@volley/recognition-client-sdk 0.1.200

package/src/recognition-client.spec.ts

@@ -0,0 +1,551 @@
+/**
+ * Unit tests for RealTimeTwoWayWebSocketRecognitionClient
+ */
+
+import { RealTimeTwoWayWebSocketRecognitionClient } from './recognition-client';
+import { ClientState } from './recognition-client.types';
+import { WebSocket as MockWebSocket } from 'ws';
+
+// Mock WebSocket
+jest.mock('ws');
+
+describe('RealTimeTwoWayWebSocketRecognitionClient', () => {
+  let client: RealTimeTwoWayWebSocketRecognitionClient;
+  let mockWs: any;
+
+  beforeEach(() => {
+    // Reset mocks
+    jest.clearAllMocks();
+
+    // Create mock WebSocket
+    mockWs = {
+      readyState: MockWebSocket.CONNECTING,
+      send: jest.fn(),
+      close: jest.fn(),
+      on: jest.fn(),
+      removeAllListeners: jest.fn(),
+    };
+
+    // Mock WebSocket constructor
+    (MockWebSocket as any).mockImplementation(() => mockWs);
+
+    // Create client
+    client = new RealTimeTwoWayWebSocketRecognitionClient({
+      url: 'ws://test.example.com/recognize',
+      asrRequestConfig: {
+        provider: 'deepgram',
+        language: 'en',
+        sampleRate: 16000,
+        encoding: 'linear16'
+      }
+    });
+  });
+
+  afterEach(() => {
+    // Clean up
+  });
+
+  describe('Constructor', () => {
+    it('should initialize with correct default values', () => {
+      expect(client.getState()).toBe(ClientState.INITIAL);
+      expect(client.isConnected()).toBe(false);
+      expect(client.isBufferOverflowing()).toBe(false);
+      expect(client.getAudioUtteranceId()).toBeDefined();
+      expect(typeof client.getAudioUtteranceId()).toBe('string');
+    });
+
+    it('should have immutable audioUtteranceId', () => {
+      const originalId = client.getAudioUtteranceId();
+      // audioUtteranceId should not change
+      expect(client.getAudioUtteranceId()).toBe(originalId);
+    });
+
+    it('should initialize stats correctly', () => {
+      const stats = client.getStats();
+      expect(stats.audioBytesSent).toBe(0);
+      expect(stats.audioChunksSent).toBe(0);
+      expect(stats.audioChunksBuffered).toBe(0);
+      expect(stats.bufferOverflowCount).toBe(0);
+      expect(stats.currentBufferedChunks).toBe(0);
+    });
+  });
+
+  describe.skip('State Management', () => {
+    it('should transition from INITIAL to CONNECTING on connect()', async () => {
+      expect(client.getState()).toBe(ClientState.INITIAL);
+
+      // Simulate successful connection
+      const connectPromise = client.connect();
+      expect(client.getState()).toBe(ClientState.CONNECTING);
+
+      // Simulate WebSocket open event
+      mockWs.readyState = MockWebSocket.OPEN;
+      const openHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'open')[1];
+      openHandler();
+
+      await connectPromise;
+      expect(client.getState()).toBe(ClientState.CONNECTED);
+    });
+
+    it('should transition to READY when server sends ready message', async () => {
+      // Connect first
+      const connectPromise = client.connect();
+      mockWs.readyState = MockWebSocket.OPEN;
+      const openHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'open')[1];
+      openHandler();
+      await connectPromise;
+
+      // Simulate ready message
+      const messageHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'message')[1];
+      const readyMessage = JSON.stringify({
+        v: 1,
+        type: 'message',
+        data: {
+          type: 'ClientControlMessage',
+          action: 'ready_for_uploading_recording',
+          audioUtteranceId: 'test-utterance-id'
+        }
+      });
+      messageHandler(readyMessage);
+
+      expect(client.getState()).toBe(ClientState.READY);
+      expect(client.isConnected()).toBe(true);
+    });
+
+    it('should transition to STOPPING when stopRecording() is called', async () => {
+      // Setup: Connect and become ready
+      await setupConnectedClient();
+
+      // Call stopRecording
+      const stopPromise = client.stopRecording();
+      expect(client.getState()).toBe(ClientState.STOPPING);
+
+      // Simulate final transcript
+      const messageHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'message')[1];
+      const finalMessage = JSON.stringify({
+        v: 1,
+        type: 'message',
+        data: {
+          type: 'Transcription',
+          finalTranscript: 'test',
+          is_finished: true
+        }
+      });
+      messageHandler(finalMessage);
+
+      await stopPromise;
