@livekit/agents 1.1.0 → 1.2.1

package/dist/inference/stt.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/inference/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame } from '@livekit/rtc-node';\nimport { type RawData, WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport {\n  STT as BaseSTT,\n  SpeechStream as BaseSpeechStream,\n  type SpeechData,\n  type SpeechEvent,\n  SpeechEventType,\n} from '../stt/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { type AudioBuffer, Event, Task, cancelAndWait, shortuuid, waitForAbort } from '../utils.js';\nimport { type AnyModels, connectWs, createAccessToken } from './utils.js';\n\nexport type DeepgramModels =\n  | 'deepgram'\n  | 'deepgram/nova-3'\n  | 'deepgram/nova-3-general'\n  | 'deepgram/nova-3-medical'\n  | 'deepgram/nova-2'\n  | 'deepgram/nova-2-general'\n  | 'deepgram/nova-2-medical'\n  | 'deepgram/nova-2-phonecall';\n\nexport type CartesiaModels = 'cartesia' | 'cartesia/ink-whisper';\n\nexport type AssemblyaiModels = 'assemblyai' | 'assemblyai/universal-streaming';\n\nexport interface CartesiaOptions {\n  min_volume?: number; // default: not specified\n  max_silence_duration_secs?: number; // default: not specified\n}\n\nexport interface DeepgramOptions {\n  filler_words?: boolean; // default: true\n  interim_results?: boolean; // default: true\n  endpointing?: number; // default: 25 (ms)\n  punctuate?: boolean; // default: false\n  smart_format?: boolean;\n  keywords?: Array<[string, number]>;\n  keyterms?: string[];\n  profanity_filter?: boolean;\n  numerals?: boolean;\n  mip_opt_out?: boolean;\n}\n\nexport interface AssemblyaiOptions {\n  format_turns?: boolean; // default: false\n  end_of_turn_confidence_threshold?: number; // default: 0.01\n  min_end_of_turn_silence_when_confident?: number; // default: 0\n  max_turn_silence?: number; // default: not specified\n  keyterms_prompt?: string[]; // default: not specified\n}\n\nexport type STTModels = DeepgramModels | CartesiaModels | AssemblyaiModels | AnyModels;\nexport type STTOptions<TModel extends STTModels> = TModel extends DeepgramModels\n  ? DeepgramOptions\n  : TModel extends CartesiaModels\n    ? CartesiaOptions\n    : TModel extends AssemblyaiModels\n      ? AssemblyaiOptions\n      : Record<string, unknown>;\n\nexport type STTLanguages = 'en' | 'de' | 'es' | 'fr' | 'ja' | 'pt' | 'zh';\nexport type STTEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: STTEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'wss://agent-gateway.livekit.cloud/v1';\nconst DEFAULT_CANCEL_TIMEOUT = 5000;\n\nexport interface InferenceSTTOptions<TModel extends STTModels> {\n  model: TModel;\n  language?: STTLanguages | string;\n  encoding: STTEncoding;\n  sampleRate: number;\n  baseURL: string;\n  apiKey: string;\n  apiSecret: string;\n  extraKwargs: STTOptions<TModel>;\n}\n\nexport class STT<TModel extends STTModels> extends BaseSTT {\n  private opts: InferenceSTTOptions<TModel>;\n  private streams: Set<SpeechStream<TModel>> = new Set();\n\n  constructor(opts: {\n    model: TModel;\n    language?: STTLanguages | string;\n    baseURL?: string;\n    encoding?: STTEncoding;\n    sampleRate?: number;\n    apiKey?: string;\n    apiSecret?: string;\n    extraKwargs?: STTOptions<TModel>;\n  }) {\n    super({ streaming: true, interimResults: true });\n\n    const {\n      model,\n      language,\n      baseURL,\n      encoding = DEFAULT_ENCODING,\n      sampleRate = DEFAULT_SAMPLE_RATE,\n      apiKey,\n      apiSecret,\n      extraKwargs = {} as STTOptions<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    this.opts = {\n      model,\n      language,\n      encoding,\n      sampleRate,\n      baseURL: lkBaseURL,\n      apiKey: lkApiKey,\n      apiSecret: lkApiSecret,\n      extraKwargs,\n    };\n  }\n\n  get label(): string {\n    return 'inference.STT';\n  }\n\n  protected async _recognize(_: AudioBuffer): Promise<SpeechEvent> {\n    throw new Error('LiveKit STT does not support batch recognition, use stream() instead');\n  }\n\n  updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n    this.opts = { ...this.opts, ...opts };\n\n    for (const stream of this.streams) {\n      stream.updateOptions(opts);\n    }\n  }\n\n  stream(options?: {\n    language?: STTLanguages | string;\n    connOptions?: APIConnectOptions;\n  }): SpeechStream<TModel> {\n    const { language, connOptions = DEFAULT_API_CONNECT_OPTIONS } = options || {};\n    const streamOpts = {\n      ...this.opts,\n      language: language ?? this.opts.language,\n    } as InferenceSTTOptions<TModel>;\n\n    const stream = new SpeechStream(this, streamOpts, connOptions);\n    this.streams.add(stream);\n\n    return stream;\n  }\n}\n\nexport class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {\n  private opts: InferenceSTTOptions<TModel>;\n  private requestId = shortuuid('stt_request_');\n  private speaking = false;\n  private speechDuration = 0;\n  private reconnectEvent = new Event();\n\n  #logger = log();\n\n  constructor(\n    sttImpl: STT<TModel>,\n    opts: InferenceSTTOptions<TModel>,\n    connOptions: APIConnectOptions,\n  ) {\n    super(sttImpl, opts.sampleRate, connOptions);\n    this.opts = opts;\n  }\n\n  get label(): string {\n    return 'inference.SpeechStream';\n  }\n\n  updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n    this.opts = { ...this.opts, ...opts };\n  }\n\n  protected async run(): Promise<void> {\n    let ws: WebSocket | null = null;\n    let closingWs = false;\n\n    this.reconnectEvent.set();\n\n    const connect = async () => {\n      const params = {\n        settings: {\n          sample_rate: String(this.opts.sampleRate),\n          encoding: this.opts.encoding,\n          extra: this.opts.extraKwargs,\n        },\n      } as Record<string, unknown>;\n\n      if (this.opts.model) {\n        params.model = this.opts.model;\n      }\n\n      if (this.opts.language) {\n        (params.settings as Record<string, unknown>).language = this.opts.language;\n      }\n\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}/stt`;\n      const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n      const socket = await connectWs(url, headers, 10000);\n      const msg = { ...params, type: 'session.create' };\n      socket.send(JSON.stringify(msg));\n\n      return socket;\n    };\n\n    const send = async (socket: WebSocket, signal: AbortSignal) => {\n      const audioStream = new AudioByteStream(\n        this.opts.sampleRate,\n        1,\n        Math.floor(this.opts.sampleRate / 20), // 50ms\n      );\n\n      for await (const ev of this.input) {\n        if (signal.aborted) break;\n        let frames: AudioFrame[];\n\n        if (ev === SpeechStream.FLUSH_SENTINEL) {\n          frames = audioStream.flush();\n        } else {\n          const frame = ev as AudioFrame;\n          frames = audioStream.write(new Int16Array(frame.data).buffer);\n        }\n\n        for (const frame of frames) {\n          this.speechDuration += frame.samplesPerChannel / frame.sampleRate;\n          const base64 = Buffer.from(frame.data.buffer).toString('base64');\n          const msg = { type: 'input_audio', audio: base64 };\n          socket.send(JSON.stringify(msg));\n        }\n      }\n\n      closingWs = true;\n      socket.send(JSON.stringify({ type: 'session.finalize' }));\n    };\n\n    const recv = async (socket: WebSocket, signal: AbortSignal) => {\n      while (!this.closed && !signal.aborted) {\n        const dataPromise = new Promise<string>((resolve, reject) => {\n          const messageHandler = (d: RawData) => {\n            resolve(d.toString());\n            removeListeners();\n          };\n          const errorHandler = (e: Error) => {\n            reject(e);\n            removeListeners();\n          };\n          const closeHandler = (code: number) => {\n            if (closingWs) {\n              resolve('');\n            } else {\n              reject(\n                new APIStatusError({\n                  message: 'LiveKit STT connection closed unexpectedly',\n                  options: { statusCode: code },\n                }),\n              );\n            }\n            removeListeners();\n          };\n          const removeListeners = () => {\n            socket.removeListener('message', messageHandler);\n            socket.removeListener('error', errorHandler);\n            socket.removeListener('close', closeHandler);\n          };\n          socket.once('message', messageHandler);\n          socket.once('error', errorHandler);\n          socket.once('close', closeHandler);\n        });\n\n        const data = await Promise.race([dataPromise, waitForAbort(signal)]);\n\n        if (!data || signal.aborted) return;\n\n        const json = JSON.parse(data);\n        const type = json.type as string | undefined;\n\n        switch (type) {\n          case 'session.created':\n          case 'session.finalized':\n          case 'session.closed':\n            break;\n          case 'interim_transcript':\n            this.processTranscript(json, false);\n            break;\n          case 'final_transcript':\n            this.processTranscript(json, true);\n            break;\n          case 'error':\n            this.#logger.error('received error from LiveKit STT: %o', json);\n            throw new APIError(`LiveKit STT returned error: ${JSON.stringify(json)}`);\n          default:\n            this.#logger.warn('received unexpected message from LiveKit STT: %o', json);\n            break;\n        }\n      }\n    };\n\n    while (true) {\n      try {\n        ws = await connect();\n\n        const sendTask = Task.from(async ({ signal }) => {\n          await send(ws!, signal);\n        });\n\n        const recvTask = Task.from(async ({ signal }) => {\n          await recv(ws!, signal);\n        });\n\n        const tasks = [sendTask, recvTask];\n        const waitReconnectTask = Task.from(async ({ signal }) => {\n          await Promise.race([this.reconnectEvent.wait(), waitForAbort(signal)]);\n        });\n\n        try {\n          await Promise.race([\n            Promise.all(tasks.map((task) => task.result)),\n            waitReconnectTask.result,\n          ]);\n\n          if (!waitReconnectTask.done) break;\n          this.reconnectEvent.clear();\n        } finally {\n          await cancelAndWait([sendTask, recvTask, waitReconnectTask], DEFAULT_CANCEL_TIMEOUT);\n        }\n      } finally {\n        try {\n          if (ws) ws.close();\n        } catch {}\n      }\n    }\n  }\n\n  private processTranscript(data: Record<string, any>, isFinal: boolean) {\n    const requestId = data.request_id ?? this.requestId;\n    const text = data.transcript ?? '';\n    const language = data.language ?? this.opts.language ?? 'en';\n\n    if (!text && !isFinal) return;\n\n    // We'll have a more accurate way of detecting when speech started when we have VAD\n    if (!this.speaking) {\n      this.speaking = true;\n      this.queue.put({ type: SpeechEventType.START_OF_SPEECH });\n    }\n\n    const speechData: SpeechData = {\n      language,\n      startTime: data.start ?? 0,\n      endTime: data.duration ?? 0,\n      confidence: data.confidence ?? 