@livekit/agents 1.0.20 → 1.0.22

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.
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/inference/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport { basic as tokenizeBasic } from '../tokenize/index.js';\nimport type { ChunkedStream } from '../tts/index.js';\nimport { SynthesizeStream as BaseSynthesizeStream, TTS as BaseTTS } from '../tts/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { shortuuid } from '../utils.js';\nimport {\n type TtsClientEvent,\n type TtsServerEvent,\n type TtsSessionCreateEvent,\n ttsClientEventSchema,\n ttsServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken } from './utils.js';\n\nexport type CartesiaModels =\n | 'cartesia'\n | 'cartesia/sonic'\n | 'cartesia/sonic-2'\n | 'cartesia/sonic-turbo';\n\nexport type ElevenlabsModels =\n | 'elevenlabs'\n | 'elevenlabs/eleven_flash_v2'\n | 'elevenlabs/eleven_flash_v2_5'\n | 'elevenlabs/eleven_turbo_v2'\n | 'elevenlabs/eleven_turbo_v2_5'\n | 'elevenlabs/eleven_multilingual_v2';\n\nexport type RimeModels = 'rime' | 'rime/mist' | 'rime/mistv2' | 'rime/arcana';\n\nexport type InworldModels = 'inworld' | 'inworld/inworld-tts-1';\n\nexport interface CartesiaOptions {\n duration?: number; // max duration of audio in seconds\n speed?: 'slow' | 'normal' | 'fast'; // default: not specified\n}\n\nexport interface ElevenlabsOptions {\n inactivity_timeout?: number; // default: 60\n apply_text_normalization?: 'auto' | 'off' | 'on'; // default: \"auto\"\n}\n\nexport interface RimeOptions {}\n\nexport interface InworldOptions {}\n\ntype _TTSModels = CartesiaModels | ElevenlabsModels | RimeModels | InworldModels;\n\nexport type TTSModels = CartesiaModels | ElevenlabsModels | RimeModels | InworldModels | AnyString;\n\nexport type ModelWithVoice = `${_TTSModels}:${string}` | TTSModels;\n\nexport type TTSOptions<TModel extends TTSModels> = TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends ElevenlabsModels\n ? ElevenlabsOptions\n : TModel extends RimeOptions\n ? RimeOptions\n : TModel extends InworldOptions\n ? InworldOptions\n : Record<string, unknown>;\n\ntype TTSEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: TTSEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'https://agent-gateway.livekit.cloud/v1';\nconst NUM_CHANNELS = 1;\nconst DEFAULT_LANGUAGE = 'en';\n\nexport interface InferenceTTSOptions<TModel extends TTSModels> {\n model?: TModel;\n voice?: string;\n language?: string;\n encoding: TTSEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: TTSOptions<TModel>;\n}\n\n/**\n * Livekit Cloud Inference TTS\n */\nexport class TTS<TModel extends TTSModels> extends BaseTTS {\n private opts: InferenceTTSOptions<TModel>;\n private streams: Set<SynthesizeStream<TModel>> = new Set();\n\n #logger = log();\n\n constructor(opts: {\n model: TModel;\n voice?: string;\n language?: string;\n baseURL?: string;\n encoding?: TTSEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: TTSOptions<TModel>;\n }) {\n const sampleRate = opts?.sampleRate ?? DEFAULT_SAMPLE_RATE;\n super(sampleRate, 1, { streaming: true });\n\n const {\n model,\n voice,\n language = DEFAULT_LANGUAGE,\n baseURL,\n encoding = DEFAULT_ENCODING,\n apiKey,\n apiSecret,\n modelOptions = {} as TTSOptions<TModel>,\n } = opts || {};\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n // read voice id from the model if provided: \"provider/model:voice_id\"\n let nextModel = model;\n let nextVoice = voice;\n if (typeof nextModel === 'string') {\n const idx = nextModel.lastIndexOf(':');\n if (idx !== -1) {\n const voiceFromModel = nextModel.slice(idx + 1);\n if (nextVoice && nextVoice !== voiceFromModel) {\n this.#logger.warn(\n '`voice` is provided via both argument and model, using the one from the argument',\n { voice: nextVoice, model: nextModel },\n );\n } else {\n nextVoice = voiceFromModel;\n }\n nextModel = nextModel.slice(0, idx) as TModel;\n }\n }\n\n this.opts = {\n model: nextModel,\n voice: nextVoice,\n language,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n };\n }\n\n get label() {\n return 'inference.TTS';\n }\n\n static fromModelString(modelString: string): TTS<AnyString> {\n if (modelString.includes(':')) {\n const [model, voice] = modelString.split(':') as [TTSModels, string];\n return new TTS({ model, voice });\n }\n return new TTS({ model: modelString });\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n synthesize(_: string): ChunkedStream {\n throw new Error('ChunkedStream is not implemented');\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream<TModel> {\n const { connOptions = DEFAULT_API_CONNECT_OPTIONS } = options || {};\n const stream = new SynthesizeStream(this, { ...this.opts }, connOptions);\n this.streams.add(stream);\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/tts`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const params = {\n type: 'session.create',\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n } as TtsSessionCreateEvent;\n\n if (this.opts.voice) params.voice = this.opts.voice;\n if (this.opts.model) params.model = this.opts.model;\n if (this.opts.language) params.language = this.opts.language;\n\n const socket = await connectWs(url, headers, timeout);\n socket.send(JSON.stringify(params));\n return socket;\n }\n\n async closeWs(ws: WebSocket) {\n await ws.close();\n }\n\n async close() {\n for (const stream of this.streams) {\n await stream.close();\n }\n this.streams.clear();\n }\n}\n\nexport class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeStream {\n private opts: InferenceTTSOptions<TModel>;\n private tts: TTS<TModel>;\n private connOptions: APIConnectOptions;\n\n #logger = log();\n\n constructor(tts: TTS<TModel>, opts: InferenceTTSOptions<TModel>, connOptions: APIConnectOptions) {\n super(tts, connOptions);\n this.opts = opts;\n this.tts = tts;\n this.connOptions = connOptions;\n }\n\n get label() {\n return 'inference.SynthesizeStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n }\n\n protected async run(): Promise<void> {\n let ws: WebSocket | null = null;\n let closing = false;\n let finalReceived = false;\n let lastFrame: AudioFrame | undefined;\n\n const sendTokenizerStream = new tokenizeBasic.SentenceTokenizer().stream();\n const eventChannel = createStreamChannel<TtsServerEvent>();\n const requestId = shortuuid('tts_request_');\n\n const resourceCleanup = () => {\n if (closing) return;\n closing = true;\n sendTokenizerStream.close();\n eventChannel.close();\n ws?.removeAllListeners();\n ws?.close();\n };\n\n const sendClientEvent = async (event: TtsClientEvent) => {\n const validatedEvent = await ttsClientEventSchema.parseAsync(event);\n if (!ws || ws.readyState !== WebSocket.OPEN) {\n this.#logger.warn('Trying to send client TTS event to a closed WebSocket');\n return;\n }\n ws.send(JSON.stringify(validatedEvent));\n };\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const createInputTask = async () => {\n for await (const data of this.input) {\n if (this.abortController.signal.aborted) break;\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n sendTokenizerStream.flush();\n continue;\n }\n sendTokenizerStream.pushText(data);\n }\n sendTokenizerStream.endInput();\n };\n\n const createSentenceStreamTask = async () => {\n for await (const ev of sendTokenizerStream) {\n if (this.abortController.signal.aborted) break;\n\n sendClientEvent({\n type: 'input_transcript',\n transcript: ev.token + ' ',\n });\n }\n\n sendClientEvent({ type: 'session.flush' });\n };\n\n const createWsListenerTask = async (ws: WebSocket) => {\n return new Promise<void>((resolve, reject) => {\n this.abortController.signal.addEventListener('abort', () => {\n resourceCleanup();\n reject(new Error('WebSocket connection aborted'));\n });\n\n ws.on('message', async (data) => {\n const eventJson = JSON.parse(data.toString()) as Record<string, unknown>;\n const validatedEvent = ttsServerEventSchema.parse(eventJson);\n eventChannel.write(validatedEvent);\n });\n\n ws.on('error', (e) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n resourceCleanup();\n reject(e);\n });\n\n ws.on('close', () => {\n resourceCleanup();\n\n if (!closing) return this.#logger.error('WebSocket closed unexpectedly');\n if (finalReceived) return resolve();\n\n reject(\n new APIStatusError({\n message: 'Gateway connection closed unexpectedly',\n options: { requestId },\n }),\n );\n });\n });\n };\n\n const createRecvTask = async () => {\n let currentSessionId: string | null = null;\n\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n while (!this.closed && !this.abortController.signal.aborted) {\n const result = await reader.read();\n if (this.abortController.signal.aborted) return;\n if (result.done) return;\n\n const serverEvent = result.value;\n switch (serverEvent.type) {\n case 'session.created':\n currentSessionId = serverEvent.session_id;\n break;\n case 'output_audio':\n const base64Data = new Int8Array(Buffer.from(serverEvent.audio, 'base64'));\n for (const frame of bstream.write(base64Data.buffer)) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n break;\n case 'done':\n finalReceived = true;\n for (const frame of bstream.flush()) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n sendLastFrame(currentSessionId!, true);\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n break;\n case 'session.closed':\n resourceCleanup();\n break;\n case 'error':\n this.#logger.error(\n { serverEvent },\n 'Received error message from LiveKit TTS WebSocket',\n );\n resourceCleanup();\n throw new APIError(`LiveKit TTS returned error: ${serverEvent.message}`);\n default:\n this.#logger.warn('Unexpected message %s', serverEvent);\n break;\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n ws = await this.tts.connectWs(this.connOptions.timeoutMs);\n\n await Promise.all([\n createInputTask(),\n createSentenceStreamTask(),\n createWsListenerTask(ws),\n createRecvTask(),\n ]);\n } catch (e) {\n this.