Remove auto-execution on Flash in the event of a 429/Quota failover (#3662)

Co-authored-by: Jenna Inouye <[email protected]>

5 files changed, 93 insertions, 61 deletions

diff --git a/packages/cli/src/ui/App.tsx b/packages/cli/src/ui/App.tsx
index 833cc2b5..e3a5eb55 100644
--- a/packages/cli/src/ui/App.tsx
+++ b/packages/cli/src/ui/App.tsx
@@ -70,6 +70,7 @@ import { UpdateNotification } from './components/UpdateNotification.js';
 import {
   isProQuotaExceededError,
   isGenericQuotaExceededError,
+  UserTierId,
 } from '@google/gemini-cli-core';
 import { checkForUpdates } from './utils/updateCheck.js';
 import ansiEscapes from 'ansi-escapes';
@@ -136,6 +137,8 @@ const App = ({ config, settings, startupWarnings = [] }: AppProps) => {
   const ctrlDTimerRef = useRef<NodeJS.Timeout | null>(null);
   const [constrainHeight, setConstrainHeight] = useState<boolean>(true);
   const [showPrivacyNotice, setShowPrivacyNotice] = useState<boolean>(false);
+  const [modelSwitchedFromQuotaError, setModelSwitchedFromQuotaError] =
+    useState<boolean>(false);
 
   const openPrivacyNotice = useCallback(() => {
     setShowPrivacyNotice(true);
@@ -251,23 +254,51 @@ const App = ({ config, settings, startupWarnings = [] }: AppProps) => {
     ): Promise<boolean> => {
       let message: string;
 
+      // For quota errors, assume FREE tier (safe default) - only show upgrade messaging to free tier users
+      // TODO: Get actual user tier from config when available
+      const userTier = undefined; // Defaults to FREE tier behavior
+      const isPaidTier =
+        userTier === UserTierId.LEGACY || userTier === UserTierId.STANDARD;
+
       // Check if this is a Pro quota exceeded error
       if (error && isProQuotaExceededError(error)) {
-        message = `⚡ You have reached your daily ${currentModel} quota limit.
+        if (isPaidTier) {
+          message = `⚡ You have reached your daily ${currentModel} quota limit.
+⚡ Automatically switching from ${currentModel} to ${fallbackModel} for the remainder of this session.
+⚡ To continue accessing the ${currentModel} model today, consider using /auth to switch to using a paid API key from AI Studio at https://aistudio.google.com/apikey`;
+        } else {
+          message = `⚡ You have reached your daily ${currentModel} quota limit.
 ⚡ Automatically switching from ${currentModel} to ${fallbackModel} for the remainder of this session.
 ⚡ To increase your limits, upgrade to a Gemini Code Assist Standard or Enterprise plan with higher limits at https://goo.gle/set-up-gemini-code-assist
 ⚡ Or you can utilize a Gemini API Key. See: https://goo.gle/gemini-cli-docs-auth#gemini-api-key
 ⚡ You can switch authentication methods by typing /auth`;
+        }
       } else if (error && isGenericQuotaExceededError(error)) {
-        message = `⚡ You have reached your daily quota limit.
+        if (isPaidTier) {
+          message = `⚡ You have reached your daily quota limit.
+⚡ Automatically switching from ${currentModel} to ${fallbackModel} for the remainder of this session.
+⚡ To continue accessing the ${currentModel} model today, consider using /auth to switch to using a paid API key from AI Studio at https://aistudio.google.com/apikey`;
+        } else {
+          message = `⚡ You have reached your daily quota limit.
 ⚡ Automatically switching from ${currentModel} to ${fallbackModel} for the remainder of this session.
 ⚡ To increase your limits, upgrade to a Gemini Code Assist Standard or Enterprise plan with higher limits at https://goo.gle/set-up-gemini-code-assist
 ⚡ Or you can utilize a Gemini API Key. See: https://goo.gle/gemini-cli-docs-auth#gemini-api-key
 ⚡ You can switch authentication methods by typing /auth`;
+        }
       } else {
-        // Default fallback message for other cases (like consecutive 429s)
-        message = `⚡ Slow response times detected.
-⚡ Automatically switching from ${currentModel} to ${fallbackModel} for faster responses for the remainder of this session.`;
+        if (isPaidTier) {
+          // Default fallback message for other cases (like consecutive 429s)
+          message = `⚡ Automatically switching from ${currentModel} to ${fallbackModel} for faster responses for the remainder of this session.
+⚡ Possible reasons for this are that you have received multiple consecutive capacity errors or you have reached your daily ${currentModel} quota limit
+⚡ To continue accessing the ${currentModel} model today, consider using /auth to switch to using a paid API key from AI Studio at https://aistudio.google.com/apikey`;
+        } else {
+          // Default fallback message for other cases (like consecutive 429s)
+          message = `⚡ Automatically switching from ${currentModel} to ${fallbackModel} for faster responses for the remainder of this session.  
+⚡ Possible reasons for this are that you have received multiple consecutive capacity errors or you have reached your daily ${currentModel} quota limit
+⚡ To increase your limits, upgrade to a Gemini Code Assist Standard or Enterprise plan with higher limits at https://goo.gle/set-up-gemini-code-assist
+⚡ Or you can utilize a Gemini API Key. See: https://goo.gle/gemini-cli-docs-auth#gemini-api-key
+⚡ You can switch authentication methods by typing /auth`;
+        }
       }
 
