütçe kataloğunda **model başına 471 satır**, **34 yinelenen havuz** ve **52 yinelenen/anahtarsız sonsuza kadar ücretsiz sağlayıcı** bulunur. Bunlar tasarım gereği farklı paydalardır; tanımlar ve havuz yinelemeleri giderilmiş hesaplamalar [Sağlayıcı Referansı](docs/reference/PROVIDER_REFERENCE.md) ve [Ücretsiz Katmanlar](docs/reference/FREE_TIERS.md) bölümünde yer alır. +> **Kanonik sohbet, medya, arama, yerel, bulut aracısı ve sistem koleksiyonlarında** toplam **384 kayıtlı sağlayıcı** bulunur; bunların **153’ü `hasFree: true` keşif meta verisini taşır**. Sohbet modeli kayıt defteri **239 sağlayıcıyı / 3.059 farklı sağlayıcı-model çiftini / 1.475 ham model kimliğini** kapsar; ayrı ücretsiz bütçe kataloğunda **model başına 471 satır**, **34 yinelenen havuz** ve **52 yinelenen/anahtarsız sonsuza kadar ücretsiz sağlayıcı** bulunur. Bunlar tasarım gereği farklı paydalardır; tanımlar ve havuz yinelemeleri giderilmiş hesaplamalar [Sağlayıcı Referansı](docs/reference/PROVIDER_REFERENCE.md) ve [Ücretsiz Katmanlar](docs/reference/FREE_TIERS.md) bölümünde yer alır.
@@ -1294,7 +1294,7 @@ aynı süreç tarafından sunulur, bu nedenle bugün için ayrı bir yalnızca C Çalışma zamanıNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 DilTypeScript 6.0 — src/ ve open-sse/ genelinde %100 TypeScript (v2.0'dan bu yana çekirdek kodda sıfır any) ÇatıNext.js 16 + React 19 + Tailwind CSS 4 - Veritabanıbetter-sqlite3 (SQLite, WAL günlüğe kaydetme) + LowDB (JSON eski sistem) — 147 alan modülü, 230 geçiş + Veritabanıbetter-sqlite3 (SQLite, WAL günlüğe kaydetme) + LowDB (JSON eski sistem) — 147 alan modülü, 232 geçiş BellekSQLite FTS5 tam metin + int8 nicemlemeli vektör gömmeleri, tür bilgili azalma ŞemalarZod 4 — MCP araç G/Ç doğrulaması + API sözleşmeleri ProtokollerMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/tr/llm.txt b/docs/i18n/tr/llm.txt index d0cbeae039c..936aa02077b 100644 --- a/docs/i18n/tr/llm.txt +++ b/docs/i18n/tr/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/uk-UA/README.md b/docs/i18n/uk-UA/README.md index 230db8cc3df..267b864111a 100644 --- a/docs/i18n/uk-UA/README.md +++ b/docs/i18n/uk-UA/README.md @@ -15,7 +15,7 @@ **Створено [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Ніколи не припиняйте програмувати. Кожен інструмент ШІ → 383 провайдер — 150+ безкоштовних — через одну кінцеву точку. Claude Code, Codex, Cursor, Cline, Copilot та Antigravity у БЕЗКОШТОВНІ Claude / GPT / Gemini з автоматичним резервним перемиканням. Послідовне стиснення RTK + Caveman заощаджує 15–95% токенів (~89% у середньому) — ліміти більше не перешкода. 383 провайдер ШІ · 150+ безкоштовних рівнів · ~1.62B безкоштовних токенів/міс. · 20 стратегій маршрутизації · $0 для початку. +OmniRoute — Ніколи не припиняйте програмувати. Кожен інструмент ШІ → 384 провайдер — 150+ безкоштовних — через одну кінцеву точку. Claude Code, Codex, Cursor, Cline, Copilot та Antigravity у БЕЗКОШТОВНІ Claude / GPT / Gemini з автоматичним резервним перемиканням. Послідовне стиснення RTK + Caveman заощаджує 15–95% токенів (~89% у середньому) — ліміти більше не перешкода. 384 провайдер ШІ · 150+ безкоштовних рівнів · ~1.62B безкоштовних токенів/міс. · 20 стратегій маршрутизації · $0 для початку.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | -------------------------------- | :-----------------------: | :---------: | :-------------------: | -| 🌐 Провайдери | 352 | **358** | 383 | +| 🌐 Провайдери | 352 | **358** | 384 | | 🧠 Унікальні ID чат-моделей | 1320 | **1374** | 1491 | | 🖼️ Міст модальностей | 🆕 vision + audio + video | ✓ | ✓ | | 📡 Безкоштовний каталог Radar | 🆕 opt-in | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Функції 🎯 Комбінації - 🌐 Провайдери + 🌐 Провайдери 🔌 CLI та MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Обіцянка — одна кінцева точка та 383 провайдер. Автоматичний резервний перехід підтримує маршрутизацію, доки доступна інша справна ціль. Шість основних переваг: відмовостійкий резервний перехід між 383 провайдером · до 95% економії токенів для відповідних робочих навантажень · $0 для початку зі 150+ безкоштовними тарифами та 52 безкоштовними назавжди провайдерами з регулярним або безключовим доступом · 39 інтеграцій CLI/агентів через одну конфігурацію · сумісність з OpenAI, Claude, Gemini та Responses API через /v1 · засоби керування для продуктивного середовища, зокрема автоматичні вимикачі, приховування TLS, MCP зі 110 інструментами, A2A, пам’ять, обмеження безпеки, оцінювання та понад 56 000 статичних тестових декларацій у понад 7 800 відстежуваних тестових файлах. +Обіцянка — одна кінцева точка та 384 провайдер. Автоматичний резервний перехід підтримує маршрутизацію, доки доступна інша справна ціль. Шість основних переваг: відмовостійкий резервний перехід між 384 провайдером · до 95% економії токенів для відповідних робочих навантажень · $0 для початку зі 150+ безкоштовними тарифами та 52 безкоштовними назавжди провайдерами з регулярним або безключовим доступом · 39 інтеграцій CLI/агентів через одну конфігурацію · сумісність з OpenAI, Claude, Gemini та Responses API через /v1 · засоби керування для продуктивного середовища, зокрема автоматичні вимикачі, приховування TLS, MCP зі 110 інструментами, A2A, пам’ять, обмеження безпеки, оцінювання та понад 56 000 статичних тестових декларацій у понад 7 800 відстежуваних тестових файлах.

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -Чим вирізняється OmniRoute — актуальний знімок функціональності порівняно з 9router, OpenRouter, CLIProxyAPI та LiteLLM за 13 можливостями. OmniRoute: 383 провайдер, понад 150 вбудованих безкоштовних тарифів, 20 стратегій маршрутизації, стиснення токенів на базі 12 рушіїв, вбудований MCP-сервер зі 110 інструментами, протокол агентів A2A, постійна пам’ять, захисні обмеження, хмарні агенти, приховування TLS-відбитка, Desktop/Termux/PWA та 67 локалей інтерфейсу i18n. OmniRoute ліцензовано за MIT і можна розгортати на власній інфраструктурі. Можливості та кількість функцій конкурентів можуть змінюватися; див. методологію за посиланням. +Чим вирізняється OmniRoute — актуальний знімок функціональності порівняно з 9router, OpenRouter, CLIProxyAPI та LiteLLM за 13 можливостями. OmniRoute: 384 провайдер, понад 150 вбудованих безкоштовних тарифів, 20 стратегій маршрутизації, стиснення токенів на базі 12 рушіїв, вбудований MCP-сервер зі 110 інструментами, протокол агентів A2A, постійна пам’ять, захисні обмеження, хмарні агенти, приховування TLS-відбитка, Desktop/Termux/PWA та 67 локалей інтерфейсу i18n. OmniRoute ліцензовано за MIT і можна розгортати на власній інфраструктурі. Можливості та кількість функцій конкурентів можуть змінюватися; див. методологію за посиланням. 📊 Повна методологія та деталі за кожною функцією порівняно з 9router, OpenRouter, CLIProxyAPI та LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar використовується за бажанням і лише чер - **🖼️ Нові кінцеві точки** — `/v1/ocr` (Mistral OCR) і `/v1/audio/translations` (у стилі Whisper) доповнюють набір медіаможливостей. → [Довідник API](docs/reference/API_REFERENCE.md) - **🎨 Генерація зображень / відео / аудіо** — один API для медіа: xAI Grok Imagine і відео Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind та провайдери мовлення, як-от ElevenLabs. → [Довідник API](docs/reference/API_REFERENCE.md) - **🌍 Розгортання та експлуатація** — `basePath` для зворотного проксі, автоматичне визначення мови браузера, відстеження пристроїв для кожного ключа, довіра MITM без root, локалізація zh-TW. → [Середовище](docs/reference/ENVIRONMENT.md) -- **🤝 Більше провайдерів та агентів** — хмарні агенти (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) із входом через браузер і OAuth, повноцінна картка Ollama, Claude Opus 5 і Sonnet 5, офіційне партнерство з Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… і оновлений **каталог із 383 провайдера**. → [Провайдери](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Більше провайдерів та агентів** — хмарні агенти (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) із входом через браузер і OAuth, повноцінна картка Ollama, Claude Opus 5 і Sonnet 5, офіційне партнерство з Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… і оновлений **каталог із 384 провайдера**. → [Провайдери](docs/reference/PROVIDER_REFERENCE.md) - **📡 Прозорість маршрутизації** — кожна відповідь містить заголовок `X-OmniRoute-Decision` із назвою стратегії/провайдера/затримки, які її обробили; нова стратегія комбінацій `cache-optimized` і фактор маршруту `cacheAffinity` в Auto-Combo спрямовують повторні запити до з’єднання, що містить кешований префікс; кінцева точка лише для читання `/v1/auto-combo/{channel}/candidates` показує актуальний пул кандидатів каналу `auto/*`. → [Автоматичні комбінації](docs/routing/AUTO-COMBO.md). - **⚡ Локальна продуктивність та інфраструктура** — локальний Redis в один клік, засоби розгортання ретрансляторів Cloudflare Workers / Deno Deploy, Bifrost і Mux як контрольовані вбудовані служби. → [Вбудовані служби](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Також у комплекті** — фреймворк плагінів і маркетплейс, фреймворки навичок Omni/Agent/GitHub, інтеграція зі сховищем Obsidian (22 інструменти MCP), сумісні з OpenAI API пакетні API та API файлів, семантичний кеш відповідей, гейміфікація з таблицями лідерів, виявлення агентів ACP (15 вбудованих агентів), запланований експорт журналів до BigQuery, паралельне розгалуження на кілька моделей `auto/chaos`, міст для бота Telegram, менеджер версій у застосунку та рейтинги безкоштовних провайдерів LMArena-ELO. → [Документація](docs/README.md) @@ -689,11 +689,11 @@ omniroute configure codex # також: claude opencode qwen aider goos
-## 🌐 383 провайдерів AI — 153 позначені каталогом як безкоштовні +## 🌐 384 провайдерів AI — 153 позначені каталогом як безкоштовні
-> **383 зареєстрований провайдер** у канонічних колекціях чатів, медіа, пошуку, локальних, хмарних агентів і систем, зокрема **153 провайдери з метаданими виявлення `hasFree: true`**. Реєстр чат-моделей охоплює **239 провайдерів / 3 059 унікальних пар провайдер-модель / 1 475 ідентифікаторів необроблених моделей**; окремий каталог безкоштовних бюджетів містить **471 рядок для окремих моделей**, **34 періодичні пули** та **52 провайдери з регулярним або без ключа безкоштовним доступом назавжди**. Це навмисно різні знаменники; визначення та розрахунки з дедуплікацією пулів наведено в [Довіднику провайдерів](docs/reference/PROVIDER_REFERENCE.md) і [Безкоштовних тарифах](docs/reference/FREE_TIERS.md). +> **384 зареєстрований провайдер** у канонічних колекціях чатів, медіа, пошуку, локальних, хмарних агентів і систем, зокрема **153 провайдери з метаданими виявлення `hasFree: true`**. Реєстр чат-моделей охоплює **239 провайдерів / 3 059 унікальних пар провайдер-модель / 1 475 ідентифікаторів необроблених моделей**; окремий каталог безкоштовних бюджетів містить **471 рядок для окремих моделей**, **34 періодичні пули** та **52 провайдери з регулярним або без ключа безкоштовним доступом назавжди**. Це навмисно різні знаменники; визначення та розрахунки з дедуплікацією пулів наведено в [Довіднику провайдерів](docs/reference/PROVIDER_REFERENCE.md) і [Безкоштовних тарифах](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 також про Середовище виконанняNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 МоваTypeScript 6.0 — 100% TypeScript у src/ і open-sse/ (нуль any у ядрі починаючи з v2.0) ФреймворкNext.js 16 + React 19 + Tailwind CSS 4 - База данихbetter-sqlite3 (SQLite, журналювання WAL) + LowDB (застарілий JSON) — 147 доменних модулів, 230 міграцій + База данихbetter-sqlite3 (SQLite, журналювання WAL) + LowDB (застарілий JSON) — 147 доменних модулів, 232 міграцій Пам’ятьПовнотекстовий пошук SQLite FTS5 + векторні ембеддинги з квантуванням int8, типізоване згасання СхемиZod 4 — перевірка введення/виведення інструментів MCP + контракти API ПротоколиMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/uk-UA/llm.txt b/docs/i18n/uk-UA/llm.txt index 843fcecc6f8..a597ed1655b 100644 --- a/docs/i18n/uk-UA/llm.txt +++ b/docs/i18n/uk-UA/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/ur/README.md b/docs/i18n/ur/README.md index a45ad746861..791b2546ae7 100644 --- a/docs/i18n/ur/README.md +++ b/docs/i18n/ur/README.md @@ -15,7 +15,7 @@ **تخلیق کار [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — کبھی کوڈنگ مت روکیں۔ ہر AI ٹول → 383 فراہم کنندگان — 150+ مفت — ایک endpoint کے ذریعے۔ Claude Code، Codex، Cursor، Cline، Copilot اور Antigravity کو خودکار fallback کے ساتھ مفت Claude / GPT / Gemini میں تبدیل کریں۔ RTK + Caveman کی تہہ دار compression سے 15–95% tokens (~89% اوسط) کی بچت — کبھی limits سے نہ ٹکرائیں۔ 383 AI فراہم کنندگان · 150+ مفت tiers · ~1.62B مفت tokens/ماہ · routing کی 20 strategies · شروع کرنے کی لاگت $0. +OmniRoute — کبھی کوڈنگ مت روکیں۔ ہر AI ٹول → 384 فراہم کنندگان — 150+ مفت — ایک endpoint کے ذریعے۔ Claude Code، Codex، Cursor، Cline، Copilot اور Antigravity کو خودکار fallback کے ساتھ مفت Claude / GPT / Gemini میں تبدیل کریں۔ RTK + Caveman کی تہہ دار compression سے 15–95% tokens (~89% اوسط) کی بچت — کبھی limits سے نہ ٹکرائیں۔ 384 AI فراہم کنندگان · 150+ مفت tiers · ~1.62B مفت tokens/ماہ · routing کی 20 strategies · شروع کرنے کی لاگت $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------ | :-------------------: | :---------: | :-------------------: | -| 🌐 فراہم کنندگان | 352 | **358** | 383 | +| 🌐 فراہم کنندگان | 352 | **358** | 384 | | 🧠 منفرد چیٹ ماڈل IDs | 1320 | **1374** | 1491 | | 🖼️ موڈیلیٹی برج | 🆕 وژن + آڈیو + ویڈیو | ✓ | ✓ | | 📡 ریڈار مفت کیٹلاگ | 🆕 آپٹ اِن | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ خصوصیات 🎯 کومبو - 🌐 فراہم کنندگان + 🌐 فراہم کنندگان 🔌 CLI اور MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -وعدہ — ایک endpoint اور 383 providers۔ جب کوئی دوسرا صحت مند target دستیاب ہو تو خودکار fallback روٹنگ جاری رکھتا ہے۔ چھ ستون: 383 providers کے درمیان resilient fallback · اہل workloads پر token کی بچت 95% تک · شروع کرنے کے لیے $0، 150+ free tiers اور 52 recurring/keyless free-forever providers کے ساتھ · ایک config کے ذریعے 39 CLI/agent integrations · /v1 پر OpenAI، Claude، Gemini اور Responses API compatibility · production controls، جن میں circuit breakers، TLS stealth، MCP کے 110 tools، A2A، memory، guardrails، evals اور 7,800+ tracked test files میں 56,000+ static test declarations شامل ہیں۔ +وعدہ — ایک endpoint اور 384 providers۔ جب کوئی دوسرا صحت مند target دستیاب ہو تو خودکار fallback روٹنگ جاری رکھتا ہے۔ چھ ستون: 384 providers کے درمیان resilient fallback · اہل workloads پر token کی بچت 95% تک · شروع کرنے کے لیے $0، 150+ free tiers اور 52 recurring/keyless free-forever providers کے ساتھ · ایک config کے ذریعے 39 CLI/agent integrations · /v1 پر OpenAI، Claude، Gemini اور Responses API compatibility · production controls، جن میں circuit breakers، TLS stealth، MCP کے 110 tools، A2A، memory، guardrails، evals اور 7,800+ tracked test files میں 56,000+ static test declarations شامل ہیں۔

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRoute کو ممتاز بنانے والی خصوصیات — 13 صلاحیتوں کے حوالے سے 9router، OpenRouter، CLIProxyAPI اور LiteLLM کے مقابلے میں ایک تاریخ شدہ فیچر اسنیپ شاٹ۔ OmniRoute میں 383 فراہم کنندگان، پہلے سے شامل 150+ مفت درجات، 20 روٹنگ حکمتِ عملیاں، 12-انجن ٹوکن کمپریشن، 110 ٹولز والا بلٹ اِن MCP سرور، A2A ایجنٹ پروٹوکول، مستقل میموری، گارڈ ریلز، کلاؤڈ ایجنٹس، TLS فنگر پرنٹ اسٹیلتھ، Desktop/Termux/PWA اور i18n UI کے لیے 67 لوکیلز شامل ہیں۔ OmniRoute MIT لائسنس یافتہ اور خود میزبانی کے قابل ہے۔ حریفوں کی صلاحیتیں اور تعداد تبدیل ہو سکتی ہیں؛ منسلک طریقۂ کار ملاحظہ کریں۔ +OmniRoute کو ممتاز بنانے والی خصوصیات — 13 صلاحیتوں کے حوالے سے 9router، OpenRouter، CLIProxyAPI اور LiteLLM کے مقابلے میں ایک تاریخ شدہ فیچر اسنیپ شاٹ۔ OmniRoute میں 384 فراہم کنندگان، پہلے سے شامل 150+ مفت درجات، 20 روٹنگ حکمتِ عملیاں، 12-انجن ٹوکن کمپریشن، 110 ٹولز والا بلٹ اِن MCP سرور، A2A ایجنٹ پروٹوکول، مستقل میموری، گارڈ ریلز، کلاؤڈ ایجنٹس، TLS فنگر پرنٹ اسٹیلتھ، Desktop/Termux/PWA اور i18n UI کے لیے 67 لوکیلز شامل ہیں۔ OmniRoute MIT لائسنس یافتہ اور خود میزبانی کے قابل ہے۔ حریفوں کی صلاحیتیں اور تعداد تبدیل ہو سکتی ہیں؛ منسلک طریقۂ کار ملاحظہ کریں۔ 📊 9router، OpenRouter، CLIProxyAPI اور LiteLLM کے مقابلے میں مکمل طریقۂ کار اور فیچر بہ فیچر تفصیل → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -609,7 +609,7 @@ Radar اختیاری ہے اور صرف GET استعمال کرتا ہے۔ OmniR - **🖼️ نئے endpoints** — `/v1/ocr` (Mistral OCR) اور `/v1/audio/translations` (Whisper-style)، میڈیا کی سطح کو مکمل کرتے ہیں۔ → [API حوالہ](docs/reference/API_REFERENCE.md) - **🎨 تصویر / ویڈیو / آڈیو تخلیق** — میڈیا کے لیے ایک API: xAI Grok Imagine اور Novita AI video، ComfyUI، Magnific، Adobe Firefly، Segmind، اور ElevenLabs جیسے speech providers۔ → [API حوالہ](docs/reference/API_REFERENCE.md) - **🌍 تعیناتی اور آپریشنز** — reverse-proxy `basePath`، براؤزر-زبان کی خودکار شناخت، فی-کلید ڈیوائس ٹریکنگ، root-less MITM trust، اور zh-TW لوکلائزیشن۔ → [ماحول](docs/reference/ENVIRONMENT.md) -- **🤝 مزید فراہم کنندگان اور ایجنٹس** — cloud agents (Codex Cloud، Cursor، Devin، Jules)، browser + OAuth login کے ساتھ Grok Build (xAI)، first-class card والا Ollama، Claude Opus 5 اور Sonnet 5، Kimi کی سرکاری شراکت داری (Code/Web/Moonshot)، Zed، Requesty، SenseNova، Yuanbao، Agnes AI… اور تازہ کردہ **383-provider catalog**۔ → [فراہم کنندگان](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 مزید فراہم کنندگان اور ایجنٹس** — cloud agents (Codex Cloud، Cursor، Devin، Jules)، browser + OAuth login کے ساتھ Grok Build (xAI)، first-class card والا Ollama، Claude Opus 5 اور Sonnet 5، Kimi کی سرکاری شراکت داری (Code/Web/Moonshot)، Zed، Requesty، SenseNova، Yuanbao، Agnes AI… اور تازہ کردہ **384-provider catalog**۔ → [فراہم کنندگان](docs/reference/PROVIDER_REFERENCE.md) - **📡 شفاف روٹنگ** — ہر response میں `X-OmniRoute-Decision` ہیڈر شامل ہوتا ہے جو اسے فراہم کرنے والی strategy/provider/latency کا نام بتاتا ہے؛ نئی `cache-optimized` combo strategy اور Auto-Combo `cacheAffinity` factor بار بار آنے والی درخواستوں کو اس connection کی طرف واپس بھیجتے ہیں جس کے پاس cached prefix موجود ہو؛ اور صرف مطالعے کے لیے `/v1/auto-combo/{channel}/candidates` endpoint کسی `auto/*` channel کے فعال candidate pool کو ظاہر کرتا ہے۔ → [آٹو-کومبو](docs/routing/AUTO-COMBO.md) - **⚡ مقامی کارکردگی اور انفرا —** ایک کلک پر مقامی Redis، Cloudflare Workers / Deno Deploy ریلے ڈپلائرز، اور Bifrost و Mux بطور زیرِ نگرانی ایمبیڈڈ سروسز۔ → [ایمبیڈڈ سروسز](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 پیکیج میں مزید شامل —** پلگ اِن فریم ورک اور مارکیٹ پلیس، Omni/Agent/GitHub اسکلز فریم ورکس، Obsidian والٹ انٹیگریشن (22 MCP ٹولز)، OpenAI سے ہم آہنگ Batch اور Files APIs، سیمینٹک رسپانس کیش، لیڈر بورڈز کے ساتھ گیمیفیکیشن، ACP ایجنٹ دریافت (15 بلٹ اِن ایجنٹس)، BigQuery کو مقررہ شیڈول کے مطابق لاگ ایکسپورٹ، `auto/chaos` متوازی ملٹی ماڈل فین آؤٹ، Telegram بوٹ برج، ایپ میں موجود ورژن مینیجر، اور LMArena-ELO مفت فراہم کنندگان کی درجہ بندیاں۔ → [دستاویزات](docs/README.md) @@ -692,11 +692,11 @@ omniroute configure codex # یہ بھی: claude opencode qwen aider goo
-## 🌐 383 AI فراہم کنندگان — 153 کیٹلاگ میں نشان زدہ مفت +## 🌐 384 AI فراہم کنندگان — 153 کیٹلاگ میں نشان زدہ مفت
-> **383 رجسٹرڈ فراہم کنندگان** کینونیکل چیٹ، میڈیا، سرچ، لوکل، کلاؤڈ-ایجنٹ اور سسٹم کلیکشنز میں شامل ہیں، جن میں **153 کے پاس `hasFree: true` دریافت میٹا ڈیٹا** موجود ہے۔ چیٹ ماڈل رجسٹری میں **239 فراہم کنندگان / 3,059 منفرد فراہم کنندہ-ماڈل جوڑے / 1,475 خام ماڈل IDs** شامل ہیں؛ جبکہ علیحدہ مفت بجٹ کیٹلاگ میں **471 فی-ماڈل قطاریں**، **34 بار بار آنے والے پولز** اور **52 بار بار آنے والے/کی لیس ہمیشہ کے لیے مفت فراہم کنندگان** شامل ہیں۔ یہ مختلف ڈینومینیٹرز جان بوجھ کر رکھے گئے ہیں؛ تعریفیں اور پول-ڈی ڈپلیکیٹ شدہ حسابات [فراہم کنندہ حوالہ](docs/reference/PROVIDER_REFERENCE.md) اور [مفت درجات](docs/reference/FREE_TIERS.md) میں موجود ہیں۔ +> **384 رجسٹرڈ فراہم کنندگان** کینونیکل چیٹ، میڈیا، سرچ، لوکل، کلاؤڈ-ایجنٹ اور سسٹم کلیکشنز میں شامل ہیں، جن میں **153 کے پاس `hasFree: true` دریافت میٹا ڈیٹا** موجود ہے۔ چیٹ ماڈل رجسٹری میں **239 فراہم کنندگان / 3,059 منفرد فراہم کنندہ-ماڈل جوڑے / 1,475 خام ماڈل IDs** شامل ہیں؛ جبکہ علیحدہ مفت بجٹ کیٹلاگ میں **471 فی-ماڈل قطاریں**، **34 بار بار آنے والے پولز** اور **52 بار بار آنے والے/کی لیس ہمیشہ کے لیے مفت فراہم کنندگان** شامل ہیں۔ یہ مختلف ڈینومینیٹرز جان بوجھ کر رکھے گئے ہیں؛ تعریفیں اور پول-ڈی ڈپلیکیٹ شدہ حسابات [فراہم کنندہ حوالہ](docs/reference/PROVIDER_REFERENCE.md) اور [مفت درجات](docs/reference/FREE_TIERS.md) میں موجود ہیں۔
@@ -1297,7 +1297,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 بھی اسے چ رن ٹائمNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 زبانTypeScript 6.0 — 100% TypeScript، src/ اور open-sse/ میں (v2.0 سے کور میں صفر any) فریم ورکNext.js 16 + React 19 + Tailwind CSS 4 - ڈیٹابیسbetter-sqlite3 (SQLite، WAL جرنلنگ) + LowDB (JSON لیگیسی) — 147 ڈومین ماڈیولز، 230 مائیگریشنز + ڈیٹابیسbetter-sqlite3 (SQLite، WAL جرنلنگ) + LowDB (JSON لیگیسی) — 147 ڈومین ماڈیولز، 232 مائیگریشنز میموریSQLite FTS5 مکمل متن + int8-quantized ویکٹر ایمبیڈنگز، ٹائپ شدہ تنزلی اسکیمازZod 4 — MCP ٹول I/O کی توثیق + API معاہدے پروٹوکولزMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/ur/llm.txt b/docs/i18n/ur/llm.txt index 142c971e2e9..c9595499eed 100644 --- a/docs/i18n/ur/llm.txt +++ b/docs/i18n/ur/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/uz/README.md b/docs/i18n/uz/README.md index 7b5b10725ad..767168686de 100644 --- a/docs/i18n/uz/README.md +++ b/docs/i18n/uz/README.md @@ -15,7 +15,7 @@ **[Diego Souza](https://github.com/diegosouzapw) tomonidan yaratilgan** -OmniRoute — Kod yozishni hech qachon to‘xtatmang. Har bir AI vositasi → bitta endpoint orqali 383 ta provayder — 150+ tasi bepul. Claude Code, Codex, Cursor, Cline, Copilot va Antigravity → bepul Claude / GPT / Gemini xizmatlariga avtomatik qayta yo‘naltirish bilan. RTK + Caveman kombinatsiyalangan siqish usuli tokenlarni 15–95% ga (~89% o‘rtacha) tejaydi — limitlarga hech qachon duch kelmang. 383 ta AI provayderi · 150+ ta bepul tarif · oyiga ~1,62 mlrd bepul token · 20 ta yo‘naltirish strategiyasi · boshlash uchun $0. +OmniRoute — Kod yozishni hech qachon to‘xtatmang. Har bir AI vositasi → bitta endpoint orqali 384 ta provayder — 150+ tasi bepul. Claude Code, Codex, Cursor, Cline, Copilot va Antigravity → bepul Claude / GPT / Gemini xizmatlariga avtomatik qayta yo‘naltirish bilan. RTK + Caveman kombinatsiyalangan siqish usuli tokenlarni 15–95% ga (~89% o‘rtacha) tejaydi — limitlarga hech qachon duch kelmang. 384 ta AI provayderi · 150+ ta bepul tarif · oyiga ~1,62 mlrd bepul token · 20 ta yo‘naltirish strategiyasi · boshlash uchun $0.