1.0,\n      text,\n    };\n\n    if (isFinal) {\n      if (this.speechDuration > 0) {\n        this.queue.put({\n          type: SpeechEventType.RECOGNITION_USAGE,\n          requestId,\n          recognitionUsage: { audioDuration: this.speechDuration },\n        });\n        this.speechDuration = 0;\n      }\n\n      this.queue.put({\n        type: SpeechEventType.FINAL_TRANSCRIPT,\n        requestId,\n        alternatives: [speechData],\n      });\n\n      if (this.speaking) {\n        this.speaking = false;\n        this.queue.put({ type: SpeechEventType.END_OF_SPEECH });\n      }\n    } else {\n      this.queue.put({\n        type: SpeechEventType.INTERIM_TRANSCRIPT,\n        requestId,\n        alternatives: [speechData],\n      });\n    }\n  }\n}\n"],"mappings":"AAGA,eAAgC;AAChC,SAAuB,iBAAiB;AACxC,SAAS,UAAU,sBAAsB;AACzC,SAAS,uBAAuB;AAChC,SAAS,WAAW;AACpB;AAAA,EACE,OAAO;AAAA,EACP,gBAAgB;AAAA,EAGhB;AAAA,OACK;AACP,SAAiC,mCAAmC;AACpE,SAA2B,OAAO,MAAM,eAAe,WAAW,oBAAoB;AACtF,SAAyB,WAAW,yBAAyB;AAsD7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,yBAAyB;AAaxB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAqC,oBAAI,IAAI;AAAA,EAErD,YAAY,MAST;AACD,UAAM,EAAE,WAAW,MAAM,gBAAgB,KAAK,CAAC;AAE/C,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX,aAAa;AAAA,MACb;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;AAEA,SAAK,OAAO;AAAA,MACV;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,MAAgB,WAAW,GAAsC;AAC/D,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAEpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,OAAO,SAGkB;AACvB,UAAM,EAAE,UAAU,cAAc,4BAA4B,IAAI,WAAW,CAAC;AAC5E,UAAM,aAAa;AAAA,MACjB,GAAG,KAAK;AAAA,MACR,UAAU,YAAY,KAAK,KAAK;AAAA,IAClC;AAEA,UAAM,SAAS,IAAI,aAAa,MAAM,YAAY,WAAW;AAC7D,SAAK,QAAQ,IAAI,MAAM;AAEvB,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAA+C,iBAAiB;AAAA,EACnE;AAAA,EACA,YAAY,UAAU,cAAc;AAAA,EACpC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,iBAAiB,IAAI,MAAM;AAAA,EAEnC,UAAU,IAAI;AAAA,EAEd,YACE,SACA,MACA,aACA;AACA,UAAM,SAAS,KAAK,YAAY,WAAW;AAC3C,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAAA,EACtC;AAAA,EAEA,MAAgB,MAAqB;AACnC,QAAI,KAAuB;AAC3B,QAAI,YAAY;AAEhB,SAAK,eAAe,IAAI;AAExB,UAAM,UAAU,YAAY;AAC1B,YAAM,SAAS;AAAA,QACb,UAAU;AAAA,UACR,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,UACxC,UAAU,KAAK,KAAK;AAAA,UACpB,OAAO,KAAK,KAAK;AAAA,QACnB;AAAA,MACF;AAEA,UAAI,KAAK,KAAK,OAAO;AACnB,eAAO,QAAQ,KAAK,KAAK;AAAA,MAC3B;AAEA,UAAI,KAAK,KAAK,UAAU;AACtB,QAAC,OAAO,SAAqC,WAAW,KAAK,KAAK;AAAA,MACpE;AAEA,UAAI,UAAU,KAAK,KAAK;AACxB,UAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,kBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,MACxC;AAEA,YAAM,QAAQ,MAAM,kBAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,YAAM,MAAM,GAAG,OAAO;AACtB,YAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,YAAM,SAAS,MAAM,UAAU,KAAK,SAAS,GAAK;AAClD,YAAM,MAAM,EAAE,GAAG,QAAQ,MAAM,iBAAiB;AAChD,aAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAE/B,aAAO;AAAA,IACT;AAEA,UAAM,OAAO,OAAO,QAAmB,WAAwB;AAC7D,YAAM,cAAc,IAAI;AAAA,QACtB,KAAK,KAAK;AAAA,QACV;AAAA,QACA,KAAK,MAAM,KAAK,KAAK,aAAa,EAAE;AAAA;AAAA,MACtC;AAEA,uBAAiB,MAAM,KAAK,OAAO;AACjC,YAAI,OAAO,QAAS;AACpB,YAAI;AAEJ,YAAI,OAAO,aAAa,gBAAgB;AACtC,mBAAS,YAAY,MAAM;AAAA,QAC7B,OAAO;AACL,gBAAM,QAAQ;AACd,mBAAS,YAAY,MAAM,IAAI,WAAW,MAAM,IAAI,EAAE,MAAM;AAAA,QAC9D;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,kBAAkB,MAAM,oBAAoB,MAAM;AACvD,gBAAM,SAAS,OAAO,KAAK,MAAM,KAAK,MAAM,EAAE,SAAS,QAAQ;AAC/D,gBAAM,MAAM,EAAE,MAAM,eAAe,OAAO,OAAO;AACjD,iBAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,QACjC;AAAA,MACF;AAEA,kBAAY;AACZ,aAAO,KAAK,KAAK,UAAU,EAAE,MAAM,mBAAmB,CAAC,CAAC;AAAA,IAC1D;AAEA,UAAM,OAAO,OAAO,QAAmB,WAAwB;AAC7D,aAAO,CAAC,KAAK,UAAU,CAAC,OAAO,SAAS;AACtC,cAAM,cAAc,IAAI,QAAgB,CAAC,SAAS,WAAW;AAC3D,gBAAM,iBAAiB,CAAC,MAAe;AACrC,oBAAQ,EAAE,SAAS,CAAC;AACpB,4BAAgB;AAAA,UAClB;AACA,gBAAM,eAAe,CAAC,MAAa;AACjC,mBAAO,CAAC;AACR,4BAAgB;AAAA,UAClB;AACA,gBAAM,eAAe,CAAC,SAAiB;AACrC,gBAAI,WAAW;AACb,sBAAQ,EAAE;AAAA,YACZ,OAAO;AACL;AAAA,gBACE,IAAI,eAAe;AAAA,kBACjB,SAAS;AAAA,kBACT,SAAS,EAAE,YAAY,KAAK;AAAA,gBAC9B,CAAC;AAAA,cACH;AAAA,YACF;AACA,4BAAgB;AAAA,UAClB;AACA,gBAAM,kBAAkB,MAAM;AAC5B,mBAAO,eAAe,WAAW,cAAc;AAC/C,mBAAO,eAAe,SAAS,YAAY;AAC3C,mBAAO,eAAe,SAAS,YAAY;AAAA,UAC7C;AACA,iBAAO,KAAK,WAAW,cAAc;AACrC,iBAAO,KAAK,SAAS,YAAY;AACjC,iBAAO,KAAK,SAAS,YAAY;AAAA,QACnC,CAAC;AAED,cAAM,OAAO,MAAM,QAAQ,KAAK,CAAC,aAAa,aAAa,MAAM,CAAC,CAAC;AAEnE,YAAI,CAAC,QAAQ,OAAO,QAAS;AAE7B,cAAM,OAAO,KAAK,MAAM,IAAI;AAC5B,cAAM,OAAO,KAAK;AAElB,gBAAQ,MAAM;AAAA,UACZ,KAAK;AAAA,UACL,KAAK;AAAA,UACL,KAAK;AACH;AAAA,UACF,KAAK;AACH,iBAAK,kBAAkB,MAAM,KAAK;AAClC;AAAA,UACF,KAAK;AACH,iBAAK,kBAAkB,MAAM,IAAI;AACjC;AAAA,UACF,KAAK;AACH,iBAAK,QAAQ,MAAM,uCAAuC,IAAI;AAC9D,kBAAM,IAAI,SAAS,+BAA+B,KAAK,UAAU,IAAI,CAAC,EAAE;AAAA,UAC1E;AACE,iBAAK,QAAQ,KAAK,oDAAoD,IAAI;AAC1E;AAAA,QACJ;AAAA,MACF;AAAA,IACF;AAEA,WAAO,MAAM;AACX,UAAI;AACF,aAAK,MAAM,QAAQ;AAEnB,cAAM,WAAW,KAAK,KAAK,OAAO,EAAE,OAAO,MAAM;AAC/C,gBAAM,KAAK,IAAK,MAAM;AAAA,QACxB,CAAC;AAED,cAAM,WAAW,KAAK,KAAK,OAAO,EAAE,OAAO,MAAM;AAC/C,gBAAM,KAAK,IAAK,MAAM;AAAA,QACxB,CAAC;AAED,cAAM,QAAQ,CAAC,UAAU,QAAQ;AACjC,cAAM,oBAAoB,KAAK,KAAK,OAAO,EAAE,OAAO,MAAM;AACxD,gBAAM,QAAQ,KAAK,CAAC,KAAK,eAAe,KAAK,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,QACvE,CAAC;AAED,YAAI;AACF,gBAAM,QAAQ,KAAK;AAAA,YACjB,QAAQ,IAAI,MAAM,IAAI,CAAC,SAAS,KAAK,MAAM,CAAC;AAAA,YAC5C,kBAAkB;AAAA,UACpB,CAAC;AAED,cAAI,CAAC,kBAAkB,KAAM;AAC7B,eAAK,eAAe,MAAM;AAAA,QAC5B,UAAE;AACA,gBAAM,cAAc,CAAC,UAAU,UAAU,iBAAiB,GAAG,sBAAsB;AAAA,QACrF;AAAA,MACF,UAAE;AACA,YAAI;AACF,cAAI,GAAI,IAAG,MAAM;AAAA,QACnB,QAAQ;AAAA,QAAC;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,kBAAkB,MAA2B,SAAkB;AACrE,UAAM,YAAY,KAAK,cAAc,KAAK;AAC1C,UAAM,OAAO,KAAK,cAAc;AAChC,UAAM,WAAW,KAAK,YAAY,KAAK,KAAK,YAAY;AAExD,QAAI,CAAC,QAAQ,CAAC,QAAS;AAGvB,QAAI,CAAC,KAAK,UAAU;AAClB,WAAK,WAAW;AAChB,WAAK,MAAM,IAAI,EAAE,MAAM,gBAAgB,gBAAgB,CAAC;AAAA,IAC1D;AAEA,UAAM,aAAyB;AAAA,MAC7B;AAAA,MACA,WAAW,KAAK,SAAS;AAAA,MACzB,SAAS,KAAK,YAAY;AAAA,MAC1B,YAAY,KAAK,cAAc;AAAA,MAC/B;AAAA,IACF;AAEA,QAAI,SAAS;AACX,UAAI,KAAK,iBAAiB,GAAG;AAC3B,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,gBAAgB;AAAA,UACtB;AAAA,UACA,kBAAkB,EAAE,eAAe,KAAK,eAAe;AAAA,QACzD,CAAC;AACD,aAAK,iBAAiB;AAAA,MACxB;AAEA,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,gBAAgB;AAAA,QACtB;AAAA,QACA,cAAc,CAAC,UAAU;AAAA,MAC3B,CAAC;AAED,UAAI,KAAK,UAAU;AACjB,aAAK,WAAW;AAChB,aAAK,MAAM,IAAI,EAAE,MAAM,gBAAgB,cAAc,CAAC;AAAA,MACxD;AAAA,IACF,OAAO;AACL,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,gBAAgB;AAAA,QACtB;AAAA,QACA,cAAc,CAAC,UAAU;AAAA,MAC3B,CAAC;AAAA,IACH;AAAA,EACF;AACF;","names":[]}
+{"version":3,"sources":["../../src/inference/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame } from '@livekit/rtc-node';\nimport type { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { type LanguageCode, areLanguagesEquivalent, normalizeLanguage } from '../language.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport {\n  STT as BaseSTT,\n  SpeechStream as BaseSpeechStream,\n  type SpeechData,\n  type SpeechEvent,\n  SpeechEventType,\n} from '../stt/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { type AudioBuffer, Event, Task, cancelAndWait, shortuuid, waitForAbort } from '../utils.js';\nimport { type TimedString, createTimedString } from '../voice/io.js';\nimport {\n  type SttServerEvent,\n  type SttTranscriptEvent,\n  sttServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken, getDefaultInferenceUrl } from './utils.js';\n\nexport type DeepgramModels =\n  | 'deepgram/flux-general'\n  | 'deepgram/nova-3'\n  | 'deepgram/nova-3-medical'\n  | 'deepgram/nova-2'\n  | 'deepgram/nova-2-medical'\n  | 'deepgram/nova-2-conversationalai'\n  | 'deepgram/nova-2-phonecall';\n\nexport type CartesiaModels = 'cartesia/ink-whisper';\n\nexport type AssemblyaiModels =\n  | 'assemblyai/universal-streaming'\n  | 'assemblyai/universal-streaming-multilingual';\n\nexport type ElevenlabsSTTModels = 'elevenlabs/scribe_v2_realtime';\n\nexport interface CartesiaOptions {\n  /** Minimum volume threshold. Default: not specified. */\n  min_volume?: number;\n  /** Maximum silence duration in seconds. Default: not specified. */\n  max_silence_duration_secs?: number;\n}\n\nexport interface DeepgramOptions {\n  /** Enable filler words. Default: true. */\n  filler_words?: boolean;\n  /** Enable interim results. Default: true. */\n  interim_results?: boolean;\n  /** Endpointing timeout in milliseconds. Default: 25. */\n  endpointing?: number;\n  /** Enable punctuation. Default: false. */\n  punctuate?: boolean;\n  /** Enable smart formatting. */\n  smart_format?: boolean;\n  /** Keywords with boost values. */\n  keywords?: Array<[string, number]>;\n  /** Key terms for recognition. */\n  keyterms?: string[];\n  /** Enable profanity filter. */\n  profanity_filter?: boolean;\n  /** Convert spoken numbers to numerals. */\n  numerals?: boolean;\n  /** Opt out of model improvement program. */\n  mip_opt_out?: boolean;\n}\n\nexport interface AssemblyAIOptions {\n  /** Enable turn formatting. Default: false. */\n  format_turns?: boolean;\n  /** End of turn confidence threshold. Default: 0.01. */\n  end_of_turn_confidence_threshold?: number;\n  /** Minimum silence duration in milliseconds when confident about end of turn. Default: 0. */\n  min_end_of_turn_silence_when_confident?: number;\n  /** Maximum turn silence in milliseconds. Default: not specified. */\n  max_turn_silence?: number;\n  /** Key terms prompt for recognition. Default: not specified. */\n  keyterms_prompt?: string[];\n}\n\nexport type STTLanguages =\n  | 'multi'\n  | 'en'\n  | 'de'\n  | 'es'\n  | 'fr'\n  | 'ja'\n  | 'pt'\n  | 'zh'\n  | 'hi'\n  | AnyString;\n\ntype _STTModels = DeepgramModels | CartesiaModels | AssemblyaiModels | ElevenlabsSTTModels;\n\nexport type STTModels = _STTModels | 'auto' | AnyString;\n\nexport type ModelWithLanguage = `${_STTModels}:${STTLanguages}` | STTModels;\n\nexport type STTOptions<TModel extends STTModels> = TModel extends DeepgramModels\n  ? DeepgramOptions\n  : TModel extends CartesiaModels\n    ? CartesiaOptions\n    : TModel extends AssemblyaiModels\n      ? AssemblyAIOptions\n      : Record<string, unknown>;\n\n/** A fallback model with optional extra configuration. Extra fields are passed through to the provider. */\nexport interface STTFallbackModel {\n  /** Model name (e.g. \"deepgram/nova-3\", \"assemblyai/universal-streaming\", \"cartesia/ink-whisper\"). */\n  model: string;\n  /** Extra configuration for the model. */\n  extraKwargs?: Record<string, unknown>;\n}\n\nexport type STTFallbackModelType = STTFallbackModel | string;\n\n/** Parse a model string into [model, language]. Language is undefined if not specified. */\nexport function parseSTTModelString(model: string): [string, LanguageCode | undefined] {\n  const idx = model.lastIndexOf(':');\n  if (idx !== -1) {\n    return [model.slice(0, idx), normalizeLanguage(model.slice(idx + 1))];\n  }\n  return [model, undefined];\n}\n\n/** Normalize a single or list of FallbackModelType into STTFallbackModel[]. */\nexport function normalizeSTTFallback(\n  fallback: STTFallbackModelType | STTFallbackModelType[],\n): STTFallbackModel[] {\n  const makeFallback = (model: STTFallbackModelType): STTFallbackModel => {\n    if (typeof model === 'string') {\n      const [name] = parseSTTModelString(model);\n      return { model: name };\n    }\n    return model;\n  };\n\n  if (Array.isArray(fallback)) {\n    return fallback.map(makeFallback);\n  }\n  return [makeFallback(fallback)];\n}\n\nexport type STTEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: STTEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_CANCEL_TIMEOUT = 5000;\n\nexport interface InferenceSTTOptions<TModel extends STTModels> {\n  model?: TModel;\n  language?: LanguageCode;\n  encoding: STTEncoding;\n  sampleRate: number;\n  baseURL: string;\n  apiKey: string;\n  apiSecret: string;\n  modelOptions: STTOptions<TModel>;\n  fallback?: STTFallbackModel[];\n  connOptions?: APIConnectOptions;\n}\n\n/**\n * Livekit Cloud Inference STT\n */\nexport class STT<TModel extends STTModels> extends BaseSTT {\n  private opts: InferenceSTTOptions<TModel>;\n  private streams: Set<SpeechStream<TModel>> = new Set();\n\n  #logger = log();\n\n  constructor(opts?: {\n    model?: ModelWithLanguage;\n    language?: STTLanguages;\n    baseURL?: string;\n    encoding?: STTEncoding;\n    sampleRate?: number;\n    apiKey?: string;\n    apiSecret?: string;\n    modelOptions?: STTOptions<TModel>;\n    fallback?: STTFallbackModelType | STTFallbackModelType[];\n    connOptions?: APIConnectOptions;\n  }) {\n    super({ streaming: true, interimResults: true, alignedTranscript: 'word' });\n\n    const {\n      model,\n      language,\n      baseURL,\n      encoding = DEFAULT_ENCODING,\n      sampleRate = DEFAULT_SAMPLE_RATE,\n      apiKey,\n      apiSecret,\n      modelOptions = {} as STTOptions<TModel>,\n      fallback,\n      connOptions,\n    } = opts || {};\n\n    const lkBaseURL = baseURL || getDefaultInferenceUrl();\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    // Parse language from model string if provided: \"provider/model:language\"\n    let nextModel = model;\n    let nextLanguage = language;\n    if (typeof nextModel === 'string') {\n      const [parsedModel, parsedLanguage] = parseSTTModelString(nextModel);\n      if (parsedLanguage !== undefined) {\n        if (nextLanguage && !areLanguagesEquivalent(nextLanguage, parsedLanguage)) {\n          this.