fix(ai): retry xAI capacity stream errors

packages/ai/src/protocols/open-responses.ts

@@ -1073,7 +1073,13 @@ const providerErrorMessage = (event: Event, fallback: string): string => {
 }
 
 export const providerFailure = (id: string, event: Event, fallback: string) => {
-  const code = event.code || event.error?.code || event.response?.error?.code || undefined
+  const code =
+    event.code ||
+    event.error?.code ||
+    event.error?.type ||
+    event.response?.error?.code ||
+    event.response?.error?.type ||
+    undefined
   const message = providerErrorMessage(event, fallback)
   const status =
     typeof event.status === "number"
@@ -1084,7 +1090,7 @@ export const providerFailure = (id: string, event: Event, fallback: string) => {
   return new AIError({
     module: id,
     method: "stream",
-    reason: classifyProviderFailure({ message, code, status }),
+    reason: classifyProviderFailure({ message, code, status, stream: true }),
   })
 }
 

packages/ai/src/provider-error.ts

@@ -74,9 +74,11 @@ const INVALID_REQUEST_CODES = new Set(["invalid_prompt", "invalid_request_error"
 const RATE_LIMIT_TEXT = /rate increased too quickly|rate[-_\s]?limit|too[_\s]?many[_\s]?requests/i
 const QUOTA_TEXT = /insufficient[-_\s]?quota|quota[-_\s]?exceeded/i
 const CONTENT_POLICY_TEXT = /content[-_\s]?policy|content_filter|safety/i
+const TRANSIENT_TEXT = /\btry again (?:later|in\b)|\b(?:currently|temporarily) at capacity\b|\bretry your request\b/i
 
 export interface ProviderFailure {
   readonly message: string
+  readonly stream?: boolean | undefined
   readonly status?: number | undefined
   readonly code?: string | undefined
   readonly retryAfterMs?: number | undefined
@@ -133,6 +135,12 @@ export function classifyProviderFailure(input: ProviderFailure): AIError["reason
       status: input.status,
       retryAfterMs: input.retryAfterMs,
     })
+  if (TRANSIENT_TEXT.test(text))
+    return new ProviderInternalReason({
+      ...common,
+      status: input.status,
+      retryAfterMs: input.retryAfterMs,
+    })
   if (input.status === 429) {
     return new RateLimitReason({
       ...common,
@@ -149,6 +157,12 @@ export function classifyProviderFailure(input: ProviderFailure): AIError["reason
   if (codes.some((code) => INVALID_REQUEST_CODES.has(code))) return new InvalidRequestReason(common)
   if (input.status === 400 || input.status === 404 || input.status === 413 || input.status === 422)
     return new InvalidRequestReason(common)
+  if (input.stream)
+    return new ProviderInternalReason({
+      ...common,
+      status: input.status,
+      retryAfterMs: input.retryAfterMs,
+    })
   return new UnknownProviderReason({ ...common, status: input.status })
 }
 

packages/ai/test/provider-error.test.ts

@@ -75,6 +75,26 @@ describe("provider error classification", () => {
     ).toEqual(["ProviderInternal", "ProviderInternal", "ProviderInternal"])
   })
 
+  test("classifies transient gateway message fallbacks as provider internal", () => {
+    expect(
+      [
+        "The model is currently at capacity due to high demand.",
+        "The service is temporarily at capacity.",
+        "Please try again in a few minutes.",
+        "Please retry your request shortly.",
+      ].map((message) => classifyProviderFailure({ message })._tag),
+    ).toEqual(["ProviderInternal", "ProviderInternal", "ProviderInternal", "ProviderInternal"])
+  })
+
+  test("classifies provider stream errors as provider internal", () => {
+    expect(
+      [
+        "The model is currently at capacity due to high demand. Please try again in a few minutes, or use a higher service tier for priority processing: https://docs.x.ai/developers/advanced-api-usage/priority-processing",
+        "The model is temporarily unavailable.",
+      ].map((message) => classifyProviderFailure({ message, stream: true })._tag),
+    ).toEqual(["ProviderInternal", "ProviderInternal"])
+  })
+
   test("classifies transient client statuses as provider internal", () => {
     expect([408, 409].map((status) => classifyProviderFailure({ message: `HTTP ${status}`, status })._tag)).toEqual([
       "ProviderInternal",

packages/ai/test/provider/openai-responses.test.ts

@@ -809,7 +809,7 @@ describe("OpenAI Responses route", () => {
         "ProviderInternal",
         "RateLimit",
         "ProviderInternal",
-        "UnknownProvider",
+        "ProviderInternal",
       ])
     }),
   )
@@ -2661,6 +2661,19 @@ describe("OpenAI Responses route", () => {
     }),
   )
 
+  it.effect("classifies code-less capacity errors as transient", () =>
+    Effect.gen(function* () {
+      const message =
+        "The model is currently at capacity due to high demand. Please try again in a few minutes, or use a higher service tier for priority processing: https://docs.x.ai/developers/advanced-api-usage/priority-processing"
+      const error = yield* LLMClient.generate(request).pipe(
+        Effect.provide(fixedResponse(sseEvents({ type: "error", message }))),
+        Effect.flip,
+      )
+
+      expect(error.reason).toMatchObject({ _tag: "ProviderInternal", message })
+    }),
+  )
+
   it.effect("falls back to error code when no message is present", () =>
     Effect.gen(function* () {
       const error = yield* LLMClient.generate(request).pipe(
@@ -2801,7 +2814,7 @@ describe("OpenAI Responses route", () => {
         Effect.flip,
       )
 
-      expect(error.reason).toMatchObject({ _tag: "UnknownProvider", message: "Something went wrong" })
+      expect(error.reason).toMatchObject({ _tag: "ProviderInternal", message: "Something went wrong" })
     }),
   )
 
@@ -2812,7 +2825,7 @@ describe("OpenAI Responses route", () => {
         Effect.flip,
       )
 
-      expect(error.reason).toMatchObject({ _tag: "UnknownProvider", message: "OpenAI Responses stream error" })
+      expect(error.reason).toMatchObject({ _tag: "ProviderInternal", message: "OpenAI Responses stream error" })
     }),
   )
 
@@ -2823,7 +2836,7 @@ describe("OpenAI Responses route", () => {
         Effect.flip,
       )
 
-      expect(error.reason).toMatchObject({ _tag: "UnknownProvider", message: "OpenAI Responses stream error" })
+      expect(error.reason).toMatchObject({ _tag: "ProviderInternal", message: "OpenAI Responses stream error" })
     }),
   )
 
@@ -2834,7 +2847,7 @@ describe("OpenAI Responses route", () => {
         Effect.flip,
       )
 
-      expect(error.reason).toMatchObject({ _tag: "UnknownProvider", message: "OpenAI Responses response failed" })
+      expect(error.reason).toMatchObject({ _tag: "ProviderInternal", message: "OpenAI Responses response failed" })
     }),
   )