@livekit/agents 1.0.36 → 1.0.38

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (149) hide show
  1. package/dist/cli.cjs.map +1 -1
  2. package/dist/inference/api_protos.cjs +68 -0
  3. package/dist/inference/api_protos.cjs.map +1 -1
  4. package/dist/inference/api_protos.d.cts +345 -4
  5. package/dist/inference/api_protos.d.ts +345 -4
  6. package/dist/inference/api_protos.d.ts.map +1 -1
  7. package/dist/inference/api_protos.js +60 -0
  8. package/dist/inference/api_protos.js.map +1 -1
  9. package/dist/inference/stt.cjs +32 -21
  10. package/dist/inference/stt.cjs.map +1 -1
  11. package/dist/inference/stt.d.ts.map +1 -1
  12. package/dist/inference/stt.js +34 -21
  13. package/dist/inference/stt.js.map +1 -1
  14. package/dist/ipc/inference_proc_executor.cjs.map +1 -1
  15. package/dist/ipc/job_proc_executor.cjs.map +1 -1
  16. package/dist/stt/stt.cjs +10 -0
  17. package/dist/stt/stt.cjs.map +1 -1
  18. package/dist/stt/stt.d.cts +12 -0
  19. package/dist/stt/stt.d.ts +12 -0
  20. package/dist/stt/stt.d.ts.map +1 -1
  21. package/dist/stt/stt.js +10 -0
  22. package/dist/stt/stt.js.map +1 -1
  23. package/dist/telemetry/traces.cjs +4 -3
  24. package/dist/telemetry/traces.cjs.map +1 -1
  25. package/dist/telemetry/traces.d.cts +2 -0
  26. package/dist/telemetry/traces.d.ts +2 -0
  27. package/dist/telemetry/traces.d.ts.map +1 -1
  28. package/dist/telemetry/traces.js +4 -3
  29. package/dist/telemetry/traces.js.map +1 -1
  30. package/dist/utils.cjs +6 -0
  31. package/dist/utils.cjs.map +1 -1
  32. package/dist/utils.d.cts +2 -0
  33. package/dist/utils.d.ts +2 -0
  34. package/dist/utils.d.ts.map +1 -1
  35. package/dist/utils.js +6 -0
  36. package/dist/utils.js.map +1 -1
  37. package/dist/voice/agent.cjs +5 -0
  38. package/dist/voice/agent.cjs.map +1 -1
  39. package/dist/voice/agent.d.ts.map +1 -1
  40. package/dist/voice/agent.js +5 -0
  41. package/dist/voice/agent.js.map +1 -1
  42. package/dist/voice/agent_activity.cjs +49 -23
  43. package/dist/voice/agent_activity.cjs.map +1 -1
  44. package/dist/voice/agent_activity.d.cts +1 -1
  45. package/dist/voice/agent_activity.d.ts +1 -1
  46. package/dist/voice/agent_activity.d.ts.map +1 -1
  47. package/dist/voice/agent_activity.js +50 -24
  48. package/dist/voice/agent_activity.js.map +1 -1
  49. package/dist/voice/agent_session.cjs +7 -5
  50. package/dist/voice/agent_session.cjs.map +1 -1
  51. package/dist/voice/agent_session.d.cts +5 -2
  52. package/dist/voice/agent_session.d.ts +5 -2
  53. package/dist/voice/agent_session.d.ts.map +1 -1
  54. package/dist/voice/agent_session.js +7 -5
  55. package/dist/voice/agent_session.js.map +1 -1
  56. package/dist/voice/audio_recognition.cjs +3 -1
  57. package/dist/voice/audio_recognition.cjs.map +1 -1
  58. package/dist/voice/audio_recognition.d.ts.map +1 -1
  59. package/dist/voice/audio_recognition.js +3 -1
  60. package/dist/voice/audio_recognition.js.map +1 -1
  61. package/dist/voice/avatar/datastream_io.cjs +6 -0
  62. package/dist/voice/avatar/datastream_io.cjs.map +1 -1
  63. package/dist/voice/avatar/datastream_io.d.cts +1 -0
  64. package/dist/voice/avatar/datastream_io.d.ts +1 -0
  65. package/dist/voice/avatar/datastream_io.d.ts.map +1 -1
  66. package/dist/voice/avatar/datastream_io.js +6 -0
  67. package/dist/voice/avatar/datastream_io.js.map +1 -1
  68. package/dist/voice/background_audio.cjs.map +1 -1
  69. package/dist/voice/generation.cjs +14 -5
  70. package/dist/voice/generation.cjs.map +1 -1
  71. package/dist/voice/generation.d.cts +3 -2
  72. package/dist/voice/generation.d.ts +3 -2
  73. package/dist/voice/generation.d.ts.map +1 -1
  74. package/dist/voice/generation.js +14 -5
  75. package/dist/voice/generation.js.map +1 -1
  76. package/dist/voice/io.cjs +12 -0
  77. package/dist/voice/io.cjs.map +1 -1
  78. package/dist/voice/io.d.cts +19 -1
  79. package/dist/voice/io.d.ts +19 -1
  80. package/dist/voice/io.d.ts.map +1 -1
  81. package/dist/voice/io.js +12 -0
  82. package/dist/voice/io.js.map +1 -1
  83. package/dist/voice/recorder_io/recorder_io.cjs +91 -28
  84. package/dist/voice/recorder_io/recorder_io.cjs.map +1 -1
  85. package/dist/voice/recorder_io/recorder_io.d.cts +7 -1
  86. package/dist/voice/recorder_io/recorder_io.d.ts +7 -1
  87. package/dist/voice/recorder_io/recorder_io.d.ts.map +1 -1
  88. package/dist/voice/recorder_io/recorder_io.js +91 -28
  89. package/dist/voice/recorder_io/recorder_io.js.map +1 -1
  90. package/dist/voice/room_io/_input.cjs +40 -11
  91. package/dist/voice/room_io/_input.cjs.map +1 -1
  92. package/dist/voice/room_io/_input.d.cts +4 -1
  93. package/dist/voice/room_io/_input.d.ts +4 -1
  94. package/dist/voice/room_io/_input.d.ts.map +1 -1
  95. package/dist/voice/room_io/_input.js +31 -2
  96. package/dist/voice/room_io/_input.js.map +1 -1
  97. package/dist/voice/room_io/_output.cjs +6 -0
  98. package/dist/voice/room_io/_output.cjs.map +1 -1
  99. package/dist/voice/room_io/_output.d.cts +1 -0
  100. package/dist/voice/room_io/_output.d.ts +1 -0
  101. package/dist/voice/room_io/_output.d.ts.map +1 -1
  102. package/dist/voice/room_io/_output.js +6 -0
  103. package/dist/voice/room_io/_output.js.map +1 -1
  104. package/dist/voice/room_io/room_io.cjs.map +1 -1
  105. package/dist/voice/room_io/room_io.d.cts +2 -2
  106. package/dist/voice/room_io/room_io.d.ts +2 -2
  107. package/dist/voice/room_io/room_io.d.ts.map +1 -1
  108. package/dist/voice/room_io/room_io.js.map +1 -1
  109. package/dist/voice/speech_handle.cjs +2 -0
  110. package/dist/voice/speech_handle.cjs.map +1 -1
  111. package/dist/voice/speech_handle.d.cts +3 -0
  112. package/dist/voice/speech_handle.d.ts +3 -0
  113. package/dist/voice/speech_handle.d.ts.map +1 -1
  114. package/dist/voice/speech_handle.js +2 -0
  115. package/dist/voice/speech_handle.js.map +1 -1
  116. package/dist/voice/testing/index.cjs +2 -0
  117. package/dist/voice/testing/index.cjs.map +1 -1
  118. package/dist/voice/testing/index.d.cts +1 -1
  119. package/dist/voice/testing/index.d.ts +1 -1
  120. package/dist/voice/testing/index.d.ts.map +1 -1
  121. package/dist/voice/testing/index.js +2 -0
  122. package/dist/voice/testing/index.js.map +1 -1
  123. package/dist/voice/testing/run_result.cjs +294 -5
  124. package/dist/voice/testing/run_result.cjs.map +1 -1
  125. package/dist/voice/testing/run_result.d.cts +149 -1
  126. package/dist/voice/testing/run_result.d.ts +149 -1
  127. package/dist/voice/testing/run_result.d.ts.map +1 -1
  128. package/dist/voice/testing/run_result.js +293 -5
  129. package/dist/voice/testing/run_result.js.map +1 -1
  130. package/package.json +1 -1
  131. package/src/inference/api_protos.ts +83 -0
  132. package/src/inference/stt.ts +39 -22
  133. package/src/stt/stt.ts +21 -0
  134. package/src/telemetry/traces.ts +6 -2
  135. package/src/utils.ts +7 -0
  136. package/src/voice/agent.ts +9 -0
  137. package/src/voice/agent_activity.ts +72 -26
  138. package/src/voice/agent_session.ts +6 -5
  139. package/src/voice/audio_recognition.ts +2 -0
  140. package/src/voice/avatar/datastream_io.ts +8 -0
  141. package/src/voice/generation.ts +24 -12
  142. package/src/voice/io.ts +27 -5
  143. package/src/voice/recorder_io/recorder_io.ts +123 -31
  144. package/src/voice/room_io/_input.ts +32 -4
  145. package/src/voice/room_io/_output.ts +8 -0
  146. package/src/voice/room_io/room_io.ts +3 -1
  147. package/src/voice/speech_handle.ts +4 -0
  148. package/src/voice/testing/index.ts +1 -0
  149. package/src/voice/testing/run_result.ts +373 -12
@@ -1 +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 { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport {\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 AnyString, 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-conversationalai'\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 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;\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\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;\n encoding: STTEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: STTOptions<TModel>;\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?: TModel;\n language?: STTLanguages;\n baseURL?: string;\n encoding?: STTEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: 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 modelOptions = {} 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 modelOptions,\n };\n }\n\n get label(): string {\n return 'inference.STT';\n }\n\n static fromModelString(modelString: string): STT<AnyString> {\n if (modelString.includes(':')) {\n const [model, language] = modelString.split(':') as [AnyString, STTLanguages];\n return new STT({ model, language });\n }\n return new STT({ model: modelString });\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 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 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 = { ...this.opts, ...opts };\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 type SttServerEvent = Record<string, any>;\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 const json = result.value;\n const type = json.type as string | undefined;\n\n switch (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(json, false);\n break;\n case 'final_transcript':\n this.processTranscript(json, true);\n break;\n case 'error':\n this.#logger.error({ error: json }, 'Received error from LiveKit STT');\n resourceCleanup();\n throw new APIError(`LiveKit STT returned error: ${JSON.stringify(json)}`);\n default:\n this.#logger.warn(\n { message: json },\n 'Received unexpected message from LiveKit STT',\n );\n break;\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n ws = await this.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: Record<string, any>, 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.