mirror of
https://github.com/diegosouzapw/OmniRoute.git
synced 2026-11-22 04:19:40 +00:00
@@ -2613,6 +2613,13 @@ APP_LOG_TO_FILE=true
|
||||
# Used by: src/lib/healthzLag.ts (event-loop stall sampler). Disabled by default;
|
||||
# logs sizes and inflight count only, never bodies or heap snapshots.
|
||||
# OMNIROUTE_EVENT_LOOP_STALL_THRESHOLD_MS=0 stops that sampler, and this diagnostic with it.
|
||||
# Optional response header for troubleshooting Codex OAuth plan eligibility.
|
||||
# OFF by default: when enabled, API callers can see the recognized subscription tier
|
||||
# of the selected Codex connection in X-OmniRoute-Selected-Connection-Plan.
|
||||
# Do not enable on public/multi-tenant gateways if this account metadata is private.
|
||||
# Arbitrary/unrecognized plan values and non-Codex requests never get this header.
|
||||
# OMNIROUTE_CODEX_PLAN_RESPONSE_HEADER=0
|
||||
|
||||
# OMNIROUTE_HEAP_DIAGNOSTICS_ENABLED=0
|
||||
# Heap-use threshold in MiB for the opt-in diagnostic (default: 900).
|
||||
# A high-heap log is rate-limited to once per 60 seconds.
|
||||
|
||||
@@ -0,0 +1 @@
|
||||
- Add an opt-in `X-OmniRoute-Selected-Connection-Plan` response header for Codex account-routing troubleshooting (`OMNIROUTE_CODEX_PLAN_RESPONSE_HEADER=1`), off by default with an explicit plan allowlist.
|
||||
@@ -929,6 +929,7 @@ REQUEST_TIMEOUT_MS (global override)
|
||||
| `OMNIROUTE_AGENT_GOAL_READINESS_MAX_TIMEOUT_MS` | `600000` | Maximum first-event readiness window for detected `/goal` agent runs or requests forced with `x-omniroute-agent-goal`. |
|
||||
| `OMNIROUTE_AGENT_GOAL_STREAM_RECOVERY` | `true` | Enable early stream recovery automatically for detected `/goal` agent runs. Set `false`/`0`/`off` to disable the goal-specific opt-in. This can only ADD recovery on top of the operator default — it never overrides an explicit `STREAM_RECOVERY_ENABLED`/DB settings opt-out. |
|
||||
| `OMNIROUTE_CODEX_DROP_NONSTANDARD_EVENTS` | `true` | Strip non-standard `codex.*` SSE events (e.g. `codex.rate_limits`) that break the OpenAI SDK's `responses.stream()` with a 502. Default ON (#11014). Set `0`/`false`/`no`/`off` to forward them. |
|
||||
| `OMNIROUTE_CODEX_PLAN_RESPONSE_HEADER` | `0` (off) | Opt-in Codex OAuth troubleshooting response header. Set to `1` to emit `X-OmniRoute-Selected-Connection-Plan` with the recognized plan tier of the selected Codex connection. Not sent for non-Codex requests or unrecognized plans. Avoid enabling on public or multi-tenant gateways where subscription metadata is private. |
|
||||
| `OMNIROUTE_CODEX_APPSERVER_WS` | _(unset)_ | Opt-in Codex app-server transport. WebSocket endpoint (`ws://`/`wss://`) of a local `codex app-server` sidecar. When set together with a token, Codex requests are routed over JSON-RPC to the sidecar instead of the HTTP Responses API. Also settable per-connection via `providerSpecificData.codexAppServerUrl`. Used by `open-sse/executors/codex/appServerConfig.ts`. |
|
||||
| `OMNIROUTE_CODEX_APPSERVER_WS_TOKEN` | _(unset)_ | Inline capability/bearer token presented to the app-server. Per-connection override: `providerSpecificData.codexAppServerToken`. |
|
||||
