sionrui/frontend/app/web-gold/src/composables/useTTS.js

/**
 * TTS (Text-to-Speech) 公共Hook
 * 支持多个供应商：CosyVoice, SiliconFlow, Azure, AWS等
 */
import { ref, computed } from 'vue'
import { message } from 'ant-design-vue'
import { VoiceService } from '@/api/voice'
import { normalizeProviderType, VOICE_PROVIDER_TYPES } from '@/config/voiceConfig'

// 兼容旧代码的导出
const TTS_PROVIDERS = VOICE_PROVIDER_TYPES

const DEFAULT_CONFIG =  {
    apiEndpoint: '/api/tik/voice/tts',
    audioFormat: 'mp3',
    supportedFormats: ['mp3', 'wav']
}

export function useTTS(options = {}) {
  const {
    provider = VOICE_PROVIDER_TYPES.SILICONFLOW,
    customConfig = {}
  } = options

  // 状态管理
  const previewAudioCache = new Map()
  const MAX_PREVIEW_CACHE_SIZE = 50
  const previewLoadingVoiceId = ref(null)
  const playingPreviewVoiceId = ref(null)
  const ttsText = ref('')
  const speechRate = ref(1.0)

  // 音频实例
  let previewAudio = null
  let previewObjectUrl = ''

  // 获取当前供应商配置
  const getProviderConfig = () => {
    return DEFAULT_CONFIG
  }

  /**
   * 播放音频预览
   * @param {string} url 音频URL
   * @param {Object} options 播放选项
   */
  const playAudioPreview = (url, options = {}) => {
    if (!url) return message.warning('暂无可试听的音频')

    try {
      previewAudio?.pause?.()
      previewAudio = null
    } catch (error) {
    }

    const audio = new Audio(url)
    const cleanup = () => {
      if (options.revokeOnEnd && url.startsWith('blob:')) {
        URL.revokeObjectURL(url)
        previewObjectUrl === url && (previewObjectUrl = '')
      }
      previewAudio = null
      options.onEnded && options.onEnded()
    }

    audio.play()
      .then(() => {
        previewAudio = audio
        audio.onended = cleanup
        audio.onerror = () => {
          cleanup()
          message.error('播放失败')
        }
      })
      .catch(err => {
        cleanup()
        message.error('播放失败')
      })
  }

  /**
   * 生成预览缓存键
   * @param {Object} voice 音色对象
   * @returns {string} 缓存键
   */
  const generatePreviewCacheKey = (voice) => {
    const voiceId = voice.voiceId || voice.rawId || voice.id
    const text = ttsText.value.trim()
    const rate = speechRate.value
    return `${voiceId}:${text}:${rate}`
  }

  /**
   * 解码并缓存Base64音频
   * @param {string} audioBase64 Base64编码的音频数据
   * @param {string} format 音频格式
   * @param {string} cacheKey 缓存键
   * @returns {Promise<Object>} 音频数据
   */
  const decodeAndCacheBase64 = async (audioBase64, format = 'mp3', cacheKey) => {
    const byteCharacters = window.atob(audioBase64)
    const byteNumbers = new Uint8Array(byteCharacters.length)
    for (let i = 0; i < byteCharacters.length; i++) {
      byteNumbers[i] = byteCharacters.charCodeAt(i)
    }
    const mime = format === 'mp3' ? 'audio/mpeg' : `audio/${format}`
    const blob = new Blob([byteNumbers], { type: mime })
    const objectUrl = URL.createObjectURL(blob)

    const audioData = { blob, objectUrl, format }

    previewAudioCache.set(cacheKey, audioData)

    if (previewAudioCache.size > MAX_PREVIEW_CACHE_SIZE) {
      const firstKey = previewAudioCache.keys().next().value
      const oldData = previewAudioCache.get(firstKey)
      URL.revokeObjectURL(oldData.objectUrl)
      previewAudioCache.delete(firstKey)
    }

    return audioData
  }

  /**
   * 播放缓存的音频
   * @param {Object} audioData 音频数据
   * @param {Function} onEnded 播放结束回调
   */
  const playCachedAudio = (audioData, onEnded) => {
    if (previewObjectUrl && previewObjectUrl !== audioData.objectUrl) {
      URL.revokeObjectURL(previewObjectUrl)
    }
    previewObjectUrl = audioData.objectUrl

    playAudioPreview(previewObjectUrl, {
      revokeOnEnd: false,
      onEnded: () => {
        onEnded && onEnded()
      }
    })
  }

  /**
   * 重置预览状态
   */
  const resetPreviewState = () => {
    previewLoadingVoiceId.value = null
    playingPreviewVoiceId.value = null
  }

