@machina.ai/cell-cli-core 1.16.0-rc2 → 1.19.4-rc1

package/dist/src/core/geminiChat.test.js

@@ -4,10 +4,10 @@
  * SPDX-License-Identifier: Apache-2.0
  */
 import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
-import { ApiError } from '@google/genai';
+import { ApiError, ThinkingLevel } from '@google/genai';
 import { GeminiChat, InvalidStreamError, StreamEventType, SYNTHETIC_THOUGHT_SIGNATURE, } from './geminiChat.js';
 import { setSimulate429 } from '../utils/testUtils.js';
-import { DEFAULT_GEMINI_FLASH_MODEL, DEFAULT_GEMINI_MODEL, PREVIEW_GEMINI_MODEL, } from '../config/models.js';
+import { DEFAULT_GEMINI_FLASH_MODEL, DEFAULT_GEMINI_MODEL, DEFAULT_THINKING_MODE, PREVIEW_GEMINI_MODEL, } from '../config/models.js';
 import { AuthType } from './contentGenerator.js';
 import { TerminalQuotaError } from '../utils/googleQuotaErrors.js';
 import { retryWithBackoff } from '../utils/retry.js';
@@ -65,7 +65,6 @@ describe('GeminiChat', () => {
     let mockContentGenerator;
     let chat;
     let mockConfig;
-    const config = {};
     beforeEach(() => {
         vi.clearAllMocks();
         vi.mocked(uiTelemetryService.setLastPromptTokenCount).mockClear();
@@ -104,6 +103,24 @@ describe('GeminiChat', () => {
             }),
             getContentGenerator: vi.fn().mockReturnValue(mockContentGenerator),
             getRetryFetchErrors: vi.fn().mockReturnValue(false),
+            modelConfigService: {
+                getResolvedConfig: vi.fn().mockImplementation((modelConfigKey) => {
+                    const thinkingConfig = modelConfigKey.model.startsWith('gemini-3')
+                        ? {
+                            thinkingLevel: ThinkingLevel.HIGH,
+                        }
+                        : {
+                            thinkingBudget: DEFAULT_THINKING_MODE,
+                        };
+                    return {
+                        model: modelConfigKey.model,
+                        generateContentConfig: {
+                            temperature: 0,
+                            thinkingConfig,
+                        },
+                    };
+                }),
+            },
             isPreviewModelBypassMode: vi.fn().mockReturnValue(false),
             setPreviewModelBypassMode: vi.fn(),
             isPreviewModelFallbackMode: vi.fn().mockReturnValue(false),
@@ -113,7 +130,7 @@ describe('GeminiChat', () => {
         // Disable 429 simulation for tests
         setSimulate429(false);
         // Reset history for each test by creating a new instance
-        chat = new GeminiChat(mockConfig, config, []);
+        chat = new GeminiChat(mockConfig);
     });
     afterEach(() => {
         vi.restoreAllMocks();
@@ -125,12 +142,12 @@ describe('GeminiChat', () => {
                 { role: 'user', parts: [{ text: 'Hello' }] },
                 { role: 'model', parts: [{ text: 'Hi there' }] },
             ];
-            const chatWithHistory = new GeminiChat(mockConfig, config, history);
+            const chatWithHistory = new GeminiChat(mockConfig, '', [], history);
             const estimatedTokens = Math.ceil(JSON.stringify(history).length / 4);
             expect(chatWithHistory.getLastPromptTokenCount()).toBe(estimatedTokens);
         });
         it('should initialize lastPromptTokenCount for empty history', () => {
-            const chatEmpty = new GeminiChat(mockConfig, config, []);
+            const chatEmpty = new GeminiChat(mockConfig);
             expect(chatEmpty.getLastPromptTokenCount()).toBe(Math.ceil(JSON.stringify([]).length / 4));
         });
     });
@@ -163,7 +180,7 @@ describe('GeminiChat', () => {
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(streamWithToolCall);
             // 2. Action & Assert: The stream processing should complete without throwing an error
             // because the presence of a tool call makes the empty final chunk acceptable.
