wellnua-lite-Robert/services/sherpaTTS.ts

/**
 * Sherpa TTS Service - Dynamic model loading
 * Uses react-native-sherpa-onnx-offline-tts with Piper VITS models
 * Models downloaded from Hugging Face on first use
 */

import { NativeEventEmitter } from 'react-native';
import TTSManager from 'react-native-sherpa-onnx-offline-tts';
import RNFS from '@dr.pogodin/react-native-fs';
import { debugLogger } from '@/services/DebugLogger';

// Only Orayan (Ryan) voice - downloaded from Hugging Face
const ORAYAN_VOICE = {
  id: 'ryan-medium',
  name: 'Ryan (Orayan)',
  description: 'Male, clear voice',
  hfRepo: 'csukuangfj/vits-piper-en_US-ryan-medium',
  modelDir: 'vits-piper-en_US-ryan-medium',
  onnxFile: 'en_US-ryan-medium.onnx',
};

interface SherpaTTSState {
  initialized: boolean;
  initializing: boolean;
  speaking: boolean;
  error: string | null;
}

// Check if native module is available
const NATIVE_MODULE_AVAILABLE = !!TTSManager;

if (NATIVE_MODULE_AVAILABLE) {
  debugLogger.info('TTS', 'TTSManager native module loaded successfully');
} else {
  debugLogger.error('TTS', 'TTSManager native module NOT available - prebuild required');
}

let ttsManagerEmitter: NativeEventEmitter | null = null;
if (NATIVE_MODULE_AVAILABLE) {
  ttsManagerEmitter = new NativeEventEmitter(TTSManager);
  debugLogger.info('TTS', 'TTS event emitter initialized');
}

let currentState: SherpaTTSState = {
  initialized: false,
  initializing: false,
  speaking: false,
  error: null,
};

// State listeners
const stateListeners: ((state: SherpaTTSState) => void)[] = [];

function updateState(updates: Partial<SherpaTTSState>) {
  currentState = { ...currentState, ...updates };
  notifyListeners();
}

function notifyListeners() {
  stateListeners.forEach(listener => listener({ ...currentState }));
}

export function addStateListener(listener: (state: SherpaTTSState) => void) {
  stateListeners.push(listener);
  listener({ ...currentState });
  return () => {
    const index = stateListeners.indexOf(listener);
    if (index >= 0) stateListeners.splice(index, 1);
  };
}

export function getState(): SherpaTTSState {
  return { ...currentState };
}

/**
 * Download TTS model from Hugging Face
 */
async function downloadModelFromHuggingFace(): Promise<boolean> {
  const extractPath = `${RNFS.DocumentDirectoryPath}/voices`;
  const modelDir = `${extractPath}/${ORAYAN_VOICE.modelDir}`;
  const modelPath = `${modelDir}/${ORAYAN_VOICE.onnxFile}`;

  // Check if already downloaded
  const exists = await RNFS.exists(modelPath);
  if (exists) {
    debugLogger.info('TTS', `Model already downloaded at: ${modelPath}`);
    return true;
  }

  debugLogger.info('TTS', `Downloading model from Hugging Face: ${ORAYAN_VOICE.hfRepo}`);
  updateState({ initializing: true, error: 'Downloading voice model...' });

  try {
    // Create directories
    await RNFS.mkdir(modelDir, { intermediateDirectories: true });

    // Download model files from Hugging Face
    const baseUrl = `https://huggingface.co/${ORAYAN_VOICE.hfRepo}/resolve/main`;

    const filesToDownload = [
      { url: `${baseUrl}/${ORAYAN_VOICE.onnxFile}`, path: modelPath },
      { url: `${baseUrl}/tokens.txt`, path: `${modelDir}/tokens.txt` },
      { url: `${baseUrl}/espeak-ng-data.tar.bz2`, path: `${modelDir}/espeak-ng-data.tar.bz2` },
    ];

    // Download each file
    for (const file of filesToDownload) {
      debugLogger.log('TTS', `Downloading: ${file.url}`);
      const downloadResult = await RNFS.downloadFile({
        fromUrl: file.url,
        toFile: file.path,
      }).promise;

      if (downloadResult.statusCode !== 200) {
        throw new Error(`Failed to download ${file.url}: ${downloadResult.statusCode}`);
      }
    }

    // Extract espeak-ng-data
    debugLogger.log('TTS', 'Extracting espeak-ng-data...');
    // Note: Extraction would need native module or untar library
    // For now, assume it's extracted manually or via separate process

    debugLogger.info('TTS', '✅ Model downloaded successfully');
    return true;

  } catch (error) {
    const errorMessage = error instanceof Error ? error.message : 'Download failed';
    debugLogger.error('TTS', `Model download failed: ${errorMessage}`, error);
    updateState({ error: errorMessage, initializing: false });
    return false;
  }
}

/**
 * Initialize Sherpa TTS with Orayan voice
 */
export async function initializeSherpaTTS(): Promise<boolean> {
  if (!NATIVE_MODULE_AVAILABLE) {
    debugLogger.error('TTS', 'Cannot initialize - native module not available');
    updateState({
      initialized: false,
      error: 'Native module not available - run npx expo prebuild and rebuild'
    });
    return false;
  }

  if (currentState.initializing) {
    debugLogger.warn('TTS', 'Already initializing - skipping duplicate call');
    return false;
  }

  debugLogger.info('TTS', `Starting initialization with voice: ${ORAYAN_VOICE.name}`);
  updateState({ initializing: true, error: null });

  try {
    // Download model if needed
    const downloaded = await downloadModelFromHuggingFace();
    if (!downloaded) {
      throw new Error('Model download failed');
    }

