whisper.rn 0.5.0-rc.0 → 0.5.0-rc.10

package/lib/commonjs/realtime-transcription/RealtimeTranscriber.js

@@ -0,0 +1,857 @@
+"use strict";
+
+Object.defineProperty(exports, "__esModule", {
+  value: true
+});
+exports.RealtimeTranscriber = void 0;
+var _SliceManager = require("./SliceManager");
+var _WavFileWriter = require("../utils/WavFileWriter");
+var _types = require("./types");
+/* eslint-disable class-methods-use-this */
+
+/**
+ * RealtimeTranscriber provides real-time audio transcription with VAD support.
+ *
+ * Features:
+ * - Automatic slice management based on duration
+ * - VAD-based speech detection and auto-slicing
+ * - Configurable auto-slice mechanism that triggers on speech_end/silence events
+ * - Memory management for audio slices
+ * - Queue-based transcription processing
+ */
+class RealtimeTranscriber {
+  callbacks = {};
+  isActive = false;
+  isTranscribing = false;
+  vadEnabled = false;
+  transcriptionQueue = [];
+  accumulatedData = new Uint8Array(0);
+  wavFileWriter = null;
+
+  // Simplified VAD state management
+  lastSpeechDetectedTime = 0;
+
+  // Track VAD state for proper event transitions
+  lastVadState = 'silence';
+
+  // Track last stats to emit only when changed
+  lastStatsSnapshot = null;
+
+  // Store transcription results by slice index
+  transcriptionResults = new Map();
+
+  // Store VAD events by slice index for inclusion in transcribe events
+  vadEvents = new Map();
+  constructor(dependencies) {
+    let options = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : {};
+    let callbacks = arguments.length > 2 && arguments[2] !== undefined ? arguments[2] : {};
+    this.whisperContext = dependencies.whisperContext;
+    this.vadContext = dependencies.vadContext;
+    this.audioStream = dependencies.audioStream;
+    this.fs = dependencies.fs;
+    this.callbacks = callbacks;
+
+    // Set default options with proper types
+    this.options = {
+      audioSliceSec: options.audioSliceSec || 30,
+      audioMinSec: options.audioMinSec || 1,
+      maxSlicesInMemory: options.maxSlicesInMemory || 3,
+      vadOptions: options.vadOptions || _types.VAD_PRESETS.default,
+      vadPreset: options.vadPreset,
+      autoSliceOnSpeechEnd: options.autoSliceOnSpeechEnd || true,
+      autoSliceThreshold: options.autoSliceThreshold || 0.5,
+      transcribeOptions: options.transcribeOptions || {},
+      initialPrompt: options.initialPrompt,
+      promptPreviousSlices: options.promptPreviousSlices ?? true,
+      audioOutputPath: options.audioOutputPath,
+      logger: options.logger || (() => {})
+    };
+
+    // Apply VAD preset if specified
+    if (this.options.vadPreset && _types.VAD_PRESETS[this.options.vadPreset]) {
+      this.options.vadOptions = {
+        ..._types.VAD_PRESETS[this.options.vadPreset],
+        ...this.options.vadOptions
+      };
+    }
+
+    // Enable VAD if context is provided and not explicitly disabled
+    this.vadEnabled = !!this.vadContext;
+
+    // Initialize managers
+    this.sliceManager = new _SliceManager.SliceManager(this.options.audioSliceSec, this.options.maxSlicesInMemory);
+
+    // Set up audio stream callbacks
+    this.audioStream.onData(this.handleAudioData.bind(this));
+    this.audioStream.onError(this.handleError.bind(this));
+    this.audioStream.onStatusChange(this.handleAudioStatusChange.bind(this));
+  }
+
+  /**
+   * Start realtime transcription
+   */
+  async start() {
+    if (this.isActive) {
+      throw new Error('Realtime transcription is already active');
+    }
+    try {
+      var _this$callbacks$onSta, _this$callbacks, _this$options$audioSt4, _this$options$audioSt5, _this$options$audioSt6, _this$options$audioSt7, _this$options$audioSt8;
+      this.isActive = true;
+      (_this$callbacks$onSta = (_this$callbacks = this.callbacks).onStatusChange) === null || _this$callbacks$onSta === void 0 ? void 0 : _this$callbacks$onSta.call(_this$callbacks, true);
+
+      // Reset all state to ensure clean start
+      this.reset();
+
+      // Initialize WAV file writer if output path is specified
+      if (this.fs && this.options.audioOutputPath) {