  /**
   * 提取ID从字符串
   * @param {string} idStr 包含前缀的ID字符串
   * @returns {number|null} 提取的ID
   */
  const extractIdFromString = (idStr) => {
    if (typeof idStr !== 'string' || !idStr.startsWith('user-')) return null
    const extractedId = parseInt(idStr.replace('user-', ''))
    return Number.isNaN(extractedId) ? null : extractedId
  }

  /**
   * 构建预览参数
   * @param {Object} voice 音色对象
   * @returns {Object|null} 预览参数
   */
  const buildPreviewParams = (voice) => {
    const configId = voice.rawId || extractIdFromString(voice.id)
    if (!configId) {
      message.error('配音配置无效')
      return null
    }
    const providerConfig = getProviderConfig()

    return {
      voiceConfigId: configId,
      inputText: ttsText.value,
      speechRate: speechRate.value || 1.0,
      audioFormat: providerConfig.audioFormat,
      timestamp: Date.now(),
      providerType: normalizeProviderType(provider)
    }
  }

  /**
   * 播放音色试听
   * @param {Object} voice 音色对象
   * @param {Function} onSuccess 成功回调
   * @param {Function} onError 错误回调
   */
  const playVoiceSample = async (voice, onSuccess, onError) => {
    if (!voice) return
    if (previewLoadingVoiceId.value === voice.id || playingPreviewVoiceId.value === voice.id) {
      return
    }
    if (playingPreviewVoiceId.value && playingPreviewVoiceId.value !== voice.id) {
      try {
        previewAudio?.pause?.()
        previewAudio = null
      } catch (error) {
      }
    }

    previewLoadingVoiceId.value = voice.id
    playingPreviewVoiceId.value = voice.id

    const cacheKey = generatePreviewCacheKey(voice)
    const cachedAudio = previewAudioCache.get(cacheKey)

    if (cachedAudio) {
      playCachedAudio(cachedAudio, resetPreviewState)
      onSuccess && onSuccess(cachedAudio)
      return
    }

    try {
      const params = buildPreviewParams(voice)
      if (!params) {
        resetPreviewState()
        onError && onError(new Error('参数构建失败'))
        return
      }

      const res = await VoiceService.preview(params)
      if (res.code !== 0) {
        message.error(res.msg || '试听失败')
        resetPreviewState()
        onError && onError(new Error(res.msg || '试听失败'))
        return
      }

      if (res.data?.audioUrl) {
        playAudioPreview(res.data.audioUrl, { onEnded: resetPreviewState })
        onSuccess && onSuccess(res.data)
      } else if (res.data?.audioBase64) {
        const audioData = await decodeAndCacheBase64(res.data.audioBase64, res.data.format, cacheKey)
        playCachedAudio(audioData, resetPreviewState)
        onSuccess && onSuccess(audioData)
      } else {
        message.error('试听失败')
        resetPreviewState()
        onError && onError(new Error('未收到音频数据'))
      }
    } catch (error) {
      message.error('试听失败')
      resetPreviewState()
      onError && onError(error)
    }
  }

  /**
   * TTS文本转语音
   * @param {Object} params TTS参数
   * @returns {Promise<Object>} TTS结果
   */
  const synthesize = async (params) => {
    const providerConfig = getProviderConfig()

    const ttsParams = {
      inputText: params.inputText || ttsText.value,
      voiceConfigId: params.voiceConfigId,
      speechRate: params.speechRate || speechRate.value,
      audioFormat: params.audioFormat || providerConfig.audioFormat,
      providerType: normalizeProviderType(provider)
    }

    return await VoiceService.synthesize(ttsParams)
  }

  /**
   * 设置文本
   * @param {string} text 要设置的文本
   */
  const setText = (text) => {
    ttsText.value = text
  }

  /**
   * 设置语速
   * @param {number} rate 语速倍率
   */
  const setSpeechRate = (rate) => {
    speechRate.value = rate
  }

  /**
   * 清除音频缓存
   */
  const clearAudioCache = () => {
    previewAudioCache.forEach((audioData) => {
      URL.revokeObjectURL(audioData.objectUrl)
    })
    previewAudioCache.clear()
  }

  /**
   * 停止当前播放
   */
  const stopCurrentPlayback = () => {
    try {
      previewAudio?.pause?.()
      previewAudio = null
    } catch (error) {
    }
  }

  return {
    // 状态
    previewLoadingVoiceId,
    playingPreviewVoiceId,
    ttsText,
    speechRate,

    // 方法
    playVoiceSample,
    synthesize,
    setText,
    setSpeechRate,
    playAudioPreview,
    clearAudioCache,
    stopCurrentPlayback,
    resetPreviewState,

    // 配置
    getProviderConfig,
    TTS_PROVIDERS,
    DEFAULT_CONFIG
  }
}

export { TTS_PROVIDERS }