#logger.error({ error: e }, 'Error in SynthesizeStream');\n } finally {\n resourceCleanup();\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,gBAA0B;AAC1B,wBAAyC;AACzC,mBAAgC;AAChC,iBAAoB;AACpB,4BAAoC;AACpC,sBAAuC;AAEvC,iBAAyE;AACzE,mBAAoE;AACpE,mBAA0B;AAC1B,wBAMO;AACP,IAAAA,gBAA6D;AAoD7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,eAAe;AACrB,MAAM,mBAAmB;AAiBlB,MAAM,YAAsC,WAAAC,IAAQ;AAAA,EACjD;AAAA,EACA,UAAyC,oBAAI,IAAI;AAAA,EAEzD,cAAU,gBAAI;AAAA,EAEd,YAAY,MAUT;AACD,UAAM,cAAa,6BAAM,eAAc;AACvC,UAAM,YAAY,GAAG,EAAE,WAAW,KAAK,CAAC;AAExC,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA;AAAA,MACA,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;AAGA,QAAI,YAAY;AAChB,QAAI,YAAY;AAChB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,iBAAiB,UAAU,MAAM,MAAM,CAAC;AAC9C,YAAI,aAAa,cAAc,gBAAgB;AAC7C,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,OAAO,WAAW,OAAO,UAAU;AAAA,UACvC;AAAA,QACF,OAAO;AACL,sBAAY;AAAA,QACd;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AAEA,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,OAAO;AAAA,MACP;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,QAAI,YAAY,SAAS,GAAG,GAAG;AAC7B,YAAM,CAAC,OAAO,KAAK,IAAI,YAAY,MAAM,GAAG;AAC5C,aAAO,IAAI,IAAI,EAAE,OAAO,MAAM,CAAC;AAAA,IACjC;AACA,WAAO,IAAI,IAAI,EAAE,OAAO,YAAY,CAAC;AAAA,EACvC;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,WAAW,GAA0B;AACnC,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AAAA,EAEA,OAAO,SAAyE;AAC9E,UAAM,EAAE,cAAc,yCAA4B,IAAI,WAAW,CAAC;AAClE,UAAM,SAAS,IAAI,iBAAiB,MAAM,EAAE,GAAG,KAAK,KAAK,GAAG,WAAW;AACvE,SAAK,QAAQ,IAAI,MAAM;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AACnD,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,UAAM,iCAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS;AAAA,MACb,MAAM;AAAA,MACN,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,MACxC,UAAU,KAAK,KAAK;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,IACnB;AAEA,QAAI,KAAK,KAAK,MAAO,QAAO,QAAQ,KAAK,KAAK;AAC9C,QAAI,KAAK,KAAK,MAAO,QAAO,QAAQ,KAAK,KAAK;AAC9C,QAAI,KAAK,KAAK,SAAU,QAAO,WAAW,KAAK,KAAK;AAEpD,UAAM,SAAS,UAAM,yBAAU,KAAK,SAAS,OAAO;AACpD,WAAO,KAAK,KAAK,UAAU,MAAM,CAAC;AAClC,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,QAAQ,IAAe;AAC3B,UAAM,GAAG,MAAM;AAAA,EACjB;AAAA,EAEA,MAAM,QAAQ;AACZ,eAAW,UAAU,KAAK,SAAS;AACjC,YAAM,OAAO,MAAM;AAAA,IACrB;AACA,SAAK,QAAQ,MAAM;AAAA,EACrB;AACF;AAEO,MAAM,yBAAmD,WAAAC,iBAAqB;AAAA,EAC3E;AAAA,EACA;AAAA,EACA;AAAA,EAER,cAAU,gBAAI;AAAA,EAEd,YAAY,KAAkB,MAAmC,aAAgC;AAC/F,UAAM,KAAK,WAAW;AACtB,SAAK,OAAO;AACZ,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAAA,EACtC;AAAA,EAEA,MAAgB,MAAqB;AACnC,QAAI,KAAuB;AAC3B,QAAI,UAAU;AACd,QAAI,gBAAgB;AACpB,QAAI;AAEJ,UAAM,sBAAsB,IAAI,gBAAAC,MAAc,kBAAkB,EAAE,OAAO;AACzE,UAAM,mBAAe,2CAAoC;AACzD,UAAM,gBAAY,wBAAU,cAAc;AAE1C,UAAM,kBAAkB,MAAM;AAC5B,UAAI,QAAS;AACb,gBAAU;AACV,0BAAoB,MAAM;AAC1B,mBAAa,MAAM;AACnB,+BAAI;AACJ,+BAAI;AAAA,IACN;AAEA,UAAM,kBAAkB,OAAO,UAA0B;AACvD,YAAM,iBAAiB,MAAM,uCAAqB,WAAW,KAAK;AAClE,UAAI,CAAC,MAAM,GAAG,eAAe,oBAAU,MAAM;AAC3C,aAAK,QAAQ,KAAK,uDAAuD;AACzE;AAAA,MACF;AACA,SAAG,KAAK,KAAK,UAAU,cAAc,CAAC;AAAA,IACxC;AAEA,UAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,UAAI,WAAW;AACb,aAAK,MAAM,IAAI,EAAE,WAAW,WAAW,OAAO,WAAW,MAAM,CAAC;AAChE,oBAAY;AAAA,MACd;AAAA,IACF;AAEA,UAAM,kBAAkB,YAAY;AAClC,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,8BAAoB,MAAM;AAC1B;AAAA,QACF;AACA,4BAAoB,SAAS,IAAI;AAAA,MACnC;AACA,0BAAoB,SAAS;AAAA,IAC/B;AAEA,UAAM,2BAA2B,YAAY;AAC3C,uBAAiB,MAAM,qBAAqB;AAC1C,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,wBAAgB;AAAA,UACd,MAAM;AAAA,UACN,YAAY,GAAG,QAAQ;AAAA,QACzB,CAAC;AAAA,MACH;AAEA,sBAAgB,EAAE,MAAM,gBAAgB,CAAC;AAAA,IAC3C;AAEA,UAAM,uBAAuB,OAAOC,QAAkB;AACpD,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,aAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,0BAAgB;AAChB,iBAAO,IAAI,MAAM,8BAA8B,CAAC;AAAA,QAClD,CAAC;AAED,QAAAA,IAAG,GAAG,WAAW,OAAO,SAAS;AAC/B,gBAAM,YAAY,KAAK,MAAM,KAAK,SAAS,CAAC;AAC5C,gBAAM,iBAAiB,uCAAqB,MAAM,SAAS;AAC3D,uBAAa,MAAM,cAAc;AAAA,QACnC,CAAC;AAED,QAAAA,IAAG,GAAG,SAAS,CAAC,MAAM;AACpB,eAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,0BAAgB;AAChB,iBAAO,CAAC;AAAA,QACV,CAAC;AAED,QAAAA,IAAG,GAAG,SAAS,MAAM;AACnB,0BAAgB;AAEhB,cAAI,CAAC,QAAS,QAAO,KAAK,QAAQ,MAAM,+BAA+B;AACvE,cAAI,cAAe,QAAO,QAAQ;AAElC;AAAA,YACE,IAAI,iCAAe;AAAA,cACjB,SAAS;AAAA,cACT,SAAS,EAAE,UAAU;AAAA,YACvB,CAAC;AAAA,UACH;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,UAAM,iBAAiB,YAAY;AACjC,UAAI,mBAAkC;AAEtC,YAAM,UAAU,IAAI,6BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,YAAM,oBAAoB,aAAa,OAAO;AAC9C,YAAM,SAAS,kBAAkB,UAAU;AAE3C,UAAI;AACF,eAAO,CAAC,KAAK,UAAU,CAAC,KAAK,gBAAgB,OAAO,SAAS;AAC3D,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,cAAI,OAAO,KAAM;AAEjB,gBAAM,cAAc,OAAO;AAC3B,kBAAQ,YAAY,MAAM;AAAA,YACxB,KAAK;AACH,iCAAmB,YAAY;AAC/B;AAAA,YACF,KAAK;AACH,oBAAM,aAAa,IAAI,UAAU,OAAO,KAAK,YAAY,OAAO,QAAQ,CAAC;AACzE,yBAAW,SAAS,QAAQ,MAAM,WAAW,MAAM,GAAG;AACpD,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA;AAAA,YACF,KAAK;AACH,8BAAgB;AAChB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA,4BAAc,kBAAmB,IAAI;AACrC,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAC7C;AAAA,YACF,KAAK;AACH,8BAAgB;AAChB;AAAA,YACF,KAAK;AACH,mBAAK,QAAQ;AAAA,gBACX,EAAE,YAAY;AAAA,gBACd;AAAA,cACF;AACA,8BAAgB;AAChB,oBAAM,IAAI,2BAAS,+BAA+B,YAAY,OAAO,EAAE;AAAA,YACzE;AACE,mBAAK,QAAQ,KAAK,yBAAyB,WAAW;AACtD;AAAA,UACJ;AAAA,QACF;AAAA,MACF,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,kBAAkB,OAAO;AAAA,QACjC,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,QACxF;AAAA,MACF;AAAA,IACF;AAEA,QAAI;AACF,WAAK,MAAM,KAAK,IAAI,UAAU,KAAK,YAAY,SAAS;AAExD,YAAM,QAAQ,IAAI;AAAA,QAChB,gBAAgB;AAAA,QAChB,yBAAyB;AAAA,QACzB,qBAAqB,EAAE;AAAA,QACvB,eAAe;AAAA,MACjB,CAAC;AAAA,IACH,SAAS,GAAG;AACV,WAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,2BAA2B;AAAA,IAC9D,UAAE;AACA,sBAAgB;AAAA,IAClB;AAAA,EACF;AACF;","names":["import_utils","BaseTTS","BaseSynthesizeStream","tokenizeBasic","ws"]}
1
+ {"version":3,"sources":["../../src/inference/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport { basic as tokenizeBasic } from '../tokenize/index.js';\nimport type { ChunkedStream } from '../tts/index.js';\nimport { SynthesizeStream as BaseSynthesizeStream, TTS as BaseTTS } from '../tts/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { shortuuid } from '../utils.js';\nimport {\n type TtsClientEvent,\n type TtsServerEvent,\n type TtsSessionCreateEvent,\n ttsClientEventSchema,\n ttsServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken } from './utils.js';\n\nexport type CartesiaModels =\n | 'cartesia'\n | 'cartesia/sonic'\n | 'cartesia/sonic-2'\n | 'cartesia/sonic-turbo';\n\nexport type ElevenlabsModels =\n | 'elevenlabs'\n | 'elevenlabs/eleven_flash_v2'\n | 'elevenlabs/eleven_flash_v2_5'\n | 'elevenlabs/eleven_turbo_v2'\n | 'elevenlabs/eleven_turbo_v2_5'\n | 'elevenlabs/eleven_multilingual_v2';\n\nexport type RimeModels = 'rime' | 'rime/mist' | 'rime/mistv2' | 'rime/arcana';\n\nexport type InworldModels = 'inworld' | 'inworld/inworld-tts-1';\n\nexport interface CartesiaOptions {\n duration?: number; // max duration of audio in seconds\n speed?: 'slow' | 'normal' | 'fast'; // default: not specified\n}\n\nexport interface ElevenlabsOptions {\n inactivity_timeout?: number; // default: 60\n apply_text_normalization?: 'auto' | 'off' | 'on'; // default: \"auto\"\n}\n\nexport interface RimeOptions {}\n\nexport interface InworldOptions {}\n\ntype _TTSModels = CartesiaModels | ElevenlabsModels | RimeModels | InworldModels;\n\nexport type TTSModels = CartesiaModels | ElevenlabsModels | RimeModels | InworldModels | AnyString;\n\nexport type ModelWithVoice = `${_TTSModels}:${string}` | TTSModels;\n\nexport type TTSOptions<TModel extends TTSModels> = TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends ElevenlabsModels\n ? ElevenlabsOptions\n : TModel extends RimeOptions\n ? RimeOptions\n : TModel extends InworldOptions\n ? InworldOptions\n : Record<string, unknown>;\n\ntype TTSEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: TTSEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'https://agent-gateway.livekit.cloud/v1';\nconst NUM_CHANNELS = 1;\nconst DEFAULT_LANGUAGE = 'en';\n\nexport interface InferenceTTSOptions<TModel extends TTSModels> {\n model?: TModel;\n voice?: string;\n language?: string;\n encoding: TTSEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: TTSOptions<TModel>;\n}\n\n/**\n * Livekit Cloud Inference TTS\n */\nexport class TTS<TModel extends TTSModels> extends BaseTTS {\n private opts: InferenceTTSOptions<TModel>;\n private streams: Set<SynthesizeStream<TModel>> = new Set();\n\n #logger = log();\n\n constructor(opts: {\n model: TModel;\n voice?: string;\n language?: string;\n baseURL?: string;\n encoding?: TTSEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: TTSOptions<TModel>;\n }) {\n const sampleRate = opts?.sampleRate ?? DEFAULT_SAMPLE_RATE;\n super(sampleRate, 1, { streaming: true });\n\n const {\n model,\n voice,\n language = DEFAULT_LANGUAGE,\n baseURL,\n encoding = DEFAULT_ENCODING,\n apiKey,\n apiSecret,\n modelOptions = {} as TTSOptions<TModel>,\n } = opts || {};\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n // read voice id from the model if provided: \"provider/model:voice_id\"\n let nextModel = model;\n let nextVoice = voice;\n if (typeof nextModel === 'string') {\n const idx = nextModel.lastIndexOf(':');\n if (idx !== -1) {\n const voiceFromModel = nextModel.slice(idx + 1);\n if (nextVoice && nextVoice !== voiceFromModel) {\n this.