Merge pull request #2710 from MoerAI/fix/rate-limit-hang

fix(runtime-fallback): detect bare 429 rate-limit signals (fixes #2677)

src/hooks/runtime-fallback/error-classifier.test.ts

@@ -31,6 +31,20 @@ describe("runtime-fallback error classifier", () => {
     expect(signal).toBeDefined()
   })
 
+  test("detects too-many-requests auto-retry status signals without countdown text", () => {
+    //#given
+    const info = {
+      status:
+        "Too Many Requests: Sorry, you've exhausted this model's rate limit. Please try a different model.",
+    }
+
+    //#when
+    const signal = extractAutoRetrySignal(info)
+
+    //#then
+    expect(signal).toBeDefined()
+  })
+
   test("treats cooling-down retry messages as retryable", () => {
     //#given
     const error = {

src/hooks/runtime-fallback/error-classifier.ts

@@ -167,7 +167,7 @@ export function extractAutoRetrySignal(info: Record<string, unknown> | undefined
   const combined = candidates.join("\n")
   if (!combined) return undefined
 
-  const isAutoRetry = AUTO_RETRY_PATTERNS.every((test) => test(combined))
+  const isAutoRetry = AUTO_RETRY_PATTERNS.some((test) => test(combined))
   if (isAutoRetry) {
     return { signal: combined }
   }

src/hooks/runtime-fallback/message-update-handler.test.ts

@@ -1,6 +1,7 @@
 import { describe, expect, it } from "bun:test"
 import type { RuntimeFallbackPluginInput } from "./types"
 import { hasVisibleAssistantResponse } from "./visible-assistant-response"
+import { extractAutoRetrySignal } from "./error-classifier"
 
 function createContext(messagesResponse: unknown): RuntimeFallbackPluginInput {
   return {
@@ -53,4 +54,29 @@ describe("hasVisibleAssistantResponse", () => {
     // then
     expect(result).toBe(true)
   })
+
+  it("#given a too-many-requests assistant reply #when visibility is checked #then it is treated as an auto-retry signal", async () => {
+    // given
+    const checkVisibleResponse = hasVisibleAssistantResponse(extractAutoRetrySignal)
+    const ctx = createContext({
+      data: [
+        { info: { role: "user" }, parts: [{ type: "text", text: "latest question" }] },
+        {
+          info: { role: "assistant" },
+          parts: [
+            {
+              type: "text",
+              text: "Too Many Requests: Sorry, you've exhausted this model's rate limit. Please try a different model.",
+            },
+          ],
+        },
+      ],
+    })
+
+    // when
+    const result = await checkVisibleResponse(ctx, "session-rate-limit", undefined)
+
+    // then
+    expect(result).toBe(false)
+  })
 })