| `OMNIROUTE_CODEX_APPSERVER_WS_TOKEN_FILE` | _(unset)_ | Path to a file holding the app-server capability token (from `codex app-server --ws-token-file`). Used when `OMNIROUTE_CODEX_APPSERVER_WS_TOKEN` is unset. Per-connection override: `providerSpecificData.codexAppServerTokenFile`. |
|
||||
|
||||
@@ -11,6 +11,7 @@ export const OMNIROUTE_RESPONSE_HEADERS = {
|
||||
fallbackAttempts: "X-OmniRoute-Fallback-Attempts",
|
||||
accountFallbacks: "X-OmniRoute-Account-Fallbacks",
|
||||
accountFallbackReasons: "X-OmniRoute-Account-Fallback-Reasons",
|
||||
selectedConnectionPlan: "X-OmniRoute-Selected-Connection-Plan",
|
||||
sameAccountRetries: "X-OmniRoute-Same-Account-Retries",
|
||||
routingWaitMs: "X-OmniRoute-Routing-Wait-Ms",
|
||||
latencyMs: "X-OmniRoute-Latency-Ms",
|
||||
|
||||
@@ -38,6 +38,7 @@ import {
|
||||
resolveAvalancheSettings,
|
||||
} from "@omniroute/open-sse/services/accountFallback/antiAvalanche.ts";
|
||||
import { estimateTokens } from "@omniroute/open-sse/services/contextManager.ts";
|
||||
import { getCodexPlanType } from "@omniroute/open-sse/services/codexPlanEligibility.ts";
|
||||
import { getCombo, getComboForModel, getModelInfo } from "../services/model";
|
||||
import { stripContextWindowSuffix } from "@omniroute/open-sse/services/model.ts";
|
||||
import { resolveBareModelToConnectionDefault } from "@omniroute/open-sse/services/model.ts";
|
||||
@@ -112,8 +113,6 @@ import { comboCheckProvider, ghComboGate } from "./chat/githubLiveCatalogFilter.
|
||||
import { markEmergencyFallback } from "./emergencyFallbackHeader.ts";
|
||||
import { resolveEmergencyFallbackDecision } from "./emergencyFallbackHeader.ts";
|
||||
import { resolveInternalUpstreamStartTimeoutMs } from "../services/internalUpstreamStartTimeout";
|
||||
import { resolveResilienceSettings } from "@/lib/resilience/settings";
|
||||
import { resolvePreferredConnectionUpstreamStartTimeoutMs } from "../services/preferredConnectionFailover";
|
||||
import { evaluateComboTargetPreflight } from "./chat/comboTargetKeyPolicy.ts";
|
||||
import * as resolvedPolicy from "./chat/resolvedModelPolicy.ts";
|
||||
import { recordGateRejection, recordQuotaParkedSkip } from "./quotaParkedSkipUsage";
|
||||
@@ -1822,13 +1821,9 @@ async function handleSingleModelChat(
|
||||
});
|
||||
|
||||
const userAgent = request?.headers?.get("user-agent") || "";
|
||||
const cachedRuntimeSettings =
|
||||
runtimeOptions.cachedSettings ?? (await getCachedSettings().catch(() => ({})));
|
||||
const baseRetrySettings = resolveCooldownAwareRetrySettings(cachedRuntimeSettings);
|
||||
// The read cache invalidates on settings writes (up to its 5s TTL).
|
||||
// No restart is needed for newly dispatched chat requests.
|
||||
const preferredFailoverSettings =
|
||||
resolveResilienceSettings(cachedRuntimeSettings).preferredAccountFailover;
|
||||
const baseRetrySettings = resolveCooldownAwareRetrySettings(
|
||||
runtimeOptions.cachedSettings ?? (await getCachedSettings().catch(() => ({})))
|
||||
);
|
||||
const retrySettings = disableCooldownAwareRetry(
|
||||
baseRetrySettings,
|
||||
provider === "claude-web" ||
|
||||
@@ -1868,14 +1863,12 @@ async function handleSingleModelChat(
|
||||
// STREAM_EARLY_EOF_SIBLING_FAILOVER_ENABLED: at most ONE sibling hop per request. Keeps the
|
||||
// original early-EOF 502 so an exhausted sibling pool surfaces it verbatim (combo detection).
