@livekit/agents 1.0.6 → 1.1.0

package/dist/inference/tts.cjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/inference/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport { basic as tokenizeBasic } from '../tokenize/index.js';\nimport {\n  SynthesizeStream as BaseSynthesizeStream,\n  TTS as BaseTTS,\n  ChunkedStream,\n} from '../tts/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { shortuuid } from '../utils.js';\nimport {\n  type TtsClientEvent,\n  type TtsServerEvent,\n  type TtsSessionCreateEvent,\n  ttsClientEventSchema,\n  ttsServerEventSchema,\n} from './api_protos.js';\nimport { type AnyModels, connectWs, createAccessToken } from './utils.js';\n\ntype _CartesiaModels = 'cartesia' | 'cartesia/sonic' | 'cartesia/sonic-2' | 'cartesia/sonic-turbo';\n\nexport type CartesiaModels = _CartesiaModels | `${_CartesiaModels}:${string}`;\n\ntype _ElevenlabsModels =\n  | 'elevenlabs'\n  | 'elevenlabs/eleven_flash_v2'\n  | 'elevenlabs/eleven_flash_v2_5'\n  | 'elevenlabs/eleven_turbo_v2'\n  | 'elevenlabs/eleven_turbo_v2_5'\n  | 'elevenlabs/eleven_multilingual_v2';\n\nexport type ElevenlabsModels = _ElevenlabsModels | `${_ElevenlabsModels}:${string}`;\n\nexport type _RimeModels = 'rime' | 'rime/mist' | 'rime/mistv2' | 'rime/arcana';\n\nexport type RimeModels = _RimeModels | `${_RimeModels}:${string}`;\n\nexport type _InworldModels = 'inworld' | 'inworld/inworld-tts-1';\n\nexport type InworldModels = _InworldModels | `${_InworldModels}:${string}`;\n\nexport interface CartesiaOptions {\n  duration?: number; // max duration of audio in seconds\n  speed?: 'slow' | 'normal' | 'fast'; // default: not specified\n}\n\nexport interface ElevenlabsOptions {\n  inactivity_timeout?: number; // default: 60\n  apply_text_normalization?: 'auto' | 'off' | 'on'; // default: \"auto\"\n}\n\nexport interface RimeOptions {}\n\nexport interface InworldOptions {}\n\nexport type TTSModels = CartesiaModels | ElevenlabsModels | RimeModels | InworldModels | AnyModels;\n\nexport type TTSOptions<TModel extends TTSModels> = TModel extends CartesiaModels\n  ? CartesiaOptions\n  : TModel extends ElevenlabsModels\n    ? ElevenlabsOptions\n    : TModel extends RimeOptions\n      ? RimeOptions\n      : TModel extends InworldOptions\n        ? InworldOptions\n        : Record<string, unknown>;\n\ntype TTSEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: TTSEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'https://agent-gateway.livekit.cloud/v1';\nconst NUM_CHANNELS = 1;\nconst DEFAULT_LANGUAGE = 'en';\n\nexport interface InferenceTTSOptions<TModel extends TTSModels> {\n  model?: TModel;\n  voice?: string;\n  language?: string;\n  encoding: TTSEncoding;\n  sampleRate: number;\n  baseURL: string;\n  apiKey: string;\n  apiSecret: string;\n  extraKwargs: TTSOptions<TModel>;\n}\n\nexport class TTS<TModel extends TTSModels> extends BaseTTS {\n  private opts: InferenceTTSOptions<TModel>;\n  private streams: Set<SynthesizeStream<TModel>> = new Set();\n\n  #logger = log();\n\n  constructor(opts: {\n    model: TModel;\n    voice?: string;\n    language?: string;\n    baseURL?: string;\n    encoding?: TTSEncoding;\n    sampleRate?: number;\n    apiKey?: string;\n    apiSecret?: string;\n    extraKwargs?: TTSOptions<TModel>;\n  }) {\n    const sampleRate = opts?.sampleRate ?? DEFAULT_SAMPLE_RATE;\n    super(sampleRate, 1, { streaming: true });\n\n    const {\n      model,\n      voice,\n      language = DEFAULT_LANGUAGE,\n      baseURL,\n      encoding = DEFAULT_ENCODING,\n      apiKey,\n      apiSecret,\n      extraKwargs = {} as TTSOptions<TModel>,\n    } = opts || {};\n\n    const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n    const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n    if (!lkApiKey) {\n      throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n    }\n\n    const lkApiSecret =\n      apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n    if (!lkApiSecret) {\n      throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n    }\n\n    // read voice id from the model if provided: \"provider/model:voice_id\"\n    let nextModel = model;\n    let nextVoice = voice;\n    if (typeof nextModel === 'string') {\n      const idx = nextModel.lastIndexOf(':');\n      if (idx !== -1) {\n        const voiceFromModel = nextModel.slice(idx + 1);\n        if (nextVoice && nextVoice !== voiceFromModel) {\n          this.#logger.warn(\n            '`voice` is provided via both argument and model, using the one from the argument',\n            { voice: nextVoice, model: nextModel },\n          );\n        } else {\n          nextVoice = voiceFromModel;\n        }\n        nextModel = nextModel.slice(0, idx) as TModel;\n      }\n    }\n\n    this.opts = {\n      model: nextModel,\n      voice: nextVoice,\n      language,\n      encoding,\n      sampleRate,\n      baseURL: lkBaseURL,\n      apiKey: lkApiKey,\n      apiSecret: lkApiSecret,\n      extraKwargs,\n    };\n  }\n\n  get label() {\n    return 'inference.TTS';\n  }\n\n  updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n    this.opts = { ...this.opts, ...opts };\n    for (const stream of this.streams) {\n      stream.updateOptions(opts);\n    }\n  }\n\n  synthesize(_: string): ChunkedStream {\n    throw new Error('ChunkedStream is not implemented');\n  }\n\n  stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream<TModel> {\n    const { connOptions = DEFAULT_API_CONNECT_OPTIONS } = options || {};\n    const stream = new SynthesizeStream(this, { ...this.opts }, connOptions);\n    this.streams.add(stream);\n    return stream;\n  }\n\n  async connectWs(timeout: number): Promise<WebSocket> {\n    let baseURL = this.opts.baseURL;\n    if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n      baseURL = baseURL.replace('http', 'ws');\n    }\n\n    const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n    const url = `${baseURL}/tts`;\n    const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n    const params = {\n      type: 'session.create',\n      sample_rate: String(this.opts.sampleRate),\n      encoding: this.opts.encoding,\n      extra: this.opts.extraKwargs,\n    } as TtsSessionCreateEvent;\n\n    if (this.opts.voice) params.voice = this.opts.voice;\n    if (this.opts.model) params.model = this.opts.model;\n    if (this.opts.language) params.language = this.opts.language;\n\n    const socket = await connectWs(url, headers, timeout);\n    socket.send(JSON.stringify(params));\n    return socket;\n  }\n\n  async closeWs(ws: WebSocket) {\n    await ws.close();\n  }\n\n  async close() {\n    for (const stream of this.streams) {\n      await stream.close();\n    }\n    this.streams.clear();\n  }\n}\n\nexport class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeStream {\n  private opts: InferenceTTSOptions<TModel>;\n  private tts: TTS<TModel>;\n  private connOptions: APIConnectOptions;\n\n  #logger = log();\n\n  constructor(tts: TTS<TModel>, opts: InferenceTTSOptions<TModel>, connOptions: APIConnectOptions) {\n    super(tts, connOptions);\n    this.opts = opts;\n    this.tts = tts;\n    this.connOptions = connOptions;\n  }\n\n  get label() {\n    return 'inference.SynthesizeStream';\n  }\n\n  updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n    this.opts = { ...this.opts, ...opts };\n  }\n\n  protected async run(): Promise<void> {\n    let ws: WebSocket | null = null;\n    let closing = false;\n    let finalReceived = false;\n    let lastFrame: AudioFrame | undefined;\n\n    const sendTokenizerStream = new tokenizeBasic.SentenceTokenizer().stream();\n    const eventChannel = createStreamChannel<TtsServerEvent>();\n    const requestId = shortuuid('tts_request_');\n\n    const resourceCleanup = () => {\n      if (closing) return;\n      closing = true;\n      sendTokenizerStream.close();\n      eventChannel.close();\n      ws?.removeAllListeners();\n      ws?.close();\n    };\n\n    const sendClientEvent = async (event: TtsClientEvent) => {\n      const validatedEvent = await ttsClientEventSchema.parseAsync(event);\n      if (!ws || ws.readyState !== WebSocket.OPEN) {\n        this.#logger.warn('Trying to send client TTS event to a closed WebSocket');\n        return;\n      }\n      ws.send(JSON.stringify(validatedEvent));\n    };\n\n    const sendLastFrame = (segmentId: string, final: boolean) => {\n      if (lastFrame) {\n        this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n        lastFrame = undefined;\n      }\n    };\n\n    const createInputTask = async () => {\n      for await (const data of this.input) {\n        if (this.abortController.signal.aborted) break;\n        if (data === SynthesizeStream.FLUSH_SENTINEL) {\n          sendTokenizerStream.flush();\n          continue;\n        }\n        sendTokenizerStream.pushText(data);\n      }\n      sendTokenizerStream.endInput();\n    };\n\n    const createSentenceStreamTask = async () => {\n      for await (const ev of sendTokenizerStream) {\n        if (this.abortController.signal.aborted) break;\n\n        sendClientEvent({\n          type: 'input_transcript',\n          transcript: ev.token + ' ',\n        });\n      }\n\n      sendClientEvent({ type: 'session.flush' });\n    };\n\n    const createWsListenerTask = async (ws: WebSocket) => {\n      return new Promise<void>((resolve, reject) => {\n        this.abortController.signal.addEventListener('abort', () => {\n          resourceCleanup();\n          reject(new Error('WebSocket connection aborted'));\n        });\n\n        ws.on('message', async (data) => {\n          const eventJson = JSON.parse(data.toString()) as Record<string, unknown>;\n          const validatedEvent = ttsServerEventSchema.parse(eventJson);\n          eventChannel.write(validatedEvent);\n        });\n\n        ws.on('error', (e) => {\n          this.