feat: fix so restart of server restarts batch jobs instead of failing them (#2755)

Merge PR #2755 — feat: fix so restart of server restarts batch jobs instead of failing them

.dockerignore

@@ -38,14 +38,11 @@ playwright-report
 blob-report
 
 # Documentation
-# Issue #2348: The Dashboard Docs viewer reads markdown from `/app/docs` at
-# runtime. The previous `docs/*` block hid every file except openapi.yaml,
-# so the in-product help screen failed with ENOENT for every page.
 # Translations (~51 MB) are excluded — the Docs viewer reads English sources.
-# v3.8.3+ (fumadocs-mdx): `npm run build` webpack-bundles docs — keep assets referenced
-# from English markdown (docs/diagrams/exported/*.svg, docs/screenshots/*.png).
+# Screenshots (~1.7 MB) and SVGs (~250 KB) are needed at build time for MDX
+# image resolution (fumadocs-mdx bundles them).
+# Raster sources under docs/diagrams/ only (exported SVGs are required).
 docs/i18n/**
-# Raster sources under docs/diagrams/ only (exported SVGs are required at build time).
 docs/diagrams/**/*.png
 docs/diagrams/**/*.jpg
 docs/diagrams/**/*.jpeg

src/sse/handlers/chat.ts

@@ -123,7 +123,7 @@ const COMBOS_CACHE_TTL_MS = 10_000;
 
 async function getCombosCachedForChat(): Promise<unknown[]> {
   const now = Date.now();
-  if (combosCachePromise !== null && now - combosCacheTs < COMBOS_CACHE_TTL_MS) {
+  if (combosCachePromise && now - combosCacheTs < COMBOS_CACHE_TTL_MS) {
     return combosCachePromise;
   }
 
@@ -153,27 +153,6 @@ function intersectAllowedConnectionIds(primary: unknown, secondary: unknown): st
 
 const PROVIDER_BREAKER_FAILURE_STATUSES = new Set([408, 500, 502, 503, 504]);
 
-/**
- * Wrap a Request-like object so callers reading `.signal` see a merged signal
- * combining the original request signal with an extra per-call signal (e.g. a
- * combo per-model timeout). Returns the original request unchanged when no
- * extra signal is provided, so the hot path stays a no-op.
- */
-function wrapRequestWithExtraSignal(request: any, extraSignal: AbortSignal | null) {
-  if (!extraSignal || !request) return request;
-  const baseSignal: AbortSignal | null | undefined = request?.signal ?? null;
-  const mergedSignal: AbortSignal = baseSignal
-    ? AbortSignal.any([baseSignal, extraSignal])
-    : extraSignal;
-  return new Proxy(request, {
-    get(target, prop, receiver) {
-      if (prop === "signal") return mergedSignal;
-      const value = target[prop];
-      return typeof value === "function" ? value.bind(target) : value;
-    },
-  });
-}
-
 /**
  * Handle chat completion request
  * Supports: OpenAI, Claude, Gemini, OpenAI Responses API formats
@@ -221,9 +200,6 @@ export async function handleChat(request: any, clientRawRequest: any = null) {
 
   // Log request endpoint and model
   const url = new URL(request.url);
-  // `let` because the middleware-hook pipeline (line ~319) may reassign this
-  // when a hook rewrites the target model. Previously declared `const`, which
-  // broke turbopack/strict-mode builds (PR [PR #2670](file:///home/diegosouzapw/dev/proxys/OmniRoute/package.json#L2670) regression).
   let modelStr = body.model;
 
   // Count messages (support both messages[] and input[] formats)
@@ -332,7 +308,7 @@ export async function handleChat(request: any, clientRawRequest: any = null) {
     >,
     model: modelStr,
     combo: undefined,
-    apiKeyInfo: apiKeyInfo as unknown as Record<string, unknown> | undefined,
+    apiKeyInfo: apiKeyInfo as Record<string, unknown> | undefined,
     log,
   });
 