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 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: 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 } 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":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAgC;AAEhC,wBAAyC;AACzC,mBAAgC;AAChC,iBAAoB;AACpB,4BAAoC;AACpC,iBAMO;AACP,mBAAoE;AACpE,mBAAsF;AACtF,IAAAA,gBAA6D;AAuE7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,yBAAyB;AAgBxB,MAAM,YAAsC,WAAAC,IAAQ;AAAA,EACjD;AAAA,EACA,UAAqC,oBAAI,IAAI;AAAA,EAErD,cAAU,gBAAI;AAAA,EAEd,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,eAAe,CAAC;AAAA,IAClB,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,OAAO,gBAAgB,aAAqC;AAC1D,QAAI,YAAY,SAAS,GAAG,GAAG;AAC7B,YAAM,CAAC,OAAO,QAAQ,IAAI,YAAY,MAAM,GAAG;AAC/C,aAAO,IAAI,IAAI,EAAE,OAAO,SAAS,CAAC;AAAA,IACpC;AACA,WAAO,IAAI,IAAI,EAAE,OAAO,YAAY,CAAC;AAAA,EACvC;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,yCAA4B,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;AAAA,EAEA,MAAM,UAAU,SAAqC;AACnD,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,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,UAAM,iCAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS,UAAM,yBAAU,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,WAAAC,aAAiB;AAAA,EACnE;AAAA,EACA,gBAAY,wBAAU,cAAc;AAAA,EACpC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,iBAAiB,IAAI,mBAAM;AAAA,EAC3B;AAAA,EACA;AAAA,EAER,cAAU,gBAAI;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,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,SAAK,eAAe,IAAI;AAAA,EAC1B;AAAA,EAEA,MAAgB,MAAqB;AACnC,WAAO,MAAM;AAEX,UAAI,KAAuB;AAC3B,UAAI,UAAU;AACd,UAAI,gBAAgB;AAGpB,YAAM,mBAAe,2CAAoC;AAEzD,YAAM,kBAAkB,MAAM;AAC5B,YAAI,QAAS;AACb,kBAAU;AACV,qBAAa,MAAM;AACnB,iCAAI;AACJ,iCAAI;AAAA,MACN;AAEA,YAAM,mBAAmB,OAAOC,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,iCAAe;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;AAEjB,kBAAM,OAAO,OAAO;AACpB,kBAAM,OAAO,KAAK;AAElB,oBAAQ,MAAM;AAAA,cACZ,KAAK;AAAA,cACL,KAAK;AACH;AAAA,cACF,KAAK;AACH,gCAAgB;AAChB,gCAAgB;AAChB;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,MAAM,KAAK;AAClC;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,MAAM,IAAI;AACjC;AAAA,cACF,KAAK;AACH,qBAAK,QAAQ,MAAM,EAAE,OAAO,KAAK,GAAG,iCAAiC;AACrE,gCAAgB;AAChB,sBAAM,IAAI,2BAAS,+BAA+B,KAAK,UAAU,IAAI,CAAC,EAAE;AAAA,cAC1E;AACE,qBAAK,QAAQ;AAAA,kBACX,EAAE,SAAS,KAAK;AAAA,kBAChB;AAAA,gBACF;AACA;AAAA,YACJ;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,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,IAAK,MAAM,GAAG,UAAU;AACxE,cAAM,iBAAiB,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,iBAAiB,IAAK,MAAM,GAAG,UAAU;AAC1F,cAAM,WAAW,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,MAAM,GAAG,UAAU;AACnE,cAAM,oBAAoB,kBAAK;AAAA,UAC7B,CAAC,EAAE,OAAO,MAAM,QAAQ,KAAK,CAAC,KAAK,eAAe,KAAK,OAAG,2BAAa,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,oBAAM;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,MAA2B,SAAkB;AAErE,QAAI,KAAK,MAAM,OAAQ;AAEvB,UAAM,YAAY,KAAK,cAAc,KAAK;AAC1C,UAAM,OAAO,KAAK,cAAc;AAChC,UAAM,WAAW,KAAK,YAAY,KAAK,KAAK,YAAY;AAExD,QAAI,CAAC,QAAQ,CAAC,QAAS;AAEvB,QAAI;AAEF,UAAI,CAAC,KAAK,UAAU;AAClB,aAAK,WAAW;AAChB,aAAK,MAAM,IAAI,EAAE,MAAM,2BAAgB,gBAAgB,CAAC;AAAA,MAC1D;AAEA,YAAM,aAAyB;AAAA,QAC7B;AAAA,QACA,WAAW,KAAK,SAAS;AAAA,QACzB,SAAS,KAAK,YAAY;AAAA,QAC1B,YAAY,KAAK,cAAc;AAAA,QAC/B;AAAA,MACF;AAEA,UAAI,SAAS;AACX,YAAI,KAAK,iBAAiB,GAAG;AAC3B,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,2BAAgB;AAAA,YACtB;AAAA,YACA,kBAAkB,EAAE,eAAe,KAAK,eAAe;AAAA,UACzD,CAAC;AACD,eAAK,iBAAiB;AAAA,QACxB;AAEA,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,2BAAgB;AAAA,UACtB;AAAA,UACA,cAAc,CAAC,UAAU;AAAA,QAC3B,CAAC;AAED,YAAI,KAAK,UAAU;AACjB,eAAK,WAAW;AAChB,eAAK,MAAM,IAAI,EAAE,MAAM,2BAAgB,cAAc,CAAC;AAAA,QACxD;AAAA,MACF,OAAO;AACL,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,2BAAgB;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":["import_utils","BaseSTT","BaseSpeechStream","ws"]}
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 { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport {\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 } from '../voice/io.js';\nimport {\n type SttServerEvent,\n type SttTranscriptEvent,\n sttServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, 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-conversationalai'\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 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;\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\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;\n encoding: STTEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: STTOptions<TModel>;\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?: TModel;\n language?: STTLanguages;\n baseURL?: string;\n encoding?: STTEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: STTOptions<TModel>;\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 } = 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 modelOptions,\n };\n }\n\n get label(): string {\n return 'inference.STT';\n }\n\n static fromModelString(modelString: string): STT<AnyString> {\n if (modelString.includes(':')) {\n const [model, language] = modelString.split(':') as [AnyString, STTLanguages];\n return new STT({ model, language });\n }\n return new STT({ model: modelString });\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 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 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 = { ...this.opts, ...opts };\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 = 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 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":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAgC;AAEhC,wBAAyC;AACzC,mBAAgC;AAChC,iBAAoB;AACpB,4BAAoC;AACpC,iBAMO;AACP,mBAAoE;AACpE,mBAAsF;AAEtF,wBAIO;AACP,IAAAA,gBAA6D;AAuE7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,yBAAyB;AAgBxB,MAAM,YAAsC,WAAAC,IAAQ;AAAA,EACjD;AAAA,EACA,UAAqC,oBAAI,IAAI;AAAA,EAErD,cAAU,gBAAI;AAAA,EAEd,YAAY,MAST;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,IAClB,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,OAAO,gBAAgB,aAAqC;AAC1D,QAAI,YAAY,SAAS,GAAG,GAAG;AAC7B,YAAM,CAAC,OAAO,QAAQ,IAAI,YAAY,MAAM,GAAG;AAC/C,aAAO,IAAI,IAAI,EAAE,OAAO,SAAS,CAAC;AAAA,IACpC;AACA,WAAO,IAAI,IAAI,EAAE,OAAO,YAAY,CAAC;AAAA,EACvC;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,yCAA4B,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;AAAA,EAEA,MAAM,UAAU,SAAqC;AACnD,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,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,UAAM,iCAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS,UAAM,yBAAU,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,WAAAC,aAAiB;AAAA,EACnE;AAAA,EACA,gBAAY,wBAAU,cAAc;AAAA,EACpC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,iBAAiB,IAAI,mBAAM;AAAA,EAC3B;AAAA,EACA;AAAA,EAER,cAAU,gBAAI;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,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,SAAK,eAAe,IAAI;AAAA,EAC1B;AAAA,EAEA,MAAgB,MAAqB;AACnC,WAAO,MAAM;AAEX,UAAI,KAAuB;AAC3B,UAAI,UAAU;AACd,UAAI,gBAAgB;AAEpB,YAAM,mBAAe,2CAAoC;AAEzD,YAAM,kBAAkB,MAAM;AAC5B,YAAI,QAAS;AACb,kBAAU;AACV,qBAAa,MAAM;AACnB,iCAAI;AACJ,iCAAI;AAAA,MACN;AAEA,YAAM,mBAAmB,OAAOC,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,iCAAe;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,uCAAqB,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,2BAAS,+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,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,IAAK,MAAM,GAAG,UAAU;AACxE,cAAM,iBAAiB,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,iBAAiB,IAAK,MAAM,GAAG,UAAU;AAC1F,cAAM,WAAW,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,MAAM,GAAG,UAAU;AACnE,cAAM,oBAAoB,kBAAK;AAAA,UAC7B,CAAC,EAAE,OAAO,MAAM,QAAQ,KAAK,CAAC,KAAK,eAAe,KAAK,OAAG,2BAAa,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,oBAAM;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,KAAK,YAAY,KAAK,KAAK,YAAY;AAExD,QAAI,CAAC,QAAQ,CAAC,QAAS;AAEvB,QAAI;AAEF,UAAI,CAAC,KAAK,UAAU;AAClB,aAAK,WAAW;AAChB,aAAK,MAAM,IAAI,EAAE,MAAM,2BAAgB,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,UAAuB;AAAA,YACtB,MAAM,KAAK;AAAA,YACX,WAAW,KAAK,QAAQ,KAAK;AAAA,YAC7B,SAAS,KAAK,MAAM,KAAK;AAAA,YACzB,iBAAiB,KAAK;AAAA,YACtB,YAAY,KAAK;AAAA,UACnB;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAS;AACX,YAAI,KAAK,iBAAiB,GAAG;AAC3B,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,2BAAgB;AAAA,YACtB;AAAA,YACA,kBAAkB,EAAE,eAAe,KAAK,eAAe;AAAA,UACzD,CAAC;AACD,eAAK,iBAAiB;AAAA,QACxB;AAEA,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,2BAAgB;AAAA,UACtB;AAAA,UACA,cAAc,CAAC,UAAU;AAAA,QAC3B,CAAC;AAED,YAAI,KAAK,UAAU;AACjB,eAAK,WAAW;AAChB,eAAK,MAAM,IAAI,EAAE,MAAM,2BAAgB,cAAc,CAAC;AAAA,QACxD;AAAA,MACF,OAAO;AACL,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,2BAAgB;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":["import_utils","BaseSTT","BaseSpeechStream","ws"]}
@@ -1 +1 @@
1
- {"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../../src/inference/stt.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAKpC,OAAO,EACL,GAAG,IAAI,OAAO,EACd,YAAY,IAAI,gBAAgB,EAEhC,KAAK,WAAW,EAEjB,MAAM,iBAAiB,CAAC;AACzB,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AAClF,OAAO,EAAE,KAAK,WAAW,EAAuD,MAAM,aAAa,CAAC;AACpG,OAAO,EAAE,KAAK,SAAS,EAAgC,MAAM,YAAY,CAAC;AAE1E,MAAM,MAAM,cAAc,GACtB,UAAU,GACV,iBAAiB,GACjB,yBAAyB,GACzB,yBAAyB,GACzB,kCAAkC,GAClC,iBAAiB,GACjB,yBAAyB,GACzB,yBAAyB,GACzB,2BAA2B,CAAC;AAEhC,MAAM,MAAM,cAAc,GAAG,UAAU,GAAG,sBAAsB,CAAC;AAEjE,MAAM,MAAM,gBAAgB,GAAG,YAAY,GAAG,gCAAgC,CAAC;AAE/E,MAAM,WAAW,eAAe;IAC9B,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,yBAAyB,CAAC,EAAE,MAAM,CAAC;CACpC;AAED,MAAM,WAAW,eAAe;IAC9B,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,QAAQ,CAAC,EAAE,KAAK,CAAC,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC,CAAC;IACnC,QAAQ,CAAC,EAAE,MAAM,EAAE,CAAC;IACpB,gBAAgB,CAAC,EAAE,OAAO,CAAC;IAC3B,QAAQ,CAAC,EAAE,OAAO,CAAC;IACnB,WAAW,CAAC,EAAE,OAAO,CAAC;CACvB;AAED,MAAM,WAAW,iBAAiB;IAChC,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,gCAAgC,CAAC,EAAE,MAAM,CAAC;IAC1C,sCAAsC,CAAC,EAAE,MAAM,CAAC;IAChD,gBAAgB,CAAC,EAAE,MAAM,CAAC;IAC1B,eAAe,CAAC,EAAE,MAAM,EAAE,CAAC;CAC5B;AAED,MAAM,MAAM,YAAY,GACpB,OAAO,GACP,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,SAAS,CAAC;AAEd,KAAK,UAAU,GAAG,cAAc,GAAG,cAAc,GAAG,gBAAgB,CAAC;AAErE,MAAM,MAAM,SAAS,GAAG,UAAU,GAAG,MAAM,GAAG,SAAS,CAAC;AAExD,MAAM,MAAM,iBAAiB,GAAG,GAAG,UAAU,IAAI,YAAY,EAAE,GAAG,SAAS,CAAC;AAE5E,MAAM,MAAM,UAAU,CAAC,MAAM,SAAS,SAAS,IAAI,MAAM,SAAS,cAAc,GAC5E,eAAe,GACf,MAAM,SAAS,cAAc,GAC3B,eAAe,GACf,MAAM,SAAS,gBAAgB,GAC7B,iBAAiB,GACjB,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAEhC,MAAM,MAAM,WAAW,GAAG,WAAW,CAAC;AAOtC,MAAM,WAAW,mBAAmB,CAAC,MAAM,SAAS,SAAS;IAC3D,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,YAAY,CAAC;IACxB,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,YAAY,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;CAClC;AAED;;GAEG;AACH,qBAAa,GAAG,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,OAAO;;IACxD,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,OAAO,CAAwC;gBAI3C,IAAI,CAAC,EAAE;QACjB,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,QAAQ,CAAC,EAAE,YAAY,CAAC;QACxB,OAAO,CAAC,EAAE,MAAM,CAAC;QACjB,QAAQ,CAAC,EAAE,WAAW,CAAC;QACvB,UAAU,CAAC,EAAE,MAAM,CAAC;QACpB,MAAM,CAAC,EAAE,MAAM,CAAC;QAChB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;KACnC;IAsCD,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,MAAM,CAAC,eAAe,CAAC,WAAW,EAAE,MAAM,GAAG,GAAG,CAAC,SAAS,CAAC;cAQ3C,UAAU,CAAC,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;IAIhE,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,UAAU,CAAC,CAAC,GAAG,IAAI;IAQ3F,MAAM,CAAC,OAAO,CAAC,EAAE;QACf,QAAQ,CAAC,EAAE,YAAY,GAAG,MAAM,CAAC;QACjC,WAAW,CAAC,EAAE,iBAAiB,CAAC;KACjC,GAAG,YAAY,CAAC,MAAM,CAAC;IAalB,SAAS,CAAC,OAAO,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC;CAgCrD;AAED,qBAAa,YAAY,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,gBAAgB;;IAC1E,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,SAAS,CAA6B;IAC9C,OAAO,CAAC,QAAQ,CAAS;IACzB,OAAO,CAAC,cAAc,CAAK;IAC3B,OAAO,CAAC,cAAc,CAAe;IACrC,OAAO,CAAC,GAAG,CAAc;IACzB,OAAO,CAAC,WAAW,CAAoB;gBAKrC,OAAO,EAAE,GAAG,CAAC,MAAM,CAAC,EACpB,IAAI,EAAE,mBAAmB,CAAC,MAAM,CAAC,EACjC,WAAW,EAAE,iBAAiB;IAQhC,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,UAAU,CAAC,CAAC,GAAG,IAAI;cAK3E,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IAiMpC,OAAO,CAAC,iBAAiB;CAgE1B"}