@@ -73,7 +73,7 @@ OMNIROUTE sizga pul tejash va ishingizni osonlashtirishda yordam bergan bo‘lsa | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ----------------------------------------- | :------------------------: | :---------: | :----------------------: | -| 🌐 Provayderlar | 352 | **358** | 383 | +| 🌐 Provayderlar | 352 | **358** | 384 | | 🧠 Noyob chat modeli identifikatorlari | 1320 | **1374** | 1491 | | 🖼️ Modalitet ko‘prigi | 🆕 ko‘rish + audio + video | ✓ | ✓ | | 📡 Radar bepul katalogi | 🆕 ixtiyoriy | ✓ | ✓ | @@ -114,7 +114,7 @@ OMNIROUTE sizga pul tejash va ishingizni osonlashtirishda yordam bergan bo‘lsa ⚙️ Imkoniyatlar 🎯 Kombolar - 🌐 Provayderlar + 🌐 Provayderlar 🔌 CLI va MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Vaʼda — Bitta endpoint va 383 ta provayder. Boshqa sogʻlom maqsad mavjud ekan, avtomatik zaxira yoʻnaltirish ishlashda davom etadi. Oltita ustun: 383 ta provayder boʻylab barqaror zaxira yoʻnaltirish · mos ish yuklarida tokenlarni 95% gacha tejash · 150 dan ortiq bepul tarif va takrorlanuvchi/kalitsiz, abadiy bepul 52 ta provayder bilan boshlash uchun $0 · bitta konfiguratsiya orqali CLI/agentlar uchun 39 ta integratsiya · /v1 manzilida OpenAI, Claude, Gemini va Responses API bilan moslik · circuit breakerlar, TLS stealth, 110 ta MCP vositasi, A2A, xotira, guardrails, evals hamda 7 800 dan ortiq kuzatiladigan test fayllaridagi 56 000 dan ortiq statik test deklaratsiyalarini oʻz ichiga olgan ishlab chiqarish boshqaruvlari. +Vaʼda — Bitta endpoint va 384 ta provayder. Boshqa sogʻlom maqsad mavjud ekan, avtomatik zaxira yoʻnaltirish ishlashda davom etadi. Oltita ustun: 384 ta provayder boʻylab barqaror zaxira yoʻnaltirish · mos ish yuklarida tokenlarni 95% gacha tejash · 150 dan ortiq bepul tarif va takrorlanuvchi/kalitsiz, abadiy bepul 52 ta provayder bilan boshlash uchun $0 · bitta konfiguratsiya orqali CLI/agentlar uchun 39 ta integratsiya · /v1 manzilida OpenAI, Claude, Gemini va Responses API bilan moslik · circuit breakerlar, TLS stealth, 110 ta MCP vositasi, A2A, xotira, guardrails, evals hamda 7 800 dan ortiq kuzatiladigan test fayllaridagi 56 000 dan ortiq statik test deklaratsiyalarini oʻz ichiga olgan ishlab chiqarish boshqaruvlari.

@@ -510,7 +510,7 @@ Barcha **20** ta strategiya — har bir combo bosqichida aralashtirib qoʻllang: -OmniRoute’ni ajratib turadigan jihatlar — 13 ta imkoniyat bo‘yicha 9router, OpenRouter, CLIProxyAPI va LiteLLM bilan taqqoslangan, sanasi ko‘rsatilgan imkoniyatlar surati. OmniRoute: 383 ta provayder, ichiga o‘rnatilgan 150+ ta bepul tarif, 20 ta marshrutlash strategiyasi, tokenlarni siqish uchun 12 ta dvigatel, 110 ta vositaga ega ichki MCP serveri, A2A agent protokoli, doimiy xotira, himoya cheklovlari, bulut agentlari, TLS fingerprint yashirinligi, Desktop/Termux/PWA va 67 ta i18n UI lokal tili. OmniRoute MIT litsenziyasi asosida tarqatiladi va o‘z serveringizda joylashtirilishi mumkin. Raqobatchilar imkoniyatlari va ko‘rsatkichlari o‘zgarishi mumkin; havoladagi metodologiyaga qarang. +OmniRoute’ni ajratib turadigan jihatlar — 13 ta imkoniyat bo‘yicha 9router, OpenRouter, CLIProxyAPI va LiteLLM bilan taqqoslangan, sanasi ko‘rsatilgan imkoniyatlar surati. OmniRoute: 384 ta provayder, ichiga o‘rnatilgan 150+ ta bepul tarif, 20 ta marshrutlash strategiyasi, tokenlarni siqish uchun 12 ta dvigatel, 110 ta vositaga ega ichki MCP serveri, A2A agent protokoli, doimiy xotira, himoya cheklovlari, bulut agentlari, TLS fingerprint yashirinligi, Desktop/Termux/PWA va 67 ta i18n UI lokal tili. OmniRoute MIT litsenziyasi asosida tarqatiladi va o‘z serveringizda joylashtirilishi mumkin. Raqobatchilar imkoniyatlari va ko‘rsatkichlari o‘zgarishi mumkin; havoladagi metodologiyaga qarang. 📊 9router, OpenRouter, CLIProxyAPI va LiteLLM bilan to‘liq metodologiya va har bir imkoniyat tafsilotlari → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar ixtiyoriy ravishda yoqiladi va faqat GET so‘rovlaridan foydalanadi. Omni - **🖼️ Yangi endpointlar** — `/v1/ocr` (Mistral OCR) va `/v1/audio/translations` (Whisper uslubidagi) media imkoniyatlarini to‘ldiradi. → [API ma’lumotnomasi](docs/reference/API_REFERENCE.md) - **🎨 Rasm / video / audio generatsiyasi** — media uchun yagona API: xAI Grok Imagine va Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind hamda ElevenLabs kabi nutq provayderlari. → [API ma’lumotnomasi](docs/reference/API_REFERENCE.md) - **🌍 Joylashtirish va operatsiyalar** — reverse-proxy `basePath`, brauzer tilini avtomatik aniqlash, kalit bo‘yicha qurilma kuzatuvi, root’siz MITM ishonchi, zh-TW lokalizatsiyasi. → [Muhit](docs/reference/ENVIRONMENT.md) -- **🤝 Ko‘proq provayderlar va agentlar** — bulut agentlari (Codex Cloud, Cursor, Devin, Jules), brauzer + OAuth orqali kirishga ega Grok Build (xAI), birinchi darajali Ollama kartasi, Claude Opus 5 va Sonnet 5, Kimi bilan rasmiy hamkorlik (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… hamda yangilangan **383 provayderli katalog**. → [Provayderlar](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Ko‘proq provayderlar va agentlar** — bulut agentlari (Codex Cloud, Cursor, Devin, Jules), brauzer + OAuth orqali kirishga ega Grok Build (xAI), birinchi darajali Ollama kartasi, Claude Opus 5 va Sonnet 5, Kimi bilan rasmiy hamkorlik (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… hamda yangilangan **384 provayderli katalog**. → [Provayderlar](docs/reference/PROVIDER_REFERENCE.md) - **📡 Yo‘naltirish shaffofligi** — har bir javobda unga xizmat ko‘rsatgan strategiya/provayder/kechikish nomlarini ko‘rsatuvchi `X-OmniRoute-Decision` sarlavhasi mavjud; yangi `cache-optimized` kombinatsiya strategiyasi va Auto-Combo `cacheAffinity` omili takroriy so‘rovlarni keshlangan prefiksni saqlayotgan ulanishga qayta yo‘naltiradi; faqat o‘qish uchun mo‘ljallangan `/v1/auto-combo/{channel}/candidates` endpointi esa `auto/*` kanalining joriy nomzodlar havzasini ochib beradi. → [Auto-Combo](docs/routing/AUTO-COMBO.md). - **⚡ Lokal unumdorlik va infratuzilma** — bir marta bosish bilan lokal Redis, Cloudflare Workers / Deno Deploy relay deployerlari, nazorat qilinadigan Bifrost va Mux o‘rnatilgan xizmatlar sifatida. → [O‘rnatilgan xizmatlar](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 To‘plamda yana** — plaginlar freymvorki va marketplace, Omni/Agent/GitHub ko‘nikmalar freymvorklari, Obsidian vault integratsiyasi (22 ta MCP vositasi), OpenAI bilan mos Batch va Files API’lari, semantik javob keshi, yetakchilar jadvali bilan geymifikatsiya, ACP agentlarini aniqlash (o‘rnatilgan 15 ta agent), BigQuery’ga rejalashtirilgan log eksporti, `auto/chaos` parallel ko‘p modellli fan-out, Telegram bot ko‘prigi, ilova ichidagi versiyalar menejeri va LMArena-ELO bepul provayderlar reytingi. → [Hujjatlar](docs/README.md) @@ -689,11 +689,11 @@ shell tarixidan tashqarida saqlaydi. → [CLI Integratsiyalari](docs/guides/CLI-
-## 🌐 383 ta AI provayderi — katalogda belgilangan 153 ta bepul +## 🌐 384 ta AI provayderi — katalogda belgilangan 153 ta bepul
-> **383 ta roʻyxatdan oʻtgan provayder** kanonik chat, media, qidiruv, lokal, bulut agenti va tizim toʻplamlari boʻylab mavjud, ulardan **153 tasi `hasFree: true` aniqlash metamaʼlumotlariga ega**. Chat modellari reyestri **239 ta provayder / 3,059 ta noyob provayder-model juftligi / 1,475 ta xom model ID**ni qamrab oladi; alohida bepul byudjet katalogida esa **har bir model uchun 471 ta qator**, **34 ta takrorlanuvchi havza** va **52 ta takrorlanuvchi/kalitsiz, abadiy bepul provayder** mavjud. Bular ataylab turli maxrajlardir; taʼriflar va havzalar dublikatlari olib tashlangan hisob-kitoblar [Provayder maʼlumotnomasi](docs/reference/PROVIDER_REFERENCE.md) hamda [Bepul tariflar](docs/reference/FREE_TIERS.md)da keltirilgan. +> **384 ta roʻyxatdan oʻtgan provayder** kanonik chat, media, qidiruv, lokal, bulut agenti va tizim toʻplamlari boʻylab mavjud, ulardan **153 tasi `hasFree: true` aniqlash metamaʼlumotlariga ega**. Chat modellari reyestri **239 ta provayder / 3,059 ta noyob provayder-model juftligi / 1,475 ta xom model ID**ni qamrab oladi; alohida bepul byudjet katalogida esa **har bir model uchun 471 ta qator**, **34 ta takrorlanuvchi havza** va **52 ta takrorlanuvchi/kalitsiz, abadiy bepul provayder** mavjud. Bular ataylab turli maxrajlardir; taʼriflar va havzalar dublikatlari olib tashlangan hisob-kitoblar [Provayder maʼlumotnomasi](docs/reference/PROVIDER_REFERENCE.md) hamda [Bepul tariflar](docs/reference/FREE_TIERS.md)da keltirilgan.
@@ -1291,7 +1291,7 @@ bitta jarayon xizmat koʻrsatadi, shu sababli hozircha faqat CLI uchun alohida p RuntimeNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 TilTypeScript 6.0 — src/ va open-sse/ bo‘ylab 100% TypeScript (v2.0 dan beri core’da nolta any) FreymvorkNext.js 16 + React 19 + Tailwind CSS 4 - Ma’lumotlar bazasibetter-sqlite3 (SQLite, WAL jurnallash) + LowDB (JSON merosiy) — 147 ta domen moduli, 230 ta migratsiya + Ma’lumotlar bazasibetter-sqlite3 (SQLite, WAL jurnallash) + LowDB (JSON merosiy) — 147 ta domen moduli, 232 ta migratsiya XotiraSQLite FTS5 to‘liq matnli qidiruvi + int8-kvantlangan vektor embeddinglari, tiplashtirilgan so‘nish SxemalarZod 4 — MCP vositalari I/O validatsiyasi + API kontraktlari ProtokollarMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/uz/llm.txt b/docs/i18n/uz/llm.txt index 5183379a067..25028a16879 100644 --- a/docs/i18n/uz/llm.txt +++ b/docs/i18n/uz/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/vi/README.md b/docs/i18n/vi/README.md index fe1a04bde7e..53441efbbc2 100644 --- a/docs/i18n/vi/README.md +++ b/docs/i18n/vi/README.md @@ -15,7 +15,7 @@ **Được tạo bởi [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Không bao giờ ngừng lập trình. Mọi công cụ AI → 383 nhà cung cấp — 150+ miễn phí — thông qua một endpoint duy nhất. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity đến Claude / GPT / Gemini MIỄN PHÍ với tính năng tự động chuyển dự phòng. Tính năng nén xếp chồng RTK + Caveman tiết kiệm 15–95% token (~89% trung bình) — không bao giờ chạm giới hạn. 383 nhà cung cấp AI · 150+ gói miễn phí · ~1,62B token miễn phí/tháng · 20 chiến lược định tuyến · Bắt đầu với $0. +OmniRoute — Không bao giờ ngừng lập trình. Mọi công cụ AI → 384 nhà cung cấp — 150+ miễn phí — thông qua một endpoint duy nhất. Claude Code, Codex, Cursor, Cline, Copilot & Antigravity đến Claude / GPT / Gemini MIỄN PHÍ với tính năng tự động chuyển dự phòng. Tính năng nén xếp chồng RTK + Caveman tiết kiệm 15–95% token (~89% trung bình) — không bao giờ chạm giới hạn. 384 nhà cung cấp AI · 150+ gói miễn phí · ~1,62B token miễn phí/tháng · 20 chiến lược định tuyến · Bắt đầu với $0.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | --------------------------------- | :----------------------------: | :---------: | :-------------------: | -| 🌐 Nhà cung cấp | 352 | **358** | 383 | +| 🌐 Nhà cung cấp | 352 | **358** | 384 | | 🧠 ID mô hình trò chuyện duy nhất | 1320 | **1374** | 1491 | | 🖼️ Cầu nối đa phương thức | 🆕 hình ảnh + âm thanh + video | ✓ | ✓ | | 📡 Danh mục miễn phí của Radar | 🆕 opt-in | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ Tính năng 🎯 Combo - 🌐 Nhà cung cấp + 🌐 Nhà cung cấp 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -Lời hứa — Một endpoint và 383 nhà cung cấp. Tự động chuyển dự phòng giúp duy trì định tuyến khi vẫn còn một mục tiêu khác đang hoạt động tốt. Sáu trụ cột: chuyển dự phòng linh hoạt trên 383 nhà cung cấp · tiết kiệm tối đa 95% token đối với các khối lượng công việc đủ điều kiện · bắt đầu với $0 cùng hơn 150 gói miễn phí và 52 nhà cung cấp miễn phí vĩnh viễn định kỳ/không cần khóa · 39 tích hợp CLI/agent thông qua một cấu hình · khả năng tương thích với OpenAI, Claude, Gemini và Responses API tại /v1 · các tính năng kiểm soát cho môi trường production, bao gồm bộ ngắt mạch, TLS stealth, MCP với 110 công cụ, A2A, bộ nhớ, guardrails, evals và hơn 56.000 khai báo kiểm thử tĩnh trên hơn 7.800 tệp kiểm thử được theo dõi. +Lời hứa — Một endpoint và 384 nhà cung cấp. Tự động chuyển dự phòng giúp duy trì định tuyến khi vẫn còn một mục tiêu khác đang hoạt động tốt. Sáu trụ cột: chuyển dự phòng linh hoạt trên 384 nhà cung cấp · tiết kiệm tối đa 95% token đối với các khối lượng công việc đủ điều kiện · bắt đầu với $0 cùng hơn 150 gói miễn phí và 52 nhà cung cấp miễn phí vĩnh viễn định kỳ/không cần khóa · 39 tích hợp CLI/agent thông qua một cấu hình · khả năng tương thích với OpenAI, Claude, Gemini và Responses API tại /v1 · các tính năng kiểm soát cho môi trường production, bao gồm bộ ngắt mạch, TLS stealth, MCP với 110 công cụ, A2A, bộ nhớ, guardrails, evals và hơn 56.000 khai báo kiểm thử tĩnh trên hơn 7.800 tệp kiểm thử được theo dõi.

@@ -510,7 +510,7 @@ Tất cả **20** chiến lược — kết hợp tùy ý theo từng bước c -Điều gì làm OmniRoute nổi bật — ảnh chụp nhanh các tính năng tại một thời điểm so với 9router, OpenRouter, CLIProxyAPI và LiteLLM trên 13 khả năng. OmniRoute: 383 nhà cung cấp, tích hợp sẵn hơn 150 gói miễn phí, 20 chiến lược định tuyến, tính năng nén token với 12 engine, máy chủ MCP tích hợp sẵn với 110 công cụ, giao thức tác nhân A2A, bộ nhớ liên tục, các cơ chế bảo vệ, tác nhân đám mây, tính năng ẩn dấu vân tay TLS, Desktop/Termux/PWA và 67 ngôn ngữ giao diện i18n. OmniRoute được cấp phép MIT và có thể tự lưu trữ. Khả năng và số lượng của các đối thủ có thể thay đổi; xem phương pháp luận được liên kết. +Điều gì làm OmniRoute nổi bật — ảnh chụp nhanh các tính năng tại một thời điểm so với 9router, OpenRouter, CLIProxyAPI và LiteLLM trên 13 khả năng. OmniRoute: 384 nhà cung cấp, tích hợp sẵn hơn 150 gói miễn phí, 20 chiến lược định tuyến, tính năng nén token với 12 engine, máy chủ MCP tích hợp sẵn với 110 công cụ, giao thức tác nhân A2A, bộ nhớ liên tục, các cơ chế bảo vệ, tác nhân đám mây, tính năng ẩn dấu vân tay TLS, Desktop/Termux/PWA và 67 ngôn ngữ giao diện i18n. OmniRoute được cấp phép MIT và có thể tự lưu trữ. Khả năng và số lượng của các đối thủ có thể thay đổi; xem phương pháp luận được liên kết. 📊 Phương pháp luận đầy đủ & chi tiết theo từng tính năng so với 9router, OpenRouter, CLIProxyAPI & LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar hoạt động theo cơ chế chọn tham gia và chỉ sử dụng GET. - **🖼️ Endpoint mới** — `/v1/ocr` (Mistral OCR) và `/v1/audio/translations` (kiểu Whisper) hoàn thiện nhóm chức năng media. → [Tham chiếu API](docs/reference/API_REFERENCE.md) - **🎨 Tạo ảnh / video / âm thanh** — một API cho media: video xAI Grok Imagine & Novita AI, ComfyUI, Magnific, Adobe Firefly, Segmind và các nhà cung cấp giọng nói như ElevenLabs. → [Tham chiếu API](docs/reference/API_REFERENCE.md) - **🌍 Triển khai & vận hành** — `basePath` reverse-proxy, tự động phát hiện ngôn ngữ trình duyệt, theo dõi thiết bị theo từng khóa, trust MITM không cần quyền root, bản địa hóa zh-TW. → [Môi trường](docs/reference/ENVIRONMENT.md) -- **🤝 Nhiều nhà cung cấp & agent hơn** — agent đám mây (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) với trình duyệt + đăng nhập OAuth, card Ollama hạng nhất, Claude Opus 5 & Sonnet 5, hợp tác chính thức với Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… và **catalog 383 nhà cung cấp** được làm mới. → [Nhà cung cấp](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Nhiều nhà cung cấp & agent hơn** — agent đám mây (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) với trình duyệt + đăng nhập OAuth, card Ollama hạng nhất, Claude Opus 5 & Sonnet 5, hợp tác chính thức với Kimi (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… và **catalog 384 nhà cung cấp** được làm mới. → [Nhà cung cấp](docs/reference/PROVIDER_REFERENCE.md) - **📡 Minh bạch định tuyến** — mọi phản hồi đều chứa header `X-OmniRoute-Decision` nêu chiến lược/nhà cung cấp/độ trễ đã phục vụ phản hồi, chiến lược combo `cache-optimized` mới cùng yếu tố route `cacheAffinity` của Auto-Combo đưa các yêu cầu lặp lại trở về kết nối đang giữ prefix đã cache, và endpoint chỉ đọc `/v1/auto-combo/{channel}/candidates` hiển thị pool ứng viên trực tiếp của một channel `auto/*`. → [Auto-Combo](docs/routing/AUTO-COMBO.md) - **⚡ Hiệu năng & hạ tầng cục bộ** — Redis cục bộ chỉ với một cú nhấp chuột, các trình triển khai relay cho Cloudflare Workers / Deno Deploy, Bifrost & Mux dưới dạng các dịch vụ nhúng được giám sát. → [Dịch vụ nhúng](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Cũng được tích hợp sẵn** — framework plugin + marketplace, các framework kỹ năng Omni/Agent/GitHub, tích hợp vault Obsidian (22 công cụ MCP), API Batch & Files tương thích với OpenAI, bộ nhớ đệm phản hồi ngữ nghĩa, tính năng game hóa với bảng xếp hạng, tính năng phát hiện tác nhân ACP (15 tác nhân tích hợp sẵn), xuất nhật ký theo lịch sang BigQuery, fan-out song song đa mô hình `auto/chaos`, cầu nối bot Telegram, trình quản lý phiên bản trong ứng dụng và bảng xếp hạng nhà cung cấp miễn phí LMArena-ELO. → [Tài liệu](docs/README.md) @@ -689,11 +689,11 @@ khỏi lịch sử shell của bạn. → [Tích hợp CLI](docs/guides/CLI-INTE
-## 🌐 383 Nhà cung cấp AI — 153 được đánh dấu miễn phí trong danh mục +## 🌐 384 Nhà cung cấp AI — 153 được đánh dấu miễn phí trong danh mục
-> **383 nhà cung cấp đã đăng ký** trên các bộ sưu tập chuẩn về trò chuyện, phương tiện, tìm kiếm, cục bộ, tác nhân đám mây và hệ thống, trong đó **153 nhà cung cấp có siêu dữ liệu khám phá `hasFree: true`**. Sổ đăng ký mô hình trò chuyện bao phủ **239 nhà cung cấp / 3,059 cặp nhà cung cấp-mô hình riêng biệt / 1,475 ID mô hình thô**; danh mục ngân sách miễn phí riêng biệt có **471 dòng cho từng mô hình**, **34 nhóm định kỳ** và **52 nhà cung cấp miễn phí mãi mãi định kỳ/không cần khóa**. Đây là các mẫu số khác nhau theo chủ đích; định nghĩa và các phép tính đã loại trùng theo nhóm nằm trong [Tài liệu tham khảo về nhà cung cấp](docs/reference/PROVIDER_REFERENCE.md) và [Các gói miễn phí](docs/reference/FREE_TIERS.md). +> **384 nhà cung cấp đã đăng ký** trên các bộ sưu tập chuẩn về trò chuyện, phương tiện, tìm kiếm, cục bộ, tác nhân đám mây và hệ thống, trong đó **153 nhà cung cấp có siêu dữ liệu khám phá `hasFree: true`**. Sổ đăng ký mô hình trò chuyện bao phủ **239 nhà cung cấp / 3,059 cặp nhà cung cấp-mô hình riêng biệt / 1,475 ID mô hình thô**; danh mục ngân sách miễn phí riêng biệt có **471 dòng cho từng mô hình**, **34 nhóm định kỳ** và **52 nhà cung cấp miễn phí mãi mãi định kỳ/không cần khóa**. Đây là các mẫu số khác nhau theo chủ đích; định nghĩa và các phép tính đã loại trùng theo nhóm nằm trong [Tài liệu tham khảo về nhà cung cấp](docs/reference/PROVIDER_REFERENCE.md) và [Các gói miễn phí](docs/reference/FREE_TIERS.md).