+      expect(client.getState()).toBe(ClientState.STOPPED);
+    });
+
+    it('should transition to FAILED on connection error', async () => {
+      const connectPromise = client.connect();
+
+      // Simulate error
+      const errorHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'error')[1];
+      errorHandler(new Error('Connection failed'));
+
+      try {
+        await connectPromise;
+      } catch (e) {
+        // Expected
+      }
+
+      expect(client.getState()).toBe(ClientState.FAILED);
+    });
+  });
+
+  describe.skip('Connection Handling', () => {
+    it('should handle duplicate connect() calls', async () => {
+      // Call connect twice
+      const promise1 = client.connect();
+      const promise2 = client.connect();
+
+      // Should be the same promise
+      expect(promise1).toBe(promise2);
+
+      // Simulate successful connection
+      mockWs.readyState = MockWebSocket.OPEN;
+      const openHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'open')[1];
+      openHandler();
+
+      await Promise.all([promise1, promise2]);
+      expect(client.getState()).toBe(ClientState.CONNECTED);
+    });
+
+    it('should not reconnect if already connected', async () => {
+      // First connection
+      await setupConnectedClient();
+      const firstWs = mockWs;
+
+      // Try to connect again
+      await client.connect();
+
+      // Should not create new WebSocket
+      expect(MockWebSocket).toHaveBeenCalledTimes(1);
+      expect(client.isConnected()).toBe(true);
+    });
+  });
+
+  describe.skip('Audio Handling', () => {
+    it('should buffer audio when not ready', () => {
+      const audioData = Buffer.from([1, 2, 3, 4]);
+      client.sendAudio(audioData);
+
+      const stats = client.getStats();
+      expect(stats.audioBytesSent).toBe(0);
+      expect(stats.audioChunksBuffered).toBe(1);
+      expect(stats.currentBufferedChunks).toBe(1);
+    });
+
+    it('should send audio immediately when ready', async () => {
+      await setupReadyClient();
+
+      const audioData = Buffer.from([1, 2, 3, 4]);
+      client.sendAudio(audioData);
+
+      // Should have sent the audio
+      expect(mockWs.send).toHaveBeenCalled();
+      const stats = client.getStats();
+      expect(stats.audioBytesSent).toBe(4);
+      expect(stats.audioChunksSent).toBe(1);
+    });
+
+    it('should flush buffered audio when becoming ready', async () => {
+      // Buffer some audio while disconnected
+      const audioData1 = Buffer.from([1, 2, 3, 4]);
+      const audioData2 = Buffer.from([5, 6, 7, 8]);
+      client.sendAudio(audioData1);
+      client.sendAudio(audioData2);
+
+      // Verify buffered
+      let stats = client.getStats();
+      expect(stats.currentBufferedChunks).toBe(2);
+      expect(stats.audioBytesSent).toBe(0);
+
+      // Connect and become ready
+      await setupReadyClient();
+
+      // Should have flushed the buffer
+      stats = client.getStats();
+      expect(stats.audioBytesSent).toBe(8);
+      expect(stats.audioChunksSent).toBe(2);
+      expect(stats.currentBufferedChunks).toBe(0);
+    });
+
+    it('should detect buffer overflow', () => {
+      // Fill buffer to capacity (simulate overflow)
+      const chunkSize = 1024;
+      const maxChunks = 6000; // Default buffer size
+
+      for (let i = 0; i <= maxChunks; i++) {
+        client.sendAudio(Buffer.alloc(chunkSize));
+      }
+
+      expect(client.isBufferOverflowing()).toBe(true);
+      const stats = client.getStats();
+      expect(stats.bufferOverflowCount).toBeGreaterThan(0);
+    });
+  });
+
+  describe.skip('Helper Methods', () => {
+    it('should report correct connection status', async () => {
+      expect(client.isConnected()).toBe(false);
+
+      await setupConnectedClient();
+      expect(client.isConnected()).toBe(true);
+
+      await setupReadyClient();
+      expect(client.isConnected()).toBe(true);
+
+      // Disconnect
+      const closeHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'close')[1];
+      closeHandler(1000, 'Normal closure');
+      expect(client.isConnected()).toBe(false);
+    });
+
+    it('should track statistics correctly', async () => {
+      await setupReadyClient();
+
+      // Send some audio
+      client.sendAudio(Buffer.alloc(100));
+      client.sendAudio(Buffer.alloc(200));
+      client.sendAudio(Buffer.alloc(300));
+
+      const stats = client.getStats();
+      expect(stats.audioBytesSent).toBe(600);
