1 files changed, 56 insertions, 2 deletions
diff --git a/packages/core/src/utils/retry.ts b/packages/core/src/utils/retry.ts
index f3f5f2d2..01651950 100644
--- a/packages/core/src/utils/retry.ts
+++ b/packages/core/src/utils/retry.ts
@@ -5,13 +5,20 @@
  */
 
 import { AuthType } from '../core/contentGenerator.js';
+import {
+  isProQuotaExceededError,
+  isGenericQuotaExceededError,
+} from './quotaErrorDetection.js';
 
 export interface RetryOptions {
   maxAttempts: number;
   initialDelayMs: number;
   maxDelayMs: number;
   shouldRetry: (error: Error) => boolean;
-  onPersistent429?: (authType?: string) => Promise<string | null>;
+  onPersistent429?: (
+    authType?: string,
+    error?: unknown,
+  ) => Promise<string | null>;
   authType?: string;
 }
 
@@ -86,6 +93,53 @@ export async function retryWithBackoff<T>(
     } catch (error) {
       const errorStatus = getErrorStatus(error);
 
+      // Check for Pro quota exceeded error first - immediate fallback for OAuth users
+      if (
+        errorStatus === 429 &&
+        authType === AuthType.LOGIN_WITH_GOOGLE &&
+        isProQuotaExceededError(error) &&
+        onPersistent429
+      ) {
+        try {
+          const fallbackModel = await onPersistent429(authType, error);
+          if (fallbackModel) {
+            // Reset attempt counter and try with new model
+            attempt = 0;
+            consecutive429Count = 0;
+            currentDelay = initialDelayMs;
+            // With the model updated, we continue to the next attempt
+            continue;
+          }
+        } catch (fallbackError) {
+          // If fallback fails, continue with original error
+          console.warn('Fallback to Flash model failed:', fallbackError);
+        }
+      }
+
+      // Check for generic quota exceeded error (but not Pro, which was handled above) - immediate fallback for OAuth users
+      if (
+        errorStatus === 429 &&
+        authType === AuthType.LOGIN_WITH_GOOGLE &&
+        !isProQuotaExceededError(error) &&
+        isGenericQuotaExceededError(error) &&
+        onPersistent429
+      ) {
+        try {
+          const fallbackModel = await onPersistent429(authType, error);
+          if (fallbackModel) {
+            // Reset attempt counter and try with new model
+            attempt = 0;
+            consecutive429Count = 0;
+            currentDelay = initialDelayMs;
+            // With the model updated, we continue to the next attempt
+            continue;
+          }
+        } catch (fallbackError) {
+          // If fallback fails, continue with original error
+          console.warn('Fallback to Flash model failed:', fallbackError);
+        }
+      }
+
       // Track consecutive 429 errors
       if (errorStatus === 429) {
         consecutive429Count++;
@@ -100,7 +154,7 @@ export async function retryWithBackoff<T>(
         authType === AuthType.LOGIN_WITH_GOOGLE
       ) {
         try {
-          const fallbackModel = await onPersistent429(authType);
+          const fallbackModel = await onPersistent429(authType, error);
           if (fallbackModel) {
             // Reset attempt counter and try with new model
             attempt = 0;