Merge pull request #3556 from MoerAI/fix/fallback-credit-exhaustion

Fixes #3519. Quota exhaustion now advances the runtime fallback chain instead of stopping. Surgical change with parallel test updates. Pre-existing prompt-text test mismatch on the older base disappeared after rebasing onto current dev — all checks green.

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

@@ -127,7 +127,7 @@ export function classifyErrorType(error: unknown): string | undefined {
     errorName?.includes("billingerror") ||
     /quota.?exceeded/i.test(message) ||
     /subscription.*quota/i.test(message) ||
-    /insufficient.?quota/i.test(message) ||
+    /insufficient.?(?:quota|balance|funds?)/i.test(message) ||
     /billing.?(?:hard.?)?limit/i.test(message) ||
     /exhausted\s+your\s+capacity/i.test(message) ||
     /out\s+of\s+credits?/i.test(message) ||
@@ -169,10 +169,9 @@ export function isRetryableError(error: unknown, retryOnErrors: number[]): boole
   }
 
   if (errorType === "quota_exceeded") {
-    // When a provider signals an auto-retry (e.g. "retrying in ~2 weeks"),
+    // Quota exhaustion means the current model/provider cannot serve requests.
-    // we should still trigger fallback to another model rather than STOP.
+    // Trigger fallback to the next configured model instead of stopping entirely.
-    const hasAutoRetrySignal = /retrying\s+in/i.test(message)
+    return true
-    return hasAutoRetrySignal
   }
 
   if (statusCode && retryOnErrors.includes(statusCode)) {

src/hooks/runtime-fallback/index.test.ts

@@ -293,7 +293,7 @@ describe("runtime-fallback", () => {
       expect(errorLog).toBeDefined()
     })
 
-    test("should NOT trigger fallback for quota exhaustion without auto-retry signal (STOP classification)", async () => {
+    test("should trigger fallback for quota exhaustion to try next configured model", async () => {
       const hook = createRuntimeFallbackHook(createMockPluginInput(), {
         config: createMockConfig({ notify_on_fallback: false }),
         pluginConfig: createMockPluginConfigWithCategoryFallback(["zai-coding-plan/glm-5.1"]),
@@ -318,11 +318,9 @@ describe("runtime-fallback", () => {
         },
       })
 
+      // quota exhaustion now triggers fallback to the next model
       const fallbackLog = logCalls.find((c) => c.msg.includes("Preparing fallback"))
-      expect(fallbackLog).toBeUndefined()
+      expect(fallbackLog).toBeDefined()
-
-      const skipLog = logCalls.find((c) => c.msg.includes("Error not retryable"))
-      expect(skipLog).toBeDefined()
     })
 
     test("should continue fallback chain when fallback model is not found", async () => {
@@ -2071,7 +2069,7 @@ describe("runtime-fallback", () => {
       expect(retriedModels).toContain("openai/gpt-5.3-codex")
     })
 
-    test("does NOT trigger fallback for quota exhaustion in error parts without auto-retry signal (STOP classification)", async () => {
+    test("triggers fallback for quota exhaustion in error parts to try next model", async () => {
       const retriedModels: string[] = []
 
       const hook = createRuntimeFallbackHook(
@@ -2119,10 +2117,8 @@ describe("runtime-fallback", () => {
         },
       })
 
-      expect(retriedModels).toHaveLength(0)
+      // quota exhaustion now triggers fallback to next configured model
-
+      expect(retriedModels.length).toBeGreaterThanOrEqual(1)
-      const skipLog = logCalls.find((c) => c.msg.includes("message.updated error not retryable"))
-      expect(skipLog).toBeDefined()
     })
 
     test("triggers fallback when message has mixed text and error parts", async () => {

src/hooks/runtime-fallback/provider-matrix.test.ts

@@ -18,7 +18,8 @@ describe("runtime-fallback provider matrix quota tests", () => {
 
       //#then
       expect(errorType).toBe("quota_exceeded")
-      expect(retryable).toBe(false)
+      // quota exhaustion triggers fallback to next configured model
+      expect(retryable).toBe(true)
     })
 
     test("classifies OpenAI billing_hard_limit error as quota_exceeded", () => {
@@ -54,7 +55,7 @@ describe("runtime-fallback provider matrix quota tests", () => {
   })
 
   describe("Anthropic provider", () => {
-    test("classifies Anthropic quota exceeded as non-retryable", () => {
+    test("classifies Anthropic quota exceeded as retryable to trigger fallback", () => {
       //#given
       const error = {
         name: "QuotaExceededError",
@@ -68,10 +69,11 @@ describe("runtime-fallback provider matrix quota tests", () => {
 