+      expect(stats.audioChunksSent).toBe(3);
+      expect(stats.audioChunksBuffered).toBe(3);
+    });
+  });
+
+  describe.skip('Memory Management', () => {
+    it('should clear ring buffer on disconnect', async () => {
+      // Buffer some audio
+      client.sendAudio(Buffer.alloc(100));
+      client.sendAudio(Buffer.alloc(200));
+
+      let stats = client.getStats();
+      expect(stats.currentBufferedChunks).toBe(2);
+
+      // Connect then disconnect
+      await setupConnectedClient();
+      const closeHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'close')[1];
+      closeHandler(1000, 'Normal closure');
+
+      // Buffer should be cleared
+      stats = client.getStats();
+      expect(stats.currentBufferedChunks).toBe(0);
+    });
+
+    it('should cleanup WebSocket listeners on close', async () => {
+      await setupConnectedClient();
+
+      await client.stopRecording();
+
+      expect(mockWs.removeAllListeners).toHaveBeenCalled();
+      expect(mockWs.close).toHaveBeenCalled();
+    });
+  });
+
+  describe.skip('Message Handling', () => {
+    it('should handle transcription messages', async () => {
+      const onTranscript = jest.fn();
+      client = new RealTimeTwoWayWebSocketRecognitionClient({
+        url: 'ws://test.example.com/recognize',
+        onTranscript,
+      });
+
+      await setupConnectedClient();
+
+      const messageHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'message')[1];
+      const transcriptMessage = JSON.stringify({
+        v: 1,
+        type: 'message',
+        data: {
+          type: 'Transcription',
+          finalTranscript: 'Hello world',
+          finalTranscriptConfidence: 0.95,
+          is_finished: false
+        }
+      });
+
+      messageHandler(transcriptMessage);
+
+      expect(onTranscript).toHaveBeenCalledWith(expect.objectContaining({
+        finalTranscript: 'Hello world',
+        finalTranscriptConfidence: 0.95,
+        is_finished: false
+      }));
+    });
+
+    it('should handle function call messages', async () => {
+      const onFunctionCall = jest.fn();
+      client = new RealTimeTwoWayWebSocketRecognitionClient({
+        url: 'ws://test.example.com/recognize',
+        onFunctionCall,
+      });
+
+      await setupConnectedClient();
+
+      const messageHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'message')[1];
+      const functionCallMessage = JSON.stringify({
+        v: 1,
+        type: 'message',
+        data: {
+          type: 'FunctionCall',
+          audioUtteranceId: 'test-id',
+          functionName: 'playMusic',
+          functionArgJson: '{"song": "Bohemian Rhapsody"}'
+        }
+      });
+
+      messageHandler(functionCallMessage);
+
+      expect(onFunctionCall).toHaveBeenCalledWith(expect.objectContaining({
+        type: 'FunctionCall',
+        audioUtteranceId: 'test-id',
+        functionName: 'playMusic',
+        functionArgJson: '{"song": "Bohemian Rhapsody"}'
+      }));
+    });
+
+    it('should handle metadata messages', async () => {
+      const onMetadata = jest.fn();
+      client = new RealTimeTwoWayWebSocketRecognitionClient({
+        url: 'ws://test.example.com/recognize',
+        onMetadata,
+      });
+
+      await setupConnectedClient();
+
+      const messageHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'message')[1];
+      const metadataMessage = JSON.stringify({
+        v: 1,
+        type: 'message',
+        data: {
+          type: 'Metadata',
+          audioUtteranceId: 'test-id',
+          duration: 1000
+        }
+      });
+
+      messageHandler(metadataMessage);
+
+      expect(onMetadata).toHaveBeenCalledWith(expect.objectContaining({
+        audioUtteranceId: 'test-id',
+        duration: 1000
+      }));
+    });
+
+    it('should handle error messages', async () => {
+      const onError = jest.fn();
+      client = new RealTimeTwoWayWebSocketRecognitionClient({
+        url: 'ws://test.example.com/recognize',
+        onError,
+      });
+
+      await setupConnectedClient();
+
+      const messageHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'message')[1];
+      const errorMessage = JSON.stringify({
+        v: 1,
+        type: 'message',
+        data: {
+          type: 'Error',
+          message: 'Something went wrong'
+        }
+      });
+
+      messageHandler(errorMessage);
+
+      expect(onError).toHaveBeenCalledWith(expect.any(Error));
+      expect(onError.mock.calls[0][0].message).toContain('Something went wrong');