-            const stream = await chat.sendMessageStream('test-model', { message: 'test message' }, 'prompt-id-tool-call-empty-end');
+            const stream = await chat.sendMessageStream({ model: 'test-model' }, 'test message', 'prompt-id-tool-call-empty-end', new AbortController().signal);
             await expect((async () => {
                 for await (const _ of stream) {
                     /* consume stream */
@@ -203,7 +220,7 @@ describe('GeminiChat', () => {
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(streamWithNoFinish);
             // 2. Action & Assert: The stream should fail because there's no finish reason.
-            const stream = await chat.sendMessageStream('gemini-2.0-flash', { message: 'test message' }, 'prompt-id-no-finish-empty-end');
+            const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'test message', 'prompt-id-no-finish-empty-end', new AbortController().signal);
             await expect((async () => {
                 for await (const _ of stream) {
                     /* consume stream */
@@ -238,7 +255,7 @@ describe('GeminiChat', () => {
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(streamWithInvalidEnd);
             // 2. Action & Assert: The stream should complete without throwing an error.
-            const stream = await chat.sendMessageStream('test-model', { message: 'test message' }, 'prompt-id-valid-then-invalid-end');
+            const stream = await chat.sendMessageStream({ model: 'test-model' }, 'test message', 'prompt-id-valid-then-invalid-end', new AbortController().signal);
             await expect((async () => {
                 for await (const _ of stream) {
                     /* consume stream */
@@ -273,7 +290,7 @@ describe('GeminiChat', () => {
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(multiChunkStream);
             // 2. Action: Send a message and consume the stream.
-            const stream = await chat.sendMessageStream('test-model', { message: 'test message' }, 'prompt-id-empty-chunk-consolidation');
+            const stream = await chat.sendMessageStream({ model: 'test-model' }, 'test message', 'prompt-id-empty-chunk-consolidation', new AbortController().signal);
             for await (const _ of stream) {
                 // Consume the stream
             }
@@ -321,7 +338,7 @@ describe('GeminiChat', () => {
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(multiChunkStream);
             // 2. Action: Send a message and consume the stream.
-            const stream = await chat.sendMessageStream('test-model', { message: 'test message' }, 'prompt-id-multi-chunk');
+            const stream = await chat.sendMessageStream({ model: 'test-model' }, 'test message', 'prompt-id-multi-chunk', new AbortController().signal);
             for await (const _ of stream) {
                 // Consume the stream to trigger history recording.
             }
@@ -357,7 +374,7 @@ describe('GeminiChat', () => {
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(mixedContentStream);
             // 2. Action: Send a message and fully consume the stream to trigger history recording.
-            const stream = await chat.sendMessageStream('test-model', { message: 'test message' }, 'prompt-id-mixed-chunk');
+            const stream = await chat.sendMessageStream({ model: 'test-model' }, 'test message', 'prompt-id-mixed-chunk', new AbortController().signal);
             for await (const _ of stream) {
                 // This loop consumes the stream.
             }
@@ -385,7 +402,7 @@ describe('GeminiChat', () => {
                     ],
                 };
             })());
-            const stream = await chat.sendMessageStream(PREVIEW_GEMINI_MODEL, { message: 'test' }, 'prompt-id-fast-retry');
+            const stream = await chat.sendMessageStream({ model: PREVIEW_GEMINI_MODEL }, 'test', 'prompt-id-fast-retry', new AbortController().signal);
             for await (const _ of stream) {
                 // consume stream
             }
@@ -405,7 +422,7 @@ describe('GeminiChat', () => {
                     ],
                 };
             })());
-            const stream = await chat.sendMessageStream(DEFAULT_GEMINI_FLASH_MODEL, { message: 'test' }, 'prompt-id-normal-retry');
+            const stream = await chat.sendMessageStream({ model: DEFAULT_GEMINI_FLASH_MODEL }, 'test', 'prompt-id-normal-retry', new AbortController().signal);
             for await (const _ of stream) {
                 // consume stream
             }
@@ -437,7 +454,7 @@ describe('GeminiChat', () => {
             }));
             // ACT
             const consumeStream = async () => {
-                const stream = await chat.sendMessageStream(PREVIEW_GEMINI_MODEL, { message: 'test' }, 'prompt-id-bypass');
+                const stream = await chat.sendMessageStream({ model: PREVIEW_GEMINI_MODEL }, 'test', 'prompt-id-bypass', new AbortController().signal);
                 // Consume the stream to trigger execution
                 for await (const _ of stream) {
                     // do nothing
@@ -484,14 +501,12 @@ describe('GeminiChat', () => {
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(emptyStreamResponse);
             // 3. Action: Send the function response back to the model and consume the stream.
