## Summary
- `GET`/`PUT /api/v1/providers/[provider]/limits` and `GET`/`PUT /api/v1/accounts/[id]/limits` now use `requireManagementAuth()`, the guard the neighbouring `/api/v1/registered-keys` and `/api/v1/management/*` routes already use. These caps are what `POST /api/v1/registered-keys` checks before it issues a key, so they take the same credential: a dashboard session, the local CLI token, or an API key with `manage`/`admin` scope. Other keys get `403` through `createErrorResponse()`.
- `POST /api/v1/issues/report` files a GitHub issue with the operator's `GITHUB_ISSUES_TOKEN`, so it now takes the same management credential.
- Allowed response shapes are unchanged. An instance with login disabled stays open, as before.
- `GET /api/v1/quotas/check` keeps its documented Bearer-key access, and a regression test pins it.
- Docs: `docs/openapi.yaml` (plus the matching four entries in `src/app/docs/lib/openapi.generated.ts`), `docs/reference/API_REFERENCE.md` and `docs/security/CORS.md` now list the auth tier for these routes, including `/api/v1/issues/report`.
## Related Issues
- None.
## Target Branch
## Summary
- Enforced at the same policy chokepoint as the daily/weekly limits, with the same fail-closed rule for unpriced usage (`USAGE_LIMIT_IGNORE_UNPRICED` opt-out). Anthropic clients keep the 400 envelope; other clients get `429 usage_limit_exceeded` with `retry-after` at the 5 h reset.
- `/v1/me/status` gains `usage.limits` (`fiveHour` / `daily` / `weekly`) for keys with usage limits on; the dashboard Limits tab gets a 5-hour quota input.
## Attribution
Thanks to [@sajjadsaharkhan](https://github.com/sajjadsaharkhan) for the original implementation.
## Changes
- `src/lib/db/migrations/217_api_key_five_hour_limit.sql` — `api_keys.five_hour_usage_limit_usd REAL` (+ column fallback in `apiKeyColumnFallbacks.ts`).
- `src/lib/db/apiKeyUsageLimitFields.ts`, `apiKeys.ts`, `apiKeys/permissionsUpdate.ts` — read/write `fiveHourUsageLimitUsd` (metadata SELECT, update path, env-key defaults).
- `src/lib/usage/apiKeyUsageLimits.ts` — 5 h window over `now - 5h` reusing `getApiKeyUsdSpendSince`; `MIN(timestamp)` for the reset; the extra query only runs when a 5 h limit is set. Exceeded message / tripped reset (daily → weekly → 5-hour precedence); `@@om-usage` text gets a 5-hour block only when the limit is configured (existing output unchanged otherwise).
- `src/shared/utils/apiKeyPolicy.ts`, `src/shared/validation/schemas/keys.ts`, `src/app/api/keys/route.ts`, `src/app/api/keys/[id]/route.ts`, `src/app/api/keys/[id]/usage-limits/route.ts`, `src/lib/usage/internalUsageCommand.ts
## Summary
- Upgrades the Microsoft 365 Copilot (`copilot-m365-web`) provider from captured-HAR credentials to a native work/school sign-in: PKCE authorization against the Microsoft identity platform, refresh-token rotation through the shared refresh mutex, and account-safe reauthentication that refuses to bind a different tenant/user to an existing connection. HAR connections keep working.
- Rebuilds the turn transport (`m365Turn.ts`): one bounded SignalR turn with size, idle and total deadlines, strict handshake, sanitized errors, and explicit classification of Copilot's final result: model licence (`403 m365_model_not_licensed`, model-scoped lock), account licence (`403 m365_account_not_licensed`), per-user throttle (`429 m365_throttled`, connection cooldown) and priority-access exhaustion (`429 priority_access_exhausted`, model-scoped, with the UTC reset as `Retry-After`).
- Copilot safety refusals (`Disengaged`) surface as `400 provider_safety_rejection`, the shared provider safety-stop label: never counted against connection health, not retried by client SDKs, and with `PROVIDER_SAFETY_STOPS_TERMINAL` enabled not replayed against another combo target.
- Registers all 34 documented model IDs plus four legacy IDs (tone, scenario and licence per model id), independently of the connected account’s entitlement. Display labels omit the redundant provider prefix and availability suffix; compatibility aliases retain their IDs. Microsoft decides availability.
- Streaming compos
Thanks @aldoeliacim.
## Current checkpoint — 2026-10-16
Draft / HOLD for admission. Current head: `fd52e61bb1499675b21d7c2c5243da04ad9bf16a`.