@@ -1295,7 +1295,7 @@ Số liệu chuẩn vào 2026-10-18: **1.976 video duy nhất** · **20.220.289 Môi trường chạyNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 Ngôn ngữTypeScript 6.0 — 100% TypeScript trong src/ và open-sse/ (không có any trong phần lõi kể từ v2.0) FrameworkNext.js 16 + React 19 + Tailwind CSS 4 - Cơ sở dữ liệubetter-sqlite3 (SQLite, ghi nhật ký WAL) + LowDB (JSON kế thừa) — 147 mô-đun miền, 230 bản di chuyển + Cơ sở dữ liệubetter-sqlite3 (SQLite, ghi nhật ký WAL) + LowDB (JSON kế thừa) — 147 mô-đun miền, 232 bản di chuyển Bộ nhớTìm kiếm toàn văn SQLite FTS5 + các vector embedding được lượng tử hóa int8, suy giảm có kiểu Lược đồZod 4 — xác thực I/O công cụ MCP + hợp đồng API Giao thứcMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/vi/llm.txt b/docs/i18n/vi/llm.txt index a54a5a4a75f..6992d2fef42 100644 --- a/docs/i18n/vi/llm.txt +++ b/docs/i18n/vi/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/yo/README.md b/docs/i18n/yo/README.md index acb11da061c..6389cf2dc16 100644 --- a/docs/i18n/yo/README.md +++ b/docs/i18n/yo/README.md @@ -15,7 +15,7 @@ **Ẹ̀dá [Diego Souza](https://github.com/diegosouzapw)** -OmniRoute — Má ṣe dá kóòdù dúró láéláé. Gbogbo irinṣẹ́ AI → àwọn olùpèsè 383 — 150+ ọ̀fẹ́ — nípasẹ̀ endpoint kan. Claude Code, Codex, Cursor, Cline, Copilot àti Antigravity sí Claude / GPT / Gemini Ọ̀fẹ́ pẹ̀lú àyípadà aládàáṣiṣẹ́. Ìfipamọ́ tó 15–95% àwọn token (~89% ní àpapọ̀) pẹ̀lú ìfọ́kànsí RTK + Caveman — má ṣe dé ààlà láéláé. Àwọn olùpèsè AI 383 · àwọn ipele ọ̀fẹ́ 150+ · ~1.62B àwọn token ọ̀fẹ́/bí oṣù · àwọn ọgbọ́n ìdarí 20 · $0 láti bẹ̀rẹ̀. +OmniRoute — Má ṣe dá kóòdù dúró láéláé. Gbogbo irinṣẹ́ AI → àwọn olùpèsè 384 — 150+ ọ̀fẹ́ — nípasẹ̀ endpoint kan. Claude Code, Codex, Cursor, Cline, Copilot àti Antigravity sí Claude / GPT / Gemini Ọ̀fẹ́ pẹ̀lú àyípadà aládàáṣiṣẹ́. Ìfipamọ́ tó 15–95% àwọn token (~89% ní àpapọ̀) pẹ̀lú ìfọ́kànsí RTK + Caveman — má ṣe dé ààlà láéláé. Àwọn olùpèsè AI 384 · àwọn ipele ọ̀fẹ́ 150+ · ~1.62B àwọn token ọ̀fẹ́/bí oṣù · àwọn ọgbọ́n ìdarí 20 · $0 láti bẹ̀rẹ̀.
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------------------------- | :--------------------: | :---------: | :--------------------: | -| 🌐 Àwọn olupèsè | 352 | **358** | 383 | +| 🌐 Àwọn olupèsè | 352 | **358** | 384 | | 🧠 Àwọn ID modèle ìjíròrò aláìlékànsí | 1320 | **1374** | 1491 | | 🖼️ Afárá ìbáṣepọ̀ | 🆕 ìran + ohun + fídíò | ✓ | ✓ | | 📡 Katalogi ọ̀fẹ́ Radar | 🆕 ìforúkọsílẹ̀ àṣàyàn | ✓ | ✓ | @@ -113,7 +113,7 @@ ⚙️ Àwọn Ẹya 🎯 Àwọn Combo - 🌐 Àwọn olùpèsè + 🌐 Àwọn olùpèsè 🔌 CLI & MCP @@ -249,7 +249,7 @@ curl http://localhost:20128/v1/chat/completions \ -Àdéhùn náà — Endpoint kan àti àwọn olupese 383. Fallback aládàáṣiṣẹ́ máa ń jẹ́ kí ìdarí tẹ̀síwájú nígbà tí ibi-afẹ́fẹ́ mìíràn tó ní ìlera bá wà. Àwọn òpó mẹ́fà: fallback tó lágbára kọjá àwọn olupese 383 · ìfipamọ́ token tó tó 95% lórí àwọn iṣẹ́ tó yẹ · $0 láti bẹ̀rẹ̀ pẹ̀lú àwọn ìpele ọ̀fẹ́ 150+ àti àwọn olupese ọ̀fẹ́ títí láé 52 tí ń tún ara wọn ṣe/tí kò nílò kọ́kọ́rọ́ · àwọn ìṣọ̀kan CLI/agent 39 nípasẹ̀ ìṣètò kan · ìbámu OpenAI, Claude, Gemini àti Responses API ní /v1 · àwọn ìṣàkóso production pẹ̀lú circuit breakers, TLS stealth, MCP 110 tools, A2A, memory, guardrails, evals àti àwọn ìkéde ìdánwò static 56,000+ káàkiri àwọn fáìlì ìdánwò 7,800+ tí a ń tọ́pin. +Àdéhùn náà — Endpoint kan àti àwọn olupese 384. Fallback aládàáṣiṣẹ́ máa ń jẹ́ kí ìdarí tẹ̀síwájú nígbà tí ibi-afẹ́fẹ́ mìíràn tó ní ìlera bá wà. Àwọn òpó mẹ́fà: fallback tó lágbára kọjá àwọn olupese 384 · ìfipamọ́ token tó tó 95% lórí àwọn iṣẹ́ tó yẹ · $0 láti bẹ̀rẹ̀ pẹ̀lú àwọn ìpele ọ̀fẹ́ 150+ àti àwọn olupese ọ̀fẹ́ títí láé 52 tí ń tún ara wọn ṣe/tí kò nílò kọ́kọ́rọ́ · àwọn ìṣọ̀kan CLI/agent 39 nípasẹ̀ ìṣètò kan · ìbámu OpenAI, Claude, Gemini àti Responses API ní /v1 · àwọn ìṣàkóso production pẹ̀lú circuit breakers, TLS stealth, MCP 110 tools, A2A, memory, guardrails, evals àti àwọn ìkéde ìdánwò static 56,000+ káàkiri àwọn fáìlì ìdánwò 7,800+ tí a ń tọ́pin.

@@ -509,7 +509,7 @@ Gbogbo ọgbọ́n **20** náà — parapọ̀ wọn bí o ṣe fẹ́ fún ìgb -Ohun tó yà OmniRoute sọ́tọ̀ — àkótán àwọn ẹ̀yà ọjọ́ kan tí a fi wé 9router, OpenRouter, CLIProxyAPI àti LiteLLM káàkiri àwọn agbára 13. OmniRoute: àwọn olùpèsè 383, àwọn ipele ọ̀fẹ́ tí ó ju 150 lọ tí a ti ṣètò sínú rẹ̀, àwọn ọgbọ́n ìtọ́jú ipa 20, ìfipamọ́ token pẹ̀lú ẹ́ńjìnì 12, olupin MCP tí a ti ṣètò sínú rẹ̀ pẹ̀lú àwọn irinṣẹ́ 110, ìlànà aṣojú A2A, ìrántí títẹpẹlẹ, àwọn ìdènà ààbò, àwọn aṣojú awọ̀n ಮೋ, ìfipamọ́ ìtọ́ka TLS, Desktop/Termux/PWA àti àwọn èdè UI i18n 67. OmniRoute ní ìwé-aṣẹ MIT, ó sì lè jẹ́ ẹni tí a gbàlejò fúnra wa. Àwọn agbára àti iye àwọn olùjà lè yí padà; wo ìlànà tí a so mọ́ ọ. +Ohun tó yà OmniRoute sọ́tọ̀ — àkótán àwọn ẹ̀yà ọjọ́ kan tí a fi wé 9router, OpenRouter, CLIProxyAPI àti LiteLLM káàkiri àwọn agbára 13. OmniRoute: àwọn olùpèsè 384, àwọn ipele ọ̀fẹ́ tí ó ju 150 lọ tí a ti ṣètò sínú rẹ̀, àwọn ọgbọ́n ìtọ́jú ipa 20, ìfipamọ́ token pẹ̀lú ẹ́ńjìnì 12, olupin MCP tí a ti ṣètò sínú rẹ̀ pẹ̀lú àwọn irinṣẹ́ 110, ìlànà aṣojú A2A, ìrántí títẹpẹlẹ, àwọn ìdènà ààbò, àwọn aṣojú awọ̀n ಮೋ, ìfipamọ́ ìtọ́ka TLS, Desktop/Termux/PWA àti àwọn èdè UI i18n 67. OmniRoute ní ìwé-aṣẹ MIT, ó sì lè jẹ́ ẹni tí a gbàlejò fúnra wa. Àwọn agbára àti iye àwọn olùjà lè yí padà; wo ìlànà tí a so mọ́ ọ. 📊 Ìlànà kíkún àti àlàyé ẹ̀yà kọ̀ọ̀kan ní ìfiwéra pẹ̀lú 9router, OpenRouter, CLIProxyAPI àti LiteLLM → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -613,7 +613,7 @@ tẹlifíṣọ̀nù lílò, tàbí ipò ìkọ̀sílẹ̀ ìkéde abẹ́lẹ̀ - **🖼️ Àwọn endpoint tuntun** — `/v1/ocr` (Mistral OCR) àti `/v1/audio/translations` (irú Whisper) ń mú àgbègbè media pé. → [Ìtọ́kasí API](docs/reference/API_REFERENCE.md) - **🎨 Ṣíṣẹ̀dá image / video / audio** — API kan fún media: xAI Grok Imagine & Novita AI video, ComfyUI, Magnific, Adobe Firefly, Segmind, àti àwọn provider speech bí ElevenLabs. → [Ìtọ́kasí API](docs/reference/API_REFERENCE.md) - **🌍 Deployment àti ops** — reverse-proxy `basePath`, ìṣàwárí browser-language aládàáṣe, device tracking fún key kọ̀ọ̀kan, MITM trust tí kò nílò root, àti ìtumọ̀ zh-TW. → [Ayíká](docs/reference/ENVIRONMENT.md) -- **🤝 Àwọn provider àti agent míì** — cloud agents (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) pẹ̀lú browser + OAuth login, kaadi Ollama tó ní àtìlẹ́yìn kíkún, Claude Opus 5 & Sonnet 5, ìfọwọ́sowọ́pọ̀ Kimi tó jẹ́ official (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… àti **catalog provider 383** tí a tún ṣe. → [Àwọn Provider](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 Àwọn provider àti agent míì** — cloud agents (Codex Cloud, Cursor, Devin, Jules), Grok Build (xAI) pẹ̀lú browser + OAuth login, kaadi Ollama tó ní àtìlẹ́yìn kíkún, Claude Opus 5 & Sonnet 5, ìfọwọ́sowọ́pọ̀ Kimi tó jẹ́ official (Code/Web/Moonshot), Zed, Requesty, SenseNova, Yuanbao, Agnes AI… àti **catalog provider 384** tí a tún ṣe. → [Àwọn Provider](docs/reference/PROVIDER_REFERENCE.md) - **📡 Ìhàn gbangba routing** — gbogbo ìdáhùn ní header `X-OmniRoute-Decision` tó ń sọ strategy/provider/latency tó pèsè rẹ̀, combo strategy `cache-optimized` tuntun + factor `cacheAffinity` Auto-Combo máa ń darí àwọn ìbéèrè tí a tún ṣe padà sí connection tó ń di prefix tí a cache mọ́, àti endpoint `/v1/auto-combo/{channel}/candidates` tí a kà nìkan ń ṣí pool candidate gidi ti channel `auto/*` hàn. → [Auto-Combo](docs/routing/AUTO-COMBO.md). - **⚡ Ìṣiṣẹ́ agbègbè tó yára & amáyédẹrùn** — Redis agbègbè pẹ̀lú títẹ bọtìnì kan, àwọn olùgbékalẹ̀ relay Cloudflare Workers / Deno Deploy, àti Bifrost & Mux gẹ́gẹ́ bí àwọn iṣẹ́ tí a fi sínú tí a ń bójú tó. → [Àwọn Iṣẹ́ Tí A Fi Sínú](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 Àwọn míì nínú àpótí** — ìlànà plugin + ọjà, àwọn ìlànà ọgbọ́n Omni/Agent/GitHub, ìṣọ̀kan Obsidian vault (àwọn irinṣẹ́ MCP 22), àwọn API Batch & Files tó bá OpenAI mu, kásì ìdáhùn ọgbọ́n, ìṣeré pẹ̀lú àwọn tábìlì aṣáájú, ìwádìí aṣojú ACP (àwọn aṣojú tí a ṣe sínú 15), ìtajà log tí a ṣètò sí BigQuery, `auto/chaos` ìfìtẹ́lẹ̀ ọ̀pọ̀ àwòṣe ní ìfarawé, afárá bot Telegram, olùṣàkóso ẹ̀dà nínú ìṣàfilọ́lẹ̀ àti àwọn ìṣàyẹ̀wò àwọn olùpèsè ọ̀fẹ́ LMArena-ELO. → [Àkọsílẹ̀](docs/README.md) @@ -697,11 +697,11 @@ nínú ìtàn shell rẹ. → [Àwọn Ìṣọ̀kan CLI](docs/guides/CLI-INTEGR
-## 🌐 Àwọn Olùpèsè AI 383 — 153 Ti A Samisi Nínú Kátálógù Gẹ́gẹ́ Bí Ọ̀fẹ́ +## 🌐 Àwọn Olùpèsè AI 384 — 153 Ti A Samisi Nínú Kátálógù Gẹ́gẹ́ Bí Ọ̀fẹ́
-> **Àwọn olùpèsè 383 tí a forúkọsílẹ̀** káàkiri àwọn àkójọpọ̀ chat, media, search, local, cloud-agent àti system tó jẹ́ àkọ́kọ́, pẹ̀lú **153 tí ó ní metadata ìwádìí `hasFree: true`**. Àkọsílẹ̀ àwọn chat model bo **àwọn olùpèsè 239 / àwọn ìbáṣepọ̀ provider-model ọ̀tọ̀ọ̀tọ̀ 3,059 / àwọn ID model aise 1,475**; kátálógù ìná ọ̀fẹ́ tó yàtọ̀ ní **àwọn ìlà fún model 471**, **àwọn adágún tó ń tún ara wọn ṣe 34** àti **àwọn olùpèsè ọ̀fẹ́ títí láé 52 tí ó ń tún ara wọn ṣe/tí kò nílò kọ́kọ́rọ́**. Àwọn iye ìpín wọ̀nyí yàtọ̀ ní ìmọ̀lára; àwọn ìtumọ̀ àti ìṣirò tí a ti yọ ìdàpọ̀ adágún kúrò wà nínú [Ìtọ́kasí Olùpèsè](docs/reference/PROVIDER_REFERENCE.md) àti [Àwọn Ìpele Ọ̀fẹ́](docs/reference/FREE_TIERS.md). +> **Àwọn olùpèsè 384 tí a forúkọsílẹ̀** káàkiri àwọn àkójọpọ̀ chat, media, search, local, cloud-agent àti system tó jẹ́ àkọ́kọ́, pẹ̀lú **153 tí ó ní metadata ìwádìí `hasFree: true`**. Àkọsílẹ̀ àwọn chat model bo **àwọn olùpèsè 239 / àwọn ìbáṣepọ̀ provider-model ọ̀tọ̀ọ̀tọ̀ 3,059 / àwọn ID model aise 1,475**; kátálógù ìná ọ̀fẹ́ tó yàtọ̀ ní **àwọn ìlà fún model 471**, **àwọn adágún tó ń tún ara wọn ṣe 34** àti **àwọn olùpèsè ọ̀fẹ́ títí láé 52 tí ó ń tún ara wọn ṣe/tí kò nílò kọ́kọ́rọ́**. Àwọn iye ìpín wọ̀nyí yàtọ̀ ní ìmọ̀lára; àwọn ìtumọ̀ àti ìṣirò tí a ti yọ ìdàpọ̀ adágún kúrò wà nínú [Ìtọ́kasí Olùpèsè](docs/reference/PROVIDER_REFERENCE.md) àti [Àwọn Ìpele Ọ̀fẹ́](docs/reference/FREE_TIERS.md).
@@ -1303,7 +1303,7 @@ port kan, nítorí náà kò sí package CLI-nìkan lọ́tọ̀ ní báyìí. Ìpẹ̀yàNode.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 ÈdèTypeScript 6.0 — 100% TypeScript káàkiri src/ àti open-sse/ (any kò sí nínú kókó láti v2.0) Àgbékalẹ̀Next.js 16 + React 19 + Tailwind CSS 4 - Àkójọpọ̀ dátàbetter-sqlite3 (SQLite, ìwé-àkọsílẹ̀ WAL) + LowDB (JSON àtijọ́) — àwọn modúlù àgbègbè 147, àwọn ìṣíkiri 230 + Àkójọpọ̀ dátàbetter-sqlite3 (SQLite, ìwé-àkọsílẹ̀ WAL) + LowDB (JSON àtijọ́) — àwọn modúlù àgbègbè 147, àwọn ìṣíkiri 232 ÌrántíÌwádìí ọ̀rọ̀-kíkún SQLite FTS5 + àwọn ìfarahàn fákítọ̀ tí a ṣe quantize sí int8, ìparẹ́ tí a tẹ̀wé Àwọn ètò àkàwéZod 4 — ìfọwọ́sí ìwọlé/jíjáde irinṣẹ́ MCP + àwọn àdéhùn API Àwọn ìlànàMCP (stdio / HTTP / SSE) + A2A v0.3 (JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/yo/llm.txt b/docs/i18n/yo/llm.txt index d640927fbf6..0efc6d8d779 100644 --- a/docs/i18n/yo/llm.txt +++ b/docs/i18n/yo/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/zh-CN/README.md b/docs/i18n/zh-CN/README.md index c210bcae940..ff9bfaffac4 100644 --- a/docs/i18n/zh-CN/README.md +++ b/docs/i18n/zh-CN/README.md @@ -15,7 +15,7 @@ **由 [Diego Souza](https://github.com/diegosouzapw) 创建** -OmniRoute — 永不停下编程的脚步。每个 AI 工具 → 383 个提供者 — 其中 150+ 个免费 — 通过一个端点接入。Claude Code、Codex、Cursor、Cline、Copilot 和 Antigravity 接入免费的 Claude / GPT / Gemini,并支持自动故障转移。RTK + Caveman 堆叠式压缩可节省 15–95% 的令牌(平均约 89%)— 永不触及限制。383 个 AI 提供者 · 150+ 个免费层级 · 每月约 16.2 亿个免费令牌 · 20 种路由策略 · $0 起步。 +OmniRoute — 永不停下编程的脚步。每个 AI 工具 → 384 个提供者 — 其中 150+ 个免费 — 通过一个端点接入。Claude Code、Codex、Cursor、Cline、Copilot 和 Antigravity 接入免费的 Claude / GPT / Gemini,并支持自动故障转移。RTK + Caveman 堆叠式压缩可节省 15–95% 的令牌(平均约 89%)— 永不触及限制。384 个 AI 提供者 · 150+ 个免费层级 · 每月约 16.2 亿个免费令牌 · 20 种路由策略 · $0 起步。
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ------------------- | :-------------------: | :---------: | :-------------------: | -| 🌐 提供者 | 352 | **358** | 383 | +| 🌐 提供者 | 352 | **358** | 384 | | 🧠 唯一聊天模型 ID | 1320 | **1374** | 1491 | | 🖼️ 模态桥接 | 🆕 视觉 + 音频 + 视频 | ✓ | ✓ | | 📡 Radar 免费目录 | 🆕 可选加入 | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ 功能 🎯 组合 - 🌐 提供者 + 🌐 提供者 🔌 CLI & MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -我们的承诺 — 一个端点和 383 个提供者。当仍有其他健康目标可用时,自动回退会持续进行路由。六大支柱:跨 383 个提供者的高韧性回退 · 符合条件的工作负载最高可节省 95% 的令牌成本 · 从 $0 开始,提供 150+ 个免费层级以及 52 个定期提供或免密钥的永久免费提供者 · 通过一个配置支持 39 个 CLI/代理集成 · 在 /v1 提供 OpenAI、Claude、Gemini 和 Responses API 兼容性 · 生产级控制能力,包括断路器、TLS 隐身、MCP 110 个工具、A2A、记忆、护栏、评估,以及分布在 7,800+ 个受跟踪测试文件中的 56,000+ 个静态测试声明。 +我们的承诺 — 一个端点和 384 个提供者。当仍有其他健康目标可用时,自动回退会持续进行路由。六大支柱:跨 384 个提供者的高韧性回退 · 符合条件的工作负载最高可节省 95% 的令牌成本 · 从 $0 开始,提供 150+ 个免费层级以及 52 个定期提供或免密钥的永久免费提供者 · 通过一个配置支持 39 个 CLI/代理集成 · 在 /v1 提供 OpenAI、Claude、Gemini 和 Responses API 兼容性 · 生产级控制能力,包括断路器、TLS 隐身、MCP 110 个工具、A2A、记忆、护栏、评估,以及分布在 7,800+ 个受跟踪测试文件中的 56,000+ 个静态测试声明。

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRoute 的独特之处——一份截至特定日期的功能快照,对比了 9router、OpenRouter、CLIProxyAPI 和 LiteLLM 在 13 项能力上的表现。OmniRoute:383 个提供者、内置 150 多个免费层级、20 种路由策略、12 引擎令牌压缩、内置 MCP 服务器(含 110 个工具)、A2A 代理协议、持久化记忆、护栏、云代理、TLS 指纹隐身、Desktop/Termux/PWA,以及支持 67 种国际化语言环境的 UI。OmniRoute 采用 MIT 许可证并支持自托管。竞争对手的能力和数量可能会发生变化;请参阅链接的方法说明。 +OmniRoute 的独特之处——一份截至特定日期的功能快照,对比了 9router、OpenRouter、CLIProxyAPI 和 LiteLLM 在 13 项能力上的表现。OmniRoute:384 个提供者、内置 150 多个免费层级、20 种路由策略、12 引擎令牌压缩、内置 MCP 服务器(含 110 个工具)、A2A 代理协议、持久化记忆、护栏、云代理、TLS 指纹隐身、Desktop/Termux/PWA,以及支持 67 种国际化语言环境的 UI。OmniRoute 采用 MIT 许可证并支持自托管。竞争对手的能力和数量可能会发生变化;请参阅链接的方法说明。 📊 与 9router、OpenRouter、CLIProxyAPI 和 LiteLLM 对比的完整方法说明及各功能详情 → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar 需主动选择启用,且仅使用 GET 请求。OmniRoute 客户端不 - **🖼️ 新端点** — `/v1/ocr`(Mistral OCR)和 `/v1/audio/translations`(Whisper 风格)完善了媒体功能。→ [API 参考](docs/reference/API_REFERENCE.md) - **🎨 图像 / 视频 / 音频生成** — 通过一个 API 处理媒体:xAI Grok Imagine 与 Novita AI 视频、ComfyUI、Magnific、Adobe Firefly、Segmind,以及 ElevenLabs 等语音提供者。→ [API 参考](docs/reference/API_REFERENCE.md) - **🌍 部署与运维** — 反向代理 `basePath`、浏览器语言自动检测、按密钥进行设备跟踪、无 root 的 MITM 信任,以及 zh-TW 本地化。→ [环境变量](docs/reference/ENVIRONMENT.md) -- **🤝 更多提供者与代理** — 云端代理(Codex Cloud、Cursor、Devin、Jules)、支持浏览器 + OAuth 登录的 Grok Build(xAI)、一等支持的 Ollama 卡片、Claude Opus 5 与 Sonnet 5、Kimi 官方合作伙伴(Code/Web/Moonshot)、Zed、Requesty、SenseNova、Yuanbao、Agnes AI……以及更新后的 **383 个提供者目录**。→ [提供者](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 更多提供者与代理** — 云端代理(Codex Cloud、Cursor、Devin、Jules)、支持浏览器 + OAuth 登录的 Grok Build(xAI)、一等支持的 Ollama 卡片、Claude Opus 5 与 Sonnet 5、Kimi 官方合作伙伴(Code/Web/Moonshot)、Zed、Requesty、SenseNova、Yuanbao、Agnes AI……以及更新后的 **384 个提供者目录**。→ [提供者](docs/reference/PROVIDER_REFERENCE.md) - **📡 路由透明度** — 每个响应都携带 `X-OmniRoute-Decision` 标头,标明为其提供服务的策略 / 提供者 / 延迟;新增的 `cache-optimized` 组合策略和 Auto-Combo `cacheAffinity` 因子会将重复请求路由回持有缓存前缀的连接;只读端点 `/v1/auto-combo/{channel}/candidates` 则公开 `auto/*` 通道的实时候选池。→ [自动组合](docs/routing/AUTO-COMBO.md)。 - **⚡ 本地性能与基础设施** — 一键启动本地 Redis、Cloudflare Workers / Deno Deploy 中继部署器,以及作为受监管嵌入式服务运行的 Bifrost 和 Mux。→ [嵌入式服务](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 开箱即用** — 插件框架与市场、Omni/Agent/GitHub 技能框架、Obsidian vault 集成(22 个 MCP 工具)、兼容 OpenAI 的 Batch 和 Files API、语义响应缓存、带排行榜的游戏化机制、ACP 代理发现(15 个内置代理)、定时将日志导出到 BigQuery、`auto/chaos` 并行多模型扇出、Telegram 机器人桥接、应用内版本管理器,以及 LMArena-ELO 免费提供者排名。→ [文档](docs/README.md) @@ -689,11 +689,11 @@ Shell 历史记录中。→ [CLI 集成](docs/guides/CLI-INTEGRATIONS.md)
-## 🌐 383 个 AI 提供者 — 目录标记为免费的 153 个 +## 🌐 384 个 AI 提供者 — 目录标记为免费的 153 个
-> **383 个已注册的提供者**分布于规范聊天、媒体、搜索、本地、云端代理和系统集合中,其中包括 **153 个携带 `hasFree: true` 发现元数据的提供者**。聊天模型注册表涵盖 **239 个提供者 / 3,059 个不同的提供者-模型对 / 1,475 个原始模型 ID**;独立的免费额度目录包含 **471 行按模型记录**、**34 个周期性额度池**以及 **52 个周期性/无需密钥的永久免费提供者**。这些分母经过有意区分;定义以及按额度池去重的计算方式请参阅[提供者参考](docs/reference/PROVIDER_REFERENCE.md)和[免费层级](docs/reference/FREE_TIERS.md)。 +> **384 个已注册的提供者**分布于规范聊天、媒体、搜索、本地、云端代理和系统集合中,其中包括 **153 个携带 `hasFree: true` 发现元数据的提供者**。聊天模型注册表涵盖 **239 个提供者 / 3,059 个不同的提供者-模型对 / 1,475 个原始模型 ID**;独立的免费额度目录包含 **471 行按模型记录**、**34 个周期性额度池**以及 **52 个周期性/无需密钥的永久免费提供者**。这些分母经过有意区分;定义以及按额度池去重的计算方式请参阅[提供者参考](docs/reference/PROVIDER_REFERENCE.md)和[免费层级](docs/reference/FREE_TIERS.md)。
@@ -1295,7 +1295,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 也会跳过 运行时Node.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 语言TypeScript 6.0 — src/ 和 open-sse/ 中 100% 使用 TypeScript(自 v2.0 起核心代码中没有 any) 框架Next.js 16 + React 19 + Tailwind CSS 4 - 数据库better-sqlite3(SQLite,WAL 日志记录)+ LowDB(JSON legacy)— 147 个领域模块,230 个迁移 + 数据库better-sqlite3(SQLite,WAL 日志记录)+ LowDB(JSON legacy)— 147 个领域模块,232 个迁移 内存SQLite FTS5 全文搜索 + int8 量化向量嵌入,类型化衰减 架构Zod 4 — MCP 工具 I/O 验证 + API 契约 协议MCP(stdio / HTTP / SSE)+ A2A v0.3(JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/zh-CN/llm.txt b/docs/i18n/zh-CN/llm.txt index 22f7d7bdf72..5d51d028ac0 100644 --- a/docs/i18n/zh-CN/llm.txt +++ b/docs/i18n/zh-CN/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/i18n/zh-TW/README.md b/docs/i18n/zh-TW/README.md index c3e4be27d2a..964f13b1758 100644 --- a/docs/i18n/zh-TW/README.md +++ b/docs/i18n/zh-TW/README.md @@ -15,7 +15,7 @@ **由 [Diego Souza](https://github.com/diegosouzapw) 建立** -OmniRoute — 永不停下寫程式的腳步。每個 AI 工具 → 383 個提供者 — 其中 150+ 個免費 — 透過單一端點。Claude Code、Codex、Cursor、Cline、Copilot 與 Antigravity 皆可連接至免費的 Claude / GPT / Gemini,並具備自動備援。RTK + Caveman 堆疊式壓縮可節省 15–95% 的 token(平均約 89%)— 永遠不會觸及限制。383 個 AI 提供者 · 150+ 個免費方案 · 每月約 16.2 億個免費 token · 20 種路由策略 · $0 即可開始。 +OmniRoute — 永不停下寫程式的腳步。每個 AI 工具 → 384 個提供者 — 其中 150+ 個免費 — 透過單一端點。Claude Code、Codex、Cursor、Cline、Copilot 與 Antigravity 皆可連接至免費的 Claude / GPT / Gemini,並具備自動備援。RTK + Caveman 堆疊式壓縮可節省 15–95% 的 token(平均約 89%)— 永遠不會觸及限制。384 個 AI 提供者 · 150+ 個免費方案 · 每月約 16.2 億個免費 token · 20 種路由策略 · $0 即可開始。
@@ -73,7 +73,7 @@ | | v3.8.50 | **v3.8.51** | `v3.8.54+` | | ----------------------- | :-------------------: | :---------: | :-------------------: | -| 🌐 提供者 | 352 | **358** | 383 | +| 🌐 提供者 | 352 | **358** | 384 | | 🧠 唯一聊天模型 ID | 1320 | **1374** | 1491 | | 🖼️ 模態橋接 | 🆕 視覺 + 音訊 + 影片 | ✓ | ✓ | | 📡 Radar 免費目錄 | 🆕 選擇加入 | ✓ | ✓ | @@ -114,7 +114,7 @@ ⚙️ 功能 🎯 組合 - 🌐 提供者 + 🌐 提供者 🔌 CLI 與 MCP @@ -250,7 +250,7 @@ curl http://localhost:20128/v1/chat/completions \ -願景 — 單一端點和 383 個提供者。只要仍有其他健康目標可用,自動備援就會持續進行路由。六大支柱:跨 383 個提供者的韌性備援 · 符合資格的工作負載最多可節省 95% 的權杖 · 150+ 個免費方案及 52 個定期提供/免金鑰永久免費的提供者,$0 即可開始 · 透過單一設定支援 39 個 CLI/代理程式整合 · 在 /v1 相容於 OpenAI、Claude、Gemini 和 Responses API · 生產環境控制項,包括斷路器、TLS 隱匿、MCP 110 個工具、A2A、記憶體、防護措施、評估,以及分布於 7,800+ 個受追蹤測試檔案中的 56,000+ 個靜態測試宣告。 +願景 — 單一端點和 384 個提供者。只要仍有其他健康目標可用,自動備援就會持續進行路由。六大支柱:跨 384 個提供者的韌性備援 · 符合資格的工作負載最多可節省 95% 的權杖 · 150+ 個免費方案及 52 個定期提供/免金鑰永久免費的提供者,$0 即可開始 · 透過單一設定支援 39 個 CLI/代理程式整合 · 在 /v1 相容於 OpenAI、Claude、Gemini 和 Responses API · 生產環境控制項,包括斷路器、TLS 隱匿、MCP 110 個工具、A2A、記憶體、防護措施、評估,以及分布於 7,800+ 個受追蹤測試檔案中的 56,000+ 個靜態測試宣告。

@@ -510,7 +510,7 @@ curl http://localhost:20128/v1/chat/completions \ -OmniRoute 的獨特之處 — 與 9router、OpenRouter、CLIProxyAPI 和 LiteLLM 在 13 項功能上的當期功能快照比較。OmniRoute:383 個提供者、內建 150 多個免費層級、20 種路由策略、12 引擎權杖壓縮、內建支援 110 個工具的 MCP 伺服器、A2A 代理程式通訊協定、持久記憶體、安全防護、雲端代理程式、TLS 指紋隱匿、Desktop/Termux/PWA,以及支援 67 種語言的 i18n UI。OmniRoute 採用 MIT 授權並可自行託管。競爭對手的功能與數量可能會變更;請參閱連結的方法論。 +OmniRoute 的獨特之處 — 與 9router、OpenRouter、CLIProxyAPI 和 LiteLLM 在 13 項功能上的當期功能快照比較。OmniRoute:384 個提供者、內建 150 多個免費層級、20 種路由策略、12 引擎權杖壓縮、內建支援 110 個工具的 MCP 伺服器、A2A 代理程式通訊協定、持久記憶體、安全防護、雲端代理程式、TLS 指紋隱匿、Desktop/Termux/PWA,以及支援 67 種語言的 i18n UI。OmniRoute 採用 MIT 授權並可自行託管。競爭對手的功能與數量可能會變更;請參閱連結的方法論。 📊 完整方法論與相較於 9router、OpenRouter、CLIProxyAPI 和 LiteLLM 的各功能詳細資訊 → [`docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md`](docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md) @@ -605,7 +605,7 @@ Radar 採選擇加入制,且僅使用 GET。OmniRoute 用戶端不會上傳提 - **🖼️ 新端點** — `/v1/ocr`(Mistral OCR)與 `/v1/audio/translations`(Whisper 風格)補足媒體功能範圍。→ [API 參考](docs/reference/API_REFERENCE.md) - **🎨 圖像/影片/音訊生成** — 以單一 API 提供媒體功能:xAI Grok Imagine 與 Novita AI 影片、ComfyUI、Magnific、Adobe Firefly、Segmind,以及 ElevenLabs 等語音提供者。→ [API 參考](docs/reference/API_REFERENCE.md) - **🌍 部署與作業** — 反向代理 `basePath`、瀏覽器語言自動偵測、每金鑰裝置追蹤、免 root 的 MITM 信任,以及 zh-TW 在地化。→ [環境](docs/reference/ENVIRONMENT.md) -- **🤝 更多提供者與代理** — 雲端代理(Codex Cloud、Cursor、Devin、Jules)、具備瀏覽器+OAuth 登入功能的 Grok Build(xAI)、一級支援卡片的 Ollama、Claude Opus 5 與 Sonnet 5、Kimi 官方合作夥伴(Code/Web/Moonshot)、Zed、Requesty、SenseNova、Yuanbao、Agnes AI……以及更新後的 **383 個提供者目錄**。→ [提供者](docs/reference/PROVIDER_REFERENCE.md) +- **🤝 更多提供者與代理** — 雲端代理(Codex Cloud、Cursor、Devin、Jules)、具備瀏覽器+OAuth 登入功能的 Grok Build(xAI)、一級支援卡片的 Ollama、Claude Opus 5 與 Sonnet 5、Kimi 官方合作夥伴(Code/Web/Moonshot)、Zed、Requesty、SenseNova、Yuanbao、Agnes AI……以及更新後的 **384 個提供者目錄**。→ [提供者](docs/reference/PROVIDER_REFERENCE.md) - **📡 路由透明度** — 每個回應都會攜帶 `X-OmniRoute-Decision` 標頭,指出提供服務的策略/提供者/延遲;新增的 `cache-optimized` 組合策略,以及 Auto-Combo 的 `cacheAffinity` 因子,會將重複請求重新路由至持有快取前綴的連線;唯讀的 `/v1/auto-combo/{channel}/candidates` 端點則會公開 `auto/*` 頻道的即時候選池。→ [Auto-Combo](docs/routing/AUTO-COMBO.md)。 - **⚡ 本機效能與基礎架構** — 一鍵部署本機 Redis、Cloudflare Workers / Deno Deploy 中繼部署工具,以及由 Bifrost 與 Mux 擔任受監督的嵌入式服務。→ [嵌入式服務](docs/frameworks/EMBEDDED-SERVICES.md) - **🧩 內建功能** — 外掛框架與市集、Omni/Agent/GitHub 技能框架、Obsidian vault 整合(22 個 MCP 工具)、相容於 OpenAI 的 Batch 與 Files API、語意回應快取、具排行榜的遊戲化功能、ACP 代理程式探索(15 個內建代理程式)、排程將日誌匯出至 BigQuery、`auto/chaos` 平行多模型扇出、Telegram 機器人橋接器、應用程式內版本管理器,以及 LMArena-ELO 免費提供者排名。→ [文件](docs/README.md) @@ -689,11 +689,11 @@ Shell 歷程記錄中。→ [CLI 整合](docs/guides/CLI-INTEGRATIONS.md)
-## 🌐 383 個 AI 提供者 — 153 個標記為免費 +## 🌐 384 個 AI 提供者 — 153 個標記為免費
-> **383 個已註冊的提供者** 分布於標準聊天、媒體、搜尋、本機、雲端代理程式和系統集合,其中包括 **153 個帶有 `hasFree: true` 探索中繼資料的提供者**。聊天模型註冊表涵蓋 **239 個提供者 / 3,059 個不同的提供者-模型配對 / 1,475 個原始模型 ID**;獨立的免費額度目錄則有 **471 個逐模型資料列**、**34 個週期性配額池**,以及 **52 個週期性/免金鑰的永久免費提供者**。這些分母是刻意區分的;定義與去除配額池重複項目的計算方式,請參閱[提供者參考](docs/reference/PROVIDER_REFERENCE.md)和[免費方案](docs/reference/FREE_TIERS.md)。 +> **384 個已註冊的提供者** 分布於標準聊天、媒體、搜尋、本機、雲端代理程式和系統集合,其中包括 **153 個帶有 `hasFree: true` 探索中繼資料的提供者**。聊天模型註冊表涵蓋 **239 個提供者 / 3,059 個不同的提供者-模型配對 / 1,475 個原始模型 ID**;獨立的免費額度目錄則有 **471 個逐模型資料列**、**34 個週期性配額池**,以及 **52 個週期性/免金鑰的永久免費提供者**。這些分母是刻意區分的;定義與去除配額池重複項目的計算方式,請參閱[提供者參考](docs/reference/PROVIDER_REFERENCE.md)和[免費方案](docs/reference/FREE_TIERS.md)。