       //#then
       expect(errorType).toBe("quota_exceeded")
-      expect(retryable).toBe(false)
+      // quota exhaustion triggers fallback to next configured model
+      expect(retryable).toBe(true)
     })
 
-    test("classifies Anthropic subscription quota as non-retryable", () => {
+    test("classifies Anthropic subscription quota as retryable to trigger fallback", () => {
       //#given
       const error = {
         name: "AI_APICallError",
@@ -85,7 +87,8 @@ describe("runtime-fallback provider matrix quota tests", () => {
 
       //#then
       expect(errorType).toBe("quota_exceeded")
-      expect(retryable).toBe(false)
+      // quota exhaustion triggers fallback to next configured model
+      expect(retryable).toBe(true)
     })
 
     test("classifies Anthropic cooling down with retry signal as retryable (auto-retry pattern)", () => {
@@ -139,7 +142,8 @@ describe("runtime-fallback provider matrix quota tests", () => {
 
       //#then
       expect(errorType).toBe("quota_exceeded")
-      expect(retryable).toBe(false)
+      // quota exhaustion triggers fallback to next configured model
+      expect(retryable).toBe(true)
     })
 
     test("classifies Google rate limit exceeded as retryable", () => {
@@ -274,7 +278,7 @@ describe("runtime-fallback provider matrix quota tests", () => {
       expect(retryable).toBe(true)
     })
 
-    test("402 payment required is NOT retryable", () => {
+    test("402 payment required triggers fallback via quota_exceeded", () => {
       //#given
       const error = { statusCode: 402, message: "Payment Required" }
 
@@ -282,7 +286,8 @@ describe("runtime-fallback provider matrix quota tests", () => {
       const retryable = isRetryableError(error, [429, 500, 502, 503, 504])
 
       //#then
-      expect(retryable).toBe(false)
+      // payment required is classified as quota_exceeded, which triggers fallback
+      expect(retryable).toBe(true)
     })
 
     test("500 server error is retryable", () => {

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

@@ -3,7 +3,7 @@ import { describe, expect, test } from "bun:test"
 import { classifyErrorType, isRetryableError } from "./error-classifier"
 
 describe("runtime-fallback quota error regressions", () => {
-  test("classifies subscription quota errors as quota_exceeded and stops retry", () => {
+  test("classifies subscription quota errors as quota_exceeded and triggers fallback", () => {
     //#given
     const error = {
       name: "AI_APICallError",
@@ -16,10 +16,11 @@ describe("runtime-fallback quota error regressions", () => {
 
     //#then
     expect(errorType).toBe("quota_exceeded")
-    expect(retryable).toBe(false)
+    // quota exhaustion should trigger fallback to the next model
+    expect(retryable).toBe(true)
   })
 
-  test("treats HTTP 402 payment required as non-retryable", () => {
+  test("treats HTTP 402 payment required as fallback-eligible", () => {
     //#given
     const error = { statusCode: 402, message: "Payment Required" }
 
@@ -27,7 +28,8 @@ describe("runtime-fallback quota error regressions", () => {
     const retryable = isRetryableError(error, [429, 500, 502, 503, 504])
 
     //#then
-    expect(retryable).toBe(false)
+    // payment failure triggers fallback to a different provider/model
+    expect(retryable).toBe(true)
   })
 
   test("keeps HTTP 429 rate limit retryable", () => {
@@ -41,7 +43,7 @@ describe("runtime-fallback quota error regressions", () => {
     expect(retryable).toBe(true)
   })
 
-  test("classifies quota error names as quota_exceeded without retry", () => {
+  test("classifies quota error names as quota_exceeded and triggers fallback", () => {
     //#given
     const error = { name: "QuotaExceededError", message: "Request failed." }
 
@@ -51,6 +53,7 @@ describe("runtime-fallback quota error regressions", () => {
 
     //#then
     expect(errorType).toBe("quota_exceeded")
-    expect(retryable).toBe(false)
+    // quota errors trigger fallback to next configured model
+    expect(retryable).toBe(true)
   })
 })