+        var _this$options$audioSt, _this$options$audioSt2, _this$options$audioSt3;
+        this.wavFileWriter = new _WavFileWriter.WavFileWriter(this.fs, this.options.audioOutputPath, {
+          sampleRate: ((_this$options$audioSt = this.options.audioStreamConfig) === null || _this$options$audioSt === void 0 ? void 0 : _this$options$audioSt.sampleRate) || 16000,
+          channels: ((_this$options$audioSt2 = this.options.audioStreamConfig) === null || _this$options$audioSt2 === void 0 ? void 0 : _this$options$audioSt2.channels) || 1,
+          bitsPerSample: ((_this$options$audioSt3 = this.options.audioStreamConfig) === null || _this$options$audioSt3 === void 0 ? void 0 : _this$options$audioSt3.bitsPerSample) || 16
+        });
+        await this.wavFileWriter.initialize();
+      }
+
+      // Start audio recording
+      await this.audioStream.initialize({
+        sampleRate: ((_this$options$audioSt4 = this.options.audioStreamConfig) === null || _this$options$audioSt4 === void 0 ? void 0 : _this$options$audioSt4.sampleRate) || 16000,
+        channels: ((_this$options$audioSt5 = this.options.audioStreamConfig) === null || _this$options$audioSt5 === void 0 ? void 0 : _this$options$audioSt5.channels) || 1,
+        bitsPerSample: ((_this$options$audioSt6 = this.options.audioStreamConfig) === null || _this$options$audioSt6 === void 0 ? void 0 : _this$options$audioSt6.bitsPerSample) || 16,
+        audioSource: ((_this$options$audioSt7 = this.options.audioStreamConfig) === null || _this$options$audioSt7 === void 0 ? void 0 : _this$options$audioSt7.audioSource) || 6,
+        bufferSize: ((_this$options$audioSt8 = this.options.audioStreamConfig) === null || _this$options$audioSt8 === void 0 ? void 0 : _this$options$audioSt8.bufferSize) || 16 * 1024
+      });
+      await this.audioStream.start();
+
+      // Emit stats update for status change
+      this.emitStatsUpdate('status_change');
+      this.log('Realtime transcription started');
+    } catch (error) {
+      var _this$callbacks$onSta2, _this$callbacks2;
+      this.isActive = false;
+      (_this$callbacks$onSta2 = (_this$callbacks2 = this.callbacks).onStatusChange) === null || _this$callbacks$onSta2 === void 0 ? void 0 : _this$callbacks$onSta2.call(_this$callbacks2, false);
+      throw error;
+    }
+  }
+
+  /**
+   * Stop realtime transcription
+   */
+  async stop() {
+    if (!this.isActive) {
+      return;
+    }
+    try {
+      var _this$callbacks$onSta3, _this$callbacks3;
+      this.isActive = false;
+
+      // Stop audio recording
+      await this.audioStream.stop();
+
+      // Process any remaining accumulated data
+      if (this.accumulatedData.length > 0) {
+        this.processAccumulatedDataForSliceManagement();
+      }
+
+      // Process any remaining queued transcriptions
+      await this.processTranscriptionQueue();
+
+      // Finalize WAV file
+      if (this.wavFileWriter) {
+        await this.wavFileWriter.finalize();
+        this.wavFileWriter = null;
+      }
+
+      // Reset all state completely
+      this.reset();
+      (_this$callbacks$onSta3 = (_this$callbacks3 = this.callbacks).onStatusChange) === null || _this$callbacks$onSta3 === void 0 ? void 0 : _this$callbacks$onSta3.call(_this$callbacks3, false);
+
+      // Emit stats update for status change
+      this.emitStatsUpdate('status_change');
+      this.log('Realtime transcription stopped');
+    } catch (error) {
+      this.handleError(`Stop error: ${error}`);
+    }
+  }
+
+  /**
+   * Handle incoming audio data from audio stream
+   */
+  handleAudioData(streamData) {
+    if (!this.isActive) {
+      return;
+    }
+    try {
+      // Write to WAV file if enabled (convert to Uint8Array for WavFileWriter)
+      if (this.wavFileWriter) {
+        this.wavFileWriter.appendAudioData(streamData.data).catch(error => {
+          this.log(`Failed to write audio to WAV file: ${error}`);
+        });
+      }
+
+      // Always accumulate data for slice management
+      this.accumulateAudioData(streamData.data);
+    } catch (error) {
+      const errorMessage = error instanceof Error ? error.message : 'Audio processing error';
