Suppress transient runner failures in channels (#87069)

src/auto-reply/reply/agent-runner-execution.test.ts

@@ -4854,6 +4854,95 @@ describe("runAgentTurnWithFallback", () => {
     },
   );
 
+  it.each(["group", "channel"] as const)(
+    "keeps classified non-transient failures visible in Discord %s chats",
+    async (chatType) => {
+      state.runEmbeddedPiAgentMock.mockRejectedValueOnce(
+        new Error('No API key found for provider "openai"'),
+      );
+
+      const runAgentTurnWithFallback = await getRunAgentTurnWithFallback();
+      const result = await runAgentTurnWithFallback(
+        createMinimalRunAgentTurnParams({
+          sessionCtx: {
+            Provider: "discord",
+            Surface: "discord",
+            ChatType: chatType,
+            GroupSubject: "agent group",
+            GroupChannel: "#general",
+            MessageSid: "msg",
+          } as unknown as TemplateContext,
+        }),
+      );
+
+      expect(result.kind).toBe("final");
+      if (result.kind === "final") {
+        expect(result.payload.text).not.toBe(SILENT_REPLY_TOKEN);
+        expect(result.payload.text).toContain('Missing API key for provider "openai"');
+      }
+    },
+  );
+
+  it.each(["group", "channel"] as const)(
+    "keeps rate-limit fallback copy out of Discord %s chats",
+    async (chatType) => {
+      state.runEmbeddedPiAgentMock.mockRejectedValueOnce(new Error("429 rate limit exceeded"));
+
+      const runAgentTurnWithFallback = await getRunAgentTurnWithFallback();
+      const result = await runAgentTurnWithFallback(
+        createMinimalRunAgentTurnParams({
+          sessionCtx: {
+            Provider: "discord",
+            Surface: "discord",
+            ChatType: chatType,
+            GroupSubject: "agent group",
+            GroupChannel: "#general",
+            MessageSid: "msg",
+          } as unknown as TemplateContext,
+        }),
+      );
+
+      expect(result.kind).toBe("final");
+      if (result.kind === "final") {
+        expect(result.payload.text).toBe(SILENT_REPLY_TOKEN);
+      }
+    },
+  );
+
+  it("surfaces rate-limit fallback copy in Discord group chats when silentReply.group is disallow", async () => {
+    state.runEmbeddedPiAgentMock.mockRejectedValueOnce(new Error("429 rate limit exceeded"));
+
+    const followupRun = createFollowupRun();
+    followupRun.run.config = {
+      agents: {
+        defaults: {
+          silentReply: { group: "disallow" },
+        },
+      },
+    };
+
+    const runAgentTurnWithFallback = await getRunAgentTurnWithFallback();
+    const result = await runAgentTurnWithFallback(
+      createMinimalRunAgentTurnParams({
+        followupRun,
+        sessionCtx: {
+          Provider: "discord",
+          Surface: "discord",
+          ChatType: "group",
+          GroupSubject: "agent group",
+          GroupChannel: "#general",
+          MessageSid: "msg",
+        } as unknown as TemplateContext,
+      }),
+    );
+
+    expect(result.kind).toBe("final");
+    if (result.kind === "final") {
+      expect(result.payload.text).not.toBe(SILENT_REPLY_TOKEN);
+      expect(result.payload.text).toContain("rate-limited");
+    }
+  });
+
   it("uses compact generic copy for raw runner failures in normal Discord direct chats", async () => {
     state.runEmbeddedPiAgentMock.mockRejectedValueOnce(
       new Error("openai-codex/gpt-5.5 ended with an incomplete terminal response"),

src/auto-reply/reply/agent-runner-execution.ts

@@ -643,12 +643,17 @@ function resolveExternalRunFailureTextForConversation(params: {
   text: string;
   sessionCtx: TemplateContext;
   isGenericRunnerFailure: boolean;
+  suppressInNonDirect?: boolean;
   cfg?: OpenClawConfig;
 }): string {
   if (!isNonDirectConversationContext(params.sessionCtx)) {
     return params.text;
   }
-  if (!params.isGenericRunnerFailure && !params.text.includes(AGENT_FAILED_BEFORE_REPLY_TEXT)) {
+  if (
+    !params.suppressInNonDirect &&
+    !params.isGenericRunnerFailure &&
+    !params.text.includes(AGENT_FAILED_BEFORE_REPLY_TEXT)
+  ) {
     return params.text;
   }
   // Match normal reply routing: default group/channel failures stay silent,
@@ -826,6 +831,7 @@ export function buildKnownAgentRunFailureReplyPayload(params: {
         text: buildRateLimitCooldownMessage(params.err),
         sessionCtx: params.sessionCtx,
         isGenericRunnerFailure: false,
+        suppressInNonDirect: true,
         cfg: params.cfg,
       }),
     });
@@ -837,6 +843,7 @@ export function buildKnownAgentRunFailureReplyPayload(params: {
         text: rateLimitOrOverloadedCopy,
         sessionCtx: params.sessionCtx,
         isGenericRunnerFailure: false,
+        suppressInNonDirect: true,
         cfg: params.cfg,
       }),
     });
@@ -2806,6 +2813,7 @@ export async function runAgentTurnWithFallback(params: {
         text: fallbackText,
         sessionCtx: params.sessionCtx,
         isGenericRunnerFailure: externalRunFailureReply?.isGenericRunnerFailure ?? false,
+        suppressInNonDirect: Boolean(isRateLimit || rateLimitOrOverloadedCopy),
         cfg: params.followupRun.run.config,
       });
 
@@ -2870,7 +2878,13 @@ export async function runAgentTurnWithFallback(params: {
       if (formattedErrorCandidate) {
         runResult.payloads = [
           markAgentRunFailureReplyPayload({
-            text: formattedErrorCandidate,
+            text: resolveExternalRunFailureTextForConversation({
+              text: formattedErrorCandidate,
+              sessionCtx: params.sessionCtx,
+              isGenericRunnerFailure: false,
+              suppressInNonDirect: true,
+              cfg: params.followupRun.run.config,
+            }),
             isError: true,
           }),
         ];