@@ -1295,7 +1295,7 @@ OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 也會略過 執行環境Node.js 22.x / 24.x LTS — >=22.22.2 <23 || >=24.0.0 <27 語言TypeScript 6.0 — 100% TypeScript 涵蓋 src/ 與 open-sse/(自 v2.0 起核心中沒有任何 any) 框架Next.js 16 + React 19 + Tailwind CSS 4 - 資料庫better-sqlite3(SQLite、WAL 日誌記錄)+ LowDB(JSON 舊版)— 147 個領域模組、230 次遷移 + 資料庫better-sqlite3(SQLite、WAL 日誌記錄)+ LowDB(JSON 舊版)— 147 個領域模組、232 次遷移 記憶體SQLite FTS5 全文搜尋 + int8 量化向量嵌入、具類型衰減 結構描述Zod 4 — MCP 工具 I/O 驗證 + API 合約 通訊協定MCP(stdio / HTTP / SSE)+ A2A v0.3(JSON-RPC 2.0 + SSE) diff --git a/docs/i18n/zh-TW/llm.txt b/docs/i18n/zh-TW/llm.txt index 94878f82fe2..ed42f3850fb 100644 --- a/docs/i18n/zh-TW/llm.txt +++ b/docs/i18n/zh-TW/llm.txt @@ -4,7 +4,7 @@ --- -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -18,7 +18,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -128,7 +128,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -281,7 +281,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -393,7 +393,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -437,7 +437,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -478,7 +478,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/docs/reference/PROVIDER_REFERENCE.md b/docs/reference/PROVIDER_REFERENCE.md index b5311702e8e..8221ca4eed3 100644 --- a/docs/reference/PROVIDER_REFERENCE.md +++ b/docs/reference/PROVIDER_REFERENCE.md @@ -10,7 +10,7 @@ lastUpdated: 2026-11-09 > Regenerate with: `npm run gen:provider-reference` > **Last generated:** 2026-11-09 -Total providers: **383**. See category breakdown below. +Total providers: **384**. See category breakdown below. ## Categories @@ -124,7 +124,7 @@ Use the dashboard at `/dashboard/providers` to enable, configure, and test each | `zai-web` | `zw` | Z.ai Web | Web cookie | [link](https://chat.z.ai) | Copy the "token" value from chat.z.ai → DevTools → Application → Local Storage. Do not copy cookies; OmniRoute handles the per-request CAPTCHA through its browser transport. | — | | `zenmux-free` | `zmf` | ZenMux Free (Web) | Web cookie | [link](https://zenmux.ai) | Login at zenmux.ai, then export all cookies using EditThisCookie or Cookie-Editor and paste the full Cookie header string here. Refresh every ~30 days. | — | -## API Key Providers (paid / paid-with-free-credits) (257) +## API Key Providers (paid / paid-with-free-credits) (258) | ID | Alias | Name | Tags | Website | Notes | |----|-------|------|------|---------|-------| @@ -303,6 +303,7 @@ Use the dashboard at `/dashboard/providers` to enable, configure, and test each | `opper` | `opper` | Opper | API key, aggregator | [link](https://opper.ai) | — | | `orcarouter` | `orcarouter` | OrcaRouter | API key | [link](https://www.orcarouter.ai) | — | | `ovhcloud` | `ovh` | OVHcloud AI | API key | [link](https://www.ovhcloud.com) | — | +| `paraloncloud` | `paralon` | ParalonCloud | API key | [link](https://paraloncloud.com) | — | | `perplexity` | `pplx` | Perplexity | API key | [link](https://www.perplexity.ai) | — | | `perplexity-agent` | `pplx-agent` | Perplexity Agent | API key | [link](https://www.perplexity.ai) | Use your Perplexity API key. OmniRoute routes Agent API model IDs through Perplexity's Responses-compatible endpoint. | | `piapi` | `pi` | PiAPI | API key, aggregator | [link](https://piapi.ai) | — | diff --git a/llm.txt b/llm.txt index 8c2714eb4ed..7d795286861 100644 --- a/llm.txt +++ b/llm.txt @@ -1,6 +1,6 @@ # OmniRoute -> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 383 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. +> OmniRoute is a free, open-source AI Gateway that acts as a universal API proxy for multi-provider LLMs. It provides smart routing, automatic fallback, load balancing, and format translation across 384 AI providers — all through a single OpenAI-compatible endpoint. Includes a built-in MCP Server (110 tools), A2A v0.3 protocol, Memory/Skills systems, Cloud Agents (codex, cursor, devin, jules), Guardrails framework, and an Electron desktop app. ## Overview @@ -14,7 +14,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo - **Runtime:** Node.js `>=22.22.2 <23 || >=24.0.0 <27`, ES Modules (`"type": "module"`) - **Framework:** Next.js 16 (App Router) with TypeScript 6 -- **Database:** SQLite via better-sqlite3 (local, zero-config, 230 migrations) +- **Database:** SQLite via better-sqlite3 (local, zero-config, 232 migrations) - **State management:** Zustand (client), SQLite (server persistence) - **UI:** React 19, Tailwind CSS 4, Recharts for analytics, @lobehub/icons for 130+ provider SVG icons - **Auth:** OAuth 2.0 (PKCE) for providers, bcrypt for local user auth @@ -124,7 +124,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo │ │ │ ├── secrets.ts # Secrets management │ │ │ ├── stateReset.ts # State reset utilities │ │ │ ├── migrationRunner.ts # Schema migration runner -│ │ │ └── migrations/ # 230 versioned SQL migration files +│ │ │ └── migrations/ # 232 versioned SQL migration files │ │ ├── evals/ # Eval runner and scheduler │ │ ├── memory/ # Persistent conversational memory │ │ │ ├── extraction.ts # Memory extraction from conversations @@ -277,7 +277,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo ## Key Features (v3.8.54) ### Core Proxy -- **383 AI providers** with automatic format translation +- **384 AI providers** with automatic format translation - **Provider categories**: Free (90+ free tiers), OAuth, API Key, Self-Hosted, Custom (OpenAI/Anthropic-compatible) - **20 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, quota-weighted, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline - **4-tier fallback**: Subscription → API Key → Cheap → Free @@ -389,7 +389,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 8. **ProviderIcon component:** Unified icon system using `@lobehub/icons` (130+ SVG) with PNG fallback and generic icon fallback chain. Used on providers, dashboard, and agents pages. -9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 230 SQL migrations. +9. **DB architecture:** `localDb.ts` is a re-export layer only — real logic lives in 147 `src/lib/db/` modules with 232 SQL migrations. 10. **Upstream headers:** Custom headers merged in executors after default auth; same header name replaces executor value. Forbidden header names in `src/shared/constants/upstreamHeaders.ts`. @@ -433,7 +433,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool 4. **Environment variables:** All configuration is in `.env` (from `.env.example`). Key vars: `PORT`, `NEXT_PUBLIC_BASE_URL`, `API_KEY`, `ADMIN_PASSWORD`. -5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 230 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. +5. **Database layer:** Operations go through `src/lib/db/` modules (147 domain-specific files, 232 migrations). `localDb.ts` is re-exports only — add new functions to the proper `db/*.ts` module. 6. **Tests** use Node.js built-in test runner + Vitest. Run `npm test`. Vitest for MCP/autoCombo (`npm run test:vitest`). Playwright for E2E (`npm run test:e2e`). Coverage gate: ratchet vs `quality-baseline.json`, absolute floor 60% statements/lines/functions/branches. @@ -474,7 +474,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool ## v3.8.x Highlights -- **383-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add +- **384-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add - **20 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay` - **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules) diff --git a/open-sse/config/embeddingRegistry.ts b/open-sse/config/embeddingRegistry.ts index 15f15f7fcc1..1bfb60120c7 100644 --- a/open-sse/config/embeddingRegistry.ts +++ b/open-sse/config/embeddingRegistry.ts @@ -585,6 +585,15 @@ export const EMBEDDING_PROVIDERS: Record = { authHeader: "bearer", models: [{ id: "qwen3-embedding", name: "Qwen3 Embedding", dimensions: 4096 }], }, + + // ParalonCloud — OpenAI-compatible /v1/embeddings behind the chat API key. + paraloncloud: { + id: "paraloncloud", + baseUrl: "https://paraloncloud.com/v1/embeddings", + authType: "apikey", + authHeader: "bearer", + models: [{ id: "multilingual-e5-small", name: "Multilingual E5 Small", dimensions: 384 }], + }, }; const EMBEDDING_PROVIDER_ALIASES: Record = { @@ -597,6 +606,7 @@ const EMBEDDING_PROVIDER_ALIASES: Record = { "lm-studio": "lmstudio", llamacpp: "llama-cpp", "llama.cpp": "llama-cpp", + paralon: "paraloncloud", gh: "github", }; diff --git a/open-sse/config/providers/index.ts b/open-sse/config/providers/index.ts index 4b12720e1af..c30fc49a650 100644 --- a/open-sse/config/providers/index.ts +++ b/open-sse/config/providers/index.ts @@ -294,6 +294,7 @@ import { syntxProvider } from "./registry/syntx/index.ts"; import { daoxeProvider } from "./registry/daoxe/index.ts"; import { gisooProvider } from "./registry/gisoo/index.ts"; import { demonrouteProvider } from "./registry/demonroute/index.ts"; +import { paraloncloudProvider } from "./registry/paraloncloud/index.ts"; import { empiriolabsProvider } from "./registry/empiriolabs/index.ts"; export const REGISTRY: Record = { @@ -593,5 +594,6 @@ export const REGISTRY: Record = { daoxe: daoxeProvider, gisoo: gisooProvider, demonroute: demonrouteProvider, + paraloncloud: paraloncloudProvider, empiriolabs: empiriolabsProvider, }; diff --git a/open-sse/config/providers/registry/paraloncloud/index.ts b/open-sse/config/providers/registry/paraloncloud/index.ts new file mode 100644 index 00000000000..b2df97329f2 --- /dev/null +++ b/open-sse/config/providers/registry/paraloncloud/index.ts @@ -0,0 +1,21 @@ +import type { RegistryEntry } from "../../shared.ts"; +import { buildOpenAiCompatibleRegistryEntry } from "../../shared.ts"; + +/** + * ParalonCloud (https://paraloncloud.com) — OpenAI-compatible `/v1` host for chat and + * embeddings. API keys come from https://paraloncloud.com/console. + * + * Embeddings (`multilingual-e5-small`) live in embeddingRegistry. `GET /v1/models` is + * key-gated, so `modelsUrl` drives key validation and dashboard discovery; the seed below + * is only the offline fallback. + */ +export const paraloncloudProvider: RegistryEntry = buildOpenAiCompatibleRegistryEntry({ + id: "paraloncloud", + alias: "paralon", + baseUrl: "https://paraloncloud.com/v1/chat/completions", + modelsUrl: "https://paraloncloud.com/v1/models", + models: [ + { id: "qwen3.8-27b", name: "Qwen3.8 27B" }, + { id: "qwen3-3b", name: "Qwen3 3B" }, + ], +}); diff --git a/open-sse/handlers/chatCore.ts b/open-sse/handlers/chatCore.ts index 0bfeb40a989..67ac7ce4c22 100644 --- a/open-sse/handlers/chatCore.ts +++ b/open-sse/handlers/chatCore.ts @@ -250,6 +250,7 @@ import { runPluginOnResponseHook } from "./chatCore/pluginOnResponse.ts"; import { applyPostTranslationCompression } from "./chatCore/postTranslationCompression.ts"; import { isJsonRecord } from "./chatCore/nonStreamingResponseParse.ts"; import { recordNonStreamingUsageStats } from "./chatCore/nonStreamingUsageStats.ts"; +import { resolveUsageAgentContext } from "./chatCore/agentContext.ts"; import { getModelNormalizeToolCallId, getModelPreserveOpenAIDeveloperRole } from "@/lib/db/models"; import { extractSessionAffinityKey } from "@/sse/services/auth"; import { createManagedLeaseFence } from "./chatCore/managedLeaseFence.ts"; @@ -532,6 +533,7 @@ async function handleChatCoreInner({ ): EffectiveServiceTier | null => resolveReportedServiceTierFor(provider, payload, maxDepth); let upstreamHttpSelection: HttpAttemptSelection | null = null; let providerResponse; + const agentContext = resolveUsageAgentContext(body, clientRawRequest?.headers, apiKeyInfo); // Failure usage record building extracted to chatCore/failureUsage.ts (#3501); the handler keeps // the fire-and-forget save + computes latencyMs, so the call sites stay byte-identical. const persistFailureUsage = ( @@ -553,6 +555,7 @@ async function handleChatCoreInner({ latencyMs: Date.now() - startTime, endpoint: endpointPath, cpaAuthIndex: readCpaAuthIndex(providerResponse), + agentContext, aggregate: aggregate ?? undefined, }) ).catch(() => {}); @@ -3316,6 +3319,7 @@ async function handleChatCoreInner({ // Non-streaming response if (!stream) { const nonStreamingOutcome = await runNonStreamingResponse({ + agentContext, reportSignatureFailure, geminiPromptCacheMode, isScopedGeminiCacheTarget, @@ -3459,6 +3463,7 @@ async function handleChatCoreInner({ } const streamingTailOutcome = await runStreamingTail({ + agentContext, upstreamHttpSelection, geminiPromptCacheMode, isScopedGeminiCacheTarget, diff --git a/open-sse/handlers/chatCore/agentContext.ts b/open-sse/handlers/chatCore/agentContext.ts new file mode 100644 index 00000000000..ac8d96002f2 --- /dev/null +++ b/open-sse/handlers/chatCore/agentContext.ts @@ -0,0 +1,239 @@ +/** + * Agent context: which coding-agent session a request belongs to and which project it works + * on, so usage can be attributed per session and per project. + * + * Resolution order (first match wins): + * project `x-omniroute-project` / `x-omniroute-project-repo` headers (set by the client's + * launch wrapper through ANTHROPIC_CUSTOM_HEADERS), then the working directory the + * agent announces in its prompt: Claude Code's "Primary working directory:" line in a + * `role: "system"` message (2.1.28x, mid-conversation-system beta), "Working + * directory:" in the top-level `system` field (older builds), or Codex's `` inside + * ``. + * session x-omniroute-session-id, then the client's own session id exactly as call logs + * record it (resolveClientSessionId in services/conversationTracker.ts: + * x-claude-code-session-id, x-opencode-session, Codex `session_id`, the Claude Code + * `metadata.user_id` JSON `session_id`), so agent_sessions.client_session_id joins + * call_logs.client_session_id. + * + * Prompt text is untrusted and can be hundreds of KB, so scanning is bounded to the first + * entries and the first MAX_SCAN_CHARS of each text, and uses indexOf rather than regexes. + */ + +import { resolveClientSessionId as resolveSharedClientSessionId } from "../../services/conversationTracker.ts"; +import { clientSessionIdFromRawRequest } from "./clientSessionId.ts"; +import { getHeaderValueCaseInsensitive } from "./headers.ts"; + +type HeaderSource = Record | Headers | null | undefined; +type JsonRecord = Record; + +export interface AgentContext { + client: string | null; + clientSessionId: string | null; + projectName: string | null; + projectRepo: string | null; + projectPath: string | null; + projectSource: "header" | "path" | null; + gitBranch: string | null; +} + +const MAX_SCAN_CHARS = 64_000; +const MAX_SCANNED_ENTRIES = 12; +const MAX_PATH_CHARS = 512; +const MAX_NAME_CHARS = 120; +const MAX_REPO_CHARS = 200; +const MAX_SESSION_ID_CHARS = 128; +const WORKING_DIRECTORY_MARKERS = ["Primary working directory:", "Working directory:"]; +const WORKTREE_SEGMENTS = ["/.claude/worktrees/", "/.worktrees/"]; +const INSTRUCTION_ROLES: ReadonlySet = new Set(["system", "developer"]); +const USER_ROLES: ReadonlySet = new Set(["user"]); + +function asRecord(value: unknown): JsonRecord { + return value && typeof value === "object" && !Array.isArray(value) ? (value as JsonRecord) : {}; +} + +/** Single line, no control characters, capped length; null when nothing usable is left. */ +function cleanValue(value: unknown, maxChars: number): string | null { + if (typeof value !== "string") return null; + const cleaned = value.replace(/[\u0000-\u001f\u007f]/g, " ").trim(); + return cleaned ? cleaned.slice(0, maxChars) : null; +} + +function readHeader(headers: HeaderSource, name: string, maxChars: number): string | null { + return cleanValue(getHeaderValueCaseInsensitive(headers, name), maxChars); +} + +function textOfContent(content: unknown): string { + if (typeof content === "string") return content.slice(0, MAX_SCAN_CHARS); + if (!Array.isArray(content)) return ""; + let text = ""; + for (const part of content) { + const partText = asRecord(part).text; + if (typeof partText === "string") text += `${partText.slice(0, MAX_SCAN_CHARS)}\n`; + if (text.length >= MAX_SCAN_CHARS) break; + } + return text.slice(0, MAX_SCAN_CHARS); +} + +/** Chat `messages` or Responses `input` entries, capped to the first few. */ +function conversationEntries(body: JsonRecord): JsonRecord[] { + const entries = Array.isArray(body.messages) + ? body.messages + : Array.isArray(body.input) + ? body.input + : []; + return entries.slice(0, MAX_SCANNED_ENTRIES).map(asRecord); +} + +function textsByRole(body: JsonRecord, roles: ReadonlySet): string[] { + return conversationEntries(body) + .filter((entry) => roles.has(String(entry.role))) + .map((entry) => textOfContent(entry.content)) + .filter(Boolean); +} + +function lineAfterMarker(text: string, marker: string): string | null { + const start = text.indexOf(marker); + if (start < 0) return null; + const valueStart = start + marker.length; + const lineEnd = text.indexOf("\n", valueStart); + return text.slice(valueStart, lineEnd < 0 ? undefined : lineEnd).trim() || null; +} + +function textBetween(text: string, open: string, close: string): string | null { + const start = text.indexOf(open); + if (start < 0) return null; + const end = text.indexOf(close, start + open.length); + return end < 0 ? null : text.slice(start + open.length, end).trim() || null; +} + +/** Absolute POSIX, home-relative or Windows drive paths only, so prose never matches. */ +function looksLikePath(value: string): boolean { + return value.startsWith("/") || value.startsWith("~") || /^[A-Za-z]:[\\/]/.test(value); +} + +function asPath(candidate: string | null): string | null { + return candidate && looksLikePath(candidate) ? cleanValue(candidate, MAX_PATH_CHARS) : null; +} + +/** Claude Code / OpenCode announce the directory in system or developer text. */ +function workingDirectoryFromInstructions(body: JsonRecord): string | null { + const texts = [textOfContent(body.system), ...textsByRole(body, INSTRUCTION_ROLES)]; + for (const text of texts) { + for (const marker of WORKING_DIRECTORY_MARKERS) { + const path = asPath(lineAfterMarker(text, marker)); + if (path) return path; + } + } + return null; +} + +/** + * Codex sends `…` as a user message. Only that envelope is + * trusted in user text, because the first user message also carries arbitrary files (CLAUDE.md). + */ +function workingDirectoryFromCodexEnvironment(body: JsonRecord): string | null { + for (const text of textsByRole(body, USER_ROLES)) { + const environment = textBetween(text, "", ""); + const path = environment ? asPath(textBetween(environment, "", "")) : null; + if (path) return path; + } + return null; +} + +/** Repository folder name for a working directory, skipping known worktree folders. */ +export function projectNameFromPath(path: string): string | null { + let normalized = path.replace(/\\/g, "/").replace(/\/+$/, ""); + for (const segment of WORKTREE_SEGMENTS) { + const index = normalized.indexOf(segment); + if (index > 0) normalized = normalized.slice(0, index); + } + const name = normalized.slice(normalized.lastIndexOf("/") + 1); + return cleanValue(name, MAX_NAME_CHARS); +} + +function isHeaders(headers: HeaderSource): headers is Headers { + return Boolean(headers) && typeof (headers as Headers).get === "function"; +} + +/** The client's own session id, read by the same resolver that fills call_logs (#17125). */ +function sharedClientSessionId(body: JsonRecord, headers: HeaderSource): string | null { + if (!isHeaders(headers)) return clientSessionIdFromRawRequest({ headers, body }); + try { + return resolveSharedClientSessionId(headers, body); + } catch { + return null; + } +} + +function resolveClientSessionId(body: JsonRecord, headers: HeaderSource): string | null { + return ( + readHeader(headers, "x-omniroute-session-id", MAX_SESSION_ID_CHARS) || + cleanValue(sharedClientSessionId(body, headers), MAX_SESSION_ID_CHARS) + ); +} + +function resolveClient(headers: HeaderSource): string | null { + const userAgent = readHeader(headers, "user-agent", 200)?.toLowerCase(); + if (!userAgent) return null; + if (userAgent.startsWith("claude-cli/") || userAgent.includes("claude-code")) { + return "claude-code"; + } + if (userAgent.startsWith("codex")) return "codex"; + if (userAgent.startsWith("opencode")) return "opencode"; + return cleanValue(userAgent.split(/[/\s]/)[0], 40); +} + +function resolveGitBranch(body: JsonRecord): string | null { + const firstUser = conversationEntries(body).find((entry) => USER_ROLES.has(String(entry.role))); + if (!firstUser) return null; + return cleanValue(lineAfterMarker(textOfContent(firstUser.content), "Current branch:"), 200); +} + +export function extractAgentContext(body: unknown, headers: HeaderSource): AgentContext { + const record = asRecord(body); + const headerProject = readHeader(headers, "x-omniroute-project", MAX_NAME_CHARS); + const projectPath = + workingDirectoryFromInstructions(record) || workingDirectoryFromCodexEnvironment(record); + const pathProject = projectPath ? projectNameFromPath(projectPath) : null; + const projectName = headerProject || pathProject; + return { + client: resolveClient(headers), + clientSessionId: resolveClientSessionId(record, headers), + projectName, + projectRepo: readHeader(headers, "x-omniroute-project-repo", MAX_REPO_CHARS), + projectPath, + projectSource: headerProject ? "header" : pathProject ? "path" : null, + gitBranch: resolveGitBranch(record), + }; +} + +/** Drops everything read from prompt text, keeping only header-supplied identifiers. */ +export function withoutPromptDerivedFields(context: AgentContext): AgentContext { + const fromHeader = context.projectSource === "header"; + return { + ...context, + projectName: fromHeader ? context.projectName : null, + projectSource: fromHeader ? "header" : null, + projectPath: null, + gitBranch: null, + }; +} + +/** + * Agent context for usage attribution. Keys with `noLog` never persist prompt-derived fields + * (working directory, branch); explicit header identifiers are still kept. + */ +export function resolveUsageAgentContext( + body: unknown, + headers: HeaderSource, + apiKeyInfo: { noLog?: boolean } | null | undefined +): AgentContext { + const context = extractAgentContext(body, headers); + return apiKeyInfo?.noLog === true ? withoutPromptDerivedFields(context) : context; +} + +export function hasAgentIdentity( + context: AgentContext | null | undefined +): context is AgentContext { + return Boolean(context?.clientSessionId || context?.projectName); +} diff --git a/open-sse/handlers/chatCore/failureUsage.ts b/open-sse/handlers/chatCore/failureUsage.ts index 70223183604..dfebe4d5c28 100644 --- a/open-sse/handlers/chatCore/failureUsage.ts +++ b/open-sse/handlers/chatCore/failureUsage.ts @@ -9,6 +9,7 @@ */ import { buildErrorBody } from "../../utils/error.ts"; +import { type AgentContext } from "./agentContext.ts"; export { readCpaAuthIndex } from "./cpaTraceAuthIndex.ts"; @@ -69,6 +70,7 @@ export function buildFailureUsageRecord(opts: { endpoint?: string | null | undefined; cpaAuthIndex?: string | null | undefined; aggregate?: FailureUsageAggregate | null; + agentContext?: AgentContext | null; }) { return { provider: opts.provider || "unknown", @@ -93,5 +95,6 @@ export function buildFailureUsageRecord(opts: { comboStrategy: opts.isCombo ? opts.comboStrategy || undefined : undefined, endpoint: opts.endpoint || undefined, cpaAuthIndex: opts.cpaAuthIndex || undefined, + agentContext: opts.agentContext ?? null, }; } diff --git a/open-sse/handlers/chatCore/nonStreamingResponse.ts b/open-sse/handlers/chatCore/nonStreamingResponse.ts index 8ff730ebea6..d42c6c3f309 100644 --- a/open-sse/handlers/chatCore/nonStreamingResponse.ts +++ b/open-sse/handlers/chatCore/nonStreamingResponse.ts @@ -108,6 +108,7 @@ export async function runNonStreamingResponse(deps: NonStreamingDeps) { effectiveModel, effectiveServiceTier: _effectiveServiceTier, emitRequestGamificationEvent, + agentContext, endpointPath, executeProviderRequest, executeRefreshCredentials, @@ -759,6 +760,7 @@ export async function runNonStreamingResponse(deps: NonStreamingDeps) { comboStrategy, endpoint: endpointPath, cpaAuthIndex: readCpaAuthIndex(providerResponse), + agentContext, }); // #12150 P1b surface 3 (fix round 1): a video-bridge-observed request's diff --git a/open-sse/handlers/chatCore/nonStreamingUsageStats.ts b/open-sse/handlers/chatCore/nonStreamingUsageStats.ts index 2d9fb125d30..1213293c2f8 