+      this.handleError(errorMessage);
+    }
+  }
+
+  /**
+   * Accumulate audio data for slice management
+   */
+  accumulateAudioData(newData) {
+    const combined = new Uint8Array(this.accumulatedData.length + newData.length);
+    combined.set(this.accumulatedData);
+    combined.set(new Uint8Array(newData), this.accumulatedData.length);
+    this.accumulatedData = combined;
+
+    // Process accumulated data when we have enough for slice management
+    const minBufferSamples = 16000 * 1; // 1 second for slice management
+    if (this.accumulatedData.length >= minBufferSamples) {
+      this.processAccumulatedDataForSliceManagement();
+    }
+  }
+
+  /**
+   * Process accumulated audio data through SliceManager
+   */
+  processAccumulatedDataForSliceManagement() {
+    if (this.accumulatedData.length === 0) {
+      return;
+    }
+
+    // Process through slice manager directly with Uint8Array
+    const result = this.sliceManager.addAudioData(this.accumulatedData);
+    if (result.slice) {
+      this.log(`Slice ${result.slice.index} ready (${result.slice.data.length} bytes)`);
+
+      // Process VAD for the slice if enabled
+      if (!this.isTranscribing && this.vadEnabled) {
+        this.processSliceVAD(result.slice).catch(error => {
+          this.handleError(`VAD processing error: ${error}`);
+        });
+      } else if (!this.isTranscribing) {
+        // If VAD is disabled, transcribe slices as they become ready
+        this.queueSliceForTranscription(result.slice).catch(error => {
+          this.handleError(`Failed to queue slice for transcription: ${error}`);
+        });
+      } else {
+        this.log(`Skipping slice ${result.slice.index} - already transcribing`);
+      }
+      this.emitStatsUpdate('memory_change');
+    }
+
+    // Clear accumulated data
+    this.accumulatedData = new Uint8Array(0);
+  }
+
+  /**
+   * Check if auto-slice should be triggered based on VAD event and timing
+   */
+  async checkAutoSlice(vadEvent, _slice) {
+    if (!this.options.autoSliceOnSpeechEnd || !this.vadEnabled) {
+      return;
+    }
+
+    // Only trigger on speech_end or silence events
+    const shouldTriggerAutoSlice = vadEvent.type === 'speech_end' || vadEvent.type === 'silence';
+    if (!shouldTriggerAutoSlice) {
+      return;
+    }
+
+    // Get current slice info from SliceManager
+    const currentSliceInfo = this.sliceManager.getCurrentSliceInfo();
+    const currentSlice = this.sliceManager.getSliceByIndex(currentSliceInfo.currentSliceIndex);
+    if (!currentSlice) {
+      return;
+    }
+
+    // Calculate current slice duration
+    const currentDuration = (Date.now() - currentSlice.startTime) / 1000; // Convert to seconds
+    const targetDuration = this.options.audioSliceSec;
+    const minDuration = this.options.audioMinSec;
+    const autoSliceThreshold = targetDuration * this.options.autoSliceThreshold;
+
+    // Check if conditions are met for auto-slice
+    const meetsMinDuration = currentDuration >= minDuration;
+    const meetsThreshold = currentDuration >= autoSliceThreshold;
+    if (meetsMinDuration && meetsThreshold) {
+      this.log(`Auto-slicing on ${vadEvent.type} at ${currentDuration.toFixed(1)}s ` + `(min: ${minDuration}s, threshold: ${autoSliceThreshold.toFixed(1)}s, target: ${targetDuration}s)`);
+
+      // Force next slice
+      await this.nextSlice();
+    } else {
+      this.log(`Auto-slice conditions not met on ${vadEvent.type}: ` + `duration=${currentDuration.toFixed(1)}s, min=${minDuration}s, threshold=${autoSliceThreshold.toFixed(1)}s ` + `(minOk=${meetsMinDuration}, thresholdOk=${meetsThreshold})`);
+    }
+  }
+
+  /**
+   * Process VAD for a completed slice
+   */
+  async processSliceVAD(slice) {
+    try {
+      var _this$callbacks$onVad, _this$callbacks4;
+      // Get audio data from the slice for VAD processing
+      const audioData = this.sliceManager.getAudioDataForTranscription(slice.index);
+      if (!audioData) {
+        this.log(`No audio data available for VAD processing of slice ${slice.index}`);
+        return;
+      }
+
+      // Convert base64 back to Uint8Array for VAD processing