    // Build paths to model files
    const extractPath = `${RNFS.DocumentDirectoryPath}/voices`;
    const modelDir = `${extractPath}/${ORAYAN_VOICE.modelDir}`;
    const modelPath = `${modelDir}/${ORAYAN_VOICE.onnxFile}`;
    const tokensPath = `${modelDir}/tokens.txt`;
    const dataDirPath = `${modelDir}/espeak-ng-data`;

    debugLogger.log('TTS', 'Model paths:', {
      model: modelPath,
      tokens: tokensPath,
      dataDir: dataDirPath
    });

    // Create config JSON for native module
    const configJSON = JSON.stringify({
      modelPath,
      tokensPath,
      dataDirPath,
    });

    debugLogger.log('TTS', `Calling TTSManager.initialize() with config JSON`);

    // Initialize native TTS
    await TTSManager.initialize(configJSON);

    updateState({
      initialized: true,
      initializing: false,
      error: null
    });

    debugLogger.info('TTS', '✅ Initialization successful');
    return true;

  } catch (error) {
    const errorMessage = error instanceof Error ? error.message : 'Unknown error';
    debugLogger.error('TTS', `Initialization failed: ${errorMessage}`, error);
    updateState({
      initialized: false,
      initializing: false,
      error: errorMessage
    });
    return false;
  }
}

/**
 * Speak text using Sherpa TTS
 */
export async function speak(
  text: string,
  options?: {
    speed?: number;
    speakerId?: number;
    onStart?: () => void;
    onDone?: () => void;
    onError?: (error: Error) => void;
  }
): Promise<void> {
  debugLogger.log('TTS', `speak() called with text: "${text.substring(0, 50)}${text.length > 50 ? '...' : ''}"`);

  if (!NATIVE_MODULE_AVAILABLE || !currentState.initialized) {
    debugLogger.error('TTS', 'Cannot speak - TTS not initialized or module unavailable');
    options?.onError?.(new Error('Sherpa TTS not initialized'));
    return;
  }

  if (!text || text.trim().length === 0) {
    debugLogger.warn('TTS', 'Empty text provided, skipping speech');
    return;
  }

  const speed = options?.speed ?? 1.0;
  const speakerId = options?.speakerId ?? 0;

  debugLogger.log('TTS', `Speech parameters: speed=${speed}, speakerId=${speakerId}`);

  try {
    updateState({ speaking: true });
    debugLogger.info('TTS', 'State updated to speaking=true, calling onStart callback');
    options?.onStart?.();

    debugLogger.log('TTS', `Calling TTSManager.generateAndPlay("${text}", ${speakerId}, ${speed})`);

    await TTSManager.generateAndPlay(text, speakerId, speed);

    debugLogger.info('TTS', '✅ Speech playback completed successfully');
    updateState({ speaking: false });
    options?.onDone?.();
  } catch (error) {
    const err = error instanceof Error ? error : new Error('TTS playback failed');
    debugLogger.error('TTS', `💥 Speech playback error: ${err.message}`, error);
    updateState({ speaking: false });
    options?.onError?.(err);
  }
}

/**
 * Stop current speech playback
 */
export async function stop(): Promise<void> {
  debugLogger.info('TTS', 'stop() called');

  if (NATIVE_MODULE_AVAILABLE && currentState.initialized) {
    try {
      debugLogger.log('TTS', 'Calling TTSManager.deinitialize() to stop playback');
      TTSManager.deinitialize();
      updateState({ speaking: false });

      // Re-initialize after stop to be ready for next speech
      debugLogger.log('TTS', 'Scheduling re-initialization in 100ms');
      setTimeout(() => {
        initializeSherpaTTS();
      }, 100);

      debugLogger.info('TTS', 'Playback stopped successfully');
    } catch (error) {
      debugLogger.error('TTS', 'Failed to stop playback', error);
    }
  } else {
    debugLogger.warn('TTS', 'Cannot stop - module not available or not initialized');
  }
}

/**
 * Deinitialize and free resources
 */
export function deinitialize(): void {
  debugLogger.info('TTS', 'deinitialize() called');

  if (NATIVE_MODULE_AVAILABLE) {
    try {
      debugLogger.log('TTS', 'Calling TTSManager.deinitialize() to free resources');
      TTSManager.deinitialize();
      debugLogger.info('TTS', 'TTS resources freed successfully');
    } catch (error) {
      debugLogger.error('TTS', 'Failed to deinitialize', error);
    }
  }
  updateState({ initialized: false, speaking: false, error: null });
}

/**
 * Check if Sherpa TTS is available (native module loaded)
 */
export function isAvailable(): boolean {
  return NATIVE_MODULE_AVAILABLE && currentState.initialized;
}

/**
 * Check if currently speaking
 */
export async function isSpeaking(): Promise<boolean> {
  return currentState.speaking;
}

// Voice selection removed - only Lessac voice is available

/**
 * Add listener for volume updates during playback
 */
export function addVolumeListener(callback: (volume: number) => void): (() => void) | null {
  if (!ttsManagerEmitter) return null;

  const subscription = ttsManagerEmitter.addListener('VolumeUpdate', (event) => {
    callback(event.volume);
  });

  return () => subscription.remove();
}

export default {
  initialize: initializeSherpaTTS,
  speak,
  stop,
  deinitialize,
  isAvailable,
  isSpeaking,
  addStateListener,
  getState,
  addVolumeListener,
};