100644 --- a/open-sse/handlers/chatCore/nonStreamingUsageStats.ts +++ b/open-sse/handlers/chatCore/nonStreamingUsageStats.ts @@ -17,6 +17,7 @@ import { COLORS } from "../../utils/stream.ts"; import { recordTokenUsage } from "../../services/tokenLimitCounter.ts"; import { computeBillableTokens } from "./upstreamTimeouts.ts"; import { type EffectiveServiceTier } from "./serviceTier.ts"; +import { type AgentContext } from "./agentContext.ts"; export type RecordNonStreamingUsageStatsContext = { traceEnabled: boolean; @@ -30,6 +31,7 @@ export type RecordNonStreamingUsageStatsContext = { comboStrategy: string | null | undefined; endpoint?: string | null | undefined; cpaAuthIndex?: string | null | undefined; + agentContext?: AgentContext | null; }; function logUsageTrace( @@ -60,6 +62,7 @@ function persistUsageRow(usage: object, ctx: RecordNonStreamingUsageStatsContext comboStrategy: ctx.isCombo ? ctx.comboStrategy || undefined : undefined, endpoint: ctx.endpoint || undefined, cpaAuthIndex: ctx.cpaAuthIndex || undefined, + agentContext: ctx.agentContext ?? null, }).catch((err) => { console.error("Failed to save usage stats:", err.message); }); diff --git a/open-sse/handlers/chatCore/streamingTail.ts b/open-sse/handlers/chatCore/streamingTail.ts index 9217a5cfe2c..7006a0f042c 100644 --- a/open-sse/handlers/chatCore/streamingTail.ts +++ b/open-sse/handlers/chatCore/streamingTail.ts @@ -107,6 +107,7 @@ export async function runStreamingTail(deps: StreamingTailDeps) { customToolNames, echoModel, effectiveModel, + agentContext, endpointPath, executeProviderRequest, executor, @@ -533,6 +534,7 @@ export async function runStreamingTail(deps: StreamingTailDeps) { comboStrategy, endpoint: endpointPath, cpaAuthIndex: readCpaAuthIndex(providerResponse), + agentContext, partialUsage: streamPartialUsage, emittedOutput: streamEmittedOutput(transformStream), }); diff --git a/open-sse/handlers/chatCore/streamingUsageStats.ts b/open-sse/handlers/chatCore/streamingUsageStats.ts index 6052a80ad23..149fbf56a25 100644 --- a/open-sse/handlers/chatCore/streamingUsageStats.ts +++ b/open-sse/handlers/chatCore/streamingUsageStats.ts @@ -16,6 +16,7 @@ import { recordKeyQuotaUsage } from "@/domain/keyQuota"; import { recordTokenUsage } from "../../services/tokenLimitCounter.ts"; import { computeBillableTokens } from "./upstreamTimeouts.ts"; import { type EffectiveServiceTier } from "./serviceTier.ts"; +import { type AgentContext } from "./agentContext.ts"; export type RecordStreamingUsageStatsContext = { provider: string | null | undefined; @@ -31,6 +32,7 @@ export type RecordStreamingUsageStatsContext = { comboStrategy: string | null | undefined; endpoint?: string | null | undefined; cpaAuthIndex?: string | null | undefined; + agentContext?: AgentContext | null; /** Tokens generated before a 499 / mid-stream failure (usage is null then). */ partialUsage?: unknown; /** Whether the stream carried output, so a combo cannot have failed over from it. */ @@ -56,6 +58,7 @@ function persistStreamingUsageRow(usage: object, ctx: RecordStreamingUsageStatsC comboStrategy: ctx.isCombo ? ctx.comboStrategy || undefined : undefined, endpoint: ctx.endpoint || undefined, cpaAuthIndex: ctx.cpaAuthIndex || undefined, + agentContext: ctx.agentContext ?? null, }).catch((err) => { console.error("Failed to save usage stats:", err.message); }); diff --git a/package.json b/package.json index 9f3dfe329c5..288bcaf1979 100644 --- a/package.json +++ b/package.json @@ -1,7 +1,7 @@ { "name": "omniroute", "version": "3.8.54", - "description": "Unified AI router with 383 providers, RTK+Caveman compression, auto fallback, MCP/A2A, desktop, PWA, and OpenAI-compatible APIs.", + "description": "Unified AI router with 384 providers, RTK+Caveman compression, auto fallback, MCP/A2A, desktop, PWA, and OpenAI-compatible APIs.", "type": "module", "bin": { "omniroute": "bin/omniroute.mjs", diff --git a/public/images/tier-flow-dark.svg b/public/images/tier-flow-dark.svg index cb0e11931e9..a473ac3a264 100644 --- a/public/images/tier-flow-dark.svg +++ b/public/images/tier-flow-dark.svg @@ -1,6 +1,6 @@ - + OmniRoute 4-tier fallback - OmniRoute 4-tier fallback: your IDE or CLI calls one local endpoint and the OmniRoute Smart Router fails over across 383 providers in 4 tiers — Tier 1 Subscription, Tier 2 API, Tier 3 Cheap, Tier 4 Free. + OmniRoute 4-tier fallback: your IDE or CLI calls one local endpoint and the OmniRoute Smart Router fails over across 384 providers in 4 tiers — Tier 1 Subscription, Tier 2 API, Tier 3 Cheap, Tier 4 Free. @@ -15,7 +15,7 @@ OmniRoute 4-tier fallback - Never stop building — automatic zero-config failover across 383 providers + Never stop building — automatic zero-config failover across 384 providers diff --git a/public/images/tier-flow-light.svg b/public/images/tier-flow-light.svg index cd71d24a05d..30a8c5d4352 100644 --- a/public/images/tier-flow-light.svg +++ b/public/images/tier-flow-light.svg @@ -1,6 +1,6 @@ - + OmniRoute 4-tier fallback - OmniRoute 4-tier fallback: your IDE or CLI calls one local endpoint and the OmniRoute Smart Router fails over across 383 providers in 4 tiers — Tier 1 Subscription, Tier 2 API, Tier 3 Cheap, Tier 4 Free. + OmniRoute 4-tier fallback: your IDE or CLI calls one local endpoint and the OmniRoute Smart Router fails over across 384 providers in 4 tiers — Tier 1 Subscription, Tier 2 API, Tier 3 Cheap, Tier 4 Free. @@ -15,7 +15,7 @@ OmniRoute 4-tier fallback - Never stop building — automatic zero-config failover across 383 providers + Never stop building — automatic zero-config failover across 384 providers diff --git a/scripts/docs/gen-readme-cli-terminal.mjs b/scripts/docs/gen-readme-cli-terminal.mjs index f43e8f0aafc..0c5ef537173 100644 --- a/scripts/docs/gen-readme-cli-terminal.mjs +++ b/scripts/docs/gen-readme-cli-terminal.mjs @@ -33,7 +33,7 @@ const INV = KIT.dark; // The gate-checked provider total (scripts/check/check-docs-counts-sync.mjs reads // " providers" from the title); the health line and the "… N more providers" footer of the // providers screen derive from it so the three numbers can never disagree. -const PROVIDER_TOTAL = 383; +const PROVIDER_TOTAL = 384; // The three screens: command, then header + 4 lines + a dim footer line. Segments are plain // strings or { t, c } with c ∈ ok | warn | err | dim. diff --git a/scripts/docs/gen-readme-comparison-table.mjs b/scripts/docs/gen-readme-comparison-table.mjs index 4f27a636c62..ea7a2511355 100644 --- a/scripts/docs/gen-readme-comparison-table.mjs +++ b/scripts/docs/gen-readme-comparison-table.mjs @@ -9,7 +9,7 @@ // the drawn text and what the outlining step shapes. The canonical numbers stay in // / because scripts/check/check-docs-counts-sync.mjs reads them from the raw file. // The table is a dated snapshot of each project's docs — cells are kept verbatim; only the -// provider total follows the live code (383), which the counts gate enforces. +// provider total follows the live code (384), which the counts gate enforces. import { writeFileSync } from "node:fs"; import { fileURLToPath } from "node:url"; import { outlineSvg } from "./readme-outline.mjs"; @@ -18,7 +18,7 @@ import { svgOpen, gridDefs, esc, afterKicker } from "./readme-diagram-theme.mjs" // Cells: a string is drawn as text; "yes" / "mid" / "no" draw the check / partial / cross glyph. // Column order: OmniRoute, 9router, OpenRouter, CLIProxyAPI, LiteLLM. const ROWS = [ - { label: "Providers", cells: ["383", "40+", "400+*", "~5", "100+"] }, + { label: "Providers", cells: ["384", "40+", "400+*", "~5", "100+"] }, { label: "Free providers built-in", cells: ["150+", "yes", "mid", "no", "no"] }, { label: "Routing strategies", cells: ["21", "2", "3", "2", "6"] }, { label: "Token compression", cells: ["12 eng", "yes", "mid", "no", "no"] }, @@ -98,7 +98,7 @@ export async function renderComparisonTable() { viewBox: "0 0 1200 780", title: "Comparison table: OmniRoute versus 9router, OpenRouter, CLIProxyAPI and LiteLLM across 13 capabilities.", - desc: "OmniRoute is the only one with the full set: 383 providers, 150+ free providers built-in, 21 routing strategies, 12-engine token compression, a built-in MCP server with 110 tools, A2A protocol, persistent memory, guardrails, cloud agents, TLS fingerprint stealth, desktop/Termux/PWA, 67 UI locales and 100% MIT self-hosted. 9router has free providers, RTK compression and translation but no MCP, A2A, memory, guardrails, cloud agents or stealth. OpenRouter is a hosted SaaS with 400+ models, guardrails and a hosted MCP but is not self-hosted and lacks A2A, memory, cloud agents and stealth. CLIProxyAPI is a light OAuth proxy with two routing strategies. LiteLLM has 100+ providers, A2A and extensive guardrails but no memory, compression, free tier, stealth or cloud agents. Static-header comparison table where each capability row fades in top to bottom; the OmniRoute column is highlighted and shows a check or a leading value in every row, while competitors show a mix of checks, partials and crosses. Light and dark follow the viewer's color scheme.", + desc: "OmniRoute is the only one with the full set: 384 providers, 150+ free providers built-in, 21 routing strategies, 12-engine token compression, a built-in MCP server with 110 tools, A2A protocol, persistent memory, guardrails, cloud agents, TLS fingerprint stealth, desktop/Termux/PWA, 67 UI locales and 100% MIT self-hosted. 9router has free providers, RTK compression and translation but no MCP, A2A, memory, guardrails, cloud agents or stealth. OpenRouter is a hosted SaaS with 400+ models, guardrails and a hosted MCP but is not self-hosted and lacks A2A, memory, cloud agents and stealth. CLIProxyAPI is a light OAuth proxy with two routing strategies. LiteLLM has 100+ providers, A2A and extensive guardrails but no memory, compression, free tier, stealth or cloud agents. Static-header comparison table where each capability row fades in top to bottom; the OmniRoute column is highlighted and shows a check or a leading value in every row, while competitors show a mix of checks, partials and crosses. Light and dark follow the viewer's color scheme.", extraCss: `.cmp-strong{font-weight:600;fill:var(--ink)}`, }) + gridDefs("cmp-grid") + diff --git a/scripts/docs/gen-readme-hero.mjs b/scripts/docs/gen-readme-hero.mjs index 2e0bf880975..9d49c4e7bd2 100644 --- a/scripts/docs/gen-readme-hero.mjs +++ b/scripts/docs/gen-readme-hero.mjs @@ -15,14 +15,14 @@ import { svgOpen, gridDefs, esc } from "./readme-diagram-theme.mjs"; export const HERO_COPY = { kicker: ["OMNIROUTE — THE FREE AI GATEWAY"], headline: ["Never stop coding."], - sub: ["Every AI tool → 383 providers — 150+ free —", "through one endpoint."], + sub: ["Every AI tool → 384 providers — 150+ free —", "through one endpoint."], tools: [ "Claude Code · Codex · Cursor · Cline · Copilot · Antigravity → FREE Claude / GPT / Gemini · auto-fallback", ], cmd: ["$ npm i -g omniroute"], foot: ["RTK + CAVEMAN · STACKED COMPRESSION · ONE ENDPOINT · /v1"], stats: [ - "383", + "384", "AI PROVIDERS", "150+", "FREE TIERS", @@ -60,8 +60,8 @@ export async function renderReadmeHero() { svgOpen({ viewBox: "0 0 1200 548", title: - "OmniRoute — Never stop coding. Every AI tool to 383 providers, 150+ free, through one endpoint.", - desc: "Hero card in the Cheaper Inference UI Kit style: Switzer drawn as outlines, IBM Plex Mono for the install command. Claims: 383 AI providers, 150+ free tiers, ~1.62B free tokens per month, 15–95% token savings, 21 routing strategies, $0 to start. Light and dark follow the viewer's color scheme.", + "OmniRoute — Never stop coding. Every AI tool to 384 providers, 150+ free, through one endpoint.", + desc: "Hero card in the Cheaper Inference UI Kit style: Switzer drawn as outlines, IBM Plex Mono for the install command. Claims: 384 AI providers, 150+ free tiers, ~1.62B free tokens per month, 15–95% token savings, 21 routing strategies, $0 to start. Light and dark follow the viewer's color scheme.", }) + gridDefs("heroGrid") + ` \n` + diff --git a/scripts/docs/gen-readme-promise-pillars.mjs b/scripts/docs/gen-readme-promise-pillars.mjs index 432350d15c8..d7fcfa34040 100644 --- a/scripts/docs/gen-readme-promise-pillars.mjs +++ b/scripts/docs/gen-readme-promise-pillars.mjs @@ -20,7 +20,7 @@ const NB = " "; export const PROMISE_PILLARS_COPY = { kicker: ["THE PROMISE"], headline: ["One endpoint. ", " Never stop building — OmniRoute picks ", "."], - headlineMarks: ["383 providers.", "the cheapest one that works"], + headlineMarks: ["384 providers.", "the cheapest one that works"], titles: [ "Never hit limits", "Save up to 95% tokens", @@ -30,7 +30,7 @@ export const PROMISE_PILLARS_COPY = { "Production-grade", ], bodies: [ - "Auto-fallback across 383 providers in", + "Auto-fallback across 384 providers in", "milliseconds. Quota out? The next provider", "takes over while a healthy target remains.", "RTK + Caveman stacked compression cuts", @@ -148,7 +148,7 @@ export async function renderReadmePromisePillars() { svgOpen({ viewBox: `0 0 ${W} ${H}`, title: - "The OmniRoute promise: one endpoint and 383 providers. Six pillars. Resilient fallback: automatic routing continues while another healthy target is available. Save up to 95 percent of eligible tokens: RTK plus Caveman stacked compression averages about 89 percent on tool-heavy sessions. Zero dollars to start: 150+ providers with a free tier and 52 recurring or keyless free-forever providers. Every tool works: 39 CLI and agent integration records, including Claude Code, Codex, Cursor, Cline, Copilot and Antigravity, through one config. One endpoint: OpenAI, Claude, Gemini and Responses API translation at /v1. Production controls: circuit breakers, TLS stealth, MCP with 110 tools, A2A, memory, guardrails, evals, and 60,000+ static test declarations across 8,400+ tracked test files.", + "The OmniRoute promise: one endpoint and 384 providers. Six pillars. Resilient fallback: automatic routing continues while another healthy target is available. Save up to 95 percent of eligible tokens: RTK plus Caveman stacked compression averages about 89 percent on tool-heavy sessions. Zero dollars to start: 150+ providers with a free tier and 52 recurring or keyless free-forever providers. Every tool works: 39 CLI and agent integration records, including Claude Code, Codex, Cursor, Cline, Copilot and Antigravity, through one config. One endpoint: OpenAI, Claude, Gemini and Responses API translation at /v1. Production controls: circuit breakers, TLS stealth, MCP with 110 tools, A2A, memory, guardrails, evals, and 60,000+ static test declarations across 8,400+ tracked test files.", desc: "Animated promise card: six pillar tiles in reading order, with a soft highlight sweeping from tile to tile in a continuous cycle.", }) + gridDefs("pp-grid") + diff --git a/skills/README.md b/skills/README.md index 061086d4865..c804063c3f2 100644 --- a/skills/README.md +++ b/skills/README.md @@ -27,7 +27,7 @@ Each manifest URL follows the pattern: | ID | Name | Description | | ---------------------- | ----------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `omni-auth` | Authentication | Manage API key authentication and session tokens. Start here to authenticate requests via Bearer token, obtain session cookies, and configure login requirements. | -| `omni-providers` | Providers | Manage provider connections, API keys, OAuth flows, and connection tests across OmniRoute's 383-provider catalog. | +| `omni-providers` | Providers | Manage provider connections, API keys, OAuth flows, and connection tests across OmniRoute's 384-provider catalog. | | `omni-models` | Models | Query available AI models across all configured providers. List models, resolve model aliases, and browse the full model catalog including provider-specific variants. | | `omni-combos-routing` | Combos & Routing | Create and manage routing combos with 19 strategies (priority, weighted, round-robin, Auto-combo, and more). Configure fallback chains, test routing outcomes, and retrieve combo metrics. | | `omni-api-keys` | API Keys | Create, list, rotate, and revoke OmniRoute API keys. Control per-key scopes, spending limits, and expiration. | diff --git a/skills/omni-auth/SKILL.md b/skills/omni-auth/SKILL.md index 063f39b6b1d..4ba80b945a4 100644 --- a/skills/omni-auth/SKILL.md +++ b/skills/omni-auth/SKILL.md @@ -176,7 +176,7 @@ See the full OpenAPI specification at `GET /api/openapi/spec` or `docs/openapi.y # OmniRoute -Local/remote AI gateway exposing OpenAI-compatible REST. One key, 383 providers, +Local/remote AI gateway exposing OpenAI-compatible REST. One key, 384 providers, auto-fallback, RTK token saver, MCP server, A2A agents. ## Setup diff --git a/skills/omni-providers/SKILL.md b/skills/omni-providers/SKILL.md index 5c7ef0e814a..9de7d3be32a 100644 --- a/skills/omni-providers/SKILL.md +++ b/skills/omni-providers/SKILL.md @@ -1,13 +1,13 @@ --- name: omni-providers -description: "Manage provider connections, API keys, OAuth flows, and connection tests via the REST API. List, add, update, remove, and test AI provider integrations across OmniRoute's 383-provider catalog." +description: "Manage provider connections, API keys, OAuth flows, and connection tests via the REST API. List, add, update, remove, and test AI provider integrations across OmniRoute's 384-provider catalog." --- ## Overview -Manage provider connections, API keys, OAuth flows, and connection tests via the REST API. List, add, update, remove, and test AI provider integrations across OmniRoute's 383-provider catalog. +Manage provider connections, API keys, OAuth flows, and connection tests via the REST API. List, add, update, remove, and test AI provider integrations across OmniRoute's 384-provider catalog. ## Authentication diff --git a/src/app/api/providers/[id]/models/discovery/providerSets.ts b/src/app/api/providers/[id]/models/discovery/providerSets.ts index 9dfb6196455..5d0eff88a6a 100644 --- a/src/app/api/providers/[id]/models/discovery/providerSets.ts +++ b/src/app/api/providers/[id]/models/discovery/providerSets.ts @@ -138,6 +138,9 @@ export const NAMED_OPENAI_STYLE_PROVIDERS = new Set([ // served from its key-gated /v1/models endpoint; the registry seed stays as the // offline fallback. "tokenharbor", + // ParalonCloud — GET https://paraloncloud.com/v1/models is key-gated; the registry + // seed is only the offline fallback. + "paraloncloud", ]); export function isNamedOpenAIStyleProvider(provider: string): boolean { diff --git a/src/lib/db/agentSessions.ts b/src/lib/db/agentSessions.ts new file mode 100644 index 00000000000..6ff0a83dc41 --- /dev/null +++ b/src/lib/db/agentSessions.ts @@ -0,0 +1,144 @@ +/** + * Agent sessions: one aggregate row per coding-agent session (see 229_agent_sessions.sql). + * Written from inside saveRequestUsage's transaction, so a usage row and its session counters + * always move together and dedup no-ops never double count. + */ + +import { createHash, randomUUID } from "node:crypto"; +import type { AgentContext } from "@omniroute/open-sse/handlers/chatCore/agentContext.ts"; + +import type { SqliteAdapter } from "./adapters/types"; + +/** Requests without a client session id join the key+project session seen this recently. */ +export const AGENT_SESSION_IDLE_WINDOW_MS = 30 * 60 * 1000; + +// A type alias (not an interface) so it stays assignable to the cost calculator's token record. +export type AgentSessionTokens = { + input: number; + output: number; + cacheRead: number; + cacheCreation: number; + reasoning: number; +}; + +export interface AgentSessionUsage { + context: AgentContext; + apiKeyId: string | null; + apiKeyName: string | null; + timestamp: string; + success: boolean; + tokens: AgentSessionTokens; + costUsd: number; + /** false when the model has no pricing row and cost_usd could not include it. */ + priced: boolean; + provider: string | null; + model: string | null; + connectionId: string | null; +} + +function newSessionId(): string { + return `as_${randomUUID().replace(/-/g, "").slice(0, 32)}`; +} + +/** Scoped to the API key, so one key can never write into another key's session. */ +function clientSessionRowId(apiKeyId: string | null, clientSessionId: string): string { + const digest = createHash("sha256") + .update(`${apiKeyId ?? ""}\x1f${clientSessionId}`) + .digest("hex"); + return `as_${digest.slice(0, 32)}`; +} + +function findIdleWindowSession(db: SqliteAdapter, usage: AgentSessionUsage): string | null { + const cutoff = new Date( + new Date(usage.timestamp).getTime() - AGENT_SESSION_IDLE_WINDOW_MS + ).toISOString(); + const row = db + .prepare( + `SELECT id FROM agent_sessions + WHERE api_key_id IS ? AND client_session_id IS NULL AND project_name = ? + AND last_seen_at >= ? + ORDER BY last_seen_at DESC LIMIT 1` + ) + .get(usage.apiKeyId, usage.context.projectName, cutoff) as { id: string } | undefined; + return row?.id ?? null; +} + +function resolveSessionId(db: SqliteAdapter, usage: AgentSessionUsage): string { + const { clientSessionId } = usage.context; + if (clientSessionId) return clientSessionRowId(usage.apiKeyId, clientSessionId); + return findIdleWindowSession(db, usage) ?? newSessionId(); +} + +const UPSERT_SQL = ` + INSERT INTO agent_sessions ( + id, api_key_id, api_key_name, client, client_session_id, project_name, project_repo, + project_path, project_source, git_branch, first_seen_at, last_seen_at, request_count, + error_count, tokens_input, tokens_output, tokens_cache_read, tokens_cache_creation, + tokens_reasoning, cost_usd, unpriced_count, last_provider, last_model, last_connection_id + ) VALUES ( + @id, @apiKeyId, @apiKeyName, @client, @clientSessionId, @projectName, @projectRepo, + @projectPath, @projectSource, @gitBranch, @timestamp, @timestamp, 1, + @errorCount, @tokensInput, @tokensOutput, @tokensCacheRead, @tokensCacheCreation, + @tokensReasoning, @costUsd, @unpricedCount, @provider, @model, @connectionId + ) + ON CONFLICT(id) DO UPDATE SET + api_key_name = COALESCE(excluded.api_key_name, api_key_name), + client = COALESCE(excluded.client, client), + project_name = COALESCE(excluded.project_name, project_name), + project_repo = COALESCE(excluded.project_repo, project_repo), + project_path = COALESCE(excluded.project_path, project_path), + project_source = COALESCE(excluded.project_source, project_source), + git_branch = COALESCE(excluded.git_branch, git_branch), + first_seen_at = MIN(first_seen_at, excluded.first_seen_at), + last_seen_at = MAX(last_seen_at, excluded.last_seen_at), + request_count = request_count + 1, + error_count = error_count + excluded.error_count, + tokens_input = tokens_input + excluded.tokens_input, + tokens_output = tokens_output + excluded.tokens_output, + tokens_cache_read = tokens_cache_read + excluded.tokens_cache_read, + tokens_cache_creation = tokens_cache_creation + excluded.tokens_cache_creation, + tokens_reasoning = tokens_reasoning + excluded.tokens_reasoning, + cost_usd = cost_usd + excluded.cost_usd, + unpriced_count = unpriced_count + excluded.unpriced_count, + last_provider = COALESCE(excluded.last_provider, last_provider), + last_model = COALESCE(excluded.last_model, last_model), + last_connection_id = COALESCE(excluded.last_connection_id, last_connection_id) +`; + +function hasTokens(tokens: AgentSessionTokens): boolean { + return tokens.input + tokens.output + tokens.cacheRead + tokens.cacheCreation > 0; +} + +/** + * Adds one request to its agent session and returns the session id. Callers must only pass + * contexts that carry a session id or a project (see hasAgentIdentity). + */ +export function recordAgentSessionUsage(db: SqliteAdapter, usage: AgentSessionUsage): string { + const id = resolveSessionId(db, usage); + const { context, tokens } = usage; + db.prepare(UPSERT_SQL).run({ + id, + apiKeyId: usage.apiKeyId, + apiKeyName: usage.apiKeyName, + client: context.client, + clientSessionId: context.clientSessionId, + projectName: context.projectName, + projectRepo: context.projectRepo, + projectPath: context.projectPath, + projectSource: context.projectSource, + gitBranch: context.gitBranch, + timestamp: usage.timestamp, + errorCount: usage.success ? 