#logger.error({ error: e }, 'WebSocket error');\n          resourceCleanup();\n          reject(e);\n        });\n\n        ws.on('close', () => {\n          resourceCleanup();\n\n          if (!closing) return this.#logger.error('WebSocket closed unexpectedly');\n          if (finalReceived) return resolve();\n\n          reject(\n            new APIStatusError({\n              message: 'Gateway connection closed unexpectedly',\n              options: { requestId },\n            }),\n          );\n        });\n      });\n    };\n\n    const createRecvTask = async () => {\n      let currentSessionId: string | null = null;\n\n      const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n      const serverEventStream = eventChannel.stream();\n      const reader = serverEventStream.getReader();\n\n      try {\n        while (!this.closed && !this.abortController.signal.aborted) {\n          const result = await reader.read();\n          if (this.abortController.signal.aborted) return;\n          if (result.done) return;\n\n          const serverEvent = result.value;\n          switch (serverEvent.type) {\n            case 'session.created':\n              currentSessionId = serverEvent.session_id;\n              break;\n            case 'output_audio':\n              const base64Data = new Int8Array(Buffer.from(serverEvent.audio, 'base64'));\n              for (const frame of bstream.write(base64Data.buffer)) {\n                sendLastFrame(currentSessionId!, false);\n                lastFrame = frame;\n              }\n              break;\n            case 'done':\n              finalReceived = true;\n              for (const frame of bstream.flush()) {\n                sendLastFrame(currentSessionId!, false);\n                lastFrame = frame;\n              }\n              sendLastFrame(currentSessionId!, true);\n              this.queue.put(SynthesizeStream.END_OF_STREAM);\n              break;\n            case 'session.closed':\n              resourceCleanup();\n              break;\n            case 'error':\n              this.#logger.error(\n                { serverEvent },\n                'Received error message from LiveKit TTS WebSocket',\n              );\n              resourceCleanup();\n              throw new APIError(`LiveKit TTS returned error: ${serverEvent.message}`);\n            default:\n              this.#logger.warn('Unexpected message %s', serverEvent);\n              break;\n          }\n        }\n      } finally {\n        reader.releaseLock();\n        try {\n          await serverEventStream.cancel();\n        } catch (e) {\n          this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n        }\n      }\n    };\n\n    try {\n      ws = await this.tts.connectWs(this.connOptions.timeoutMs);\n\n      await Promise.all([\n        createInputTask(),\n        createSentenceStreamTask(),\n        createWsListenerTask(ws),\n        createRecvTask(),\n      ]);\n    } catch (e) {\n      this.#logger.error('Error in SynthesizeStream', { error: e });\n    } finally {\n      resourceCleanup();\n    }\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,gBAA0B;AAC1B,wBAAyC;AACzC,mBAAgC;AAChC,iBAAoB;AACpB,4BAAoC;AACpC,sBAAuC;AACvC,iBAIO;AACP,mBAAoE;AACpE,mBAA0B;AAC1B,wBAMO;AACP,IAAAA,gBAA6D;AAoD7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,eAAe;AACrB,MAAM,mBAAmB;AAclB,MAAM,YAAsC,WAAAC,IAAQ;AAAA,EACjD;AAAA,EACA,UAAyC,oBAAI,IAAI;AAAA,EAEzD,cAAU,gBAAI;AAAA,EAEd,YAAY,MAUT;AACD,UAAM,cAAa,6BAAM,eAAc;AACvC,UAAM,YAAY,GAAG,EAAE,WAAW,KAAK,CAAC;AAExC,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA;AAAA,MACA,cAAc,CAAC;AAAA,IACjB,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,WAAW,QAAQ,IAAI,yBAAyB;AAClE,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAGA,QAAI,YAAY;AAChB,QAAI,YAAY;AAChB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,iBAAiB,UAAU,MAAM,MAAM,CAAC;AAC9C,YAAI,aAAa,cAAc,gBAAgB;AAC7C,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,OAAO,WAAW,OAAO,UAAU;AAAA,UACvC;AAAA,QACF,OAAO;AACL,sBAAY;AAAA,QACd;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AAEA,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,OAAO;AAAA,MACP;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,WAAW,GAA0B;AACnC,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AAAA,EAEA,OAAO,SAAyE;AAC9E,UAAM,EAAE,cAAc,yCAA4B,IAAI,WAAW,CAAC;AAClE,UAAM,SAAS,IAAI,iBAAiB,MAAM,EAAE,GAAG,KAAK,KAAK,GAAG,WAAW;AACvE,SAAK,QAAQ,IAAI,MAAM;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AACnD,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,UAAM,iCAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS;AAAA,MACb,MAAM;AAAA,MACN,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,MACxC,UAAU,KAAK,KAAK;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,IACnB;AAEA,QAAI,KAAK,KAAK,MAAO,QAAO,QAAQ,KAAK,KAAK;AAC9C,QAAI,KAAK,KAAK,MAAO,QAAO,QAAQ,KAAK,KAAK;AAC9C,QAAI,KAAK,KAAK,SAAU,QAAO,WAAW,KAAK,KAAK;AAEpD,UAAM,SAAS,UAAM,yBAAU,KAAK,SAAS,OAAO;AACpD,WAAO,KAAK,KAAK,UAAU,MAAM,CAAC;AAClC,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,QAAQ,IAAe;AAC3B,UAAM,GAAG,MAAM;AAAA,EACjB;AAAA,EAEA,MAAM,QAAQ;AACZ,eAAW,UAAU,KAAK,SAAS;AACjC,YAAM,OAAO,MAAM;AAAA,IACrB;AACA,SAAK,QAAQ,MAAM;AAAA,EACrB;AACF;AAEO,MAAM,yBAAmD,WAAAC,iBAAqB;AAAA,EAC3E;AAAA,EACA;AAAA,EACA;AAAA,EAER,cAAU,gBAAI;AAAA,EAEd,YAAY,KAAkB,MAAmC,aAAgC;AAC/F,UAAM,KAAK,WAAW;AACtB,SAAK,OAAO;AACZ,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAAA,EACtC;AAAA,EAEA,MAAgB,MAAqB;AACnC,QAAI,KAAuB;AAC3B,QAAI,UAAU;AACd,QAAI,gBAAgB;AACpB,QAAI;AAEJ,UAAM,sBAAsB,IAAI,gBAAAC,MAAc,kBAAkB,EAAE,OAAO;AACzE,UAAM,mBAAe,2CAAoC;AACzD,UAAM,gBAAY,wBAAU,cAAc;AAE1C,UAAM,kBAAkB,MAAM;AAC5B,UAAI,QAAS;AACb,gBAAU;AACV,0BAAoB,MAAM;AAC1B,mBAAa,MAAM;AACnB,+BAAI;AACJ,+BAAI;AAAA,IACN;AAEA,UAAM,kBAAkB,OAAO,UAA0B;AACvD,YAAM,iBAAiB,MAAM,uCAAqB,WAAW,KAAK;AAClE,UAAI,CAAC,MAAM,GAAG,eAAe,oBAAU,MAAM;AAC3C,aAAK,QAAQ,KAAK,uDAAuD;AACzE;AAAA,MACF;AACA,SAAG,KAAK,KAAK,UAAU,cAAc,CAAC;AAAA,IACxC;AAEA,UAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,UAAI,WAAW;AACb,aAAK,MAAM,IAAI,EAAE,WAAW,WAAW,OAAO,WAAW,MAAM,CAAC;AAChE,oBAAY;AAAA,MACd;AAAA,IACF;AAEA,UAAM,kBAAkB,YAAY;AAClC,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,8BAAoB,MAAM;AAC1B;AAAA,QACF;AACA,4BAAoB,SAAS,IAAI;AAAA,MACnC;AACA,0BAAoB,SAAS;AAAA,IAC/B;AAEA,UAAM,2BAA2B,YAAY;AAC3C,uBAAiB,MAAM,qBAAqB;AAC1C,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,wBAAgB;AAAA,UACd,MAAM;AAAA,UACN,YAAY,GAAG,QAAQ;AAAA,QACzB,CAAC;AAAA,MACH;AAEA,sBAAgB,EAAE,MAAM,gBAAgB,CAAC;AAAA,IAC3C;AAEA,UAAM,uBAAuB,OAAOC,QAAkB;AACpD,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,aAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,0BAAgB;AAChB,iBAAO,IAAI,MAAM,8BAA8B,CAAC;AAAA,QAClD,CAAC;AAED,QAAAA,IAAG,GAAG,WAAW,OAAO,SAAS;AAC/B,gBAAM,YAAY,KAAK,MAAM,KAAK,SAAS,CAAC;AAC5C,gBAAM,iBAAiB,uCAAqB,MAAM,SAAS;AAC3D,uBAAa,MAAM,cAAc;AAAA,QACnC,CAAC;AAED,QAAAA,IAAG,GAAG,SAAS,CAAC,MAAM;AACpB,eAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,0BAAgB;AAChB,iBAAO,CAAC;AAAA,QACV,CAAC;AAED,QAAAA,IAAG,GAAG,SAAS,MAAM;AACnB,0BAAgB;AAEhB,cAAI,CAAC,QAAS,QAAO,KAAK,QAAQ,MAAM,+BAA+B;AACvE,cAAI,cAAe,QAAO,QAAQ;AAElC;AAAA,YACE,IAAI,iCAAe;AAAA,cACjB,SAAS;AAAA,cACT,SAAS,EAAE,UAAU;AAAA,YACvB,CAAC;AAAA,UACH;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,UAAM,iBAAiB,YAAY;AACjC,UAAI,mBAAkC;AAEtC,YAAM,UAAU,IAAI,6BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,YAAM,oBAAoB,aAAa,OAAO;AAC9C,YAAM,SAAS,kBAAkB,UAAU;AAE3C,UAAI;AACF,eAAO,CAAC,KAAK,UAAU,CAAC,KAAK,gBAAgB,OAAO,SAAS;AAC3D,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,cAAI,OAAO,KAAM;AAEjB,gBAAM,cAAc,OAAO;AAC3B,kBAAQ,YAAY,MAAM;AAAA,YACxB,KAAK;AACH,iCAAmB,YAAY;AAC/B;AAAA,YACF,KAAK;AACH,oBAAM,aAAa,IAAI,UAAU,OAAO,KAAK,YAAY,OAAO,QAAQ,CAAC;AACzE,yBAAW,SAAS,QAAQ,MAAM,WAAW,MAAM,GAAG;AACpD,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA;AAAA,YACF,KAAK;AACH,8BAAgB;AAChB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA,4BAAc,kBAAmB,IAAI;AACrC,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAC7C;AAAA,YACF,KAAK;AACH,8BAAgB;AAChB;AAAA,YACF,KAAK;AACH,mBAAK,QAAQ;AAAA,gBACX,EAAE,YAAY;AAAA,gBACd;AAAA,cACF;AACA,8BAAgB;AAChB,oBAAM,IAAI,2BAAS,+BAA+B,YAAY,OAAO,EAAE;AAAA,YACzE;AACE,mBAAK,QAAQ,KAAK,yBAAyB,WAAW;AACtD;AAAA,UACJ;AAAA,QACF;AAAA,MACF,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,kBAAkB,OAAO;AAAA,QACjC,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,QACxF;AAAA,MACF;AAAA,IACF;AAEA,QAAI;AACF,WAAK,MAAM,KAAK,IAAI,UAAU,KAAK,YAAY,SAAS;AAExD,YAAM,QAAQ,IAAI;AAAA,QAChB,gBAAgB;AAAA,QAChB,yBAAyB;AAAA,QACzB,qBAAqB,EAAE;AAAA,QACvB,eAAe;AAAA,MACjB,CAAC;AAAA,IACH,SAAS,GAAG;AACV,WAAK,QAAQ,MAAM,6BAA6B,EAAE,OAAO,EAAE,CAAC;AAAA,IAC9D,UAAE;AACA,sBAAgB;AAAA,IAClB;AAAA,EACF;AACF;","names":["import_utils","BaseTTS","BaseSynthesizeStream","tokenizeBasic","ws"]}