+
+      // Detect speech in the slice
+      const vadEvent = await this.detectSpeech(audioData, slice.index);
+      vadEvent.timestamp = Date.now();
+
+      // Store VAD event for inclusion in transcribe event
+      this.vadEvents.set(slice.index, vadEvent);
+
+      // Emit VAD event
+      (_this$callbacks$onVad = (_this$callbacks4 = this.callbacks).onVad) === null || _this$callbacks$onVad === void 0 ? void 0 : _this$callbacks$onVad.call(_this$callbacks4, vadEvent);
+
+      // Check if auto-slice should be triggered
+      await this.checkAutoSlice(vadEvent, slice);
+
+      // Check if speech was detected and if we should transcribe
+      const isSpeech = vadEvent.type === 'speech_start' || vadEvent.type === 'speech_continue';
+      const isSpeechEnd = vadEvent.type === 'speech_end';
+      if (isSpeech) {
+        const minDuration = this.options.audioMinSec;
+        // Check minimum duration requirement
+        const speechDuration = slice.data.length / 16000 / 2; // Convert bytes to seconds (16kHz, 16-bit)
+
+        if (speechDuration >= minDuration) {
+          this.log(`Speech detected in slice ${slice.index}, queueing for transcription`);
+          await this.queueSliceForTranscription(slice);
+        } else {
+          this.log(`Speech too short in slice ${slice.index} (${speechDuration.toFixed(2)}s < ${minDuration}s), skipping`);
+        }
+      } else if (isSpeechEnd) {
+        this.log(`Speech ended in slice ${slice.index}`);
+        // For speech_end events, we might want to queue the slice for transcription
+        // to capture the final part of the speech segment
+        const speechDuration = slice.data.length / 16000 / 2; // Convert bytes to seconds
+        const minDuration = this.options.audioMinSec;
+        if (speechDuration >= minDuration) {
+          this.log(`Speech end detected in slice ${slice.index}, queueing final segment for transcription`);
+          await this.queueSliceForTranscription(slice);
+        } else {
+          this.log(`Speech end segment too short in slice ${slice.index} (${speechDuration.toFixed(2)}s < ${minDuration}s), skipping`);
+        }
+      } else {
+        this.log(`No speech detected in slice ${slice.index}`);
+      }
+
+      // Emit stats update for VAD change
+      this.emitStatsUpdate('vad_change');
+    } catch (error) {
+      this.handleError(`VAD processing error for slice ${slice.index}: ${error}`);
+    }
+  }
+
+  /**
+   * Queue a slice for transcription
+   */
+  async queueSliceForTranscription(slice) {
+    try {
+      // Get audio data from the slice
+      const audioData = this.sliceManager.getAudioDataForTranscription(slice.index);
+      if (!audioData) {
+        this.log(`No audio data available for slice ${slice.index}`);
+        return;
+      }
+      if (this.callbacks.onBeginTranscribe) {
+        const shouldTranscribe = (await this.callbacks.onBeginTranscribe({
+          sliceIndex: slice.index,
+          audioData,
+          duration: slice.data.length / 16000 / 2 * 1000,
+          // Convert to milliseconds
+          vadEvent: this.vadEvents.get(slice.index)
+        })) ?? true;
+        if (!shouldTranscribe) {
+          this.log(`User callback declined transcription for slice ${slice.index}`);
+          return;
+        }
+      }
+
+      // Add to transcription queue
+      this.transcriptionQueue.unshift({
+        sliceIndex: slice.index,
+        audioData
+      });
+      this.log(`Queued slice ${slice.index} for transcription (${slice.data.length} samples)`);
+      await this.processTranscriptionQueue();
+    } catch (error) {
+      this.handleError(`Failed to queue slice for transcription: ${error}`);
+    }
+  }
+
+  /**
+   * Detect speech using VAD context
+   */
+  async detectSpeech(audioData, sliceIndex) {
+    if (!this.vadContext) {
+      // When no VAD context is available, assume speech is always detected
+      // but still follow the state machine pattern
+      const currentTimestamp = Date.now();
+
+      // Assume speech is always detected when no VAD context