-            const stream = await chat.sendMessageStream('gemini-2.0-flash', {
-                message: {
-                    functionResponse: {
-                        name: 'find_restaurant',
-                        response: { name: 'Vesuvio' },
-                    },
+            const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, {
+                functionResponse: {
+                    name: 'find_restaurant',
+                    response: { name: 'Vesuvio' },
                 },
-            }, 'prompt-id-stream-1');
+            }, 'prompt-id-stream-1', new AbortController().signal);
             // 4. Assert: The stream processing should throw an InvalidStreamError.
             await expect((async () => {
                 for await (const _ of stream) {
@@ -522,7 +537,7 @@ describe('GeminiChat', () => {
                 };
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(streamWithToolCall);
-            const stream = await chat.sendMessageStream('test-model', { message: 'test' }, 'prompt-id-1');
+            const stream = await chat.sendMessageStream({ model: 'test-model' }, 'test message', 'prompt-id-1', new AbortController().signal);
             // Should not throw an error
             await expect((async () => {
                 for await (const _ of stream) {
@@ -546,7 +561,7 @@ describe('GeminiChat', () => {
                 };
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(streamWithoutFinishReason);
-            const stream = await chat.sendMessageStream('gemini-2.0-flash', { message: 'test' }, 'prompt-id-1');
+            const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'test message', 'prompt-id-1', new AbortController().signal);
             await expect((async () => {
                 for await (const _ of stream) {
                     // consume stream
@@ -569,7 +584,7 @@ describe('GeminiChat', () => {
                 };
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(streamWithEmptyResponse);
-            const stream = await chat.sendMessageStream('gemini-2.0-flash', { message: 'test' }, 'prompt-id-1');
+            const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'test message', 'prompt-id-1', new AbortController().signal);
             await expect((async () => {
                 for await (const _ of stream) {
                     // consume stream
@@ -592,7 +607,7 @@ describe('GeminiChat', () => {
                 };
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(validStream);
-            const stream = await chat.sendMessageStream('test-model', { message: 'test' }, 'prompt-id-1');
+            const stream = await chat.sendMessageStream({ model: 'test-model' }, 'test message', 'prompt-id-1', new AbortController().signal);
             // Should not throw an error
             await expect((async () => {
                 for await (const _ of stream) {
@@ -616,7 +631,7 @@ describe('GeminiChat', () => {
                 };
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(streamWithMalformedFunctionCall);
-            const stream = await chat.sendMessageStream('gemini-2.5-pro', { message: 'test' }, 'prompt-id-malformed');
+            const stream = await chat.sendMessageStream({ model: 'gemini-2.5-pro' }, 'test', 'prompt-id-malformed', new AbortController().signal);
             // Should throw an error
             await expect((async () => {
                 for await (const _ of stream) {
@@ -650,7 +665,7 @@ describe('GeminiChat', () => {
                 };
             })());
             // 2. Send a message
-            const stream = await chat.sendMessageStream('gemini-2.5-pro', { message: 'test retry' }, 'prompt-id-retry-malformed');
+            const stream = await chat.sendMessageStream({ model: 'gemini-2.5-pro' }, 'test retry', 'prompt-id-retry-malformed', new AbortController().signal);
             const events = [];
             for await (const event of stream) {
                 events.push(event);
@@ -688,7 +703,7 @@ describe('GeminiChat', () => {
                 };
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(response);
-            const stream = await chat.sendMessageStream('test-model', { message: 'hello' }, 'prompt-id-1');
+            const stream = await chat.sendMessageStream({ model: 'test-model' }, 'hello', 'prompt-id-1', new AbortController().signal);
             for await (const _ of stream) {
                 // consume stream
             }
@@ -700,9 +715,69 @@ describe('GeminiChat', () => {
                         parts: [{ text: 'hello' }],
                     },
                 ],
-                config: {},
+                config: {
+                    systemInstruction: '',
+                    tools: [],
+                    temperature: 0,
+                    thinkingConfig: {
+                        thinkingBudget: DEFAULT_THINKING_MODE,
+                    },
+                    abortSignal: expect.any(AbortSignal),
+                },
             }, 'prompt-id-1');
         });
+        it('should use thinkingLevel and remove thinkingBudget for gemini-3 models', async () => {
+            const response = (async function* () {
+                yield {
+                    candidates: [
+                        {
+                            content: { parts: [{ text: 'response' }], role: 'model' },
+                            finishReason: 'STOP',
+                        },