#logger.warn(\n '`voice` is provided via both argument and model, using the one from the argument',\n { voice: nextVoice, model: nextModel },\n );\n } else {\n nextVoice = voiceFromModel;\n }\n nextModel = nextModel.slice(0, idx) as TModel;\n }\n }\n\n this.opts = {\n model: nextModel,\n voice: nextVoice,\n language,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n };\n }\n\n get label() {\n return 'inference.TTS';\n }\n\n static fromModelString(modelString: string): TTS<AnyString> {\n if (modelString.includes(':')) {\n const [model, voice] = modelString.split(':') as [TTSModels, string];\n return new TTS({ model, voice });\n }\n return new TTS({ model: modelString });\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n synthesize(_: string): ChunkedStream {\n throw new Error('ChunkedStream is not implemented');\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream<TModel> {\n const { connOptions = DEFAULT_API_CONNECT_OPTIONS } = options || {};\n const stream = new SynthesizeStream(this, { ...this.opts }, connOptions);\n this.streams.add(stream);\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/tts`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const params = {\n type: 'session.create',\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n } as TtsSessionCreateEvent;\n\n if (this.opts.voice) params.voice = this.opts.voice;\n if (this.opts.model) params.model = this.opts.model;\n if (this.opts.language) params.language = this.opts.language;\n\n const socket = await connectWs(url, headers, timeout);\n socket.send(JSON.stringify(params));\n return socket;\n }\n\n async closeWs(ws: WebSocket) {\n await ws.close();\n }\n\n async close() {\n for (const stream of this.streams) {\n await stream.close();\n }\n this.streams.clear();\n }\n}\n\nexport class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeStream {\n private opts: InferenceTTSOptions<TModel>;\n private tts: TTS<TModel>;\n private connOptions: APIConnectOptions;\n\n #logger = log();\n\n constructor(tts: TTS<TModel>, opts: InferenceTTSOptions<TModel>, connOptions: APIConnectOptions) {\n super(tts, connOptions);\n this.opts = opts;\n this.tts = tts;\n this.connOptions = connOptions;\n }\n\n get label() {\n return 'inference.SynthesizeStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n }\n\n protected async run(): Promise<void> {\n let ws: WebSocket | null = null;\n let closing = false;\n let finalReceived = false;\n let lastFrame: AudioFrame | undefined;\n\n const sendTokenizerStream = new tokenizeBasic.SentenceTokenizer().stream();\n const eventChannel = createStreamChannel<TtsServerEvent>();\n const requestId = shortuuid('tts_request_');\n\n const resourceCleanup = () => {\n if (closing) return;\n closing = true;\n sendTokenizerStream.close();\n eventChannel.close();\n ws?.removeAllListeners();\n ws?.close();\n };\n\n const sendClientEvent = async (event: TtsClientEvent) => {\n // Don't send events to a closed WebSocket or aborted controller\n if (this.abortController.signal.aborted || closing) return;\n\n const validatedEvent = await ttsClientEventSchema.parseAsync(event);\n if (!ws || ws.readyState !== WebSocket.OPEN) {\n this.#logger.warn('Trying to send client TTS event to a closed WebSocket');\n return;\n }\n ws.send(JSON.stringify(validatedEvent));\n };\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const createInputTask = async () => {\n for await (const data of this.input) {\n if (this.abortController.signal.aborted) break;\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n sendTokenizerStream.flush();\n continue;\n }\n sendTokenizerStream.pushText(data);\n }\n sendTokenizerStream.endInput();\n };\n\n const createSentenceStreamTask = async () => {\n for await (const ev of sendTokenizerStream) {\n if (this.abortController.signal.aborted) break;\n\n sendClientEvent({\n type: 'input_transcript',\n transcript: ev.token + ' ',\n });\n }\n\n sendClientEvent({ type: 'session.flush' });\n };\n\n const createWsListenerTask = async (ws: WebSocket) => {\n return new Promise<void>((resolve, reject) => {\n this.abortController.signal.addEventListener('abort', () => {\n resourceCleanup();\n resolve(); // Abort is triggered by close(), which is a normal shutdown, not an error\n });\n\n ws.on('message', async (data) => {\n const eventJson = JSON.parse(data.toString()) as Record<string, unknown>;\n const validatedEvent = ttsServerEventSchema.parse(eventJson);\n eventChannel.write(validatedEvent);\n });\n\n ws.on('error', (e) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n resourceCleanup();\n reject(e);\n });\n\n ws.on('close', () => {\n resourceCleanup();\n\n if (!closing) return this.#logger.error('WebSocket closed unexpectedly');\n if (finalReceived) return resolve();\n\n reject(\n new APIStatusError({\n message: 'Gateway connection closed unexpectedly',\n options: { requestId },\n }),\n );\n });\n });\n };\n\n const createRecvTask = async () => {\n let currentSessionId: string | null = null;\n\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n while (!this.closed && !this.abortController.signal.aborted) {\n const result = await reader.read();\n if (this.abortController.signal.aborted) return;\n if (result.done) return;\n\n const serverEvent = result.value;\n switch (serverEvent.type) {\n case 'session.created':\n currentSessionId = serverEvent.session_id;\n break;\n case 'output_audio':\n const base64Data = new Int8Array(Buffer.from(serverEvent.audio, 'base64'));\n for (const frame of bstream.write(base64Data.buffer)) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n break;\n case 'done':\n finalReceived = true;\n for (const frame of bstream.flush()) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n sendLastFrame(currentSessionId!, true);\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n break;\n case 'session.closed':\n resourceCleanup();\n break;\n case 'error':\n this.#logger.error(\n { serverEvent },\n 'Received error message from LiveKit TTS WebSocket',\n );\n resourceCleanup();\n throw new APIError(`LiveKit TTS returned error: ${serverEvent.message}`);\n default:\n this.#logger.warn('Unexpected message %s', serverEvent);\n break;\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n ws = await this.tts.connectWs(this.connOptions.timeoutMs);\n\n await Promise.all([\n createInputTask(),\n createSentenceStreamTask(),\n createWsListenerTask(ws),\n createRecvTask(),\n ]);\n } finally {\n resourceCleanup();\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,gBAA0B;AAC1B,wBAAyC;AACzC,mBAAgC;AAChC,iBAAoB;AACpB,4BAAoC;AACpC,sBAAuC;AAEvC,iBAAyE;AACzE,mBAAoE;AACpE,mBAA0B;AAC1B,wBAMO;AACP,IAAAA,gBAA6D;AAoD7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,eAAe;AACrB,MAAM,mBAAmB;AAiBlB,MAAM,YAAsC,WAAAC,IAAQ;AAAA,EACjD;AAAA,EACA,UAAyC,oBAAI,IAAI;AAAA,EAEzD,cAAU,gBAAI;AAAA,EAEd,YAAY,MAUT;AACD,UAAM,cAAa,6BAAM,eAAc;AACvC,UAAM,YAAY,GAAG,EAAE,WAAW,KAAK,CAAC;AAExC,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA;AAAA,MACA,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;AAGA,QAAI,YAAY;AAChB,QAAI,YAAY;AAChB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,iBAAiB,UAAU,MAAM,MAAM,CAAC;AAC9C,YAAI,aAAa,cAAc,gBAAgB;AAC7C,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,OAAO,WAAW,OAAO,UAAU;AAAA,UACvC;AAAA,QACF,OAAO;AACL,sBAAY;AAAA,QACd;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AAEA,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,OAAO;AAAA,MACP;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,QAAI,YAAY,SAAS,GAAG,GAAG;AAC7B,YAAM,CAAC,OAAO,KAAK,IAAI,YAAY,MAAM,GAAG;AAC5C,aAAO,IAAI,IAAI,EAAE,OAAO,MAAM,CAAC;AAAA,IACjC;AACA,WAAO,IAAI,IAAI,EAAE,OAAO,YAAY,CAAC;AAAA,EACvC;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,WAAW,GAA0B;AACnC,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AAAA,EAEA,OAAO,SAAyE;AAC9E,UAAM,EAAE,cAAc,yCAA4B,IAAI,WAAW,CAAC;AAClE,UAAM,SAAS,IAAI,iBAAiB,MAAM,EAAE,GAAG,KAAK,KAAK,GAAG,WAAW;AACvE,SAAK,QAAQ,IAAI,MAAM;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AACnD,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,UAAM,iCAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS;AAAA,MACb,MAAM;AAAA,MACN,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,MACxC,UAAU,KAAK,KAAK;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,IACnB;AAEA,QAAI,KAAK,KAAK,MAAO,QAAO,QAAQ,KAAK,KAAK;AAC9C,QAAI,KAAK,KAAK,MAAO,QAAO,QAAQ,KAAK,KAAK;AAC9C,QAAI,KAAK,KAAK,SAAU,QAAO,WAAW,KAAK,KAAK;AAEpD,UAAM,SAAS,UAAM,yBAAU,KAAK,SAAS,OAAO;AACpD,WAAO,KAAK,KAAK,UAAU,MAAM,CAAC;AAClC,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,QAAQ,IAAe;AAC3B,UAAM,GAAG,MAAM;AAAA,EACjB;AAAA,EAEA,MAAM,QAAQ;AACZ,eAAW,UAAU,KAAK,SAAS;AACjC,YAAM,OAAO,MAAM;AAAA,IACrB;AACA,SAAK,QAAQ,MAAM;AAAA,EACrB;AACF;AAEO,MAAM,yBAAmD,WAAAC,iBAAqB;AAAA,EAC3E;AAAA,EACA;AAAA,EACA;AAAA,EAER,cAAU,gBAAI;AAAA,EAEd,YAAY,KAAkB,MAAmC,aAAgC;AAC/F,UAAM,KAAK,WAAW;AACtB,SAAK,OAAO;AACZ,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAAA,EACtC;AAAA,EAEA,MAAgB,MAAqB;AACnC,QAAI,KAAuB;AAC3B,QAAI,UAAU;AACd,QAAI,gBAAgB;AACpB,QAAI;AAEJ,UAAM,sBAAsB,IAAI,gBAAAC,MAAc,kBAAkB,EAAE,OAAO;AACzE,UAAM,mBAAe,2CAAoC;AACzD,UAAM,gBAAY,wBAAU,cAAc;AAE1C,UAAM,kBAAkB,MAAM;AAC5B,UAAI,QAAS;AACb,gBAAU;AACV,0BAAoB,MAAM;AAC1B,mBAAa,MAAM;AACnB,+BAAI;AACJ,+BAAI;AAAA,IACN;AAEA,UAAM,kBAAkB,OAAO,UAA0B;AAEvD,UAAI,KAAK,gBAAgB,OAAO,WAAW,QAAS;AAEpD,YAAM,iBAAiB,MAAM,uCAAqB,WAAW,KAAK;AAClE,UAAI,CAAC,MAAM,GAAG,eAAe,oBAAU,MAAM;AAC3C,aAAK,QAAQ,KAAK,uDAAuD;AACzE;AAAA,MACF;AACA,SAAG,KAAK,KAAK,UAAU,cAAc,CAAC;AAAA,IACxC;AAEA,UAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,UAAI,WAAW;AACb,aAAK,MAAM,IAAI,EAAE,WAAW,WAAW,OAAO,WAAW,MAAM,CAAC;AAChE,oBAAY;AAAA,MACd;AAAA,IACF;AAEA,UAAM,kBAAkB,YAAY;AAClC,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,8BAAoB,MAAM;AAC1B;AAAA,QACF;AACA,4BAAoB,SAAS,IAAI;AAAA,MACnC;AACA,0BAAoB,SAAS;AAAA,IAC/B;AAEA,UAAM,2BAA2B,YAAY;AAC3C,uBAAiB,MAAM,qBAAqB;AAC1C,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,wBAAgB;AAAA,UACd,MAAM;AAAA,UACN,YAAY,GAAG,QAAQ;AAAA,QACzB,CAAC;AAAA,MACH;AAEA,sBAAgB,EAAE,MAAM,gBAAgB,CAAC;AAAA,IAC3C;AAEA,UAAM,uBAAuB,OAAOC,QAAkB;AACpD,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,aAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,0BAAgB;AAChB,kBAAQ;AAAA,QACV,CAAC;AAED,QAAAA,IAAG,GAAG,WAAW,OAAO,SAAS;AAC/B,gBAAM,YAAY,KAAK,MAAM,KAAK,SAAS,CAAC;AAC5C,gBAAM,iBAAiB,uCAAqB,MAAM,SAAS;AAC3D,uBAAa,MAAM,cAAc;AAAA,QACnC,CAAC;AAED,QAAAA,IAAG,GAAG,SAAS,CAAC,MAAM;AACpB,eAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,0BAAgB;AAChB,iBAAO,CAAC;AAAA,QACV,CAAC;AAED,QAAAA,IAAG,GAAG,SAAS,MAAM;AACnB,0BAAgB;AAEhB,cAAI,CAAC,QAAS,QAAO,KAAK,QAAQ,MAAM,+BAA+B;AACvE,cAAI,cAAe,QAAO,QAAQ;AAElC;AAAA,YACE,IAAI,iCAAe;AAAA,cACjB,SAAS;AAAA,cACT,SAAS,EAAE,UAAU;AAAA,YACvB,CAAC;AAAA,UACH;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,UAAM,iBAAiB,YAAY;AACjC,UAAI,mBAAkC;AAEtC,YAAM,UAAU,IAAI,6BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,YAAM,oBAAoB,aAAa,OAAO;AAC9C,YAAM,SAAS,kBAAkB,UAAU;AAE3C,UAAI;AACF,eAAO,CAAC,KAAK,UAAU,CAAC,KAAK,gBAAgB,OAAO,SAAS;AAC3D,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,cAAI,OAAO,KAAM;AAEjB,gBAAM,cAAc,OAAO;AAC3B,kBAAQ,YAAY,MAAM;AAAA,YACxB,KAAK;AACH,iCAAmB,YAAY;AAC/B;AAAA,YACF,KAAK;AACH,oBAAM,aAAa,IAAI,UAAU,OAAO,KAAK,YAAY,OAAO,QAAQ,CAAC;AACzE,yBAAW,SAAS,QAAQ,MAAM,WAAW,MAAM,GAAG;AACpD,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA;AAAA,YACF,KAAK;AACH,8BAAgB;AAChB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA,4BAAc,kBAAmB,IAAI;AACrC,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAC7C;AAAA,YACF,KAAK;AACH,8BAAgB;AAChB;AAAA,YACF,KAAK;AACH,mBAAK,QAAQ;AAAA,gBACX,EAAE,YAAY;AAAA,gBACd;AAAA,cACF;AACA,8BAAgB;AAChB,oBAAM,IAAI,2BAAS,+BAA+B,YAAY,OAAO,EAAE;AAAA,YACzE;AACE,mBAAK,QAAQ,KAAK,yBAAyB,WAAW;AACtD;AAAA,UACJ;AAAA,QACF;AAAA,MACF,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,kBAAkB,OAAO;AAAA,QACjC,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,QACxF;AAAA,MACF;AAAA,IACF;AAEA,QAAI;AACF,WAAK,MAAM,KAAK,IAAI,UAAU,KAAK,YAAY,SAAS;AAExD,YAAM,QAAQ,IAAI;AAAA,QAChB,gBAAgB;AAAA,QAChB,yBAAyB;AAAA,QACzB,qBAAqB,EAAE;AAAA,QACvB,eAAe;AAAA,MACjB,CAAC;AAAA,IACH,UAAE;AACA,sBAAgB;AAAA,IAClB;AAAA,EACF;AACF;","names":["import_utils","BaseTTS","BaseSynthesizeStream","tokenizeBasic","ws"]}
@@ -1 +1 @@
1
- {"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../../src/inference/tts.ts"],"names":[],"mappings":"AAIA,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAM/B,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AACrD,OAAO,EAAE,gBAAgB,IAAI,oBAAoB,EAAE,GAAG,IAAI,OAAO,EAAE,MAAM,iBAAiB,CAAC;AAC3F,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AASlF,OAAO,EAAE,KAAK,SAAS,EAAgC,MAAM,YAAY,CAAC;AAE1E,MAAM,MAAM,cAAc,GACtB,UAAU,GACV,gBAAgB,GAChB,kBAAkB,GAClB,sBAAsB,CAAC;AAE3B,MAAM,MAAM,gBAAgB,GACxB,YAAY,GACZ,4BAA4B,GAC5B,8BAA8B,GAC9B,4BAA4B,GAC5B,8BAA8B,GAC9B,mCAAmC,CAAC;AAExC,MAAM,MAAM,UAAU,GAAG,MAAM,GAAG,WAAW,GAAG,aAAa,GAAG,aAAa,CAAC;AAE9E,MAAM,MAAM,aAAa,GAAG,SAAS,GAAG,uBAAuB,CAAC;AAEhE,MAAM,WAAW,eAAe;IAC9B,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,KAAK,CAAC,EAAE,MAAM,GAAG,QAAQ,GAAG,MAAM,CAAC;CACpC;AAED,MAAM,WAAW,iBAAiB;IAChC,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,wBAAwB,CAAC,EAAE,MAAM,GAAG,KAAK,GAAG,IAAI,CAAC;CAClD;AAED,MAAM,WAAW,WAAW;CAAG;AAE/B,MAAM,WAAW,cAAc;CAAG;AAElC,KAAK,UAAU,GAAG,cAAc,GAAG,gBAAgB,GAAG,UAAU,GAAG,aAAa,CAAC;AAEjF,MAAM,MAAM,SAAS,GAAG,cAAc,GAAG,gBAAgB,GAAG,UAAU,GAAG,aAAa,GAAG,SAAS,CAAC;AAEnG,MAAM,MAAM,cAAc,GAAG,GAAG,UAAU,IAAI,MAAM,EAAE,GAAG,SAAS,CAAC;AAEnE,MAAM,MAAM,UAAU,CAAC,MAAM,SAAS,SAAS,IAAI,MAAM,SAAS,cAAc,GAC5E,eAAe,GACf,MAAM,SAAS,gBAAgB,GAC7B,iBAAiB,GACjB,MAAM,SAAS,WAAW,GACxB,WAAW,GACX,MAAM,SAAS,cAAc,GAC3B,cAAc,GACd,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAElC,KAAK,WAAW,GAAG,WAAW,CAAC;AAQ/B,MAAM,WAAW,mBAAmB,CAAC,MAAM,SAAS,SAAS;IAC3D,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,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,CAA4C;gBAI/C,IAAI,EAAE;QAChB,KAAK,EAAE,MAAM,CAAC;QACd,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,QAAQ,CAAC,EAAE,MAAM,CAAC;QAClB,OAAO,CAAC,EAAE,MAAM,CAAC;QACjB,QAAQ,CAAC,EAAE,WAAW,CAAC;QACvB,UAAU,CAAC,EAAE,MAAM,CAAC;QACpB,MAAM,CAAC,EAAE,MAAM,CAAC;QAChB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;KACnC;IA2DD,IAAI,KAAK,WAER;IAED,MAAM,CAAC,eAAe,CAAC,WAAW,EAAE,MAAM,GAAG,GAAG,CAAC,SAAS,CAAC;IAQ3D,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,OAAO,GAAG,UAAU,CAAC,CAAC;IAO9F,UAAU,CAAC,CAAC,EAAE,MAAM,GAAG,aAAa;IAIpC,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,gBAAgB,CAAC,MAAM,CAAC;IAOzE,SAAS,CAAC,OAAO,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC;IA0B9C,OAAO,CAAC,EAAE,EAAE,SAAS;IAIrB,KAAK;CAMZ;AAED,qBAAa,gBAAgB,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,oBAAoB;;IAClF,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,GAAG,CAAc;IACzB,OAAO,CAAC,WAAW,CAAoB;gBAI3B,GAAG,EAAE,GAAG,CAAC,MAAM,CAAC,EAAE,IAAI,EAAE,mBAAmB,CAAC,MAAM,CAAC,EAAE,WAAW,EAAE,iBAAiB;IAO/F,IAAI,KAAK,WAER;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,OAAO,GAAG,UAAU,CAAC,CAAC;cAI9E,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;CAyKrC"}
1
+ {"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../../src/inference/tts.ts"],"names":[],"mappings":"AAIA,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAM/B,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AACrD,OAAO,EAAE,gBAAgB,IAAI,oBAAoB,EAAE,GAAG,IAAI,OAAO,EAAE,MAAM,iBAAiB,CAAC;AAC3F,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AASlF,OAAO,EAAE,KAAK,SAAS,EAAgC,MAAM,YAAY,CAAC;AAE1E,MAAM,MAAM,cAAc,GACtB,UAAU,GACV,gBAAgB,GAChB,kBAAkB,GAClB,sBAAsB,CAAC;AAE3B,MAAM,MAAM,gBAAgB,GACxB,YAAY,GACZ,4BAA4B,GAC5B,8BAA8B,GAC9B,4BAA4B,GAC5B,8BAA8B,GAC9B,mCAAmC,CAAC;AAExC,MAAM,MAAM,UAAU,GAAG,MAAM,GAAG,WAAW,GAAG,aAAa,GAAG,aAAa,CAAC;AAE9E,MAAM,MAAM,aAAa,GAAG,SAAS,GAAG,uBAAuB,CAAC;AAEhE,MAAM,WAAW,eAAe;IAC9B,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,KAAK,CAAC,EAAE,MAAM,GAAG,QAAQ,GAAG,MAAM,CAAC;CACpC;AAED,MAAM,WAAW,iBAAiB;IAChC,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,wBAAwB,CAAC,EAAE,MAAM,GAAG,KAAK,GAAG,IAAI,CAAC;CAClD;AAED,MAAM,WAAW,WAAW;CAAG;AAE/B,MAAM,WAAW,cAAc;CAAG;AAElC,KAAK,UAAU,GAAG,cAAc,GAAG,gBAAgB,GAAG,UAAU,GAAG,aAAa,CAAC;AAEjF,MAAM,MAAM,SAAS,GAAG,cAAc,GAAG,gBAAgB,GAAG,UAAU,GAAG,aAAa,GAAG,SAAS,CAAC;AAEnG,MAAM,MAAM,cAAc,GAAG,GAAG,UAAU,IAAI,MAAM,EAAE,GAAG,SAAS,CAAC;AAEnE,MAAM,MAAM,UAAU,CAAC,MAAM,SAAS,SAAS,IAAI,MAAM,SAAS,cAAc,GAC5E,eAAe,GACf,MAAM,SAAS,gBAAgB,GAC7B,iBAAiB,GACjB,MAAM,SAAS,WAAW,GACxB,WAAW,GACX,MAAM,SAAS,cAAc,GAC3B,cAAc,GACd,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAElC,KAAK,WAAW,GAAG,WAAW,CAAC;AAQ/B,MAAM,WAAW,mBAAmB,CAAC,MAAM,SAAS,SAAS;IAC3D,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,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,CAA4C;gBAI/C,IAAI,EAAE;QAChB,KAAK,EAAE,MAAM,CAAC;QACd,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,QAAQ,CAAC,EAAE,MAAM,CAAC;QAClB,OAAO,CAAC,EAAE,MAAM,CAAC;QACjB,QAAQ,CAAC,EAAE,WAAW,CAAC;QACvB,UAAU,CAAC,EAAE,MAAM,CAAC;QACpB,MAAM,CAAC,EAAE,MAAM,CAAC;QAChB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;KACnC;IA2DD,IAAI,KAAK,WAER;IAED,MAAM,CAAC,eAAe,CAAC,WAAW,EAAE,MAAM,GAAG,GAAG,CAAC,SAAS,CAAC;IAQ3D,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,OAAO,GAAG,UAAU,CAAC,CAAC;IAO9F,UAAU,CAAC,CAAC,EAAE,MAAM,GAAG,aAAa;IAIpC,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,gBAAgB,CAAC,MAAM,CAAC;IAOzE,SAAS,CAAC,OAAO,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC;IA0B9C,OAAO,CAAC,EAAE,EAAE,SAAS;IAIrB,KAAK;CAMZ;AAED,qBAAa,gBAAgB,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,oBAAoB;;IAClF,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,GAAG,CAAc;IACzB,OAAO,CAAC,WAAW,CAAoB;gBAI3B,GAAG,EAAE,GAAG,CAAC,MAAM,CAAC,EAAE,IAAI,EAAE,mBAAmB,CAAC,MAAM,CAAC,EAAE,WAAW,EAAE,iBAAiB;IAO/F,IAAI,KAAK,WAER;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,OAAO,GAAG,UAAU,CAAC,CAAC;cAI9E,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;CA0KrC"}
@@ -162,6 +162,7 @@ class SynthesizeStream extends BaseSynthesizeStream {
162
162
  ws == null ? void 0 : ws.close();
163
163
  };
164
164
  const sendClientEvent = async (event) => {
165
+ if (this.abortController.signal.aborted || closing) return;
165
166
  const validatedEvent = await ttsClientEventSchema.parseAsync(event);
166
167
  if (!ws || ws.readyState !== WebSocket.OPEN) {
167
168
  this.#logger.warn("Trying to send client TTS event to a closed WebSocket");
@@ -200,7 +201,7 @@ class SynthesizeStream extends BaseSynthesizeStream {
200
201
  return new Promise((resolve, reject) => {
201
202
  this.abortController.signal.addEventListener("abort", () => {
202
203
  resourceCleanup();
203
- reject(new Error("WebSocket connection aborted"));
204
+ resolve();
204
205
  });
205
206
  ws2.on("message", async (data) => {
206
207
  const eventJson = JSON.parse(data.toString());
@@ -288,8 +289,6 @@ class SynthesizeStream extends BaseSynthesizeStream {
288
289
  createWsListenerTask(ws),
289
290
  createRecvTask()
290
291
  ]);
291
- } catch (e) {
292
- this.#logger.error({ error: e }, "Error in SynthesizeStream");
293
292
  } finally {
294
293
  resourceCleanup();
295
294
  }