+      const vadEventType = this.lastVadState === 'silence' ? 'speech_start' : 'speech_continue';
+
+      // Update VAD state
+      this.lastVadState = 'speech';
+      const {
+        sampleRate = 16000
+      } = this.options.audioStreamConfig || {};
+      return {
+        type: vadEventType,
+        lastSpeechDetectedTime: 0,
+        timestamp: currentTimestamp,
+        confidence: 1.0,
+        duration: audioData.length / sampleRate / 2,
+        // Convert bytes to seconds
+        sliceIndex
+      };
+    }
+    try {
+      const audioBuffer = audioData.buffer;
+
+      // Use VAD context to detect speech segments
+      const vadSegments = await this.vadContext.detectSpeechData(audioBuffer, this.options.vadOptions);
+
+      // Calculate confidence based on speech segments
+      let confidence = 0.0;
+      let lastSpeechDetectedTime = 0;
+      if (vadSegments && vadSegments.length > 0) {
+        var _vadSegments;
+        // If there are speech segments, calculate average confidence
+        const totalTime = vadSegments.reduce((sum, segment) => sum + (segment.t1 - segment.t0), 0);
+        const audioDuration = audioData.length / 16000 / 2; // Convert bytes to seconds
+        confidence = totalTime > 0 ? Math.min(totalTime / audioDuration, 1.0) : 0.0;
+        lastSpeechDetectedTime = ((_vadSegments = vadSegments[vadSegments.length - 1]) === null || _vadSegments === void 0 ? void 0 : _vadSegments.t1) || -1;
+      }
+      const threshold = this.options.vadOptions.threshold || 0.5;
+      let isSpeech = confidence > threshold;
+      const currentTimestamp = Date.now();
+
+      // Determine VAD event type based on current and previous state
+      let vadEventType;
+      if (isSpeech) {
+        vadEventType = this.lastVadState === 'silence' ? 'speech_start' : 'speech_continue';
+        const minDuration = this.options.audioMinSec;
+        // Check if this is a new speech detection (different from last detected time)
+        if (lastSpeechDetectedTime === this.lastSpeechDetectedTime || (lastSpeechDetectedTime - this.lastSpeechDetectedTime) / 100 < minDuration) {
+          if (this.lastVadState === 'silence') vadEventType = 'silence';
+          if (this.lastVadState === 'speech') vadEventType = 'speech_end';
+          isSpeech = false;
+          confidence = 0.0;
+        }
+        this.lastSpeechDetectedTime = lastSpeechDetectedTime;
+      } else {
+        vadEventType = this.lastVadState === 'speech' ? 'speech_end' : 'silence';
+      }
+
+      // Update VAD state for next detection
+      this.lastVadState = isSpeech ? 'speech' : 'silence';
+      const {
+        sampleRate = 16000
+      } = this.options.audioStreamConfig || {};
+      return {
+        type: vadEventType,
+        lastSpeechDetectedTime,
+        timestamp: currentTimestamp,
+        confidence,
+        duration: audioData.length / sampleRate / 2,
+        // Convert bytes to seconds
+        sliceIndex,
+        currentThreshold: threshold
+      };
+    } catch (error) {
+      this.log(`VAD detection error: ${error}`);
+      // Re-throw the error so it can be handled by the caller
+      throw error;
+    }
+  }
+  isProcessingTranscriptionQueue = false;
+
+  /**
+   * Process the transcription queue
+   */
+  async processTranscriptionQueue() {
+    if (this.isProcessingTranscriptionQueue) return;
+    this.isProcessingTranscriptionQueue = true;
+    while (this.transcriptionQueue.length > 0) {
+      const item = this.transcriptionQueue.shift();
+      this.transcriptionQueue = []; // Old items are not needed anymore
+      if (item) {
+        // eslint-disable-next-line no-await-in-loop
+        await this.processTranscription(item).catch(error => {
+          this.handleError(`Transcription error: ${error}`);
+        });
+      }
+    }
+    this.isProcessingTranscriptionQueue = false;
+  }
+
+  /**
+   * Build prompt from initial prompt and previous slices
+   */
+  buildPrompt(currentSliceIndex) {
+    const promptParts = [];
+
+    // Add initial prompt if provided
+    if (this.options.initialPrompt) {
+      promptParts.push(this.options.initialPrompt);
+    }
+
