fix: complete fast mode fallback and status wiring

vincentkoc · vincentkoc · commit 7f5423ca9717 · 2026-06-22T09:09:48.000+08:00
diff --git a/src/acp/translator.ts b/src/acp/translator.ts
@@ -1498,6 +1498,7 @@ export class AcpGatewayAgent implements Agent {
       modelProvider: session.modelProvider,
       model: session.model,
       fastMode: session.fastMode,
+      effectiveFastMode: session.effectiveFastMode,
       verboseLevel: session.verboseLevel,
       traceLevel: session.traceLevel,
       reasoningLevel: session.reasoningLevel,
diff --git a/src/agents/agent-command.ts b/src/agents/agent-command.ts
@@ -1926,6 +1926,7 @@ async function agentCommandInternal(
                 fastMode === "auto"
                   ? (opts.fastModeAutoOnSeconds ?? fastModeState.fastAutoOnSeconds)
                   : fastModeState.fastAutoOnSeconds,
+              isFinalFallbackAttempt: runOptions?.isFinalFallbackAttempt,
               timeoutMs,
               runTimeoutOverrideMs,
               runId,
diff --git a/src/agents/command/attempt-execution.ts b/src/agents/command/attempt-execution.ts
@@ -478,6 +478,7 @@ export function runAgentAttempt(params: {
   fastMode?: FastMode;
   fastModeStartedAtMs?: number;
   fastModeAutoOnSeconds?: number;
+  isFinalFallbackAttempt?: boolean;
   timeoutMs: number;
   runTimeoutOverrideMs?: number;
   runId: string;
@@ -790,6 +791,7 @@ export function runAgentAttempt(params: {
     fastMode: params.fastMode,
     fastModeStartedAtMs: params.fastModeStartedAtMs,
     fastModeAutoOnSeconds: params.fastModeAutoOnSeconds,
+    isFinalFallbackAttempt: params.isFinalFallbackAttempt,
     verboseLevel: params.resolvedVerboseLevel,
     bashElevated: params.opts.bashElevated,
     timeoutMs: params.timeoutMs,
diff --git a/src/agents/embedded-agent-runner/run.ts b/src/agents/embedded-agent-runner/run.ts
@@ -4123,7 +4123,9 @@ async function runEmbeddedAgentInternal(
           };
         }
       } finally {
-        await maybeEmitFastModeAutoResetBestEffort();
+        if (params.isFinalFallbackAttempt !== false) {
+          await maybeEmitFastModeAutoResetBestEffort();
+        }
         forgetPromptBuildDrainCacheForRun(params.runId);
         stopRuntimeAuthRefreshTimer();
         await runAgentCleanupStep({
diff --git a/src/agents/embedded-agent-runner/run/params.ts b/src/agents/embedded-agent-runner/run/params.ts
@@ -151,6 +151,8 @@ export type RunEmbeddedAgentParams = {
   fastModeAutoOnSeconds?: number;
   /** Shared notification state for nested harnesses that can observe the same tool boundary. */
   fastModeAutoProgressState?: FastModeAutoProgressState;
+  /** True when the outer model fallback loop has reached its final candidate. */
+  isFinalFallbackAttempt?: boolean;
   verboseLevel?: VerboseLevel;
   reasoningLevel?: ReasoningLevel;
   toolResultFormat?: ToolResultFormat;
diff --git a/src/agents/model-fallback.ts b/src/agents/model-fallback.ts
@@ -168,6 +168,7 @@ export function isFallbackSummaryError(err: unknown): err is FallbackSummaryErro
 
 export type ModelFallbackRunOptions = {
   allowTransientCooldownProbe?: boolean;
+  isFinalFallbackAttempt?: boolean;
 };
 
 type ModelFallbackRuntimeContext = {
@@ -1657,7 +1658,10 @@ async function runWithModelFallbackInternal<T>(
       run: params.run,
       ...candidate,
       attempts,
-      options: runOptions,
+      options: {
+        ...runOptions,
+        isFinalFallbackAttempt: i + 1 === candidates.length,
+      },
       // Only the outer fallback loop knows another candidate remains. Carry
       // that fact through this attempt so the embedded runner does not freeze
       // the shared lane before the next candidate can run.
diff --git a/src/status/status-message.ts b/src/status/status-message.ts
@@ -9,6 +9,7 @@ import {
 import { resolveContextTokensForModel } from "../agents/context.js";
 import { DEFAULT_CONTEXT_TOKENS, DEFAULT_MODEL, DEFAULT_PROVIDER } from "../agents/defaults.js";
 import { resolveExtraParams } from "../agents/embedded-agent-runner/extra-params.js";
+import { resolveFastModeState } from "../agents/fast-mode.js";
 import { resolveModelAuthMode } from "../agents/model-auth.js";
 import {
   areRuntimeModelRefsEquivalent,
@@ -25,7 +26,6 @@ import {
   formatProviderModelRef,
   resolveSelectedAndActiveModel,
 } from "../auto-reply/model-runtime.js";
-import { formatFastModeStatusValue } from "../shared/fast-mode.js";
 import type {
   ElevatedLevel,
   ReasoningLevel,
@@ -56,6 +56,7 @@ import {
 } from "../media-understanding/runner.entries.js";
 import type { MediaUnderstandingDecision } from "../media-understanding/types.js";
 import { resolveAgentIdFromSessionKey } from "../routing/session-key.js";
+import { formatFastModeStatusValue } from "../shared/fast-mode.js";
 import { resolveStatusTtsSnapshot } from "../tts/status-config.js";
 import {
   estimateUsageCost,
@@ -882,6 +883,13 @@ export function buildStatusMessage(args: StatusArgs): string {
   const verboseLevel =
     args.resolvedVerbose ?? args.sessionEntry?.verboseLevel ?? args.agent?.verboseDefault ?? "off";
   const fastMode = args.resolvedFast ?? args.sessionEntry?.fastMode ?? false;
+  const fastModeState = resolveFastModeState({
+    cfg: args.config,
+    provider: activeProvider,
+    model: activeModel,
+    agentId: args.agentId,
+    sessionEntry: args.sessionEntry,
+  });
   const reasoningLevel =
     args.resolvedReasoning ??
     args.sessionEntry?.reasoningLevel ??
@@ -970,7 +978,10 @@ export function buildStatusMessage(args: StatusArgs): string {
     `Execution: ${execution.label}`,
     `Runtime: ${agentRuntimeLabel}`,
     `Think: ${thinkLevel}`,
-    `Fast: ${formatFastModeStatusValue({ mode: fastMode })}`,
+    `Fast: ${formatFastModeStatusValue({
+      mode: fastMode,
+      fastAutoOnSeconds: fastModeState.fastAutoOnSeconds,
+    })}`,
     textVerbosity ? `Text: ${textVerbosity}` : null,
     verboseLabel,
     traceLabel,
