@machina.ai/cell-cli 1.56.0-rc1 → 1.59.0-rc1

package/dist/src/ui/hooks/useGeminiStream.test.js

@@ -31,6 +31,11 @@ const MockedGeminiClientClass = vi.hoisted(() => vi.fn().mockImplementation(func
     this.startChat = mockStartChat;
     this.sendMessageStream = mockSendMessageStream;
     this.addHistory = vi.fn();
+    let mockHistory = [];
+    this.getHistory = vi.fn().mockImplementation(() => mockHistory);
+    this.setHistory = vi.fn().mockImplementation((newHistory) => {
+        mockHistory = [...newHistory];
+    });
     this.generateContent = vi.fn().mockResolvedValue({
         candidates: [
             { content: { parts: [{ text: 'Got it. Focusing on tests only.' }] } },
@@ -554,6 +559,12 @@ describe('useGeminiStream', () => {
                 0,
             ];
         });
+        mockSendMessageStream.mockReturnValue((async function* () {
+            yield {
+                type: ServerGeminiEventType.Content,
+                value: 'Visible response text',
+            };
+        })());
         await renderHookWithProviders(() => useGeminiStream(new MockedGeminiClientClass(mockConfig), [], mockAddItem, mockConfig, mockLoadedSettings, mockOnDebugMessage, mockHandleSlashCommand, false, () => 'vscode', () => { }, () => Promise.resolve(), false, () => { }, () => { }, () => { }, 80, 24));
         // Trigger the onComplete callback with completed tools
         await act(async () => {
@@ -646,7 +657,73 @@ describe('useGeminiStream', () => {
             input: sentParts,
         });
     });
-    it('should handle all tool calls being cancelled', async () => {
+    it('should auto-nudge the model when tool execution succeeds but model stream is empty', async () => {
+        const toolCallResponseParts = [{ text: 'tool final response' }];
+        const completedToolCalls = [
+            {
+                request: {
+                    callId: 'call1',
+                    name: 'tool1',
+                    args: {},
+                    isClientInitiated: false,
+                    prompt_id: 'prompt-id-ack',
+                },
+                status: 'success',
+                responseSubmittedToGemini: false,
+                response: {
+                    callId: 'call1',
+                    responseParts: toolCallResponseParts,
+                    errorType: undefined,
+                },
+                tool: {
+                    displayName: 'MockTool',
+                },
+                invocation: {
+                    getDescription: () => `Mock description`,
+                },
+            },
+        ];
+        let callCount = 0;
+        mockSendMessageStream.mockImplementation(() => {
+            callCount += 1;
+            if (callCount === 1) {
+                return (async function* () { })();
+            }
+            else {
+                return (async function* () {
+                    yield {
+                        type: ServerGeminiEventType.Content,
+                        value: 'I have analyzed the empty response. Here is the final answer.',
+                    };
+                })();
+            }
+        });
+        let capturedOnComplete = null;
+        mockUseToolScheduler.mockImplementation((onComplete) => {
+            capturedOnComplete = onComplete;
+            return [
+                [],
+                mockScheduleToolCalls,
+                mockMarkToolsAsSubmitted,
+                vi.fn(),
+                mockCancelAllToolCalls,
+                0,
+            ];
+        });
+        await renderHookWithProviders(() => useGeminiStream(new MockedGeminiClientClass(mockConfig), [], mockAddItem, mockConfig, mockLoadedSettings, mockOnDebugMessage, mockHandleSlashCommand, false, () => 'vscode', () => { }, () => Promise.resolve(), false, () => { }, () => { }, () => { }, 80, 24, undefined, () => 'focus on tests only'));
+        await act(async () => {
+            if (capturedOnComplete) {
+                await new Promise((resolve) => setTimeout(resolve, 0));
+                await capturedOnComplete(completedToolCalls);
+            }
+        });
+        await waitFor(() => {
+            expect(mockSendMessageStream).toHaveBeenCalledTimes(2);
+        });
+        const sentParts = mockSendMessageStream.mock.calls[1][0];
+        expect(sentParts[0].text).toContain('[System: You successfully executed a tool but returned an empty response. Please analyze the tool output and explain your progress or final answer.]');
+    });
+    it('should handle all tool calls being cancelled by rolling back the history', async () => {
         const cancelledToolCalls = [
             {
                 request: {
@@ -696,6 +773,7 @@ describe('useGeminiStream', () => {
             },
         ];
         const client = new MockedGeminiClientClass(mockConfig);
+        client.setHistory([{ role: 'user', parts: [{ text: 'User prompt' }] }]);
         // Capture the onComplete callback
         let capturedOnComplete = null;
         mockUseToolScheduler.mockImplementation((onComplete) => {
@@ -709,7 +787,20 @@ describe('useGeminiStream', () => {
                 0,
             ];
         });
-        await renderHookWithProviders(() => useGeminiStream(client, [], mockAddItem, mockConfig, mockLoadedSettings, mockOnDebugMessage, mockHandleSlashCommand, false, () => 'vscode', () => { }, () => Promise.resolve(), false, () => { }, () => { }, () => { }, 80, 24));
+        const { result } = await renderHookWithProviders(() => useGeminiStream(client, [], mockAddItem, mockConfig, mockLoadedSettings, mockOnDebugMessage, mockHandleSlashCommand, false, () => 'vscode', () => { }, () => Promise.resolve(), false, () => { }, () => { }, () => { }, 80, 24));