package/dist/inference/tts.d.cts

@@ -0,0 +1,75 @@
+import { WebSocket } from 'ws';
+import { SynthesizeStream as BaseSynthesizeStream, TTS as BaseTTS, ChunkedStream } from '../tts/index.js';
+import { type APIConnectOptions } from '../types.js';
+import { type AnyModels } from './utils.js';
+type _CartesiaModels = 'cartesia' | 'cartesia/sonic' | 'cartesia/sonic-2' | 'cartesia/sonic-turbo';
+export type CartesiaModels = _CartesiaModels | `${_CartesiaModels}:${string}`;
+type _ElevenlabsModels = 'elevenlabs' | 'elevenlabs/eleven_flash_v2' | 'elevenlabs/eleven_flash_v2_5' | 'elevenlabs/eleven_turbo_v2' | 'elevenlabs/eleven_turbo_v2_5' | 'elevenlabs/eleven_multilingual_v2';
+export type ElevenlabsModels = _ElevenlabsModels | `${_ElevenlabsModels}:${string}`;
+export type _RimeModels = 'rime' | 'rime/mist' | 'rime/mistv2' | 'rime/arcana';
+export type RimeModels = _RimeModels | `${_RimeModels}:${string}`;
+export type _InworldModels = 'inworld' | 'inworld/inworld-tts-1';
+export type InworldModels = _InworldModels | `${_InworldModels}:${string}`;
+export interface CartesiaOptions {
+    duration?: number;
+    speed?: 'slow' | 'normal' | 'fast';
+}
+export interface ElevenlabsOptions {
+    inactivity_timeout?: number;
+    apply_text_normalization?: 'auto' | 'off' | 'on';
+}
+export interface RimeOptions {
+}
+export interface InworldOptions {
+}
+export type TTSModels = CartesiaModels | ElevenlabsModels | RimeModels | InworldModels | AnyModels;
+export type TTSOptions<TModel extends TTSModels> = TModel extends CartesiaModels ? CartesiaOptions : TModel extends ElevenlabsModels ? ElevenlabsOptions : TModel extends RimeOptions ? RimeOptions : TModel extends InworldOptions ? InworldOptions : Record<string, unknown>;
+type TTSEncoding = 'pcm_s16le';
+export interface InferenceTTSOptions<TModel extends TTSModels> {
+    model?: TModel;
+    voice?: string;
+    language?: string;
+    encoding: TTSEncoding;
+    sampleRate: number;
+    baseURL: string;
+    apiKey: string;
+    apiSecret: string;
+    extraKwargs: TTSOptions<TModel>;
+}
+export declare class TTS<TModel extends TTSModels> extends BaseTTS {
+    #private;
+    private opts;
+    private streams;
+    constructor(opts: {
+        model: TModel;
+        voice?: string;
+        language?: string;
+        baseURL?: string;
+        encoding?: TTSEncoding;
+        sampleRate?: number;
+        apiKey?: string;
+        apiSecret?: string;
+        extraKwargs?: TTSOptions<TModel>;
+    });
+    get label(): string;
+    updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>): void;
+    synthesize(_: string): ChunkedStream;
+    stream(options?: {
+        connOptions?: APIConnectOptions;
+    }): SynthesizeStream<TModel>;
+    connectWs(timeout: number): Promise<WebSocket>;
+    closeWs(ws: WebSocket): Promise<void>;
+    close(): Promise<void>;
+}
+export declare class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeStream {
+    #private;
+    private opts;
+    private tts;
+    private connOptions;
+    constructor(tts: TTS<TModel>, opts: InferenceTTSOptions<TModel>, connOptions: APIConnectOptions);
+    get label(): string;
+    updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>): void;
+    protected run(): Promise<void>;
+}
+export {};
+//# sourceMappingURL=tts.d.ts.map

package/dist/inference/tts.d.ts

@@ -0,0 +1,75 @@
+import { WebSocket } from 'ws';
+import { SynthesizeStream as BaseSynthesizeStream, TTS as BaseTTS, ChunkedStream } from '../tts/index.js';
+import { type APIConnectOptions } from '../types.js';
+import { type AnyModels } from './utils.js';
+type _CartesiaModels = 'cartesia' | 'cartesia/sonic' | 'cartesia/sonic-2' | 'cartesia/sonic-turbo';
+export type CartesiaModels = _CartesiaModels | `${_CartesiaModels}:${string}`;
+type _ElevenlabsModels = 'elevenlabs' | 'elevenlabs/eleven_flash_v2' | 'elevenlabs/eleven_flash_v2_5' | 'elevenlabs/eleven_turbo_v2' | 'elevenlabs/eleven_turbo_v2_5' | 'elevenlabs/eleven_multilingual_v2';
+export type ElevenlabsModels = _ElevenlabsModels | `${_ElevenlabsModels}:${string}`;
+export type _RimeModels = 'rime' | 'rime/mist' | 'rime/mistv2' | 'rime/arcana';
+export type RimeModels = _RimeModels | `${_RimeModels}:${string}`;
+export type _InworldModels = 'inworld' | 'inworld/inworld-tts-1';
+export type InworldModels = _InworldModels | `${_InworldModels}:${string}`;
+export interface CartesiaOptions {
+    duration?: number;
+    speed?: 'slow' | 'normal' | 'fast';
+}
+export interface ElevenlabsOptions {
+    inactivity_timeout?: number;
+    apply_text_normalization?: 'auto' | 'off' | 'on';
+}
+export interface RimeOptions {
+}
+export interface InworldOptions {
+}
+export type TTSModels = CartesiaModels | ElevenlabsModels | RimeModels | InworldModels | AnyModels;
+export type TTSOptions<TModel extends TTSModels> = TModel extends CartesiaModels ? CartesiaOptions : TModel extends ElevenlabsModels ? ElevenlabsOptions : TModel extends RimeOptions ? RimeOptions : TModel extends InworldOptions ? InworldOptions : Record<string, unknown>;
+type TTSEncoding = 'pcm_s16le';
+export interface InferenceTTSOptions<TModel extends TTSModels> {
+    model?: TModel;
+    voice?: string;
+    language?: string;
+    encoding: TTSEncoding;
+    sampleRate: number;
+    baseURL: string;
+    apiKey: string;
+    apiSecret: string;
+    extraKwargs: TTSOptions<TModel>;
+}
+export declare class TTS<TModel extends TTSModels> extends BaseTTS {
+    #private;
+    private opts;
+    private streams;
+    constructor(opts: {
+        model: TModel;
+        voice?: string;
+        language?: string;
+        baseURL?: string;
+        encoding?: TTSEncoding;
+        sampleRate?: number;
+        apiKey?: string;
+        apiSecret?: string;
+        extraKwargs?: TTSOptions<TModel>;
+    });
+    get label(): string;
+    updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>): void;
+    synthesize(_: string): ChunkedStream;
+    stream(options?: {
+        connOptions?: APIConnectOptions;
+    }): SynthesizeStream<TModel>;
+    connectWs(timeout: number): Promise<WebSocket>;
+    closeWs(ws: WebSocket): Promise<void>;
+    close(): Promise<void>;
+}
+export declare class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeStream {
+    #private;
+    private opts;
+    private tts;
+    private connOptions;
+    constructor(tts: TTS<TModel>, opts: InferenceTTSOptions<TModel>, connOptions: APIConnectOptions);
+    get label(): string;
+    updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>): void;
+    protected run(): Promise<void>;
+}
+export {};
+//# sourceMappingURL=tts.d.ts.map