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/inference/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport { basic as tokenizeBasic } from '../tokenize/index.js';\nimport type { ChunkedStream } from '../tts/index.js';\nimport { SynthesizeStream as BaseSynthesizeStream, TTS as BaseTTS } from '../tts/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { shortuuid } from '../utils.js';\nimport {\n type TtsClientEvent,\n type TtsServerEvent,\n type TtsSessionCreateEvent,\n ttsClientEventSchema,\n ttsServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken } from './utils.js';\n\nexport type CartesiaModels =\n | 'cartesia'\n | 'cartesia/sonic'\n | 'cartesia/sonic-2'\n | 'cartesia/sonic-turbo';\n\nexport type ElevenlabsModels =\n | 'elevenlabs'\n | 'elevenlabs/eleven_flash_v2'\n | 'elevenlabs/eleven_flash_v2_5'\n | 'elevenlabs/eleven_turbo_v2'\n | 'elevenlabs/eleven_turbo_v2_5'\n | 'elevenlabs/eleven_multilingual_v2';\n\nexport type RimeModels = 'rime' | 'rime/mist' | 'rime/mistv2' | 'rime/arcana';\n\nexport type InworldModels = 'inworld' | 'inworld/inworld-tts-1';\n\nexport interface CartesiaOptions {\n duration?: number; // max duration of audio in seconds\n speed?: 'slow' | 'normal' | 'fast'; // default: not specified\n}\n\nexport interface ElevenlabsOptions {\n inactivity_timeout?: number; // default: 60\n apply_text_normalization?: 'auto' | 'off' | 'on'; // default: \"auto\"\n}\n\nexport interface RimeOptions {}\n\nexport interface InworldOptions {}\n\ntype _TTSModels = CartesiaModels | ElevenlabsModels | RimeModels | InworldModels;\n\nexport type TTSModels = CartesiaModels | ElevenlabsModels | RimeModels | InworldModels | AnyString;\n\nexport type ModelWithVoice = `${_TTSModels}:${string}` | TTSModels;\n\nexport type TTSOptions<TModel extends TTSModels> = TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends ElevenlabsModels\n ? ElevenlabsOptions\n : TModel extends RimeOptions\n ? RimeOptions\n : TModel extends InworldOptions\n ? InworldOptions\n : Record<string, unknown>;\n\ntype TTSEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: TTSEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'https://agent-gateway.livekit.cloud/v1';\nconst NUM_CHANNELS = 1;\nconst DEFAULT_LANGUAGE = 'en';\n\nexport interface InferenceTTSOptions<TModel extends TTSModels> {\n model?: TModel;\n voice?: string;\n language?: string;\n encoding: TTSEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: TTSOptions<TModel>;\n}\n\n/**\n * Livekit Cloud Inference TTS\n */\nexport class TTS<TModel extends TTSModels> extends BaseTTS {\n private opts: InferenceTTSOptions<TModel>;\n private streams: Set<SynthesizeStream<TModel>> = new Set();\n\n #logger = log();\n\n constructor(opts: {\n model: TModel;\n voice?: string;\n language?: string;\n baseURL?: string;\n encoding?: TTSEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: TTSOptions<TModel>;\n }) {\n const sampleRate = opts?.sampleRate ?? DEFAULT_SAMPLE_RATE;\n super(sampleRate, 1, { streaming: true });\n\n const {\n model,\n voice,\n language = DEFAULT_LANGUAGE,\n baseURL,\n encoding = DEFAULT_ENCODING,\n apiKey,\n apiSecret,\n modelOptions = {} as TTSOptions<TModel>,\n } = opts || {};\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n // read voice id from the model if provided: \"provider/model:voice_id\"\n let nextModel = model;\n let nextVoice = voice;\n if (typeof nextModel === 'string') {\n const idx = nextModel.lastIndexOf(':');\n if (idx !== -1) {\n const voiceFromModel = nextModel.slice(idx + 1);\n if (nextVoice && nextVoice !== voiceFromModel) {\n this.#logger.warn(\n '`voice` is provided via both argument and model, using the one from the argument',\n { voice: nextVoice, model: nextModel },\n );\n } else {\n nextVoice = voiceFromModel;\n }\n nextModel = nextModel.slice(0, idx) as TModel;\n }\n }\n\n this.opts = {\n model: nextModel,\n voice: nextVoice,\n language,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n };\n }\n\n get label() {\n return 'inference.TTS';\n }\n\n static fromModelString(modelString: string): TTS<AnyString> {\n if (modelString.includes(':')) {\n const [model, voice] = modelString.split(':') as [TTSModels, string];\n return new TTS({ model, voice });\n }\n return new TTS({ model: modelString });\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n synthesize(_: string): ChunkedStream {\n throw new Error('ChunkedStream is not implemented');\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream<TModel> {\n const { connOptions = DEFAULT_API_CONNECT_OPTIONS } = options || {};\n const stream = new SynthesizeStream(this, { ...this.opts }, connOptions);\n this.streams.add(stream);\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/tts`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const params = {\n type: 'session.create',\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n } as TtsSessionCreateEvent;\n\n if (this.opts.voice) params.voice = this.opts.voice;\n if (this.opts.model) params.model = this.opts.model;\n if (this.opts.language) params.language = this.opts.language;\n\n const socket = await connectWs(url, headers, timeout);\n socket.send(JSON.stringify(params));\n return socket;\n }\n\n async closeWs(ws: WebSocket) {\n await ws.close();\n }\n\n async close() {\n for (const stream of this.streams) {\n await stream.close();\n }\n this.streams.clear();\n }\n}\n\nexport class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeStream {\n private opts: InferenceTTSOptions<TModel>;\n private tts: TTS<TModel>;\n private connOptions: APIConnectOptions;\n\n #logger = log();\n\n constructor(tts: TTS<TModel>, opts: InferenceTTSOptions<TModel>, connOptions: APIConnectOptions) {\n super(tts, connOptions);\n this.opts = opts;\n this.tts = tts;\n this.connOptions = connOptions;\n }\n\n get label() {\n return 'inference.SynthesizeStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n }\n\n protected async run(): Promise<void> {\n let ws: WebSocket | null = null;\n let closing = false;\n let finalReceived = false;\n let lastFrame: AudioFrame | undefined;\n\n const sendTokenizerStream = new tokenizeBasic.SentenceTokenizer().stream();\n const eventChannel = createStreamChannel<TtsServerEvent>();\n const requestId = shortuuid('tts_request_');\n\n const resourceCleanup = () => {\n if (closing) return;\n closing = true;\n sendTokenizerStream.close();\n eventChannel.close();\n ws?.removeAllListeners();\n ws?.close();\n };\n\n const sendClientEvent = async (event: TtsClientEvent) => {\n const validatedEvent = await ttsClientEventSchema.parseAsync(event);\n if (!ws || ws.readyState !== WebSocket.OPEN) {\n this.#logger.warn('Trying to send client TTS event to a closed WebSocket');\n return;\n }\n ws.send(JSON.stringify(validatedEvent));\n };\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const createInputTask = async () => {\n for await (const data of this.input) {\n if (this.abortController.signal.aborted) break;\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n sendTokenizerStream.flush();\n continue;\n }\n sendTokenizerStream.pushText(data);\n }\n sendTokenizerStream.endInput();\n };\n\n const createSentenceStreamTask = async () => {\n for await (const ev of sendTokenizerStream) {\n if (this.abortController.signal.aborted) break;\n\n sendClientEvent({\n type: 'input_transcript',\n transcript: ev.token + ' ',\n });\n }\n\n sendClientEvent({ type: 'session.flush' });\n };\n\n const createWsListenerTask = async (ws: WebSocket) => {\n return new Promise<void>((resolve, reject) => {\n this.abortController.signal.addEventListener('abort', () => {\n resourceCleanup();\n reject(new Error('WebSocket connection aborted'));\n });\n\n ws.on('message', async (data) => {\n const eventJson = JSON.parse(data.toString()) as Record<string, unknown>;\n const validatedEvent = ttsServerEventSchema.parse(eventJson);\n eventChannel.write(validatedEvent);\n });\n\n ws.on('error', (e) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n resourceCleanup();\n reject(e);\n });\n\n ws.on('close', () => {\n resourceCleanup();\n\n if (!closing) return this.#logger.error('WebSocket closed unexpectedly');\n if (finalReceived) return resolve();\n\n reject(\n new APIStatusError({\n message: 'Gateway connection closed unexpectedly',\n options: { requestId },\n }),\n );\n });\n });\n };\n\n const createRecvTask = async () => {\n let currentSessionId: string | null = null;\n\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n while (!this.closed && !this.abortController.signal.aborted) {\n const result = await reader.read();\n if (this.abortController.signal.aborted) return;\n if (result.done) return;\n\n const serverEvent = result.value;\n switch (serverEvent.type) {\n case 'session.created':\n currentSessionId = serverEvent.session_id;\n break;\n case 'output_audio':\n const base64Data = new Int8Array(Buffer.from(serverEvent.audio, 'base64'));\n for (const frame of bstream.write(base64Data.buffer)) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n break;\n case 'done':\n finalReceived = true;\n for (const frame of bstream.flush()) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n sendLastFrame(currentSessionId!, true);\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n break;\n case 'session.closed':\n resourceCleanup();\n break;\n case 'error':\n this.#logger.error(\n { serverEvent },\n 'Received error message from LiveKit TTS WebSocket',\n );\n resourceCleanup();\n throw new APIError(`LiveKit TTS returned error: ${serverEvent.message}`);\n default:\n this.#logger.warn('Unexpected message %s', serverEvent);\n break;\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n ws = await this.tts.connectWs(this.connOptions.timeoutMs);\n\n await Promise.all([\n createInputTask(),\n createSentenceStreamTask(),\n createWsListenerTask(ws),\n createRecvTask(),\n ]);\n } catch (e) {\n this.#logger.error({ error: e }, 'Error in SynthesizeStream');\n } finally {\n resourceCleanup();\n }\n }\n}\n"],"mappings":"AAIA,SAAS,iBAAiB;AAC1B,SAAS,UAAU,sBAAsB;AACzC,SAAS,uBAAuB;AAChC,SAAS,WAAW;AACpB,SAAS,2BAA2B;AACpC,SAAS,SAAS,qBAAqB;AAEvC,SAAS,oBAAoB,sBAAsB,OAAO,eAAe;AACzE,SAAiC,mCAAmC;AACpE,SAAS,iBAAiB;AAC1B;AAAA,EAIE;AAAA,EACA;AAAA,OACK;AACP,SAAyB,WAAW,yBAAyB;AAoD7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,eAAe;AACrB,MAAM,mBAAmB;AAiBlB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAyC,oBAAI,IAAI;AAAA,EAEzD,UAAU,IAAI;AAAA,EAEd,YAAY,MAUT;AACD,UAAM,cAAa,6BAAM,eAAc;AACvC,UAAM,YAAY,GAAG,EAAE,WAAW,KAAK,CAAC;AAExC,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA;AAAA,MACA,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;AAGA,QAAI,YAAY;AAChB,QAAI,YAAY;AAChB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,iBAAiB,UAAU,MAAM,MAAM,CAAC;AAC9C,YAAI,aAAa,cAAc,gBAAgB;AAC7C,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,OAAO,WAAW,OAAO,UAAU;AAAA,UACvC;AAAA,QACF,OAAO;AACL,sBAAY;AAAA,QACd;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AAEA,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,OAAO;AAAA,MACP;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,QAAI,YAAY,SAAS,GAAG,GAAG;AAC7B,YAAM,CAAC,OAAO,KAAK,IAAI,YAAY,MAAM,GAAG;AAC5C,aAAO,IAAI,IAAI,EAAE,OAAO,MAAM,CAAC;AAAA,IACjC;AACA,WAAO,IAAI,IAAI,EAAE,OAAO,YAAY,CAAC;AAAA,EACvC;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,WAAW,GAA0B;AACnC,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AAAA,EAEA,OAAO,SAAyE;AAC9E,UAAM,EAAE,cAAc,4BAA4B,IAAI,WAAW,CAAC;AAClE,UAAM,SAAS,IAAI,iBAAiB,MAAM,EAAE,