+                    ],
+                };
+            })();
+            vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(response);
+            const stream = await chat.sendMessageStream({ model: 'gemini-3-test-only-model-string-for-testing' }, 'hello', 'prompt-id-thinking-level', new AbortController().signal);
+            for await (const _ of stream) {
+                // consume stream
+            }
+            expect(mockContentGenerator.generateContentStream).toHaveBeenCalledWith(expect.objectContaining({
+                model: 'gemini-3-test-only-model-string-for-testing',
+                config: expect.objectContaining({
+                    thinkingConfig: {
+                        thinkingBudget: undefined,
+                        thinkingLevel: ThinkingLevel.HIGH,
+                    },
+                }),
+            }), 'prompt-id-thinking-level');
+        });
+        it('should use thinkingBudget and remove thinkingLevel for non-gemini-3 models', async () => {
+            const response = (async function* () {
+                yield {
+                    candidates: [
+                        {
+                            content: { parts: [{ text: 'response' }], role: 'model' },
+                            finishReason: 'STOP',
+                        },
+                    ],
+                };
+            })();
+            vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(response);
+            const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'hello', 'prompt-id-thinking-budget', new AbortController().signal);
+            for await (const _ of stream) {
+                // consume stream
+            }
+            expect(mockContentGenerator.generateContentStream).toHaveBeenCalledWith(expect.objectContaining({
+                model: 'gemini-2.0-flash',
+                config: expect.objectContaining({
+                    thinkingConfig: {
+                        thinkingBudget: DEFAULT_THINKING_MODE,
+                        thinkingLevel: undefined,
+                    },
+                }),
+            }), 'prompt-id-thinking-budget');
+        });
     });
     describe('addHistory', () => {
         it('should add a new content item to the history', () => {
@@ -740,7 +815,7 @@ describe('GeminiChat', () => {
                     candidates: [{ content: { parts: [{ text: '' }] } }],
                 };
             })());
-            const stream = await chat.sendMessageStream('gemini-1.5-pro', { message: 'test' }, 'prompt-id-no-retry');
+            const stream = await chat.sendMessageStream({ model: 'gemini-1.5-pro' }, 'test', 'prompt-id-no-retry', new AbortController().signal);
             await expect((async () => {
                 for await (const _ of stream) {
                     // Must loop to trigger the internal logic that throws.
@@ -773,7 +848,7 @@ describe('GeminiChat', () => {
                 };
             })());
             // ACT: Send a message and collect all events from the stream.
-            const stream = await chat.sendMessageStream('gemini-2.0-flash', { message: 'test' }, 'prompt-id-yield-retry');
+            const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'test message', 'prompt-id-yield-retry', new AbortController().signal);
             const events = [];
             for await (const event of stream) {
                 events.push(event);
@@ -805,7 +880,7 @@ describe('GeminiChat', () => {
                     ],
                 };
             })());
-            const stream = await chat.sendMessageStream('gemini-2.0-flash', { message: 'test' }, 'prompt-id-retry-success');
+            const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'test', 'prompt-id-retry-success', new AbortController().signal);
             const chunks = [];
             for await (const chunk of stream) {
                 chunks.push(chunk);
@@ -856,7 +931,7 @@ describe('GeminiChat', () => {
                     ],
                 };
             })());
-            const stream = await chat.sendMessageStream('gemini-2.0-flash', { message: 'test', config: { temperature: 0.5 } }, 'prompt-id-retry-temperature');
+            const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'test message', 'prompt-id-retry-temperature', new AbortController().signal);
             for await (const _ of stream) {
                 // consume stream
             }
@@ -864,7 +939,7 @@ describe('GeminiChat', () => {
             // First call should have original temperature
             expect(mockContentGenerator.generateContentStream).toHaveBeenNthCalledWith(1, expect.objectContaining({
                 config: expect.objectContaining({
-                    temperature: 0.5,
+                    temperature: 0,
                 }),
             }), 'prompt-id-retry-temperature');
             // Second call (retry) should have temperature 1
@@ -887,7 +962,7 @@ describe('GeminiChat', () => {
                     ],
                 };
             })());
-            const stream = await chat.sendMessageStream('gemini-2.0-flash', { message: 'test' }, 'prompt-id-retry-fail');
+            const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'test', 'prompt-id-retry-fail', new AbortController().signal);
             await expect(async () => {
                 for await (const _ of stream) {
                     // Must loop to trigger the internal logic that throws.