+        // Call submitQuery to populate the user turn and set historyLengthAfterUserPromptRef
+        await act(async () => {
+            // eslint-disable-next-line @typescript-eslint/no-floating-promises
+            result.current.submitQuery('User prompt');
+        });
+        // Model issues a functionCall request, which appends to history
+        client.setHistory([
+            { role: 'user', parts: [{ text: 'User prompt' }] },
+            {
+                role: 'model',
+                parts: [{ functionCall: { name: 'testTool', args: {} } }],
+            },
+        ]);
         // Trigger the onComplete callback with cancelled tools
         await act(async () => {
             if (capturedOnComplete) {
@@ -720,21 +811,12 @@ describe('useGeminiStream', () => {
         });
         await waitFor(() => {
             expect(mockMarkToolsAsSubmitted).toHaveBeenCalledWith(['topic1', '1']);
-            expect(client.addHistory).toHaveBeenCalledWith({
-                role: 'user',
-                parts: [
-                    {
-                        functionResponse: {
-                            name: UPDATE_TOPIC_TOOL_NAME,
-                            id: 'topic1',
-                            response: {},
-                        },
-                    },
-                    { text: CoreToolCallStatus.Cancelled },
-                ],
-            });
-            // Ensure we do NOT call back to the API
-            expect(mockSendMessageStream).not.toHaveBeenCalled();
+            // Should NOT have appended cancellations via addHistory
+            expect(client.addHistory).not.toHaveBeenCalled();
+            // Should have rolled history back to pre-model length (1)
+            expect(client.getHistory().length).toBe(1);
+            // Ensure we do NOT call back to the API a second time (only the initial user turn was sent)
+            expect(mockSendMessageStream).toHaveBeenCalledTimes(1);
         });
     });
     it('should record tool responses in history when the model was switched due to a quota error', async () => {
@@ -972,7 +1054,7 @@ describe('useGeminiStream', () => {
         expect(failureHintIndex).toBe(-1);
         expect(noteIndex).toBeLessThan(stopIndex);
     });
-    it('should group multiple cancelled tool call responses into a single history entry', async () => {
+    it('should rollback multiple cancelled tool calls rather than appending them to history', async () => {
         const cancelledToolCall1 = {
             request: {
                 callId: 'cancel-1',
@@ -1033,6 +1115,7 @@ describe('useGeminiStream', () => {
         };
         const allCancelledTools = [cancelledToolCall1, cancelledToolCall2];
         const client = new MockedGeminiClientClass(mockConfig);
+        client.setHistory([{ role: 'user', parts: [{ text: 'User prompt' }] }]);
         let capturedOnComplete = null;
         mockUseToolScheduler.mockImplementation((onComplete) => {
             capturedOnComplete = onComplete;
@@ -1045,7 +1128,17 @@ describe('useGeminiStream', () => {
                 0,
             ];
         });
-        await renderHookWithProviders(() => useGeminiStream(client, [], mockAddItem, mockConfig, mockLoadedSettings, mockOnDebugMessage, mockHandleSlashCommand, false, () => 'vscode', () => { }, () => Promise.resolve(), false, () => { }, () => { }, () => { }, 80, 24));
+        const { result } = await renderHookWithProviders(() => useGeminiStream(client, [], mockAddItem, mockConfig, mockLoadedSettings, mockOnDebugMessage, mockHandleSlashCommand, false, () => 'vscode', () => { }, () => Promise.resolve(), false, () => { }, () => { }, () => { }, 80, 24));
+        // Call submitQuery to populate the user turn and set historyLengthAfterUserPromptRef
+        await act(async () => {
+            // eslint-disable-next-line @typescript-eslint/no-floating-promises
+            result.current.submitQuery('User prompt');
+        });
+        // Model issues model turns, which appends to history
+        client.setHistory([
+            { role: 'user', parts: [{ text: 'User prompt' }] },
+            { role: 'model', parts: [{ functionCall: { name: 'toolA', args: {} } }] },
+        ]);
         // Trigger the onComplete callback with multiple cancelled tools
         await act(async () => {
             if (capturedOnComplete) {
@@ -1060,18 +1153,12 @@ describe('useGeminiStream', () => {
                 'cancel-1',
                 'cancel-2',
             ]);
-            // Crucially, addHistory should be called only ONCE
-            expect(client.addHistory).toHaveBeenCalledTimes(1);
-            // And that single call should contain BOTH function responses
-            expect(client.addHistory).toHaveBeenCalledWith({
-                role: 'user',
-                parts: [
-                    ...cancelledToolCall1.response.responseParts,
-                    ...cancelledToolCall2.response.responseParts,
-                ],
-            });
+            // Crucially, addHistory should NOT be called
+            expect(client.addHistory).not.toHaveBeenCalled();
+            // Instead, history should be rolled back to pre-model length (1)
+            expect(client.getHistory().length).toBe(1);
             // No message should be sent back to the API for a turn with only cancellations
-            expect(mockSendMessageStream).not.toHaveBeenCalled();
+            expect(mockSendMessageStream).toHaveBeenCalledTimes(1);
         });
     });
     it('should not flicker streaming state to Idle between tool completion and submission', async () => {