@@ -493,11 +469,6 @@ export async function handleChat(request: any, clientRawRequest: any = null) {
         resolvedModel,
         {
           sessionKey: sessionAffinityKey,
-          sessionAffinityTtlMs: Number.isFinite(
-            Number((settings as any)?.codexSessionAffinityTtlMs)
-          )
-            ? Number((settings as any).codexSessionAffinityTtlMs)
-            : null,
           ...(target?.connectionId ? { forcedConnectionId: target.connectionId } : {}),
         }
       );
@@ -530,15 +501,13 @@ export async function handleChat(request: any, clientRawRequest: any = null) {
           stepId?: string | null;
           allowedConnectionIds?: string[] | null;
           failoverBeforeRetry?: boolean;
-          trafficType?: "production" | "shadow";
-          modelAbortSignal?: AbortSignal | null;
         }
       ) =>
         handleSingleModelChat(
           b,
           m,
           clientRawRequest,
-          wrapRequestWithExtraSignal(request, target?.modelAbortSignal ?? null),
+          request,
           combo.name,
           apiKeyInfo,
           telemetry,
@@ -551,7 +520,6 @@ export async function handleChat(request: any, clientRawRequest: any = null) {
             comboStepId: target?.stepId || null,
             comboExecutionKey: target?.executionKey || target?.stepId || null,
             skipUpstreamRetry: target?.failoverBeforeRetry ?? false,
-            trafficType: target?.trafficType === "shadow" ? "shadow" : "production",
             preselectedCredentials: comboPreselectedCredentials.get(
               getComboCredentialCacheKey(m, target)
             ),
@@ -566,7 +534,7 @@ export async function handleChat(request: any, clientRawRequest: any = null) {
       allCombos,
       apiKeyAllowedConnections: apiKeyInfo?.allowedConnections ?? null,
       relayOptions:
-        combo.strategy === "context-relay" || combo.strategy === "auto"
+        combo.strategy === "context-relay"
           ? {
               sessionId,
               config: relayConfig,
@@ -685,7 +653,6 @@ async function handleSingleModelChat(
     skipUpstreamRetry?: boolean;
     preselectedCredentials?: any;
     cachedSettings?: any;
-    trafficType?: "production" | "shadow";
   } = {},
   comboStrategy: string | null = null,
   isCombo: boolean = false
@@ -709,12 +676,21 @@ async function handleSingleModelChat(
     return handleComboChat({
       body,
       combo: redirectCombo,
-      handleSingleModel: (b: any, m: string, target?: any) =>
+      handleSingleModel: (
+        b: any,
+        m: string,
+        target?: {
+          connectionId?: string | null;
+          executionKey?: string | null;
+          stepId?: string | null;
+          failoverBeforeRetry?: boolean;
+        }
+      ) =>
         handleSingleModelChat(
           b,
           m,
           clientRawRequest,
-          wrapRequestWithExtraSignal(request, target?.modelAbortSignal ?? null),
+          request,
           redirectCombo.name ?? modelStr,
           apiKeyInfo,
           telemetry,
@@ -726,7 +702,6 @@ async function handleSingleModelChat(
             comboStepId: null,
             comboExecutionKey: null,
             skipUpstreamRetry: target?.failoverBeforeRetry ?? false,
-            trafficType: target?.trafficType === "shadow" ? "shadow" : "production",
           },
           redirectCombo.strategy ?? "priority",
           false
@@ -742,7 +717,6 @@ async function handleSingleModelChat(
 
   const { provider, model, sourceFormat, targetFormat, extendedContext, apiFormat } = resolved;
   const forceLiveComboTest = runtimeOptions.forceLiveComboTest === true;
-  const isShadowTraffic = runtimeOptions.trafficType === "shadow";
   const hasForcedConnection =
     typeof runtimeOptions.forcedConnectionId === "string" &&
     runtimeOptions.forcedConnectionId.trim().length > 0;
@@ -837,11 +811,6 @@ async function handleSingleModelChat(
               {
                 sessionKey: runtimeOptions.sessionAffinityKey ?? runtimeOptions.sessionId ?? null,
                 excludeConnectionIds: Array.from(excludedConnectionIds),
-                sessionAffinityTtlMs: Number.isFinite(
-                  Number((runtimeOptions.cachedSettings as any)?.codexSessionAffinityTtlMs)
-                )
-                  ? Number((runtimeOptions.cachedSettings as any).codexSessionAffinityTtlMs)
-                  : null,
                 ...(forceLiveComboTest
                   ? {
                       allowSuppressedConnections: true,
@@ -953,11 +922,7 @@ async function handleSingleModelChat(
           ...(workspaceId ? { workspaceId } : {}),
         });
       }
-      if (
-        !isShadowTraffic &&
-        runtimeOptions.sessionId &&
-        body?._omnirouteInternalRequest !== "context-handoff"
-      ) {
+      if (runtimeOptions.sessionId && body?._omnirouteInternalRequest !== "context-handoff") {
         touchSession(runtimeOptions.sessionId, credentials.connectionId);
         startQuotaMonitor(
           runtimeOptions.sessionId,
@@ -972,7 +937,7 @@ async function handleSingleModelChat(
       // 4. Execute chat via core after breaker gate checks (with optional TLS tracking)
       if (telemetry) telemetry.startPhase("connect");
       const { result, tlsFingerprintUsed } = await executeChatWithBreaker({
-        bypassCircuitBreaker: forceLiveComboTest || hasForcedConnection || isShadowTraffic,
+        bypassCircuitBreaker: forceLiveComboTest || hasForcedConnection,
         breaker,
         body: requestBody,
         provider,
@@ -994,7 +959,6 @@ async function handleSingleModelChat(
         providerProfile,
         cachedSettings: runtimeOptions.cachedSettings,
         skipUpstreamRetry: runtimeOptions.skipUpstreamRetry ?? false,
-        trafficType: isShadowTraffic ? "shadow" : "production",
       });
       if (telemetry) telemetry.endPhase();
 