#logger.warn(\n            '`language` is provided via both argument and model, using the one from the argument',\n            { language: nextLanguage, model: nextModel },\n          );\n        } else {\n          nextLanguage = parsedLanguage as STTLanguages;\n        }\n        nextModel = parsedModel as TModel;\n      }\n    }\n    const normalizedFallback = fallback ? normalizeSTTFallback(fallback) : undefined;\n\n    this.opts = {\n      model: nextModel as TModel,\n      language: nextLanguage ? normalizeLanguage(nextLanguage) : undefined,\n      encoding,\n      sampleRate,\n      baseURL: lkBaseURL,\n      apiKey: lkApiKey,\n      apiSecret: lkApiSecret,\n      modelOptions,\n      fallback: normalizedFallback,\n      connOptions: connOptions ?? DEFAULT_API_CONNECT_OPTIONS,\n    };\n  }\n\n  get label(): string {\n    return 'inference.STT';\n  }\n\n  get model(): string {\n    return this.opts.model ?? 'auto';\n  }\n\n  get provider(): string {\n    return 'livekit';\n  }\n\n  static fromModelString(modelString: string): STT<AnyString> {\n    const [model, language] = parseSTTModelString(modelString);\n    return new STT({ model, language });\n  }\n\n  protected async _recognize(_: AudioBuffer): Promise<SpeechEvent> {\n    throw new Error('LiveKit STT does not support batch recognition, use stream() instead');\n  }\n\n  updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n    this.opts = {\n      ...this.opts,\n      ...opts,\n      language: opts.language !== undefined ? normalizeLanguage(opts.language) : this.opts.language,\n    };\n\n    for (const stream of this.streams) {\n      stream.updateOptions(opts);\n    }\n  }\n\n  stream(options?: {\n    language?: STTLanguages | string;\n    connOptions?: APIConnectOptions;\n  }): SpeechStream<TModel> {\n    const { language, connOptions = this.opts.connOptions ?? DEFAULT_API_CONNECT_OPTIONS } =\n      options || {};\n    const streamOpts = {\n      ...this.opts,\n      language: language !== undefined ? normalizeLanguage(language) : this.opts.language,\n    } as InferenceSTTOptions<TModel>;\n\n    const stream = new SpeechStream(this, streamOpts, connOptions);\n    this.streams.add(stream);\n\n    return stream;\n  }\n\n  async connectWs(timeout: number): Promise<WebSocket> {\n    const params = {\n      settings: {\n        sample_rate: String(this.opts.sampleRate),\n        encoding: this.opts.encoding,\n        extra: this.opts.modelOptions,\n      },\n    } as Record<string, unknown>;\n\n    if (this.opts.model && this.opts.model !== 'auto') {\n      params.model = this.opts.model;\n    }\n\n    if (this.opts.language) {\n      (params.settings as Record<string, unknown>).language = this.opts.language;\n    }\n\n    if (this.opts.fallback?.length) {\n      params.fallback = {\n        models: this.opts.fallback.map((m) => ({\n          model: m.model,\n          extra: m.extraKwargs ?? {},\n        })),\n      };\n    }\n\n    if (this.opts.connOptions) {\n      params.connection = {\n        timeout: this.opts.connOptions.timeoutMs / 1000,\n        retries: this.opts.connOptions.maxRetry,\n      };\n    }\n\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}/stt`;\n    const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n    const socket = await connectWs(url, headers, timeout);\n    const msg = { ...params, type: 'session.create' };\n    socket.send(JSON.stringify(msg));\n\n    return socket;\n  }\n}\n\nexport class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {\n  private opts: InferenceSTTOptions<TModel>;\n  private requestId = shortuuid('stt_request_');\n  private speaking = false;\n  private speechDuration = 0;\n  private reconnectEvent = new Event();\n  private stt: STT<TModel>;\n  private connOptions: APIConnectOptions;\n\n  #logger = log();\n\n  constructor(\n    sttImpl: STT<TModel>,\n    opts: InferenceSTTOptions<TModel>,\n    connOptions: APIConnectOptions,\n  ) {\n    super(sttImpl, opts.sampleRate, connOptions);\n    this.opts = opts;\n    this.stt = sttImpl;\n    this.connOptions = connOptions;\n  }\n\n  get label(): string {\n    return 'inference.SpeechStream';\n  }\n\n  updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n    this.opts = {\n      ...this.opts,\n      ...opts,\n      language: opts.language !== undefined ? normalizeLanguage(opts.language) : this.opts.language,\n    };\n    this.reconnectEvent.set();\n  }\n\n  protected async run(): Promise<void> {\n    while (true) {\n      // Create fresh resources for each connection attempt\n      let ws: WebSocket | null = null;\n      let closing = false;\n      let finalReceived = false;\n\n      const eventChannel = createStreamChannel<SttServerEvent>();\n\n      const resourceCleanup = () => {\n        if (closing) return;\n        closing = true;\n        eventChannel.close();\n        ws?.removeAllListeners();\n        ws?.close();\n      };\n\n      const createWsListener = async (ws: WebSocket, signal: AbortSignal) => {\n        return new Promise<void>((resolve, reject) => {\n          const onAbort = () => {\n            resourceCleanup();\n            reject(new Error('WebSocket connection aborted'));\n          };\n\n          signal.addEventListener('abort', onAbort, { once: true });\n\n          ws.on('message', (data) => {\n            const json = JSON.parse(data.toString()) as SttServerEvent;\n            eventChannel.write(json);\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', (code: number) => {\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: 'LiveKit STT connection closed unexpectedly',\n                options: { statusCode: code },\n              }),\n            );\n          });\n        });\n      };\n\n      const send = async (socket: WebSocket, signal: AbortSignal) => {\n        const audioStream = new AudioByteStream(\n          this.opts.sampleRate,\n          1,\n          Math.floor(this.opts.sampleRate / 20), // 50ms\n        );\n\n        // Create abort promise once to avoid memory leak\n        const abortPromise = new Promise<never>((_, reject) => {\n          if (signal.aborted) {\n            return reject(new Error('Send aborted'));\n          }\n          const onAbort = () => reject(new Error('Send aborted'));\n          signal.addEventListener('abort', onAbort, { once: true });\n        });\n\n        // Manual iteration to support cancellation\n        const iterator = this.input[Symbol.asyncIterator]();\n        try {\n          while (true) {\n            const result = await Promise.race([iterator.next(), abortPromise]);\n\n            if (result.done) break;\n            const ev = result.value;\n\n            let frames: AudioFrame[];\n            if (ev === SpeechStream.FLUSH_SENTINEL) {\n              frames = audioStream.flush();\n            } else {\n              const frame = ev as AudioFrame;\n              frames = audioStream.write(new Int16Array(frame.data).buffer);\n            }\n\n            for (const frame of frames) {\n              this.speechDuration += frame.samplesPerChannel / frame.sampleRate;\n              const base64 = Buffer.from(frame.data.buffer).toString('base64');\n              const msg = { type: 'input_audio', audio: base64 };\n              socket.send(JSON.stringify(msg));\n            }\n          }\n\n          closing = true;\n          socket.send(JSON.stringify({ type: 'session.finalize' }));\n        } catch (e) {\n          if ((e as Error).message === 'Send aborted') {\n            // Expected abort, don't log\n            return;\n          }\n          throw e;\n        }\n      };\n\n      const recv = async (signal: AbortSignal) => {\n        const serverEventStream = eventChannel.stream();\n        const reader = serverEventStream.getReader();\n\n        try {\n          while (!this.closed && !signal.aborted) {\n            const result = await reader.read();\n            if (signal.aborted) return;\n            if (result.done) return;\n\n            // Parse and validate with Zod schema\n            const parseResult = await sttServerEventSchema.safeParseAsync(result.value);\n            if (!parseResult.success) {\n              this.#logger.warn(\n                { error: parseResult.error, rawData: result.value },\n                'Failed to parse STT server event',\n              );\n              continue;\n            }\n\n            const event: SttServerEvent = parseResult.data;\n\n            switch (event.type) {\n              case 'session.created':\n              case 'session.finalized':\n                break;\n              case 'session.closed':\n                finalReceived = true;\n                resourceCleanup();\n                break;\n              case 'interim_transcript':\n                this.processTranscript(event, false);\n                break;\n              case 'final_transcript':\n                this.processTranscript(event, true);\n                break;\n              case 'error':\n                this.#logger.error({ error: event }, 'Received error from LiveKit STT');\n                resourceCleanup();\n                throw new APIError(`LiveKit STT returned error: ${JSON.stringify(event)}`);\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.stt.connectWs(this.connOptions.timeoutMs);\n\n        const controller = this.abortController; // Use base class abortController for proper cancellation\n        const sendTask = Task.from(({ signal }) => send(ws!, signal), controller);\n        const wsListenerTask = Task.from(({ signal }) => createWsListener(ws!, signal), controller);\n        const recvTask = Task.from(({ signal }) => recv(signal), controller);\n        const waitReconnectTask = Task.from(\n          ({ signal }) => Promise.race([this.reconnectEvent.wait(), waitForAbort(signal)]),\n          controller,\n        );\n\n        try {\n          await Promise.race([\n            Promise.all([sendTask.result, wsListenerTask.result, recvTask.result]),\n            waitReconnectTask.result,\n          ]);\n\n          // If reconnect didn't trigger, tasks finished - exit loop\n          if (!waitReconnectTask.done) break;\n\n          // Reconnect triggered - clear event and continue loop\n          this.reconnectEvent.clear();\n        } finally {\n          // Cancel all tasks to ensure cleanup\n          await cancelAndWait(\n            [sendTask, wsListenerTask, recvTask, waitReconnectTask],\n            DEFAULT_CANCEL_TIMEOUT,\n          );\n          resourceCleanup();\n        }\n      } finally {\n        // Ensure cleanup even if connectWs throws\n        resourceCleanup();\n      }\n    }\n  }\n\n  private processTranscript(data: SttTranscriptEvent, isFinal: boolean) {\n    // Check if queue is closed to avoid race condition during disconnect\n    if (this.queue.closed) return;\n\n    const requestId = data.session_id || this.requestId;\n    const text = data.transcript;\n    const language = normalizeLanguage(data.language || this.opts.language || 'en');\n\n    if (!text && !isFinal) return;\n\n    try {\n      // We'll have a more accurate way of detecting when speech started when we have VAD\n      if (!this.speaking) {\n        this.speaking = true;\n        this.queue.put({ type: SpeechEventType.START_OF_SPEECH });\n      }\n\n      const speechData: SpeechData = {\n        language,\n        startTime: this.startTimeOffset + data.start,\n        endTime: this.startTimeOffset + data.start + data.duration,\n        confidence: data.confidence,\n        text,\n        words: data.words.map(\n          (word): TimedString =>\n            createTimedString({\n              text: word.word,\n              startTime: word.start + this.startTimeOffset,\n              endTime: word.end + this.startTimeOffset,\n              startTimeOffset: this.startTimeOffset,\n              confidence: word.confidence,\n            }),\n        ),\n      };\n\n      if (isFinal) {\n        if (this.speechDuration > 0) {\n          this.queue.put({\n            type: SpeechEventType.RECOGNITION_USAGE,\n            requestId,\n            recognitionUsage: { audioDuration: this.speechDuration },\n          });\n          this.speechDuration = 0;\n        }\n\n        this.queue.put({\n          type: SpeechEventType.FINAL_TRANSCRIPT,\n          requestId,\n          alternatives: [speechData],\n        });\n\n        if (this.speaking) {\n          this.speaking = false;\n          this.queue.put({ type: SpeechEventType.END_OF_SPEECH });\n        }\n      } else {\n        this.queue.put({\n          type: SpeechEventType.INTERIM_TRANSCRIPT,\n          requestId,\n          alternatives: [speechData],\n        });\n      }\n    } catch (e) {\n      if (e instanceof Error && e.message.includes('Queue is closed')) {\n        // Expected behavior on disconnect, log as warning\n        this.