+    // Add previous slice results if enabled
+    if (this.options.promptPreviousSlices) {
+      // Get transcription results from previous slices (up to the current slice)
+      const previousResults = Array.from(this.transcriptionResults.entries()).filter(_ref => {
+        let [sliceIndex] = _ref;
+        return sliceIndex < currentSliceIndex;
+      }).sort((_ref2, _ref3) => {
+        let [a] = _ref2;
+        let [b] = _ref3;
+        return a - b;
+      }) // Sort by slice index
+      .map(_ref4 => {
+        var _result$transcribeEve;
+        let [, result] = _ref4;
+        return (_result$transcribeEve = result.transcribeEvent.data) === null || _result$transcribeEve === void 0 ? void 0 : _result$transcribeEve.result;
+      }).filter(result => Boolean(result)); // Filter out empty results with type guard
+
+      if (previousResults.length > 0) {
+        promptParts.push(...previousResults);
+      }
+    }
+    return promptParts.join(' ') || undefined;
+  }
+
+  /**
+   * Process a single transcription
+   */
+  async processTranscription(item) {
+    if (!this.isActive) {
+      return;
+    }
+    this.isTranscribing = true;
+
+    // Emit stats update for status change
+    this.emitStatsUpdate('status_change');
+    const startTime = Date.now();
+    try {
+      var _this$callbacks$onTra, _this$callbacks5;
+      // Build prompt from initial prompt and previous slices
+      const prompt = this.buildPrompt(item.sliceIndex);
+      const audioBuffer = item.audioData.buffer;
+      const {
+        promise
+      } = this.whisperContext.transcribeData(audioBuffer, {
+        ...this.options.transcribeOptions,
+        prompt,
+        // Include the constructed prompt
+        onProgress: undefined // Disable progress for realtime
+      });
+
+      const result = await promise;
+      const endTime = Date.now();
+
+      // Create transcribe event
+      const {
+        sampleRate = 16000
+      } = this.options.audioStreamConfig || {};
+      const transcribeEvent = {
+        type: 'transcribe',
+        sliceIndex: item.sliceIndex,
+        data: result,
+        isCapturing: this.audioStream.isRecording(),
+        processTime: endTime - startTime,
+        recordingTime: item.audioData.length / (sampleRate / 1000) / 2,
+        // ms,
+        memoryUsage: this.sliceManager.getMemoryUsage(),
+        vadEvent: this.vadEvents.get(item.sliceIndex)
+      };
+
+      // Save transcription results
+      const slice = this.sliceManager.getSliceByIndex(item.sliceIndex);
+      if (slice) {
+        this.transcriptionResults.set(item.sliceIndex, {
+          slice: {
+            // Don't keep data in the slice
+            index: slice.index,
+            sampleCount: slice.sampleCount,
+            startTime: slice.startTime,
+            endTime: slice.endTime,
+            isProcessed: slice.isProcessed,
+            isReleased: slice.isReleased
+          },
+          transcribeEvent
+        });
+      }
+
+      // Emit transcribe event
+      (_this$callbacks$onTra = (_this$callbacks5 = this.callbacks).onTranscribe) === null || _this$callbacks$onTra === void 0 ? void 0 : _this$callbacks$onTra.call(_this$callbacks5, transcribeEvent);
+      this.vadEvents.delete(item.sliceIndex);
+
+      // Emit stats update for memory/slice changes
+      this.emitStatsUpdate('memory_change');
+      this.log(`Transcribed speech segment ${item.sliceIndex}: "${result.result}"`);
+    } catch (error) {
+      var _this$callbacks$onTra2, _this$callbacks6;
+      // Emit error event to transcribe callback
+      const errorEvent = {
+        type: 'error',
+        sliceIndex: item.sliceIndex,
+        data: undefined,
+        isCapturing: this.audioStream.isRecording(),
+        processTime: Date.now() - startTime,
+        recordingTime: 0,
+        memoryUsage: this.sliceManager.getMemoryUsage(),
+        vadEvent: this.vadEvents.get(item.sliceIndex)
+      };
+      (_this$callbacks$onTra2 = (_this$callbacks6 = this.callbacks).onTranscribe) === null || _this$callbacks$onTra2 === void 0 ? void 0 : _this$callbacks$onTra2.call(_this$callbacks6, errorEvent);