The owner selected this stack for v3.8.52. Original ancestry is preserved.
### Repairs and scoped evidence
- Empty focused selection no longer revokes a still-retained artifact; later reads, deletion and expiry remain available. Reproduced RED then GREEN on the designated VPS.
- Diagnostic logging strips ephemeral drill-down handles while preserving runtime metadata.
- Both public drill-down route aliases now belong to the existing video endpoint category. Restricted keys are checked for GET and DELETE; unrestricted-key behavior is preserved.
- Added in-flight cancellation/deletion, empty-selection idle expiry and warm-result-cache consent/ownership tests.
- Clean `3d9c50197d`: 518 backend tests / 63 files and 33 actual UI tests / 6 files PASS, no failures, cancellations or skips. Core/dashboard typechecks and changelog gate PASS.
- Five drill-down controls reviewed across 66 non-English locales; 64 catalogs changed, two already-translated catalogs preserved. Clean `fd52e61bb1`: all eight official UI/docs gates PASS without skip: key completeness, translation ratio, new-key coverage, UI coverage, value drift, ICU, the actual 33-test UI lane and docs-all. Source fingerprints remained clean.
### Still required
Generated skills and strict documentation translation drift are FAIL on the recorded candidate; full current-base/combined-candidate admission, re
## Summary
`searxng-search` is `fallbackOnly` (#9543) so the catalog default `http://localhost:8888/search` is never auto-selected, and #10976 skips that default outright. Both guard the *unconfigured* case. An operator who runs their own SearXNG and stores a connection with its real base URL still only got it after every keyed provider: with a Brave, Tavily or Firecrawl key also configured, `/v1/search` and the `web_search` fallback always picked the paid provider, and the local, free, unrate-limited engine served nothing.
This adds a registry flag, `primaryWhenConfigured`, which only `searxng-search` sets. When the provider's stored connection names a real `baseUrl` (anything but the catalog default), auto-select ranks it by cost alongside the keyed providers. That puts the $0 SearXNG first, with the cheapest keyed provider as its failover. With no connection, or with the catalog default, nothing changes. A connection that stores a token is auto-selected only over `https://` or loopback `http://`; on remote plain `http://` it stays explicit-only, so auto-selection never sends the bearer token in cleartext (tokenless cluster-internal `http://` is still auto-selected). `duckduckgo-free` stays last-resort only, and an explicit `provider` behaves as before.
- `open-sse/config/searchRegistry.ts`: `primaryWhenConfigured`, `isOperatorConfiguredPrimary`, `autoSearchCandidates`, `admitsAutoCandidate`.
- `src/app/api/v1/search/route.ts` and `src/lib/search/executeWebSearch.ts`: bo
Thanks @aldoeliacim.
## What
`GET /api/init` was in `PUBLIC_API_ROUTES_EXACT` and the handler had no auth of its own, so an anonymous request passed the middleware even with `requireLogin=true`. The first call that gets through starts the cloud-sync scheduler, model sync, and the budget-reset, token-health and log_export jobs. Boot already starts these from `instrumentation-node.ts`, which calls `ensureCloudSyncInitialized()` directly. The only caller of the route is the CLI command `get-api-init`.
On a headless server (`OMNIROUTE_HEADLESS=1`) boot skips cloud-sync, so there the anonymous call did real work: it started everything boot had left off.
Change:
- `/api/init` is removed from the public list, so it is now a MANAGEMENT route. The handler also calls `requireManagementAuth(request)` before doing anything.
- With `requireLogin=true`, an anonymous request now gets 401 and starts nothing, in headless mode too.
- Headless behaviour is otherwise unchanged. An authenticated `/api/init` still starts cloud-sync and its three jobs later, as before; it is the only way to get those jobs on a headless server. `HEADLESS.md` now says so.
- Docs: `AUTHZ_GUIDE.md` (the public list and the `PUBLIC` row of the route-class table), `API_REFERENCE.md`, `HEADLESS.md`. `tests/unit/authz/public-policy.test.ts` used `/api/init` as its sample public path and now uses `/api/auth/status`.
Thanks @HouMinXi.
Validated in the full merge-train on the .113 (spawn and endpoint-adapter fixes re-tested on the merged head). Live Gemini Live check deferred to the release-generation phase by the owner.
Keep every Exa highlight in provider_raw.
Evidence: focused tests on the combined board 62/62; file-size, changelog integrity, check-deps, lint, typecheck:core, open-sse-typecheck and api-typecheck OK against origin/release/v3.8.53.