package/dist/inference/tts.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../../src/inference/tts.ts"],"names":[],"mappings":"AAIA,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAM/B,OAAO,EACL,gBAAgB,IAAI,oBAAoB,EACxC,GAAG,IAAI,OAAO,EACd,aAAa,EACd,MAAM,iBAAiB,CAAC;AACzB,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AASlF,OAAO,EAAE,KAAK,SAAS,EAAgC,MAAM,YAAY,CAAC;AAE1E,KAAK,eAAe,GAAG,UAAU,GAAG,gBAAgB,GAAG,kBAAkB,GAAG,sBAAsB,CAAC;AAEnG,MAAM,MAAM,cAAc,GAAG,eAAe,GAAG,GAAG,eAAe,IAAI,MAAM,EAAE,CAAC;AAE9E,KAAK,iBAAiB,GAClB,YAAY,GACZ,4BAA4B,GAC5B,8BAA8B,GAC9B,4BAA4B,GAC5B,8BAA8B,GAC9B,mCAAmC,CAAC;AAExC,MAAM,MAAM,gBAAgB,GAAG,iBAAiB,GAAG,GAAG,iBAAiB,IAAI,MAAM,EAAE,CAAC;AAEpF,MAAM,MAAM,WAAW,GAAG,MAAM,GAAG,WAAW,GAAG,aAAa,GAAG,aAAa,CAAC;AAE/E,MAAM,MAAM,UAAU,GAAG,WAAW,GAAG,GAAG,WAAW,IAAI,MAAM,EAAE,CAAC;AAElE,MAAM,MAAM,cAAc,GAAG,SAAS,GAAG,uBAAuB,CAAC;AAEjE,MAAM,MAAM,aAAa,GAAG,cAAc,GAAG,GAAG,cAAc,IAAI,MAAM,EAAE,CAAC;AAE3E,MAAM,WAAW,eAAe;IAC9B,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,KAAK,CAAC,EAAE,MAAM,GAAG,QAAQ,GAAG,MAAM,CAAC;CACpC;AAED,MAAM,WAAW,iBAAiB;IAChC,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,wBAAwB,CAAC,EAAE,MAAM,GAAG,KAAK,GAAG,IAAI,CAAC;CAClD;AAED,MAAM,WAAW,WAAW;CAAG;AAE/B,MAAM,WAAW,cAAc;CAAG;AAElC,MAAM,MAAM,SAAS,GAAG,cAAc,GAAG,gBAAgB,GAAG,UAAU,GAAG,aAAa,GAAG,SAAS,CAAC;AAEnG,MAAM,MAAM,UAAU,CAAC,MAAM,SAAS,SAAS,IAAI,MAAM,SAAS,cAAc,GAC5E,eAAe,GACf,MAAM,SAAS,gBAAgB,GAC7B,iBAAiB,GACjB,MAAM,SAAS,WAAW,GACxB,WAAW,GACX,MAAM,SAAS,cAAc,GAC3B,cAAc,GACd,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAElC,KAAK,WAAW,GAAG,WAAW,CAAC;AAQ/B,MAAM,WAAW,mBAAmB,CAAC,MAAM,SAAS,SAAS;IAC3D,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,WAAW,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;CACjC;AAED,qBAAa,GAAG,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,OAAO;;IACxD,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,OAAO,CAA4C;gBAI/C,IAAI,EAAE;QAChB,KAAK,EAAE,MAAM,CAAC;QACd,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,QAAQ,CAAC,EAAE,MAAM,CAAC;QAClB,OAAO,CAAC,EAAE,MAAM,CAAC;QACjB,QAAQ,CAAC,EAAE,WAAW,CAAC;QACvB,UAAU,CAAC,EAAE,MAAM,CAAC;QACpB,MAAM,CAAC,EAAE,MAAM,CAAC;QAChB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,WAAW,CAAC,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;KAClC;IA2DD,IAAI,KAAK,WAER;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,OAAO,GAAG,UAAU,CAAC,CAAC;IAO9F,UAAU,CAAC,CAAC,EAAE,MAAM,GAAG,aAAa;IAIpC,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,gBAAgB,CAAC,MAAM,CAAC;IAOzE,SAAS,CAAC,OAAO,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC;IA0B9C,OAAO,CAAC,EAAE,EAAE,SAAS;IAIrB,KAAK;CAMZ;AAED,qBAAa,gBAAgB,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,oBAAoB;;IAClF,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,GAAG,CAAc;IACzB,OAAO,CAAC,WAAW,CAAoB;gBAI3B,GAAG,EAAE,GAAG,CAAC,MAAM,CAAC,EAAE,IAAI,EAAE,mBAAmB,CAAC,MAAM,CAAC,EAAE,WAAW,EAAE,iBAAiB;IAO/F,IAAI,KAAK,WAER;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,OAAO,GAAG,UAAU,CAAC,CAAC;cAI9E,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;CAyKrC"}

