fix(claude): classify native subscription quota 429 (#8628)

Co-authored-by: Escalada Online <aescaladaonline@gmail.com>
Co-authored-by: diegosouzapw <diegosouzapw@users.noreply.github.com>

open-sse/services/accountFallback.ts

@@ -1578,7 +1578,8 @@ export function checkFallbackError(
       const subResult = buildSubscriptionQuotaFallback(
         errorStr,
         getUpstreamRetryHintMs,
-        parseRetryFromErrorText
+        parseRetryFromErrorText,
+        provider
       );
       if (subResult) return subResult;
     }

open-sse/services/quotaTextCooldowns.ts

@@ -28,13 +28,18 @@ export interface QuotaTextFallback {
 // for the 5-hour subscription quota. Without a dedicated branch the request
 // falls through to the generic 429 retry path (~5s base cooldown).
 
-export function isSubscriptionQuotaText(lower: string): boolean {
+export function isSubscriptionQuotaText(lower: string, provider?: string | null): boolean {
   return (
     lower.includes("usage limit reached") ||
     lower.includes("usage limit has been") ||
     lower.includes("claude pro usage limit") ||
     lower.includes("you've reached your usage limit") ||
-    lower.includes("you have reached your usage limit")
+    lower.includes("you have reached your usage limit") ||
+    // Native Claude OAuth uses this otherwise-generic 429 wording for an
+    // exhausted subscription window. Keep it provider-scoped: other upstreams
+    // can use the same phrase for a short RPM throttle.
+    (provider === "claude" &&
+      lower.includes("this request would exceed your account's rate limit"))
   );
 }
 
@@ -55,9 +60,10 @@ const SUBSCRIPTION_QUOTA_COOLDOWN_MS = 60 * 60 * 1000; // 1 hour
 export function buildSubscriptionQuotaFallback(
   errorStr: string,
   getUpstreamRetryHintMs: () => number | null,
-  parseRetryFromErrorText: (text: string) => number | null
+  parseRetryFromErrorText: (text: string) => number | null,
+  provider?: string | null
 ): QuotaTextFallback | null {
-  if (!isSubscriptionQuotaText(errorStr.toLowerCase())) return null;
+  if (!isSubscriptionQuotaText(errorStr.toLowerCase(), provider)) return null;
   const hintMs = getUpstreamRetryHintMs();
   const bodyHint = parseRetryFromErrorText(errorStr);
   return {

src/sse/handlers/chat.ts

@@ -95,6 +95,7 @@ import { getComboFailureLogError } from "./comboFailureLogging";
 
 // Pipeline integration — wired modules
 import { classify429FromError, type FailureKind } from "@/shared/utils/classify429";
+import { isSubscriptionQuotaText } from "@omniroute/open-sse/services/quotaTextCooldowns.ts";
 import { resolveUseUpstream429BreakerHints } from "@/shared/utils/providerHints";
 import { getCircuitBreaker, isLocalStreamLifecycleError } from "../../shared/utils/circuitBreaker";
 import { markAccountExhaustedFrom429 } from "../../domain/quotaCache";
@@ -1725,7 +1726,9 @@ async function handleSingleModelChat(
       const errorStr = String(result.rawMessage ?? result.error ?? "");
       const failureKind =
         result.status === 429
-          ? classify429FromError({ status: result.status, message: errorStr })
+          ? isSubscriptionQuotaText(errorStr.toLowerCase(), provider)
+            ? "quota_exhausted"
+            : classify429FromError({ status: result.status, message: errorStr })
           : undefined;
       if (result.status === 429 && isDailyQuotaExhausted(errorStr)) {
         // Parse which model is quota-limited

tests/unit/account-fallback-anthropic-quota.test.ts

@@ -10,6 +10,8 @@ import assert from "node:assert/strict";
 
 const { classifyErrorText, parseRetryFromErrorText, checkFallbackError, getProviderProfile } =
   await import("../../open-sse/services/accountFallback.ts");
+const { isSubscriptionQuotaText } =
+  await import("../../open-sse/services/quotaTextCooldowns.ts");
 const { RateLimitReason } = await import("../../open-sse/config/constants.ts");
 
 test("#2321 classifyErrorText flags 'Usage Limit Reached' as QUOTA_EXHAUSTED", () => {
@@ -34,6 +36,12 @@ test("#2321 classifyErrorText still returns RATE_LIMIT_EXCEEDED for generic rate
   assert.equal(out, RateLimitReason.RATE_LIMIT_EXCEEDED);
 });
 
+test("Claude native account rate-limit text is subscription quota only for Claude", () => {
+  const text = "This request would exceed your account's rate limit. Please try again later.";
+  assert.equal(isSubscriptionQuotaText(text.toLowerCase(), "claude"), true);
+  assert.equal(isSubscriptionQuotaText(text.toLowerCase(), "antigravity"), false);
+});
+
 test("#2321 parseRetryFromErrorText extracts an ISO timestamp", () => {
   const future = new Date(Date.now() + 60 * 60 * 1000).toISOString();
   const ms = parseRetryFromErrorText(`Usage Limit Reached. Try again at ${future}`);
@@ -70,6 +78,18 @@ test("#2321 checkFallbackError returns ~1h cooldown for OAuth 429 + Usage Limit
   );
 });
 
+test("Claude native account rate-limit text gets subscription quota cooldown", () => {
+  const out = checkFallbackError(
+    429,
+    "This request would exceed your account's rate limit. Please try again later.",
+    0,
+    null,
+    "claude"
+  );
+  assert.equal(out.reason, RateLimitReason.QUOTA_EXHAUSTED);
+  assert.ok(out.cooldownMs >= 5 * 60 * 1000, `expected long cooldown, got ${out.cooldownMs}ms`);
+});
+
 test("#2321 checkFallbackError ignores ISO timestamp when upstream retry hints are disabled", () => {
   const future = new Date(Date.now() + 45 * 60 * 1000).toISOString();
   const out = checkFallbackError(