GAAG,KAAK,KAAK,GAAG,WAAW;AACvE,SAAK,QAAQ,IAAI,MAAM;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AACnD,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,MAAM,kBAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS;AAAA,MACb,MAAM;AAAA,MACN,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,MACxC,UAAU,KAAK,KAAK;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,IACnB;AAEA,QAAI,KAAK,KAAK,MAAO,QAAO,QAAQ,KAAK,KAAK;AAC9C,QAAI,KAAK,KAAK,MAAO,QAAO,QAAQ,KAAK,KAAK;AAC9C,QAAI,KAAK,KAAK,SAAU,QAAO,WAAW,KAAK,KAAK;AAEpD,UAAM,SAAS,MAAM,UAAU,KAAK,SAAS,OAAO;AACpD,WAAO,KAAK,KAAK,UAAU,MAAM,CAAC;AAClC,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,QAAQ,IAAe;AAC3B,UAAM,GAAG,MAAM;AAAA,EACjB;AAAA,EAEA,MAAM,QAAQ;AACZ,eAAW,UAAU,KAAK,SAAS;AACjC,YAAM,OAAO,MAAM;AAAA,IACrB;AACA,SAAK,QAAQ,MAAM;AAAA,EACrB;AACF;AAEO,MAAM,yBAAmD,qBAAqB;AAAA,EAC3E;AAAA,EACA;AAAA,EACA;AAAA,EAER,UAAU,IAAI;AAAA,EAEd,YAAY,KAAkB,MAAmC,aAAgC;AAC/F,UAAM,KAAK,WAAW;AACtB,SAAK,OAAO;AACZ,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAAA,EACtC;AAAA,EAEA,MAAgB,MAAqB;AACnC,QAAI,KAAuB;AAC3B,QAAI,UAAU;AACd,QAAI,gBAAgB;AACpB,QAAI;AAEJ,UAAM,sBAAsB,IAAI,cAAc,kBAAkB,EAAE,OAAO;AACzE,UAAM,eAAe,oBAAoC;AACzD,UAAM,YAAY,UAAU,cAAc;AAE1C,UAAM,kBAAkB,MAAM;AAC5B,UAAI,QAAS;AACb,gBAAU;AACV,0BAAoB,MAAM;AAC1B,mBAAa,MAAM;AACnB,+BAAI;AACJ,+BAAI;AAAA,IACN;AAEA,UAAM,kBAAkB,OAAO,UAA0B;AACvD,YAAM,iBAAiB,MAAM,qBAAqB,WAAW,KAAK;AAClE,UAAI,CAAC,MAAM,GAAG,eAAe,UAAU,MAAM;AAC3C,aAAK,QAAQ,KAAK,uDAAuD;AACzE;AAAA,MACF;AACA,SAAG,KAAK,KAAK,UAAU,cAAc,CAAC;AAAA,IACxC;AAEA,UAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,UAAI,WAAW;AACb,aAAK,MAAM,IAAI,EAAE,WAAW,WAAW,OAAO,WAAW,MAAM,CAAC;AAChE,oBAAY;AAAA,MACd;AAAA,IACF;AAEA,UAAM,kBAAkB,YAAY;AAClC,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,8BAAoB,MAAM;AAC1B;AAAA,QACF;AACA,4BAAoB,SAAS,IAAI;AAAA,MACnC;AACA,0BAAoB,SAAS;AAAA,IAC/B;AAEA,UAAM,2BAA2B,YAAY;AAC3C,uBAAiB,MAAM,qBAAqB;AAC1C,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,wBAAgB;AAAA,UACd,MAAM;AAAA,UACN,YAAY,GAAG,QAAQ;AAAA,QACzB,CAAC;AAAA,MACH;AAEA,sBAAgB,EAAE,MAAM,gBAAgB,CAAC;AAAA,IAC3C;AAEA,UAAM,uBAAuB,OAAOA,QAAkB;AACpD,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,aAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,0BAAgB;AAChB,iBAAO,IAAI,MAAM,8BAA8B,CAAC;AAAA,QAClD,CAAC;AAED,QAAAA,IAAG,GAAG,WAAW,OAAO,SAAS;AAC/B,gBAAM,YAAY,KAAK,MAAM,KAAK,SAAS,CAAC;AAC5C,gBAAM,iBAAiB,qBAAqB,MAAM,SAAS;AAC3D,uBAAa,MAAM,cAAc;AAAA,QACnC,CAAC;AAED,QAAAA,IAAG,GAAG,SAAS,CAAC,MAAM;AACpB,eAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,0BAAgB;AAChB,iBAAO,CAAC;AAAA,QACV,CAAC;AAED,QAAAA,IAAG,GAAG,SAAS,MAAM;AACnB,0BAAgB;AAEhB,cAAI,CAAC,QAAS,QAAO,KAAK,QAAQ,MAAM,+BAA+B;AACvE,cAAI,cAAe,QAAO,QAAQ;AAElC;AAAA,YACE,IAAI,eAAe;AAAA,cACjB,SAAS;AAAA,cACT,SAAS,EAAE,UAAU;AAAA,YACvB,CAAC;AAAA,UACH;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,UAAM,iBAAiB,YAAY;AACjC,UAAI,mBAAkC;AAEtC,YAAM,UAAU,IAAI,gBAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,YAAM,oBAAoB,aAAa,OAAO;AAC9C,YAAM,SAAS,kBAAkB,UAAU;AAE3C,UAAI;AACF,eAAO,CAAC,KAAK,UAAU,CAAC,KAAK,gBAAgB,OAAO,SAAS;AAC3D,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,cAAI,OAAO,KAAM;AAEjB,gBAAM,cAAc,OAAO;AAC3B,kBAAQ,YAAY,MAAM;AAAA,YACxB,KAAK;AACH,iCAAmB,YAAY;AAC/B;AAAA,YACF,KAAK;AACH,oBAAM,aAAa,IAAI,UAAU,OAAO,KAAK,YAAY,OAAO,QAAQ,CAAC;AACzE,yBAAW,SAAS,QAAQ,MAAM,WAAW,MAAM,GAAG;AACpD,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA;AAAA,YACF,KAAK;AACH,8BAAgB;AAChB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA,4BAAc,kBAAmB,IAAI;AACrC,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAC7C;AAAA,YACF,KAAK;AACH,8BAAgB;AAChB;AAAA,YACF,KAAK;AACH,mBAAK,QAAQ;AAAA,gBACX,EAAE,YAAY;AAAA,gBACd;AAAA,cACF;AACA,8BAAgB;AAChB,oBAAM,IAAI,SAAS,+BAA+B,YAAY,OAAO,EAAE;AAAA,YACzE;AACE,mBAAK,QAAQ,KAAK,yBAAyB,WAAW;AACtD;AAAA,UACJ;AAAA,QACF;AAAA,MACF,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,kBAAkB,OAAO;AAAA,QACjC,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,QACxF;AAAA,MACF;AAAA,IACF;AAEA,QAAI;AACF,WAAK,MAAM,KAAK,IAAI,UAAU,KAAK,YAAY,SAAS;AAExD,YAAM,QAAQ,IAAI;AAAA,QAChB,gBAAgB;AAAA,QAChB,yBAAyB;AAAA,QACzB,qBAAqB,EAAE;AAAA,QACvB,eAAe;AAAA,MACjB,CAAC;AAAA,IACH,SAAS,GAAG;AACV,WAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,2BAA2B;AAAA,IAC9D,UAAE;AACA,sBAAgB;AAAA,IAClB;AAAA,EACF;AACF;","names":["ws"]}
1
+ {"version":3,"sources":["../../src/inference/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport { basic as tokenizeBasic } from '../tokenize/index.js';\nimport type { ChunkedStream } from '../tts/index.js';\nimport { SynthesizeStream as BaseSynthesizeStream, TTS as BaseTTS } from '../tts/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { shortuuid } from '../utils.js';\nimport {\n type TtsClientEvent,\n type TtsServerEvent,\n type TtsSessionCreateEvent,\n ttsClientEventSchema,\n ttsServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken } from './utils.js';\n\nexport type CartesiaModels =\n | 'cartesia'\n | 'cartesia/sonic'\n | 'cartesia/sonic-2'\n | 'cartesia/sonic-turbo';\n\nexport type ElevenlabsModels =\n | 'elevenlabs'\n | 'elevenlabs/eleven_flash_v2'\n | 'elevenlabs/eleven_flash_v2_5'\n | 'elevenlabs/eleven_turbo_v2'\n | 'elevenlabs/eleven_turbo_v2_5'\n | 'elevenlabs/eleven_multilingual_v2';\n\nexport type RimeModels = 'rime' | 'rime/mist' | 'rime/mistv2' | 'rime/arcana';\n\nexport type InworldModels = 'inworld' | 'inworld/inworld-tts-1';\n\nexport interface CartesiaOptions {\n duration?: number; // max duration of audio in seconds\n speed?: 'slow' | 'normal' | 'fast'; // default: not specified\n}\n\nexport interface ElevenlabsOptions {\n inactivity_timeout?: number; // default: 60\n apply_text_normalization?: 'auto' | 'off' | 'on'; // default: \"auto\"\n}\n\nexport interface RimeOptions {}\n\nexport interface InworldOptions {}\n\ntype _TTSModels = CartesiaModels | ElevenlabsModels | RimeModels | InworldModels;\n\nexport type TTSModels = CartesiaModels | ElevenlabsModels | RimeModels | InworldModels | AnyString;\n\nexport type ModelWithVoice = `${_TTSModels}:${string}` | TTSModels;\n\nexport type TTSOptions<TModel extends TTSModels> = TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends ElevenlabsModels\n ? ElevenlabsOptions\n : TModel extends RimeOptions\n ? RimeOptions\n : TModel extends InworldOptions\n ? InworldOptions\n : Record<string, unknown>;\n\ntype TTSEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: TTSEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'https://agent-gateway.livekit.cloud/v1';\nconst NUM_CHANNELS = 1;\nconst DEFAULT_LANGUAGE = 'en';\n\nexport interface InferenceTTSOptions<TModel extends TTSModels> {\n model?: TModel;\n voice?: string;\n language?: string;\n encoding: TTSEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: TTSOptions<TModel>;\n}\n\n/**\n * Livekit Cloud Inference TTS\n */\nexport class TTS<TModel extends TTSModels> extends BaseTTS {\n private opts: InferenceTTSOptions<TModel>;\n private streams: Set<SynthesizeStream<TModel>> = new Set();\n\n #logger = log();\n\n constructor(opts: {\n model: TModel;\n voice?: string;\n language?: string;\n baseURL?: string;\n encoding?: TTSEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: TTSOptions<TModel>;\n }) {\n const sampleRate = opts?.sampleRate ?? DEFAULT_SAMPLE_RATE;\n super(sampleRate, 1, { streaming: true });\n\n const {\n model,\n voice,\n language = DEFAULT_LANGUAGE,\n baseURL,\n encoding = DEFAULT_ENCODING,\n apiKey,\n apiSecret,\n modelOptions = {} as TTSOptions<TModel>,\n } = opts || {};\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n // read voice id from the model if provided: \"provider/model:voice_id\"\n let nextModel = model;\n let nextVoice = voice;\n if (typeof nextModel === 'string') {\n const idx = nextModel.lastIndexOf(':');\n if (idx !== -1) {\n const voiceFromModel = nextModel.slice(idx + 1);\n if (nextVoice && nextVoice !== voiceFromModel) {\n this.#logger.warn(\n '`voice` is provided via both argument and model, using the one from the argument',\n { voice: nextVoice, model: nextModel },\n );\n } else {\n nextVoice = voiceFromModel;\n }\n nextModel = nextModel.slice(0, idx) as TModel;\n }\n }\n\n this.opts = {\n model: nextModel,\n voice: nextVoice,\n language,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n };\n }\n\n get label() {\n return 'inference.TTS';\n }\n\n static fromModelString(modelString: string): TTS<AnyString> {\n if (modelString.includes(':')) {\n const [model, voice] = modelString.split(':') as [TTSModels, string];\n return new TTS({ model, voice });\n }\n return new TTS({ model: modelString });\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n synthesize(_: string): ChunkedStream {\n throw new Error('ChunkedStream is not implemented');\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream<TModel> {\n const { connOptions = DEFAULT_API_CONNECT_OPTIONS } = options || {};\n const stream = new SynthesizeStream(this, { ...this.opts }, connOptions);\n this.streams.add(stream);\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/tts`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const params = {\n type: 'session.create',\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n } as TtsSessionCreateEvent;\n\n if (this.opts.voice) params.voice = this.opts.voice;\n if (this.opts.model) params.model = this.opts.model;\n if (this.opts.language) params.language = this.opts.language;\n\n const socket = await connectWs(url, headers, timeout);\n socket.send(JSON.stringify(params));\n return socket;\n }\n\n async closeWs(ws: WebSocket) {\n await ws.close();\n }\n\n async close() {\n for (const stream of this.streams) {\n await stream.close();\n }\n this.streams.clear();\n }\n}\n\nexport class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeStream {\n private opts: InferenceTTSOptions<TModel>;\n private tts: TTS<TModel>;\n private connOptions: APIConnectOptions;\n\n #logger = log();\n\n constructor(tts: TTS<TModel>, opts: InferenceTTSOptions<TModel>, connOptions: APIConnectOptions) {\n super(tts, connOptions);\n this.opts = opts;\n this.tts = tts;\n this.connOptions = connOptions;\n }\n\n get label() {\n return 'inference.SynthesizeStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n }\n\n protected async run(): Promise<void> {\n let ws: WebSocket | null = null;\n let closing = false;\n let finalReceived = false;\n let lastFrame: AudioFrame | undefined;\n\n const sendTokenizerStream = new tokenizeBasic.SentenceTokenizer().stream();\n const eventChannel = createStreamChannel<TtsServerEvent>();\n const requestId = shortuuid('tts_request_');\n\n const resourceCleanup = () => {\n if (closing) return;\n closing = true;\n sendTokenizerStream.close();\n eventChannel.close();\n ws?.removeAllListeners();\n ws?.close();\n };\n\n const sendClientEvent = async (event: TtsClientEvent) => {\n // Don't send events to a closed WebSocket or aborted controller\n if (this.abortController.signal.aborted || closing) return;\n\n const validatedEvent = await ttsClientEventSchema.parseAsync(event);\n if (!ws || ws.readyState !