@@ -936,7 +1011,7 @@ describe('GeminiChat', () => {
             it('should not retry on 400 Bad Request errors', async () => {
                 const error400 = new ApiError({ message: 'Bad Request', status: 400 });
                 vi.mocked(mockContentGenerator.generateContentStream).mockRejectedValue(error400);
-                const stream = await chat.sendMessageStream('gemini-2.0-flash', { message: 'test' }, 'prompt-id-400');
+                const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'test message', 'prompt-id-400', new AbortController().signal);
                 await expect((async () => {
                     for await (const _ of stream) {
                         /* consume stream */
@@ -959,7 +1034,7 @@ describe('GeminiChat', () => {
                         ],
                     };
                 })());
-                const stream = await chat.sendMessageStream('test-model', { message: 'test' }, 'prompt-id-429-retry');
+                const stream = await chat.sendMessageStream({ model: 'test-model' }, 'test message', 'prompt-id-429-retry', new AbortController().signal);
                 const events = [];
                 for await (const event of stream) {
                     events.push(event);
@@ -988,7 +1063,7 @@ describe('GeminiChat', () => {
                         ],
                     };
                 })());
-                const stream = await chat.sendMessageStream('test-model', { message: 'test' }, 'prompt-id-500-retry');
+                const stream = await chat.sendMessageStream({ model: 'test-model' }, 'test message', 'prompt-id-500-retry', new AbortController().signal);
                 const events = [];
                 for await (const event of stream) {
                     events.push(event);
@@ -1024,7 +1099,7 @@ describe('GeminiChat', () => {
                         throw error;
                     }
                 });
-                const stream = await chat.sendMessageStream('test-model', { message: 'test' }, 'prompt-id-fetch-error-retry');
+                const stream = await chat.sendMessageStream({ model: 'test-model' }, 'test message', 'prompt-id-fetch-error-retry', new AbortController().signal);
                 const events = [];
                 for await (const event of stream) {
                     events.push(event);
@@ -1067,7 +1142,7 @@ describe('GeminiChat', () => {
             };
         })());
         // 3. Send a new message
-        const stream = await chat.sendMessageStream('gemini-2.0-flash', { message: 'Second question' }, 'prompt-id-retry-existing');
+        const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'Second question', 'prompt-id-retry-existing', new AbortController().signal);
         for await (const _ of stream) {
             // consume stream
         }
@@ -1119,7 +1194,7 @@ describe('GeminiChat', () => {
             };
         })());
         // 2. Call the method and consume the stream.
-        const stream = await chat.sendMessageStream('gemini-2.0-flash', { message: 'test empty stream' }, 'prompt-id-empty-stream');
+        const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'test empty stream', 'prompt-id-empty-stream', new AbortController().signal);
         const chunks = [];
         for await (const chunk of stream) {
             chunks.push(chunk);
@@ -1180,11 +1255,11 @@ describe('GeminiChat', () => {
             .mockResolvedValueOnce(firstStreamGenerator)
             .mockResolvedValueOnce(secondStreamGenerator);
         // 3. Start the first stream and consume only the first chunk to pause it
-        const firstStream = await chat.sendMessageStream('test-model', { message: 'first' }, 'prompt-1');
+        const firstStream = await chat.sendMessageStream({ model: 'test-model' }, 'first', 'prompt-1', new AbortController().signal);
         const firstStreamIterator = firstStream[Symbol.asyncIterator]();
         await firstStreamIterator.next();
         // 4. While the first stream is paused, start the second call. It will block.
-        const secondStreamPromise = chat.sendMessageStream('test-model', { message: 'second' }, 'prompt-2');
+        const secondStreamPromise = chat.sendMessageStream({ model: 'test-model' }, 'second', 'prompt-2', new AbortController().signal);
         // 5. Assert that only one API call has been made so far.
         expect(mockContentGenerator.generateContentStream).toHaveBeenCalledTimes(1);
         // 6. Unblock and fully consume the first stream to completion.
