docs(ai): describe quota codes provider-neutrally

packages/ai/src/provider-error.ts

@@ -59,7 +59,7 @@ export const isContextOverflowFailure = (failure: unknown) =>
     : Schema.is(ProviderErrorEvent)(failure) && failure.classification === "context-overflow"
 
 const decodeJson = Schema.decodeUnknownOption(Schema.fromJsonString(Schema.Unknown))
-// OpenCode Zen reports account caps as typed 429/402 errors that are not throttles.
+// Account caps that providers and gateways report as typed 429/402 errors, not throttles.
 const QUOTA_CODES = new Set([
   "insufficient_quota",
   "usage_not_included",
@@ -67,6 +67,8 @@ const QUOTA_CODES = new Set([
   "gousagelimiterror",
   "freeusagelimiterror",
   "creditlimitexceeded",
+  "organization_spend_limit_exceeded",
+  "project_spend_limit_exceeded",
 ])
 const AUTH_CODES = new Set(["authentication_error", "permission_error"])
 const SERVER_CODES = new Set([

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

@@ -278,6 +278,19 @@ describe("provider error classification", () => {
     ).toEqual(Array(6).fill("QuotaExceeded"))
   })
 
+  test("classifies OpenAI hard spend limits as quota rather than throttling", () => {
+    expect(
+      ["organization_spend_limit_exceeded", "project_spend_limit_exceeded"].map(
+        (code) =>
+          classifyProviderFailure({
+            message: "Your project has reached its configured enforced spend limit.",
+            status: 429,
+            rawBody: JSON.stringify({ error: { code, message: "Spend limit reached." } }),
+          })._tag,
+      ),
+    ).toEqual(["QuotaExceeded", "QuotaExceeded"])
+  })
+
   test("does not let substituted server codes make a 4xx retryable", () => {
     const openai = { error: { type: "server_error", message: "Upstream request failed: Model is unavailable." } }
     const anthropic = {