fix(core): raise stream idle timeout default to 5m and hint the env knob (#6107)

The streaming inactivity watchdog aborted any stream silent for 120000ms,
which breaks large-prompt ingest, /compress reprocessing, and long thinking
phases on local models and reasoning proxies that legitimately emit no SSE
deltas for over two minutes. Raise DEFAULT_STREAM_IDLE_TIMEOUT_MS to 300000
(5 minutes) and append an actionable hint to the timeout error pointing at
QWEN_STREAM_IDLE_TIMEOUT_MS (or 0 to disable).

Refs #5975

Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>
Co-authored-by: Shaojin Wen <shaojin.wensj@alibaba-inc.com>

packages/core/src/core/openaiContentGenerator/constants.ts

@@ -1,8 +1,10 @@
 export const DEFAULT_TIMEOUT = 120000;
 // Inactivity (no-chunk) timeout for streaming responses. The SDK `timeout`
 // only bounds connect + first response, so a stream that returns 200 then
-// goes silent is otherwise unbounded; this watchdog aborts it.
-export const DEFAULT_STREAM_IDLE_TIMEOUT_MS = 120000;
+// goes silent is otherwise unbounded. The 4-minute default gives large-prompt
+// ingest and long thinking phases room while staying below
+// LoggingContentGenerator's stream-span idle timer, so this watchdog fires first.
+export const DEFAULT_STREAM_IDLE_TIMEOUT_MS = 240000;
 // Env override (deployment knob) for the streaming inactivity timeout, so a
 // daemon deployment can tune it without code — the same way the QWEN_SERVE_*
 // params are set. An explicit ContentGeneratorConfig.streamIdleTimeoutMs still

packages/core/src/core/openaiContentGenerator/pipeline.test.ts

@@ -3148,7 +3148,9 @@ describe('ContentGenerationPipeline', () => {
       const err = await captured;
       expect(err).toBeInstanceOf(StreamInactivityTimeoutError);
       expect((err as Error).message).toBe(
-        'No stream activity for 1000ms after 0 chunks (stream lifetime: 1000ms)',
+        'No stream activity for 1000ms after 0 chunks ' +
+          '(stream lifetime: 1000ms). Set QWEN_STREAM_IDLE_TIMEOUT_MS ' +
+          'to increase this window (or 0 to disable it).',
       );
       expect(err).toMatchObject({ code: 'ETIMEDOUT' });
       expect((err as StreamInactivityTimeoutError).chunksReceived).toBe(0);
@@ -3157,6 +3159,29 @@ describe('ContentGenerationPipeline', () => {
       expect(mockErrorHandler.handle).not.toHaveBeenCalled();
     });
 
+    it('includes the idle detail and env override hint in timeout errors', async () => {
+      const gated = gatedStream(); // never push/end → silent
+      (mockClient.chat.completions.create as Mock).mockResolvedValue(
+        gated.stream,
+      );
+      const p = buildPipeline(1000);
+      const gen = await p.executeStream(
+        streamingRequest(new AbortController().signal),
+        'id',
+      );
+      const captured = (async () => {
+        for await (const _ of gen) {
+          /* drain */
+        }
+      })().catch((e: unknown) => e);
+      await vi.advanceTimersByTimeAsync(1000);
+      const err = await captured;
+      expect(err).toBeInstanceOf(StreamInactivityTimeoutError);
+      const message = (err as Error).message;
+      expect(message).toContain('No stream activity for 1000ms after 0 chunks');
+      expect(message).toContain('QWEN_STREAM_IDLE_TIMEOUT_MS');
+    });
+
     it('uses the default stream idle timeout when no override is configured', async () => {
       const gated = gatedStream(); // never push/end → silent
       (mockClient.chat.completions.create as Mock).mockResolvedValue(
@@ -3524,7 +3549,7 @@ describe('ContentGenerationPipeline', () => {
       (mockClient.chat.completions.create as Mock).mockResolvedValue(
         gated.stream,
       );
-      const p = buildPipeline(); // no config; invalid env → default (120000ms)
+      const p = buildPipeline(); // no config; invalid env → default
       const gen = await p.executeStream(
         streamingRequest(new AbortController().signal),
         'id',
@@ -3557,7 +3582,7 @@ describe('ContentGenerationPipeline', () => {
       (mockClient.chat.completions.create as Mock).mockResolvedValue(
         gated.stream,
       );
-      const p = buildPipeline(); // no config; oversized env → default (120000ms)
+      const p = buildPipeline(); // no config; oversized env → default
       const gen = await p.executeStream(
         streamingRequest(new AbortController().signal),
         'id',
@@ -3657,7 +3682,7 @@ describe('ContentGenerationPipeline', () => {
       (mockClient.chat.completions.create as Mock).mockResolvedValue(
         gated.stream,
       );
-      // Config is oversized → rejected; env = 4000 → used (not default 120000).
+      // Config is oversized → rejected; env = 4000 → used (not default).
       const p = buildPipeline(MAX_STREAM_IDLE_TIMEOUT_MS + 1);
       const gen = await p.executeStream(
         streamingRequest(new AbortController().signal),
@@ -3673,7 +3698,7 @@ describe('ContentGenerationPipeline', () => {
       expect(settled).toBe(false); // not yet at the env value
       await vi.advanceTimersByTimeAsync(1);
       await consume;
-      expect(settled).toBe(true); // trips at 4000ms from the env (not 120000 default)
+      expect(settled).toBe(true); // trips at 4000ms from the env (not default)
     });
 
     it('disables the watchdog when QWEN_STREAM_IDLE_TIMEOUT_MS=0', async () => {

packages/core/src/core/openaiContentGenerator/pipeline.ts

@@ -56,7 +56,10 @@ export class StreamInactivityTimeoutError extends Error {
     readonly streamLifetimeMs: number,
   ) {
     super(
-      `No stream activity for ${idleMs}ms after ${chunksReceived} chunks (stream lifetime: ${streamLifetimeMs}ms)`,
+      `No stream activity for ${idleMs}ms after ${chunksReceived} chunks ` +
+        `(stream lifetime: ${streamLifetimeMs}ms). Set ` +
+        `${QWEN_STREAM_IDLE_TIMEOUT_MS_ENV} to increase this window ` +
+        `(or 0 to disable it).`,
     );
     this.name = 'StreamInactivityTimeoutError';
   }