Subagents: stop retrying external completion timeouts (openclaw#41235) (openclaw#43847)

vincentkoc · web-flow · commit 8ad0ca309e84 · 2026-03-12T11:03:06.000-04:00
* Changelog: add subagent announce timeout note

* Tests: cover subagent completion timeout no-retry

* Subagents: stop retrying external completion timeouts

* Config: update subagent announce timeout default docs

* Tests: use fake timers for subagent timeout retry guard
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -34,6 +34,7 @@ Docs: https://docs.openclaw.ai
 - Mattermost/block streaming: fix duplicate message delivery (one threaded, one top-level) when block streaming is active by excluding `replyToId` from the block reply dedup key and adding an explicit `threading` dock to the Mattermost plugin. (#41362) Thanks @mathiasnagler and @vincentkoc.
 - BlueBubbles/self-chat echo dedupe: drop reflected duplicate webhook copies only when a matching `fromMe` event was just seen for the same chat, body, and timestamp, preventing self-chat loops without broad webhook suppression. Related to #32166. (#38442) Thanks @vincentkoc.
 - Models/Kimi Coding: send `anthropic-messages` tools in native Anthropic format again so `kimi-coding` stops degrading tool calls into XML/plain-text pseudo invocations instead of real `tool_use` blocks. (#38669, #39907, #40552) Thanks @opriz.
+- Subagents/completion announce retries: raise the default announce timeout to 90 seconds and stop retrying gateway-timeout failures for externally delivered completion announces, preventing duplicate user-facing completion messages after slow gateway responses. Fixes #41235. Thanks @vasujain00 and @vincentkoc.
 - Sandbox/write: preserve pinned mutation-helper payload stdin so sandboxed `write` no longer reports success while creating empty files. (#43876) Thanks @glitch418x.
 - Gateway/main-session routing: keep TUI and other `mode:UI` main-session sends on the internal surface when `deliver` is enabled, so replies no longer inherit the session's persisted Telegram/WhatsApp route. (#43918) Thanks @obviyus.
 - Doctor/gateway service audit: canonicalize service entrypoint paths before comparing them so symlink-vs-realpath installs no longer trigger false "entrypoint does not match the current install" repair prompts. (#43882) Thanks @ngutman.
diff --git a/src/agents/subagent-announce.timeout.test.ts b/src/agents/subagent-announce.timeout.test.ts
@@ -8,6 +8,12 @@ type GatewayCall = {
 };
 
 const gatewayCalls: GatewayCall[] = [];
+let callGatewayImpl: (request: GatewayCall) => Promise<unknown> = async (request) => {
+  if (request.method === "chat.history") {
+    return { messages: [] };
+  }
+  return {};
+};
 let sessionStore: Record<string, Record<string, unknown>> = {};
 let configOverride: ReturnType<(typeof import("../config/config.js"))["loadConfig"]> = {
   session: {
@@ -27,10 +33,7 @@ let fallbackRequesterResolution: {
 vi.mock("../gateway/call.js", () => ({
   callGateway: vi.fn(async (request: GatewayCall) => {
     gatewayCalls.push(request);
-    if (request.method === "chat.history") {
-      return { messages: [] };
-    }
-    return {};
+    return await callGatewayImpl(request);
   }),
 }));
 
@@ -120,6 +123,12 @@ function findGatewayCall(predicate: (call: GatewayCall) => boolean): GatewayCall
 describe("subagent announce timeout config", () => {
   beforeEach(() => {
     gatewayCalls.length = 0;
+    callGatewayImpl = async (request) => {
+      if (request.method === "chat.history") {
+        return { messages: [] };
+      }
+      return {};
+    };
     sessionStore = {};
     configOverride = {
       session: defaultSessionConfig,
@@ -131,13 +140,13 @@ describe("subagent announce timeout config", () => {
     fallbackRequesterResolution = null;
   });
 
-  it("uses 60s timeout by default for direct announce agent call", async () => {
+  it("uses 90s timeout by default for direct announce agent call", async () => {
     await runAnnounceFlowForTest("run-default-timeout");
 
     const directAgentCall = findGatewayCall(
       (call) => call.method === "agent" && call.expectFinal === true,
     );
-    expect(directAgentCall?.timeoutMs).toBe(60_000);
+    expect(directAgentCall?.timeoutMs).toBe(90_000);
   });
 
   it("honors configured announce timeout for direct announce agent call", async () => {
@@ -166,6 +175,35 @@ describe("subagent announce timeout config", () => {
     expect(completionDirectAgentCall?.timeoutMs).toBe(90_000);
   });
 