       // Add message to UI history
@@ -278,7 +309,14 @@ const App = ({ config, settings, startupWarnings = [] }: AppProps) => {
         },
         Date.now(),
       );
-      return true; // Always accept the fallback
+
+      // Set the flag to prevent tool continuation
+      setModelSwitchedFromQuotaError(true);
+      // Set global quota error flag to prevent Flash model calls
+      config.setQuotaErrorOccurred(true);
+      // Switch model for future use but return false to stop current retry
+      config.setModel(fallbackModel);
+      return false; // Don't continue with current prompt
     };
 
     config.setFlashFallbackHandler(flashFallbackHandler);
@@ -445,6 +483,8 @@ const App = ({ config, settings, startupWarnings = [] }: AppProps) => {
     getPreferredEditor,
     onAuthError,
     performMemoryRefresh,
+    modelSwitchedFromQuotaError,
+    setModelSwitchedFromQuotaError,
   );
   pendingHistoryItems.push(...pendingGeminiHistoryItems);
   const { elapsedTime, currentLoadingPhrase } =
diff --git a/packages/cli/src/ui/hooks/useGeminiStream.test.tsx b/packages/cli/src/ui/hooks/useGeminiStream.test.tsx
index fc6f93c5..62ade50f 100644
--- a/packages/cli/src/ui/hooks/useGeminiStream.test.tsx
+++ b/packages/cli/src/ui/hooks/useGeminiStream.test.tsx
@@ -301,6 +301,8 @@ describe('useGeminiStream', () => {
       getUsageStatisticsEnabled: () => true,
       getDebugMode: () => false,
       addHistory: vi.fn(),
+      setQuotaErrorOccurred: vi.fn(),
+      getQuotaErrorOccurred: vi.fn(() => false),
     } as unknown as Config;
     mockOnDebugMessage = vi.fn();
     mockHandleSlashCommand = vi.fn().mockResolvedValue(false);
@@ -386,6 +388,8 @@ describe('useGeminiStream', () => {
           () => 'vscode' as EditorType,
           () => {},
           () => Promise.resolve(),
+          false,
+          () => {},
         );
       },
       {
@@ -518,6 +522,8 @@ describe('useGeminiStream', () => {
         () => 'vscode' as EditorType,
         () => {},
         () => Promise.resolve(),
+        false,
+        () => {},
       ),
     );
 
@@ -582,6 +588,8 @@ describe('useGeminiStream', () => {
         () => 'vscode' as EditorType,
         () => {},
         () => Promise.resolve(),
+        false,
+        () => {},
       ),
     );
 
@@ -675,6 +683,8 @@ describe('useGeminiStream', () => {
         () => 'vscode' as EditorType,
         () => {},
         () => Promise.resolve(),
+        false,
+        () => {},
       ),
     );
 
@@ -775,6 +785,8 @@ describe('useGeminiStream', () => {
         () => 'vscode' as EditorType,
         () => {},
         () => Promise.resolve(),
+        false,
+        () => {},
       ),
     );
 
@@ -1063,6 +1075,8 @@ describe('useGeminiStream', () => {
           () => 'vscode' as EditorType,
           () => {},
           mockPerformMemoryRefresh,
+          false,
+          () => {},
         ),
       );
 
