fix: preserve provider error status (#39976)

packages/ai/src/provider-error.ts

@@ -16,7 +16,6 @@ import {
 
 const patterns = [
   /prompt is too long/i,
-  /request_too_large/i,
   /input is too long for requested model/i,
   /exceeds the context window/i,
   /exceeds (?:the )?(?:model'?s )?maximum context length(?: of [\d,]+ tokens?|\s*\([\d,]+\))/i,
@@ -33,7 +32,6 @@ const patterns = [
   /context window exceeds limit/i,
   /exceeded model token limit/i,
   /context[_ ]length[_ ]exceeded/i,
-  /request entity too large/i,
   /context length is only \d+ tokens/i,
   /input length.*exceeds.*context length/i,
   /prompt too long; exceeded (?:max )?context length/i,
@@ -44,11 +42,15 @@ const patterns = [
   /token limit exceeded/i,
 ]
 
+const payloadPatterns = [/request_too_large/i, /request entity too large/i, /payload too large/i, /request too large/i]
+
 const exclusions = [/^(throttling error|service unavailable):/i, /rate limit/i, /too many requests/i]
 
 export const isContextOverflow = (message: string) =>
   !exclusions.some((pattern) => pattern.test(message)) &&
-  (patterns.some((pattern) => pattern.test(message)) || /^4(00|13)\s*(status code)?\s*\(no body\)/i.test(message))
+  (patterns.some((pattern) => pattern.test(message)) || /^400\s*(status code)?\s*\(no body\)/i.test(message))
+
+export const isPayloadTooLarge = (message: string) => payloadPatterns.some((pattern) => pattern.test(message))
 
 export const isContextOverflowFailure = (failure: unknown) =>
   failure instanceof LLMError
@@ -100,6 +102,8 @@ export function classifyProviderFailure(input: ProviderFailure): LLMError["reaso
       isContextOverflow(text))
   )
     return new InvalidRequestReason({ ...common, classification: "context-overflow" })
+  if (input.status === 413 || isPayloadTooLarge(text))
+    return new InvalidRequestReason({ ...common, classification: "payload-too-large" })
   if (CONTENT_POLICY_TEXT.test(text)) return new ContentPolicyReason(common)
   if (codes.some((code) => QUOTA_CODES.has(code)) || (input.status === 429 && QUOTA_TEXT.test(text)))
     return new QuotaExceededReason(common)
@@ -142,12 +146,7 @@ export function classifyProviderFailure(input: ProviderFailure): LLMError["reaso
       retryAfterMs: input.retryAfterMs,
     })
   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
-  )
+  if (input.status === 400 || input.status === 404 || input.status === 413 || input.status === 422)
     return new InvalidRequestReason(common)
   return new UnknownProviderReason({ ...common, status: input.status })
 }

packages/ai/src/schema/errors.ts

@@ -2,7 +2,7 @@ import { Schema } from "effect"
 import { Tool } from "@opencode-ai/schema/tool"
 import { ModelID, ProviderID, ProviderMetadata, RouteID } from "./ids"
 
-export const ProviderFailureClassification = Schema.Literal("context-overflow")
+export const ProviderFailureClassification = Schema.Literals(["context-overflow", "payload-too-large"])
 export type ProviderFailureClassification = typeof ProviderFailureClassification.Type
 
 export class HttpRequestDetails extends Schema.Class<HttpRequestDetails>("LLM.HttpRequestDetails")({

packages/ai/test/executor.test.ts

@@ -85,14 +85,17 @@ describe("RequestExecutor", () => {
     ),
   )
 
-  it.effect("does not classify generic HTTP 413 payload errors as context overflow", () =>
+  it.effect("classifies generic HTTP 413 payload errors", () =>
     Effect.gen(function* () {
       const executor = yield* RequestExecutor.Service
       const error = yield* executor.execute(request).pipe(Effect.flip)
 
       expectLLMError(error)
-      expect(error.reason).toMatchObject({ _tag: "InvalidRequest" })
-      expect("classification" in error.reason ? error.reason.classification : undefined).toBeUndefined()
+      expect(error.reason).toMatchObject({
+        _tag: "InvalidRequest",
+        classification: "payload-too-large",
+        http: { response: { status: 413 } },
+      })
     }).pipe(Effect.provide(responsesLayer([new Response("request too large", { status: 413 })]))),
   )
 

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

@@ -6,7 +6,6 @@ describe("provider error classification", () => {
   test("classifies provider token limit messages as context overflow", () => {
     const messages = [
       "tokens in request more than max tokens allowed",
-      '{"error":{"type":"request_too_large","message":"Request exceeds the maximum size"}}',
       "Requested token count exceeds the model's maximum context length of 131072 tokens.",
       "Input length (265330) exceeds model's maximum context length (262144).",
       "Input length 131393 exceeds the maximum allowed input length of 131040 tokens.",
@@ -19,6 +18,24 @@ describe("provider error classification", () => {
     expect(messages.every(isContextOverflow)).toBe(true)
   })
 
+  test("classifies request size failures separately from context overflow", () => {
+    const failures = [
+      classifyProviderFailure({ message: "request too large", status: 413 }),
+      classifyProviderFailure({
+        message: '{"error":{"type":"request_too_large","message":"Request exceeds the maximum size"}}',
+        status: 400,
+      }),
+      classifyProviderFailure({ message: "upstream request entity too large", status: 502 }),
+    ]
+
+    expect(failures).toEqual(
+      failures.map((failure) =>
+        expect.objectContaining({ _tag: "InvalidRequest", classification: "payload-too-large" }),
+      ),
+    )
+    expect(isContextOverflow("413 status code (no body)")).toBe(false)
+  })
+
   test("does not classify rate limits as context overflow", () => {
     const messages = [
       "Throttling error: Too many tokens, please wait before trying again.",
@@ -59,9 +76,10 @@ describe("provider error classification", () => {
   })
 
   test("classifies transient client statuses as provider internal", () => {
-    expect(
-      [408, 409].map((status) => classifyProviderFailure({ message: `HTTP ${status}`, status })._tag),
-    ).toEqual(["ProviderInternal", "ProviderInternal"])
+    expect([408, 409].map((status) => classifyProviderFailure({ message: `HTTP ${status}`, status })._tag)).toEqual([
+      "ProviderInternal",
+      "ProviderInternal",
+    ])
   })
 
   test("classifies nested provider codes when a top-level code is also present", () => {