1
+ {"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../../src/inference/stt.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAKpC,OAAO,EACL,GAAG,IAAI,OAAO,EACd,YAAY,IAAI,gBAAgB,EAEhC,KAAK,WAAW,EAEjB,MAAM,iBAAiB,CAAC;AACzB,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AAClF,OAAO,EAAE,KAAK,WAAW,EAAuD,MAAM,aAAa,CAAC;AAOpG,OAAO,EAAE,KAAK,SAAS,EAAgC,MAAM,YAAY,CAAC;AAE1E,MAAM,MAAM,cAAc,GACtB,UAAU,GACV,iBAAiB,GACjB,yBAAyB,GACzB,yBAAyB,GACzB,kCAAkC,GAClC,iBAAiB,GACjB,yBAAyB,GACzB,yBAAyB,GACzB,2BAA2B,CAAC;AAEhC,MAAM,MAAM,cAAc,GAAG,UAAU,GAAG,sBAAsB,CAAC;AAEjE,MAAM,MAAM,gBAAgB,GAAG,YAAY,GAAG,gCAAgC,CAAC;AAE/E,MAAM,WAAW,eAAe;IAC9B,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,yBAAyB,CAAC,EAAE,MAAM,CAAC;CACpC;AAED,MAAM,WAAW,eAAe;IAC9B,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,QAAQ,CAAC,EAAE,KAAK,CAAC,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC,CAAC;IACnC,QAAQ,CAAC,EAAE,MAAM,EAAE,CAAC;IACpB,gBAAgB,CAAC,EAAE,OAAO,CAAC;IAC3B,QAAQ,CAAC,EAAE,OAAO,CAAC;IACnB,WAAW,CAAC,EAAE,OAAO,CAAC;CACvB;AAED,MAAM,WAAW,iBAAiB;IAChC,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,gCAAgC,CAAC,EAAE,MAAM,CAAC;IAC1C,sCAAsC,CAAC,EAAE,MAAM,CAAC;IAChD,gBAAgB,CAAC,EAAE,MAAM,CAAC;IAC1B,eAAe,CAAC,EAAE,MAAM,EAAE,CAAC;CAC5B;AAED,MAAM,MAAM,YAAY,GACpB,OAAO,GACP,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,SAAS,CAAC;AAEd,KAAK,UAAU,GAAG,cAAc,GAAG,cAAc,GAAG,gBAAgB,CAAC;AAErE,MAAM,MAAM,SAAS,GAAG,UAAU,GAAG,MAAM,GAAG,SAAS,CAAC;AAExD,MAAM,MAAM,iBAAiB,GAAG,GAAG,UAAU,IAAI,YAAY,EAAE,GAAG,SAAS,CAAC;AAE5E,MAAM,MAAM,UAAU,CAAC,MAAM,SAAS,SAAS,IAAI,MAAM,SAAS,cAAc,GAC5E,eAAe,GACf,MAAM,SAAS,cAAc,GAC3B,eAAe,GACf,MAAM,SAAS,gBAAgB,GAC7B,iBAAiB,GACjB,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAEhC,MAAM,MAAM,WAAW,GAAG,WAAW,CAAC;AAOtC,MAAM,WAAW,mBAAmB,CAAC,MAAM,SAAS,SAAS;IAC3D,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,YAAY,CAAC;IACxB,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,YAAY,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;CAClC;AAED;;GAEG;AACH,qBAAa,GAAG,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,OAAO;;IACxD,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,OAAO,CAAwC;gBAI3C,IAAI,CAAC,EAAE;QACjB,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,QAAQ,CAAC,EAAE,YAAY,CAAC;QACxB,OAAO,CAAC,EAAE,MAAM,CAAC;QACjB,QAAQ,CAAC,EAAE,WAAW,CAAC;QACvB,UAAU,CAAC,EAAE,MAAM,CAAC;QACpB,MAAM,CAAC,EAAE,MAAM,CAAC;QAChB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;KACnC;IAsCD,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,MAAM,CAAC,eAAe,CAAC,WAAW,EAAE,MAAM,GAAG,GAAG,CAAC,SAAS,CAAC;cAQ3C,UAAU,CAAC,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;IAIhE,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,UAAU,CAAC,CAAC,GAAG,IAAI;IAQ3F,MAAM,CAAC,OAAO,CAAC,EAAE;QACf,QAAQ,CAAC,EAAE,YAAY,GAAG,MAAM,CAAC;QACjC,WAAW,CAAC,EAAE,iBAAiB,CAAC;KACjC,GAAG,YAAY,CAAC,MAAM,CAAC;IAalB,SAAS,CAAC,OAAO,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC;CAgCrD;AAED,qBAAa,YAAY,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,gBAAgB;;IAC1E,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,SAAS,CAA6B;IAC9C,OAAO,CAAC,QAAQ,CAAS;IACzB,OAAO,CAAC,cAAc,CAAK;IAC3B,OAAO,CAAC,cAAc,CAAe;IACrC,OAAO,CAAC,GAAG,CAAc;IACzB,OAAO,CAAC,WAAW,CAAoB;gBAKrC,OAAO,EAAE,GAAG,CAAC,MAAM,CAAC,EACpB,IAAI,EAAE,mBAAmB,CAAC,MAAM,CAAC,EACjC,WAAW,EAAE,iBAAiB;IAQhC,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,UAAU,CAAC,CAAC,GAAG,IAAI;cAK3E,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IAmMpC,OAAO,CAAC,iBAAiB;CAyE1B"}
@@ -10,6 +10,9 @@ import {
10
10
  } from "../stt/index.js";
11
11
  import { DEFAULT_API_CONNECT_OPTIONS } from "../types.js";
12
12
  import { Event, Task, cancelAndWait, shortuuid, waitForAbort } from "../utils.js";
13
+ import {
14
+ sttServerEventSchema
15
+ } from "./api_protos.js";
13
16
  import { connectWs, createAccessToken } from "./utils.js";
14
17
  const DEFAULT_ENCODING = "pcm_s16le";
15
18
  const DEFAULT_SAMPLE_RATE = 16e3;
@@ -20,7 +23,7 @@ class STT extends BaseSTT {
20
23
  streams = /* @__PURE__ */ new Set();
21
24
  #logger = log();
22
25
  constructor(opts) {
23
- super({ streaming: true, interimResults: true });
26
+ super({ streaming: true, interimResults: true, alignedTranscript: "word" });
24
27
  const {
25
28
  model,
26
29
  language,
@@ -222,9 +225,16 @@ class SpeechStream extends BaseSpeechStream {
222
225
  const result = await reader.read();
223
226
  if (signal.aborted) return;
224
227
  if (result.done) return;
225
- const json = result.value;
226
- const type = json.type;
227
- switch (type) {
228
+ const parseResult = await sttServerEventSchema.safeParseAsync(result.value);
229
+ if (!parseResult.success) {
230
+ this.#logger.warn(
231
+ { error: parseResult.error, rawData: result.value },
232
+ "Failed to parse STT server event"
233
+ );
234
+ continue;
235
+ }
236
+ const event = parseResult.data;
237
+ switch (event.type) {
228
238
  case "session.created":
229
239
  case "session.finalized":
230
240
  break;
@@ -233,21 +243,15 @@ class SpeechStream extends BaseSpeechStream {
233
243
  resourceCleanup();
234
244
  break;
235
245
  case "interim_transcript":
236
- this.processTranscript(json, false);
246
+ this.processTranscript(event, false);
237
247
  break;
238
248
  case "final_transcript":
239
- this.processTranscript(json, true);
249
+ this.processTranscript(event, true);
240
250
  break;
241
251
  case "error":
242
- this.#logger.error({ error: json }, "Received error from LiveKit STT");
252
+ this.#logger.error({ error: event }, "Received error from LiveKit STT");
243
253
  resourceCleanup();
244
- throw new APIError(`LiveKit STT returned error: ${JSON.stringify(json)}`);
245
- default:
246
- this.#logger.warn(
247
- { message: json },
248
- "Received unexpected message from LiveKit STT"
249
- );
250
- break;
254
+ throw new APIError(`LiveKit STT returned error: ${JSON.stringify(event)}`);
251
255
  }
252
256
  }
253
257
  } finally {
@@ -290,9 +294,9 @@ class SpeechStream extends BaseSpeechStream {
290
294
  }
291
295
  processTranscript(data, isFinal) {
292
296
  if (this.queue.closed) return;
293
- const requestId = data.request_id ?? this.requestId;
294
- const text = data.transcript ?? "";
295
- const language = data.language ?? this.opts.language ?? "en";
297
+ const requestId = data.session_id || this.requestId;
298
+ const text = data.transcript;
299
+ const language = data.language || this.opts.language || "en";
296
300
  if (!text && !isFinal) return;
297
301
  try {
298
302
  if (!this.speaking) {
@@ -301,10 +305,19 @@ class SpeechStream extends BaseSpeechStream {
301
305
  }
302
306
  const speechData = {
303
307
  language,
304
- startTime: data.start ?? 0,
305
- endTime: data.duration ?? 0,
306
- confidence: data.confidence ?? 1,
307
- text
308
+ startTime: this.startTimeOffset + data.start,
309
+ endTime: this.startTimeOffset + data.start + data.duration,
310
+ confidence: data.confidence,
311
+ text,
312
+ words: data.words.map(
313
+ (word) => ({
314
+ text: word.word,
315
+ startTime: word.start + this.startTimeOffset,
316
+ endTime: word.end + this.startTimeOffset,
317
+ startTimeOffset: this.startTimeOffset,
318
+ confidence: word.confidence
319
+ })
320
+ )
308
321
  };
309
322
  if (isFinal) {
310
323
  if (this.speechDuration > 0) {
@@ -1 +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 { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport {\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 AnyString, 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-conversationalai'\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 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;\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\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;\n encoding: STTEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: STTOptions<TModel>;\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?: TModel;\n language?: STTLanguages;\n baseURL?: string;\n encoding?: STTEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: 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 modelOptions = {} 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 modelOptions,\n };\n }\n\n get label(): string {\n return 'inference.STT';\n }\n\n static fromModelString(modelString: string): STT<AnyString> {\n if (modelString.includes(':')) {\n const [model, language] = modelString.split(':') as [AnyString, STTLanguages];\n return new STT({ model, language });\n }\n return new STT({ model: modelString });\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 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 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 = { ...this.opts, ...opts };\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 type SttServerEvent = Record<string, any>;\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 const json = result.value;\n const type = json.type as string | undefined;\n\n switch (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(json, false);\n break;\n case 'final_transcript':\n this.processTranscript(json, true);\n break;\n case 'error':\n this.#logger.error({ error: json }, 'Received error from LiveKit STT');\n resourceCleanup();\n throw new APIError(`LiveKit STT returned error: ${JSON.stringify(json)}`);\n default:\n this.#logger.warn(\n { message: json },\n 'Received unexpected message from LiveKit STT',\n );\n break;\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n ws = await this.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: Record<string, any>, 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.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 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: 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 } 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,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,SAAyB,WAAW,yBAAyB;AAuE7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,yBAAyB;AAgBxB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAqC,oBAAI,IAAI;AAAA,EAErD,UAAU,IAAI;AAAA,EAEd,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,eAAe,CAAC;AAAA,IAClB,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,OAAO,gBAAgB,aAAqC;AAC1D,QAAI,YAAY,SAAS,GAAG,GAAG;AAC7B,YAAM,CAAC,OAAO,QAAQ,IAAI,YAAY,MAAM,GAAG;AAC/C,aAAO,IAAI,IAAI,EAAE,OAAO,SAAS,CAAC;AAAA,IACpC;AACA,WAAO,IAAI,IAAI,EAAE,OAAO,YAAY,CAAC;AAAA,EACvC;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;AAAA,EAEA,MAAM,UAAU,SAAqC;AACnD,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,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,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,SAAK,eAAe,IAAI;AAAA,EAC1B;AAAA,EAEA,MAAgB,MAAqB;AACnC,WAAO,MAAM;AAEX,UAAI,KAAuB;AAC3B,UAAI,UAAU;AACd,UAAI,gBAAgB;AAGpB,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;AAEjB,kBAAM,OAAO,OAAO;AACpB,kBAAM,OAAO,KAAK;AAElB,oBAAQ,MAAM;AAAA,cACZ,KAAK;AAAA,cACL,KAAK;AACH;AAAA,cACF,KAAK;AACH,gCAAgB;AAChB,gCAAgB;AAChB;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,MAAM,KAAK;AAClC;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,MAAM,IAAI;AACjC;AAAA,cACF,KAAK;AACH,qBAAK,QAAQ,MAAM,EAAE,OAAO,KAAK,GAAG,iCAAiC;AACrE,gCAAgB;AAChB,sBAAM,IAAI,SAAS,+BAA+B,KAAK,UAAU,IAAI,CAAC,EAAE;AAAA,cAC1E;AACE,qBAAK,QAAQ;AAAA,kBACX,EAAE,SAAS,KAAK;AAAA,kBAChB;AAAA,gBACF;AACA;AAAA,YACJ;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,MAA2B,SAAkB;AAErE,QAAI,KAAK,MAAM,OAAQ;AAEvB,UAAM,YAAY,KAAK,cAAc,KAAK;AAC1C,UAAM,OAAO,KAAK,cAAc;AAChC,UAAM,WAAW,KAAK,YAAY,KAAK,KAAK,YAAY;AAExD,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,SAAS;AAAA,QACzB,SAAS,KAAK,YAAY;AAAA,QAC1B,YAAY,KAAK,cAAc;AAAA,QAC/B;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"]}