0 : 1, + tokensInput: tokens.input, + tokensOutput: tokens.output, + tokensCacheRead: tokens.cacheRead, + tokensCacheCreation: tokens.cacheCreation, + tokensReasoning: tokens.reasoning, + costUsd: usage.costUsd, + unpricedCount: !usage.priced && hasTokens(tokens) ? 1 : 0, + provider: usage.provider, + model: usage.model, + connectionId: usage.connectionId, + }); + return id; +} diff --git a/src/lib/db/migrationRunner/constants.ts b/src/lib/db/migrationRunner/constants.ts index 8902de4d79b..58dd46b995f 100644 --- a/src/lib/db/migrationRunner/constants.ts +++ b/src/lib/db/migrationRunner/constants.ts @@ -479,6 +479,93 @@ export const RENAMED_MIGRATION_COMPATIBILITY = [ toVersion: "233", toName: "api_key_self_service_settings", }, + // Agent-session builds (#14833) shipped agent_sessions + usage_history.agent_session_id + // as 190/191, 197/198, 198/199, 208/209, 214/215, 230/231 and 229/230. Upstream owns all of + // those slots; the migrations now live at 234/235. Each entry is name-locked. + { + fromVersion: "190", + fromName: "agent_sessions", + toVersion: "234", + toName: "agent_sessions", + }, + { + fromVersion: "191", + fromName: "usage_history_agent_session_id", + toVersion: "235", + toName: "usage_history_agent_session_id", + }, + { + fromVersion: "197", + fromName: "agent_sessions", + toVersion: "234", + toName: "agent_sessions", + }, + { + fromVersion: "198", + fromName: "usage_history_agent_session_id", + toVersion: "235", + toName: "usage_history_agent_session_id", + }, + { + fromVersion: "198", + fromName: "agent_sessions", + toVersion: "234", + toName: "agent_sessions", + }, + { + fromVersion: "199", + fromName: "usage_history_agent_session_id", + toVersion: "235", + toName: "usage_history_agent_session_id", + }, + { + fromVersion: "208", + fromName: "agent_sessions", + toVersion: "234", + toName: "agent_sessions", + }, + { + fromVersion: "209", + fromName: "usage_history_agent_session_id", + toVersion: "235", + toName: "usage_history_agent_session_id", + }, + { + fromVersion: "214", + fromName: "agent_sessions", + toVersion: "234", + toName: "agent_sessions", + }, + { + fromVersion: "215", + fromName: "usage_history_agent_session_id", + toVersion: "235", + toName: "usage_history_agent_session_id", + }, + { + fromVersion: "229", + fromName: "agent_sessions", + toVersion: "234", + toName: "agent_sessions", + }, + { + fromVersion: "230", + fromName: "usage_history_agent_session_id", + toVersion: "235", + toName: "usage_history_agent_session_id", + }, + { + fromVersion: "230", + fromName: "agent_sessions", + toVersion: "234", + toName: "agent_sessions", + }, + { + fromVersion: "231", + fromName: "usage_history_agent_session_id", + toVersion: "235", + toName: "usage_history_agent_session_id", + }, ] as const; export const LEGACY_VERSION_SLOT_MIGRATIONS = [ diff --git a/src/lib/db/migrations/234_agent_sessions.sql b/src/lib/db/migrations/234_agent_sessions.sql new file mode 100644 index 00000000000..f6d60a2243f --- /dev/null +++ b/src/lib/db/migrations/234_agent_sessions.sql @@ -0,0 +1,39 @@ +-- One row per coding-agent session, aggregated at write time so per-member and per-project +-- reports never scan the raw request tables (call_logs keeps only days; usage_history rolls up +-- without api_key_id). A session is identified by the client's own session id (Claude Code +-- x-claude-code-session-id, Codex, OpenCode) scoped to the API key; without one, requests from +-- the same key and project within 30 minutes share a session. project_* come from the +-- x-omniroute-project headers or the working directory announced in the prompt. +CREATE TABLE IF NOT EXISTS agent_sessions ( + id TEXT PRIMARY KEY, + api_key_id TEXT, + api_key_name TEXT, + client TEXT, + client_session_id TEXT, + project_name TEXT, + project_repo TEXT, + project_path TEXT, + project_source TEXT, + git_branch TEXT, + first_seen_at TEXT NOT NULL, + last_seen_at TEXT NOT NULL, + request_count INTEGER NOT NULL DEFAULT 0, + error_count INTEGER NOT NULL DEFAULT 0, + tokens_input INTEGER NOT NULL DEFAULT 0, + tokens_output INTEGER NOT NULL DEFAULT 0, + tokens_cache_read INTEGER NOT NULL DEFAULT 0, + tokens_cache_creation INTEGER NOT NULL DEFAULT 0, + tokens_reasoning INTEGER NOT NULL DEFAULT 0, + cost_usd REAL NOT NULL DEFAULT 0, + unpriced_count INTEGER NOT NULL DEFAULT 0, + last_provider TEXT, + last_model TEXT, + last_connection_id TEXT +); + +CREATE INDEX IF NOT EXISTS idx_agent_sessions_key_last_seen + ON agent_sessions(api_key_id, last_seen_at); +CREATE INDEX IF NOT EXISTS idx_agent_sessions_project_last_seen + ON agent_sessions(project_name, last_seen_at); +CREATE INDEX IF NOT EXISTS idx_agent_sessions_last_seen + ON agent_sessions(last_seen_at); diff --git a/src/lib/db/migrations/235_usage_history_agent_session_id.sql b/src/lib/db/migrations/235_usage_history_agent_session_id.sql new file mode 100644 index 00000000000..0678e4cbfbd --- /dev/null +++ b/src/lib/db/migrations/235_usage_history_agent_session_id.sql @@ -0,0 +1,6 @@ +-- Links each usage row to its agent session (234_agent_sessions) so reports can filter requests +-- by project or session over any date-time range. NULL for traffic without an agent identity. +-- Kept as its own migration: the runner records a file as applied when an ALTER hits +-- "duplicate column name", which would otherwise roll back the agent_sessions table with it. +ALTER TABLE usage_history ADD COLUMN agent_session_id TEXT; +CREATE INDEX IF NOT EXISTS idx_uh_agent_session ON usage_history(agent_session_id); diff --git a/src/lib/usage/usageHistory.ts b/src/lib/usage/usageHistory.ts index aa97284ff14..cb279a26890 100644 --- a/src/lib/usage/usageHistory.ts +++ b/src/lib/usage/usageHistory.ts @@ -19,6 +19,7 @@ import { buildLatencyStatsEntry, createLatencyBucket, mapUsageHistoryRow, + normalizeServiceTier, resolvePositiveOption, toNumber, toStringOrNull, @@ -33,8 +34,14 @@ import { storeCompletedDetail, getCompletedDetails, } from "./completedRequestDetails"; +import * as sessions from "./usageHistory/agentSessionUsage"; import { shouldPersistToDisk } from "./migrations"; -import { attemptWrite, isWriteLockError, scheduleWriteRetry } from "./usageWriteRetry"; +import { + attemptWrite, + isWriteLockError, + scheduleWriteRetry, + type UsageWriteOptions, +} from "./usageWriteRetry"; import { getCallLogApiKeyContext } from "./callLogApiKeyContext"; export type PendingRequestMetadata = { @@ -974,6 +981,8 @@ export interface UsageEntry { endpoint?: string | null; /** Opaque CLIProxyAPI auth_index. Never a label, path, token, or email. */ cpaAuthIndex?: string | null; + /** Coding-agent session and project of the request; attributes the row to an agent session. */ + agentContext?: sessions.AgentContext | null; } /** @@ -1011,14 +1020,26 @@ export async function saveRequestUsage(entry: UsageEntry, options: SaveRequestUs apiKeyName: entry.apiKeyName || apiKeyContext?.apiKeyName || null, }; + // Agent-session counters are priced before the write (pricing is async) and replayed + // unchanged by the deferred retry; requests without an agent identity skip the await. + let writeOptions: UsageWriteOptions = options; try { - attemptWrite(frozenEntry, options); + if (frozenEntry.agentContext) { + const serviceTier = normalizeServiceTier(entry.serviceTier ?? entry.service_tier); + const agentSession = await sessions.buildAgentSessionUsage( + frozenEntry, + frozenEntry.timestamp, + serviceTier + ); + writeOptions = { ...options, agentSession }; + } + attemptWrite(frozenEntry, writeOptions); } catch (error) { if (!isWriteLockError(error)) { console.error("Failed to save usage stats:", error); return; } - scheduleWriteRetry(frozenEntry, options); + scheduleWriteRetry(frozenEntry, writeOptions); } } diff --git a/src/lib/usage/usageHistory/agentSessionUsage.ts b/src/lib/usage/usageHistory/agentSessionUsage.ts new file mode 100644 index 00000000000..e36fedeb45c --- /dev/null +++ b/src/lib/usage/usageHistory/agentSessionUsage.ts @@ -0,0 +1,101 @@ +/** + * Agent-session attribution for saveRequestUsage (#14833): builds the per-request session + * counters (priced at request time) and upserts the session inside the caller's transaction. + * Both steps are best-effort — a failure degrades to an unattributed usage_history row and + * never drops the row itself. + */ +import { resolveProviderId } from "@/shared/constants/providers"; +import { + hasAgentIdentity, + type AgentContext, +} from "@omniroute/open-sse/handlers/chatCore/agentContext.ts"; +import { + recordAgentSessionUsage, + type AgentSessionTokens, + type AgentSessionUsage, +} from "../../db/agentSessions"; +import { calculateCostDetailed } from "../costCalculator"; +import { + getLoggedInputTokens, + getLoggedOutputTokens, + getPromptCacheCreationTokens, + getPromptCacheReadTokens, + getReasoningTokens, +} from "../tokenAccounting"; + +export type { AgentContext }; + +export type AgentSessionUsageInput = { + agentContext?: AgentContext | null; + tokens?: unknown; + provider?: string | null; + model?: string | null; + apiKeyId?: string | null; + apiKeyName?: string | null; + connectionId?: string | null; + success?: boolean | null; +}; + +/** + * Upsert the request's agent session inside the caller's transaction; null when it has none. + * Errors the caller retries as a whole (`isRetryable`, e.g. a database lock) are rethrown so + * the row and its session are written together on the retry. + */ +export function recordAgentSession( + db: Parameters[0], + usage: AgentSessionUsage | null, + isRetryable: (error: unknown) => boolean = () => false +): string | null { + if (!usage) return null; + try { + return recordAgentSessionUsage(db, usage); + } catch (error) { + if (isRetryable(error)) throw error; + console.warn("Failed to record agent session; saving the usage row without it:", error); + return null; + } +} + +/** Session counters for this request, priced now so reports keep the price at request time. */ +export async function buildAgentSessionUsage( + entry: AgentSessionUsageInput, + timestamp: string, + serviceTier: string +): Promise { + if (!hasAgentIdentity(entry.agentContext)) return null; + // Same token accounting as the usage_history row the session is linked to. + const tokens: AgentSessionTokens = { + input: getLoggedInputTokens(entry.tokens), + output: getLoggedOutputTokens(entry.tokens), + cacheRead: getPromptCacheReadTokens(entry.tokens), + cacheCreation: getPromptCacheCreationTokens(entry.tokens), + reasoning: getReasoningTokens(entry.tokens), + }; + const provider = entry.provider ? resolveProviderId(entry.provider) : null; + const model = entry.model || null; + let pricing: { costUsd: number; priced: boolean }; + try { + pricing = await calculateCostDetailed(provider || "", model || "", tokens, { + provider, + model, + serviceTier, + }); + } catch (error) { + // Pricing only feeds the session counters; never let it drop the usage_history row. + console.warn("Failed to price agent session usage; saving the usage row without it:", error); + return null; + } + return { + context: entry.agentContext, + apiKeyId: entry.apiKeyId || null, + apiKeyName: entry.apiKeyName || null, + timestamp, + success: entry.success !== false, + tokens, + costUsd: pricing.costUsd, + priced: pricing.priced, + provider, + model, + connectionId: entry.connectionId || null, + }; +} diff --git a/src/lib/usage/usageWriteRetry.ts b/src/lib/usage/usageWriteRetry.ts index d0298234e45..c3676202d34 100644 --- a/src/lib/usage/usageWriteRetry.ts +++ b/src/lib/usage/usageWriteRetry.ts @@ -32,6 +32,8 @@ import { getReasoningTokens, } from "./tokenAccounting"; import { backfillDuplicateUsageEntry } from "./usageHistory/duplicateEntry"; +import { recordAgentSession } from "./usageHistory/agentSessionUsage"; +import type { AgentSessionUsage } from "../db/agentSessions"; import type { UsageEntry } from "./usageHistory"; /** Delay before the single deferred attempt. Internal, not user-facing. */ @@ -67,9 +69,14 @@ function isLockMessage(message: string): boolean { return /database(?: table| schema)? is (?:locked|busy)|SQLITE_BUSY|SQLITE_LOCKED/i.test(message); } -/** Legacy natural-key dedup stays on unless the caller owns a physical attempt. */ -type UsageWriteOptions = { +/** + * Legacy natural-key dedup stays on unless the caller owns a physical attempt. + * `agentSession` holds the request's agent-session counters, priced by the caller before + * the write; the session is upserted in the same transaction as the row it is linked to. + */ +export type UsageWriteOptions = { deduplicate?: boolean; + agentSession?: AgentSessionUsage | null; }; /** @@ -81,7 +88,7 @@ type UsageWriteOptions = { */ export function attemptWrite(entry: UsageEntry, options: UsageWriteOptions = {}): void { const db = getDbInstance(); - const prepared = prepareEntry(db, entry); + const prepared = prepareEntry(db, entry, options.agentSession ?? null); let inserted = false; const deduplicate = options.deduplicate !== false; @@ -109,6 +116,7 @@ type PreparedEntry = { billingTeamId: string | null; tokensInput: number; tokensOutput: number; + agentSession: AgentSessionUsage | null; accountIdentity: { accountKey: string | null; accountLabel: string | null; @@ -116,7 +124,11 @@ type PreparedEntry = { }; }; -function prepareEntry(db: ReturnType, entry: UsageEntry): PreparedEntry { +function prepareEntry( + db: ReturnType, + entry: UsageEntry, + agentSession: AgentSessionUsage | null +): PreparedEntry { const timestamp = entry.timestamp || new Date().toISOString(); const connection = entry.connectionId ? (db.prepare("SELECT * FROM provider_connections WHERE id = ?").get(entry.connectionId) as @@ -131,6 +143,7 @@ function prepareEntry(db: ReturnType, entry: UsageEntry): billingTeamId: resolveBillingTeamIdForApiKeyAt(entry.apiKeyId, timestamp), tokensInput: getLoggedInputTokens(entry.tokens), tokensOutput: getLoggedOutputTokens(entry.tokens), + agentSession, accountIdentity: connection ? resolveUsageAccountIdentity(connection) : resolveOrphanedUsageAccountIdentity(entry.provider, entry.connectionId), @@ -168,14 +181,17 @@ function insertEntry(db: ReturnType, prepared: PreparedEnt accountIdentity, } = prepared; const { providerCredits, providerCostUsd } = readReportedCost(entry.provider, entry.tokens); + // Only a row that is actually inserted counts toward its session (dedup no-ops do not). + // A lock fails the whole write, so the deferred retry replays the row and its session. + const agentSessionId = recordAgentSession(db, prepared.agentSession, isWriteLockError); db.prepare( ` INSERT INTO usage_history (provider, model, connection_id, account_key, account_label, account_label_priority, api_key_id, api_key_name, billing_team_id, tokens_input, tokens_output, tokens_cache_read, tokens_cache_creation, tokens_reasoning, service_tier, status, success, latency_ms, ttft_ms, error_code, combo_strategy, endpoint, - cpa_auth_index, timestamp, provider_credits, provider_cost_usd) - VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?) + cpa_auth_index, agent_session_id, timestamp, provider_credits, provider_cost_usd) + VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?) ` ).run( entry.provider ? resolveProviderId(entry.provider) : null, @@ -201,6 +217,7 @@ function insertEntry(db: ReturnType, prepared: PreparedEnt entry.comboStrategy || entry.combo_strategy || null, entry.endpoint || null, entry.cpaAuthIndex || null, + agentSessionId, timestamp, providerCredits, providerCostUsd diff --git a/src/shared/constants/agentSkills.ts b/src/shared/constants/agentSkills.ts index 6aaed16e0b7..34e2cecab6d 100644 --- a/src/shared/constants/agentSkills.ts +++ b/src/shared/constants/agentSkills.ts @@ -55,7 +55,7 @@ export const CURATED_SKILLS: CuratedSkillEntry[] = [ id: "omni-providers", name: "Providers", description: - "Manage provider connections, API keys, OAuth flows, and connection tests via the REST API. List, add, update, remove, and test AI provider integrations across OmniRoute's 383-provider catalog.", + "Manage provider connections, API keys, OAuth flows, and connection tests via the REST API. List, add, update, remove, and test AI provider integrations across OmniRoute's 384-provider catalog.", category: "api", area: "providers", icon: "key", diff --git a/src/shared/constants/config.ts b/src/shared/constants/config.ts index 92743c15861..886ccffe393 100644 --- a/src/shared/constants/config.ts +++ b/src/shared/constants/config.ts @@ -74,6 +74,7 @@ export const PROVIDER_ENDPOINTS = { demonroute: "https://api.demonroute.com/v1/chat/completions", daoxe: "https://api.daoxe.com/v1/chat/completions", gisoo: "https://gisoo.pro/api/v1/chat/completions", + paraloncloud: "https://paraloncloud.com/v1/chat/completions", empiriolabs: "https://api.empiriolabs.ai/v1/chat/completions", openai: "https://api.openai.com/v1/chat/completions", anthropic: "https://api.anthropic.com/v1/messages", diff --git a/src/shared/constants/providers/apikey/inference-hosts.ts b/src/shared/constants/providers/apikey/inference-hosts.ts index 1093972b84f..4d6ac06dd5e 100644 --- a/src/shared/constants/providers/apikey/inference-hosts.ts +++ b/src/shared/constants/providers/apikey/inference-hosts.ts @@ -158,6 +158,20 @@ export const APIKEY_PROVIDERS_INFERENCE = { hasFree: true, freeNote: "$1-5 trial credits on signup for serverless inference", }, + // ParalonCloud (https://paraloncloud.com) — OpenAI-compatible /v1 for chat + embeddings. + paraloncloud: { + id: "paraloncloud", + serviceKinds: ["llm", "embedding"], + alias: "paralon", + name: "ParalonCloud", + icon: "cloud", + color: "#2EE6A6", + textIcon: "PC", + website: "https://paraloncloud.com", + notice: { + apiKeyUrl: "https://paraloncloud.com/console", + }, + }, "ollama-cloud": { id: "ollama-cloud", serviceKinds: ["llm"], diff --git a/tests/snapshots/provider/translate-path.json b/tests/snapshots/provider/translate-path.json index f69f6b9b61e..67a0b87f986 100644 --- a/tests/snapshots/provider/translate-path.json +++ b/tests/snapshots/provider/translate-path.json @@ -5120,6 +5120,29 @@ "stream": "https://oai.endpoints.kepler.ai.cloud.ovh.net/v1/chat/completions" } }, + "paraloncloud": { + "format": "openai", + "headers": { + "apiKey": { + "Accept": "text/event-stream", + "Authorization": "Bearer ", + "Content-Type": "application/json" + }, + "nonStream": { + "Authorization": "Bearer ", + "Content-Type": "application/json" + }, + "oauth": { + "Accept": "text/event-stream", + "Authorization": "Bearer ", + "Content-Type": "application/json" + } + }, + "url": { + "nonStream": "https://paraloncloud.com/v1/chat/completions", + "stream": "https://paraloncloud.com/v1/chat/completions" + } + }, "perplexity": { "format": "openai", "headers": { diff --git a/tests/unit/agent-context-chatcore-usage.test.ts b/tests/unit/agent-context-chatcore-usage.test.ts new file mode 100644 index 00000000000..12d00dba432 --- /dev/null +++ b/tests/unit/agent-context-chatcore-usage.test.ts @@ -0,0 +1,151 @@ +// @ts-nocheck +// End to end: the agent context resolved in handleChatCore must reach saveRequestUsage on every +// path that persists usage — streaming (chatCore/streamingTail.ts), non-streaming +// (chatCore/nonStreamingResponse.ts) and upstream failures (persistFailureUsage). The helper +// tests exercise the leaves in isolation; this guards the wiring between chatCore and them, which +// a conflict-only resolution of #14725's split silently dropped. +import test from "node:test"; +import assert from "node:assert/strict"; +import fs from "node:fs"; +import path from "node:path"; + +const scratchDir = process.env.DATA_DIR || path.resolve("_artifacts/tests"); +fs.mkdirSync(scratchDir, { recursive: true }); +const TEST_DATA_DIR = fs.mkdtempSync(path.join(scratchDir, "omniroute-agent-ctx-chatcore-")); +process.env.DATA_DIR = TEST_DATA_DIR; + +const core = await import("../../src/lib/db/core.ts"); +const { handleChatCore } = await import("../../open-sse/handlers/chatCore.ts"); + +const originalFetch = globalThis.fetch; +const noopLog = () => ({ debug() {}, info() {}, warn() {}, error() {} }); +const sleep = (ms) => new Promise((resolve) => setTimeout(resolve, ms)); + +test.after(() => { + globalThis.fetch = originalFetch; + core.resetDbInstance(); + fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 }); +}); + +function sse(text) { + return new Response(text, { status: 200, headers: { "Content-Type": "text/event-stream" } }); +} + +function chatStream() { + const chunk = (delta, extra = {}) => + `data: ${JSON.stringify({ id: "c1", object: "chat.completion.chunk", created: 1, model: "m", choices: [{ index: 0, delta, finish_reason: null }], ...extra })}\n\n`; + return sse( + chunk({ role: "assistant", content: "" }) + + chunk({ content: "Hello" }) + + `data: ${JSON.stringify({ id: "c1", object: "chat.completion.chunk", created: 1, model: "m", choices: [{ index: 0, delta: {}, finish_reason: "stop" }], usage: { prompt_tokens: 5, completion_tokens: 1, total_tokens: 6 } })}\n\ndata: [DONE]\n\n` + ); +} + +function chatJson() { + return new Response( + JSON.stringify({ + id: "c2", + object: "chat.completion", + created: 1, + model: "gpt-4o-mini", + choices: [ + { index: 0, message: { role: "assistant", content: "Hello" }, finish_reason: "stop" }, + ], + usage: { prompt_tokens: 5, completion_tokens: 1, total_tokens: 6 }, + }), + { status: 200, headers: { "Content-Type": "application/json" } } + ); +} + +function upstreamError() { + return new Response(JSON.stringify({ error: { message: "bad request", type: "invalid" } }), { + status: 400, + headers: { "Content-Type": "application/json" }, + }); +} + +async function drain(response) { + if (!response?.body) return; + const reader = response.body.getReader(); + for (;;) { + const { done } = await reader.read(); + if (done) break; + } +} + +async function sessionRowFor(sessionId, timeoutMs = 15000) { + const db = core.getDbInstance(); + const deadline = Date.now() + timeoutMs; + while (Date.now() < deadline) { + const row = db + .prepare( + `SELECT s.client_session_id, s.project_name, s.request_count, s.error_count, + (SELECT COUNT(*) FROM usage_history u WHERE u.agent_session_id = s.id) AS usage_rows + FROM agent_sessions s WHERE s.client_session_id = ?