#logger.warn(\n          { err: e },\n          'Queue closed during transcript processing (expected during disconnect)',\n        );\n      } else {\n        this.#logger.error({ err: e }, 'Error putting transcript to queue');\n      }\n    }\n  }\n}\n"],"mappings":"AAGA,eAAgC;AAEhC,SAAS,UAAU,sBAAsB;AACzC,SAAS,uBAAuB;AAChC,SAA4B,wBAAwB,yBAAyB;AAC7E,SAAS,WAAW;AACpB,SAAS,2BAA2B;AACpC;AAAA,EACE,OAAO;AAAA,EACP,gBAAgB;AAAA,EAGhB;AAAA,OACK;AACP,SAAiC,mCAAmC;AACpE,SAA2B,OAAO,MAAM,eAAe,WAAW,oBAAoB;AACtF,SAA2B,yBAAyB;AACpD;AAAA,EAGE;AAAA,OACK;AACP,SAAyB,WAAW,mBAAmB,8BAA8B;AAmG9E,SAAS,oBAAoB,OAAmD;AACrF,QAAM,MAAM,MAAM,YAAY,GAAG;AACjC,MAAI,QAAQ,IAAI;AACd,WAAO,CAAC,MAAM,MAAM,GAAG,GAAG,GAAG,kBAAkB,MAAM,MAAM,MAAM,CAAC,CAAC,CAAC;AAAA,EACtE;AACA,SAAO,CAAC,OAAO,MAAS;AAC1B;AAGO,SAAS,qBACd,UACoB;AACpB,QAAM,eAAe,CAAC,UAAkD;AACtE,QAAI,OAAO,UAAU,UAAU;AAC7B,YAAM,CAAC,IAAI,IAAI,oBAAoB,KAAK;AACxC,aAAO,EAAE,OAAO,KAAK;AAAA,IACvB;AACA,WAAO;AAAA,EACT;AAEA,MAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,WAAO,SAAS,IAAI,YAAY;AAAA,EAClC;AACA,SAAO,CAAC,aAAa,QAAQ,CAAC;AAChC;AAIA,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,yBAAyB;AAkBxB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAqC,oBAAI,IAAI;AAAA,EAErD,UAAU,IAAI;AAAA,EAEd,YAAY,MAWT;AACD,UAAM,EAAE,WAAW,MAAM,gBAAgB,MAAM,mBAAmB,OAAO,CAAC;AAE1E,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX,aAAa;AAAA,MACb;AAAA,MACA;AAAA,MACA,eAAe,CAAC;AAAA,MAChB;AAAA,MACA;AAAA,IACF,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,WAAW,uBAAuB;AACpD,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,eAAe;AACnB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,CAAC,aAAa,cAAc,IAAI,oBAAoB,SAAS;AACnE,UAAI,mBAAmB,QAAW;AAChC,YAAI,gBAAgB,CAAC,uBAAuB,cAAc,cAAc,GAAG;AACzE,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,UAAU,cAAc,OAAO,UAAU;AAAA,UAC7C;AAAA,QACF,OAAO;AACL,yBAAe;AAAA,QACjB;AACA,oBAAY;AAAA,MACd;AAAA,IACF;AACA,UAAM,qBAAqB,WAAW,qBAAqB,QAAQ,IAAI;AAEvE,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,UAAU,eAAe,kBAAkB,YAAY,IAAI;AAAA,MAC3D;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,MACA,UAAU;AAAA,MACV,aAAa,eAAe;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO,KAAK,KAAK,SAAS;AAAA,EAC5B;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,WAAW;AACzD,WAAO,IAAI,IAAI,EAAE,OAAO,SAAS,CAAC;AAAA,EACpC;AAAA,EAEA,MAAgB,WAAW,GAAsC;AAC/D,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO;AAAA,MACV,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UAAU,KAAK,aAAa,SAAY,kBAAkB,KAAK,QAAQ,IAAI,KAAK,KAAK;AAAA,IACvF;AAEA,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,OAAO,SAGkB;AACvB,UAAM,EAAE,UAAU,cAAc,KAAK,KAAK,eAAe,4BAA4B,IACnF,WAAW,CAAC;AACd,UAAM,aAAa;AAAA,MACjB,GAAG,KAAK;AAAA,MACR,UAAU,aAAa,SAAY,kBAAkB,QAAQ,IAAI,KAAK,KAAK;AAAA,IAC7E;AAEA,UAAM,SAAS,IAAI,aAAa,MAAM,YAAY,WAAW;AAC7D,SAAK,QAAQ,IAAI,MAAM;AAEvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AA5SvD;AA6SI,UAAM,SAAS;AAAA,MACb,UAAU;AAAA,QACR,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,QACxC,UAAU,KAAK,KAAK;AAAA,QACpB,OAAO,KAAK,KAAK;AAAA,MACnB;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,SAAS,KAAK,KAAK,UAAU,QAAQ;AACjD,aAAO,QAAQ,KAAK,KAAK;AAAA,IAC3B;AAEA,QAAI,KAAK,KAAK,UAAU;AACtB,MAAC,OAAO,SAAqC,WAAW,KAAK,KAAK;AAAA,IACpE;AAEA,SAAI,UAAK,KAAK,aAAV,mBAAoB,QAAQ;AAC9B,aAAO,WAAW;AAAA,QAChB,QAAQ,KAAK,KAAK,SAAS,IAAI,CAAC,OAAO;AAAA,UACrC,OAAO,EAAE;AAAA,UACT,OAAO,EAAE,eAAe,CAAC;AAAA,QAC3B,EAAE;AAAA,MACJ;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,aAAa;AACzB,aAAO,aAAa;AAAA,QAClB,SAAS,KAAK,KAAK,YAAY,YAAY;AAAA,QAC3C,SAAS,KAAK,KAAK,YAAY;AAAA,MACjC;AAAA,IACF;AAEA,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,MAAM,UAAU,KAAK,SAAS,OAAO;AACpD,UAAM,MAAM,EAAE,GAAG,QAAQ,MAAM,iBAAiB;AAChD,WAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAE/B,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAA+C,iBAAiB;AAAA,EACnE;AAAA,EACA,YAAY,UAAU,cAAc;AAAA,EACpC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,iBAAiB,IAAI,MAAM;AAAA,EAC3B;AAAA,EACA;AAAA,EAER,UAAU,IAAI;AAAA,EAEd,YACE,SACA,MACA,aACA;AACA,UAAM,SAAS,KAAK,YAAY,WAAW;AAC3C,SAAK,OAAO;AACZ,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO;AAAA,MACV,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UAAU,KAAK,aAAa,SAAY,kBAAkB,KAAK,QAAQ,IAAI,KAAK,KAAK;AAAA,IACvF;AACA,SAAK,eAAe,IAAI;AAAA,EAC1B;AAAA,EAEA,MAAgB,MAAqB;AACnC,WAAO,MAAM;AAEX,UAAI,KAAuB;AAC3B,UAAI,UAAU;AACd,UAAI,gBAAgB;AAEpB,YAAM,eAAe,oBAAoC;AAEzD,YAAM,kBAAkB,MAAM;AAC5B,YAAI,QAAS;AACb,kBAAU;AACV,qBAAa,MAAM;AACnB,iCAAI;AACJ,iCAAI;AAAA,MACN;AAEA,YAAM,mBAAmB,OAAOA,KAAe,WAAwB;AACrE,eAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,gBAAM,UAAU,MAAM;AACpB,4BAAgB;AAChB,mBAAO,IAAI,MAAM,8BAA8B,CAAC;AAAA,UAClD;AAEA,iBAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAExD,UAAAA,IAAG,GAAG,WAAW,CAAC,SAAS;AACzB,kBAAM,OAAO,KAAK,MAAM,KAAK,SAAS,CAAC;AACvC,yBAAa,MAAM,IAAI;AAAA,UACzB,CAAC;AAED,UAAAA,IAAG,GAAG,SAAS,CAAC,MAAM;AACpB,iBAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,4BAAgB;AAChB,mBAAO,CAAC;AAAA,UACV,CAAC;AAED,UAAAA,IAAG,GAAG,SAAS,CAAC,SAAiB;AAC/B,4BAAgB;AAEhB,gBAAI,CAAC,QAAS,QAAO,KAAK,QAAQ,MAAM,+BAA+B;AACvE,gBAAI,cAAe,QAAO,QAAQ;AAElC;AAAA,cACE,IAAI,eAAe;AAAA,gBACjB,SAAS;AAAA,gBACT,SAAS,EAAE,YAAY,KAAK;AAAA,cAC9B,CAAC;AAAA,YACH;AAAA,UACF,CAAC;AAAA,QACH,CAAC;AAAA,MACH;AAEA,YAAM,OAAO,OAAO,QAAmB,WAAwB;AAC7D,cAAM,cAAc,IAAI;AAAA,UACtB,KAAK,KAAK;AAAA,UACV;AAAA,UACA,KAAK,MAAM,KAAK,KAAK,aAAa,EAAE;AAAA;AAAA,QACtC;AAGA,cAAM,eAAe,IAAI,QAAe,CAAC,GAAG,WAAW;AACrD,cAAI,OAAO,SAAS;AAClB,mBAAO,OAAO,IAAI,MAAM,cAAc,CAAC;AAAA,UACzC;AACA,gBAAM,UAAU,MAAM,OAAO,IAAI,MAAM,cAAc,CAAC;AACtD,iBAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAAA,QAC1D,CAAC;AAGD,cAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAClD,YAAI;AACF,iBAAO,MAAM;AACX,kBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,SAAS,KAAK,GAAG,YAAY,CAAC;AAEjE,gBAAI,OAAO,KAAM;AACjB,kBAAM,KAAK,OAAO;AAElB,gBAAI;AACJ,gBAAI,OAAO,aAAa,gBAAgB;AACtC,uBAAS,YAAY,MAAM;AAAA,YAC7B,OAAO;AACL,oBAAM,QAAQ;AACd,uBAAS,YAAY,MAAM,IAAI,WAAW,MAAM,IAAI,EAAE,MAAM;AAAA,YAC9D;AAEA,uBAAW,SAAS,QAAQ;AAC1B,mBAAK,kBAAkB,MAAM,oBAAoB,MAAM;AACvD,oBAAM,SAAS,OAAO,KAAK,MAAM,KAAK,MAAM,EAAE,SAAS,QAAQ;AAC/D,oBAAM,MAAM,EAAE,MAAM,eAAe,OAAO,OAAO;AACjD,qBAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,YACjC;AAAA,UACF;AAEA,oBAAU;AACV,iBAAO,KAAK,KAAK,UAAU,EAAE,MAAM,mBAAmB,CAAC,CAAC;AAAA,QAC1D,SAAS,GAAG;AACV,cAAK,EAAY,YAAY,gBAAgB;AAE3C;AAAA,UACF;AACA,gBAAM;AAAA,QACR;AAAA,MACF;AAEA,YAAM,OAAO,OAAO,WAAwB;AAC1C,cAAM,oBAAoB,aAAa,OAAO;AAC9C,cAAM,SAAS,kBAAkB,UAAU;AAE3C,YAAI;AACF,iBAAO,CAAC,KAAK,UAAU,CAAC,OAAO,SAAS;AACtC,kBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,gBAAI,OAAO,QAAS;AACpB,gBAAI,OAAO,KAAM;AAGjB,kBAAM,cAAc,MAAM,qBAAqB,eAAe,OAAO,KAAK;AAC1E,gBAAI,CAAC,YAAY,SAAS;AACxB,mBAAK,QAAQ;AAAA,gBACX,EAAE,OAAO,YAAY,OAAO,SAAS,OAAO,MAAM;AAAA,gBAClD;AAAA,cACF;AACA;AAAA,YACF;AAEA,kBAAM,QAAwB,YAAY;AAE1C,oBAAQ,MAAM,MAAM;AAAA,cAClB,KAAK;AAAA,cACL,KAAK;AACH;AAAA,cACF,KAAK;AACH,gCAAgB;AAChB,gCAAgB;AAChB;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,OAAO,KAAK;AACnC;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,OAAO,IAAI;AAClC;AAAA,cACF,KAAK;AACH,qBAAK,QAAQ,MAAM,EAAE,OAAO,MAAM,GAAG,iCAAiC;AACtE,gCAAgB;AAChB,sBAAM,IAAI,SAAS,+BAA+B,KAAK,UAAU,KAAK,CAAC,EAAE;AAAA,YAC7E;AAAA,UACF;AAAA,QACF,UAAE;AACA,iBAAO,YAAY;AACnB,cAAI;AACF,kBAAM,kBAAkB,OAAO;AAAA,UACjC,SAAS,GAAG;AACV,iBAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,UACxF;AAAA,QACF;AAAA,MACF;AAEA,UAAI;AACF,aAAK,MAAM,KAAK,IAAI,UAAU,KAAK,YAAY,SAAS;AAExD,cAAM,aAAa,KAAK;AACxB,cAAM,WAAW,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,IAAK,MAAM,GAAG,UAAU;AACxE,cAAM,iBAAiB,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,iBAAiB,IAAK,MAAM,GAAG,UAAU;AAC1F,cAAM,WAAW,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,MAAM,GAAG,UAAU;AACnE,cAAM,oBAAoB,KAAK;AAAA,UAC7B,CAAC,EAAE,OAAO,MAAM,QAAQ,KAAK,CAAC,KAAK,eAAe,KAAK,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,UAC/E;AAAA,QACF;AAEA,YAAI;AACF,gBAAM,QAAQ,KAAK;AAAA,YACjB,QAAQ,IAAI,CAAC,SAAS,QAAQ,eAAe,QAAQ,SAAS,MAAM,CAAC;AAAA,YACrE,kBAAkB;AAAA,UACpB,CAAC;AAGD,cAAI,CAAC,kBAAkB,KAAM;AAG7B,eAAK,eAAe,MAAM;AAAA,QAC5B,UAAE;AAEA,gBAAM;AAAA,YACJ,CAAC,UAAU,gBAAgB,UAAU,iBAAiB;AAAA,YACtD;AAAA,UACF;AACA,0BAAgB;AAAA,QAClB;AAAA,MACF,UAAE;AAEA,wBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,kBAAkB,MAA0B,SAAkB;AAEpE,QAAI,KAAK,MAAM,OAAQ;AAEvB,UAAM,YAAY,KAAK,cAAc,KAAK;AAC1C,UAAM,OAAO,KAAK;AAClB,UAAM,WAAW,kBAAkB,KAAK,YAAY,KAAK,KAAK,YAAY,IAAI;AAE9E,QAAI,CAAC,QAAQ,CAAC,QAAS;AAEvB,QAAI;AAEF,UAAI,CAAC,KAAK,UAAU;AAClB,aAAK,WAAW;AAChB,aAAK,MAAM,IAAI,EAAE,MAAM,gBAAgB,gBAAgB,CAAC;AAAA,MAC1D;AAEA,YAAM,aAAyB;AAAA,QAC7B;AAAA,QACA,WAAW,KAAK,kBAAkB,KAAK;AAAA,QACvC,SAAS,KAAK,kBAAkB,KAAK,QAAQ,KAAK;AAAA,QAClD,YAAY,KAAK;AAAA,QACjB;AAAA,QACA,OAAO,KAAK,MAAM;AAAA,UAChB,CAAC,SACC,kBAAkB;AAAA,YAChB,MAAM,KAAK;AAAA,YACX,WAAW,KAAK,QAAQ,KAAK;AAAA,YAC7B,SAAS,KAAK,MAAM,KAAK;AAAA,YACzB,iBAAiB,KAAK;AAAA,YACtB,YAAY,KAAK;AAAA,UACnB,CAAC;AAAA,QACL;AAAA,MACF;AAEA,UAAI,SAAS;AACX,YAAI,KAAK,iBAAiB,GAAG;AAC3B,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,gBAAgB;AAAA,YACtB;AAAA,YACA,kBAAkB,EAAE,eAAe,KAAK,eAAe;AAAA,UACzD,CAAC;AACD,eAAK,iBAAiB;AAAA,QACxB;AAEA,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,gBAAgB;AAAA,UACtB;AAAA,UACA,cAAc,CAAC,UAAU;AAAA,QAC3B,CAAC;AAED,YAAI,KAAK,UAAU;AACjB,eAAK,WAAW;AAChB,eAAK,MAAM,IAAI,EAAE,MAAM,gBAAgB,cAAc,CAAC;AAAA,QACxD;AAAA,MACF,OAAO;AACL,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,gBAAgB;AAAA,UACtB;AAAA,UACA,cAAc,CAAC,UAAU;AAAA,QAC3B,CAAC;AAAA,MACH;AAAA,IACF,SAAS,GAAG;AACV,UAAI,aAAa,SAAS,EAAE,QAAQ,SAAS,iBAAiB,GAAG;AAE/D,aAAK,QAAQ;AAAA,UACX,EAAE,KAAK,EAAE;AAAA,UACT;AAAA,QACF;AAAA,MACF,OAAO;AACL,aAAK,QAAQ,MAAM,EAAE,KAAK,EAAE,GAAG,mCAAmC;AAAA,MACpE;AAAA,IACF;AAAA,EACF;AACF;","names":["ws"]}

package/dist/inference/stt.test.cjs

@@ -0,0 +1,218 @@
+"use strict";
+var import_vitest = require("vitest");
+var import_language = require("../language.cjs");
+var import_log = require("../log.cjs");
+var import_types = require("../types.cjs");
+var import_stt = require("./stt.cjs");
+(0, import_vitest.beforeAll)(() => {
+  (0, import_log.initializeLogger)({ level: "silent", pretty: false });
+});
+function makeStt(overrides = {}) {
+  const defaults = {
+    model: "deepgram",
+    apiKey: "test-key",
+    apiSecret: "test-secret",
+    baseURL: "https://example.livekit.cloud"
+  };
+  return new import_stt.STT({ ...defaults, ...overrides });
+}
+(0, import_vitest.describe)("parseSTTModelString", () => {
+  (0, import_vitest.it)("simple model without language", () => {
+    const [model, language] = (0, import_stt.parseSTTModelString)("deepgram");
+    (0, import_vitest.expect)(model).toBe("deepgram");
+    (0, import_vitest.expect)(language).toBeUndefined();
+  });
+  (0, import_vitest.it)("model with language suffix", () => {
+    const [model, language] = (0, import_stt.parseSTTModelString)("deepgram:en");
+    (0, import_vitest.expect)(model).toBe("deepgram");
+    (0, import_vitest.expect)(language).toBe("en");
+  });
+  (0, import_vitest.it)("normalizes language suffixes", () => {
+    const [model, language] = (0, import_stt.parseSTTModelString)("deepgram:english");
+    (0, import_vitest.expect)(model).toBe("deepgram");
+    (0, import_vitest.expect)(language).toBe("en");
+  });
+  (0, import_vitest.it)("provider/model format without language", () => {
+    const [model, language] = (0, import_stt.parseSTTModelString)("deepgram/nova-3");
+    (0, import_vitest.expect)(model).toBe("deepgram/nova-3");
+    (0, import_vitest.expect)(language).toBeUndefined();
+  });
+  (0, import_vitest.it)("provider/model format with language", () => {
+    const [model, language] = (0, import_stt.parseSTTModelString)("deepgram/nova-3:en");
+    (0, import_vitest.expect)(model).toBe("deepgram/nova-3");
+    (0, import_vitest.expect)(language).toBe("en");
+  });
+  import_vitest.it.each([
+    ["cartesia/ink-whisper:de", "cartesia/ink-whisper", "de"],
+    ["assemblyai:es", "assemblyai", "es"],