+    });
+
+    it('should handle primitive message data without crashing', async () => {
+      const onError = jest.fn();
+      client = new RealTimeTwoWayWebSocketRecognitionClient({
+        url: 'ws://test.example.com/recognize',
+        onError,
+      });
+
+      await setupConnectedClient();
+
+      const messageHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'message')[1];
+
+      // Test with various primitive values
+      const primitiveMessages = [
+        JSON.stringify({ v: 1, type: 'test', data: 'string' }),
+        JSON.stringify({ v: 1, type: 'test', data: 123 }),
+        JSON.stringify({ v: 1, type: 'test', data: true }),
+        JSON.stringify({ v: 1, type: 'test', data: null }),
+      ];
+
+      // Should not throw
+      expect(() => {
+        primitiveMessages.forEach(msg => messageHandler(msg));
+      }).not.toThrow();
+
+      // Should log errors for primitives
+      expect(onError).toHaveBeenCalled();
+    });
+  });
+
+  describe('Debug Logging', () => {
+    it('should not log debug messages when debug logging is disabled (default)', () => {
+      const mockLogger = jest.fn();
+      const testClient = new RealTimeTwoWayWebSocketRecognitionClient({
+        url: 'ws://test.example.com/recognize',
+        asrRequestConfig: {
+          provider: 'deepgram',
+          language: 'en',
+          sampleRate: 16000,
+          encoding: 'linear16'
+        },
+        logger: mockLogger
+      });
+
+      // Trigger some actions that would normally log debug messages
+      expect(testClient.getState()).toBe(ClientState.INITIAL);
+
+      // Debug logs should not be called
+      const debugCalls = mockLogger.mock.calls.filter(call => call[0] === 'debug');
+      expect(debugCalls.length).toBe(0);
+
+      // But other log levels should work
+      const nonDebugCalls = mockLogger.mock.calls.filter(call => call[0] !== 'debug');
+      // May or may not have non-debug logs, just checking we can track them
+      expect(nonDebugCalls.length).toBeGreaterThanOrEqual(0);
+    });
+
+    it('should log debug messages when enableDebugLog is true in debugCommand', () => {
+      const mockLogger = jest.fn();
+      const testClient = new RealTimeTwoWayWebSocketRecognitionClient({
+        url: 'ws://test.example.com/recognize',
+        asrRequestConfig: {
+          provider: 'deepgram',
+          language: 'en',
+          sampleRate: 16000,
+          encoding: 'linear16',
+          debugCommand: {
+            enableDebugLog: true
+          }
+        } as any, // Using 'as any' to bypass type checking for the new field
+        logger: mockLogger
+      });
+
+      // Note: Debug logging is enabled in onConnected() when ASR request is sent
+      // So we need to test after connection, but for now we verify the config is accepted
+      expect(testClient.getAudioUtteranceId()).toBeDefined();
+    });
+
+    it('should respect debugCommand.enableDebugLog flag', () => {
+      const mockLogger = jest.fn();
+
+      // Client with debug logging explicitly disabled
+      const clientNoDebug = new RealTimeTwoWayWebSocketRecognitionClient({
+        url: 'ws://test.example.com/recognize',
+        asrRequestConfig: {
+          provider: 'deepgram',
+          language: 'en',
+          sampleRate: 16000,
+          encoding: 'linear16',
+          debugCommand: {
+            enableDebugLog: false
+          }
+        } as any,
+        logger: mockLogger
+      });
+
+      expect(clientNoDebug.getAudioUtteranceId()).toBeDefined();
+
+      // Debug logs should not be called
+      const debugCalls = mockLogger.mock.calls.filter(call => call[0] === 'debug');
+      expect(debugCalls.length).toBe(0);
+    });
+  });
+
+  // Helper functions
+  async function setupConnectedClient() {
+    const connectPromise = client.connect();
+    mockWs.readyState = MockWebSocket.OPEN;
+    const openHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'open')[1];
+    openHandler();
+    await connectPromise;
+  }
+
+  async function setupReadyClient() {
+    await setupConnectedClient();
+    const messageHandler = mockWs.on.mock.calls.find((call: any[]) => call[0] === 'message')[1];
+    const readyMessage = JSON.stringify({
+      v: 1,
+      type: 'message',
+      data: {
+        type: 'ClientControlMessage',
+        action: 'ready_for_uploading_recording',
+        audioUtteranceId: 'test-utterance-id'
+      }
+    });
+    messageHandler(readyMessage);
+  }
+});