`GET /api/models?withLimits=1` exposes each token limit together with the source that supplied it; the parameter is opt-in and Zod-validated.
Evidence: focused tests passed on a combined board (model-limit-provenance.test.ts); lint, typecheck:core and check-deps OK; open-sse-typecheck OK; api-typecheck reports only errors in files this batch does not touch, reproduced on the pure tip; file-size OK; changelog integrity OK against origin/release/v3.8.53.
⚠️ base-red inherited: #17219
Merged via /merge-batch (2026-10-17, round 4). Validated as submitted on a combined board of 33 PRs over the current release/v3.8.52 tip: typecheck:core, open-sse and API-route typechecks, eslint on changed files, file-size, complexity, cognitive-complexity, dead-code, boundary 774/774, changelog-integrity, env-doc-sync, docs-sync, agent-skills-sync (Generated 0), provider-consistency and i18n new-key coverage all green; focused tests 401/406 node:test + 11/11 vitest (the only reds were the #16988 fixture, fixed separately). Remaining CI reds are the inherited base-red #15306. Thanks @PrinceThummar011 for the contribution!
Merged via /merge-batch (2026-10-17 wave). Reconciled on the current tip; PROVIDER_REFERENCE.md regenerated (gen:provider-reference, 375 unique IDs) and the public provider count bumped 374 -> 375 (owner-approved, count-only: every protected file — README, AGENTS.md, llm.txt + 66 mirrors, package.json, skills omni-auth/omni-providers SKILL.md — differs from the tip only in digits, verified file by file). Focused tests 129/129; provider-consistency OK (375); docs-counts no strict drift (2 soft executor-count drifts inherited); agent-skills-sync Generated 0; docs-sync PASS; changelog-integrity OK; boundary 774/774; api-typecheck within baseline. Remaining CI reds are the inherited base-red #15306.
Merged via /merge-batch (2026-10-17 wave). Reconciled on the current tip and regenerated skills/omni-models/SKILL.md through the generator (owner-approved protected surface; generated text only: the GET /v1/models filter note and the new POST /api/v1/models/query section). Focused tests 22/22 (discovery filters + query); check-api-typecheck OK (the earlier 3x TS2339 in catalogDiscovery.ts are fixed), agent-skills-sync Generated 0, changelog-integrity OK, boundary 774/774, complexity/cognitive within baseline. Remaining CI reds are the inherited base-red #15306.
Merged via /merge-batch (2026-10-17 wave, maintainer PRs). Validated on the 2026-10-17 combined board of 26 maintainer PRs over release/v3.8.52@3c3edb9ef: typecheck:core, open-sse typecheck, eslint on changed files, file-size, changelog-integrity, complexity, cognitive-complexity, env-doc-sync and agent-skills-sync all green; focused tests 618/618 (node:test) + 17/17 (vitest UI); boundary 774/774 on the combined tree (#16810 removes one engine->app edge, #16933 adds one, merged in that order); api-typecheck, dead-code, provider-consistency, docs-sync and i18n new-key coverage checked. Remaining CI reds are the inherited base-red #15306.
Merged via /merge-batch (2026-10-17 wave, maintainer PRs). Validated on the 2026-10-17 combined board of 26 maintainer PRs over release/v3.8.52@3c3edb9ef: typecheck:core, open-sse typecheck, eslint on changed files, file-size, changelog-integrity, complexity, cognitive-complexity, env-doc-sync and agent-skills-sync all green; focused tests 618/618 (node:test) + 17/17 (vitest UI); boundary 774/774 on the combined tree (#16810 removes one engine->app edge, #16933 adds one, merged in that order); api-typecheck, dead-code, provider-consistency, docs-sync and i18n new-key coverage checked. Remaining CI reds are the inherited base-red #15306.
⚠️ base-red inherited: #15306
## Summary
Authorization boundary fix on the destructive side of the backup surface. Restoring a backup became operator-only in #16899; destroying restore points still accepted any management credential: `DELETE /api/db-backups` (cleanup — keep the latest N / drop older than D days) gated on `isAuthenticated()` only, so a `manage`-scoped API key or an `oma_` access token could delete every stored restore point but one. Restore points are the recovery path after exactly that kind of leaked machine credential.
**`PATCH /api/db-backups` is included on purpose and flagged for the owner's review.** The owner approved gating `DELETE`; `PATCH` is the closure of the same hole: it persists the same `keepLatest` / `retentionDays`, and `backupDbFile()` calls `cleanupDbBackups()` with no arguments after every backup (`src/lib/db/backup.ts:414`), which falls back to those persisted values (`backup.ts:84-91`). Gating `DELETE` alone is bypassable — a machine credential saves `keepLatest: 1`, then triggers a manual backup (`PUT`, still management-reachable) or simply waits for the next automatic one, and the pruning happens anyway.