` + ) + .get(sessionId); + if (row && row.usage_rows > 0) return row; + await sleep(25); + } + return null; +} + +async function run({ stream, upstream, sessionId }) { + globalThis.fetch = async () => upstream(); + const body = { model: "gpt-4o-mini", stream, messages: [{ role: "user", content: "hi" }] }; + const result = await handleChatCore({ + body: structuredClone(body), + modelInfo: { provider: "openai", model: "gpt-4o-mini", extendedContext: false }, + credentials: { apiKey: "sk-test", providerSpecificData: {} }, + log: noopLog(), + clientRawRequest: { + endpoint: "/v1/chat/completions", + body: structuredClone(body), + headers: new Headers({ + accept: stream ? "text/event-stream" : "application/json", + "user-agent": "claude-cli/2.1.282 (external, cli)", + "x-claude-code-session-id": sessionId, + "x-omniroute-project": "acme", + }), + }, + connectionId: "openai-conn", + }); + if (result.success) await drain(result.response); + return { result, row: await sessionRowFor(sessionId) }; +} + +test("streaming usage is attributed to the request's agent session", async () => { + const { result, row } = await run({ + stream: true, + upstream: chatStream, + sessionId: "stream-session", + }); + assert.equal(result.success, true, JSON.stringify(result.error ?? result.status)); + assert.ok(row, "expected an agent_sessions row linked to the streamed usage row"); + assert.equal(row.project_name, "acme"); + assert.equal(row.request_count, 1); + assert.equal(row.usage_rows, 1); +}); + +test("non-streaming usage is attributed to the request's agent session", async () => { + const { result, row } = await run({ + stream: false, + upstream: chatJson, + sessionId: "json-session", + }); + assert.equal(result.success, true, JSON.stringify(result.error ?? result.status)); + assert.ok(row, "expected an agent_sessions row linked to the non-streamed usage row"); + assert.equal(row.project_name, "acme"); + assert.equal(row.usage_rows, 1); +}); + +test("an upstream failure is attributed to the request's agent session", async () => { + const { result, row } = await run({ + stream: false, + upstream: upstreamError, + sessionId: "failure-session", + }); + assert.equal(result.success, false); + assert.ok(row, "expected an agent_sessions row linked to the failure usage row"); + assert.equal(row.error_count >= 1, true); +}); diff --git a/tests/unit/agent-context.test.ts b/tests/unit/agent-context.test.ts new file mode 100644 index 00000000000..8d6211a44f8 --- /dev/null +++ b/tests/unit/agent-context.test.ts @@ -0,0 +1,245 @@ +// Agent context extraction: which coding-agent session a request belongs to and which project +// it works on. Payload shapes mirror real captured traffic (Claude Code 2.1.282 sends its +// working directory in a role:"system" message; Codex sends ). +import test from "node:test"; +import assert from "node:assert/strict"; + +const { extractAgentContext, hasAgentIdentity, projectNameFromPath, resolveUsageAgentContext } = + await import("../../open-sse/handlers/chatCore/agentContext.ts"); +const { forwardOpencodeClientHeaders } = await import("../../open-sse/utils/opencodeHeaders.ts"); +const { clientSessionIdFromRawRequest } = + await import("../../open-sse/handlers/chatCore/clientSessionId.ts"); + +const CLAUDE_SESSION = "fedc860f-461d-4b8b-866d-77477e1d4872"; + +function claudeCodeBody(workingDirectory: string) { + return { + model: "claude-opus-5.5", + metadata: { + user_id: JSON.stringify({ + device_id: "d".repeat(64), + account_uuid: "", + session_id: CLAUDE_SESSION, + }), + }, + system: [{ type: "text", text: "You are an interactive agent." }], + messages: [ + { + role: "user", + content: [ + { + type: "text", + text: "\n# gitStatus\nCurrent branch: fix/token-limits\n", + }, + { type: "text", text: "Run pwd" }, + ], + }, + { + role: "system", + content: [ + { + type: "text", + text: `# Environment\n - Primary working directory: ${workingDirectory}\n - Platform: darwin`, + }, + ], + }, + ], + }; +} + +const claudeHeaders = { + "user-agent": "claude-cli/2.1.282 (external, cli)", + "x-claude-code-session-id": CLAUDE_SESSION, +}; + +test("Claude Code request yields client, session, worktree-safe project and branch", () => { + const context = extractAgentContext( + claudeCodeBody("/Users/dev/work/acme/.claude/worktrees/token-limits"), + new Headers(claudeHeaders) + ); + assert.deepEqual(context, { + client: "claude-code", + clientSessionId: CLAUDE_SESSION, + projectName: "acme", + projectRepo: null, + projectPath: "/Users/dev/work/acme/.claude/worktrees/token-limits", + projectSource: "path", + gitBranch: "fix/token-limits", + }); +}); + +test("explicit project headers win over the prompt-derived project", () => { + const context = extractAgentContext(claudeCodeBody("/Users/dev/work/acme"), { + ...claudeHeaders, + "x-omniroute-project": "omniroute", + "X-OmniRoute-Project-Repo": "github.com/diegosouzapw/OmniRoute", + }); + assert.equal(context.projectName, "omniroute"); + assert.equal(context.projectSource, "header"); + assert.equal(context.projectRepo, "github.com/diegosouzapw/OmniRoute"); + assert.equal(context.projectPath, "/Users/dev/work/acme"); +}); + +test("session id falls back to metadata.user_id JSON when the header is absent", () => { + const context = extractAgentContext(claudeCodeBody("/srv/app"), {}); + assert.equal(context.clientSessionId, CLAUDE_SESSION); +}); + +test("Codex environment_context in a Responses input yields project and session", () => { + const body = { + input: [ + { role: "developer", content: [{ type: "input_text", text: "You are Codex." }] }, + { + role: "user", + content: [ + { + type: "input_text", + text: "\n /home/dev/billing\n zsh\n", + }, + ], + }, + ], + }; + const context = extractAgentContext(body, { + "user-agent": "codex_cli_rs/0.40.0 (Mac OS 15.1; arm64)", + session_id: "codex-123", + }); + assert.equal(context.client, "codex"); + assert.equal(context.clientSessionId, "codex-123"); + assert.equal(context.projectName, "billing"); + assert.equal(context.projectPath, "/home/dev/billing"); +}); + +test("agent session id is the same client session id that call logs record", () => { + // agent_sessions.client_session_id must join call_logs.client_session_id (#17125), so both + // read the id through the same resolver for every supported client. + const requests: Array<{ client: string; headers: Record; body: unknown }> = [ + { client: "claude-code (header)", headers: claudeHeaders, body: claudeCodeBody("/srv/app") }, + { + client: "claude-code (metadata only)", + headers: { "user-agent": "claude-cli/2.1.282 (external, cli)" }, + body: claudeCodeBody("/srv/app"), + }, + { + client: "opencode", + headers: { "user-agent": "opencode/1.2.3", "x-opencode-session": "ses_opencode_1" }, + body: { messages: [{ role: "user", content: "hi" }] }, + }, + { + client: "codex", + headers: { "user-agent": "codex_cli_rs/0.40.0", session_id: "codex-session-9" }, + body: { input: [] }, + }, + ]; + for (const request of requests) { + const callLogId = clientSessionIdFromRawRequest({ + headers: request.headers, + body: request.body, + }); + assert.ok(callLogId, request.client); + assert.equal( + extractAgentContext(request.body, request.headers).clientSessionId, + callLogId, + request.client + ); + assert.equal( + extractAgentContext(request.body, new Headers(request.headers)).clientSessionId, + callLogId, + request.client + ); + } +}); + +test("an explicit x-omniroute-session-id still wins over the client's own session id", () => { + const context = extractAgentContext(claudeCodeBody("/srv/app"), { + ...claudeHeaders, + "x-omniroute-session-id": "pinned-session", + }); + assert.equal(context.clientSessionId, "pinned-session"); +}); + +test("working directory is read from the top-level system field of older builds", () => { + const context = extractAgentContext( + { system: "\nWorking directory: C:\\Users\\dev\\payments-api\n", messages: [] }, + {} + ); + assert.equal(context.projectName, "payments-api"); +}); + +test("prompt text that is not an announced working directory never becomes a project", () => { + const cases = [ + // A user file (e.g. CLAUDE.md) that happens to contain the marker is not trusted. + { messages: [{ role: "user", content: "Docs say:\nWorking directory: /etc/secret" }] }, + // Prose after the marker is not a path. + { system: "Working directory: the repository root", messages: [] }, + // A bare outside Codex's environment envelope is ignored. + { messages: [{ role: "user", content: "/tmp/x" }] }, + ]; + for (const body of cases) { + const context = extractAgentContext(body, {}); + assert.equal(context.projectName, null, JSON.stringify(body)); + assert.equal(context.projectPath, null, JSON.stringify(body)); + } +}); + +test("header values are single-line, control-free and length-capped", () => { + const context = extractAgentContext( + {}, + { "x-omniroute-project": ` evil\r\nx-injected: 1${"a".repeat(300)}` } + ); + assert.ok(context.projectName); + assert.ok(!/[\r\n]/.test(context.projectName)); + assert.equal(context.projectName.length, 120); +}); + +test("scanning stops at the bound, so a marker buried past it is ignored", () => { + const huge = `${"x".repeat(200_000)}\nPrimary working directory: /far/away\n`; + const context = extractAgentContext({ messages: [{ role: "system", content: huge }] }, {}); + assert.equal(context.projectPath, null); +}); + +test("plain API traffic carries no agent identity", () => { + const context = extractAgentContext( + { model: "gpt-4o-mini", messages: [{ role: "user", content: "hi" }] }, + { "user-agent": "curl/8.7.1" } + ); + assert.equal(hasAgentIdentity(context), false); + assert.equal(context.client, "curl"); +}); + +test("noLog keys keep header identifiers but drop everything read from the prompt", () => { + const body = claudeCodeBody("/Users/dev/work/acme"); + const withoutHeader = resolveUsageAgentContext(body, claudeHeaders, { noLog: true }); + assert.equal(withoutHeader.projectName, null); + assert.equal(withoutHeader.projectPath, null); + assert.equal(withoutHeader.gitBranch, null); + assert.equal(withoutHeader.clientSessionId, CLAUDE_SESSION); + + const withHeader = resolveUsageAgentContext( + body, + { ...claudeHeaders, "x-omniroute-project": "acme" }, + { noLog: true } + ); + assert.equal(withHeader.projectName, "acme"); + assert.equal(withHeader.projectPath, null); +}); + +test("projectNameFromPath strips worktree folders and trailing slashes", () => { + const cases: Array<[string, string]> = [ + ["/Users/dev/acme/", "acme"], + ["/Users/dev/acme/.worktrees/feature-x", "acme"], + ["C:\\work\\acme\\.claude\\worktrees\\fix", "acme"], + ]; + for (const [path, expected] of cases) assert.equal(projectNameFromPath(path), expected, path); +}); + +test("project headers are never forwarded to the upstream provider", () => { + const upstream: Record = {}; + forwardOpencodeClientHeaders(upstream, { + "x-omniroute-project": "acme", + "x-omniroute-project-repo": "github.com/org/acme", + }); + const names = Object.keys(upstream).map((name) => name.toLowerCase()); + assert.ok(!names.includes("x-omniroute-project")); + assert.ok(!names.includes("x-omniroute-project-repo")); +}); diff --git a/tests/unit/agent-sessions-usage.test.ts b/tests/unit/agent-sessions-usage.test.ts new file mode 100644 index 00000000000..3dfa81d604c --- /dev/null +++ b/tests/unit/agent-sessions-usage.test.ts @@ -0,0 +1,242 @@ +// Agent sessions: saveRequestUsage aggregates each attributed request into its session row +// (229_agent_sessions) and links the usage row to it (230_usage_history_agent_session_id). +import test from "node:test"; +import assert from "node:assert/strict"; +import fs from "node:fs"; +import path from "node:path"; +import Database from "better-sqlite3"; + +const scratchDir = process.env.DATA_DIR || path.resolve("_artifacts/tests"); +fs.mkdirSync(scratchDir, { recursive: true }); +const TEST_DATA_DIR = fs.mkdtempSync(path.join(scratchDir, "omniroute-agent-sessions-")); +process.env.DATA_DIR = TEST_DATA_DIR; + +const core = await import("../../src/lib/db/core.ts"); +const settingsDb = await import("../../src/lib/db/settings.ts"); +const usageHistory = await import("../../src/lib/usage/usageHistory.ts"); + +test.after(() => { + core.resetDbInstance(); + fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 }); +}); + +type SessionRow = { + id: string; + api_key_id: string | null; + client: string | null; + project_name: string | null; + project_path: string | null; + git_branch: string | null; + first_seen_at: string; + last_seen_at: string; + request_count: number; + error_count: number; + tokens_input: number; + tokens_output: number; + tokens_cache_read: number; + cost_usd: number; + unpriced_count: number; + last_model: string | null; +}; + +function agentContext(overrides: Record = {}) { + return { + client: "claude-code", + clientSessionId: null, + projectName: "acme", + projectRepo: null, + projectPath: "/Users/dev/acme", + projectSource: "path" as const, + gitBranch: "main", + ...overrides, + }; +} + +async function recordUsage(entry: Record) { + await usageHistory.saveRequestUsage({ + provider: "openai", + model: "gpt-4o-mini", + tokens: { input: 1000, output: 500, cacheRead: 200 }, + success: true, + latencyMs: 10, + apiKeyId: "key-alice", + apiKeyName: "Alice", + ...entry, + }); +} + +function sessionsFor(apiKeyId: string): SessionRow[] { + return core + .getDbInstance() + .prepare("SELECT * FROM agent_sessions WHERE api_key_id IS ? ORDER BY first_seen_at") + .all(apiKeyId) as SessionRow[]; +} + +function usageSessionIds(apiKeyId: string): Array { + return ( + core + .getDbInstance() + .prepare("SELECT agent_session_id FROM usage_history WHERE api_key_id IS ? ORDER BY id") + .all(apiKeyId) as Array<{ agent_session_id: string | null }> + ).map((row) => row.agent_session_id); +} + +test.before(async () => { + await settingsDb.updatePricing({ + openai: { + "gpt-4o-mini": { input: 1, output: 2, cached: 0.5, cache_creation: 1, reasoning: 2 }, + }, + }); +}); + +test("requests of one client session aggregate into a single priced session row", async () => { + const context = agentContext({ clientSessionId: "sess-1" }); + await recordUsage({ agentContext: context, timestamp: "2026-09-25T10:00:00.000Z" }); + await recordUsage({ + agentContext: { ...context, gitBranch: "feature/x" }, + timestamp: "2026-09-25T10:05:00.000Z", + success: false, + tokens: { input: 0, output: 0 }, + }); + + const [session] = sessionsFor("key-alice"); + assert.equal(sessionsFor("key-alice").length, 1); + assert.equal(session.request_count, 2); + assert.equal(session.error_count, 1); + assert.equal(session.tokens_input, 1000); + assert.equal(session.tokens_output, 500); + assert.equal(session.tokens_cache_read, 200); + assert.equal(session.first_seen_at, "2026-09-25T10:00:00.000Z"); + assert.equal(session.last_seen_at, "2026-09-25T10:05:00.000Z"); + assert.equal(session.git_branch, "feature/x"); + assert.equal(session.project_name, "acme"); + assert.equal(session.client, "claude-code"); + assert.ok(session.cost_usd > 0, "cost is priced at write time"); + assert.equal(session.unpriced_count, 0); + assert.deepEqual(usageSessionIds("key-alice"), [session.id, session.id]); +}); + +test("the same client session id under another API key is a separate session", async () => { + await recordUsage({ + apiKeyId: "key-bob", + apiKeyName: "Bob", + agentContext: agentContext({ clientSessionId: "sess-1" }), + timestamp: "2026-09-25T10:01:00.000Z", + }); + const [bobSession] = sessionsFor("key-bob"); + const [aliceSession] = sessionsFor("key-alice"); + assert.ok(bobSession); + assert.notEqual(bobSession.id, aliceSession.id); + assert.equal(aliceSession.request_count, 2, "Bob's request never touches Alice's session"); +}); + +test("without a client session id, one key and project share a session within the idle window", async () => { + const context = agentContext({ projectName: "billing", projectPath: "/srv/billing" }); + await recordUsage({ + apiKeyId: "key-carol", + agentContext: context, + timestamp: "2026-09-25T09:00:00.000Z", + }); + await recordUsage({ + apiKeyId: "key-carol", + agentContext: context, + timestamp: "2026-09-25T09:29:00.000Z", + }); + await recordUsage({ + apiKeyId: "key-carol", + agentContext: context, + timestamp: "2026-09-25T10:30:00.000Z", + }); + + const sessions = sessionsFor("key-carol"); + assert.deepEqual( + sessions.map((session) => session.request_count), + [2, 1], + "a gap longer than 30 minutes starts a new session" + ); +}); + +test("a deduplicated usage write does not count the request twice", async () => { + const entry = { + apiKeyId: "key-dave", + agentContext: agentContext({ clientSessionId: "sess-dave" }), + timestamp: "2026-09-25T11:00:00.000Z", + }; + await recordUsage(entry); + await recordUsage(entry); + const [session] = sessionsFor("key-dave"); + assert.equal(session.request_count, 1); +}); + +test("unpriced models are counted instead of silently adding $0", async () => { + await recordUsage({ + apiKeyId: "key-erin", + model: "model-without-pricing-xyz", + agentContext: agentContext({ clientSessionId: "sess-erin" }), + timestamp: "2026-09-25T12:00:00.000Z", + }); + const [session] = sessionsFor("key-erin"); + assert.equal(session.unpriced_count, 1); + assert.equal(session.cost_usd, 0); +}); + +test("traffic without an agent identity creates no session", async () => { + await recordUsage({ apiKeyId: "key-frank", timestamp: "2026-09-25T13:00:00.000Z" }); + await recordUsage({ + apiKeyId: "key-frank", + agentContext: agentContext({ projectName: null, projectPath: null, projectSource: null }), + timestamp: "2026-09-25T13:01:00.000Z", + }); + assert.equal(sessionsFor("key-frank").length, 0); + assert.deepEqual(usageSessionIds("key-frank"), [null, null]); +}); + +test("a failing agent-session write never drops the usage_history row", async () => { + const db = core.getDbInstance(); + // Simulate agent_sessions being unavailable (missing table / failed migration). + db.exec("ALTER TABLE agent_sessions RENAME TO agent_sessions_unavailable"); + try { + await recordUsage({ + apiKeyId: "key-isolation", + apiKeyName: "Isolation", + timestamp: "2026-10-08T10:00:00.000Z", + agentContext: agentContext({ clientSessionId: "isolation-session" }), + }); + } finally { + db.exec("ALTER TABLE agent_sessions_unavailable RENAME TO agent_sessions"); + } + assert.deepEqual(usageSessionIds("key-isolation"), [null]); + assert.deepEqual(sessionsFor("key-isolation"), []); +}); + +test("a write deferred by a database lock records the row and its session once, together", async () => { + const warnings: unknown[] = []; + const warn = console.warn; + console.warn = (...args: unknown[]) => { + warnings.push(args); + }; + // A rival connection holds the write lock, so the first attempt fails and is retried later. + const locker = new Database(path.join(TEST_DATA_DIR, "storage.sqlite")); + locker.exec("BEGIN IMMEDIATE"); + try { + await recordUsage({ + apiKeyId: "key-locked", + timestamp: "2026-10-20T10:00:00.000Z", + agentContext: agentContext({ clientSessionId: "locked-session" }), + }); + assert.deepEqual(usageSessionIds("key-locked"), []); + } finally { + locker.exec("ROLLBACK"); + locker.close(); + } + const deadline = Date.now() + 5000; + while (usageSessionIds("key-locked").length === 0 && Date.now() < deadline) { + await new Promise((resolve) => setTimeout(resolve, 50)); + } + console.warn = warn; + const [session] = sessionsFor("key-locked"); + assert.equal(session?.request_count, 1); + assert.deepEqual(usageSessionIds("key-locked"), [session.id]); + // The lock is not reported as a failed session write that dropped the attribution. + assert.deepEqual(warnings, []); +}); diff --git a/tests/unit/db-migrationrunner-constants-split.test.ts b/tests/unit/db-migrationrunner-constants-split.test.ts index f4ae2175954..35dd64f37ff 100644 --- a/tests/unit/db-migrationrunner-constants-split.test.ts +++ b/tests/unit/db-migrationrunner-constants-split.test.ts @@ -77,8 +77,28 @@ describe("migrationRunner/constants — exact small-table snapshots", () => { // ── large tables — count + shape + spot-checks (corruption guard) ───────────── describe("migrationRunner/constants — large-table integrity", () => { - it("RENAMED_MIGRATION_COMPATIBILITY has 73 well-formed entries", () => { - assert.equal(RENAMED_MIGRATION_COMPATIBILITY.length, 73); + it("RENAMED_MIGRATION_COMPATIBILITY has 87 well-formed entries", () => { + assert.equal(RENAMED_MIGRATION_COMPATIBILITY.length, 87); + assert.deepEqual( + RENAMED_MIGRATION_COMPATIBILITY.filter((entry) => entry.fromName === "agent_sessions"), + ["190", "197", "198", "208", "214", "229", "230"].map((fromVersion) => ({ + fromVersion, + fromName: "agent_sessions", + toVersion: "234", + toName: "agent_sessions", + })) + ); + assert.deepEqual( + RENAMED_MIGRATION_COMPATIBILITY.filter( + (entry) => entry.fromName === "usage_history_agent_session_id" + ), + ["191", "198", "199", "209", "215", "230", "231"].map((fromVersion) => ({ + fromVersion, + fromName: "usage_history_agent_session_id", + toVersion: "235", + toName: "usage_history_agent_session_id", + })) + ); assert.deepEqual( RENAMED_MIGRATION_COMPATIBILITY.filter((entry) => entry.fromName === "daily_cached_input"), ["206", "213", "214", "215", "217", "218", "223", "224", "225", "226", "228", "229"].map( @@ -165,93 +185,93 @@ describe("migrationRunner/constants — large-table integrity", () => { toName: "inspector_custom_hosts", } ); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-15), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-29), { fromVersion: "134", fromName: "ccr_blocks", toVersion: "139", toName: "ccr_blocks", }); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-14), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-28), { fromVersion: "139", fromName: "job_registry", toVersion: "146", toName: "job_registry", }); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-13), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-27), { fromVersion: "143", fromName: "radar_local_model_state", toVersion: "153", toName: "radar_local_model_state", }); // npm 3.8.50 shipped model_capabilities as 163; it now lives at 169 - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-12), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-26), { fromVersion: "163", fromName: "model_capabilities", toVersion: "169", toName: "model_capabilities", }); // #12036: renamed migrations 056/073/077/101 appended as compatibility renames - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-11), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-25), { fromVersion: "056", fromName: "provider_default", toVersion: "056", toName: "mcp_accessibility_compression", }); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-10), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-24), { fromVersion: "073", fromName: "discovery_results", toVersion: "073", toName: "per_model_token_limits", }); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-9), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-23), { fromVersion: "077", fromName: "plugin_metrics", toVersion: "077", toName: "api_key_stream_default_mode", }); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-8), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-22), { fromVersion: "101", fromName: "proxy_pool_rotation", toVersion: "101", toName: "api_key_usage_limits", }); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-7), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-21), { fromVersion: "194", fromName: "api_key_self_service_settings", toVersion: "233", toName: "api_key_self_service_settings", }); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-6), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-20), { fromVersion: "197", fromName: "api_key_self_service_settings", toVersion: "233", toName: "api_key_self_service_settings", }); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-5), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-19), { fromVersion: "207", fromName: "api_key_self_service_settings", toVersion: "233", toName: "api_key_self_service_settings", }); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-4), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-18), { fromVersion: "206", fromName: "recreate_turn_nodes_content_hash_index", toVersion: "211", toName: "recreate_turn_nodes_content_hash_index", }); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-3), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-17), { fromVersion: "213", fromName: "api_key_self_service_settings", toVersion: "233", toName: "api_key_self_service_settings", }); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-2), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-16), { fromVersion: "228", fromName: "api_key_self_service_settings", toVersion: "233", toName: "api_key_self_service_settings", }); - assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-1), { + assert.deepEqual(RENAMED_MIGRATION_COMPATIBILITY.at(-15), { fromVersion: "229", fromName: "api_key_self_service_settings", toVersion: "233", diff --git a/tests/unit/db/migration-233-stack-shift.test.ts b/tests/unit/db/migration-233-stack-shift.test.ts new file mode 100644 index 00000000000..d3a37e241f5 --- /dev/null +++ b/tests/unit/db/migration-233-stack-shift.test.ts @@ -0,0 +1,99 @@ +// The self-service + agent-session stack (#14771, #14833) was published with its migrations at +// 229–231 and later at 228–230, while the release filled those slots with its own migrations. +// The stack now lives at 233 (self-service settings, landed with #14771), 234 (agent_sessions) +// and 235 (usage_history.agent_session_id). A database that ran either published layout must +// move every stack row to its new slot without losing data, and still receive the upstream +// migrations of the vacated slots. +import test from "node:test"; +import assert from "node:assert/strict"; +import fs from "node:fs"; +import path from "node:path"; + +const scratchDir = process.env.DATA_DIR || path.resolve("_artifacts/tests"); +fs.mkdirSync(scratchDir, { recursive: true }); +const TEST_DATA_DIR = fs.mkdtempSync(path.join(scratchDir, "omniroute-stack-shift-233-")); +process.env.DATA_DIR = TEST_DATA_DIR; + +const core = await import("../../../src/lib/db/core.ts"); +const { runMigrations } = await import("../../../src/lib/db/migrationRunner.ts"); + +function resetStorage() { + core.resetDbInstance(); + fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 }); + fs.mkdirSync(TEST_DATA_DIR, { recursive: true }); +} + +test.beforeEach(resetStorage); +test.after(() => { + core.resetDbInstance(); + fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 }); +}); + +const PUBLISHED_LAYOUTS = [ + { selfService: "229", sessions: "230", usage: "231" }, + { selfService: "228", sessions: "229", usage: "230" }, +]; + +for (const layout of PUBLISHED_LAYOUTS) { + const slots = [layout.selfService, layout.sessions, layout.usage]; + test(`a database migrated by the ${slots.join("/")} stack build shifts its rows to 233/234/235`, () => { + const db = core.getDbInstance(); + const now = new Date().toISOString(); + db.prepare( + "INSERT INTO agent_sessions (id, api_key_id, first_seen_at, last_seen_at, request_count) VALUES (?, ?, ?, ?, 2)" + ).run("kept-session", "key-1", now, now); + const upstream = Object.fromEntries( + ( + db + .prepare( + `SELECT version, name FROM _omniroute_migrations WHERE version IN (${slots + .map(() => "?") + .join(", ")})` + ) + .all(...slots) as Array<{ version: string; name: string }> + ).map((row) => [row.version, row.name]) + ); + + // Rewind the ledger to the published layout: the stack's own names in the upstream slots, + // which never ran there, and nothing at 233–235. + db.exec(` + DELETE FROM _omniroute_migrations WHERE version IN ('${slots.join("', '")}'); + UPDATE _omniroute_migrations SET version = '${layout.selfService}' + WHERE version = '233' AND name = 'api_key_self_service_settings'; + UPDATE _omniroute_migrations SET version = '${layout.sessions}' + WHERE version = '234' AND name = 'agent_sessions'; + UPDATE _omniroute_migrations SET version = '${layout.usage}' + WHERE version = '235' AND name = 'usage_history_agent_session_id'; + `); + + runMigrations(db); + + const ledger = db + .prepare( + `SELECT version, name FROM _omniroute_migrations WHERE version IN (${[ + ...slots, + "233", + "234", + "235", + ] + .map(() => "?") + .join(", ")}) ORDER BY version` + ) + .all(...slots, "233", "234", "235"); + assert.deepEqual(ledger, [ + ...[...slots].sort().map((version) => ({ version, name: upstream[version] })), + { version: "233", name: "api_key_self_service_settings" }, + { version: "234", name: "agent_sessions" }, + { version: "235", name: "usage_history_agent_session_id" }, + ]); + const columns = db.prepare("PRAGMA table_info(usage_history)").all() as Array<{ + name: string; + }>; + assert.ok(columns.some((column) => column.name === "agent_session_id")); + // A second boot finds nothing to repair. + assert.doesNotThrow(() => runMigrations(db)); + assert.deepEqual(db.prepare("SELECT id, request_count FROM agent_sessions").all(), [ + { id: "kept-session", request_count: 2 }, + ]); + }); +} diff --git a/tests/unit/db/migration-234-agent-sessions-renumbered.test.ts b/tests/unit/db/migration-234-agent-sessions-renumbered.test.ts new file mode 100644 index 00000000000..5ca01552efa --- /dev/null +++ b/tests/unit/db/migration-234-agent-sessions-renumbered.test.ts @@ -0,0 +1,174 @@ +// Earlier #14833 builds shipped agent_sessions / usage_history.agent_session_id as 190/191, +// 197/198, 198/199, 208/209, 214/215, 230/231 and 229/230 — slots upstream assigns to its own +// migrations. A database that ran one of those builds must move its rows to 234/235, keep its +// sessions, and still receive the upstream migrations of those slots (the runner treats a +// version as applied by number). +import test from "node:test"; +import assert from "node:assert/strict"; +import fs from "node:fs"; +import path from "node:path"; + +const scratchDir = process.env.DATA_DIR || path.resolve("_artifacts/tests"); +fs.mkdirSync(scratchDir, { recursive: true }); +const TEST_DATA_DIR = fs.mkdtempSync(path.join(scratchDir, "omniroute-agent-sessions-renumber-")); +process.env.DATA_DIR = TEST_DATA_DIR; + +const core = await import("../../../src/lib/db/core.ts"); +const { runMigrations } = await import("../../../src/lib/db/migrationRunner.ts"); + +type Db = ReturnType; + +function resetStorage() { + core.resetDbInstance(); + fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 }); + fs.mkdirSync(TEST_DATA_DIR, { recursive: true }); +} + +test.beforeEach(resetStorage); +test.after(() => { + core.resetDbInstance(); + fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 }); +}); + +function columns(db: Db, table: string): string[] { + return (db.prepare(`PRAGMA table_info(${table})`).all() as { name: string }[]).map( + (column) => column.name + ); +} + +function hasObject(db: Db, type: "table" | "index", name: string): boolean { + return Boolean( + db.prepare("SELECT 1 FROM sqlite_master WHERE type = ? AND name = ?").get(type, name) + ); +} + +const HISTORICAL_BUILDS = [ + { + sessions: "190", + usage: "191", + upstream: ["call_logs_content_provenance", "call_logs_resilience_actions"], + undo: ` + ALTER TABLE call_logs DROP COLUMN has_content; + ALTER TABLE call_logs DROP COLUMN usage_provenance; + ALTER TABLE call_logs DROP COLUMN resilience_actions; + `, + restored: (db: Db) => + ["has_content", "usage_provenance", "resilience_actions"].every((name) => + columns(db, "call_logs").includes(name) + ), + }, + { + sessions: "197", + usage: "198", + upstream: ["proxy_subscription_core_config", "proxy_subscription_core_binary"], + undo: ` + ALTER TABLE proxy_subscriptions DROP COLUMN core_config_path; + ALTER TABLE proxy_subscriptions DROP COLUMN selector_last_switch_kind; + ALTER TABLE proxy_subscriptions DROP COLUMN core_binary_path; + `, + restored: (db: Db) => + ["core_config_path", "selector_last_switch_kind", "core_binary_path"].every((name) => + columns(db, "proxy_subscriptions").includes(name) + ), + }, + { + sessions: "198", + usage: "199", + upstream: ["proxy_subscription_core_binary", "proxy_operator_egress"], + undo: ` + ALTER TABLE proxy_subscriptions DROP COLUMN core_binary_path; + DROP TABLE proxy_operator_egress; + `, + restored: (db: Db) => + columns(db, "proxy_subscriptions").includes("core_binary_path") && + hasObject(db, "table", "proxy_operator_egress"), + }, + { + sessions: "208", + usage: "209", + upstream: [ + "proxy_subscription_selector_last_switch_by_selector", + "call_logs_artifact_relpath_index", + ], + undo: ` + ALTER TABLE proxy_subscriptions DROP COLUMN selector_last_switch_by_selector; + DROP INDEX idx_cl_artifact_relpath; + `, + restored: (db: Db) => + columns(db, "proxy_subscriptions").includes("selector_last_switch_by_selector") && + hasObject(db, "index", "idx_cl_artifact_relpath"), + }, + { + sessions: "214", + usage: "215", + upstream: ["call_logs_client_session_id", "team_cost_centers"], + undo: ` + DROP INDEX idx_cl_client_session_id; + ALTER TABLE call_logs DROP COLUMN client_session_id; + DROP INDEX idx_usage_history_billing_team_time; + `, + restored: (db: Db) => + columns(db, "call_logs").includes("client_session_id") && + hasObject(db, "index", "idx_usage_history_billing_team_time"), + }, + { + // 230's CHECK-constrained call_logs columns cannot be dropped; it re-runs as a + // duplicate-column no-op, so only 229's column is removed and checked here. + sessions: "229", + usage: "230", + upstream: ["api_key_five_hour_limit", "call_logs_cache_write_tiers"], + undo: ` + ALTER TABLE api_keys DROP COLUMN five_hour_usage_limit_usd; + `, + restored: (db: Db) => columns(db, "api_keys").includes("five_hour_usage_limit_usd"), + }, + { + sessions: "230", + usage: "231", + upstream: ["call_logs_cache_write_tiers", "daily_cached_input"], + undo: ` + ALTER TABLE daily_usage_summary DROP COLUMN total_cache_read_tokens; + `, + restored: (db: Db) => columns(db, "daily_usage_summary").includes("total_cache_read_tokens"), + }, +]; + +for (const build of HISTORICAL_BUILDS) { + test(`agent-session migrations ${build.sessions}/${build.usage} move to 234/235 and free the slots for ${build.upstream.join(" + ")}`, () => { + const db = core.getDbInstance(); + const now = new Date().toISOString(); + db.prepare( + "INSERT INTO agent_sessions (id, api_key_id, first_seen_at, last_seen_at, request_count) VALUES (?, ?, ?, ?, 3)" + ).run("kept-session", "key-1", now, now); + // Rewind the ledger to what that build left behind: its own names in the upstream slots, + // which never ran there, and nothing at 234/235. + db.exec(` + DELETE FROM _omniroute_migrations WHERE version IN ('${build.sessions}', '${build.usage}'); + UPDATE _omniroute_migrations SET version = '${build.sessions}' + WHERE version = '234' AND name = 'agent_sessions'; + UPDATE _omniroute_migrations SET version = '${build.usage}' + WHERE version = '235' AND name = 'usage_history_agent_session_id'; + ${build.undo} + `); + assert.equal(build.restored(db), false); + + runMigrations(db); + + const ledger = db + .prepare( + "SELECT version, name FROM _omniroute_migrations WHERE version IN (?, ?, '234', '235') ORDER BY version" + ) + .all(build.sessions, build.usage); + assert.deepEqual(ledger, [ + { version: build.sessions, name: build.upstream[0] }, + { version: build.usage, name: build.upstream[1] }, + { version: "234", name: "agent_sessions" }, + { version: "235", name: "usage_history_agent_session_id" }, + ]); + assert.equal(build.restored(db), true); + assert.ok(columns(db, "usage_history").includes("agent_session_id")); + assert.deepEqual(db.prepare("SELECT id, request_count FROM agent_sessions").all(), [ + { id: "kept-session", request_count: 3 }, + ]); + }); +} diff --git a/tests/unit/db/migration-234-agent-sessions.test.ts b/tests/unit/db/migration-234-agent-sessions.test.ts new file mode 100644 index 00000000000..c638f20e84d --- /dev/null +++ b/tests/unit/db/migration-234-agent-sessions.test.ts @@ -0,0 +1,112 @@ +// 234 creates agent_sessions; 235 links usage_history rows to it. They are separate files because +// the runner records a file as applied when an ALTER hits "duplicate column name" and rolls the +// rest of that file back: a database that already has the column must still get the table. +import test from "node:test"; +import assert from "node:assert/strict"; +import fs from "node:fs"; +import path from "node:path"; +import { fileURLToPath } from "node:url"; +import Database from "better-sqlite3"; + +const repoMigrations = path.join( + path.dirname(fileURLToPath(import.meta.url)), + "../../../src/lib/db/migrations" +); +const MIGRATION_FILES = ["234_agent_sessions.sql", "235_usage_history_agent_session_id.sql"]; +const scratchDir = process.env.DATA_DIR || path.resolve("_artifacts/tests"); +fs.mkdirSync(scratchDir, { recursive: true }); +const migrationsDir = fs.mkdtempSync(path.join(scratchDir, "omniroute-migration-234-")); +for (const file of MIGRATION_FILES) { + fs.copyFileSync(path.join(repoMigrations, file), path.join(migrationsDir, file)); +} +const originalMigrationsDir = process.env.OMNIROUTE_MIGRATIONS_DIR; +process.env.OMNIROUTE_MIGRATIONS_DIR = migrationsDir; + +const { runMigrations } = await import("../../../src/lib/db/migrationRunner.ts"); + +test.after(() => { + fs.rmSync(migrationsDir, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 }); + if (originalMigrationsDir === undefined) delete process.env.OMNIROUTE_MIGRATIONS_DIR; + else process.env.OMNIROUTE_MIGRATIONS_DIR = originalMigrationsDir; +}); + +function openDb(withAgentSessionColumn: boolean): Database.Database { + const db = new Database(":memory:"); + db.exec( + `CREATE TABLE usage_history (id INTEGER PRIMARY KEY, api_key_id TEXT, timestamp TEXT${ + withAgentSessionColumn ? ", agent_session_id TEXT" : "" + });` + ); + return db; +} + +function tableExists(db: Database.Database, name: string): boolean { + return Boolean( + db.prepare("SELECT 1 FROM sqlite_master WHERE type = 'table' AND name = ?").get(name) + ); +} + +function usageColumns(db: Database.Database): string[] { + return (db.prepare("PRAGMA table_info(usage_history)").all() as Array<{ name: string }>).map( + (column) => column.name + ); +} + +function appliedVersions(db: Database.Database): string[] { + return ( + db.prepare("SELECT version FROM _omniroute_migrations ORDER BY version").all() as Array<{ + version: string; + }> + ).map((row) => row.version); +} + +test("an older database gains agent_sessions and usage_history.agent_session_id; a rerun is a no-op", () => { + const db = openDb(false); + try { + assert.equal(runMigrations(db, { isNewDb: true }), 2); + assert.ok(tableExists(db, "agent_sessions")); + assert.ok(usageColumns(db).includes("agent_session_id")); + assert.equal(runMigrations(db, { isNewDb: true }), 0); + assert.deepEqual(appliedVersions(db), ["234", "235"]); + } finally { + db.close(); + } +}); + +test("a database that already has the column still gets the agent_sessions table", () => { + const db = openDb(true); + try { + assert.equal(runMigrations(db, { isNewDb: true }), 2); + assert.ok(tableExists(db, "agent_sessions")); + assert.equal(usageColumns(db).filter((name) => name === "agent_session_id").length, 1); + assert.deepEqual(appliedVersions(db), ["234", "235"]); + } finally { + db.close(); + } +}); + +test("235 does not duplicate the (api_key_id, timestamp) index that 051 already creates", () => { + const db = openDb(false); + try { + // 051_hot_path_db_indexes already ships this exact index on every install. + db.exec( + "CREATE INDEX idx_usage_history_api_key_id_timestamp ON usage_history(api_key_id, timestamp)" + ); + runMigrations(db, { isNewDb: true }); + const indexes = db.prepare("PRAGMA index_list(usage_history)").all() as Array<{ name: string }>; + const sameColumns = indexes.filter((index) => { + const columns = ( + db.prepare(`PRAGMA index_info(${JSON.stringify(index.name)})`).all() as Array<{ + name: string; + }> + ).map((column) => column.name); + return columns.join(",") === "api_key_id,timestamp"; + }); + assert.deepEqual( + sameColumns.map((index) => index.name), + ["idx_usage_history_api_key_id_timestamp"] + ); + } finally { + db.close(); + } +}); diff --git a/tests/unit/provider-node-reserved-prefix.test.ts b/tests/unit/provider-node-reserved-prefix.test.ts index 30ea2c9d7b7..fb3a1123840 100644 --- a/tests/unit/provider-node-reserved-prefix.test.ts +++ b/tests/unit/provider-node-reserved-prefix.test.ts @@ -226,7 +226,8 @@ test("shared set size includes live REGISTRY and retired Designer + Felo + Qwen // MiniMax Code (#17286) registers id "minimax-code" and alias "mmcode" — two // more (436 -> 438). // EmpirioLabs AI (#17209) registers id "empiriolabs" with the same alias (+1: 438 -> 439). - assert.equal(RESERVED_PREFIX_COUNT, 439); + // ParalonCloud registers id "paraloncloud" plus alias "paralon" (+2: 439 -> 441). + assert.equal(RESERVED_PREFIX_COUNT, 441); assert.equal(RESERVED_PROVIDER_PREFIXES.has("ainetcafe"), true); assert.equal(RESERVED_PROVIDER_PREFIXES.has("tokenharbor"), true); assert.equal(RESERVED_PROVIDER_PREFIXES.has("runcrate"), true); diff --git a/tests/unit/provider-paraloncloud-registry.test.ts b/tests/unit/provider-paraloncloud-registry.test.ts new file mode 100644 index 00000000000..9d194096c51 --- /dev/null +++ b/tests/unit/provider-paraloncloud-registry.test.ts @@ -0,0 +1,75 @@ +// ParalonCloud (https://paraloncloud.com) — OpenAI-compatible API-key host covering chat +// (/v1/chat/completions), embeddings (/v1/embeddings) and a key-gated /v1/models list. +import test from "node:test"; +import assert from "node:assert/strict"; + +const { APIKEY_PROVIDERS, resolveProviderId } = + await import("../../src/shared/constants/providers.ts"); +const { PROVIDER_ENDPOINTS } = await import("../../src/shared/constants/config.ts"); +const { REGISTRY: providerRegistry } = await import("../../open-sse/config/providerRegistry.ts"); +const { resolveProviderAlias } = await import("../../open-sse/services/providerAlias.ts"); +const { getEmbeddingProvider, getEmbeddingDimension, parseEmbeddingModel } = + await import("../../open-sse/config/embeddingRegistry.ts"); +const { isNamedOpenAIStyleProvider } = + await import("../../src/app/api/providers/[id]/models/discovery/providerSets.ts"); +const { DefaultExecutor, getExecutor } = await import("../../open-sse/executors/index.ts"); + +const PARALON_API = "https://paraloncloud.com/v1"; + +test("paraloncloud is registered as an API-key provider in the UI catalog", () => { + const entry = APIKEY_PROVIDERS.paraloncloud; + assert.ok(entry, "APIKEY_PROVIDERS.paraloncloud must be defined"); + assert.equal(entry.id, "paraloncloud"); + assert.equal(entry.alias, "paralon"); + assert.equal(entry.name, "ParalonCloud"); + assert.equal(entry.website, "https://paraloncloud.com"); + assert.equal(entry.textIcon, "PC"); + assert.deepEqual(entry.serviceKinds, ["llm", "embedding"]); + assert.equal(entry.notice?.apiKeyUrl, "https://paraloncloud.com/console"); +}); + +test("paraloncloud chat registry entry is OpenAI-compatible with live model discovery", () => { + const entry = providerRegistry.paraloncloud; + assert.ok(entry, "providerRegistry.paraloncloud must be defined"); + assert.equal(entry.id, "paraloncloud"); + assert.equal(entry.alias, "paralon"); + assert.equal(entry.format, "openai"); + assert.equal(entry.executor, "default"); + assert.equal(entry.authType, "apikey"); + assert.equal(entry.authHeader, "bearer"); + assert.equal(entry.baseUrl, `${PARALON_API}/chat/completions`); + assert.equal(entry.modelsUrl, `${PARALON_API}/models`); + assert.equal(PROVIDER_ENDPOINTS.paraloncloud, `${PARALON_API}/chat/completions`); + const ids = entry.models.map((m: { id: string }) => m.id); + assert.deepEqual(ids, ["qwen3.8-27b", "qwen3-3b"]); + assert.equal(isNamedOpenAIStyleProvider("paraloncloud"), true); +}); + +test("paralon alias resolves to the paraloncloud provider id", () => { + assert.equal(resolveProviderId("paralon"), "paraloncloud"); + assert.equal(resolveProviderAlias("paralon"), "paraloncloud"); + assert.equal(resolveProviderAlias("paraloncloud"), "paraloncloud"); +}); + +test("paraloncloud default executor targets the chat completions endpoint", async () => { + assert.ok((await getExecutor("paraloncloud")) instanceof DefaultExecutor); + assert.equal( + new DefaultExecutor("paraloncloud").buildUrl("qwen3-3b", false), + `${PARALON_API}/chat/completions` + ); +}); + +test("paraloncloud resolves in the embedding registry", () => { + const embedding = getEmbeddingProvider("paraloncloud"); + assert.equal(embedding?.baseUrl, `${PARALON_API}/embeddings`); + assert.equal(embedding?.authHeader, "bearer"); + assert.equal(getEmbeddingDimension("paraloncloud/multilingual-e5-small"), 384); + assert.deepEqual(parseEmbeddingModel("paraloncloud/multilingual-e5-small"), { + provider: "paraloncloud", + model: "multilingual-e5-small", + }); + assert.deepEqual(parseEmbeddingModel("paralon/multilingual-e5-small"), { + provider: "paraloncloud", + model: "multilingual-e5-small", + }); +}); diff --git a/tests/unit/providers-constants-split.test.ts b/tests/unit/providers-constants-split.test.ts index 019c9c688df..1f0f122c113 100644 --- a/tests/unit/providers-constants-split.test.ts +++ b/tests/unit/providers-constants-split.test.ts @@ -65,7 +65,8 @@ const P = await import("../../src/shared/constants/providers.ts"); // Gisoo (#18112) adds one apikey/gateways entry — 255. // deAPI adds one apikey/specialty-media entry — 256. // EmpirioLabs AI (#17209) adds one apikey entry in its own extract, apikey/empiriolabs.ts: 257. -const APIKEY_PROVIDER_COUNT = 257; +// ParalonCloud adds one apikey/inference-hosts entry — 258. +const APIKEY_PROVIDER_COUNT = 258; test("barrel still exports every catalog + key helpers", () => { for (const name of [