package/dist/inference/tts.js

@@ -0,0 +1,299 @@
+import { WebSocket } from "ws";
+import { APIError, APIStatusError } from "../_exceptions.js";
+import { AudioByteStream } from "../audio.js";
+import { log } from "../log.js";
+import { createStreamChannel } from "../stream/stream_channel.js";
+import { basic as tokenizeBasic } from "../tokenize/index.js";
+import {
+  SynthesizeStream as BaseSynthesizeStream,
+  TTS as BaseTTS,
+  ChunkedStream
+} from "../tts/index.js";
+import { DEFAULT_API_CONNECT_OPTIONS } from "../types.js";
+import { shortuuid } from "../utils.js";
+import {
+  ttsClientEventSchema,
+  ttsServerEventSchema
+} from "./api_protos.js";
+import { connectWs, createAccessToken } from "./utils.js";
+const DEFAULT_ENCODING = "pcm_s16le";
+const DEFAULT_SAMPLE_RATE = 16e3;
+const DEFAULT_BASE_URL = "https://agent-gateway.livekit.cloud/v1";
+const NUM_CHANNELS = 1;
+const DEFAULT_LANGUAGE = "en";
+class TTS extends BaseTTS {
+  opts;
+  streams = /* @__PURE__ */ new Set();
+  #logger = log();
+  constructor(opts) {
+    const sampleRate = (opts == null ? void 0 : opts.sampleRate) ?? DEFAULT_SAMPLE_RATE;
+    super(sampleRate, 1, { streaming: true });
+    const {
+      model,
+      voice,
+      language = DEFAULT_LANGUAGE,
+      baseURL,
+      encoding = DEFAULT_ENCODING,
+      apiKey,
+      apiSecret,
+      extraKwargs = {}
+    } = opts || {};
+    const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;
+    const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;
+    if (!lkApiKey) {
+      throw new Error("apiKey is required: pass apiKey or set LIVEKIT_API_KEY");
+    }
+    const lkApiSecret = apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;
+    if (!lkApiSecret) {
+      throw new Error("apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET");
+    }
+    let nextModel = model;
+    let nextVoice = voice;
+    if (typeof nextModel === "string") {
+      const idx = nextModel.lastIndexOf(":");
+      if (idx !== -1) {
+        const voiceFromModel = nextModel.slice(idx + 1);
+        if (nextVoice && nextVoice !== voiceFromModel) {
+          this.#logger.warn(
+            "`voice` is provided via both argument and model, using the one from the argument",
+            { voice: nextVoice, model: nextModel }
+          );
+        } else {
+          nextVoice = voiceFromModel;
+        }
+        nextModel = nextModel.slice(0, idx);
+      }
+    }
+    this.opts = {
+      model: nextModel,
+      voice: nextVoice,
+      language,
+      encoding,
+      sampleRate,
+      baseURL: lkBaseURL,
+      apiKey: lkApiKey,
+      apiSecret: lkApiSecret,
+      extraKwargs
+    };
+  }
+  get label() {
+    return "inference.TTS";
+  }
+  updateOptions(opts) {
+    this.opts = { ...this.opts, ...opts };
+    for (const stream of this.streams) {
+      stream.updateOptions(opts);
+    }
+  }
+  synthesize(_) {
+    throw new Error("ChunkedStream is not implemented");
+  }
+  stream(options) {
+    const { connOptions = DEFAULT_API_CONNECT_OPTIONS } = options || {};
+    const stream = new SynthesizeStream(this, { ...this.opts }, connOptions);
+    this.streams.add(stream);
+    return stream;
+  }
+  async connectWs(timeout) {
+    let baseURL = this.opts.baseURL;
+    if (baseURL.startsWith("http://") || baseURL.startsWith("https://")) {
+      baseURL = baseURL.replace("http", "ws");
+    }
+    const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);
+    const url = `${baseURL}/tts`;
+    const headers = { Authorization: `Bearer ${token}` };
+    const params = {
+      type: "session.create",
+      sample_rate: String(this.opts.sampleRate),
+      encoding: this.opts.encoding,
+      extra: this.opts.extraKwargs
+    };
+    if (this.opts.voice) params.voice = this.opts.voice;
+    if (this.opts.model) params.model = this.opts.model;
+    if (this.opts.language) params.language = this.opts.language;
+    const socket = await connectWs(url, headers, timeout);
+    socket.send(JSON.stringify(params));
+    return socket;
+  }
+  async closeWs(ws) {
+    await ws.close();
+  }
+  async close() {
+    for (const stream of this.streams) {
+      await stream.close();
+    }
+    this.streams.clear();
+  }
+}
+class SynthesizeStream extends BaseSynthesizeStream {
+  opts;
+  tts;
+  connOptions;
+  #logger = log();
+  constructor(tts, opts, connOptions) {
+    super(tts, connOptions);
+    this.opts = opts;
+    this.tts = tts;
+    this.connOptions = connOptions;
+  }
+  get label() {
+    return "inference.SynthesizeStream";
+  }
+  updateOptions(opts) {
+    this.opts = { ...this.opts, ...opts };
+  }
+  async run() {
+    let ws = null;
+    let closing = false;
+    let finalReceived = false;
+    let lastFrame;
+    const sendTokenizerStream = new tokenizeBasic.SentenceTokenizer().stream();
+    const eventChannel = createStreamChannel();
+    const requestId = shortuuid("tts_request_");
+    const resourceCleanup = () => {
+      if (closing) return;
+      closing = true;
+      sendTokenizerStream.close();
+      eventChannel.close();
+      ws == null ? void 0 : ws.removeAllListeners();
+      ws == null ? void 0 : ws.close();
+    };
+    const sendClientEvent = async (event) => {
+      const validatedEvent = await ttsClientEventSchema.parseAsync(event);
+      if (!ws || ws.readyState !== WebSocket.OPEN) {
+        this.#logger.warn("Trying to send client TTS event to a closed WebSocket");
+        return;
+      }
+      ws.send(JSON.stringify(validatedEvent));
+    };
+    const sendLastFrame = (segmentId, final) => {
+      if (lastFrame) {
+        this.queue.put({ requestId, segmentId, frame: lastFrame, final });
+        lastFrame = void 0;
+      }
+    };
+    const createInputTask = async () => {
+      for await (const data of this.input) {
+        if (this.abortController.signal.aborted) break;
+        if (data === SynthesizeStream.FLUSH_SENTINEL) {
+          sendTokenizerStream.flush();
+          continue;
+        }
+        sendTokenizerStream.pushText(data);
+      }
+      sendTokenizerStream.endInput();
+    };
+    const createSentenceStreamTask = async () => {
+      for await (const ev of sendTokenizerStream) {
+        if (this.abortController.signal.aborted) break;
+        sendClientEvent({
+          type: "input_transcript",
+          transcript: ev.token + " "
+        });
+      }
+      sendClientEvent({ type: "session.flush" });
+    };
+    const createWsListenerTask = async (ws2) => {
+      return new Promise((resolve, reject) => {
+        this.abortController.signal.addEventListener("abort", () => {
+          resourceCleanup();
+          reject(new Error("WebSocket connection aborted"));
+        });
+        ws2.on("message", async (data) => {
+          const eventJson = JSON.parse(data.toString());
+          const validatedEvent = ttsServerEventSchema.parse(eventJson);
+          eventChannel.write(validatedEvent);
+        });
+        ws2.on("error", (e) => {
+          this.#logger.error({ error: e }, "WebSocket error");
+          resourceCleanup();
+          reject(e);
+        });
+        ws2.on("close", () => {
+          resourceCleanup();
+          if (!closing) return this.#logger.error("WebSocket closed unexpectedly");
+          if (finalReceived) return resolve();
+          reject(
+            new APIStatusError({
+              message: "Gateway connection closed unexpectedly",
+              options: { requestId }
+            })
+          );
+        });
+      });
+    };
+    const createRecvTask = async () => {
+      let currentSessionId = null;
+      const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);
+      const serverEventStream = eventChannel.stream();
+      const reader = serverEventStream.getReader();
+      try {
+        while (!this.closed && !this.abortController.signal.aborted) {
+          const result = await reader.read();
+          if (this.abortController.signal.aborted) return;
+          if (result.done) return;
+          const serverEvent = result.value;
+          switch (serverEvent.type) {
+            case "session.created":
+              currentSessionId = serverEvent.session_id;
+              break;
+            case "output_audio":
+              const base64Data = new Int8Array(Buffer.from(serverEvent.audio, "base64"));
+              for (const frame of bstream.write(base64Data.buffer)) {
+                sendLastFrame(currentSessionId, false);
+                lastFrame = frame;
+              }
+              break;
+            case "done":
+              finalReceived = true;
+              for (const frame of bstream.flush()) {
+                sendLastFrame(currentSessionId, false);
+                lastFrame = frame;
+              }
+              sendLastFrame(currentSessionId, true);
+              this.queue.put(SynthesizeStream.END_OF_STREAM);
+              break;
+            case "session.closed":
+              resourceCleanup();
+              break;
+            case "error":
+              this.#logger.error(
+                { serverEvent },
+                "Received error message from LiveKit TTS WebSocket"
+              );
+              resourceCleanup();
+              throw new APIError(`LiveKit TTS returned error: ${serverEvent.message}`);
+            default:
+              this.#logger.warn("Unexpected message %s", serverEvent);
+              break;
+          }
+        }
+      } finally {
+        reader.releaseLock();
+        try {
+          await serverEventStream.cancel();
+        } catch (e) {
+          this.#logger.debug("Error cancelling serverEventStream (may already be cancelled):", e);
+        }
+      }
+    };
+    try {
+      ws = await this.tts.connectWs(this.connOptions.timeoutMs);
+      await Promise.all([
+        createInputTask(),
+        createSentenceStreamTask(),
+        createWsListenerTask(ws),
+        createRecvTask()
+      ]);
+    } catch (e) {
+      this.#logger.error("Error in SynthesizeStream", { error: e });
+    } finally {
+      resourceCleanup();
+    }
+  }
+}
+export {
+  SynthesizeStream,
+  TTS
+};
+//# sourceMappingURL=tts.js.map