Both handlers now go through the existing `requireOperatorCredential()` (`src/lib/api/operatorCredential.ts`, #16680) with route-local 403 messages — no new export was added to the shared helper and `backup.ts` semantics are untouched.
## Behavior before / after
| Handler | Credential | Before | After |
| --- | --- | --- | --- |
| `DELETE /api/db-backups` | none | 401 | 401 (unchanged) |
| `DELETE /api/db-backups` | `manage` API key / `oma_` admin token | **200 — backups pruned** | **403**, settings and files untouched |
| `DELETE /api/db-backups` | inference-only API key | 403 (`manage` scope) | 403 (unchanged) |
| `DELETE /api/db-backups` | dashboard session / login disabled | 200 | 200 (unchanged) |
| `DELETE /api/db-backups` | pipeline-stamped local CLI subject | 401 (`isAuthenticated()` never honored the stamp) | 200 |
| `PATCH /api/db-backups` | none | 401 | 401 (unchanged) |
| `PATCH /api/db-backups` | `manage` API key / `oma_` admin token | **200 — retention persisted (drives automatic pruning)** | **403**, settings untouched |
| `PATCH /api/db-backups` | dashboard session / login disabled | 200 | 200 (unchanged) |
| `PATCH /api/db-backups` | pipeline-stamped local CLI subject | 401 | 200 |
| `GET` / `PUT /api/db-backups` | `manage` API key | 200 | 200 (unchanged — not destructive, covered by positive tests) |
The CLI has no generated command for `PATCH`/`DELETE /api/db-backups` (`bin/cli/api-commands/system.mjs` only exposes `get-`/`post-api-db-backups`; `bin/cli/commands/backup.mjs` prunes its own local export directory), so nothing in `bin/` changes.
## Tests
New sibling file `tests/unit/backup-cleanup-require-operator.test.ts` (same style as `tests/unit/backup-restore-require-operator.test.ts`): handlers invoked directly with a crafted `Request`, isolated `DATA_DIR`, four seeded restore points, retention settings read back before/after. Matrix per handler: no credential → 401; `manage` key, inference-only key, admin `oma_` token → 403 with the retention settings unchanged and all four files still on disk; dashboard session / local CLI subject / login disabled → 200 and the handler really does its job (PATCH persists and never prunes; DELETE prunes down to the requested count). Positive `GET` and `PUT` with a `manage` key → 200.
- **RED** (commit `3429f715c0`, on the unfixed route): `ℹ tests 17 / pass 5 / fail 12` — root failure `actual: 200, expected: 403` on the manage-key `DELETE`; the unguarded handler then persisted `1/1` and pruned all four files, cascading into every later "nothing changed" assertion; plus `401 vs 200` for the local CLI subject on both handlers.
- **GREEN** (fix commit `b9aba368db`): `17/17`; on the merged tree together with `backup-restore-require-operator` + `db-backup-extended`: `56/56`.
- **Related suites** (`grep -rlE "db-backup|backup-restore|backup-export|operatorCredential|credential-export|cli-backup" tests/` → 30 files): 29 node:test files in 3 batches → 75/75, 117/117, 127/127; 1 vitest file (`tests/unit/ui/system-storage-backup-auth-envelope-14846.test.tsx`) → 2/2. No assertion removed or weakened anywhere.
## Validation
| Check | Result |
| --- | --- |
| `npx prettier --check` (7 touched files) | OK |
| `npx eslint --suppressions-location config/quality/eslint-suppressions.json` (route + test) | 0 errors |
| `check-complexity-ratchets.mjs --base-ref <merge-base>` | OK (complexity 0 / cognitive 0 on new code) |
| `check:file-size --base-ref <merge-base>` | OK (no frozen file grown) |
| `scripts/dev/typecheck-touched.sh --base origin/release/v3.8.52` | 0 errors in touched files; 1 error in `tests/helpers/managementSession.ts(62,5)` TS2322 — file byte-identical to the base (inherited, same helper #16899 imports) |
| `check:openapi-security-tiers` / `check:api-docs-refs` / `check:openapi-coverage` / `check:changelog-integrity` | PASS |
| `check:docs-all` | docs-sync, frontmatter, doc-links PASS; `check:fabricated-docs` fails on `/api/balance` in `docs/reference/ENVIRONMENT.md:368` — file identical to the base (inherited) |
| i18n doc mirrors (`npm run i18n:run -- --files=<3 docs>`) | translation backend answered **404** (`DEPLOYMENT_NOT_FOUND`), nothing written. The drift gate was already red on the base with 34 drifted sources, and all three docs I edited were already among them (state hash ≠ base content), so this PR adds no new drift entry; the 65 mirrors still need a real `i18n:run` once the backend is back. No blanket `--adopt` (it would mask the 34 genuine stale sources). |
## Removed / rewritten assertions, allowlist and suppression deltas
None. `config/quality/eslint-suppressions.json` and `config/quality/test-masking-allowlist.json` untouched.