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 { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport {\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 } from '../voice/io.js';\nimport {\n type SttServerEvent,\n type SttTranscriptEvent,\n sttServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, 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-conversationalai'\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 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;\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\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;\n encoding: STTEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: STTOptions<TModel>;\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?: TModel;\n language?: STTLanguages;\n baseURL?: string;\n encoding?: STTEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: STTOptions<TModel>;\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 } = 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 modelOptions,\n };\n }\n\n get label(): string {\n return 'inference.STT';\n }\n\n static fromModelString(modelString: string): STT<AnyString> {\n if (modelString.includes(':')) {\n const [model, language] = modelString.split(':') as [AnyString, STTLanguages];\n return new STT({ model, language });\n }\n return new STT({ model: modelString });\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 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 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 = { ...this.opts, ...opts };\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 = 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 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,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;AAEtF;AAAA,EAGE;AAAA,OACK;AACP,SAAyB,WAAW,yBAAyB;AAuE7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,yBAAyB;AAgBxB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAqC,oBAAI,IAAI;AAAA,EAErD,UAAU,IAAI;AAAA,EAEd,YAAY,MAST;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,IAClB,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,OAAO,gBAAgB,aAAqC;AAC1D,QAAI,YAAY,SAAS,GAAG,GAAG;AAC7B,YAAM,CAAC,OAAO,QAAQ,IAAI,YAAY,MAAM,GAAG;AAC/C,aAAO,IAAI,IAAI,EAAE,OAAO,SAAS,CAAC;AAAA,IACpC;AACA,WAAO,IAAI,IAAI,EAAE,OAAO,YAAY,CAAC;AAAA,EACvC;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;AAAA,EAEA,MAAM,UAAU,SAAqC;AACnD,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,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,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,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,KAAK,YAAY,KAAK,KAAK,YAAY;AAExD,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,UAAuB;AAAA,YACtB,MAAM,KAAK;AAAA,YACX,WAAW,KAAK,QAAQ,KAAK;AAAA,YAC7B,SAAS,KAAK,MAAM,KAAK;AAAA,YACzB,iBAAiB,KAAK;AAAA,YACtB,YAAY,KAAK;AAAA,UACnB;AAAA,QACF;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"]}
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/ipc/inference_proc_executor.ts","../../../node_modules/.pnpm/tsup@8.4.0_@microsoft+api-extractor@7.43.7_@types+node@22.15.30__postcss@8.5.6_tsx@4.20.4_typescript@5.4.5/node_modules/tsup/assets/cjs_shims.js"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { ChildProcess } from 'node:child_process';\nimport { fork } from 'node:child_process';\nimport { extname } from 'node:path';\nimport { log } from '../log.js';\nimport { shortuuid } from '../utils.js';\nimport type { InferenceExecutor } from './inference_executor.js';\nimport type { IPCMessage } from './message.js';\nimport { SupervisedProc } from './supervised_proc.js';\n\nclass PendingInference {\n promise = new Promise<{ requestId: string; data: unknown; error?: Error }>((resolve) => {\n this.resolve = resolve;\n });\n resolve(arg: { requestId: string; data: unknown; error?: Error }) {\n arg;\n }\n}\n\nconst currentFileExtension = extname(import.meta.url);\n\nexport class InferenceProcExecutor extends SupervisedProc implements InferenceExecutor {\n #runners: { [id: string]: string };\n #activeRequests: { [id: string]: PendingInference } = {};\n #logger = log();\n\n constructor({\n runners,\n initializeTimeout,\n closeTimeout,\n memoryWarnMB,\n memoryLimitMB,\n pingInterval,\n pingTimeout,\n highPingThreshold,\n }: {\n runners: { [id: string]: string };\n initializeTimeout: number;\n closeTimeout: number;\n memoryWarnMB: number;\n memoryLimitMB: number;\n pingInterval: number;\n pingTimeout: number;\n highPingThreshold: number;\n }) {\n super(\n initializeTimeout,\n closeTimeout,\n memoryWarnMB,\n memoryLimitMB,\n pingInterval,\n pingTimeout,\n highPingThreshold,\n );\n this.#runners = runners;\n }\n\n createProcess(): ChildProcess {\n const forkUrl = new URL(`./inference_proc_lazy_main${currentFileExtension}`, import.meta.url);\n\n // When running via tsx/ts-node (TypeScript files), we need to inherit the parent's\n // execArgv so the child process can also execute TypeScript with the same loader\n const isTypeScript = currentFileExtension === '.ts';\n const forkOptions = isTypeScript ? { execArgv: process.execArgv } : undefined;\n\n return fork(forkUrl, [JSON.stringify(this.#runners)], forkOptions);\n }\n\n async mainTask(proc: ChildProcess) {\n proc.on('message', (msg: IPCMessage) => {\n switch (msg.case) {\n case 'inferenceResponse':\n const res = this.#activeRequests[msg.value.requestId];\n delete this.#activeRequests[msg.value.requestId];\n if (!res) {\n this.#logger\n .child({ requestId: msg.value.requestId })\n .warn('received unexpected inference response');\n return;\n }\n\n res.resolve(msg.value);\n }\n });\n }\n\n async doInference(method: string, data: unknown): Promise<unknown> {\n const requestId = shortuuid('inference_req_');\n const fut = new PendingInference();\n this.proc!.send({ case: 'inferenceRequest', value: { requestId, method, data } });\n this.#activeRequests[requestId] = fut;\n\n const res = await fut.promise;\n if (res.error) {\n throw new Error(`inference of ${method} failed: ${res.error}`);\n }\n return res.data;\n }\n}\n","// Shim globals in cjs bundle\n// There's a weird bug that esbuild will always inject importMetaUrl\n// if we export it as `const importMetaUrl = ... __filename ...`\n// But using a function will not cause this issue\n\nconst getImportMetaUrl = () =>\n typeof document === 'undefined'\n ? new URL(`file:${__filename}`).href\n : (document.currentScript && document.currentScript.src) ||\n new URL('main.js', document.baseURI).href\n\nexport const importMetaUrl = /* @__PURE__ */ getImportMetaUrl()\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;ACKA,IAAM,mBAAmB,MACvB,OAAO,aAAa,cAChB,IAAI,IAAI,QAAQ,UAAU,EAAE,EAAE,OAC7B,SAAS,iBAAiB,SAAS,cAAc,OAClD,IAAI,IAAI,WAAW,SAAS,OAAO,EAAE;AAEpC,IAAM,gBAAgC,iCAAiB;ADP9D,gCAAqB;AACrB,uBAAwB;AACxB,iBAAoB;AACpB,mBAA0B;AAG1B,6BAA+B;AAE/B,MAAM,iBAAiB;AAAA,EACrB,UAAU,IAAI,QAA6D,CAAC,YAAY;AACtF,SAAK,UAAU;AAAA,EACjB,CAAC;AAAA,EACD,QAAQ,KAA0D;AAChE;AAAA,EACF;AACF;AAEA,MAAM,2BAAuB,0BAAQ,aAAe;AAE7C,MAAM,8BAA8B,sCAA4C;AAAA,EACrF;AAAA,EACA,kBAAsD,CAAC;AAAA,EACvD,cAAU,gBAAI;AAAA,EAEd,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GASG;AACD;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,SAAK,WAAW;AAAA,EAClB;AAAA,EAEA,gBAA8B;AAC5B,UAAM,UAAU,IAAI,IAAI,6BAA6B,oBAAoB,IAAI,aAAe;AAI5F,UAAM,eAAe,yBAAyB;AAC9C,UAAM,cAAc,eAAe,EAAE,UAAU,QAAQ,SAAS,IAAI;AAEpE,eAAO,gCAAK,SAAS,CAAC,KAAK,UAAU,KAAK,QAAQ,CAAC,GAAG,WAAW;AAAA,EACnE;AAAA,EAEA,MAAM,SAAS,MAAoB;AACjC,SAAK,GAAG,WAAW,CAAC,QAAoB;AACtC,cAAQ,IAAI,MAAM;AAAA,QAChB,KAAK;AACH,gBAAM,MAAM,KAAK,gBAAgB,IAAI,MAAM,SAAS;AACpD,iBAAO,KAAK,gBAAgB,IAAI,MAAM,SAAS;AAC/C,cAAI,CAAC,KAAK;AACR,iBAAK,QACF,MAAM,EAAE,WAAW,IAAI,MAAM,UAAU,CAAC,EACxC,KAAK,wCAAwC;AAChD;AAAA,UACF;AAEA,cAAI,QAAQ,IAAI,KAAK;AAAA,MACzB;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,YAAY,QAAgB,MAAiC;AACjE,UAAM,gBAAY,wBAAU,gBAAgB;AAC5C,UAAM,MAAM,IAAI,iBAAiB;AACjC,SAAK,KAAM,KAAK,EAAE,MAAM,oBAAoB,OAAO,EAAE,WAAW,QAAQ,KAAK,EAAE,CAAC;AAChF,SAAK,gBAAgB,SAAS,IAAI;AAElC,UAAM,MAAM,MAAM,IAAI;AACtB,QAAI,IAAI,OAAO;AACb,YAAM,IAAI,MAAM,gBAAgB,MAAM,YAAY,IAAI,KAAK,EAAE;AAAA,IAC/D;AACA,WAAO,IAAI;AAAA,EACb;AACF;","names":[]}
1
+ {"version":3,"sources":["../../src/ipc/inference_proc_executor.ts","../../../node_modules/.pnpm/tsup@8.4.0_@microsoft+api-extractor@7.43.7_@types+node@22.15.30__postcss@8.5.6_tsx@4.21.0_typescript@5.4.5/node_modules/tsup/assets/cjs_shims.js"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { ChildProcess } from 'node:child_process';\nimport { fork } from 'node:child_process';\nimport { extname } from 'node:path';\nimport { log } from '../log.js';\nimport { shortuuid } from '../utils.js';\nimport type { InferenceExecutor } from './inference_executor.js';\nimport type { IPCMessage } from './message.js';\nimport { SupervisedProc } from './supervised_proc.js';\n\nclass PendingInference {\n promise = new Promise<{ requestId: string; data: unknown; error?: Error }>((resolve) => {\n this.resolve = resolve;\n });\n resolve(arg: { requestId: string; data: unknown; error?: Error }) {\n arg;\n }\n}\n\nconst currentFileExtension = extname(import.meta.url);\n\nexport class InferenceProcExecutor extends SupervisedProc implements InferenceExecutor {\n #runners: { [id: string]: string };\n #activeRequests: { [id: string]: PendingInference } = {};\n #logger = log();\n\n constructor({\n runners,\n initializeTimeout,\n closeTimeout,\n memoryWarnMB,\n memoryLimitMB,\n pingInterval,\n pingTimeout,\n highPingThreshold,\n }: {\n runners: { [id: string]: string };\n initializeTimeout: number;\n closeTimeout: number;\n memoryWarnMB: number;\n memoryLimitMB: number;\n pingInterval: number;\n pingTimeout: number;\n highPingThreshold: number;\n }) {\n super(\n initializeTimeout,\n closeTimeout,\n memoryWarnMB,\n memoryLimitMB,\n pingInterval,\n pingTimeout,\n highPingThreshold,\n );\n this.#runners = runners;\n }\n\n createProcess(): ChildProcess {\n const forkUrl = new URL(`./inference_proc_lazy_main${currentFileExtension}`, import.meta.url);\n\n // When running via tsx/ts-node (TypeScript files), we need to inherit the parent's\n // execArgv so the child process can also execute TypeScript with the same loader\n const isTypeScript = currentFileExtension === '.ts';\n const forkOptions = isTypeScript ? { execArgv: process.execArgv } : undefined;\n\n return fork(forkUrl, [JSON.stringify(this.