package/dist/inference/tts.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/inference/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport { basic as tokenizeBasic } from '../tokenize/index.js';\nimport {\n  SynthesizeStream as BaseSynthesizeStream,\n  TTS as BaseTTS,\n  ChunkedStream,\n} from '../tts/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { shortuuid } from '../utils.js';\nimport {\n  type TtsClientEvent,\n  type TtsServerEvent,\n  type TtsSessionCreateEvent,\n  ttsClientEventSchema,\n  ttsServerEventSchema,\n} from './api_protos.js';\nimport { type AnyModels, connectWs, createAccessToken } from './utils.js';\n\ntype _CartesiaModels = 'cartesia' | 'cartesia/sonic' | 'cartesia/sonic-2' | 'cartesia/sonic-turbo';\n\nexport type CartesiaModels = _CartesiaModels | `${_CartesiaModels}:${string}`;\n\ntype _ElevenlabsModels =\n  | 'elevenlabs'\n  | 'elevenlabs/eleven_flash_v2'\n  | 'elevenlabs/eleven_flash_v2_5'\n  | 'elevenlabs/eleven_turbo_v2'\n  | 'elevenlabs/eleven_turbo_v2_5'\n  | 'elevenlabs/eleven_multilingual_v2';\n\nexport type ElevenlabsModels = _ElevenlabsModels | `${_ElevenlabsModels}:${string}`;\n\nexport type _RimeModels = 'rime' | 'rime/mist' | 'rime/mistv2' | 'rime/arcana';\n\nexport type RimeModels = _RimeModels | `${_RimeModels}:${string}`;\n\nexport type _InworldModels = 'inworld' | 'inworld/inworld-tts-1';\n\nexport type InworldModels = _InworldModels | `${_InworldModels}:${string}`;\n\nexport interface CartesiaOptions {\n  duration?: number; // max duration of audio in seconds\n  speed?: 'slow' | 'normal' | 'fast'; // default: not specified\n}\n\nexport interface ElevenlabsOptions {\n  inactivity_timeout?: number; // default: 60\n  apply_text_normalization?: 'auto' | 'off' | 'on'; // default: \"auto\"\n}\n\nexport interface RimeOptions {}\n\nexport interface InworldOptions {}\n\nexport type TTSModels = CartesiaModels | ElevenlabsModels | RimeModels | InworldModels | AnyModels;\n\nexport type TTSOptions<TModel extends TTSModels> = TModel extends CartesiaModels\n  ? CartesiaOptions\n  : TModel extends ElevenlabsModels\n    ? ElevenlabsOptions\n    : TModel extends RimeOptions\n      ? RimeOptions\n      : TModel extends InworldOptions\n        ? InworldOptions\n        : Record<string, unknown>;\n\ntype TTSEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: TTSEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'https://agent-gateway.livekit.cloud/v1';\nconst NUM_CHANNELS = 1;\nconst DEFAULT_LANGUAGE = 'en';\n\nexport interface InferenceTTSOptions<TModel extends TTSModels> {\n  model?: TModel;\n  voice?: string;\n  language?: string;\n  encoding: TTSEncoding;\n  sampleRate: number;\n  baseURL: string;\n  apiKey: string;\n  apiSecret: string;\n  extraKwargs: TTSOptions<TModel>;\n}\n\nexport class TTS<TModel extends TTSModels> extends BaseTTS {\n  private opts: InferenceTTSOptions<TModel>;\n  private streams: Set<SynthesizeStream<TModel>> = new Set();\n\n  #logger = log();\n\n  constructor(opts: {\n    model: TModel;\n    voice?: string;\n    language?: string;\n    baseURL?: string;\n    encoding?: TTSEncoding;\n    sampleRate?: number;\n    apiKey?: string;\n    apiSecret?: string;\n    extraKwargs?: TTSOptions<TModel>;\n  }) {\n    const sampleRate = opts?.sampleRate ?? DEFAULT_SAMPLE_RATE;\n    super(sampleRate, 1, { streaming: true });\n\n    const {\n      model,\n      voice,\n      language = DEFAULT_LANGUAGE,\n      baseURL,\n      encoding = DEFAULT_ENCODING,\n      apiKey,\n      apiSecret,\n      extraKwargs = {} as TTSOptions<TModel>,\n    } = opts || {};\n\n    const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n    const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n    if (!lkApiKey) {\n      throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n    }\n\n    const lkApiSecret =\n      apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n    if (!lkApiSecret) {\n      throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n    }\n\n    // read voice id from the model if provided: \"provider/model:voice_id\"\n    let nextModel = model;\n    let nextVoice = voice;\n    if (typeof nextModel === 'string') {\n      const idx = nextModel.lastIndexOf(':');\n      if (idx !== -1) {\n        const voiceFromModel = nextModel.slice(idx + 1);\n        if (nextVoice && nextVoice !== voiceFromModel) {\n          this.#logger.warn(\n            '`voice` is provided via both argument and model, using the one from the argument',\n            { voice: nextVoice, model: nextModel },\n          );\n        } else {\n          nextVoice = voiceFromModel;\n        }\n        nextModel = nextModel.slice(0, idx) as TModel;\n      }\n    }\n\n    this.opts = {\n      model: nextModel,\n      voice: nextVoice,\n      language,\n      encoding,\n      sampleRate,\n      baseURL: lkBaseURL,\n      apiKey: lkApiKey,\n      apiSecret: lkApiSecret,\n      extraKwargs,\n    };\n  }\n\n  get label() {\n    return 'inference.TTS';\n  }\n\n  updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n    this.opts = { ...this.opts, ...opts };\n    for (const stream of this.streams) {\n      stream.updateOptions(opts);\n    }\n  }\n\n  synthesize(_: string): ChunkedStream {\n    throw new Error('ChunkedStream is not implemented');\n  }\n\n  stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream<TModel> {\n    const { connOptions = DEFAULT_API_CONNECT_OPTIONS } = options || {};\n    const stream = new SynthesizeStream(this, { ...this.opts }, connOptions);\n    this.streams.add(stream);\n    return stream;\n  }\n\n  async connectWs(timeout: number): Promise<WebSocket> {\n    let baseURL = this.opts.baseURL;\n    if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n      baseURL = baseURL.replace('http', 'ws');\n    }\n\n    const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n    const url = `${baseURL}/tts`;\n    const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n    const params = {\n      type: 'session.create',\n      sample_rate: String(this.opts.sampleRate),\n      encoding: this.opts.encoding,\n      extra: this.opts.extraKwargs,\n    } as TtsSessionCreateEvent;\n\n    if (this.opts.voice) params.voice = this.opts.voice;\n    if (this.opts.model) params.model = this.opts.model;\n    if (this.opts.language) params.language = this.opts.language;\n\n    const socket = await connectWs(url, headers, timeout);\n    socket.send(JSON.stringify(params));\n    return socket;\n  }\n\n  async closeWs(ws: WebSocket) {\n    await ws.close();\n  }\n\n  async close() {\n    for (const stream of this.streams) {\n      await stream.close();\n    }\n    this.streams.clear();\n  }\n}\n\nexport class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeStream {\n  private opts: InferenceTTSOptions<TModel>;\n  private tts: TTS<TModel>;\n  private connOptions: APIConnectOptions;\n\n  #logger = log();\n\n  constructor(tts: TTS<TModel>, opts: InferenceTTSOptions<TModel>, connOptions: APIConnectOptions) {\n    super(tts, connOptions);\n    this.opts = opts;\n    this.tts = tts;\n    this.connOptions = connOptions;\n  }\n\n  get label() {\n    return 'inference.SynthesizeStream';\n  }\n\n  updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n    this.opts = { ...this.opts, ...opts };\n  }\n\n  protected async run(): Promise<void> {\n    let ws: WebSocket | null = null;\n    let closing = false;\n    let finalReceived = false;\n    let lastFrame: AudioFrame | undefined;\n\n    const sendTokenizerStream = new tokenizeBasic.SentenceTokenizer().stream();\n    const eventChannel = createStreamChannel<TtsServerEvent>();\n    const requestId = shortuuid('tts_request_');\n\n    const resourceCleanup = () => {\n      if (closing) return;\n      closing = true;\n      sendTokenizerStream.close();\n      eventChannel.close();\n      ws?.removeAllListeners();\n      ws?.close();\n    };\n\n    const sendClientEvent = async (event: TtsClientEvent) => {\n      const validatedEvent = await ttsClientEventSchema.parseAsync(event);\n      if (!ws || ws.readyState !== WebSocket.OPEN) {\n        this.#logger.warn('Trying to send client TTS event to a closed WebSocket');\n        return;\n      }\n      ws.send(JSON.stringify(validatedEvent));\n    };\n\n    const sendLastFrame = (segmentId: string, final: boolean) => {\n      if (lastFrame) {\n        this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n        lastFrame = undefined;\n      }\n    };\n\n    const createInputTask = async () => {\n      for await (const data of this.input) {\n        if (this.abortController.signal.aborted) break;\n        if (data === SynthesizeStream.FLUSH_SENTINEL) {\n          sendTokenizerStream.flush();\n          continue;\n        }\n        sendTokenizerStream.pushText(data);\n      }\n      sendTokenizerStream.endInput();\n    };\n\n    const createSentenceStreamTask = async () => {\n      for await (const ev of sendTokenizerStream) {\n        if (this.abortController.signal.aborted) break;\n\n        sendClientEvent({\n          type: 'input_transcript',\n          transcript: ev.token + ' ',\n        });\n      }\n\n      sendClientEvent({ type: 'session.flush' });\n    };\n\n    const createWsListenerTask = async (ws: WebSocket) => {\n      return new Promise<void>((resolve, reject) => {\n        this.abortController.signal.addEventListener('abort', () => {\n          resourceCleanup();\n          reject(new Error('WebSocket connection aborted'));\n        });\n\n        ws.on('message', async (data) => {\n          const eventJson = JSON.parse(data.toString()) as Record<string, unknown>;\n          const validatedEvent = ttsServerEventSchema.parse(eventJson);\n          eventChannel.write(validatedEvent);\n        });\n\n        ws.on('error', (e) => {\n          this.