## References
Refs #16899 (precedent: restoring a backup requires an operator credential). No issue to close.
## Out of scope
- `docs/openapi.yaml` has no `delete` operation for `/api/db-backups` at all (only `get`/`post`/`patch`); adding one would change what the next `npm run build:cli-api` generates (a new `delete-api-db-backups` CLI command + locale keys), so the 403 is documented on `patch` only.
- `bin/cli/api-commands/system.mjs` still describes `post-api-db-backups` as "Create database backup" (the pre-#16899 summary; `POST` is the restore-by-id).
- Inherited base reds named above (#15306, `/api/balance` doc claim, `managementSession.ts` TS2322, i18n drift of 34 sources).
Merge-batch Track B. Narrowed the settings admin override from an ADMIN_MUTATION_PREFIXES prefix to an exact-path list (ADMIN_MUTATION_EXACT_PATHS: /api/settings, /api/settings/require-login) so other /api/settings/* sub-routes keep the write scope (owner decision). New tests: inference for sub-routes + real requireManagementAuth gate (write oma_ token passes /api/settings/proxies, 403 on PATCH /api/settings). Focused: 91/91 (security-settings-operator, settings-require-login-operator, access-scopes-infer, login-bootstrap-route, mcp-route-scope-carveout, cli-login-push-remote) + 38/38 (oidc-settings-*, settings-import-json-session-state, settings-route-password, api/settings-audit). Gates: typecheck:core OK, open-sse typecheck OK, complexity/cognitive/file-size/dead-code OK, eslint+prettier clean, skill generator Generated: 0 after --apply (omni-settings/references/endpoints.md). Boundary 774/771 inherited from base (unchanged).
Merged via /merge-batch (2026-10-16 wave). Validated in the 20-PR combined board, then reconciled with #16898 (which landed first): the key-metadata SELECT keeps #16898's hash-only lookup (WHERE key_hash = ?) and adds this PR's forced_model column — byte-identical to the board-validated file. Focused tests: 61/61 (forced-model + access form), 11/11 (#16898 at-rest + hash-only suites on the merged tree), 18/18 vitest UI; typecheck:core OK. Remaining CI reds are inherited base-red #15306.
Merged via /merge-batch (2026-10-16 wave, maintainer PRs). Validated on one combined board of 20 PRs over release/v3.8.52@76e905b15: typecheck:core, open-sse typecheck, eslint on changed files, file-size, changelog-integrity, complexity, cognitive-complexity, env-doc-sync and agent-skills-sync all green; focused tests 263/263 (node:test) + 26/26 (vitest UI). Remaining CI reds are the inherited base-red #15306.
Require management scope for /api/v1/registered-keys, matching /api/keys.
Evidence: focused tests 181 passed, 0 failed, in the combined run; file-size OK (176 frozen files, cap 1440); changelog integrity OK against origin/release/v3.8.52.
⚠️ base-red inherited: #15306
Thanks @yourspraveen.
Validated in local merge-train merge-train-20261016-085341 @ d0bbc5fd with file-size ceilings raised 20% (#16831, cap 1440). Static gates green. The only failing tests were in tests/unit/chat-cooldown-aware-retry.test.ts, which belongs to #16305 and was not merged. Inherited typecheck and i18n drift match the tip (#15306).
Validated on local board 917689fb (batch-owner-1015) with the owner wave: static gates plus focused tests. This squash is that same head (f83c6d94de).
⚠️ base-red inherited: #15306
Thanks @diegosouzapw.
* fix(security): require an operator credential to export stored credentials
GET /api/settings/export-json returns every provider connection decrypted plus
the raw API keys, and GET /api/db-backups/export and /api/db-backups/exportAll
ship the whole SQLite database. All three accepted any credential that passed
isAuthenticated(), so an API key with the `manage` (or `admin`) scope could
download every stored secret.