+      this.vadEvents.delete(item.sliceIndex);
+      this.handleError(`Transcription failed for speech segment ${item.sliceIndex}: ${error}`);
+    } finally {
+      // Check if we should continue processing queue
+      if (this.transcriptionQueue.length > 0) {
+        await this.processTranscriptionQueue();
+      } else {
+        this.isTranscribing = false;
+      }
+    }
+  }
+
+  /**
+   * Handle audio status changes
+   */
+  handleAudioStatusChange(isRecording) {
+    this.log(`Audio recording: ${isRecording ? 'started' : 'stopped'}`);
+  }
+
+  /**
+   * Handle errors from components
+   */
+  handleError(error) {
+    var _this$callbacks$onErr, _this$callbacks7;
+    this.log(`Error: ${error}`);
+    (_this$callbacks$onErr = (_this$callbacks7 = this.callbacks).onError) === null || _this$callbacks$onErr === void 0 ? void 0 : _this$callbacks$onErr.call(_this$callbacks7, error);
+  }
+
+  /**
+   * Update callbacks
+   */
+  updateCallbacks(callbacks) {
+    this.callbacks = {
+      ...this.callbacks,
+      ...callbacks
+    };
+  }
+
+  /**
+   * Update VAD options dynamically
+   */
+  updateVadOptions(options) {
+    this.options.vadOptions = {
+      ...this.options.vadOptions,
+      ...options
+    };
+  }
+
+  /**
+   * Update auto-slice options dynamically
+   */
+  updateAutoSliceOptions(options) {
+    if (options.autoSliceOnSpeechEnd !== undefined) {
+      this.options.autoSliceOnSpeechEnd = options.autoSliceOnSpeechEnd;
+    }
+    if (options.autoSliceThreshold !== undefined) {
+      this.options.autoSliceThreshold = options.autoSliceThreshold;
+    }
+    this.log(`Auto-slice options updated: enabled=${this.options.autoSliceOnSpeechEnd}, threshold=${this.options.autoSliceThreshold}`);
+  }
+
+  /**
+   * Get current statistics
+   */
+  getStatistics() {
+    return {
+      isActive: this.isActive,
+      isTranscribing: this.isTranscribing,
+      vadEnabled: this.vadEnabled,
+      audioStats: {
+        isRecording: this.audioStream.isRecording(),
+        accumulatedSamples: this.accumulatedData.length
+      },
+      vadStats: this.vadEnabled ? {
+        enabled: true,
+        contextAvailable: !!this.vadContext,
+        lastSpeechDetectedTime: this.lastSpeechDetectedTime
+      } : null,
+      sliceStats: this.sliceManager.getCurrentSliceInfo(),
+      autoSliceConfig: {
+        enabled: this.options.autoSliceOnSpeechEnd,
+        threshold: this.options.autoSliceThreshold,
+        targetDuration: this.options.audioSliceSec,
+        minDuration: this.options.audioMinSec
+      }
+    };
+  }
+
+  /**
+   * Get all transcription results
+   */
+  getTranscriptionResults() {
+    return Array.from(this.transcriptionResults.values());
+  }
+
+  /**
+   * Force move to the next slice, finalizing the current one regardless of capacity
+   */
+  async nextSlice() {
+    var _this$callbacks$onTra3, _this$callbacks8;
+    if (!this.isActive) {
+      this.log('Cannot force next slice - transcriber is not active');
+      return;
+    }
+
+    // Emit start event to indicate slice processing has started
+    const startEvent = {
+      type: 'start',
+      sliceIndex: -1,
+      // Use -1 to indicate forced slice
+      data: undefined,
+      isCapturing: this.audioStream.isRecording(),
+      processTime: 0,
+      recordingTime: 0,
+      memoryUsage: this.sliceManager.getMemoryUsage()
+    };
+    (_this$callbacks$onTra3 = (_this$callbacks8 = this.callbacks).onTranscribe) === null || _this$callbacks$onTra3 === void 0 ? void 0 : _this$callbacks$onTra3.call(_this$callbacks8, startEvent);
+
+    // Check if there are pending transcriptions or currently transcribing
+    if (this.isTranscribing || this.transcriptionQueue.length > 0) {
+      this.log('Waiting for pending transcriptions to complete before forcing next slice...');
+
+      // Wait for current transcription queue to be processed
+      await this.processTranscriptionQueue();
+    }
+    const result = this.sliceManager.forceNextSlice();
+    if (result.slice) {
+      this.log(`Forced slice ${result.slice.index} ready (${result.slice.data.length} bytes)`);
+