|
||||
let earlyEofOriginal: Response | null = null;
|
||||
// Preferred connection start timeout is a single-hop routing hint, never a cooldown.
|
||||
let preferredTimeoutOriginal: Response | null = null;
|
||||
let preferredFailoverUsed = false;
|
||||
const sameAccountTransportRetries = new Map<string, number>();
|
||||
// An obsolete Codex selection/response is not a fault of the current account. Reselect once
|
||||
// without cooling or excluding its connection ID; never replay a successful response.
|
||||
let staleCodexReselected = false;
|
||||
let accountFallbacks = 0;
|
||||
let selectedConnectionPlan: string | null = null;
|
||||
let sameAccountRetries = 0;
|
||||
let routingWaitMs = 0;
|
||||
const accountFallbackReasons: string[] = [];
|
||||
@@ -1887,6 +1880,7 @@ async function handleSingleModelChat(
|
||||
};
|
||||
const routingTelemetry = () => ({
|
||||
accountFallbacks,
|
||||
selectedConnectionPlan,
|
||||
accountFallbackReasons,
|
||||
sameAccountRetries,
|
||||
routingWaitMs,
|
||||
@@ -2027,7 +2021,6 @@ async function handleSingleModelChat(
|
||||
"allExpired" in credentials ||
|
||||
!credentials.connectionId
|
||||
) {
|
||||
if (preferredTimeoutOriginal) return preferredTimeoutOriginal;
|
||||
if (earlyEofOriginal) return withRoutingTelemetry(earlyEofOriginal, null);
|
||||
if (!credentials?.allRateLimited && agy.earliestRetryHintAtMs !== null)
|
||||
return agyLease.buildAntigravityPoolBusyResponse(agy.earliestRetryHintAtMs);
|
||||
@@ -2126,6 +2119,8 @@ async function handleSingleModelChat(
|
||||
excludedConnectionIds.size > 0
|
||||
? Array.from(excludedConnectionIds)[excludedConnectionIds.size - 1]
|
||||
: null;
|
||||
// No account was selected for this response; don't report a previous attempt's plan.
|
||||
selectedConnectionPlan = null;
|
||||
return withRoutingTelemetry(noCredsRes, lastFailedConnectionId);
|
||||
}
|
||||
|
||||
@@ -2148,6 +2143,9 @@ async function handleSingleModelChat(
|
||||
}
|
||||
pendingStreamRetryConnectionId = null;
|
||||
}
|
||||
// Refresh on every selected credential so fallback never reports a stale plan.
|
||||
selectedConnectionPlan =
|
||||
provider === "codex" ? getCodexPlanType(credentials.providerSpecificData) : null;
|
||||
const accountId = credentials.connectionId.slice(0, 8);
|
||||
const releaseOAuthSession = credentials.releaseOAuthSession ?? (() => {});
|
||||
// Undefined whenever the lease flag is off, which makes every release/hold below a no-op.
|
||||
@@ -2310,21 +2308,6 @@ async function handleSingleModelChat(
|
||||
reselectPoolMember?: () => Promise<unknown>;
|
||||
} = { proxy: null };
|
||||
const proxyStartTime = Date.now();
|
||||
// A preferred API-key account can be given one shorter time-to-headers
|
||||
// deadline when a vetted sibling exists. Pins/leases/combo/Muse are excluded.