They now go through requireOperatorCredential() and accept only the instance
operator: a signed-in dashboard session, the local CLI subject (pipeline-
stamped or validated on loopback), or an instance with login disabled. API
keys of any scope and `oma_` access tokens get 403; no credential is still 401.
The operator check already existed twice (sync token issuance and admin CLI
token minting). requireOperatorCredential() in src/lib/api/operatorCredential.ts
now wraps the management-auth step plus resolveOperatorCredential(), and the
sync token issuer is rebuilt on it (with the internal-service token opt-in it
already accepted), so issuance keeps the same issuer kinds and messages.
Imports (import-json, db-backups import/restore) are unchanged.
* docs(changelog): number the credential-export fragment
* docs(changelog): link the credential-export fragment to its PR
Thanks @PrinceThummar011. Wave 27 boarded this pull request cleanly on release/v3.8.52 at f8d11c9326. The focused unit tests for the files it touches passed on that combined tree (116 passed in the second run, plus the auto-combo weight tests and the skills characterization tests). Changelog integrity passed. The only file-size finding is executeTargetAttempt.ts, already over its frozen ceiling on the release tip and unchanged by this diff.
⚠️ base-red inherited: #15306Closes#15530
Merged with the owner's 2026-10-10 hold/backlog wave. Reworked to keep frozen files within config/quality/file-size-baseline.json after #16070 ratcheted it down (logic extracted into new modules, no baseline raised). Validated together with #16147/#16112/#16110/#15790 on top of release/v3.8.52: typecheck:core, open-sse and API-route typecheck, eslint on changed files, complexity/cognitive ratchets, changelog-integrity green; 117/117 changed tests; only file-size overrun is the inherited executeTargetAttempt.ts. Warning: base-red inherited: #15306.
Thanks @pacocartones.
Wave 26 boarded this pull request cleanly on release/v3.8.52 at 7a0bb875. The focused unit tests for the files it touches passed on that combined tree. This diff does not push a frozen file past its ceiling.
⚠️ base-red inherited: #15306
* feat(release): nightly version and npm dist-tag resolver for the LTS rail
scripts/release/dist-tag.mjs is the single source of the channel rules in
docs/ops/RELEASE_STRATEGY.md: -nightly.* -> nightly (never latest),
rc/beta/alpha (and any other pre-release) -> next, a stable of an older
major than the highest stable tag -> lts, everything else -> latest, with
the publish guard demoting a non-highest stable to historic.
scripts/release/compute-nightly-version.mjs builds
<core>-nightly.<YYYYMMDD>.<sha7> (UTC, validated semver, all-digit SHA
prefixed with g) for the develop nightly build.
* ci(release): resolve npm dist-tag via dist-tag.mjs and add dormant LTS workflows
npm-publish.yml: tag=auto now goes through scripts/release/dist-tag.mjs, so a
4.0.0-nightly.* build resolves to nightly instead of falling through to the
stable path (where it could claim latest), and a 3.9.x patch published after
the 4.0 GA resolves to lts. Stable 3.8.x keeps latest/historic exactly as
before; nightly and lts join the explicit dispatch choices.
Dormant workflows (never fire until stable/v3 / develop exist):
- forward-port.yml: cherry-pick -x of every stable/v3 commit onto
forward-port/<sha7> from develop, original author kept plus a
Co-authored-by trailer; a conflict opens a draft PR with instructions.
- validate-stable-pr.yml: stable/v3 PRs need a fix|docs|chore(i18n)|
feat(providers) title or the lts label.
- nightly-v4-build.yml: computes <core>-nightly.<date>.<sha7>, runs
build:release + check:pack-artifact, keeps the tarball as an artifact and
publishes --tag nightly --provenance only when vars.NIGHTLY_PUBLISH is true.
tests/unit/release-channel-workflows.test.ts executes the npm-publish resolve
step and the stable-PR check against scratch inputs and pins the dormancy,
permissions and SHA pins of the new workflows.
* feat(api): expose release channel and npm dist-tags on /api/system/version
GET /api/system/version gains two additive fields:
- releaseChannel: latest | next | nightly | lts for the running build,
resolved by src/lib/system/releaseChannel.ts with the same rules as
scripts/release/dist-tag.mjs (parity test in release-channel.test.ts);
- channels: { latest, next?, nightly?, lts? } from npm view omniroute
dist-tags (registry HTTP fallback), cached with the same TTL, single-flight
and refresh semantics as the latest-version lookup (both caches now share
one createCachedLookup helper and are cleared together).