+  it("does not retry gateway timeout for externally delivered completion announces", async () => {
+    vi.useFakeTimers();
+    try {
+      callGatewayImpl = async (request) => {
+        if (request.method === "chat.history") {
+          return { messages: [] };
+        }
+        throw new Error("gateway timeout after 90000ms");
+      };
+
+      await expect(
+        runAnnounceFlowForTest("run-completion-timeout-no-retry", {
+          requesterOrigin: {
+            channel: "telegram",
+            to: "12345",
+          },
+          expectsCompletionMessage: true,
+        }),
+      ).resolves.toBe(false);
+
+      const directAgentCalls = gatewayCalls.filter(
+        (call) => call.method === "agent" && call.expectFinal === true,
+      );
+      expect(directAgentCalls).toHaveLength(1);
+    } finally {
+      vi.useRealTimers();
+    }
+  });
+
   it("regression, skips parent announce while descendants are still pending", async () => {
     requesterDepthResolver = () => 1;
     pendingDescendantRuns = 2;
diff --git a/src/agents/subagent-announce.ts b/src/agents/subagent-announce.ts
@@ -51,8 +51,9 @@ import { isAnnounceSkip } from "./tools/sessions-send-helpers.js";
 
 const FAST_TEST_MODE = process.env.OPENCLAW_TEST_FAST === "1";
 const FAST_TEST_RETRY_INTERVAL_MS = 8;
-const DEFAULT_SUBAGENT_ANNOUNCE_TIMEOUT_MS = 60_000;
+const DEFAULT_SUBAGENT_ANNOUNCE_TIMEOUT_MS = 90_000;
 const MAX_TIMER_SAFE_TIMEOUT_MS = 2_147_000_000;
+const GATEWAY_TIMEOUT_PATTERN = /gateway timeout/i;
 let subagentRegistryRuntimePromise: Promise<
   typeof import("./subagent-registry-runtime.js")
 > | null = null;
@@ -107,7 +108,7 @@ const TRANSIENT_ANNOUNCE_DELIVERY_ERROR_PATTERNS: readonly RegExp[] = [
   /no active .* listener/i,
   /gateway not connected/i,
   /gateway closed \(1006/i,
-  /gateway timeout/i,
+  GATEWAY_TIMEOUT_PATTERN,
   /\b(econnreset|econnrefused|etimedout|enotfound|ehostunreach|network error)\b/i,
 ];
 
@@ -133,6 +134,11 @@ function isTransientAnnounceDeliveryError(error: unknown): boolean {
   return TRANSIENT_ANNOUNCE_DELIVERY_ERROR_PATTERNS.some((re) => re.test(message));
 }
 
+function isGatewayTimeoutError(error: unknown): boolean {
+  const message = summarizeDeliveryError(error);
+  return Boolean(message) && GATEWAY_TIMEOUT_PATTERN.test(message);
+}
+
 async function waitForAnnounceRetryDelay(ms: number, signal?: AbortSignal): Promise<void> {
   if (ms <= 0) {
     return;
@@ -160,6 +166,7 @@ async function waitForAnnounceRetryDelay(ms: number, signal?: AbortSignal): Prom
 
 async function runAnnounceDeliveryWithRetry<T>(params: {
   operation: string;
+  noRetryOnGatewayTimeout?: boolean;
   signal?: AbortSignal;
   run: () => Promise<T>;
 }): Promise<T> {
@@ -171,6 +178,9 @@ async function runAnnounceDeliveryWithRetry<T>(params: {
     try {
       return await params.run();
     } catch (err) {
+      if (params.noRetryOnGatewayTimeout && isGatewayTimeoutError(err)) {
+        throw err;
+      }
       const delayMs = DIRECT_ANNOUNCE_TRANSIENT_RETRY_DELAYS_MS[retryIndex];
       if (delayMs == null || !isTransientAnnounceDeliveryError(err) || params.signal?.aborted) {
         throw err;
@@ -789,6 +799,7 @@ async function sendSubagentAnnounceDirectly(params: {
       operation: params.expectsCompletionMessage
         ? "completion direct announce agent call"
         : "direct announce agent call",
+      noRetryOnGatewayTimeout: params.expectsCompletionMessage && shouldDeliverExternally,
       signal: params.signal,
       run: async () =>
         await callGateway({
diff --git a/src/config/types.agent-defaults.ts b/src/config/types.agent-defaults.ts
@@ -279,7 +279,7 @@ export type AgentDefaultsConfig = {
     thinking?: string;
     /** Default run timeout in seconds for spawned sub-agents (0 = no timeout). */
     runTimeoutSeconds?: number;
-    /** Gateway timeout in ms for sub-agent announce delivery calls (default: 60000). */
+    /** Gateway timeout in ms for sub-agent announce delivery calls (default: 90000). */
     announceTimeoutMs?: number;
   };
   /** Optional sandbox settings for non-main sessions. */