+    ["deepgram/nova-2-medical:ja", "deepgram/nova-2-medical", "ja"],
+    ["deepgram/nova-3:multi", "deepgram/nova-3", "multi"],
+    ["cartesia:zh", "cartesia", "zh"]
+  ])("various providers and languages: %s", (modelStr, expectedModel, expectedLang) => {
+    const [model, language] = (0, import_stt.parseSTTModelString)(modelStr);
+    (0, import_vitest.expect)(model).toBe(expectedModel);
+    (0, import_vitest.expect)(language).toBe(expectedLang);
+  });
+  (0, import_vitest.it)("auto model without language", () => {
+    const [model, language] = (0, import_stt.parseSTTModelString)("auto");
+    (0, import_vitest.expect)(model).toBe("auto");
+    (0, import_vitest.expect)(language).toBeUndefined();
+  });
+  (0, import_vitest.it)("auto model with language", () => {
+    const [model, language] = (0, import_stt.parseSTTModelString)("auto:pt");
+    (0, import_vitest.expect)(model).toBe("auto");
+    (0, import_vitest.expect)(language).toBe("pt");
+  });
+});
+(0, import_vitest.describe)("normalizeSTTFallback", () => {
+  (0, import_vitest.it)("single string model", () => {
+    const result = (0, import_stt.normalizeSTTFallback)("deepgram/nova-3");
+    (0, import_vitest.expect)(result).toEqual([{ model: "deepgram/nova-3" }]);
+  });
+  (0, import_vitest.it)("single FallbackModel dict", () => {
+    const fallback = { model: "deepgram/nova-3" };
+    const result = (0, import_stt.normalizeSTTFallback)(fallback);
+    (0, import_vitest.expect)(result).toEqual([{ model: "deepgram/nova-3" }]);
+  });
+  (0, import_vitest.it)("list of string models", () => {
+    const result = (0, import_stt.normalizeSTTFallback)(["deepgram/nova-3", "cartesia/ink-whisper"]);
+    (0, import_vitest.expect)(result).toEqual([{ model: "deepgram/nova-3" }, { model: "cartesia/ink-whisper" }]);
+  });
+  (0, import_vitest.it)("list of FallbackModel dicts", () => {
+    const fallbacks = [{ model: "deepgram/nova-3" }, { model: "assemblyai" }];
+    const result = (0, import_stt.normalizeSTTFallback)(fallbacks);
+    (0, import_vitest.expect)(result).toEqual([{ model: "deepgram/nova-3" }, { model: "assemblyai" }]);
+  });
+  (0, import_vitest.it)("mixed list of strings and dicts", () => {
+    const result = (0, import_stt.normalizeSTTFallback)([
+      "deepgram/nova-3",
+      { model: "cartesia/ink-whisper" },
+      "assemblyai"
+    ]);
+    (0, import_vitest.expect)(result).toEqual([
+      { model: "deepgram/nova-3" },
+      { model: "cartesia/ink-whisper" },
+      { model: "assemblyai" }
+    ]);
+  });
+  (0, import_vitest.it)("string with language suffix discards language", () => {
+    const result = (0, import_stt.normalizeSTTFallback)("deepgram/nova-3:en");
+    (0, import_vitest.expect)(result).toEqual([{ model: "deepgram/nova-3" }]);
+  });
+  (0, import_vitest.it)("FallbackModel with extraKwargs is preserved", () => {
+    const fallback = {
+      model: "deepgram/nova-3",
+      extraKwargs: { keywords: [["livekit", 1.5]], punctuate: true }
+    };
+    const result = (0, import_stt.normalizeSTTFallback)(fallback);
+    (0, import_vitest.expect)(result).toEqual([
+      {
+        model: "deepgram/nova-3",
+        extraKwargs: { keywords: [["livekit", 1.5]], punctuate: true }
+      }
+    ]);
+  });
+  (0, import_vitest.it)("list with extraKwargs preserved", () => {
+    const result = (0, import_stt.normalizeSTTFallback)([
+      { model: "deepgram/nova-3", extraKwargs: { punctuate: true } },
+      "cartesia/ink-whisper",
+      { model: "assemblyai", extraKwargs: { format_turns: true } }
+    ]);
+    (0, import_vitest.expect)(result).toEqual([
+      { model: "deepgram/nova-3", extraKwargs: { punctuate: true } },
+      { model: "cartesia/ink-whisper" },
+      { model: "assemblyai", extraKwargs: { format_turns: true } }
+    ]);
+  });
+  (0, import_vitest.it)("empty list returns empty list", () => {
+    const result = (0, import_stt.normalizeSTTFallback)([]);
+    (0, import_vitest.expect)(result).toEqual([]);
+  });
+  (0, import_vitest.it)("multiple colons in model string splits on last", () => {
+    const result = (0, import_stt.normalizeSTTFallback)("some:model:part:fr");
+    (0, import_vitest.expect)(result).toEqual([{ model: "some:model:part" }]);
+  });
+});
+(0, import_vitest.describe)("STT constructor fallback and connOptions", () => {
+  (0, import_vitest.it)("normalizes language in constructor and model string", () => {
+    const stt = makeStt({ model: "deepgram/nova-3:english" });
+    (0, import_vitest.expect)(stt["opts"].language).toBe("en");
+  });
+  (0, import_vitest.it)("prefers explicit normalized language over model suffix", () => {
+    const stt = makeStt({ model: "deepgram/nova-3:english", language: "en_US" });
+    (0, import_vitest.expect)(stt["opts"].language).toBe((0, import_language.normalizeLanguage)("en_US"));
+  });
+  (0, import_vitest.it)("fallback not given defaults to undefined", () => {
+    const stt = makeStt();
+    (0, import_vitest.expect)(stt["opts"].fallback).toBeUndefined();
+  });
+  (0, import_vitest.it)("fallback single string is normalized", () => {
+    const stt = makeStt({ fallback: "cartesia/ink-whisper" });
+    (0, import_vitest.expect)(stt["opts"].fallback).toEqual([{ model: "cartesia/ink-whisper" }]);
+  });
+  (0, import_vitest.it)("fallback list of strings is normalized", () => {
+    const stt = makeStt({ fallback: ["deepgram/nova-3", "assemblyai"] });
+    (0, import_vitest.expect)(stt["opts"].fallback).toEqual([{ model: "deepgram/nova-3" }, { model: "assemblyai" }]);
+  });
+  (0, import_vitest.it)("fallback single FallbackModel is normalized to list", () => {
+    const stt = makeStt({ fallback: { model: "deepgram/nova-3" } });
+    (0, import_vitest.expect)(stt["opts"].fallback).toEqual([{ model: "deepgram/nova-3" }]);
+  });
+  (0, import_vitest.it)("fallback with extraKwargs is preserved", () => {
+    const stt = makeStt({
+      fallback: {
+        model: "deepgram/nova-3",
+        extraKwargs: { punctuate: true, keywords: [["livekit", 1.5]] }
+      }
+    });
+    (0, import_vitest.expect)(stt["opts"].fallback).toEqual([
+      {
+        model: "deepgram/nova-3",
+        extraKwargs: { punctuate: true, keywords: [["livekit", 1.5]] }
+      }
+    ]);
+  });
+  (0, import_vitest.it)("fallback mixed list is normalized", () => {
+    const stt = makeStt({
+      fallback: [
+        "deepgram/nova-3",
+        { model: "cartesia", extraKwargs: { min_volume: 0.5 } },
+        "assemblyai"
+      ]
+    });
+    (0, import_vitest.expect)(stt["opts"].fallback).toEqual([
+      { model: "deepgram/nova-3" },
+      { model: "cartesia", extraKwargs: { min_volume: 0.5 } },
+      { model: "assemblyai" }
+    ]);
+  });
+  (0, import_vitest.it)("fallback string with language discards language", () => {
+    const stt = makeStt({ fallback: "deepgram/nova-3:en" });
+    (0, import_vitest.expect)(stt["opts"].fallback).toEqual([{ model: "deepgram/nova-3" }]);
+  });
+  (0, import_vitest.it)("connOptions not given uses default", () => {
+    const stt = makeStt();
+    (0, import_vitest.expect)(stt["opts"].connOptions).toEqual(import_types.DEFAULT_API_CONNECT_OPTIONS);
+  });
+  (0, import_vitest.it)("connOptions custom timeout", () => {
+    const custom = { timeoutMs: 3e4, maxRetry: 3, retryIntervalMs: 2e3 };
+    const stt = makeStt({ connOptions: custom });
+    (0, import_vitest.expect)(stt["opts"].connOptions).toEqual(custom);
+    (0, import_vitest.expect)(stt["opts"].connOptions.timeoutMs).toBe(3e4);
+  });
+  (0, import_vitest.it)("connOptions custom maxRetry", () => {
+    const custom = { timeoutMs: 1e4, maxRetry: 5, retryIntervalMs: 2e3 };
+    const stt = makeStt({ connOptions: custom });
+    (0, import_vitest.expect)(stt["opts"].connOptions).toEqual(custom);
+    (0, import_vitest.expect)(stt["opts"].connOptions.maxRetry).toBe(5);
+  });
+  (0, import_vitest.it)("connOptions full custom", () => {
+    const custom = { timeoutMs: 6e4, maxRetry: 10, retryIntervalMs: 2e3 };
+    const stt = makeStt({ connOptions: custom });
+    (0, import_vitest.expect)(stt["opts"].connOptions).toEqual(custom);
+    (0, import_vitest.expect)(stt["opts"].connOptions.timeoutMs).toBe(6e4);
+    (0, import_vitest.expect)(stt["opts"].connOptions.maxRetry).toBe(10);
+    (0, import_vitest.expect)(stt["opts"].connOptions.retryIntervalMs).toBe(2e3);
+  });
+});
+//# sourceMappingURL=stt.test.cjs.map

package/dist/inference/stt.test.cjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/inference/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { beforeAll, describe, expect, it } from 'vitest';\nimport { normalizeLanguage } from '../language.js';\nimport { initializeLogger } from '../log.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { STT, type STTFallbackModel, normalizeSTTFallback, parseSTTModelString } from './stt.js';\n\nbeforeAll(() => {\n  initializeLogger({ level: 'silent', pretty: false });\n});\n\n/** Helper to create STT with required credentials. */\nfunction makeStt(overrides: Record<string, unknown> = {}) {\n  const defaults = {\n    model: 'deepgram' as const,\n    apiKey: 'test-key',\n    apiSecret: 'test-secret',\n    baseURL: 'https://example.livekit.cloud',\n  };\n  return new STT({ ...defaults, ...overrides });\n}\n\ndescribe('parseSTTModelString', () => {\n  it('simple model without language', () => {\n    const [model, language] = parseSTTModelString('deepgram');\n    expect(model).toBe('deepgram');\n    expect(language).toBeUndefined();\n  });\n\n  it('model with language suffix', () => {\n    const [model, language] = parseSTTModelString('deepgram:en');\n    expect(model).toBe('deepgram');\n    expect(language).toBe('en');\n  });\n\n  it('normalizes language suffixes', () => {\n    const [model, language] = parseSTTModelString('deepgram:english');\n    expect(model).toBe('deepgram');\n    expect(language).toBe('en');\n  });\n\n  it('provider/model format without language', () => {\n    const [model, language] = parseSTTModelString('deepgram/nova-3');\n    expect(model).toBe('deepgram/nova-3');\n    expect(language).toBeUndefined();\n  });\n\n  it('provider/model format with language', () => {\n    const [model, language] = parseSTTModelString('deepgram/nova-3:en');\n    expect(model).toBe('deepgram/nova-3');\n    expect(language).toBe('en');\n  });\n\n  it.each([\n    ['cartesia/ink-whisper:de', 'cartesia/ink-whisper', 'de'],\n    ['assemblyai:es', 'assemblyai', 'es'],\n    ['deepgram/nova-2-medical:ja', 'deepgram/nova-2-medical', 'ja'],\n    ['deepgram/nova-3:multi', 'deepgram/nova-3', 'multi'],\n    ['cartesia:zh', 'cartesia', 'zh'],\n  ])('various providers and languages: %s', (modelStr, expectedModel, expectedLang) => {\n    const [model, language] = parseSTTModelString(modelStr);\n    expect(model).toBe(expectedModel);\n    expect(language).toBe(expectedLang);\n  });\n\n  it('auto model without language', () => {\n    const [model, language] = parseSTTModelString('auto');\n    expect(model).toBe('auto');\n    expect(language).toBeUndefined();\n  });\n\n  it('auto model with language', () => {\n    const [model, language] = parseSTTModelString('auto:pt');\n    expect(model).toBe('auto');\n    expect(language).toBe('pt');\n  });\n});\n\ndescribe('normalizeSTTFallback', () => {\n  it('single string model', () => {\n    const result = normalizeSTTFallback('deepgram/nova-3');\n    expect(result).toEqual([{ model: 'deepgram/nova-3' }]);\n  });\n\n  it('single FallbackModel dict', () => {\n    const fallback: STTFallbackModel = { model: 'deepgram/nova-3' };\n    const result = normalizeSTTFallback(fallback);\n    expect(result).toEqual([{ model: 'deepgram/nova-3' }]);\n  });\n\n  it('list of string models', () => {\n    const result = normalizeSTTFallback(['deepgram/nova-3', 'cartesia/ink-whisper']);\n    expect(result).toEqual([{ model: 'deepgram/nova-3' }, { model: 'cartesia/ink-whisper' }]);\n  });\n\n  it('list of FallbackModel dicts', () => {\n    const fallbacks: STTFallbackModel[] = [{ model: 'deepgram/nova-3' }, { model: 'assemblyai' }];\n    const result = normalizeSTTFallback(fallbacks);\n    expect(result).toEqual([{ model: 'deepgram/nova-3' }, { model: 'assemblyai' }]);\n  });\n\n  it('mixed list of strings and dicts', () => {\n    const result = normalizeSTTFallback([\n      'deepgram/nova-3',\n      { model: 