== WebSocket.OPEN) {\n this.#logger.warn('Trying to send client TTS event to a closed WebSocket');\n return;\n }\n ws.send(JSON.stringify(validatedEvent));\n };\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const createInputTask = async () => {\n for await (const data of this.input) {\n if (this.abortController.signal.aborted) break;\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n sendTokenizerStream.flush();\n continue;\n }\n sendTokenizerStream.pushText(data);\n }\n sendTokenizerStream.endInput();\n };\n\n const createSentenceStreamTask = async () => {\n for await (const ev of sendTokenizerStream) {\n if (this.abortController.signal.aborted) break;\n\n sendClientEvent({\n type: 'input_transcript',\n transcript: ev.token + ' ',\n });\n }\n\n sendClientEvent({ type: 'session.flush' });\n };\n\n const createWsListenerTask = async (ws: WebSocket) => {\n return new Promise<void>((resolve, reject) => {\n this.abortController.signal.addEventListener('abort', () => {\n resourceCleanup();\n resolve(); // Abort is triggered by close(), which is a normal shutdown, not an error\n });\n\n ws.on('message', async (data) => {\n const eventJson = JSON.parse(data.toString()) as Record<string, unknown>;\n const validatedEvent = ttsServerEventSchema.parse(eventJson);\n eventChannel.write(validatedEvent);\n });\n\n ws.on('error', (e) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n resourceCleanup();\n reject(e);\n });\n\n ws.on('close', () => {\n resourceCleanup();\n\n if (!closing) return this.#logger.error('WebSocket closed unexpectedly');\n if (finalReceived) return resolve();\n\n reject(\n new APIStatusError({\n message: 'Gateway connection closed unexpectedly',\n options: { requestId },\n }),\n );\n });\n });\n };\n\n const createRecvTask = async () => {\n let currentSessionId: string | null = null;\n\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n while (!this.closed && !this.abortController.signal.aborted) {\n const result = await reader.read();\n if (this.abortController.signal.aborted) return;\n if (result.done) return;\n\n const serverEvent = result.value;\n switch (serverEvent.type) {\n case 'session.created':\n currentSessionId = serverEvent.session_id;\n break;\n case 'output_audio':\n const base64Data = new Int8Array(Buffer.from(serverEvent.audio, 'base64'));\n for (const frame of bstream.write(base64Data.buffer)) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n break;\n case 'done':\n finalReceived = true;\n for (const frame of bstream.flush()) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n sendLastFrame(currentSessionId!, true);\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n break;\n case 'session.closed':\n resourceCleanup();\n break;\n case 'error':\n this.#logger.error(\n { serverEvent },\n 'Received error message from LiveKit TTS WebSocket',\n );\n resourceCleanup();\n throw new APIError(`LiveKit TTS returned error: ${serverEvent.message}`);\n default:\n this.#logger.warn('Unexpected message %s', serverEvent);\n break;\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n ws = await this.tts.connectWs(this.connOptions.timeoutMs);\n\n await Promise.all([\n createInputTask(),\n createSentenceStreamTask(),\n createWsListenerTask(ws),\n createRecvTask(),\n ]);\n } finally {\n resourceCleanup();\n }\n }\n}\n"],"mappings":"AAIA,SAAS,iBAAiB;AAC1B,SAAS,UAAU,sBAAsB;AACzC,SAAS,uBAAuB;AAChC,SAAS,WAAW;AACpB,SAAS,2BAA2B;AACpC,SAAS,SAAS,qBAAqB;AAEvC,SAAS,oBAAoB,sBAAsB,OAAO,eAAe;AACzE,SAAiC,mCAAmC;AACpE,SAAS,iBAAiB;AAC1B;AAAA,EAIE;AAAA,EACA;AAAA,OACK;AACP,SAAyB,WAAW,yBAAyB;AAoD7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,eAAe;AACrB,MAAM,mBAAmB;AAiBlB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAyC,oBAAI,IAAI;AAAA,EAEzD,UAAU,IAAI;AAAA,EAEd,YAAY,MAUT;AACD,UAAM,cAAa,6BAAM,eAAc;AACvC,UAAM,YAAY,GAAG,EAAE,WAAW,KAAK,CAAC;AAExC,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA;AAAA,MACA,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;AAGA,QAAI,YAAY;AAChB,QAAI,YAAY;AAChB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,iBAAiB,UAAU,MAAM,MAAM,CAAC;AAC9C,YAAI,aAAa,cAAc,gBAAgB;AAC7C,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,OAAO,WAAW,OAAO,UAAU;AAAA,UACvC;AAAA,QACF,OAAO;AACL,sBAAY;AAAA,QACd;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AAEA,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,OAAO;AAAA,MACP;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,QAAI,YAAY,SAAS,GAAG,GAAG;AAC7B,YAAM,CAAC,OAAO,KAAK,IAAI,YAAY,MAAM,GAAG;AAC5C,aAAO,IAAI,IAAI,EAAE,OAAO,MAAM,CAAC;AAAA,IACjC;AACA,WAAO,IAAI,IAAI,EAAE,OAAO,YAAY,CAAC;AAAA,EACvC;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,WAAW,GAA0B;AACnC,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AAAA,EAEA,OAAO,SAAyE;AAC9E,UAAM,EAAE,cAAc,4BAA4B,IAAI,WAAW,CAAC;AAClE,UAAM,SAAS,IAAI,iBAAiB,MAAM,EAAE,GAAG,KAAK,KAAK,GAAG,WAAW;AACvE,SAAK,QAAQ,IAAI,MAAM;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AACnD,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,MAAM,kBAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS;AAAA,MACb,MAAM;AAAA,MACN,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,MACxC,UAAU,KAAK,KAAK;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,IACnB;AAEA,QAAI,KAAK,KAAK,MAAO,QAAO,QAAQ,KAAK,KAAK;AAC9C,QAAI,KAAK,KAAK,MAAO,QAAO,QAAQ,KAAK,KAAK;AAC9C,QAAI,KAAK,KAAK,SAAU,QAAO,WAAW,KAAK,KAAK;AAEpD,UAAM,SAAS,MAAM,UAAU,KAAK,SAAS,OAAO;AACpD,WAAO,KAAK,KAAK,UAAU,MAAM,CAAC;AAClC,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,QAAQ,IAAe;AAC3B,UAAM,GAAG,MAAM;AAAA,EACjB;AAAA,EAEA,MAAM,QAAQ;AACZ,eAAW,UAAU,KAAK,SAAS;AACjC,YAAM,OAAO,MAAM;AAAA,IACrB;AACA,SAAK,QAAQ,MAAM;AAAA,EACrB;AACF;AAEO,MAAM,yBAAmD,qBAAqB;AAAA,EAC3E;AAAA,EACA;AAAA,EACA;AAAA,EAER,UAAU,IAAI;AAAA,EAEd,YAAY,KAAkB,MAAmC,aAAgC;AAC/F,UAAM,KAAK,WAAW;AACtB,SAAK,OAAO;AACZ,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAAA,EACtC;AAAA,EAEA,MAAgB,MAAqB;AACnC,QAAI,KAAuB;AAC3B,QAAI,UAAU;AACd,QAAI,gBAAgB;AACpB,QAAI;AAEJ,UAAM,sBAAsB,IAAI,cAAc,kBAAkB,EAAE,OAAO;AACzE,UAAM,eAAe,oBAAoC;AACzD,UAAM,YAAY,UAAU,cAAc;AAE1C,UAAM,kBAAkB,MAAM;AAC5B,UAAI,QAAS;AACb,gBAAU;AACV,0BAAoB,MAAM;AAC1B,mBAAa,MAAM;AACnB,+BAAI;AACJ,+BAAI;AAAA,IACN;AAEA,UAAM,kBAAkB,OAAO,UAA0B;AAEvD,UAAI,KAAK,gBAAgB,OAAO,WAAW,QAAS;AAEpD,YAAM,iBAAiB,MAAM,qBAAqB,WAAW,KAAK;AAClE,UAAI,CAAC,MAAM,GAAG,eAAe,UAAU,MAAM;AAC3C,aAAK,QAAQ,KAAK,uDAAuD;AACzE;AAAA,MACF;AACA,SAAG,KAAK,KAAK,UAAU,cAAc,CAAC;AAAA,IACxC;AAEA,UAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,UAAI,WAAW;AACb,aAAK,MAAM,IAAI,EAAE,WAAW,WAAW,OAAO,WAAW,MAAM,CAAC;AAChE,oBAAY;AAAA,MACd;AAAA,IACF;AAEA,UAAM,kBAAkB,YAAY;AAClC,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,8BAAoB,MAAM;AAC1B;AAAA,QACF;AACA,4BAAoB,SAAS,IAAI;AAAA,MACnC;AACA,0BAAoB,SAAS;AAAA,IAC/B;AAEA,UAAM,2BAA2B,YAAY;AAC3C,uBAAiB,MAAM,qBAAqB;AAC1C,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,wBAAgB;AAAA,UACd,MAAM;AAAA,UACN,YAAY,GAAG,QAAQ;AAAA,QACzB,CAAC;AAAA,MACH;AAEA,sBAAgB,EAAE,MAAM,gBAAgB,CAAC;AAAA,IAC3C;AAEA,UAAM,uBAAuB,OAAOA,QAAkB;AACpD,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,aAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,0BAAgB;AAChB,kBAAQ;AAAA,QACV,CAAC;AAED,QAAAA,IAAG,GAAG,WAAW,OAAO,SAAS;AAC/B,gBAAM,YAAY,KAAK,MAAM,KAAK,SAAS,CAAC;AAC5C,gBAAM,iBAAiB,qBAAqB,MAAM,SAAS;AAC3D,uBAAa,MAAM,cAAc;AAAA,QACnC,CAAC;AAED,QAAAA,IAAG,GAAG,SAAS,CAAC,MAAM;AACpB,eAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,0BAAgB;AAChB,iBAAO,CAAC;AAAA,QACV,CAAC;AAED,QAAAA,IAAG,GAAG,SAAS,MAAM;AACnB,0BAAgB;AAEhB,cAAI,CAAC,QAAS,QAAO,KAAK,QAAQ,MAAM,+BAA+B;AACvE,cAAI,cAAe,QAAO,QAAQ;AAElC;AAAA,YACE,IAAI,eAAe;AAAA,cACjB,SAAS;AAAA,cACT,SAAS,EAAE,UAAU;AAAA,YACvB,CAAC;AAAA,UACH;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,UAAM,iBAAiB,YAAY;AACjC,UAAI,mBAAkC;AAEtC,YAAM,UAAU,IAAI,gBAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,YAAM,oBAAoB,aAAa,OAAO;AAC9C,YAAM,SAAS,kBAAkB,UAAU;AAE3C,UAAI;AACF,eAAO,CAAC,KAAK,UAAU,CAAC,KAAK,gBAAgB,OAAO,SAAS;AAC3D,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,cAAI,OAAO,KAAM;AAEjB,gBAAM,cAAc,OAAO;AAC3B,kBAAQ,YAAY,MAAM;AAAA,YACxB,KAAK;AACH,iCAAmB,YAAY;AAC/B;AAAA,YACF,KAAK;AACH,oBAAM,aAAa,IAAI,UAAU,OAAO,KAAK,YAAY,OAAO,QAAQ,CAAC;AACzE,yBAAW,SAAS,QAAQ,MAAM,WAAW,MAAM,GAAG;AACpD,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA;AAAA,YACF,KAAK;AACH,8BAAgB;AAChB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA,4BAAc,kBAAmB,IAAI;AACrC,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAC7C;AAAA,YACF,KAAK;AACH,8BAAgB;AAChB;AAAA,YACF,KAAK;AACH,mBAAK,QAAQ;AAAA,gBACX,EAAE,YAAY;AAAA,gBACd;AAAA,cACF;AACA,8BAAgB;AAChB,oBAAM,IAAI,SAAS,+BAA+B,YAAY,OAAO,EAAE;AAAA,YACzE;AACE,mBAAK,QAAQ,KAAK,yBAAyB,WAAW;AACtD;AAAA,UACJ;AAAA,QACF;AAAA,MACF,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,kBAAkB,OAAO;AAAA,QACjC,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,QACxF;AAAA,MACF;AAAA,IACF;AAEA,QAAI;AACF,WAAK,MAAM,KAAK,IAAI,UAAU,KAAK,YAAY,SAAS;AAExD,YAAM,QAAQ,IAAI;AAAA,QAChB,gBAAgB;AAAA,QAChB,yBAAyB;AAAA,QACzB,qBAAqB,EAAE;AAAA,QACvB,eAAe;AAAA,MACjB,CAAC;AAAA,IACH,UAAE;AACA,sBAAgB;AAAA,IAClB;AAAA,EACF;AACF;","names":["ws"]}