@@ -1113,6 +1127,8 @@ describe('useGeminiStream', () => {
           () => 'vscode' as EditorType,
           () => {},
           () => Promise.resolve(),
+          false,
+          () => {},
         ),
       );
 
diff --git a/packages/cli/src/ui/hooks/useGeminiStream.ts b/packages/cli/src/ui/hooks/useGeminiStream.ts
index 550cab86..d32c9ffa 100644
--- a/packages/cli/src/ui/hooks/useGeminiStream.ts
+++ b/packages/cli/src/ui/hooks/useGeminiStream.ts
@@ -90,6 +90,8 @@ export const useGeminiStream = (
   getPreferredEditor: () => EditorType | undefined,
   onAuthError: () => void,
   performMemoryRefresh: () => Promise<void>,
+  modelSwitchedFromQuotaError: boolean,
+  setModelSwitchedFromQuotaError: React.Dispatch<React.SetStateAction<boolean>>,
 ) => {
   const [initError, setInitError] = useState<string | null>(null);
   const abortControllerRef = useRef<AbortController | null>(null);
@@ -494,6 +496,12 @@ export const useGeminiStream = (
       const userMessageTimestamp = Date.now();
       setShowHelp(false);
 
+      // Reset quota error flag when starting a new query (not a continuation)
+      if (!options?.isContinuation) {
+        setModelSwitchedFromQuotaError(false);
+        config.setQuotaErrorOccurred(false);
+      }
+
       abortControllerRef.current = new AbortController();
       const abortSignal = abortControllerRef.current.signal;
       turnCancelledRef.current = false;
@@ -552,6 +560,7 @@ export const useGeminiStream = (
     [
       streamingState,
       setShowHelp,
+      setModelSwitchedFromQuotaError,
       prepareQueryForGemini,
       processGeminiStreamEvents,
       pendingHistoryItemRef,
@@ -668,6 +677,12 @@ export const useGeminiStream = (
       );
 
       markToolsAsSubmitted(callIdsToMarkAsSubmitted);
+
+      // Don't continue if model was switched due to quota error
+      if (modelSwitchedFromQuotaError) {
+        return;
+      }
+
       submitQuery(mergePartListUnions(responsesToSend), {
         isContinuation: true,
       });
@@ -678,6 +693,7 @@ export const useGeminiStream = (
       markToolsAsSubmitted,
       geminiClient,
       performMemoryRefresh,
+      modelSwitchedFromQuotaError,
     ],
   );
 
diff --git a/packages/cli/src/ui/utils/errorParsing.test.ts b/packages/cli/src/ui/utils/errorParsing.test.ts
index 3d228efb..770dffad 100644
--- a/packages/cli/src/ui/utils/errorParsing.test.ts
+++ b/packages/cli/src/ui/utils/errorParsing.test.ts
@@ -39,7 +39,7 @@ describe('parseAndFormatApiError', () => {
     );
     expect(result).toContain('[API Error: Rate limit exceeded');
     expect(result).toContain(
-      'Slow response times detected. Switching to the gemini-2.5-flash model',
+      'Possible quota limitations in place or slow response times detected. Switching to the gemini-2.5-flash model',
     );
   });
 
@@ -55,7 +55,7 @@ describe('parseAndFormatApiError', () => {
     );
     expect(result).toContain('[API Error: Rate limit exceeded');
     expect(result).toContain(
-      'Slow response times detected. Switching to the gemini-2.5-flash model',
+      'Possible quota limitations in place or slow response times detected. Switching to the gemini-2.5-flash model',
     );
   });
 
@@ -169,7 +169,7 @@ describe('parseAndFormatApiError', () => {
     );
     expect(result).toContain('[API Error: Rate limit exceeded');
     expect(result).toContain(
-      'Slow response times detected. Switching to the gemini-2.5-flash model',
+      'Possible quota limitations in place or slow response times detected. Switching to the gemini-2.5-flash model',
     );
     expect(result).not.toContain(
       'You have reached your daily gemini-2.5-pro quota limit',
@@ -262,21 +262,17 @@ describe('parseAndFormatApiError', () => {
     );
   });
 
-  it('should handle different Gemini version strings in Pro quota exceeded errors', () => {
-    const errorMessage15 =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'Gemini 1.5 Pro Requests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
+  it('should handle different Gemini 2.5 version strings in Pro quota exceeded errors', () => {
+    const errorMessage25 =
+      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'Gemini 2.5 Pro Requests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
     const errorMessagePreview =
       'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'Gemini 2.5-preview Pro Requests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
-    const errorMessageBeta =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'Gemini beta-3.0 Pro Requests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
-    const errorMessageExperimental =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'Gemini experimental-v2 Pro Requests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
 