@@ -1021,13 +985,11 @@ async function handleSingleModelChat(
       });
 
       if (result.success) {
-        if (!isShadowTraffic) {
-          clearModelLock(provider, credentials.connectionId, model);
-        }
-        if (!forceLiveComboTest && !isShadowTraffic) {
+        clearModelLock(provider, credentials.connectionId, model);
+        if (!forceLiveComboTest) {
           breaker._onSuccess();
         }
-        if (!isShadowTraffic && injectedHandoff && runtimeOptions.sessionId && comboName) {
+        if (injectedHandoff && runtimeOptions.sessionId && comboName) {
           deleteHandoff(runtimeOptions.sessionId, comboName);
         }
         if (telemetry) telemetry.startPhase("finalize");
@@ -1111,7 +1073,7 @@ async function handleSingleModelChat(
 
       // Emergency fallback for budget exhaustion (402 / billing / quota keywords):
       // reroute to a free model (default provider/model: nvidia + openai/gpt-oss-120b) exactly once.
-      if (!isShadowTraffic && !isCombo && !runtimeOptions.emergencyFallbackTried) {
+      if (!runtimeOptions.emergencyFallbackTried) {
         const fallbackDecision = shouldUseFallback(
           Number(result.status || 0),
           String(result.error || ""),
@@ -1178,11 +1140,7 @@ async function handleSingleModelChat(
       // Daily quota lockout overrides subsequent rate_limited lockout, ensuring lockout until tomorrow 0:00
       let dailyQuotaExhausted = false;
       const errorStr = String(result.error || "");
-      if (isCombo && comboStrategy !== "context-relay" && result.status === 429) {
-        return result.response;
-      }
-
-      if (!isShadowTraffic && result.status === 429 && isDailyQuotaExhausted(errorStr)) {
+      if (result.status === 429 && isDailyQuotaExhausted(errorStr)) {
         // Parse which model is quota-limited
         const match = errorStr.match(/today's quota for model ([^,]+)/);
         const limitedModel = match ? match[1].trim() : model;
@@ -1214,7 +1172,7 @@ async function handleSingleModelChat(
       // 7. Mark account as quota-exhausted only for explicit long-window quota signals.
       // A plain 429/high-traffic response should trigger fallback/cooldown, not poison
       // quotaCache as exhausted for 5 minutes while usage quota may still be available.
-      if (!isShadowTraffic && !dailyQuotaExhausted) {
+      if (!dailyQuotaExhausted) {
         const passthroughModels = credentials.providerSpecificData?.passthroughModels;
         const failureKind =
           result.status === 429
@@ -1238,17 +1196,16 @@ async function handleSingleModelChat(
       const is401 = result.status === 401;
       const skipConnectionDisable = is401 && hasExtraKeys;
 
-      const { shouldFallback, cooldownMs } =
-        skipConnectionDisable || isShadowTraffic
-          ? { shouldFallback: false, cooldownMs: 0 }
-          : await markAccountUnavailable(
-              credentials.connectionId,
-              result.status,
-              result.error,
-              provider,
-              model,
-              providerProfile
-            );
+      const { shouldFallback, cooldownMs } = skipConnectionDisable
+        ? { shouldFallback: false, cooldownMs: 0 }
+        : await markAccountUnavailable(
+            credentials.connectionId,
+            result.status,
+            result.error,
+            provider,
+            model,
+            providerProfile
+          );
 
       if (shouldFallback) {
         if (Number.isFinite(cooldownMs) && cooldownMs > 0) {
@@ -1266,7 +1223,6 @@ async function handleSingleModelChat(
 
       if (
         !forceLiveComboTest &&
-        !isShadowTraffic &&
         !isCombo &&
         PROVIDER_BREAKER_FAILURE_STATUSES.has(Number(result.status))
       ) {