#logger.error({ error: e }, 'WebSocket error');\n          resourceCleanup();\n          reject(e);\n        });\n\n        ws.on('close', () => {\n          resourceCleanup();\n\n          if (!closing) return this.#logger.error('WebSocket closed unexpectedly');\n          if (finalReceived) return resolve();\n\n          reject(\n            new APIStatusError({\n              message: 'Gateway connection closed unexpectedly',\n              options: { requestId },\n            }),\n          );\n        });\n      });\n    };\n\n    const createRecvTask = async () => {\n      let currentSessionId: string | null = null;\n\n      const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n      const serverEventStream = eventChannel.stream();\n      const reader = serverEventStream.getReader();\n\n      try {\n        while (!this.closed && !this.abortController.signal.aborted) {\n          const result = await reader.read();\n          if (this.abortController.signal.aborted) return;\n          if (result.done) return;\n\n          const serverEvent = result.value;\n          switch (serverEvent.type) {\n            case 'session.created':\n              currentSessionId = serverEvent.session_id;\n              break;\n            case 'output_audio':\n              const base64Data = new Int8Array(Buffer.from(serverEvent.audio, 'base64'));\n              for (const frame of bstream.write(base64Data.buffer)) {\n                sendLastFrame(currentSessionId!, false);\n                lastFrame = frame;\n              }\n              break;\n            case 'done':\n              finalReceived = true;\n              for (const frame of bstream.flush()) {\n                sendLastFrame(currentSessionId!, false);\n                lastFrame = frame;\n              }\n              sendLastFrame(currentSessionId!, true);\n              this.queue.put(SynthesizeStream.END_OF_STREAM);\n              break;\n            case 'session.closed':\n              resourceCleanup();\n              break;\n            case 'error':\n              this.#logger.error(\n                { serverEvent },\n                'Received error message from LiveKit TTS WebSocket',\n              );\n              resourceCleanup();\n              throw new APIError(`LiveKit TTS returned error: ${serverEvent.message}`);\n            default:\n              this.#logger.warn('Unexpected message %s', serverEvent);\n              break;\n          }\n        }\n      } finally {\n        reader.releaseLock();\n        try {\n          await serverEventStream.cancel();\n        } catch (e) {\n          this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n        }\n      }\n    };\n\n    try {\n      ws = await this.tts.connectWs(this.connOptions.timeoutMs);\n\n      await Promise.all([\n        createInputTask(),\n        createSentenceStreamTask(),\n        createWsListenerTask(ws),\n        createRecvTask(),\n      ]);\n    } catch (e) {\n      this.#logger.error('Error in SynthesizeStream', { error: e });\n    } finally {\n      resourceCleanup();\n    }\n  }\n}\n"],"mappings":"AAIA,SAAS,iBAAiB;AAC1B,SAAS,UAAU,sBAAsB;AACzC,SAAS,uBAAuB;AAChC,SAAS,WAAW;AACpB,SAAS,2BAA2B;AACpC,SAAS,SAAS,qBAAqB;AACvC;AAAA,EACE,oBAAoB;AAAA,EACpB,OAAO;AAAA,EACP;AAAA,OACK;AACP,SAAiC,mCAAmC;AACpE,SAAS,iBAAiB;AAC1B;AAAA,EAIE;AAAA,EACA;AAAA,OACK;AACP,SAAyB,WAAW,yBAAyB;AAoD7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,eAAe;AACrB,MAAM,mBAAmB;AAclB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAyC,oBAAI,IAAI;AAAA,EAEzD,UAAU,IAAI;AAAA,EAEd,YAAY,MAUT;AACD,UAAM,cAAa,6BAAM,eAAc;AACvC,UAAM,YAAY,GAAG,EAAE,WAAW,KAAK,CAAC;AAExC,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA;AAAA,MACA,cAAc,CAAC;AAAA,IACjB,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,WAAW,QAAQ,IAAI,yBAAyB;AAClE,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAGA,QAAI,YAAY;AAChB,QAAI,YAAY;AAChB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,iBAAiB,UAAU,MAAM,MAAM,CAAC;AAC9C,YAAI,aAAa,cAAc,gBAAgB;AAC7C,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,OAAO,WAAW,OAAO,UAAU;AAAA,UACvC;AAAA,QACF,OAAO;AACL,sBAAY;AAAA,QACd;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AAEA,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,OAAO;AAAA,MACP;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,WAAW,GAA0B;AACnC,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AAAA,EAEA,OAAO,SAAyE;AAC9E,UAAM,EAAE,cAAc,4BAA4B,IAAI,WAAW,CAAC;AAClE,UAAM,SAAS,IAAI,iBAAiB,MAAM,EAAE,GAAG,KAAK,KAAK,GAAG,WAAW;AACvE,SAAK,QAAQ,IAAI,MAAM;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AACnD,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,MAAM,kBAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS;AAAA,MACb,MAAM;AAAA,MACN,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,MACxC,UAAU,KAAK,KAAK;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,IACnB;AAEA,QAAI,KAAK,KAAK,MAAO,QAAO,QAAQ,KAAK,KAAK;AAC9C,QAAI,KAAK,KAAK,MAAO,QAAO,QAAQ,KAAK,KAAK;AAC9C,QAAI,KAAK,KAAK,SAAU,QAAO,WAAW,KAAK,KAAK;AAEpD,UAAM,SAAS,MAAM,UAAU,KAAK,SAAS,OAAO;AACpD,WAAO,KAAK,KAAK,UAAU,MAAM,CAAC;AAClC,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,QAAQ,IAAe;AAC3B,UAAM,GAAG,MAAM;AAAA,EACjB;AAAA,EAEA,MAAM,QAAQ;AACZ,eAAW,UAAU,KAAK,SAAS;AACjC,YAAM,OAAO,MAAM;AAAA,IACrB;AACA,SAAK,QAAQ,MAAM;AAAA,EACrB;AACF;AAEO,MAAM,yBAAmD,qBAAqB;AAAA,EAC3E;AAAA,EACA;AAAA,EACA;AAAA,EAER,UAAU,IAAI;AAAA,EAEd,YAAY,KAAkB,MAAmC,aAAgC;AAC/F,UAAM,KAAK,WAAW;AACtB,SAAK,OAAO;AACZ,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAAA,EACtC;AAAA,EAEA,MAAgB,MAAqB;AACnC,QAAI,KAAuB;AAC3B,QAAI,UAAU;AACd,QAAI,gBAAgB;AACpB,QAAI;AAEJ,UAAM,sBAAsB,IAAI,cAAc,kBAAkB,EAAE,OAAO;AACzE,UAAM,eAAe,oBAAoC;AACzD,UAAM,YAAY,UAAU,cAAc;AAE1C,UAAM,kBAAkB,MAAM;AAC5B,UAAI,QAAS;AACb,gBAAU;AACV,0BAAoB,MAAM;AAC1B,mBAAa,MAAM;AACnB,+BAAI;AACJ,+BAAI;AAAA,IACN;AAEA,UAAM,kBAAkB,OAAO,UAA0B;AACvD,YAAM,iBAAiB,MAAM,qBAAqB,WAAW,KAAK;AAClE,UAAI,CAAC,MAAM,GAAG,eAAe,UAAU,MAAM;AAC3C,aAAK,QAAQ,KAAK,uDAAuD;AACzE;AAAA,MACF;AACA,SAAG,KAAK,KAAK,UAAU,cAAc,CAAC;AAAA,IACxC;AAEA,UAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,UAAI,WAAW;AACb,aAAK,MAAM,IAAI,EAAE,WAAW,WAAW,OAAO,WAAW,MAAM,CAAC;AAChE,oBAAY;AAAA,MACd;AAAA,IACF;AAEA,UAAM,kBAAkB,YAAY;AAClC,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,8BAAoB,MAAM;AAC1B;AAAA,QACF;AACA,4BAAoB,SAAS,IAAI;AAAA,MACnC;AACA,0BAAoB,SAAS;AAAA,IAC/B;AAEA,UAAM,2BAA2B,YAAY;AAC3C,uBAAiB,MAAM,qBAAqB;AAC1C,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,wBAAgB;AAAA,UACd,MAAM;AAAA,UACN,YAAY,GAAG,QAAQ;AAAA,QACzB,CAAC;AAAA,MACH;AAEA,sBAAgB,EAAE,MAAM,gBAAgB,CAAC;AAAA,IAC3C;AAEA,UAAM,uBAAuB,OAAOA,QAAkB;AACpD,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,aAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,0BAAgB;AAChB,iBAAO,IAAI,MAAM,8BAA8B,CAAC;AAAA,QAClD,CAAC;AAED,QAAAA,IAAG,GAAG,WAAW,OAAO,SAAS;AAC/B,gBAAM,YAAY,KAAK,MAAM,KAAK,SAAS,CAAC;AAC5C,gBAAM,iBAAiB,qBAAqB,MAAM,SAAS;AAC3D,uBAAa,MAAM,cAAc;AAAA,QACnC,CAAC;AAED,QAAAA,IAAG,GAAG,SAAS,CAAC,MAAM;AACpB,eAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,0BAAgB;AAChB,iBAAO,CAAC;AAAA,QACV,CAAC;AAED,QAAAA,IAAG,GAAG,SAAS,MAAM;AACnB,0BAAgB;AAEhB,cAAI,CAAC,QAAS,QAAO,KAAK,QAAQ,MAAM,+BAA+B;AACvE,cAAI,cAAe,QAAO,QAAQ;AAElC;AAAA,YACE,IAAI,eAAe;AAAA,cACjB,SAAS;AAAA,cACT,SAAS,EAAE,UAAU;AAAA,YACvB,CAAC;AAAA,UACH;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,UAAM,iBAAiB,YAAY;AACjC,UAAI,mBAAkC;AAEtC,YAAM,UAAU,IAAI,gBAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,YAAM,oBAAoB,aAAa,OAAO;AAC9C,YAAM,SAAS,kBAAkB,UAAU;AAE3C,UAAI;AACF,eAAO,CAAC,KAAK,UAAU,CAAC,KAAK,gBAAgB,OAAO,SAAS;AAC3D,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,cAAI,OAAO,KAAM;AAEjB,gBAAM,cAAc,OAAO;AAC3B,kBAAQ,YAAY,MAAM;AAAA,YACxB,KAAK;AACH,iCAAmB,YAAY;AAC/B;AAAA,YACF,KAAK;AACH,oBAAM,aAAa,IAAI,UAAU,OAAO,KAAK,YAAY,OAAO,QAAQ,CAAC;AACzE,yBAAW,SAAS,QAAQ,MAAM,WAAW,MAAM,GAAG;AACpD,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA;AAAA,YACF,KAAK;AACH,8BAAgB;AAChB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA,4BAAc,kBAAmB,IAAI;AACrC,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAC7C;AAAA,YACF,KAAK;AACH,8BAAgB;AAChB;AAAA,YACF,KAAK;AACH,mBAAK,QAAQ;AAAA,gBACX,EAAE,YAAY;AAAA,gBACd;AAAA,cACF;AACA,8BAAgB;AAChB,oBAAM,IAAI,SAAS,+BAA+B,YAAY,OAAO,EAAE;AAAA,YACzE;AACE,mBAAK,QAAQ,KAAK,yBAAyB,WAAW;AACtD;AAAA,UACJ;AAAA,QACF;AAAA,MACF,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,kBAAkB,OAAO;AAAA,QACjC,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,QACxF;AAAA,MACF;AAAA,IACF;AAEA,QAAI;AACF,WAAK,MAAM,KAAK,IAAI,UAAU,KAAK,YAAY,SAAS;AAExD,YAAM,QAAQ,IAAI;AAAA,QAChB,gBAAgB;AAAA,QAChB,yBAAyB;AAAA,QACzB,qBAAqB,EAAE;AAAA,QACvB,eAAe;AAAA,MACjB,CAAC;AAAA,IACH,SAAS,GAAG;AACV,WAAK,QAAQ,MAAM,6BAA6B,EAAE,OAAO,EAAE,CAAC;AAAA,IAC9D,UAAE;AACA,sBAAgB;AAAA,IAClB;AAAA,EACF;AACF;","names":["ws"]}

