fix(active-memory): preserve setup grace for embedded recall

Apply setupGraceTimeoutMs to the embedded recall runner as well as the outer Active Memory watchdog.

Co-authored-by: volcano303 <75143900+volcano303@users.noreply.github.com>

CHANGELOG.md

@@ -19,6 +19,7 @@ Docs: https://docs.openclaw.ai
 
 ### Fixes
 
+- Active Memory: apply `setupGraceTimeoutMs` to the embedded recall runner as well as the outer prompt-build watchdog, so very-cold first recalls keep the configured setup grace end-to-end. (#74480) Thanks @volcano303.
 - Plugins/tools: keep disabled bundled tool plugins out of explicit runtime allowlist ownership and fall back from loaded-but-empty channel registries to tool-bearing plugin registries, so Active Memory can use bundled `memory-core` search/get tools even when `memory-lancedb` is disabled. Fixes #76603. Thanks @jwong-art.
 - Channels/QQ Bot: resolve structured `clientSecret` SecretRefs before QQ token exchange, expose the QQ Bot secret contract to secrets tooling, and reject legacy `secretref:/...` marker strings. (#74772) Thanks @xialonglee.
 - Plugins/externalization: keep official ACPX, Google Chat, and LINE install specs on production package names, leaving beta-tag probing to the explicit OpenClaw beta update channel. Thanks @vincentkoc.

docs/concepts/active-memory.md

@@ -658,13 +658,9 @@ blocking prompt-build hook budget by default and move cold-start setup grace
 behind explicit `setupGraceTimeoutMs` config, so the plugin no longer silently
 extends 15000 ms configs to 45000 ms on the main lane."_
 
-The embedded recall runner currently still receives the raw `timeoutMs` value
-as its inner budget; the in-flight fix to extend that with `setupGraceTimeoutMs`
-is tracked at [#74480](https://github.com/openclaw/openclaw/pull/74480). Until
-that lands, very-cold first recalls can still time out at the inner layer even
-with `setupGraceTimeoutMs` set — though the outer-layer setting still
-substantially mitigates the symptom by giving the prompt-build hook room to
-cover the warm-up window.
+The embedded recall runner uses the same effective timeout budget, so
+`setupGraceTimeoutMs` covers both the outer prompt-build watchdog and the inner
+blocking recall run.
 
 For resource-tight gateways where cold-start latency is a known trade-off,
 lower values (5000–15000 ms) work too — the trade-off is a higher chance of
@@ -735,8 +731,7 @@ default `memory-core` path uses `memory_search`; `memory-lancedb` uses
     around the first eligible reply after a restart.
 
     See [Cold-start grace](#cold-start-grace) under Recommended setup for the
-    recommended `setupGraceTimeoutMs` value (and the open caveat about the
-    embedded recall budget tracked at #74480).
+    recommended `setupGraceTimeoutMs` value.
 
   </Accordion>
 </AccordionGroup>

extensions/active-memory/index.test.ts

@@ -2256,7 +2256,7 @@ describe("active-memory plugin", () => {
     );
 
     expect(result?.prependContext).toContain("remember the ramen place");
-    expect(runEmbeddedPiAgent.mock.calls.at(-1)?.[0]?.timeoutMs).toBe(CONFIGURED_TIMEOUT_MS);
+    expect(runEmbeddedPiAgent.mock.calls.at(-1)?.[0]?.timeoutMs).toBe(CONFIGURED_TIMEOUT_MS + 100);
     const infoLines = vi
       .mocked(api.logger.info)
       .mock.calls.map((call: unknown[]) => String(call[0]));

extensions/active-memory/index.ts

@@ -2273,6 +2273,7 @@ async function runRecallSubagent(params: {
 
   try {
     const embeddedConfig = applyActiveMemoryRuntimeConfigSnapshot(params.api.config, params.config);
+    const embeddedTimeoutMs = params.config.timeoutMs + params.config.setupGraceTimeoutMs;
     const result = await params.api.runtime.agent.runEmbeddedPiAgent({
       sessionId: subagentSessionId,
       sessionKey: subagentSessionKey,
@@ -2286,7 +2287,7 @@ async function runRecallSubagent(params: {
       prompt,
       provider: modelRef.provider,
       model: modelRef.model,
-      timeoutMs: params.config.timeoutMs,
+      timeoutMs: embeddedTimeoutMs,
       runId: subagentSessionId,
       trigger: "manual",
       toolsAllow: ["memory_recall", "memory_search", "memory_get"],