The existing channel field keeps meaning the deployment mode
(npm/source/docker-compose) that the dashboard updater reads, so the shape is
unchanged for current consumers. Documented in API_REFERENCE.md and as an
additive response schema in openapi.yaml.
Batch-merged with the owner's 2026-10-09 PR wave (/merge-batch). Validated in one combined tree of 57 boarded PRs on top of release/v3.8.52 (35ef635c): typecheck:core, check:open-sse-typecheck, check-deps, changelog-integrity, complexity + cognitive ratchets, check:docs-all, i18n UI coverage, ESLint on all 247 changed source files — all green for this wave; 89 focused unit test files (739/744 pass under node:test, the 4 UI .tsx files re-run under vitest 13/13, check-deps fixed in #15912). Remaining reds in that tree belong only to held/draft PRs outside this wave (#15790 file-size, #15780 open-sse types).
Validated in a combined board with the other PRs of this wave on release/v3.8.52: the PR's own tests pass; typecheck:core, open-sse typecheck, api-typecheck, complexity, changelog-integrity, env-doc-sync, test-discovery, vitest-exclusions, error-helper, test-masking, lockfile, deps, gate-manifest, cli-i18n and i18n coverage green. Any file-size ceiling or stryker registration this wave needs lands in the wave's follow-up PR. Thank you for the contribution!
Validated in a combined board with the other PRs of this wave on release/v3.8.52: the PR's own tests pass; typecheck:core, open-sse typecheck, api-typecheck, complexity, changelog-integrity, env-doc-sync, test-discovery, vitest-exclusions, error-helper, test-masking, lockfile, deps, gate-manifest, cli-i18n and i18n coverage green. Any file-size ceiling or stryker registration this wave needs lands in the wave's follow-up PR. Thank you for the contribution!
Validated in a combined board with the other PRs of this wave on release/v3.8.52: the PR's own tests pass; typecheck:core, open-sse typecheck, api-typecheck, complexity, changelog-integrity, env-doc-sync, test-discovery, vitest-exclusions, error-helper, test-masking, lockfile, deps, gate-manifest, cli-i18n and i18n coverage green. Any file-size ceiling or stryker registration this wave needs lands in the wave's follow-up PR. Thank you for the contribution!
Onda A 2026-10-07. Combined board e1b721a1 validated this PR together with the rest of the wave, then this squash lands the contributor head unchanged.
- typecheck:core passed
- complexity 2993 (baseline 3218)
- cognitive complexity 1366 (baseline 1437)
- changelog integrity passed
- file-size passed after the authorized ceiling bump; that bump ships in the wave follow-up, not in this squash
- focused tests: 318 passed in the batch run. The Devin anthropic price assertion already mismatches CLAUDE_SONNET_5_PRICING on origin/release/v3.8.52, and this wave does not change that constant. The combo-metadata miss did not reproduce when that file ran alone.
fix(reasoning): honor explicit opt-outs and Claude classifier defaults
Thanks @D4rk4.
⚠️ base-red inherited: #15306
Onda A 2026-10-07. Combined board e1b721a1 validated this PR together with the rest of the wave, then this squash lands the contributor head unchanged.
- typecheck:core passed
- complexity 2993 (baseline 3218)
- cognitive complexity 1366 (baseline 1437)
- changelog integrity passed
- file-size passed after the authorized ceiling bump; that bump ships in the wave follow-up, not in this squash
- focused tests: 318 passed in the batch run. The Devin anthropic price assertion already mismatches CLAUDE_SONNET_5_PRICING on origin/release/v3.8.52, and this wave does not change that constant. The combo-metadata miss did not reproduce when that file ran alone.
feat(api): add ZenMux image generation support
Thanks @JeanRGW.
⚠️ base-red inherited: #15306
Onda 3 (woodsonl) validada num único board em `.claude/worktrees/batch-w3` sobre `origin/release/v3.8.52` (`4f4c983347b9`), 27 PRs, superfícies protegidas 0.
Evidência: `typecheck:core` limpo. i18n drift limpo no board. changelog-integrity ok. file-size ok (170 congelados). complexity 2975/3218 e cognitive-complexity 1357/1437 dentro do baseline. node:test focado: 1868 passaram, 0 falharam. Vitest da UI: 150 passaram em 12 arquivos. Os 38 do RTK caem no `useCallback` nulo de `node_modules/node_modules/react` via zustand, o mesmo React duplicado desta worktree. Onde dois PRs mexiam no mesmo save, o board ficou com os dois comportamentos: escrita parcial de engine por id e a fila/resync do PR que entrou depois.