+      // Process VAD for the slice if enabled
+      if (!this.isTranscribing && this.vadEnabled) {
+        this.processSliceVAD(result.slice).catch(error => {
+          this.handleError(`VAD processing error: ${error}`);
+        });
+      } else if (!this.isTranscribing) {
+        // If VAD is disabled, transcribe slices as they become ready
+        this.queueSliceForTranscription(result.slice).catch(error => {
+          this.handleError(`Failed to queue slice for transcription: ${error}`);
+        });
+      } else {
+        this.log(`Skipping slice ${result.slice.index} - already transcribing`);
+      }
+      this.emitStatsUpdate('memory_change');
+    } else {
+      this.log('Forced next slice but no slice data to process');
+    }
+  }
+
+  /**
+   * Reset all components
+   */
+  reset() {
+    this.sliceManager.reset();
+    this.transcriptionQueue = [];
+    this.isTranscribing = false;
+    this.accumulatedData = new Uint8Array(0);
+
+    // Reset simplified VAD state
+    this.lastSpeechDetectedTime = -1;
+    this.lastVadState = 'silence';
+
+    // Reset stats snapshot for clean start
+    this.lastStatsSnapshot = null;
+
+    // Cancel WAV file writing if in progress
+    if (this.wavFileWriter) {
+      this.wavFileWriter.cancel().catch(error => {
+        this.log(`Failed to cancel WAV file writing: ${error}`);
+      });
+      this.wavFileWriter = null;
+    }
+
+    // Clear transcription results
+    this.transcriptionResults.clear();
+
+    // Clear VAD events
+    this.vadEvents.clear();
+  }
+
+  /**
+   * Release all resources
+   */
+  async release() {
+    var _this$wavFileWriter;
+    if (this.isActive) {
+      await this.stop();
+    }
+    await this.audioStream.release();
+    await ((_this$wavFileWriter = this.wavFileWriter) === null || _this$wavFileWriter === void 0 ? void 0 : _this$wavFileWriter.finalize());
+    this.vadContext = undefined;
+  }
+
+  /**
+   * Emit stats update event if stats have changed significantly
+   */
+  emitStatsUpdate(eventType) {
+    const currentStats = this.getStatistics();
+
+    // Check if stats have changed significantly
+    if (!this.lastStatsSnapshot || RealtimeTranscriber.shouldEmitStatsUpdate(currentStats, this.lastStatsSnapshot)) {
+      var _this$callbacks$onSta4, _this$callbacks9;
+      const statsEvent = {
+        timestamp: Date.now(),
+        type: eventType,
+        data: currentStats
+      };
+      (_this$callbacks$onSta4 = (_this$callbacks9 = this.callbacks).onStatsUpdate) === null || _this$callbacks$onSta4 === void 0 ? void 0 : _this$callbacks$onSta4.call(_this$callbacks9, statsEvent);
+      this.lastStatsSnapshot = {
+        ...currentStats
+      };
+    }
+  }
+
+  /**
+   * Determine if stats update should be emitted
+   */
+  static shouldEmitStatsUpdate(current, previous) {
+    var _current$sliceStats, _current$sliceStats$m, _previous$sliceStats, _previous$sliceStats$;
+    // Always emit on status changes
+    if (current.isActive !== previous.isActive || current.isTranscribing !== previous.isTranscribing) {
+      return true;
+    }
+
+    // Emit on significant memory changes (>10% or >5MB)
+    const currentMemory = ((_current$sliceStats = current.sliceStats) === null || _current$sliceStats === void 0 ? void 0 : (_current$sliceStats$m = _current$sliceStats.memoryUsage) === null || _current$sliceStats$m === void 0 ? void 0 : _current$sliceStats$m.estimatedMB) || 0;
+    const previousMemory = ((_previous$sliceStats = previous.sliceStats) === null || _previous$sliceStats === void 0 ? void 0 : (_previous$sliceStats$ = _previous$sliceStats.memoryUsage) === null || _previous$sliceStats$ === void 0 ? void 0 : _previous$sliceStats$.estimatedMB) || 0;
+    const memoryDiff = Math.abs(currentMemory - previousMemory);
+    if (memoryDiff > 5 || previousMemory > 0 && memoryDiff / previousMemory > 0.1) {
+      return true;
+    }
+    return false;
+  }
+
+  /**
+   * Logger function
+   */
+  log(message) {
+    this.options.logger(`[RealtimeTranscriber] ${message}`);
+  }
+}
+exports.RealtimeTranscriber = RealtimeTranscriber;
+//# sourceMappingURL=RealtimeTranscriber.js.map