pediatric-ai-scribe-v3/src/utils/ttsProvider.js

const { getLiteLLMHeaders } = require('./litellm');

function parseList(value) {
  return String(value || '')
    .split(',')
    .map(function(item) { return item.trim(); })
    .filter(Boolean);
}

var KITTEN_TTS_VOICES = ['Bella', 'Jasper', 'Luna', 'Bruno', 'Rosie', 'Hugo', 'Kiki', 'Leo'];
var SUPERTONIC_TTS_VOICES = ['F1', 'F2', 'F3', 'F4', 'F5', 'M1', 'M2', 'M3', 'M4', 'M5'];
var GROQ_ORPHEUS_ENGLISH_VOICES = ['autumn', 'diana', 'hannah', 'austin', 'daniel', 'troy'];
var GROQ_ORPHEUS_ARABIC_VOICES = ['abdullah', 'fahad', 'sultan', 'lulwa', 'noura', 'aisha'];

function uniqueList(values) {
  var seen = new Set();
  return (values || []).filter(function(value) {
    if (!value || seen.has(value)) return false;
    seen.add(value);
    return true;
  });
}

function getTTSProvider() {
  var env = process.env.TTS_PROVIDER;
  if (env === 'litellm') return 'litellm';
  if (process.env.LITELLM_API_BASE) return 'litellm';
  return 'none';
}

function getTTSEnvProvider() {
  return process.env.TTS_PROVIDER || 'auto';
}

function getTTSVoiceLists() {
  return {
    litellm: uniqueList(parseList(process.env.LITELLM_TTS_VOICES).concat(KITTEN_TTS_VOICES, SUPERTONIC_TTS_VOICES, GROQ_ORPHEUS_ENGLISH_VOICES, GROQ_ORPHEUS_ARABIC_VOICES))
  };
}

function getLiteLLMTTSModelFamily(model) {
  var id = String(model || '').toLowerCase();
  if (id === 'local-kitten-tts') return 'kitten';
  if (id === 'local-supertonic-tts') return 'supertonic';
  if (id === 'local-kokoro-tts') return 'kokoro';
  if (id === 'groq-orpheus-english' || id === 'canopylabs/orpheus-v1-english') return 'groq-orpheus-english';
  if (id === 'groq-orpheus-arabic-saudi' || id === 'canopylabs/orpheus-arabic-saudi') return 'groq-orpheus-arabic';
  return 'unknown';
}

function getLiteLLMTTSRequestOptions(model) {
  var family = getLiteLLMTTSModelFamily(model);
  if (family === 'groq-orpheus-english' || family === 'groq-orpheus-arabic') {
    return { response_format: 'wav' };
  }
  return {};
}

function isLiteLLMTTSVoiceCompatible(model, voice) {
  if (!voice) return true;
  var family = getLiteLLMTTSModelFamily(model);
  if (family === 'kitten') return KITTEN_TTS_VOICES.indexOf(voice) !== -1;
  if (family === 'supertonic') return SUPERTONIC_TTS_VOICES.indexOf(voice) !== -1;
  if (family === 'groq-orpheus-english') return GROQ_ORPHEUS_ENGLISH_VOICES.indexOf(String(voice).toLowerCase()) !== -1;
  if (family === 'groq-orpheus-arabic') return GROQ_ORPHEUS_ARABIC_VOICES.indexOf(String(voice).toLowerCase()) !== -1;
  if (family === 'kokoro') {
    return KITTEN_TTS_VOICES.indexOf(voice) === -1 && SUPERTONIC_TTS_VOICES.indexOf(voice) === -1 &&
      GROQ_ORPHEUS_ENGLISH_VOICES.indexOf(String(voice).toLowerCase()) === -1 && GROQ_ORPHEUS_ARABIC_VOICES.indexOf(String(voice).toLowerCase()) === -1;
  }
  return true;
}

function getLiteLLMTTSVoicesForModel(model, opts) {
  opts = opts || {};
  var family = getLiteLLMTTSModelFamily(model);
  var voices = [];
  if (family === 'kitten') voices = KITTEN_TTS_VOICES.slice();
  else if (family === 'supertonic') voices = SUPERTONIC_TTS_VOICES.slice();
  else if (family === 'groq-orpheus-english') voices = GROQ_ORPHEUS_ENGLISH_VOICES.slice();
  else if (family === 'groq-orpheus-arabic') voices = GROQ_ORPHEUS_ARABIC_VOICES.slice();
  else voices = parseList(process.env.LITELLM_TTS_VOICES);