O hash de `.i18n-state.json` do guia de quality-gates entra no follow-up, fora do squash.
⚠️ base-red inherited: #15306 (timeouts de suíte no tip, não neste diff).
Obrigado, @woodsonl.
Adds the opt-in 'omniroute models test-add' operation and POST /api/provider-models/validate-and-add: a model is registered only after bounded generation, streamed synthetic tool-call and continuation proofs on the selected connection. Protected surface (9 generated skills/**/SKILL.md + endpoints.md, flag documentation only): explicit owner approval given in chat; diff reviewed for external execution (none beyond a localhost curl example). Tip merged cleanly. Independent validation: 126/126 tests across 16 files, typecheck:core, open-sse typecheck, eslint on touched files, check:cycles, check:docs-all, file-size, changelog-integrity and complexity green. Route reviewed: strict Zod schema with allowInference literal true, management auth always required, errors via buildErrorBody, no client-supplied base URL. Not validated: live provider run on the VPS.
/api/cloud/auth keeps maskedApiKey/projectId for manage/admin keys only and honours allowedConnections. 4 failures on the tip, green with the change. Thank you @HouMinXi!
Boarded on the release/v3.8.51 tip together with the 2026-09-28 release-drain batch (30 PRs): typecheck:core, check:open-sse-typecheck, file-size, migration numbering, i18n keys and ESLint on the changed files are clean, and this PR's own test files (1) pass on the combined tree. Thank you @QuangBlue!
Merge-batch 2026-09-23 (PRs do mantenedor, Trilha B, retrabalhada).
Um worker corrigiu o defeito apontado na revisão direto no branch da PR; o detalhe está na seção "Rework (merge-batch 2026-09-23)" do corpo, com a prova de red→green. Revalidado numa árvore combinada das PRs da 2ª leva da Trilha B (#14415#14472#14468#14493#14403) sobre o tip atual de `release/v3.8.51`:
- 104/104 testes node focados + 46/46 vitest (as 6 suítes de UI que saíram da quarentena); `check:provider-consistency` OK (359)
- `typecheck:core`: só o herdado `cliproxyAccountHealth.ts:157`; `check:open-sse-typecheck`: só o herdado `auggie.ts` (base-red #14547, corrigido pela #14511)
- eslint nos arquivos alterados limpo; file-size OK
Merged. Thank you, @RaviTharuma — and thank you for naming whose design this is instead of quietly absorbing it.
Two good calls landed here:
- **Lossy stays off the default path.** Session dedup and whitespace folding still run when compression is on; summaries, relevance filters and style rewrites now require an explicit `x-omniroute-compression: allow-lossy`, `engine:<id>`, or a named combo. Silently rewriting a user's prompt is exactly the class of default this project should not have.
- **Combo 400 handling became a table instead of another string guard.** `statusDecisionTable.ts` with `advance` | `stop` rows is reviewable in a way a growing chain of substring checks never is: model-scoped wrappers advance, `Invalid message format` still stops, and a 400 no longer opens the provider-wide breaker.
Migrating the seeded Standard Savings combo to `session-dedup` + `lite` only when it is still the untouched RTK + Caveman seed — and leaving a renamed or edited combo alone — is the right conservative shape for a seed migration.
Validation before merge: boarded onto the current release tip together with the other seven PRs of your batch — merged clean, no conflict against the tip or any sibling. The combined board ran 81 focused tests across the batch with 0 failures, including `combo-status-decision-table`, the four compression suites and `strategySelector`; `check:file-size`, `check:changelog-integrity` and `check:migration-numbering` green.
Co-authored-by: Bob.Hou <19586012+HouMinXi@users.noreply.github.com>
Co-authored-by: Koosha Paridehpour <42529354+KooshaPari@users.noreply.github.com>
* feat(i18n): add Bosnian (bs) localization
* docs(i18n): re-adopt mirrors reformatted by the pre-commit hook
* test(i18n): count 67 locale files and accept the Bosnian 'Model' cognate; restore the bs ICU placeholder quotes
The Bosnian locale adds a 67th messages file (agnes-cn-provider pins the
count), 'Model' is the Bosnian word for the Recent Requests legend (same
cognate as hr/sr) and the retranslation had dropped the ICU literal quotes
around <your OmniRoute API key> in ccOnboardingKeyPlaceholder.
* docs(i18n): re-adopt mirrors reformatted by the pre-commit hook
* docs(i18n): regenerate the language bars and adopt the bar-only source edits after the base merge
* docs(i18n): re-adopt mirrors reformatted by the pre-commit hook
---------
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>