#runners)], forkOptions);\n }\n\n async mainTask(proc: ChildProcess) {\n proc.on('message', (msg: IPCMessage) => {\n switch (msg.case) {\n case 'inferenceResponse':\n const res = this.#activeRequests[msg.value.requestId];\n delete this.#activeRequests[msg.value.requestId];\n if (!res) {\n this.#logger\n .child({ requestId: msg.value.requestId })\n .warn('received unexpected inference response');\n return;\n }\n\n res.resolve(msg.value);\n }\n });\n }\n\n async doInference(method: string, data: unknown): Promise<unknown> {\n const requestId = shortuuid('inference_req_');\n const fut = new PendingInference();\n this.proc!.send({ case: 'inferenceRequest', value: { requestId, method, data } });\n this.#activeRequests[requestId] = fut;\n\n const res = await fut.promise;\n if (res.error) {\n throw new Error(`inference of ${method} failed: ${res.error}`);\n }\n return res.data;\n }\n}\n","// Shim globals in cjs bundle\n// There's a weird bug that esbuild will always inject importMetaUrl\n// if we export it as `const importMetaUrl = ... __filename ...`\n// But using a function will not cause this issue\n\nconst getImportMetaUrl = () =>\n typeof document === 'undefined'\n ? new URL(`file:${__filename}`).href\n : (document.currentScript && document.currentScript.src) ||\n new URL('main.js', document.baseURI).href\n\nexport const importMetaUrl = /* @__PURE__ */ getImportMetaUrl()\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;ACKA,IAAM,mBAAmB,MACvB,OAAO,aAAa,cAChB,IAAI,IAAI,QAAQ,UAAU,EAAE,EAAE,OAC7B,SAAS,iBAAiB,SAAS,cAAc,OAClD,IAAI,IAAI,WAAW,SAAS,OAAO,EAAE;AAEpC,IAAM,gBAAgC,iCAAiB;ADP9D,gCAAqB;AACrB,uBAAwB;AACxB,iBAAoB;AACpB,mBAA0B;AAG1B,6BAA+B;AAE/B,MAAM,iBAAiB;AAAA,EACrB,UAAU,IAAI,QAA6D,CAAC,YAAY;AACtF,SAAK,UAAU;AAAA,EACjB,CAAC;AAAA,EACD,QAAQ,KAA0D;AAChE;AAAA,EACF;AACF;AAEA,MAAM,2BAAuB,0BAAQ,aAAe;AAE7C,MAAM,8BAA8B,sCAA4C;AAAA,EACrF;AAAA,EACA,kBAAsD,CAAC;AAAA,EACvD,cAAU,gBAAI;AAAA,EAEd,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GASG;AACD;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,SAAK,WAAW;AAAA,EAClB;AAAA,EAEA,gBAA8B;AAC5B,UAAM,UAAU,IAAI,IAAI,6BAA6B,oBAAoB,IAAI,aAAe;AAI5F,UAAM,eAAe,yBAAyB;AAC9C,UAAM,cAAc,eAAe,EAAE,UAAU,QAAQ,SAAS,IAAI;AAEpE,eAAO,gCAAK,SAAS,CAAC,KAAK,UAAU,KAAK,QAAQ,CAAC,GAAG,WAAW;AAAA,EACnE;AAAA,EAEA,MAAM,SAAS,MAAoB;AACjC,SAAK,GAAG,WAAW,CAAC,QAAoB;AACtC,cAAQ,IAAI,MAAM;AAAA,QAChB,KAAK;AACH,gBAAM,MAAM,KAAK,gBAAgB,IAAI,MAAM,SAAS;AACpD,iBAAO,KAAK,gBAAgB,IAAI,MAAM,SAAS;AAC/C,cAAI,CAAC,KAAK;AACR,iBAAK,QACF,MAAM,EAAE,WAAW,IAAI,MAAM,UAAU,CAAC,EACxC,KAAK,wCAAwC;AAChD;AAAA,UACF;AAEA,cAAI,QAAQ,IAAI,KAAK;AAAA,MACzB;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,YAAY,QAAgB,MAAiC;AACjE,UAAM,gBAAY,wBAAU,gBAAgB;AAC5C,UAAM,MAAM,IAAI,iBAAiB;AACjC,SAAK,KAAM,KAAK,EAAE,MAAM,oBAAoB,OAAO,EAAE,WAAW,QAAQ,KAAK,EAAE,CAAC;AAChF,SAAK,gBAAgB,SAAS,IAAI;AAElC,UAAM,MAAM,MAAM,IAAI;AACtB,QAAI,IAAI,OAAO;AACb,YAAM,IAAI,MAAM,gBAAgB,MAAM,YAAY,IAAI,KAAK,EAAE;AAAA,IAC/D;AACA,WAAO,IAAI;AAAA,EACb;AACF;","names":[]}
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/ipc/job_proc_executor.ts","../../../node_modules/.pnpm/tsup@8.4.0_@microsoft+api-extractor@7.43.7_@types+node@22.15.30__postcss@8.5.6_tsx@4.20.4_typescript@5.4.5/node_modules/tsup/assets/cjs_shims.js"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { ChildProcess } from 'node:child_process';\nimport { fork } from 'node:child_process';\nimport { extname } from 'node:path';\nimport type { RunningJobInfo } from '../job.js';\nimport { log } from '../log.js';\nimport type { InferenceExecutor } from './inference_executor.js';\nimport type { JobExecutor } from './job_executor.js';\nimport { JobStatus } from './job_executor.js';\nimport type { IPCMessage } from './message.js';\nimport { SupervisedProc } from './supervised_proc.js';\n\nconst currentFileExtension = extname(import.meta.url);\n\nexport class JobProcExecutor extends SupervisedProc implements JobExecutor {\n #userArgs?: any; // eslint-disable-line @typescript-eslint/no-explicit-any\n #jobStatus?: JobStatus;\n #runningJob?: RunningJobInfo;\n #agent: string;\n #inferenceExecutor?: InferenceExecutor;\n #inferenceTasks: Promise<void>[] = [];\n #logger = log();\n\n constructor(\n agent: string,\n inferenceExecutor: InferenceExecutor | undefined,\n initializeTimeout: number,\n closeTimeout: number,\n memoryWarnMB: number,\n memoryLimitMB: number,\n pingInterval: number,\n pingTimeout: number,\n highPingThreshold: number,\n ) {\n super(\n initializeTimeout,\n closeTimeout,\n memoryWarnMB,\n memoryLimitMB,\n pingInterval,\n pingTimeout,\n highPingThreshold,\n );\n this.#agent = agent;\n this.#inferenceExecutor = inferenceExecutor;\n }\n\n get status(): JobStatus {\n if (this.#jobStatus) {\n return this.#jobStatus;\n }\n throw new Error('job status not available');\n }\n\n get userArguments(): any {\n return this.#userArgs;\n }\n\n set userArguments(args: any) {\n this.#userArgs = args;\n }\n\n get runningJob(): RunningJobInfo | undefined {\n return this.#runningJob;\n }\n\n createProcess(): ChildProcess {\n const forkUrl = new URL(`./job_proc_lazy_main${currentFileExtension}`, import.meta.url);\n\n // When running via tsx/ts-node (TypeScript files), we need to inherit the parent's\n // execArgv so the child process can also execute TypeScript with the same loader\n const isTypeScript = currentFileExtension === '.ts';\n const forkOptions = isTypeScript ? { execArgv: process.execArgv } : undefined;\n\n return fork(forkUrl, [this.#agent], forkOptions);\n }\n\n async mainTask(proc: ChildProcess) {\n proc.on('message', (msg: IPCMessage) => {\n switch (msg.case) {\n case 'inferenceRequest':\n this.#inferenceTasks.push(this.#doInferenceTask(proc, msg.value));\n }\n });\n }\n\n async #doInferenceTask(\n proc: ChildProcess,\n req: { method: string; requestId: string; data: unknown },\n ) {\n if (!this.#inferenceExecutor) {\n this.#logger.warn('inference request received but no inference executor');\n proc.send({\n case: 'inferenceResponse',\n value: { requestId: req.requestId, error: new Error('no inference executor') },\n });\n return;\n }\n\n try {\n const data = await this.#inferenceExecutor.doInference(req.method, req.data);\n proc.send({ case: 'inferenceResponse', value: { requestId: req.requestId, data } });\n } catch (error) {\n proc.send({ case: 'inferenceResponse', value: { requestId: req.requestId, error } });\n }\n }\n\n async launchJob(info: RunningJobInfo) {\n if (this.#runningJob) {\n throw Error('process already has a running job');\n }\n if (!this.init.done) {\n throw Error('process not initialized');\n }\n this.#jobStatus = JobStatus.RUNNING;\n this.#runningJob = info;\n\n this.proc!.send({ case: 'startJobRequest', value: { runningJob: info } });\n }\n}\n","// Shim globals in cjs bundle\n// There's a weird bug that esbuild will always inject importMetaUrl\n// if we export it as `const importMetaUrl = ... __filename ...`\n// But using a function will not cause this issue\n\nconst getImportMetaUrl = () =>\n typeof document === 'undefined'\n ? new URL(`file:${__filename}`).href\n : (document.currentScript && document.currentScript.src) ||\n new URL('main.js', document.baseURI).href\n\nexport const importMetaUrl = /* @__PURE__ */ getImportMetaUrl()\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;ACKA,IAAM,mBAAmB,MACvB,OAAO,aAAa,cAChB,IAAI,IAAI,QAAQ,UAAU,EAAE,EAAE,OAC7B,SAAS,iBAAiB,SAAS,cAAc,OAClD,IAAI,IAAI,WAAW,SAAS,OAAO,EAAE;AAEpC,IAAM,gBAAgC,iCAAiB;ADP9D,gCAAqB;AACrB,uBAAwB;AAExB,iBAAoB;AAGpB,0BAA0B;AAE1B,6BAA+B;AAE/B,MAAM,2BAAuB,0BAAQ,aAAe;AAE7C,MAAM,wBAAwB,sCAAsC;AAAA,EACzE;AAAA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,kBAAmC,CAAC;AAAA,EACpC,cAAU,gBAAI;AAAA,EAEd,YACE,OACA,mBACA,mBACA,cACA,cACA,eACA,cACA,aACA,mBACA;AACA;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,SAAK,SAAS;AACd,SAAK,qBAAqB;AAAA,EAC5B;AAAA,EAEA,IAAI,SAAoB;AACtB,QAAI,KAAK,YAAY;AACnB,aAAO,KAAK;AAAA,IACd;AACA,UAAM,IAAI,MAAM,0BAA0B;AAAA,EAC5C;AAAA,EAEA,IAAI,gBAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAAc,MAAW;AAC3B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,IAAI,aAAyC;AAC3C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,gBAA8B;AAC5B,UAAM,UAAU,IAAI,IAAI,uBAAuB,oBAAoB,IAAI,aAAe;AAItF,UAAM,eAAe,yBAAyB;AAC9C,UAAM,cAAc,eAAe,EAAE,UAAU,QAAQ,SAAS,IAAI;AAEpE,eAAO,gCAAK,SAAS,CAAC,KAAK,MAAM,GAAG,WAAW;AAAA,EACjD;AAAA,EAEA,MAAM,SAAS,MAAoB;AACjC,SAAK,GAAG,WAAW,CAAC,QAAoB;AACtC,cAAQ,IAAI,MAAM;AAAA,QAChB,KAAK;AACH,eAAK,gBAAgB,KAAK,KAAK,iBAAiB,MAAM,IAAI,KAAK,CAAC;AAAA,MACpE;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,iBACJ,MACA,KACA;AACA,QAAI,CAAC,KAAK,oBAAoB;AAC5B,WAAK,QAAQ,KAAK,sDAAsD;AACxE,WAAK,KAAK;AAAA,QACR,MAAM;AAAA,QACN,OAAO,EAAE,WAAW,IAAI,WAAW,OAAO,IAAI,MAAM,uBAAuB,EAAE;AAAA,MAC/E,CAAC;AACD;AAAA,IACF;AAEA,QAAI;AACF,YAAM,OAAO,MAAM,KAAK,mBAAmB,YAAY,IAAI,QAAQ,IAAI,IAAI;AAC3E,WAAK,KAAK,EAAE,MAAM,qBAAqB,OAAO,EAAE,WAAW,IAAI,WAAW,KAAK,EAAE,CAAC;AAAA,IACpF,SAAS,OAAO;AACd,WAAK,KAAK,EAAE,MAAM,qBAAqB,OAAO,EAAE,WAAW,IAAI,WAAW,MAAM,EAAE,CAAC;AAAA,IACrF;AAAA,EACF;AAAA,EAEA,MAAM,UAAU,MAAsB;AACpC,QAAI,KAAK,aAAa;AACpB,YAAM,MAAM,mCAAmC;AAAA,IACjD;AACA,QAAI,CAAC,KAAK,KAAK,MAAM;AACnB,YAAM,MAAM,yBAAyB;AAAA,IACvC;AACA,SAAK,aAAa,8BAAU;AAC5B,SAAK,cAAc;AAEnB,SAAK,KAAM,KAAK,EAAE,MAAM,mBAAmB,OAAO,EAAE,YAAY,KAAK,EAAE,CAAC;AAAA,EAC1E;AACF;","names":[]}
1
+ {"version":3,"sources":["../../src/ipc/job_proc_executor.ts","../../../node_modules/.pnpm/tsup@8.4.0_@microsoft+api-extractor@7.43.7_@types+node@22.15.30__postcss@8.5.6_tsx@4.21.0_typescript@5.4.5/node_modules/tsup/assets/cjs_shims.js"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { ChildProcess } from 'node:child_process';\nimport { fork } from 'node:child_process';\nimport { extname } from 'node:path';\nimport type { RunningJobInfo } from '../job.js';\nimport { log } from '../log.js';\nimport type { InferenceExecutor } from './inference_executor.js';\nimport type { JobExecutor } from './job_executor.js';\nimport { JobStatus } from './job_executor.js';\nimport type { IPCMessage } from './message.js';\nimport { SupervisedProc } from './supervised_proc.js';\n\nconst currentFileExtension = extname(import.meta.url);\n\nexport class JobProcExecutor extends SupervisedProc implements JobExecutor {\n #userArgs?: any; // eslint-disable-line @typescript-eslint/no-explicit-any\n #jobStatus?: JobStatus;\n #runningJob?: RunningJobInfo;\n #agent: string;\n #inferenceExecutor?: InferenceExecutor;\n #inferenceTasks: Promise<void>[] = [];\n #logger = log();\n\n constructor(\n agent: string,\n inferenceExecutor: InferenceExecutor | undefined,\n initializeTimeout: number,\n closeTimeout: number,\n memoryWarnMB: number,\n memoryLimitMB: number,\n pingInterval: number,\n pingTimeout: number,\n highPingThreshold: number,\n ) {\n super(\n initializeTimeout,\n closeTimeout,\n memoryWarnMB,\n memoryLimitMB,\n pingInterval,\n pingTimeout,\n highPingThreshold,\n );\n this.#agent = agent;\n this.#inferenceExecutor = inferenceExecutor;\n }\n\n get status(): JobStatus {\n if (this.#jobStatus) {\n return this.#jobStatus;\n }\n throw new Error('job status not available');\n }\n\n get userArguments(): any {\n return this.