  [opts.currentVoice, process.env.LITELLM_TTS_VOICE].forEach(function(voice) {
    if (isLiteLLMTTSVoiceCompatible(model, voice)) voices.push(voice);
  });
  return uniqueList(voices.filter(function(voice) { return isLiteLLMTTSVoiceCompatible(model, voice); }));
}

function isLiteLLMTTSModel(model) {
  var mode = model && model.model_info && model.model_info.mode ? String(model.model_info.mode) : '';
  return mode === 'audio_speech';
}

function getLiteLLMTTSModels(models) {
  return (models || [])
    .filter(isLiteLLMTTSModel)
    .map(function(model) { return model && (model.id || model.model_name) ? (model.id || model.model_name) : String(model || ''); });
}

function pushUniqueTTSItem(items, item) {
  if (!item || !item.id) return;
  if (items.some(function(existing) { return existing.id === item.id && existing.kind === item.kind; })) return;
  items.push(item);
}

function getLiteLLMTTSDiscoveryItems(models, opts) {
  opts = opts || {};
  var items = [];
  getLiteLLMTTSModels(models).forEach(function(id) {
    pushUniqueTTSItem(items, { id: id, name: id, source: 'gateway-api', kind: 'model' });
  });
  if (opts.currentModel) {
    pushUniqueTTSItem(items, { id: opts.currentModel, name: opts.currentModel, source: 'configured-model', kind: 'model' });
  }
  if (opts.currentVoice) {
    pushUniqueTTSItem(items, { id: opts.currentVoice, name: opts.currentVoice, source: 'configured-voice', kind: 'voice' });
  }
  getTTSVoiceLists().litellm.forEach(function(voice) {
    pushUniqueTTSItem(items, { id: voice, name: voice, source: 'configured-voice-list', kind: 'voice' });
  });
  if (getLiteLLMTTSModels(models).indexOf('local-kitten-tts') !== -1 || opts.currentModel === 'local-kitten-tts') {
    KITTEN_TTS_VOICES.forEach(function(voice) {
      pushUniqueTTSItem(items, { id: voice, name: 'Kitten ' + voice, source: 'local-kitten-tts', kind: 'voice' });
    });
  }
  if (getLiteLLMTTSModels(models).indexOf('local-supertonic-tts') !== -1 || opts.currentModel === 'local-supertonic-tts') {
    SUPERTONIC_TTS_VOICES.forEach(function(voice) {
      pushUniqueTTSItem(items, { id: voice, name: 'Supertonic ' + voice, source: 'local-supertonic-tts', kind: 'voice' });
    });
  }
  if (getLiteLLMTTSModels(models).indexOf('groq-orpheus-english') !== -1 || opts.currentModel === 'groq-orpheus-english') {
    GROQ_ORPHEUS_ENGLISH_VOICES.forEach(function(voice) {
      pushUniqueTTSItem(items, { id: voice, name: 'Groq Orpheus ' + voice, source: 'groq-orpheus-english', kind: 'voice' });
    });
  }
  if (getLiteLLMTTSModels(models).indexOf('groq-orpheus-arabic-saudi') !== -1 || opts.currentModel === 'groq-orpheus-arabic-saudi') {
    GROQ_ORPHEUS_ARABIC_VOICES.forEach(function(voice) {
      pushUniqueTTSItem(items, { id: voice, name: 'Groq Orpheus Arabic ' + voice, source: 'groq-orpheus-arabic-saudi', kind: 'voice' });
    });
  }
  return items;
}

module.exports = {
  getTTSEnvProvider,
  getLiteLLMTTSDiscoveryItems,
  getLiteLLMHeaders,
  getLiteLLMTTSModels,
  getLiteLLMTTSRequestOptions,
  getLiteLLMTTSVoicesForModel,
  getTTSProvider,
  getTTSVoiceLists,
  isLiteLLMTTSVoiceCompatible,
  isLiteLLMTTSModel
};