package/dist/inference/utils.cjs

@@ -0,0 +1,76 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+var utils_exports = {};
+__export(utils_exports, {
+  connectWs: () => connectWs,
+  createAccessToken: () => createAccessToken
+});
+module.exports = __toCommonJS(utils_exports);
+var import_livekit_server_sdk = require("livekit-server-sdk");
+var import_ws = require("ws");
+var import__ = require("../index.cjs");
+async function createAccessToken(apiKey, apiSecret, ttl = 600) {
+  const token = new import_livekit_server_sdk.AccessToken(apiKey, apiSecret, { identity: "agent", ttl });
+  token.addInferenceGrant({ perform: true });
+  return await token.toJwt();
+}
+async function connectWs(url, headers, timeoutMs) {
+  return new Promise((resolve, reject) => {
+    const socket = new import_ws.WebSocket(url, { headers });
+    const timeout = setTimeout(() => {
+      reject(new import__.APIConnectionError({ message: "Timeout connecting to LiveKit WebSocket" }));
+    }, timeoutMs);
+    const onOpen = () => {
+      clearTimeout(timeout);
+      resolve(socket);
+    };
+    const onError = (err) => {
+      clearTimeout(timeout);
+      if (err && typeof err === "object" && "code" in err && err.code === 429) {
+        reject(
+          new import__.APIStatusError({
+            message: "LiveKit gateway quota exceeded",
+            options: { statusCode: 429 }
+          })
+        );
+      } else {
+        reject(new import__.APIConnectionError({ message: "Error connecting to LiveKit WebSocket" }));
+      }
+    };
+    const onClose = (code) => {
+      clearTimeout(timeout);
+      if (code !== 1e3) {
+        reject(
+          new import__.APIConnectionError({
+            message: "Connection closed unexpectedly"
+          })
+        );
+      }
+    };
+    socket.once("open", onOpen);
+    socket.once("error", onError);
+    socket.once("close", onClose);
+  });
+}
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  connectWs,
+  createAccessToken
+});
+//# sourceMappingURL=utils.cjs.map

package/dist/inference/utils.cjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/inference/utils.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { AccessToken } from 'livekit-server-sdk';\nimport { WebSocket } from 'ws';\nimport { APIConnectionError, APIStatusError } from '../index.js';\n\nexport type AnyModels = string & NonNullable<unknown>;\n\nexport async function createAccessToken(\n  apiKey: string,\n  apiSecret: string,\n  ttl: number = 600,\n): Promise<string> {\n  const token = new AccessToken(apiKey, apiSecret, { identity: 'agent', ttl });\n  token.addInferenceGrant({ perform: true });\n\n  return await token.toJwt();\n}\n\nexport async function connectWs(\n  url: string,\n  headers: Record<string, string>,\n  timeoutMs: number,\n): Promise<WebSocket> {\n  return new Promise<WebSocket>((resolve, reject) => {\n    const socket = new WebSocket(url, { headers: headers });\n\n    const timeout = setTimeout(() => {\n      reject(new APIConnectionError({ message: 'Timeout connecting to LiveKit WebSocket' }));\n    }, timeoutMs);\n\n    const onOpen = () => {\n      clearTimeout(timeout);\n      resolve(socket);\n    };\n\n    const onError = (err: unknown) => {\n      clearTimeout(timeout);\n      if (err && typeof err === 'object' && 'code' in err && (err as any).code === 429) {\n        reject(\n          new APIStatusError({\n            message: 'LiveKit gateway quota exceeded',\n            options: { statusCode: 429 },\n          }),\n        );\n      } else {\n        reject(new APIConnectionError({ message: 'Error connecting to LiveKit WebSocket' }));\n      }\n    };\n\n    const onClose = (code: number) => {\n      clearTimeout(timeout);\n      if (code !== 1000) {\n        reject(\n          new APIConnectionError({\n            message: 'Connection closed unexpectedly',\n          }),\n        );\n      }\n    };\n    socket.once('open', onOpen);\n    socket.once('error', onError);\n    socket.once('close', onClose);\n  });\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,gCAA4B;AAC5B,gBAA0B;AAC1B,eAAmD;AAInD,eAAsB,kBACpB,QACA,WACA,MAAc,KACG;AACjB,QAAM,QAAQ,IAAI,sCAAY,QAAQ,WAAW,EAAE,UAAU,SAAS,IAAI,CAAC;AAC3E,QAAM,kBAAkB,EAAE,SAAS,KAAK,CAAC;AAEzC,SAAO,MAAM,MAAM,MAAM;AAC3B;AAEA,eAAsB,UACpB,KACA,SACA,WACoB;AACpB,SAAO,IAAI,QAAmB,CAAC,SAAS,WAAW;AACjD,UAAM,SAAS,IAAI,oBAAU,KAAK,EAAE,QAAiB,CAAC;AAEtD,UAAM,UAAU,WAAW,MAAM;AAC/B,aAAO,IAAI,4BAAmB,EAAE,SAAS,0CAA0C,CAAC,CAAC;AAAA,IACvF,GAAG,SAAS;AAEZ,UAAM,SAAS,MAAM;AACnB,mBAAa,OAAO;AACpB,cAAQ,MAAM;AAAA,IAChB;AAEA,UAAM,UAAU,CAAC,QAAiB;AAChC,mBAAa,OAAO;AACpB,UAAI,OAAO,OAAO,QAAQ,YAAY,UAAU,OAAQ,IAAY,SAAS,KAAK;AAChF;AAAA,UACE,IAAI,wBAAe;AAAA,YACjB,SAAS;AAAA,YACT,SAAS,EAAE,YAAY,IAAI;AAAA,UAC7B,CAAC;AAAA,QACH;AAAA,MACF,OAAO;AACL,eAAO,IAAI,4BAAmB,EAAE,SAAS,wCAAwC,CAAC,CAAC;AAAA,MACrF;AAAA,IACF;AAEA,UAAM,UAAU,CAAC,SAAiB;AAChC,mBAAa,OAAO;AACpB,UAAI,SAAS,KAAM;AACjB;AAAA,UACE,IAAI,4BAAmB;AAAA,YACrB,SAAS;AAAA,UACX,CAAC;AAAA,QACH;AAAA,MACF;AAAA,IACF;AACA,WAAO,KAAK,QAAQ,MAAM;AAC1B,WAAO,KAAK,SAAS,OAAO;AAC5B,WAAO,KAAK,SAAS,OAAO;AAAA,EAC9B,CAAC;AACH;","names":[]}

package/dist/inference/utils.d.cts

@@ -0,0 +1,5 @@
+import { WebSocket } from 'ws';
+export type AnyModels = string & NonNullable<unknown>;
+export declare function createAccessToken(apiKey: string, apiSecret: string, ttl?: number): Promise<string>;
+export declare function connectWs(url: string, headers: Record<string, string>, timeoutMs: number): Promise<WebSocket>;
+//# sourceMappingURL=utils.d.ts.map

package/dist/inference/utils.d.ts

@@ -0,0 +1,5 @@
+import { WebSocket } from 'ws';
+export type AnyModels = string & NonNullable<unknown>;
+export declare function createAccessToken(apiKey: string, apiSecret: string, ttl?: number): Promise<string>;
+export declare function connectWs(url: string, headers: Record<string, string>, timeoutMs: number): Promise<WebSocket>;
+//# sourceMappingURL=utils.d.ts.map

package/dist/inference/utils.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"utils.d.ts","sourceRoot":"","sources":["../../src/inference/utils.ts"],"names":[],"mappings":"AAIA,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAG/B,MAAM,MAAM,SAAS,GAAG,MAAM,GAAG,WAAW,CAAC,OAAO,CAAC,CAAC;AAEtD,wBAAsB,iBAAiB,CACrC,MAAM,EAAE,MAAM,EACd,SAAS,EAAE,MAAM,EACjB,GAAG,GAAE,MAAY,GAChB,OAAO,CAAC,MAAM,CAAC,CAKjB;AAED,wBAAsB,SAAS,CAC7B,GAAG,EAAE,MAAM,EACX,OAAO,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,EAC/B,SAAS,EAAE,MAAM,GAChB,OAAO,CAAC,SAAS,CAAC,CAyCpB"}