#userArgs;\n }\n\n set userArguments(args: any) {\n this.#userArgs = args;\n }\n\n get runningJob(): RunningJobInfo | undefined {\n return this.#runningJob;\n }\n\n createProcess(): ChildProcess {\n const forkUrl = new URL(`./job_proc_lazy_main${currentFileExtension}`, import.meta.url);\n\n // When running via tsx/ts-node (TypeScript files), we need to inherit the parent's\n // execArgv so the child process can also execute TypeScript with the same loader\n const isTypeScript = currentFileExtension === '.ts';\n const forkOptions = isTypeScript ? { execArgv: process.execArgv } : undefined;\n\n return fork(forkUrl, [this.#agent], forkOptions);\n }\n\n async mainTask(proc: ChildProcess) {\n proc.on('message', (msg: IPCMessage) => {\n switch (msg.case) {\n case 'inferenceRequest':\n this.#inferenceTasks.push(this.#doInferenceTask(proc, msg.value));\n }\n });\n }\n\n async #doInferenceTask(\n proc: ChildProcess,\n req: { method: string; requestId: string; data: unknown },\n ) {\n if (!this.#inferenceExecutor) {\n this.#logger.warn('inference request received but no inference executor');\n proc.send({\n case: 'inferenceResponse',\n value: { requestId: req.requestId, error: new Error('no inference executor') },\n });\n return;\n }\n\n try {\n const data = await this.#inferenceExecutor.doInference(req.method, req.data);\n proc.send({ case: 'inferenceResponse', value: { requestId: req.requestId, data } });\n } catch (error) {\n proc.send({ case: 'inferenceResponse', value: { requestId: req.requestId, error } });\n }\n }\n\n async launchJob(info: RunningJobInfo) {\n if (this.#runningJob) {\n throw Error('process already has a running job');\n }\n if (!this.init.done) {\n throw Error('process not initialized');\n }\n this.#jobStatus = JobStatus.RUNNING;\n this.#runningJob = info;\n\n this.proc!.send({ case: 'startJobRequest', value: { runningJob: info } });\n }\n}\n","// Shim globals in cjs bundle\n// There's a weird bug that esbuild will always inject importMetaUrl\n// if we export it as `const importMetaUrl = ... __filename ...`\n// But using a function will not cause this issue\n\nconst getImportMetaUrl = () =>\n typeof document === 'undefined'\n ? new URL(`file:${__filename}`).href\n : (document.currentScript && document.currentScript.src) ||\n new URL('main.js', document.baseURI).href\n\nexport const importMetaUrl = /* @__PURE__ */ getImportMetaUrl()\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;ACKA,IAAM,mBAAmB,MACvB,OAAO,aAAa,cAChB,IAAI,IAAI,QAAQ,UAAU,EAAE,EAAE,OAC7B,SAAS,iBAAiB,SAAS,cAAc,OAClD,IAAI,IAAI,WAAW,SAAS,OAAO,EAAE;AAEpC,IAAM,gBAAgC,iCAAiB;ADP9D,gCAAqB;AACrB,uBAAwB;AAExB,iBAAoB;AAGpB,0BAA0B;AAE1B,6BAA+B;AAE/B,MAAM,2BAAuB,0BAAQ,aAAe;AAE7C,MAAM,wBAAwB,sCAAsC;AAAA,EACzE;AAAA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,kBAAmC,CAAC;AAAA,EACpC,cAAU,gBAAI;AAAA,EAEd,YACE,OACA,mBACA,mBACA,cACA,cACA,eACA,cACA,aACA,mBACA;AACA;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,SAAK,SAAS;AACd,SAAK,qBAAqB;AAAA,EAC5B;AAAA,EAEA,IAAI,SAAoB;AACtB,QAAI,KAAK,YAAY;AACnB,aAAO,KAAK;AAAA,IACd;AACA,UAAM,IAAI,MAAM,0BAA0B;AAAA,EAC5C;AAAA,EAEA,IAAI,gBAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAAc,MAAW;AAC3B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,IAAI,aAAyC;AAC3C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,gBAA8B;AAC5B,UAAM,UAAU,IAAI,IAAI,uBAAuB,oBAAoB,IAAI,aAAe;AAItF,UAAM,eAAe,yBAAyB;AAC9C,UAAM,cAAc,eAAe,EAAE,UAAU,QAAQ,SAAS,IAAI;AAEpE,eAAO,gCAAK,SAAS,CAAC,KAAK,MAAM,GAAG,WAAW;AAAA,EACjD;AAAA,EAEA,MAAM,SAAS,MAAoB;AACjC,SAAK,GAAG,WAAW,CAAC,QAAoB;AACtC,cAAQ,IAAI,MAAM;AAAA,QAChB,KAAK;AACH,eAAK,gBAAgB,KAAK,KAAK,iBAAiB,MAAM,IAAI,KAAK,CAAC;AAAA,MACpE;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,iBACJ,MACA,KACA;AACA,QAAI,CAAC,KAAK,oBAAoB;AAC5B,WAAK,QAAQ,KAAK,sDAAsD;AACxE,WAAK,KAAK;AAAA,QACR,MAAM;AAAA,QACN,OAAO,EAAE,WAAW,IAAI,WAAW,OAAO,IAAI,MAAM,uBAAuB,EAAE;AAAA,MAC/E,CAAC;AACD;AAAA,IACF;AAEA,QAAI;AACF,YAAM,OAAO,MAAM,KAAK,mBAAmB,YAAY,IAAI,QAAQ,IAAI,IAAI;AAC3E,WAAK,KAAK,EAAE,MAAM,qBAAqB,OAAO,EAAE,WAAW,IAAI,WAAW,KAAK,EAAE,CAAC;AAAA,IACpF,SAAS,OAAO;AACd,WAAK,KAAK,EAAE,MAAM,qBAAqB,OAAO,EAAE,WAAW,IAAI,WAAW,MAAM,EAAE,CAAC;AAAA,IACrF;AAAA,EACF;AAAA,EAEA,MAAM,UAAU,MAAsB;AACpC,QAAI,KAAK,aAAa;AACpB,YAAM,MAAM,mCAAmC;AAAA,IACjD;AACA,QAAI,CAAC,KAAK,KAAK,MAAM;AACnB,YAAM,MAAM,yBAAyB;AAAA,IACvC;AACA,SAAK,aAAa,8BAAU;AAC5B,SAAK,cAAc;AAEnB,SAAK,KAAM,KAAK,EAAE,MAAM,mBAAmB,OAAO,EAAE,YAAY,KAAK,EAAE,CAAC;AAAA,EAC1E;AACF;","names":[]}
package/dist/stt/stt.cjs CHANGED
@@ -82,6 +82,7 @@ class SpeechStream {
82
82
  deferredInputStream;
83
83
  logger = (0, import_log.log)();
84
84
  _connOptions;
85
+ _startTimeOffset = 0;
85
86
  abortController = new AbortController();
86
87
  constructor(stt, sampleRate, connectionOptions = import_types.DEFAULT_API_CONNECT_OPTIONS) {
87
88
  this.#stt = stt;
@@ -181,6 +182,15 @@ class SpeechStream {
181
182
  get abortSignal() {
182
183
  return this.abortController.signal;
183
184
  }
185
+ get startTimeOffset() {
186
+ return this._startTimeOffset;
187
+ }
188
+ set startTimeOffset(value) {
189
+ if (value < 0) {
190
+ throw new Error("startTimeOffset must be non-negative");
191
+ }
192
+ this._startTimeOffset = value;
193
+ }
184
194
  updateInputStream(audioStream) {
185
195
  this.deferredInputStream.setSource(audioStream);
186
196
  }
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/stt/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame, AudioResampler } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { calculateAudioDurationSeconds } from '../audio.js';\nimport { log } from '../log.js';\nimport type { STTMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS, intervalForRetry } from '../types.js';\nimport type { AudioBuffer } from '../utils.js';\nimport { AsyncIterableQueue, delay, startSoon, toError } from '../utils.js';\n\n/** Indicates start/middle/end of speech */\nexport enum SpeechEventType {\n /**\n * Indicate the start of speech.\n * If the STT doesn't support this event, this will be emitted at the same time\n * as the first INTERIM_TRANSCRIPT.\n */\n START_OF_SPEECH = 0,\n /**\n * Interim transcript, useful for real-time transcription.\n */\n INTERIM_TRANSCRIPT = 1,\n /**\n * Final transcript, emitted when the STT is confident enough that a certain\n * portion of the speech will not change.\n */\n FINAL_TRANSCRIPT = 2,\n /**\n * Indicate the end of speech, emitted when the user stops speaking.\n * The first alternative is a combination of all the previous FINAL_TRANSCRIPT events.\n */\n END_OF_SPEECH = 3,\n /** Usage event, emitted periodically to indicate usage metrics. */\n RECOGNITION_USAGE = 4,\n /**\n * Preflight transcript, emitted before final transcript when STT has high confidence\n * but hasn't fully committed yet. Includes all pre-committed transcripts including\n * final transcript from the previous STT run.\n */\n PREFLIGHT_TRANSCRIPT = 5,\n}\n\n/** SpeechData contains metadata about this {@link SpeechEvent}. */\nexport interface SpeechData {\n language: string;\n text: string;\n startTime: number;\n endTime: number;\n confidence: number;\n}\n\nexport interface RecognitionUsage {\n audioDuration: number;\n}\n\n/** SpeechEvent is a packet of speech-to-text data. */\nexport interface SpeechEvent {\n type: SpeechEventType;\n alternatives?: [SpeechData, ...SpeechData[]];\n requestId?: string;\n recognitionUsage?: RecognitionUsage;\n}\n\n/**\n * Describes the capabilities of the STT provider.\n *\n * @remarks\n * At present, the framework only supports providers that have a streaming endpoint.\n */\nexport interface STTCapabilities {\n streaming: boolean;\n interimResults: boolean;\n}\n\nexport interface STTError {\n type: 'stt_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type STTCallbacks = {\n ['metrics_collected']: (metrics: STTMetrics) => void;\n ['error']: (error: STTError) => void;\n};\n\n/**\n * An instance of a speech-to-text adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child STT class, which inherits this class's methods.\n */\nexport abstract class STT extends (EventEmitter as new () => TypedEmitter<STTCallbacks>) {\n abstract label: string;\n #capabilities: STTCapabilities;\n\n constructor(capabilities: STTCapabilities) {\n super();\n this.#capabilities = capabilities;\n }\n\n /** Returns this STT's capabilities */\n get capabilities(): STTCapabilities {\n return this.#capabilities;\n }\n\n /** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */\n async recognize(frame: AudioBuffer, abortSignal?: AbortSignal): Promise<SpeechEvent> {\n const startTime = process.hrtime.bigint();\n const event = await this._recognize(frame, abortSignal);\n const durationMs = Number((process.hrtime.bigint() - startTime) / BigInt(1000000));\n this.emit('metrics_collected', {\n type: 'stt_metrics',\n requestId: event.requestId ?? '',\n timestamp: Date.now(),\n durationMs,\n label: this.label,\n audioDurationMs: Math.round(calculateAudioDurationSeconds(frame) * 1000),\n streamed: false,\n });\n return event;\n }\n\n protected abstract _recognize(\n frame: AudioBuffer,\n abortSignal?: AbortSignal,\n ): Promise<SpeechEvent>;\n\n /**\n * Returns a {@link SpeechStream} that can be used to push audio frames and receive\n * transcriptions\n *\n * @param options - Optional configuration including connection options\n */\n abstract stream(options?: { connOptions?: APIConnectOptions }): SpeechStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a speech-to-text stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives[0].text)\n * }\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SpeechStream class, which inherits this class's methods.\n */\nexport abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent> {\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n protected input = new AsyncIterableQueue<AudioFrame | typeof SpeechStream.FLUSH_SENTINEL>();\n protected output = new AsyncIterableQueue<SpeechEvent>();\n protected queue = new AsyncIterableQueue<SpeechEvent>();\n protected neededSampleRate?: number;\n protected resampler?: AudioResampler;\n abstract label: string;\n protected closed = false;\n #stt: STT;\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n private logger = log();\n private _connOptions: APIConnectOptions;\n\n protected abortController = new AbortController();\n\n constructor(\n stt: STT,\n sampleRate?: number,\n connectionOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n ) {\n this.#stt = stt;\n this._connOptions = connectionOptions;\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n this.neededSampleRate = sampleRate;\n this.monitorMetrics();\n this.pumpInput();\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n startSoon(() => this.mainTask().finally(() => this.queue.close()));\n }\n\n private async mainTask() {\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await this.run();\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = intervalForRetry(this._connOptions, i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to recognize speech after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#stt.label, attempt: i + 1, error },\n `failed to recognize speech, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#stt.emit('error', {\n type: 'stt_error',\n timestamp: Date.now(),\n label: this.