'cartesia/ink-whisper' } as STTFallbackModel,\n      'assemblyai',\n    ]);\n    expect(result).toEqual([\n      { model: 'deepgram/nova-3' },\n      { model: 'cartesia/ink-whisper' },\n      { model: 'assemblyai' },\n    ]);\n  });\n\n  it('string with language suffix discards language', () => {\n    const result = normalizeSTTFallback('deepgram/nova-3:en');\n    expect(result).toEqual([{ model: 'deepgram/nova-3' }]);\n  });\n\n  it('FallbackModel with extraKwargs is preserved', () => {\n    const fallback: STTFallbackModel = {\n      model: 'deepgram/nova-3',\n      extraKwargs: { keywords: [['livekit', 1.5]], punctuate: true },\n    };\n    const result = normalizeSTTFallback(fallback);\n    expect(result).toEqual([\n      {\n        model: 'deepgram/nova-3',\n        extraKwargs: { keywords: [['livekit', 1.5]], punctuate: true },\n      },\n    ]);\n  });\n\n  it('list with extraKwargs preserved', () => {\n    const result = normalizeSTTFallback([\n      { model: 'deepgram/nova-3', extraKwargs: { punctuate: true } } as STTFallbackModel,\n      'cartesia/ink-whisper',\n      { model: 'assemblyai', extraKwargs: { format_turns: true } } as STTFallbackModel,\n    ]);\n    expect(result).toEqual([\n      { model: 'deepgram/nova-3', extraKwargs: { punctuate: true } },\n      { model: 'cartesia/ink-whisper' },\n      { model: 'assemblyai', extraKwargs: { format_turns: true } },\n    ]);\n  });\n\n  it('empty list returns empty list', () => {\n    const result = normalizeSTTFallback([]);\n    expect(result).toEqual([]);\n  });\n\n  it('multiple colons in model string splits on last', () => {\n    const result = normalizeSTTFallback('some:model:part:fr');\n    expect(result).toEqual([{ model: 'some:model:part' }]);\n  });\n});\n\ndescribe('STT constructor fallback and connOptions', () => {\n  it('normalizes language in constructor and model string', () => {\n    const stt = makeStt({ model: 'deepgram/nova-3:english' });\n    expect(stt['opts'].language).toBe('en');\n  });\n\n  it('prefers explicit normalized language over model suffix', () => {\n    const stt = makeStt({ model: 'deepgram/nova-3:english', language: 'en_US' });\n    expect(stt['opts'].language).toBe(normalizeLanguage('en_US'));\n  });\n\n  it('fallback not given defaults to undefined', () => {\n    const stt = makeStt();\n    expect(stt['opts'].fallback).toBeUndefined();\n  });\n\n  it('fallback single string is normalized', () => {\n    const stt = makeStt({ fallback: 'cartesia/ink-whisper' });\n    expect(stt['opts'].fallback).toEqual([{ model: 'cartesia/ink-whisper' }]);\n  });\n\n  it('fallback list of strings is normalized', () => {\n    const stt = makeStt({ fallback: ['deepgram/nova-3', 'assemblyai'] });\n    expect(stt['opts'].fallback).toEqual([{ model: 'deepgram/nova-3' }, { model: 'assemblyai' }]);\n  });\n\n  it('fallback single FallbackModel is normalized to list', () => {\n    const stt = makeStt({ fallback: { model: 'deepgram/nova-3' } });\n    expect(stt['opts'].fallback).toEqual([{ model: 'deepgram/nova-3' }]);\n  });\n\n  it('fallback with extraKwargs is preserved', () => {\n    const stt = makeStt({\n      fallback: {\n        model: 'deepgram/nova-3',\n        extraKwargs: { punctuate: true, keywords: [['livekit', 1.5]] },\n      },\n    });\n    expect(stt['opts'].fallback).toEqual([\n      {\n        model: 'deepgram/nova-3',\n        extraKwargs: { punctuate: true, keywords: [['livekit', 1.5]] },\n      },\n    ]);\n  });\n\n  it('fallback mixed list is normalized', () => {\n    const stt = makeStt({\n      fallback: [\n        'deepgram/nova-3',\n        { model: 'cartesia', extraKwargs: { min_volume: 0.5 } },\n        'assemblyai',\n      ],\n    });\n    expect(stt['opts'].fallback).toEqual([\n      { model: 'deepgram/nova-3' },\n      { model: 'cartesia', extraKwargs: { min_volume: 0.5 } },\n      { model: 'assemblyai' },\n    ]);\n  });\n\n  it('fallback string with language discards language', () => {\n    const stt = makeStt({ fallback: 'deepgram/nova-3:en' });\n    expect(stt['opts'].fallback).toEqual([{ model: 'deepgram/nova-3' }]);\n  });\n\n  it('connOptions not given uses default', () => {\n    const stt = makeStt();\n    expect(stt['opts'].connOptions).toEqual(DEFAULT_API_CONNECT_OPTIONS);\n  });\n\n  it('connOptions custom timeout', () => {\n    const custom: APIConnectOptions = { timeoutMs: 30000, maxRetry: 3, retryIntervalMs: 2000 };\n    const stt = makeStt({ connOptions: custom });\n    expect(stt['opts'].connOptions).toEqual(custom);\n    expect(stt['opts'].connOptions!.timeoutMs).toBe(30000);\n  });\n\n  it('connOptions custom maxRetry', () => {\n    const custom: APIConnectOptions = { timeoutMs: 10000, maxRetry: 5, retryIntervalMs: 2000 };\n    const stt = makeStt({ connOptions: custom });\n    expect(stt['opts'].connOptions).toEqual(custom);\n    expect(stt['opts'].connOptions!.maxRetry).toBe(5);\n  });\n\n  it('connOptions full custom', () => {\n    const custom: APIConnectOptions = { timeoutMs: 60000, maxRetry: 10, retryIntervalMs: 2000 };\n    const stt = makeStt({ connOptions: custom });\n    expect(stt['opts'].connOptions).toEqual(custom);\n    expect(stt['opts'].connOptions!.timeoutMs).toBe(60000);\n    expect(stt['opts'].connOptions!.maxRetry).toBe(10);\n    expect(stt['opts'].connOptions!.retryIntervalMs).toBe(2000);\n  });\n});\n"],"mappings":";AAGA,oBAAgD;AAChD,sBAAkC;AAClC,iBAAiC;AACjC,mBAAoE;AACpE,iBAAsF;AAAA,IAEtF,yBAAU,MAAM;AACd,mCAAiB,EAAE,OAAO,UAAU,QAAQ,MAAM,CAAC;AACrD,CAAC;AAGD,SAAS,QAAQ,YAAqC,CAAC,GAAG;AACxD,QAAM,WAAW;AAAA,IACf,OAAO;AAAA,IACP,QAAQ;AAAA,IACR,WAAW;AAAA,IACX,SAAS;AAAA,EACX;AACA,SAAO,IAAI,eAAI,EAAE,GAAG,UAAU,GAAG,UAAU,CAAC;AAC9C;AAAA,IAEA,wBAAS,uBAAuB,MAAM;AACpC,wBAAG,iCAAiC,MAAM;AACxC,UAAM,CAAC,OAAO,QAAQ,QAAI,gCAAoB,UAAU;AACxD,8BAAO,KAAK,EAAE,KAAK,UAAU;AAC7B,8BAAO,QAAQ,EAAE,cAAc;AAAA,EACjC,CAAC;AAED,wBAAG,8BAA8B,MAAM;AACrC,UAAM,CAAC,OAAO,QAAQ,QAAI,gCAAoB,aAAa;AAC3D,8BAAO,KAAK,EAAE,KAAK,UAAU;AAC7B,8BAAO,QAAQ,EAAE,KAAK,IAAI;AAAA,EAC5B,CAAC;AAED,wBAAG,gCAAgC,MAAM;AACvC,UAAM,CAAC,OAAO,QAAQ,QAAI,gCAAoB,kBAAkB;AAChE,8BAAO,KAAK,EAAE,KAAK,UAAU;AAC7B,8BAAO,QAAQ,EAAE,KAAK,IAAI;AAAA,EAC5B,CAAC;AAED,wBAAG,0CAA0C,MAAM;AACjD,UAAM,CAAC,OAAO,QAAQ,QAAI,gCAAoB,iBAAiB;AAC/D,8BAAO,KAAK,EAAE,KAAK,iBAAiB;AACpC,8BAAO,QAAQ,EAAE,cAAc;AAAA,EACjC,CAAC;AAED,wBAAG,uCAAuC,MAAM;AAC9C,UAAM,CAAC,OAAO,QAAQ,QAAI,gCAAoB,oBAAoB;AAClE,8BAAO,KAAK,EAAE,KAAK,iBAAiB;AACpC,8BAAO,QAAQ,EAAE,KAAK,IAAI;AAAA,EAC5B,CAAC;AAED,mBAAG,KAAK;AAAA,IACN,CAAC,2BAA2B,wBAAwB,IAAI;AAAA,IACxD,CAAC,iBAAiB,cAAc,IAAI;AAAA,IACpC,CAAC,8BAA8B,2BAA2B,IAAI;AAAA,IAC9D,CAAC,yBAAyB,mBAAmB,OAAO;AAAA,IACpD,CAAC,eAAe,YAAY,IAAI;AAAA,EAClC,CAAC,EAAE,uCAAuC,CAAC,UAAU,eAAe,iBAAiB;AACnF,UAAM,CAAC,OAAO,QAAQ,QAAI,gCAAoB,QAAQ;AACtD,8BAAO,KAAK,EAAE,KAAK,aAAa;AAChC,8BAAO,QAAQ,EAAE,KAAK,YAAY;AAAA,EACpC,CAAC;AAED,wBAAG,+BAA+B,MAAM;AACtC,UAAM,CAAC,OAAO,QAAQ,QAAI,gCAAoB,MAAM;AACpD,8BAAO,KAAK,EAAE,KAAK,MAAM;AACzB,8BAAO,QAAQ,EAAE,cAAc;AAAA,EACjC,CAAC;AAED,wBAAG,4BAA4B,MAAM;AACnC,UAAM,CAAC,OAAO,QAAQ,QAAI,gCAAoB,SAAS;AACvD,8BAAO,KAAK,EAAE,KAAK,MAAM;AACzB,8BAAO,QAAQ,EAAE,KAAK,IAAI;AAAA,EAC5B,CAAC;AACH,CAAC;AAAA,IAED,wBAAS,wBAAwB,MAAM;AACrC,wBAAG,uBAAuB,MAAM;AAC9B,UAAM,aAAS,iCAAqB,iBAAiB;AACrD,8BAAO,MAAM,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,CAAC,CAAC;AAAA,EACvD,CAAC;AAED,wBAAG,6BAA6B,MAAM;AACpC,UAAM,WAA6B,EAAE,OAAO,kBAAkB;AAC9D,UAAM,aAAS,iCAAqB,QAAQ;AAC5C,8BAAO,MAAM,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,CAAC,CAAC;AAAA,EACvD,CAAC;AAED,wBAAG,yBAAyB,MAAM;AAChC,UAAM,aAAS,iCAAqB,CAAC,mBAAmB,sBAAsB,CAAC;AAC/E,8BAAO,MAAM,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,GAAG,EAAE,OAAO,uBAAuB,CAAC,CAAC;AAAA,EAC1F,CAAC;AAED,wBAAG,+BAA+B,MAAM;AACtC,UAAM,YAAgC,CAAC,EAAE,OAAO,kBAAkB,GAAG,EAAE,OAAO,aAAa,CAAC;AAC5F,UAAM,aAAS,iCAAqB,SAAS;AAC7C,8BAAO,MAAM,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,GAAG,EAAE,OAAO,aAAa,CAAC,CAAC;AAAA,EAChF,CAAC;AAED,wBAAG,mCAAmC,MAAM;AAC1C,UAAM,aAAS,iCAAqB;AAAA,MAClC;AAAA,MACA,EAAE,OAAO,uBAAuB;AAAA,MAChC;AAAA,IACF,CAAC;AACD,8BAAO,MAAM,EAAE,QAAQ;AAAA,MACrB,EAAE,OAAO,kBAAkB;AAAA,MAC3B,EAAE,OAAO,uBAAuB;AAAA,MAChC,EAAE,OAAO,aAAa;AAAA,IACxB,CAAC;AAAA,EACH,CAAC;AAED,wBAAG,iDAAiD,MAAM;AACxD,UAAM,aAAS,iCAAqB,oBAAoB;AACxD,8BAAO,MAAM,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,CAAC,CAAC;AAAA,EACvD,CAAC;AAED,wBAAG,+CAA+C,MAAM;AACtD,UAAM,WAA6B;AAAA,MACjC,OAAO;AAAA,MACP,aAAa,EAAE,UAAU,CAAC,CAAC,WAAW,GAAG,CAAC,GAAG,WAAW,KAAK;AAAA,IAC/D;AACA,UAAM,aAAS,iCAAqB,QAAQ;AAC5C,8BAAO,MAAM,EAAE,QAAQ;AAAA,MACrB;AAAA,QACE,OAAO;AAAA,QACP,aAAa,EAAE,UAAU,CAAC,CAAC,WAAW,GAAG,CAAC,GAAG,WAAW,KAAK;AAAA,MAC/D;AAAA,IACF,CAAC;AAAA,EACH,CAAC;AAED,wBAAG,mCAAmC,MAAM;AAC1C,UAAM,aAAS,iCAAqB;AAAA,MAClC,EAAE,OAAO,mBAAmB,aAAa,EAAE,WAAW,KAAK,EAAE;AAAA,MAC7D;AAAA,MACA,EAAE,OAAO,cAAc,aAAa,EAAE,cAAc,KAAK,EAAE;AAAA,IAC7D,CAAC;AACD,8BAAO,MAAM,EAAE,QAAQ;AAAA,MACrB,EAAE,OAAO,mBAAmB,aAAa,EAAE,WAAW,KAAK,EAAE;AAAA,MAC7D,EAAE,OAAO,uBAAuB;AAAA,MAChC,EAAE,OAAO,cAAc,aAAa,EAAE,cAAc,KAAK,EAAE;AAAA,IAC7D,CAAC;AAAA,EACH,CAAC;AAED,wBAAG,iCAAiC,MAAM;AACxC,UAAM,aAAS,iCAAqB,CAAC,CAAC;AACtC,8BAAO,MAAM,EAAE,QAAQ,CAAC,CAAC;AAAA,EAC3B,CAAC;AAED,wBAAG,kDAAkD,MAAM;AACzD,UAAM,aAAS,iCAAqB,oBAAoB;AACxD,8BAAO,MAAM,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,CAAC,CAAC;AAAA,EACvD,CAAC;AACH,CAAC;AAAA,IAED,wBAAS,4CAA4C,MAAM;AACzD,wBAAG,uDAAuD,MAAM;AAC9D,UAAM,MAAM,QAAQ,EAAE,OAAO,0BAA0B,CAAC;AACxD,8BAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,KAAK,IAAI;AAAA,EACxC,CAAC;AAED,wBAAG,0DAA0D,MAAM;AACjE,UAAM,MAAM,QAAQ,EAAE,OAAO,2BAA2B,UAAU,QAAQ,CAAC;AAC3E,8BAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,SAAK,mCAAkB,OAAO,CAAC;AAAA,EAC9D,CAAC;AAED,wBAAG,4CAA4C,MAAM;AACnD,UAAM,MAAM,QAAQ;AACpB,8BAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,cAAc;AAAA,EAC7C,CAAC;AAED,wBAAG,wCAAwC,MAAM;AAC/C,UAAM,MAAM,QAAQ,EAAE,UAAU,uBAAuB,CAAC;AACxD,8BAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,QAAQ,CAAC,EAAE,OAAO,uBAAuB,CAAC,CAAC;AAAA,EAC1E,CAAC;AAED,wBAAG,0CAA0C,MAAM;AACjD,UAAM,MAAM,QAAQ,EAAE,UAAU,CAAC,mBAAmB,YAAY,EAAE,CAAC;AACnE,8BAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,GAAG,EAAE,OAAO,aAAa,CAAC,CAAC;AAAA,EAC9F,CAAC;AAED,wBAAG,uDAAuD,MAAM;AAC9D,UAAM,MAAM,QAAQ,EAAE,UAAU,EAAE,OAAO,kBAAkB,EAAE,CAAC;AAC9D,8BAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,CAAC,CAAC;AAAA,EACrE,CAAC;AAED,wBAAG,0CAA0C,MAAM;AACjD,UAAM,MAAM,QAAQ;AAAA,MAClB,UAAU;AAAA,QACR,OAAO;AAAA,QACP,aAAa,EAAE,WAAW,MAAM,UAAU,CAAC,CAAC,WAAW,GAAG,CAAC,EAAE;AAAA,MAC/D;AAAA,IACF,CAAC;AACD,8BAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,QAAQ;AAAA,MACnC;AAAA,QACE,OAAO;AAAA,QACP,aAAa,EAAE,WAAW,MAAM,UAAU,CAAC,CAAC,WAAW,GAAG,CAAC,EAAE;AAAA,MAC/D;AAAA,IACF,CAAC;AAAA,EACH,CAAC;AAED,wBAAG,qCAAqC,MAAM;AAC5C,UAAM,MAAM,QAAQ;AAAA,MAClB,UAAU;AAAA,QACR;AAAA,QACA,EAAE,OAAO,YAAY,aAAa,EAAE,YAAY,IAAI,EAAE;AAAA,QACtD;AAAA,MACF;AAAA,IACF,CAAC;AACD,8BAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,QAAQ;AAAA,MACnC,EAAE,OAAO,kBAAkB;AAAA,MAC3B,EAAE,OAAO,YAAY,aAAa,EAAE,YAAY,IAAI,EAAE;AAAA,MACtD,EAAE,OAAO,aAAa;AAAA,IACxB,CAAC;AAAA,EACH,CAAC;AAED,wBAAG,mDAAmD,MAAM;AAC1D,UAAM,MAAM,QAAQ,EAAE,UAAU,qBAAqB,CAAC;AACtD,8BAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,CAAC,CAAC;AAAA,EACrE,CAAC;AAED,wBAAG,sCAAsC,MAAM;AAC7C,UAAM,MAAM,QAAQ;AACpB,8BAAO,IAAI,MAAM,EAAE,WAAW,EAAE,QAAQ,wCAA2B;AAAA,EACrE,CAAC;AAED,wBAAG,8BAA8B,MAAM;AACrC,UAAM,SAA4B,EAAE,WAAW,KAAO,UAAU,GAAG,iBAAiB,IAAK;AACzF,UAAM,MAAM,QAAQ,EAAE,aAAa,OAAO,CAAC;AAC3C,8BAAO,IAAI,MAAM,EAAE,WAAW,EAAE,QAAQ,MAAM;AAC9C,8BAAO,IAAI,MAAM,EAAE,YAAa,SAAS,EAAE,KAAK,GAAK;AAAA,EACvD,CAAC;AAED,wBAAG,+BAA+B,MAAM;AACtC,UAAM,SAA4B,EAAE,WAAW,KAAO,UAAU,GAAG,iBAAiB,IAAK;AACzF,UAAM,MAAM,QAAQ,EAAE,aAAa,OAAO,CAAC;AAC3C,8BAAO,IAAI,MAAM,EAAE,WAAW,EAAE,QAAQ,MAAM;AAC9C,8BAAO,IAAI,MAAM,EAAE,YAAa,QAAQ,EAAE,KAAK,CAAC;AAAA,EAClD,CAAC;AAED,wBAAG,2BAA2B,MAAM;AAClC,UAAM,SAA4B,EAAE,WAAW,KAAO,UAAU,IAAI,iBAAiB,IAAK;AAC1F,UAAM,MAAM,QAAQ,EAAE,aAAa,OAAO,CAAC;AAC3C,8BAAO,IAAI,MAAM,EAAE,WAAW,EAAE,QAAQ,MAAM;AAC9C,8BAAO,IAAI,MAAM,EAAE,YAAa,SAAS,EAAE,KAAK,GAAK;AACrD,8BAAO,IAAI,MAAM,EAAE,YAAa,QAAQ,EAAE,KAAK,EAAE;AACjD,8BAAO,IAAI,MAAM,EAAE,YAAa,eAAe,EAAE,KAAK,GAAI;AAAA,EAC5D,CAAC;AACH,CAAC;","names":[]}

package/dist/inference/stt.test.js

@@ -0,0 +1,217 @@
+import { beforeAll, describe, expect, it } from "vitest";
+import { normalizeLanguage } from "../language.js";
+import { initializeLogger } from "../log.js";
+import { DEFAULT_API_CONNECT_OPTIONS } from "../types.js";