-    const result15 = parseAndFormatApiError(
-      errorMessage15,
+    const result25 = parseAndFormatApiError(
+      errorMessage25,
       AuthType.LOGIN_WITH_GOOGLE,
       undefined,
-      'gemini-1.5-pro',
+      'gemini-2.5-pro',
       DEFAULT_GEMINI_FLASH_MODEL,
     );
     const resultPreview = parseAndFormatApiError(
@@ -286,45 +282,19 @@ describe('parseAndFormatApiError', () => {
       'gemini-2.5-preview-pro',
       DEFAULT_GEMINI_FLASH_MODEL,
     );
-    const resultBeta = parseAndFormatApiError(
-      errorMessageBeta,
-      AuthType.LOGIN_WITH_GOOGLE,
-      undefined,
-      'gemini-beta-3.0-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
-    const resultExperimental = parseAndFormatApiError(
-      errorMessageExperimental,
-      AuthType.LOGIN_WITH_GOOGLE,
-      undefined,
-      'gemini-experimental-v2-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
 
-    expect(result15).toContain(
-      'You have reached your daily gemini-1.5-pro quota limit',
+    expect(result25).toContain(
+      'You have reached your daily gemini-2.5-pro quota limit',
     );
     expect(resultPreview).toContain(
       'You have reached your daily gemini-2.5-preview-pro quota limit',
     );
-    expect(resultBeta).toContain(
-      'You have reached your daily gemini-beta-3.0-pro quota limit',
-    );
-    expect(resultExperimental).toContain(
-      'You have reached your daily gemini-experimental-v2-pro quota limit',
-    );
-    expect(result15).toContain(
+    expect(result25).toContain(
       'upgrade to a Gemini Code Assist Standard or Enterprise plan',
     );
     expect(resultPreview).toContain(
       'upgrade to a Gemini Code Assist Standard or Enterprise plan',
     );
-    expect(resultBeta).toContain(
-      'upgrade to a Gemini Code Assist Standard or Enterprise plan',
-    );
-    expect(resultExperimental).toContain(
-      'upgrade to a Gemini Code Assist Standard or Enterprise plan',
-    );
   });
 
   it('should not match non-Pro models with similar version strings', () => {
@@ -339,16 +309,6 @@ describe('parseAndFormatApiError', () => {
         "Quota exceeded for quota metric 'Gemini 2.5-preview Flash Requests' and limit",
       ),
     ).toBe(false);
-    expect(
-      isProQuotaExceededError(
-        "Quota exceeded for quota metric 'Gemini beta-3.0 Flash Requests' and limit",
-      ),
-    ).toBe(false);
-    expect(
-      isProQuotaExceededError(
-        "Quota exceeded for quota metric 'Gemini experimental-v2 Flash Requests' and limit",
-      ),
-    ).toBe(false);
 
     // Test other model types
     expect(
diff --git a/packages/cli/src/ui/utils/errorParsing.ts b/packages/cli/src/ui/utils/errorParsing.ts
index 555d5e4e..5031bc0a 100644
--- a/packages/cli/src/ui/utils/errorParsing.ts
+++ b/packages/cli/src/ui/utils/errorParsing.ts
@@ -19,7 +19,7 @@ import {
 const getRateLimitErrorMessageGoogleFree = (
   fallbackModel: string = DEFAULT_GEMINI_FLASH_MODEL,
 ) =>
-  `\nSlow response times detected. Switching to the ${fallbackModel} model for the rest of this session.`;
+  `\nPossible quota limitations in place or slow response times detected. Switching to the ${fallbackModel} model for the rest of this session.`;
 
 const getRateLimitErrorMessageGoogleProQuotaFree = (
   currentModel: string = DEFAULT_GEMINI_MODEL,
@@ -34,7 +34,7 @@ const getRateLimitErrorMessageGoogleGenericQuotaFree = () =>
 const getRateLimitErrorMessageGooglePaid = (
   fallbackModel: string = DEFAULT_GEMINI_FLASH_MODEL,
 ) =>
-  `\nSlow response times detected. Switching to the ${fallbackModel} model for the rest of this session. We appreciate you for choosing Gemini Code Assist and the Gemini CLI.`;
+  `\nPossible quota limitations in place or slow response times detected. Switching to the ${fallbackModel} model for the rest of this session. We appreciate you for choosing Gemini Code Assist and the Gemini CLI.`;
 
 const getRateLimitErrorMessageGoogleProQuotaPaid = (
   currentModel: string = DEFAULT_GEMINI_MODEL,
@@ -53,7 +53,7 @@ const RATE_LIMIT_ERROR_MESSAGE_VERTEX =
 const getRateLimitErrorMessageDefault = (
   fallbackModel: string = DEFAULT_GEMINI_FLASH_MODEL,
 ) =>
-  `\nSlow response times detected. Switching to the ${fallbackModel} model for the rest of this session.`;
+  `\nPossible quota limitations in place or slow response times detected. Switching to the ${fallbackModel} model for the rest of this session.`;
 
 function getRateLimitMessage(
   authType?: AuthType,