#stt.label,\n error,\n recoverable,\n });\n }\n\n protected async pumpInput() {\n // TODO(AJS-35): Implement STT with webstreams API\n const inputStream = this.deferredInputStream.stream;\n const reader = inputStream.getReader();\n\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done) break;\n this.pushFrame(value);\n }\n } catch (error) {\n this.logger.error('Error in STTStream mainTask:', error);\n } finally {\n reader.releaseLock();\n }\n }\n\n protected async monitorMetrics() {\n for await (const event of this.queue) {\n if (!this.output.closed) {\n try {\n this.output.put(event);\n } catch (e) {\n if (e instanceof Error && e.message.includes('Queue is closed')) {\n this.logger.warn(\n { err: e },\n 'Queue closed during transcript processing (expected during disconnect)',\n );\n }\n }\n }\n if (event.type !== SpeechEventType.RECOGNITION_USAGE) continue;\n const metrics: STTMetrics = {\n type: 'stt_metrics',\n timestamp: Date.now(),\n requestId: event.requestId!,\n durationMs: 0,\n label: this.#stt.label,\n audioDurationMs: Math.round(event.recognitionUsage!.audioDuration * 1000),\n streamed: true,\n };\n this.#stt.emit('metrics_collected', metrics);\n }\n if (!this.output.closed) {\n this.output.close();\n }\n }\n\n protected abstract run(): Promise<void>;\n\n protected get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n updateInputStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputStream() {\n this.deferredInputStream.detachSource();\n }\n\n /** Push an audio frame to the STT */\n pushFrame(frame: AudioFrame) {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n if (this.neededSampleRate && frame.sampleRate !== this.neededSampleRate) {\n if (!this.resampler) {\n this.resampler = new AudioResampler(frame.sampleRate, this.neededSampleRate);\n }\n }\n\n if (this.resampler) {\n const frames = this.resampler.push(frame);\n for (const frame of frames) {\n this.input.put(frame);\n }\n } else {\n this.input.put(frame);\n }\n }\n\n /** Flush the STT, causing it to process all pending text */\n flush() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.put(SpeechStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SpeechEvent>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the STT stream */\n close() {\n if (!this.input.closed) this.input.close();\n if (!this.queue.closed) this.queue.close();\n if (!this.output.closed) this.output.close();\n if (!this.abortController.signal.aborted) this.abortController.abort();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): SpeechStream {\n return this;\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAgD;AAEhD,yBAA6B;AAE7B,wBAA6C;AAC7C,mBAA8C;AAC9C,iBAAoB;AAEpB,6BAAuC;AACvC,mBAAsF;AAEtF,mBAA8D;AAGvD,IAAK,kBAAL,kBAAKA,qBAAL;AAML,EAAAA,kCAAA,qBAAkB,KAAlB;AAIA,EAAAA,kCAAA,wBAAqB,KAArB;AAKA,EAAAA,kCAAA,sBAAmB,KAAnB;AAKA,EAAAA,kCAAA,mBAAgB,KAAhB;AAEA,EAAAA,kCAAA,uBAAoB,KAApB;AAMA,EAAAA,kCAAA,0BAAuB,KAAvB;AA5BU,SAAAA;AAAA,GAAA;AAmFL,MAAe,YAAa,gCAAsD;AAAA,EAEvF;AAAA,EAEA,YAAY,cAA+B;AACzC,UAAM;AACN,SAAK,gBAAgB;AAAA,EACvB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,MAAM,UAAU,OAAoB,aAAiD;AACnF,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,UAAM,QAAQ,MAAM,KAAK,WAAW,OAAO,WAAW;AACtD,UAAM,aAAa,QAAQ,QAAQ,OAAO,OAAO,IAAI,aAAa,OAAO,GAAO,CAAC;AACjF,SAAK,KAAK,qBAAqB;AAAA,MAC7B,MAAM;AAAA,MACN,WAAW,MAAM,aAAa;AAAA,MAC9B,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,OAAO,KAAK;AAAA,MACZ,iBAAiB,KAAK,UAAM,4CAA8B,KAAK,IAAI,GAAI;AAAA,MACvE,UAAU;AAAA,IACZ,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EAeA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAkBO,MAAe,aAA2D;AAAA,EAC/E,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,gCAAoE;AAAA,EAChF,SAAS,IAAI,gCAAgC;AAAA,EAC7C,QAAQ,IAAI,gCAAgC;AAAA,EAC5C;AAAA,EACA;AAAA,EAEA,SAAS;AAAA,EACnB;AAAA,EACQ;AAAA,EACA,aAAS,gBAAI;AAAA,EACb;AAAA,EAEE,kBAAkB,IAAI,gBAAgB;AAAA,EAEhD,YACE,KACA,YACA,oBAAuC,0CACvC;AACA,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,8CAAmC;AAClE,SAAK,mBAAmB;AACxB,SAAK,eAAe;AACpB,SAAK,UAAU;AAMf,gCAAU,MAAM,KAAK,SAAS,EAAE,QAAQ,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EACnE;AAAA,EAEA,MAAc,WAAW;AACvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,4BAAU;AAC7B,gBAAM,oBAAgB,+BAAiB,KAAK,cAAc,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,qCAAmB;AAAA,cAC3B,SAAS,oCAAoC,KAAK,aAAa,WAAW,CAAC;AAAA,cAC3E,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,2CAA2C,aAAa;AAAA,YAC1D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,sBAAM,oBAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,WAAO,sBAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAgB,YAAY;AAE1B,UAAM,cAAc,KAAK,oBAAoB;AAC7C,UAAM,SAAS,YAAY,UAAU;AAErC,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AACV,aAAK,UAAU,KAAK;AAAA,MACtB;AAAA,IACF,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,gCAAgC,KAAK;AAAA,IACzD,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,YAAI;AACF,eAAK,OAAO,IAAI,KAAK;AAAA,QACvB,SAAS,GAAG;AACV,cAAI,aAAa,SAAS,EAAE,QAAQ,SAAS,iBAAiB,GAAG;AAC/D,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,EAAE;AAAA,cACT;AAAA,YACF;AAAA,UACF;AAAA,QACF;AAAA,MACF;AACA,UAAI,MAAM,SAAS,0BAAmC;AACtD,YAAM,UAAsB;AAAA,QAC1B,MAAM;AAAA,QACN,WAAW,KAAK,IAAI;AAAA,QACpB,WAAW,MAAM;AAAA,QACjB,YAAY;AAAA,QACZ,OAAO,KAAK,KAAK;AAAA,QACjB,iBAAiB,KAAK,MAAM,MAAM,iBAAkB,gBAAgB,GAAI;AAAA,QACxE,UAAU;AAAA,MACZ;AACA,WAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,IAC7C;AACA,QAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,WAAK,OAAO,MAAM;AAAA,IACpB;AAAA,EACF;AAAA,EAIA,IAAc,cAA2B;AACvC,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA,EAEA,kBAAkB,aAAyC;AACzD,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,oBAAoB;AAClB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA;AAAA,EAGA,UAAU,OAAmB;AAC3B,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,QAAI,KAAK,oBAAoB,MAAM,eAAe,KAAK,kBAAkB;AACvE,UAAI,CAAC,KAAK,WAAW;AACnB,aAAK,YAAY,IAAI,+BAAe,MAAM,YAAY,KAAK,gBAAgB;AAAA,MAC7E;AAAA,IACF;AAEA,QAAI,KAAK,WAAW;AAClB,YAAM,SAAS,KAAK,UAAU,KAAK,KAAK;AACxC,iBAAWC,UAAS,QAAQ;AAC1B,aAAK,MAAM,IAAIA,MAAK;AAAA,MACtB;AAAA,IACF,OAAO;AACL,WAAK,MAAM,IAAI,KAAK;AAAA,IACtB;AAAA,EACF;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,aAAa,cAAc;AAAA,EAC5C;AAAA;AAAA,EAGA,WAAW;AACT,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA6C;AAC3C,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,QAAI,CAAC,KAAK,gBAAgB,OAAO,QAAS,MAAK,gBAAgB,MAAM;AACrE,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAkB;AACrC,WAAO;AAAA,EACT;AACF;","names":["SpeechEventType","frame"]}
1
+ {"version":3,"sources":["../../src/stt/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame, AudioResampler } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { calculateAudioDurationSeconds } from '../audio.js';\nimport { log } from '../log.js';\nimport type { STTMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS, intervalForRetry } from '../types.js';\nimport type { AudioBuffer } from '../utils.js';\nimport { AsyncIterableQueue, delay, startSoon, toError } from '../utils.js';\nimport type { TimedString } from '../voice/index.js';\n\n/** Indicates start/middle/end of speech */\nexport enum SpeechEventType {\n /**\n * Indicate the start of speech.\n * If the STT doesn't support this event, this will be emitted at the same time\n * as the first INTERIM_TRANSCRIPT.\n */\n START_OF_SPEECH = 0,\n /**\n * Interim transcript, useful for real-time transcription.\n */\n INTERIM_TRANSCRIPT = 1,\n /**\n * Final transcript, emitted when the STT is confident enough that a certain\n * portion of the speech will not change.\n */\n FINAL_TRANSCRIPT = 2,\n /**\n * Indicate the end of speech, emitted when the user stops speaking.\n * The first alternative is a combination of all the previous FINAL_TRANSCRIPT events.\n */\n END_OF_SPEECH = 3,\n /** Usage event, emitted periodically to indicate usage metrics. */\n RECOGNITION_USAGE = 4,\n /**\n * Preflight transcript, emitted before final transcript when STT has high confidence\n * but hasn't fully committed yet. Includes all pre-committed transcripts including\n * final transcript from the previous STT run.\n */\n PREFLIGHT_TRANSCRIPT = 5,\n}\n\n/** SpeechData contains metadata about this {@link SpeechEvent}. */\nexport interface SpeechData {\n language: string;\n text: string;\n startTime: number;\n endTime: number;\n confidence: number;\n words?: TimedString[];\n}\n\nexport interface RecognitionUsage {\n audioDuration: number;\n}\n\n/** SpeechEvent is a packet of speech-to-text data. */\nexport interface SpeechEvent {\n type: SpeechEventType;\n alternatives?: [SpeechData, ...SpeechData[]];\n requestId?: string;\n recognitionUsage?: RecognitionUsage;\n}\n\n/**\n * Describes the capabilities of the STT provider.\n *\n * @remarks\n * At present, the framework only supports providers that have a streaming endpoint.\n */\nexport interface STTCapabilities {\n streaming: boolean;\n interimResults: boolean;\n /**\n * Whether this STT supports aligned transcripts with word/chunk timestamps.\n * - 'word': Provider returns word-level timestamps\n * - 'chunk': Provider returns chunk-level timestamps (e.g., sentence/phrase boundaries)\n * - false: Provider does not support aligned transcripts\n */\n alignedTranscript?: 'word' | 'chunk' | false;\n}\n\nexport interface STTError {\n type: 'stt_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type STTCallbacks = {\n ['metrics_collected']: (metrics: STTMetrics) => void;\n ['error']: (error: STTError) => void;\n};\n\n/**\n * An instance of a speech-to-text adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child STT class, which inherits this class's methods.\n */\nexport abstract class STT extends (EventEmitter as new () => TypedEmitter<STTCallbacks>) {\n abstract label: string;\n #capabilities: STTCapabilities;\n\n constructor(capabilities: STTCapabilities) {\n super();\n this.#capabilities = capabilities;\n }\n\n /** Returns this STT's capabilities */\n get capabilities(): STTCapabilities {\n return this.#capabilities;\n }\n\n /** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */\n async recognize(frame: AudioBuffer, abortSignal?: AbortSignal): Promise<SpeechEvent> {\n const startTime = process.hrtime.bigint();\n const event = await this._recognize(frame, abortSignal);\n const durationMs = Number((process.hrtime.bigint() - startTime) / BigInt(1000000));\n this.emit('metrics_collected', {\n type: 'stt_metrics',\n requestId: event.requestId ?? '',\n timestamp: Date.now(),\n durationMs,\n label: this.label,\n audioDurationMs: Math.round(calculateAudioDurationSeconds(frame) * 1000),\n streamed: false,\n });\n return event;\n }\n\n protected abstract _recognize(\n frame: AudioBuffer,\n abortSignal?: AbortSignal,\n ): Promise<SpeechEvent>;\n\n /**\n * Returns a {@link SpeechStream} that can be used to push audio frames and receive\n * transcriptions\n *\n * @param options - Optional configuration including connection options\n */\n abstract stream(options?: { connOptions?: APIConnectOptions }): SpeechStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a speech-to-text stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives[0].text)\n * }\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SpeechStream class, which inherits this class's methods.