+import { STT, normalizeSTTFallback, parseSTTModelString } from "./stt.js";
+beforeAll(() => {
+  initializeLogger({ level: "silent", pretty: false });
+});
+function makeStt(overrides = {}) {
+  const defaults = {
+    model: "deepgram",
+    apiKey: "test-key",
+    apiSecret: "test-secret",
+    baseURL: "https://example.livekit.cloud"
+  };
+  return new STT({ ...defaults, ...overrides });
+}
+describe("parseSTTModelString", () => {
+  it("simple model without language", () => {
+    const [model, language] = parseSTTModelString("deepgram");
+    expect(model).toBe("deepgram");
+    expect(language).toBeUndefined();
+  });
+  it("model with language suffix", () => {
+    const [model, language] = parseSTTModelString("deepgram:en");
+    expect(model).toBe("deepgram");
+    expect(language).toBe("en");
+  });
+  it("normalizes language suffixes", () => {
+    const [model, language] = parseSTTModelString("deepgram:english");
+    expect(model).toBe("deepgram");
+    expect(language).toBe("en");
+  });
+  it("provider/model format without language", () => {
+    const [model, language] = parseSTTModelString("deepgram/nova-3");
+    expect(model).toBe("deepgram/nova-3");
+    expect(language).toBeUndefined();
+  });
+  it("provider/model format with language", () => {
+    const [model, language] = parseSTTModelString("deepgram/nova-3:en");
+    expect(model).toBe("deepgram/nova-3");
+    expect(language).toBe("en");
+  });
+  it.each([
+    ["cartesia/ink-whisper:de", "cartesia/ink-whisper", "de"],
+    ["assemblyai:es", "assemblyai", "es"],
+    ["deepgram/nova-2-medical:ja", "deepgram/nova-2-medical", "ja"],
+    ["deepgram/nova-3:multi", "deepgram/nova-3", "multi"],
+    ["cartesia:zh", "cartesia", "zh"]
+  ])("various providers and languages: %s", (modelStr, expectedModel, expectedLang) => {
+    const [model, language] = parseSTTModelString(modelStr);
+    expect(model).toBe(expectedModel);
+    expect(language).toBe(expectedLang);
+  });
+  it("auto model without language", () => {
+    const [model, language] = parseSTTModelString("auto");
+    expect(model).toBe("auto");
+    expect(language).toBeUndefined();
+  });
+  it("auto model with language", () => {
+    const [model, language] = parseSTTModelString("auto:pt");
+    expect(model).toBe("auto");
+    expect(language).toBe("pt");
+  });
+});
+describe("normalizeSTTFallback", () => {
+  it("single string model", () => {
+    const result = normalizeSTTFallback("deepgram/nova-3");
+    expect(result).toEqual([{ model: "deepgram/nova-3" }]);
+  });
+  it("single FallbackModel dict", () => {
+    const fallback = { model: "deepgram/nova-3" };
+    const result = normalizeSTTFallback(fallback);
+    expect(result).toEqual([{ model: "deepgram/nova-3" }]);
+  });
+  it("list of string models", () => {
+    const result = normalizeSTTFallback(["deepgram/nova-3", "cartesia/ink-whisper"]);
+    expect(result).toEqual([{ model: "deepgram/nova-3" }, { model: "cartesia/ink-whisper" }]);
+  });
+  it("list of FallbackModel dicts", () => {
+    const fallbacks = [{ model: "deepgram/nova-3" }, { model: "assemblyai" }];
+    const result = normalizeSTTFallback(fallbacks);
+    expect(result).toEqual([{ model: "deepgram/nova-3" }, { model: "assemblyai" }]);
+  });
+  it("mixed list of strings and dicts", () => {
+    const result = normalizeSTTFallback([
+      "deepgram/nova-3",
+      { model: "cartesia/ink-whisper" },
+      "assemblyai"
+    ]);
+    expect(result).toEqual([
+      { model: "deepgram/nova-3" },
+      { model: "cartesia/ink-whisper" },
+      { model: "assemblyai" }
+    ]);
+  });
+  it("string with language suffix discards language", () => {
+    const result = normalizeSTTFallback("deepgram/nova-3:en");
+    expect(result).toEqual([{ model: "deepgram/nova-3" }]);
+  });
+  it("FallbackModel with extraKwargs is preserved", () => {
+    const fallback = {
+      model: "deepgram/nova-3",
+      extraKwargs: { keywords: [["livekit", 1.5]], punctuate: true }
+    };
+    const result = normalizeSTTFallback(fallback);
+    expect(result).toEqual([
+      {
+        model: "deepgram/nova-3",
+        extraKwargs: { keywords: [["livekit", 1.5]], punctuate: true }
+      }
+    ]);
+  });
+  it("list with extraKwargs preserved", () => {
+    const result = normalizeSTTFallback([
+      { model: "deepgram/nova-3", extraKwargs: { punctuate: true } },
+      "cartesia/ink-whisper",
+      { model: "assemblyai", extraKwargs: { format_turns: true } }
+    ]);
+    expect(result).toEqual([
+      { model: "deepgram/nova-3", extraKwargs: { punctuate: true } },
+      { model: "cartesia/ink-whisper" },
+      { model: "assemblyai", extraKwargs: { format_turns: true } }
+    ]);
+  });
+  it("empty list returns empty list", () => {
+    const result = normalizeSTTFallback([]);
+    expect(result).toEqual([]);
+  });
+  it("multiple colons in model string splits on last", () => {
+    const result = normalizeSTTFallback("some:model:part:fr");
+    expect(result).toEqual([{ model: "some:model:part" }]);
+  });
+});
+describe("STT constructor fallback and connOptions", () => {
+  it("normalizes language in constructor and model string", () => {
+    const stt = makeStt({ model: "deepgram/nova-3:english" });
+    expect(stt["opts"].language).toBe("en");
+  });
+  it("prefers explicit normalized language over model suffix", () => {
+    const stt = makeStt({ model: "deepgram/nova-3:english", language: "en_US" });
+    expect(stt["opts"].language).toBe(normalizeLanguage("en_US"));
+  });
+  it("fallback not given defaults to undefined", () => {
+    const stt = makeStt();
+    expect(stt["opts"].fallback).toBeUndefined();
+  });
+  it("fallback single string is normalized", () => {
+    const stt = makeStt({ fallback: "cartesia/ink-whisper" });
+    expect(stt["opts"].fallback).toEqual([{ model: "cartesia/ink-whisper" }]);
+  });
+  it("fallback list of strings is normalized", () => {
+    const stt = makeStt({ fallback: ["deepgram/nova-3", "assemblyai"] });
+    expect(stt["opts"].fallback).toEqual([{ model: "deepgram/nova-3" }, { model: "assemblyai" }]);
+  });
+  it("fallback single FallbackModel is normalized to list", () => {
+    const stt = makeStt({ fallback: { model: "deepgram/nova-3" } });
+    expect(stt["opts"].fallback).toEqual([{ model: "deepgram/nova-3" }]);
+  });
+  it("fallback with extraKwargs is preserved", () => {
+    const stt = makeStt({
+      fallback: {
+        model: "deepgram/nova-3",
+        extraKwargs: { punctuate: true, keywords: [["livekit", 1.5]] }
+      }
+    });
+    expect(stt["opts"].fallback).toEqual([
+      {
+        model: "deepgram/nova-3",
+        extraKwargs: { punctuate: true, keywords: [["livekit", 1.5]] }
+      }
+    ]);
+  });
+  it("fallback mixed list is normalized", () => {
+    const stt = makeStt({
+      fallback: [
+        "deepgram/nova-3",
+        { model: "cartesia", extraKwargs: { min_volume: 0.5 } },
+        "assemblyai"
+      ]
+    });
+    expect(stt["opts"].fallback).toEqual([
+      { model: "deepgram/nova-3" },
+      { model: "cartesia", extraKwargs: { min_volume: 0.5 } },
+      { model: "assemblyai" }
+    ]);
+  });
+  it("fallback string with language discards language", () => {
+    const stt = makeStt({ fallback: "deepgram/nova-3:en" });
+    expect(stt["opts"].fallback).toEqual([{ model: "deepgram/nova-3" }]);
+  });
+  it("connOptions not given uses default", () => {
+    const stt = makeStt();
+    expect(stt["opts"].connOptions).toEqual(DEFAULT_API_CONNECT_OPTIONS);
+  });
+  it("connOptions custom timeout", () => {
+    const custom = { timeoutMs: 3e4, maxRetry: 3, retryIntervalMs: 2e3 };
+    const stt = makeStt({ connOptions: custom });
+    expect(stt["opts"].connOptions).toEqual(custom);
+    expect(stt["opts"].connOptions.timeoutMs).toBe(3e4);
+  });
+  it("connOptions custom maxRetry", () => {
+    const custom = { timeoutMs: 1e4, maxRetry: 5, retryIntervalMs: 2e3 };
+    const stt = makeStt({ connOptions: custom });
+    expect(stt["opts"].connOptions).toEqual(custom);
+    expect(stt["opts"].connOptions.maxRetry).toBe(5);
+  });
+  it("connOptions full custom", () => {
+    const custom = { timeoutMs: 6e4, maxRetry: 10, retryIntervalMs: 2e3 };
+    const stt = makeStt({ connOptions: custom });
+    expect(stt["opts"].connOptions).toEqual(custom);
+    expect(stt["opts"].connOptions.timeoutMs).toBe(6e4);
+    expect(stt["opts"].connOptions.maxRetry).toBe(10);
+    expect(stt["opts"].connOptions.retryIntervalMs).toBe(2e3);
+  });
+});
+//# sourceMappingURL=stt.test.js.map

package/dist/inference/stt.test.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/inference/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { beforeAll, describe, expect, it } from 'vitest';\nimport { normalizeLanguage } from '../language.js';\nimport { initializeLogger } from '../log.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { STT, type STTFallbackModel, normalizeSTTFallback, parseSTTModelString } from './stt.js';\n\nbeforeAll(() => {\n  initializeLogger({ level: 'silent', pretty: false });\n});\n\n/** Helper to create STT with required credentials. */\nfunction makeStt(overrides: Record<string, unknown> = {}) {\n  const defaults = {\n    model: 'deepgram' as const,\n    apiKey: 'test-key',\n    apiSecret: 'test-secret',\n    baseURL: 'https://example.livekit.cloud',\n  };\n  return new STT({ ...defaults, ...overrides });\n}\n\ndescribe('parseSTTModelString', () => {\n  it('simple model without language', () => {\n    const [model, language] = parseSTTModelString('deepgram');\n    expect(model).toBe('deepgram');\n    expect(language).toBeUndefined();\n  });\n\n  it('model with language suffix', () => {\n    const [model, language] = parseSTTModelString('deepgram:en');\n    expect(model).toBe('deepgram');\n    expect(language).toBe('en');\n  });\n\n  it('normalizes language suffixes', () => {\n    const [model, language] = parseSTTModelString('deepgram:english');\n    expect(model).toBe('deepgram');\n    expect(language).toBe('en');\n  });\n\n  it('provider/model format without language', () => {\n    const [model, language] = parseSTTModelString('deepgram/nova-3');\n    expect(model).toBe('deepgram/nova-3');\n    expect(language).toBeUndefined();\n  });\n\n  it('provider/model format with language', () => {\n    const [model, language] = parseSTTModelString('deepgram/nova-3:en');\n    expect(model).toBe('deepgram/nova-3');\n    expect(language).toBe('en');\n  });\n\n  it.each([\n    ['cartesia/ink-whisper:de', 'cartesia/ink-whisper', 'de'],\n    ['assemblyai:es', 'assemblyai', 'es'],\n    ['deepgram/nova-2-medical:ja', 'deepgram/nova-2-medical', 'ja'],\n    ['deepgram/nova-3:multi', 'deepgram/nova-3', 'multi'],\n    ['cartesia:zh', 'cartesia', 'zh'],\n  ])('various providers and languages: %s', (modelStr, expectedModel, expectedLang) => {\n    const [model, language] = parseSTTModelString(modelStr);\n    expect(model).toBe(expectedModel);\n    expect(language).toBe(expectedLang);\n  });\n\n  it('auto model without language', () => {\n    const [model, language] = parseSTTModelString('auto');\n    expect(model).toBe('auto');\n    expect(language).toBeUndefined();\n  });\n\n  it('auto model with language', () => {\n    const [model, language] = parseSTTModelString('auto:pt');\n    expect(model).toBe('auto');\n    expect(language).toBe('pt');\n  });\n});\n\ndescribe('normalizeSTTFallback', () => {\n  it('single string model', () => {\n    const result = normalizeSTTFallback('deepgram/nova-3');\n    expect(result).toEqual([{ model: 'deepgram/nova-3' }]);\n  });\n\n  it('single FallbackModel dict', () => {\n    const fallback: STTFallbackModel = { model: 'deepgram/nova-3' };\n    const result = normalizeSTTFallback(fallback);\n    expect(result).toEqual([{ model: 'deepgram/nova-3' }]);\n  });\n\n  it('list of string models', () => {\n    const result = normalizeSTTFallback(['deepgram/nova-3', 'cartesia/ink-whisper']);\n    expect(result).toEqual([{ model: 'deepgram/nova-3' }, { model: 'cartesia/ink-whisper' }]);\n  });\n\n  it('list of FallbackModel dicts', () => {\n    const fallbacks: STTFallbackModel[] = [{ model: 'deepgram/nova-3' }, { model: 'assemblyai' }];\n    const result = normalizeSTTFallback(fallbacks);\n    expect(result).toEqual([{ model: 'deepgram/nova-3' }, { model: 