|
||||
const preferredDeadline = resolvePreferredConnectionUpstreamStartTimeoutMs(
|
||||
provider,
|
||||
credentials,
|
||||
requestUpstreamStartTimeoutMs,
|
||||
effectiveModel,
|
||||
preferredFailoverUsed ||
|
||||
hasForcedConnection ||
|
||||
runtimeOptions.managedLease ||
|
||||
museOwner ||
|
||||
isCombo
|
||||
? { ...preferredFailoverSettings, enabled: false }
|
||||
: preferredFailoverSettings
|
||||
);
|
||||
// 4. Execute chat via core after breaker gate checks (with optional TLS tracking)
|
||||
if (telemetry) telemetry.startPhase("connect");
|
||||
let execution: Awaited<ReturnType<typeof dispatchChatWithAffinityEviction>>;
|
||||
@@ -2372,7 +2355,7 @@ async function handleSingleModelChat(
|
||||
videoBridgeLog: runtimeOptions.videoBridgeLog,
|
||||
previousResponseResumed: runtimeOptions.previousResponseResumed,
|
||||
fallbackAttempts: runtimeOptions.fallbackAttempts,
|
||||
requestUpstreamStartTimeoutMs: preferredDeadline.timeoutMs,
|
||||
requestUpstreamStartTimeoutMs,
|
||||
beforeUpstreamAttempt:
|
||||
museOwner && muse.museBeforeUpstreamAttempt(museOwner, credentials),
|
||||
forcedConnectionId: hasForcedConnection ? forcedConnectionId : null, // #14116
|
||||
@@ -2491,23 +2474,6 @@ async function handleSingleModelChat(
|
||||
// A shorter *preferred* deadline may perform one in-request sibling hop,
|
||||
// without recording a connection failure or evicting session affinity.
|
||||
if (result.errorCode === REQUEST_SCOPED_UPSTREAM_TIMEOUT_CODE) {
|
||||
if (
|
||||
preferredDeadline.mayRotateOnRequestTimeout &&
|
||||
!preferredFailoverUsed &&
|
||||
requestSignal?.aborted !== true
|
||||
) {
|
||||
preferredFailoverUsed = true;
|
||||
preferredTimeoutOriginal = withSelectedConnectionHeader(
|
||||
result.response,
|
||||
credentials.connectionId
|
||||
);
|
||||
excludedConnectionIds.add(credentials.connectionId);
|
||||
log.info(
|
||||
"AUTH",
|
||||
"Preferred account response-start timed out; trying one eligible sibling"
|
||||
);
|
||||
continue;
|
||||
}
|
||||
return withSelectedConnectionHeader(result.response, credentials.connectionId);
|
||||
}
|
||||
// Never let generic account fallback or affinity eviction move a Muse caller.
|
||||
|
||||
@@ -10,6 +10,7 @@ import { inheritResponsePolicies } from "./responsePolicies.ts";
|
||||
export type AccountRoutingTelemetry = {
|
||||
accountFallbacks?: number;
|
||||
accountFallbackReasons?: readonly string[];
|
||||
selectedConnectionPlan?: string | null;
|
||||
sameAccountRetries?: number;
|
||||
routingWaitMs?: number;
|
||||
};
|
||||
@@ -37,6 +38,30 @@ function fallbackReasons(values: readonly string[] | undefined): string[] {
|
||||
.filter((reason) => SAFE_FALLBACK_REASONS.has(reason) || /^http_[1-5][0-9]{2}$/.test(reason));
|
||||
}
|
||||
|
||||
// Only explicitly enabled server operators can expose recognized subscription tiers.
|
||||
// An arbitrary provider field is never copied into a response header.