@@ -1224,7 +1299,7 @@ describe('GeminiChat', () => {
             vi.mocked(mockContentGenerator.generateContentStream).mockImplementation(async () => (async function* () {
                 yield mockResponse;
             })());
-            const stream = await chat.sendMessageStream('test-model', { message: 'test' }, 'prompt-id-res3');
+            const stream = await chat.sendMessageStream({ model: 'test-model' }, 'test message', 'prompt-id-res3', new AbortController().signal);
             for await (const _ of stream) {
                 // consume stream
             }
@@ -1287,7 +1362,7 @@ describe('GeminiChat', () => {
                 isInFallbackModeSpy.mockReturnValue(true);
                 return true; // Signal retry
             });
-            const stream = await chat.sendMessageStream('test-model', { message: 'trigger 429' }, 'prompt-id-fb1');
+            const stream = await chat.sendMessageStream({ model: 'test-model' }, 'trigger 429', 'prompt-id-fb1', new AbortController().signal);
             // Consume stream to trigger logic
             for await (const _ of stream) {
                 // no-op
@@ -1299,11 +1374,69 @@ describe('GeminiChat', () => {
             const modelTurn = history[1];
             expect(modelTurn.parts[0].text).toBe('Success on retry');
         });
+        it('should switch to DEFAULT_GEMINI_FLASH_MODEL and use thinkingBudget when falling back from a gemini-3 model', async () => {
+            // ARRANGE
+            const authType = AuthType.LOGIN_WITH_GOOGLE;
+            vi.mocked(mockConfig.getContentGeneratorConfig).mockReturnValue({
+                authType,
+            });
+            // Initial state: Not in fallback mode
+            const isInFallbackModeSpy = vi.spyOn(mockConfig, 'isInFallbackMode');
+            isInFallbackModeSpy.mockReturnValue(false);
+            // Mock API calls:
+            // 1. Fails with 429 (simulating gemini-3 failure)
+            // 2. Succeeds (simulating fallback success)
+            vi.mocked(mockContentGenerator.generateContentStream)
+                .mockRejectedValueOnce(error429)
+                .mockResolvedValueOnce((async function* () {
+                yield {
+                    candidates: [
+                        {
+                            content: { parts: [{ text: 'Fallback success' }] },
+                            finishReason: 'STOP',
+                        },
+                    ],
+                };
+            })());
+            // Mock handleFallback to enable fallback mode and signal retry
+            mockHandleFallback.mockImplementation(async () => {
+                isInFallbackModeSpy.mockReturnValue(true); // Next call will see fallback mode = true
+                return true;
+            });
+            // ACT
+            const stream = await chat.sendMessageStream({ model: 'gemini-3-test-model' }, // Start with a gemini-3 model
+            'test fallback thinking', 'prompt-id-fb3', new AbortController().signal);
+            for await (const _ of stream) {
+                // consume stream
+            }
+            // ASSERT
+            expect(mockContentGenerator.generateContentStream).toHaveBeenCalledTimes(2);
+            // First call: gemini-3 model, thinkingLevel set
+            expect(mockContentGenerator.generateContentStream).toHaveBeenNthCalledWith(1, expect.objectContaining({
+                model: 'gemini-3-test-model',
+                config: expect.objectContaining({
+                    thinkingConfig: {
+                        thinkingBudget: undefined,
+                        thinkingLevel: ThinkingLevel.HIGH,
+                    },
+                }),
+            }), 'prompt-id-fb3');
+            // Second call: DEFAULT_GEMINI_FLASH_MODEL (due to fallback), thinkingBudget set (due to fix)
+            expect(mockContentGenerator.generateContentStream).toHaveBeenNthCalledWith(2, expect.objectContaining({
+                model: DEFAULT_GEMINI_FLASH_MODEL,
+                config: expect.objectContaining({
+                    thinkingConfig: {
+                        thinkingBudget: DEFAULT_THINKING_MODE,
+                        thinkingLevel: undefined,
+                    },
+                }),
+            }), 'prompt-id-fb3');
+        });
         it('should stop retrying if handleFallback returns false (e.g., auth intent)', async () => {
             vi.mocked(mockConfig.getModel).mockReturnValue('gemini-pro');
             vi.mocked(mockContentGenerator.generateContentStream).mockRejectedValue(error429);
             mockHandleFallback.mockResolvedValue(false);