\n */\nexport abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent> {\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n protected input = new AsyncIterableQueue<AudioFrame | typeof SpeechStream.FLUSH_SENTINEL>();\n protected output = new AsyncIterableQueue<SpeechEvent>();\n protected queue = new AsyncIterableQueue<SpeechEvent>();\n protected neededSampleRate?: number;\n protected resampler?: AudioResampler;\n abstract label: string;\n protected closed = false;\n #stt: STT;\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n private logger = log();\n private _connOptions: APIConnectOptions;\n private _startTimeOffset: number = 0;\n\n protected abortController = new AbortController();\n\n constructor(\n stt: STT,\n sampleRate?: number,\n connectionOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n ) {\n this.#stt = stt;\n this._connOptions = connectionOptions;\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n this.neededSampleRate = sampleRate;\n this.monitorMetrics();\n this.pumpInput();\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n startSoon(() => this.mainTask().finally(() => this.queue.close()));\n }\n\n private async mainTask() {\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await this.run();\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = intervalForRetry(this._connOptions, i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to recognize speech after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#stt.label, attempt: i + 1, error },\n `failed to recognize speech, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#stt.emit('error', {\n type: 'stt_error',\n timestamp: Date.now(),\n label: this.#stt.label,\n error,\n recoverable,\n });\n }\n\n protected async pumpInput() {\n // TODO(AJS-35): Implement STT with webstreams API\n const inputStream = this.deferredInputStream.stream;\n const reader = inputStream.getReader();\n\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done) break;\n this.pushFrame(value);\n }\n } catch (error) {\n this.logger.error('Error in STTStream mainTask:', error);\n } finally {\n reader.releaseLock();\n }\n }\n\n protected async monitorMetrics() {\n for await (const event of this.queue) {\n if (!this.output.closed) {\n try {\n this.output.put(event);\n } catch (e) {\n if (e instanceof Error && e.message.includes('Queue is closed')) {\n this.logger.warn(\n { err: e },\n 'Queue closed during transcript processing (expected during disconnect)',\n );\n }\n }\n }\n if (event.type !== SpeechEventType.RECOGNITION_USAGE) continue;\n const metrics: STTMetrics = {\n type: 'stt_metrics',\n timestamp: Date.now(),\n requestId: event.requestId!,\n durationMs: 0,\n label: this.#stt.label,\n audioDurationMs: Math.round(event.recognitionUsage!.audioDuration * 1000),\n streamed: true,\n };\n this.#stt.emit('metrics_collected', metrics);\n }\n if (!this.output.closed) {\n this.output.close();\n }\n }\n\n protected abstract run(): Promise<void>;\n\n protected get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n get startTimeOffset(): number {\n return this._startTimeOffset;\n }\n\n set startTimeOffset(value: number) {\n if (value < 0) {\n throw new Error('startTimeOffset must be non-negative');\n }\n this._startTimeOffset = value;\n }\n\n updateInputStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputStream() {\n this.deferredInputStream.detachSource();\n }\n\n /** Push an audio frame to the STT */\n pushFrame(frame: AudioFrame) {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n\n if (this.neededSampleRate && frame.sampleRate !== this.neededSampleRate) {\n if (!this.resampler) {\n this.resampler = new AudioResampler(frame.sampleRate, this.neededSampleRate);\n }\n }\n\n if (this.resampler) {\n const frames = this.resampler.push(frame);\n for (const frame of frames) {\n this.input.put(frame);\n }\n } else {\n this.input.put(frame);\n }\n }\n\n /** Flush the STT, causing it to process all pending text */\n flush() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.put(SpeechStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SpeechEvent>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the STT stream */\n close() {\n if (!this.input.closed) this.input.close();\n if (!this.queue.closed) this.queue.close();\n if (!this.output.closed) this.output.close();\n if (!this.abortController.signal.aborted) this.abortController.abort();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): SpeechStream {\n return this;\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAgD;AAEhD,yBAA6B;AAE7B,wBAA6C;AAC7C,mBAA8C;AAC9C,iBAAoB;AAEpB,6BAAuC;AACvC,mBAAsF;AAEtF,mBAA8D;AAIvD,IAAK,kBAAL,kBAAKA,qBAAL;AAML,EAAAA,kCAAA,qBAAkB,KAAlB;AAIA,EAAAA,kCAAA,wBAAqB,KAArB;AAKA,EAAAA,kCAAA,sBAAmB,KAAnB;AAKA,EAAAA,kCAAA,mBAAgB,KAAhB;AAEA,EAAAA,kCAAA,uBAAoB,KAApB;AAMA,EAAAA,kCAAA,0BAAuB,KAAvB;AA5BU,SAAAA;AAAA,GAAA;AA2FL,MAAe,YAAa,gCAAsD;AAAA,EAEvF;AAAA,EAEA,YAAY,cAA+B;AACzC,UAAM;AACN,SAAK,gBAAgB;AAAA,EACvB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,MAAM,UAAU,OAAoB,aAAiD;AACnF,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,UAAM,QAAQ,MAAM,KAAK,WAAW,OAAO,WAAW;AACtD,UAAM,aAAa,QAAQ,QAAQ,OAAO,OAAO,IAAI,aAAa,OAAO,GAAO,CAAC;AACjF,SAAK,KAAK,qBAAqB;AAAA,MAC7B,MAAM;AAAA,MACN,WAAW,MAAM,aAAa;AAAA,MAC9B,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,OAAO,KAAK;AAAA,MACZ,iBAAiB,KAAK,UAAM,4CAA8B,KAAK,IAAI,GAAI;AAAA,MACvE,UAAU;AAAA,IACZ,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EAeA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAkBO,MAAe,aAA2D;AAAA,EAC/E,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,gCAAoE;AAAA,EAChF,SAAS,IAAI,gCAAgC;AAAA,EAC7C,QAAQ,IAAI,gCAAgC;AAAA,EAC5C;AAAA,EACA;AAAA,EAEA,SAAS;AAAA,EACnB;AAAA,EACQ;AAAA,EACA,aAAS,gBAAI;AAAA,EACb;AAAA,EACA,mBAA2B;AAAA,EAEzB,kBAAkB,IAAI,gBAAgB;AAAA,EAEhD,YACE,KACA,YACA,oBAAuC,0CACvC;AACA,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,8CAAmC;AAClE,SAAK,mBAAmB;AACxB,SAAK,eAAe;AACpB,SAAK,UAAU;AAMf,gCAAU,MAAM,KAAK,SAAS,EAAE,QAAQ,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EACnE;AAAA,EAEA,MAAc,WAAW;AACvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,4BAAU;AAC7B,gBAAM,oBAAgB,+BAAiB,KAAK,cAAc,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,qCAAmB;AAAA,cAC3B,SAAS,oCAAoC,KAAK,aAAa,WAAW,CAAC;AAAA,cAC3E,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,2CAA2C,aAAa;AAAA,YAC1D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,sBAAM,oBAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,WAAO,sBAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAgB,YAAY;AAE1B,UAAM,cAAc,KAAK,oBAAoB;AAC7C,UAAM,SAAS,YAAY,UAAU;AAErC,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AACV,aAAK,UAAU,KAAK;AAAA,MACtB;AAAA,IACF,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,gCAAgC,KAAK;AAAA,IACzD,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,YAAI;AACF,eAAK,OAAO,IAAI,KAAK;AAAA,QACvB,SAAS,GAAG;AACV,cAAI,aAAa,SAAS,EAAE,QAAQ,SAAS,iBAAiB,GAAG;AAC/D,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,EAAE;AAAA,cACT;AAAA,YACF;AAAA,UACF;AAAA,QACF;AAAA,MACF;AACA,UAAI,MAAM,SAAS,0BAAmC;AACtD,YAAM,UAAsB;AAAA,QAC1B,MAAM;AAAA,QACN,WAAW,KAAK,IAAI;AAAA,QACpB,WAAW,MAAM;AAAA,QACjB,YAAY;AAAA,QACZ,OAAO,KAAK,KAAK;AAAA,QACjB,iBAAiB,KAAK,MAAM,MAAM,iBAAkB,gBAAgB,GAAI;AAAA,QACxE,UAAU;AAAA,MACZ;AACA,WAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,IAC7C;AACA,QAAI,CAAC,KAAK,OAAO,QAAQ;AACvB,WAAK,OAAO,MAAM;AAAA,IACpB;AAAA,EACF;AAAA,EAIA,IAAc,cAA2B;AACvC,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA,EAEA,IAAI,kBAA0B;AAC5B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,gBAAgB,OAAe;AACjC,QAAI,QAAQ,GAAG;AACb,YAAM,IAAI,MAAM,sCAAsC;AAAA,IACxD;AACA,SAAK,mBAAmB;AAAA,EAC1B;AAAA,EAEA,kBAAkB,aAAyC;AACzD,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,oBAAoB;AAClB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA;AAAA,EAGA,UAAU,OAAmB;AAC3B,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AAEA,QAAI,KAAK,oBAAoB,MAAM,eAAe,KAAK,kBAAkB;AACvE,UAAI,CAAC,KAAK,WAAW;AACnB,aAAK,YAAY,IAAI,+BAAe,MAAM,YAAY,KAAK,gBAAgB;AAAA,MAC7E;AAAA,IACF;AAEA,QAAI,KAAK,WAAW;AAClB,YAAM,SAAS,KAAK,UAAU,KAAK,KAAK;AACxC,iBAAWC,UAAS,QAAQ;AAC1B,aAAK,MAAM,IAAIA,MAAK;AAAA,MACtB;AAAA,IACF,OAAO;AACL,WAAK,MAAM,IAAI,KAAK;AAAA,IACtB;AAAA,EACF;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,aAAa,cAAc;AAAA,EAC5C;AAAA;AAAA,EAGA,WAAW;AACT,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA6C;AAC3C,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,QAAI,CAAC,KAAK,gBAAgB,OAAO,QAAS,MAAK,gBAAgB,MAAM;AACrE,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAkB;AACrC,WAAO;AAAA,EACT;AACF;","names":["SpeechEventType","frame"]}
@@ -6,6 +6,7 @@ import type { STTMetrics } from '../metrics/base.js';
6
6
  import { type APIConnectOptions } from '../types.js';
7
7
  import type { AudioBuffer } from '../utils.js';
8
8
  import { AsyncIterableQueue } from '../utils.js';
9
+ import type { TimedString } from '../voice/index.js';
9
10
  /** Indicates start/middle/end of speech */
10
11
  export declare enum SpeechEventType {
11
12
  /**
@@ -44,6 +45,7 @@ export interface SpeechData {
44
45
  startTime: number;
45
46
  endTime: number;
46
47
  confidence: number;
48
+ words?: TimedString[];
47
49
  }
48
50
  export interface RecognitionUsage {
49
51
  audioDuration: number;
@@ -64,6 +66,13 @@ export interface SpeechEvent {
64
66
  export interface STTCapabilities {
65
67
  streaming: boolean;
66
68
  interimResults: boolean;
69
+ /**
70
+ * Whether this STT supports aligned transcripts with word/chunk timestamps.
71
+ * - 'word': Provider returns word-level timestamps
72
+ * - 'chunk': Provider returns chunk-level timestamps (e.g., sentence/phrase boundaries)
73
+ * - false: Provider does not support aligned transcripts
74
+ */
75
+ alignedTranscript?: 'word' | 'chunk' | false;
67
76
  }
68
77
  export interface STTError {
69
78
  type: 'stt_error';
@@ -133,6 +142,7 @@ export declare abstract class SpeechStream implements AsyncIterableIterator<Spee
133
142
  private deferredInputStream;
134
143
  private logger;
135
144
  private _connOptions;
145
+ private _startTimeOffset;
136
146
  protected abortController: AbortController;
137
147
  constructor(stt: STT, sampleRate?: number, connectionOptions?: APIConnectOptions);
138
148
  private mainTask;
@@ -141,6 +151,8 @@ export declare abstract class SpeechStream implements AsyncIterableIterator<Spee
141
151
  protected monitorMetrics(): Promise<void>;
142
152
  protected abstract run(): Promise<void>;
143
153
  protected get abortSignal(): AbortSignal;
154
+ get startTimeOffset(): number;
155
+ set startTimeOffset(value: number);
144
156
  updateInputStream(audioStream: ReadableStream<AudioFrame>): void;
145
157
  detachInputStream(): void;
146
158
  /** Push an audio frame to the STT */