'assemblyai' }]);\n  });\n\n  it('mixed list of strings and dicts', () => {\n    const result = normalizeSTTFallback([\n      'deepgram/nova-3',\n      { model: 'cartesia/ink-whisper' } as STTFallbackModel,\n      'assemblyai',\n    ]);\n    expect(result).toEqual([\n      { model: 'deepgram/nova-3' },\n      { model: 'cartesia/ink-whisper' },\n      { model: 'assemblyai' },\n    ]);\n  });\n\n  it('string with language suffix discards language', () => {\n    const result = normalizeSTTFallback('deepgram/nova-3:en');\n    expect(result).toEqual([{ model: 'deepgram/nova-3' }]);\n  });\n\n  it('FallbackModel with extraKwargs is preserved', () => {\n    const fallback: STTFallbackModel = {\n      model: 'deepgram/nova-3',\n      extraKwargs: { keywords: [['livekit', 1.5]], punctuate: true },\n    };\n    const result = normalizeSTTFallback(fallback);\n    expect(result).toEqual([\n      {\n        model: 'deepgram/nova-3',\n        extraKwargs: { keywords: [['livekit', 1.5]], punctuate: true },\n      },\n    ]);\n  });\n\n  it('list with extraKwargs preserved', () => {\n    const result = normalizeSTTFallback([\n      { model: 'deepgram/nova-3', extraKwargs: { punctuate: true } } as STTFallbackModel,\n      'cartesia/ink-whisper',\n      { model: 'assemblyai', extraKwargs: { format_turns: true } } as STTFallbackModel,\n    ]);\n    expect(result).toEqual([\n      { model: 'deepgram/nova-3', extraKwargs: { punctuate: true } },\n      { model: 'cartesia/ink-whisper' },\n      { model: 'assemblyai', extraKwargs: { format_turns: true } },\n    ]);\n  });\n\n  it('empty list returns empty list', () => {\n    const result = normalizeSTTFallback([]);\n    expect(result).toEqual([]);\n  });\n\n  it('multiple colons in model string splits on last', () => {\n    const result = normalizeSTTFallback('some:model:part:fr');\n    expect(result).toEqual([{ model: 'some:model:part' }]);\n  });\n});\n\ndescribe('STT constructor fallback and connOptions', () => {\n  it('normalizes language in constructor and model string', () => {\n    const stt = makeStt({ model: 'deepgram/nova-3:english' });\n    expect(stt['opts'].language).toBe('en');\n  });\n\n  it('prefers explicit normalized language over model suffix', () => {\n    const stt = makeStt({ model: 'deepgram/nova-3:english', language: 'en_US' });\n    expect(stt['opts'].language).toBe(normalizeLanguage('en_US'));\n  });\n\n  it('fallback not given defaults to undefined', () => {\n    const stt = makeStt();\n    expect(stt['opts'].fallback).toBeUndefined();\n  });\n\n  it('fallback single string is normalized', () => {\n    const stt = makeStt({ fallback: 'cartesia/ink-whisper' });\n    expect(stt['opts'].fallback).toEqual([{ model: 'cartesia/ink-whisper' }]);\n  });\n\n  it('fallback list of strings is normalized', () => {\n    const stt = makeStt({ fallback: ['deepgram/nova-3', 'assemblyai'] });\n    expect(stt['opts'].fallback).toEqual([{ model: 'deepgram/nova-3' }, { model: 'assemblyai' }]);\n  });\n\n  it('fallback single FallbackModel is normalized to list', () => {\n    const stt = makeStt({ fallback: { model: 'deepgram/nova-3' } });\n    expect(stt['opts'].fallback).toEqual([{ model: 'deepgram/nova-3' }]);\n  });\n\n  it('fallback with extraKwargs is preserved', () => {\n    const stt = makeStt({\n      fallback: {\n        model: 'deepgram/nova-3',\n        extraKwargs: { punctuate: true, keywords: [['livekit', 1.5]] },\n      },\n    });\n    expect(stt['opts'].fallback).toEqual([\n      {\n        model: 'deepgram/nova-3',\n        extraKwargs: { punctuate: true, keywords: [['livekit', 1.5]] },\n      },\n    ]);\n  });\n\n  it('fallback mixed list is normalized', () => {\n    const stt = makeStt({\n      fallback: [\n        'deepgram/nova-3',\n        { model: 'cartesia', extraKwargs: { min_volume: 0.5 } },\n        'assemblyai',\n      ],\n    });\n    expect(stt['opts'].fallback).toEqual([\n      { model: 'deepgram/nova-3' },\n      { model: 'cartesia', extraKwargs: { min_volume: 0.5 } },\n      { model: 'assemblyai' },\n    ]);\n  });\n\n  it('fallback string with language discards language', () => {\n    const stt = makeStt({ fallback: 'deepgram/nova-3:en' });\n    expect(stt['opts'].fallback).toEqual([{ model: 'deepgram/nova-3' }]);\n  });\n\n  it('connOptions not given uses default', () => {\n    const stt = makeStt();\n    expect(stt['opts'].connOptions).toEqual(DEFAULT_API_CONNECT_OPTIONS);\n  });\n\n  it('connOptions custom timeout', () => {\n    const custom: APIConnectOptions = { timeoutMs: 30000, maxRetry: 3, retryIntervalMs: 2000 };\n    const stt = makeStt({ connOptions: custom });\n    expect(stt['opts'].connOptions).toEqual(custom);\n    expect(stt['opts'].connOptions!.timeoutMs).toBe(30000);\n  });\n\n  it('connOptions custom maxRetry', () => {\n    const custom: APIConnectOptions = { timeoutMs: 10000, maxRetry: 5, retryIntervalMs: 2000 };\n    const stt = makeStt({ connOptions: custom });\n    expect(stt['opts'].connOptions).toEqual(custom);\n    expect(stt['opts'].connOptions!.maxRetry).toBe(5);\n  });\n\n  it('connOptions full custom', () => {\n    const custom: APIConnectOptions = { timeoutMs: 60000, maxRetry: 10, retryIntervalMs: 2000 };\n    const stt = makeStt({ connOptions: custom });\n    expect(stt['opts'].connOptions).toEqual(custom);\n    expect(stt['opts'].connOptions!.timeoutMs).toBe(60000);\n    expect(stt['opts'].connOptions!.maxRetry).toBe(10);\n    expect(stt['opts'].connOptions!.retryIntervalMs).toBe(2000);\n  });\n});\n"],"mappings":"AAGA,SAAS,WAAW,UAAU,QAAQ,UAAU;AAChD,SAAS,yBAAyB;AAClC,SAAS,wBAAwB;AACjC,SAAiC,mCAAmC;AACpE,SAAS,KAA4B,sBAAsB,2BAA2B;AAEtF,UAAU,MAAM;AACd,mBAAiB,EAAE,OAAO,UAAU,QAAQ,MAAM,CAAC;AACrD,CAAC;AAGD,SAAS,QAAQ,YAAqC,CAAC,GAAG;AACxD,QAAM,WAAW;AAAA,IACf,OAAO;AAAA,IACP,QAAQ;AAAA,IACR,WAAW;AAAA,IACX,SAAS;AAAA,EACX;AACA,SAAO,IAAI,IAAI,EAAE,GAAG,UAAU,GAAG,UAAU,CAAC;AAC9C;AAEA,SAAS,uBAAuB,MAAM;AACpC,KAAG,iCAAiC,MAAM;AACxC,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,UAAU;AACxD,WAAO,KAAK,EAAE,KAAK,UAAU;AAC7B,WAAO,QAAQ,EAAE,cAAc;AAAA,EACjC,CAAC;AAED,KAAG,8BAA8B,MAAM;AACrC,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,aAAa;AAC3D,WAAO,KAAK,EAAE,KAAK,UAAU;AAC7B,WAAO,QAAQ,EAAE,KAAK,IAAI;AAAA,EAC5B,CAAC;AAED,KAAG,gCAAgC,MAAM;AACvC,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,kBAAkB;AAChE,WAAO,KAAK,EAAE,KAAK,UAAU;AAC7B,WAAO,QAAQ,EAAE,KAAK,IAAI;AAAA,EAC5B,CAAC;AAED,KAAG,0CAA0C,MAAM;AACjD,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,iBAAiB;AAC/D,WAAO,KAAK,EAAE,KAAK,iBAAiB;AACpC,WAAO,QAAQ,EAAE,cAAc;AAAA,EACjC,CAAC;AAED,KAAG,uCAAuC,MAAM;AAC9C,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,oBAAoB;AAClE,WAAO,KAAK,EAAE,KAAK,iBAAiB;AACpC,WAAO,QAAQ,EAAE,KAAK,IAAI;AAAA,EAC5B,CAAC;AAED,KAAG,KAAK;AAAA,IACN,CAAC,2BAA2B,wBAAwB,IAAI;AAAA,IACxD,CAAC,iBAAiB,cAAc,IAAI;AAAA,IACpC,CAAC,8BAA8B,2BAA2B,IAAI;AAAA,IAC9D,CAAC,yBAAyB,mBAAmB,OAAO;AAAA,IACpD,CAAC,eAAe,YAAY,IAAI;AAAA,EAClC,CAAC,EAAE,uCAAuC,CAAC,UAAU,eAAe,iBAAiB;AACnF,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,QAAQ;AACtD,WAAO,KAAK,EAAE,KAAK,aAAa;AAChC,WAAO,QAAQ,EAAE,KAAK,YAAY;AAAA,EACpC,CAAC;AAED,KAAG,+BAA+B,MAAM;AACtC,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,MAAM;AACpD,WAAO,KAAK,EAAE,KAAK,MAAM;AACzB,WAAO,QAAQ,EAAE,cAAc;AAAA,EACjC,CAAC;AAED,KAAG,4BAA4B,MAAM;AACnC,UAAM,CAAC,OAAO,QAAQ,IAAI,oBAAoB,SAAS;AACvD,WAAO,KAAK,EAAE,KAAK,MAAM;AACzB,WAAO,QAAQ,EAAE,KAAK,IAAI;AAAA,EAC5B,CAAC;AACH,CAAC;AAED,SAAS,wBAAwB,MAAM;AACrC,KAAG,uBAAuB,MAAM;AAC9B,UAAM,SAAS,qBAAqB,iBAAiB;AACrD,WAAO,MAAM,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,CAAC,CAAC;AAAA,EACvD,CAAC;AAED,KAAG,6BAA6B,MAAM;AACpC,UAAM,WAA6B,EAAE,OAAO,kBAAkB;AAC9D,UAAM,SAAS,qBAAqB,QAAQ;AAC5C,WAAO,MAAM,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,CAAC,CAAC;AAAA,EACvD,CAAC;AAED,KAAG,yBAAyB,MAAM;AAChC,UAAM,SAAS,qBAAqB,CAAC,mBAAmB,sBAAsB,CAAC;AAC/E,WAAO,MAAM,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,GAAG,EAAE,OAAO,uBAAuB,CAAC,CAAC;AAAA,EAC1F,CAAC;AAED,KAAG,+BAA+B,MAAM;AACtC,UAAM,YAAgC,CAAC,EAAE,OAAO,kBAAkB,GAAG,EAAE,OAAO,aAAa,CAAC;AAC5F,UAAM,SAAS,qBAAqB,SAAS;AAC7C,WAAO,MAAM,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,GAAG,EAAE,OAAO,aAAa,CAAC,CAAC;AAAA,EAChF,CAAC;AAED,KAAG,mCAAmC,MAAM;AAC1C,UAAM,SAAS,qBAAqB;AAAA,MAClC;AAAA,MACA,EAAE,OAAO,uBAAuB;AAAA,MAChC;AAAA,IACF,CAAC;AACD,WAAO,MAAM,EAAE,QAAQ;AAAA,MACrB,EAAE,OAAO,kBAAkB;AAAA,MAC3B,EAAE,OAAO,uBAAuB;AAAA,MAChC,EAAE,OAAO,aAAa;AAAA,IACxB,CAAC;AAAA,EACH,CAAC;AAED,KAAG,iDAAiD,MAAM;AACxD,UAAM,SAAS,qBAAqB,oBAAoB;AACxD,WAAO,MAAM,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,CAAC,CAAC;AAAA,EACvD,CAAC;AAED,KAAG,+CAA+C,MAAM;AACtD,UAAM,WAA6B;AAAA,MACjC,OAAO;AAAA,MACP,aAAa,EAAE,UAAU,CAAC,CAAC,WAAW,GAAG,CAAC,GAAG,WAAW,KAAK;AAAA,IAC/D;AACA,UAAM,SAAS,qBAAqB,QAAQ;AAC5C,WAAO,MAAM,EAAE,QAAQ;AAAA,MACrB;AAAA,QACE,OAAO;AAAA,QACP,aAAa,EAAE,UAAU,CAAC,CAAC,WAAW,GAAG,CAAC,GAAG,WAAW,KAAK;AAAA,MAC/D;AAAA,IACF,CAAC;AAAA,EACH,CAAC;AAED,KAAG,mCAAmC,MAAM;AAC1C,UAAM,SAAS,qBAAqB;AAAA,MAClC,EAAE,OAAO,mBAAmB,aAAa,EAAE,WAAW,KAAK,EAAE;AAAA,MAC7D;AAAA,MACA,EAAE,OAAO,cAAc,aAAa,EAAE,cAAc,KAAK,EAAE;AAAA,IAC7D,CAAC;AACD,WAAO,MAAM,EAAE,QAAQ;AAAA,MACrB,EAAE,OAAO,mBAAmB,aAAa,EAAE,WAAW,KAAK,EAAE;AAAA,MAC7D,EAAE,OAAO,uBAAuB;AAAA,MAChC,EAAE,OAAO,cAAc,aAAa,EAAE,cAAc,KAAK,EAAE;AAAA,IAC7D,CAAC;AAAA,EACH,CAAC;AAED,KAAG,iCAAiC,MAAM;AACxC,UAAM,SAAS,qBAAqB,CAAC,CAAC;AACtC,WAAO,MAAM,EAAE,QAAQ,CAAC,CAAC;AAAA,EAC3B,CAAC;AAED,KAAG,kDAAkD,MAAM;AACzD,UAAM,SAAS,qBAAqB,oBAAoB;AACxD,WAAO,MAAM,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,CAAC,CAAC;AAAA,EACvD,CAAC;AACH,CAAC;AAED,SAAS,4CAA4C,MAAM;AACzD,KAAG,uDAAuD,MAAM;AAC9D,UAAM,MAAM,QAAQ,EAAE,OAAO,0BAA0B,CAAC;AACxD,WAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,KAAK,IAAI;AAAA,EACxC,CAAC;AAED,KAAG,0DAA0D,MAAM;AACjE,UAAM,MAAM,QAAQ,EAAE,OAAO,2BAA2B,UAAU,QAAQ,CAAC;AAC3E,WAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,KAAK,kBAAkB,OAAO,CAAC;AAAA,EAC9D,CAAC;AAED,KAAG,4CAA4C,MAAM;AACnD,UAAM,MAAM,QAAQ;AACpB,WAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,cAAc;AAAA,EAC7C,CAAC;AAED,KAAG,wCAAwC,MAAM;AAC/C,UAAM,MAAM,QAAQ,EAAE,UAAU,uBAAuB,CAAC;AACxD,WAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,QAAQ,CAAC,EAAE,OAAO,uBAAuB,CAAC,CAAC;AAAA,EAC1E,CAAC;AAED,KAAG,0CAA0C,MAAM;AACjD,UAAM,MAAM,QAAQ,EAAE,UAAU,CAAC,mBAAmB,YAAY,EAAE,CAAC;AACnE,WAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,GAAG,EAAE,OAAO,aAAa,CAAC,CAAC;AAAA,EAC9F,CAAC;AAED,KAAG,uDAAuD,MAAM;AAC9D,UAAM,MAAM,QAAQ,EAAE,UAAU,EAAE,OAAO,kBAAkB,EAAE,CAAC;AAC9D,WAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,CAAC,CAAC;AAAA,EACrE,CAAC;AAED,KAAG,0CAA0C,MAAM;AACjD,UAAM,MAAM,QAAQ;AAAA,MAClB,UAAU;AAAA,QACR,OAAO;AAAA,QACP,aAAa,EAAE,WAAW,MAAM,UAAU,CAAC,CAAC,WAAW,GAAG,CAAC,EAAE;AAAA,MAC/D;AAAA,IACF,CAAC;AACD,WAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,QAAQ;AAAA,MACnC;AAAA,QACE,OAAO;AAAA,QACP,aAAa,EAAE,WAAW,MAAM,UAAU,CAAC,CAAC,WAAW,GAAG,CAAC,EAAE;AAAA,MAC/D;AAAA,IACF,CAAC;AAAA,EACH,CAAC;AAED,KAAG,qCAAqC,MAAM;AAC5C,UAAM,MAAM,QAAQ;AAAA,MAClB,UAAU;AAAA,QACR;AAAA,QACA,EAAE,OAAO,YAAY,aAAa,EAAE,YAAY,IAAI,EAAE;AAAA,QACtD;AAAA,MACF;AAAA,IACF,CAAC;AACD,WAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,QAAQ;AAAA,MACnC,EAAE,OAAO,kBAAkB;AAAA,MAC3B,EAAE,OAAO,YAAY,aAAa,EAAE,YAAY,IAAI,EAAE;AAAA,MACtD,EAAE,OAAO,aAAa;AAAA,IACxB,CAAC;AAAA,EACH,CAAC;AAED,KAAG,mDAAmD,MAAM;AAC1D,UAAM,MAAM,QAAQ,EAAE,UAAU,qBAAqB,CAAC;AACtD,WAAO,IAAI,MAAM,EAAE,QAAQ,EAAE,QAAQ,CAAC,EAAE,OAAO,kBAAkB,CAAC,CAAC;AAAA,EACrE,CAAC;AAED,KAAG,sCAAsC,MAAM;AAC7C,UAAM,MAAM,QAAQ;AACpB,WAAO,IAAI,MAAM,EAAE,WAAW,EAAE,QAAQ,2BAA2B;AAAA,EACrE,CAAC;AAED,KAAG,8BAA8B,MAAM;AACrC,UAAM,SAA4B,EAAE,WAAW,KAAO,UAAU,GAAG,iBAAiB,IAAK;AACzF,UAAM,MAAM,QAAQ,EAAE,aAAa,OAAO,CAAC;AAC3C,WAAO,IAAI,MAAM,EAAE,WAAW,EAAE,QAAQ,MAAM;AAC9C,WAAO,IAAI,MAAM,EAAE,YAAa,SAAS,EAAE,KAAK,GAAK;AAAA,EACvD,CAAC;AAED,KAAG,+BAA+B,MAAM;AACtC,UAAM,SAA4B,EAAE,WAAW,KAAO,UAAU,GAAG,iBAAiB,IAAK;AACzF,UAAM,MAAM,QAAQ,EAAE,aAAa,OAAO,CAAC;AAC3C,WAAO,IAAI,MAAM,EAAE,WAAW,EAAE,QAAQ,MAAM;AAC9C,WAAO,IAAI,MAAM,EAAE,YAAa,QAAQ,EAAE,KAAK,CAAC;AAAA,EAClD,CAAC;AAED,KAAG,2BAA2B,MAAM;AAClC,UAAM,SAA4B,EAAE,WAAW,KAAO,UAAU,IAAI,iBAAiB,IAAK;AAC1F,UAAM,MAAM,QAAQ,EAAE,aAAa,OAAO,CAAC;AAC3C,WAAO,IAAI,MAAM,EAAE,WAAW,EAAE,QAAQ,MAAM;AAC9C,WAAO,IAAI,MAAM,EAAE,YAAa,SAAS,EAAE,KAAK,GAAK;AACrD,WAAO,IAAI,MAAM,EAAE,YAAa,QAAQ,EAAE,KAAK,EAAE;AACjD,WAAO,IAAI,MAAM,EAAE,YAAa,eAAe,EAAE,KAAK,GAAI;AAAA,EAC5D,CAAC;AACH,CAAC;","names":[]}