|
||||
const CODEX_PLAN_TIERS = new Set([
|
||||
"free",
|
||||
"go",
|
||||
"plus",
|
||||
"pro",
|
||||
"prolite",
|
||||
"promax",
|
||||
"team",
|
||||
"chatgptteam",
|
||||
"business",
|
||||
"enterprise",
|
||||
"edu",
|
||||
"education",
|
||||
]);
|
||||
|
||||
function selectedCodexPlan(value: string | null | undefined): string | null {
|
||||
if (process.env.OMNIROUTE_CODEX_PLAN_RESPONSE_HEADER !== "1") return null;
|
||||
if (typeof value !== "string") return null;
|
||||
const normalized = value.trim().toLowerCase().replace(/[\s_-]+/g, "");
|
||||
return CODEX_PLAN_TIERS.has(normalized) ? normalized : null;
|
||||
}
|
||||
|
||||
export function withSelectedConnectionHeader(
|
||||
response: Response,
|
||||
connectionId: string | null | undefined,
|
||||
@@ -44,15 +69,24 @@ export function withSelectedConnectionHeader(
|
||||
): Response {
|
||||
if (!response) return response;
|
||||
const fallbacks = nonnegativeCount(telemetry?.accountFallbacks);
|
||||
const plan = selectedCodexPlan(telemetry?.selectedConnectionPlan);
|
||||
const retries = nonnegativeCount(telemetry?.sameAccountRetries);
|
||||
const waitMs = nonnegativeCount(telemetry?.routingWaitMs);
|
||||
const reasons = fallbackReasons(telemetry?.accountFallbackReasons);
|
||||
if (!connectionId && fallbacks === 0 && retries === 0 && waitMs === 0 && reasons.length === 0) {
|
||||
if (
|
||||
!connectionId &&
|
||||
fallbacks === 0 &&
|
||||
retries === 0 &&
|
||||
waitMs === 0 &&
|
||||
reasons.length === 0 &&
|
||||
!plan
|
||||
) {
|
||||
return response;
|
||||
}
|
||||
|
||||
const apply = (headers: Headers) => {
|
||||
if (connectionId) headers.set("X-OmniRoute-Selected-Connection-Id", connectionId);
|
||||
if (plan) headers.set(OMNIROUTE_RESPONSE_HEADERS.selectedConnectionPlan, plan);
|
||||
if (fallbacks > 0) headers.set(OMNIROUTE_RESPONSE_HEADERS.accountFallbacks, String(fallbacks));
|
||||
if (reasons.length) {
|
||||
headers.set(OMNIROUTE_RESPONSE_HEADERS.accountFallbackReasons, reasons.join(","));
|
||||
|
||||
@@ -32,6 +32,38 @@ test("telemetry headers are bounded and never echo untrusted errors", () => {
|
||||
assert.equal(response.headers.get("X-OmniRoute-Selected-Connection-Plan"), null);
|
||||
});
|
||||
|
||||
test("Codex account plan diagnostics are off by default and allowlisted when enabled", () => {
|
||||
const key = "OMNIROUTE_CODEX_PLAN_RESPONSE_HEADER";
|
||||
const before = process.env[key];
|
||||
try {
|
||||
delete process.env[key];
|
||||
const hidden = withSelectedConnectionHeader(new Response("ok"), "c", {
|
||||
selectedConnectionPlan: "Plus",
|
||||
});
|
||||
assert.equal(hidden.headers.get("X-OmniRoute-Selected-Connection-Plan"), null);
|
||||
|
||||
process.env[key] = "1";
|
||||
const visible = withSelectedConnectionHeader(new Response("ok"), "c", {
|
||||
selectedConnectionPlan: "Plus",
|
||||
});
|
||||
assert.equal(visible.headers.get("X-OmniRoute-Selected-Connection-Plan"), "plus");
|
||||
const normalized = withSelectedConnectionHeader(new Response("ok"), "c", {
|
||||
selectedConnectionPlan: "Pro Max",
|
||||
});
|
||||
assert.equal(normalized.headers.get("X-OmniRoute-Selected-Connection-Plan"), "promax");
|
||||
|
||||
for (const unsafe of ["custom-internal-secret", "Plus\r\nInjected: true", ""]) {
|
||||
const response = withSelectedConnectionHeader(new Response("ok"), "c", {
|
||||
selectedConnectionPlan: unsafe,
|
||||
});
|
||||
assert.equal(response.headers.get("X-OmniRoute-Selected-Connection-Plan"), null);
|
||||
}
|
||||
} finally {
|
||||
if (before === undefined) delete process.env[key];
|
||||
else process.env[key] = before;
|
||||
}
|
||||
});
|
||||
|
||||
test("immutable response clone retains trusted response policies alongside telemetry", async () => {
|
||||
const { markProviderProbeResponse, isProviderProbeResponse } =
|
||||
await import("../../src/shared/utils/providerProbeResult.ts");
|
||||
|
||||
Reference in New Issue
Block a user