-            const stream = await chat.sendMessageStream('gemini-2.0-flash', { message: 'test stop' }, 'prompt-id-fb2');
+            const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'test stop', 'prompt-id-fb2', new AbortController().signal);
             await expect((async () => {
                 for await (const _ of stream) {
                     /* consume stream */
@@ -1347,7 +1480,7 @@ describe('GeminiChat', () => {
             };
         })());
         // Send a message and consume the stream
-        const stream = await chat.sendMessageStream('gemini-2.0-flash', { message: 'test' }, 'prompt-id-discard-test');
+        const stream = await chat.sendMessageStream({ model: 'gemini-2.0-flash' }, 'test message', 'prompt-id-discard-test', new AbortController().signal);
         const events = [];
         for await (const event of stream) {
             events.push(event);
@@ -1411,7 +1544,7 @@ describe('GeminiChat', () => {
                 };
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(stream);
-            await chat.sendMessageStream('test-model', { message: 'test' }, 'prompt-id-preview-model-reset');
+            await chat.sendMessageStream({ model: 'test-model' }, 'test', 'prompt-id-preview-model-reset', new AbortController().signal);
             expect(mockConfig.setPreviewModelBypassMode).toHaveBeenCalledWith(false);
         });
         it('should reset previewModelFallbackMode to false upon successful Preview Model usage', async () => {
@@ -1426,7 +1559,7 @@ describe('GeminiChat', () => {
                 };
             })();
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(stream);
-            const resultStream = await chat.sendMessageStream(PREVIEW_GEMINI_MODEL, { message: 'test' }, 'prompt-id-preview-model-healing');
+            const resultStream = await chat.sendMessageStream({ model: PREVIEW_GEMINI_MODEL }, 'test', 'prompt-id-preview-model-healing', new AbortController().signal);
             for await (const _ of resultStream) {
                 // consume stream
             }
@@ -1446,7 +1579,7 @@ describe('GeminiChat', () => {
             vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(stream);
             // Simulate bypass mode being active (downgrade happened)
             vi.mocked(mockConfig.isPreviewModelBypassMode).mockReturnValue(true);
-            const resultStream = await chat.sendMessageStream(PREVIEW_GEMINI_MODEL, { message: 'test' }, 'prompt-id-bypass-no-healing');
+            const resultStream = await chat.sendMessageStream({ model: PREVIEW_GEMINI_MODEL }, 'test', 'prompt-id-bypass-no-healing', new AbortController().signal);
             for await (const _ of resultStream) {
                 // consume stream
             }
@@ -1455,7 +1588,7 @@ describe('GeminiChat', () => {
     });
     describe('ensureActiveLoopHasThoughtSignatures', () => {
         it('should add thoughtSignature to the first functionCall in each model turn of the active loop', () => {
-            const chat = new GeminiChat(mockConfig, {}, []);
+            const chat = new GeminiChat(mockConfig, '', [], []);
             const history = [
                 { role: 'user', parts: [{ text: 'Old message' }] },
                 {
@@ -1504,7 +1637,7 @@ describe('GeminiChat', () => {
             expect(newContents[5]?.parts?.[1]).not.toHaveProperty('thoughtSignature');
         });
         it('should not modify contents if there is no user text message', () => {
-            const chat = new GeminiChat(mockConfig, {}, []);
+            const chat = new GeminiChat(mockConfig, '', [], []);
             const history = [
                 {
                     role: 'user',
@@ -1520,13 +1653,13 @@ describe('GeminiChat', () => {
             expect(newContents[1]?.parts?.[0]).not.toHaveProperty('thoughtSignature');
         });
         it('should handle an empty history', () => {
-            const chat = new GeminiChat(mockConfig, {}, []);
+            const chat = new GeminiChat(mockConfig, '', []);
             const history = [];
             const newContents = chat.ensureActiveLoopHasThoughtSignatures(history);
             expect(newContents).toEqual([]);
         });
         it('should handle history with only a user message', () => {
-            const chat = new GeminiChat(mockConfig, {}, []);
+            const chat = new GeminiChat(mockConfig, '', []);
             const history = [{ role: 'user', parts: [{ text: 'Hello' }] }];
             const newContents = chat.ensureActiveLoopHasThoughtSignatures(history);
             expect(newContents).toEqual(history);