package/dist/stt/stt.cjs CHANGED
@@ -149,7 +149,18 @@ class SpeechStream {
149
149
  }
150
150
  async monitorMetrics() {
151
151
  for await (const event of this.queue) {
152
- this.output.put(event);
152
+ if (!this.output.closed) {
153
+ try {
154
+ this.output.put(event);
155
+ } catch (e) {
156
+ if (e instanceof Error && e.message.includes("Queue is closed")) {
157
+ this.logger.warn(
158
+ { err: e },
159
+ "Queue closed during transcript processing (expected during disconnect)"
160
+ );
161
+ }
162
+ }
163
+ }
153
164
  if (event.type !== 4 /* RECOGNITION_USAGE */) continue;
154
165
  const metrics = {
155
166
  type: "stt_metrics",
@@ -162,7 +173,9 @@ class SpeechStream {
162
173
  };
163
174
  this.#stt.emit("metrics_collected", metrics);
164
175
  }
165
- this.output.close();
176
+ if (!this.output.closed) {
177
+ this.output.close();
178
+ }
166
179
  }
167
180
  updateInputStream(audioStream) {
168
181
  this.deferredInputStream.setSource(audioStream);
@@ -217,9 +230,9 @@ class SpeechStream {
217
230
  }
218
231
  /** Close both the input and output of the STT stream */
219
232
  close() {
220
- this.input.close();
221
- this.queue.close();
222
- this.output.close();
233
+ if (!this.input.closed) this.input.close();
234
+ if (!this.queue.closed) this.queue.close();
235
+ if (!this.output.closed) this.output.close();
223
236
  this.closed = true;
224
237
  }
225
238
  [Symbol.asyncIterator]() {
@@ -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 } 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): Promise<SpeechEvent> {\n const startTime = process.hrtime.bigint();\n const event = await this._recognize(frame);\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 protected abstract _recognize(frame: AudioBuffer): Promise<SpeechEvent>;\n\n /**\n * Returns a {@link SpeechStream} that can be used to push audio frames and receive\n * transcriptions\n */\n abstract stream(): 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 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().then(() => 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 = this._connOptions._intervalForRetry(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 this.output.put(event);\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 this.output.close();\n }\n\n protected abstract run(): Promise<void>;\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 this.input.close();\n this.queue.close();\n this.output.close();\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,mBAAoE;AAEpE,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,OAA0C;AACxD,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,UAAM,QAAQ,MAAM,KAAK,WAAW,KAAK;AACzC,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,EASA,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,EAER,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,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;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,gBAAgB,KAAK,aAAa,kBAAkB,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,WAAK,OAAO,IAAI,KAAK;AACrB,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,SAAK,OAAO,MAAM;AAAA,EACpB;AAAA,EAIA,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,SAAK,MAAM,MAAM;AACjB,SAAK,MAAM,MAAM;AACjB,SAAK,OAAO,MAAM;AAClB,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 } 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): Promise<SpeechEvent> {\n const startTime = process.hrtime.bigint();\n const event = await this._recognize(frame);\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 protected abstract _recognize(frame: AudioBuffer): Promise<SpeechEvent>;\n\n /**\n * Returns a {@link SpeechStream} that can be used to push audio frames and receive\n * transcriptions\n */\n abstract stream(): 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 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().then(() => 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 = this._connOptions._intervalForRetry(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 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 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,mBAAoE;AAEpE,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,OAA0C;AACxD,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,UAAM,QAAQ,MAAM,KAAK,WAAW,KAAK;AACzC,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,EASA,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,EAER,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,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;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,gBAAgB,KAAK,aAAa,kBAAkB,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,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,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAkB;AACrC,WAAO;AAAA,EACT;AACF;","names":["SpeechEventType","frame"]}
@@ -1 +1 @@
1
- {"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../../src/stt/stt.ts"],"names":[],"mappings":";AAGA,OAAO,EAAE,KAAK,UAAU,EAAE,cAAc,EAAE,MAAM,mBAAmB,CAAC;AACpE,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAEhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AAItD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,oBAAoB,CAAC;AAErD,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AAClF,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,aAAa,CAAC;AAC/C,OAAO,EAAE,kBAAkB,EAA6B,MAAM,aAAa,CAAC;AAE5E,2CAA2C;AAC3C,oBAAY,eAAe;IACzB;;;;OAIG;IACH,eAAe,IAAI;IACnB;;OAEG;IACH,kBAAkB,IAAI;IACtB;;;OAGG;IACH,gBAAgB,IAAI;IACpB;;;OAGG;IACH,aAAa,IAAI;IACjB,mEAAmE;IACnE,iBAAiB,IAAI;IACrB;;;;OAIG;IACH,oBAAoB,IAAI;CACzB;AAED,mEAAmE;AACnE,MAAM,WAAW,UAAU;IACzB,QAAQ,EAAE,MAAM,CAAC;IACjB,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,EAAE,MAAM,CAAC;IAClB,OAAO,EAAE,MAAM,CAAC;IAChB,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,gBAAgB;IAC/B,aAAa,EAAE,MAAM,CAAC;CACvB;AAED,sDAAsD;AACtD,MAAM,WAAW,WAAW;IAC1B,IAAI,EAAE,eAAe,CAAC;IACtB,YAAY,CAAC,EAAE,CAAC,UAAU,EAAE,GAAG,UAAU,EAAE,CAAC,CAAC;IAC7C,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,gBAAgB,CAAC,EAAE,gBAAgB,CAAC;CACrC;AAED;;;;;GAKG;AACH,MAAM,WAAW,eAAe;IAC9B,SAAS,EAAE,OAAO,CAAC;IACnB,cAAc,EAAE,OAAO,CAAC;CACzB;AAED,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,WAAW,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;IAClB,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,KAAK,CAAC;IACb,WAAW,EAAE,OAAO,CAAC;CACtB;AAED,MAAM,MAAM,YAAY,GAAG;IACzB,CAAC,mBAAmB,CAAC,EAAE,CAAC,OAAO,EAAE,UAAU,KAAK,IAAI,CAAC;IACrD,CAAC,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,CAAC;CACtC,CAAC;kCAS2D,aAAa,YAAY,CAAC;AAPvF;;;;;;GAMG;AACH,8BAAsB,GAAI,SAAQ,QAAsD;;IACtF,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;gBAGX,YAAY,EAAE,eAAe;IAKzC,sCAAsC;IACtC,IAAI,YAAY,IAAI,eAAe,CAElC;IAED,8FAA8F;IACxF,SAAS,CAAC,KAAK,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;IAezD,SAAS,CAAC,QAAQ,CAAC,UAAU,CAAC,KAAK,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;IAEvE;;;OAGG;IACH,QAAQ,CAAC,MAAM,IAAI,YAAY;IAEzB,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;CAG7B;AAED;;;;;;;;;;;;;;;GAeG;AACH,8BAAsB,YAAa,YAAW,qBAAqB,CAAC,WAAW,CAAC;;IAC9E,SAAS,CAAC,MAAM,CAAC,QAAQ,CAAC,cAAc,gBAA4B;IACpE,SAAS,CAAC,KAAK,sEAA6E;IAC5F,SAAS,CAAC,MAAM,kCAAyC;IACzD,SAAS,CAAC,KAAK,kCAAyC;IACxD,SAAS,CAAC,gBAAgB,CAAC,EAAE,MAAM,CAAC;IACpC,SAAS,CAAC,SAAS,CAAC,EAAE,cAAc,CAAC;IACrC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,SAAS,CAAC,MAAM,UAAS;IAEzB,OAAO,CAAC,mBAAmB,CAAqC;IAChE,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,YAAY,CAAoB;gBAGtC,GAAG,EAAE,GAAG,EACR,UAAU,CAAC,EAAE,MAAM,EACnB,iBAAiB,GAAE,iBAA+C;YAgBtD,QAAQ;IAqCtB,OAAO,CAAC,SAAS;cAUD,SAAS;cAkBT,cAAc;IAkB9B,SAAS,CAAC,QAAQ,CAAC,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IAEvC,iBAAiB,CAAC,WAAW,EAAE,cAAc,CAAC,UAAU,CAAC;IAIzD,iBAAiB;IAIjB,qCAAqC;IACrC,SAAS,CAAC,KAAK,EAAE,UAAU;IAwB3B,4DAA4D;IAC5D,KAAK;IAUL,2DAA2D;IAC3D,QAAQ;IAUR,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,WAAW,CAAC,CAAC;IAI5C,wDAAwD;IACxD,KAAK;IAOL,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,YAAY;CAGvC"}
1
+ {"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../../src/stt/stt.ts"],"names":[],"mappings":";AAGA,OAAO,EAAE,KAAK,UAAU,EAAE,cAAc,EAAE,MAAM,mBAAmB,CAAC;AACpE,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAEhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AAItD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,oBAAoB,CAAC;AAErD,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AAClF,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,aAAa,CAAC;AAC/C,OAAO,EAAE,kBAAkB,EAA6B,MAAM,aAAa,CAAC;AAE5E,2CAA2C;AAC3C,oBAAY,eAAe;IACzB;;;;OAIG;IACH,eAAe,IAAI;IACnB;;OAEG;IACH,kBAAkB,IAAI;IACtB;;;OAGG;IACH,gBAAgB,IAAI;IACpB;;;OAGG;IACH,aAAa,IAAI;IACjB,mEAAmE;IACnE,iBAAiB,IAAI;IACrB;;;;OAIG;IACH,oBAAoB,IAAI;CACzB;AAED,mEAAmE;AACnE,MAAM,WAAW,UAAU;IACzB,QAAQ,EAAE,MAAM,CAAC;IACjB,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,EAAE,MAAM,CAAC;IAClB,OAAO,EAAE,MAAM,CAAC;IAChB,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,gBAAgB;IAC/B,aAAa,EAAE,MAAM,CAAC;CACvB;AAED,sDAAsD;AACtD,MAAM,WAAW,WAAW;IAC1B,IAAI,EAAE,eAAe,CAAC;IACtB,YAAY,CAAC,EAAE,CAAC,UAAU,EAAE,GAAG,UAAU,EAAE,CAAC,CAAC;IAC7C,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,gBAAgB,CAAC,EAAE,gBAAgB,CAAC;CACrC;AAED;;;;;GAKG;AACH,MAAM,WAAW,eAAe;IAC9B,SAAS,EAAE,OAAO,CAAC;IACnB,cAAc,EAAE,OAAO,CAAC;CACzB;AAED,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,WAAW,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;IAClB,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,KAAK,CAAC;IACb,WAAW,EAAE,OAAO,CAAC;CACtB;AAED,MAAM,MAAM,YAAY,GAAG;IACzB,CAAC,mBAAmB,CAAC,EAAE,CAAC,OAAO,EAAE,UAAU,KAAK,IAAI,CAAC;IACrD,CAAC,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,CAAC;CACtC,CAAC;kCAS2D,aAAa,YAAY,CAAC;AAPvF;;;;;;GAMG;AACH,8BAAsB,GAAI,SAAQ,QAAsD;;IACtF,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;gBAGX,YAAY,EAAE,eAAe;IAKzC,sCAAsC;IACtC,IAAI,YAAY,IAAI,eAAe,CAElC;IAED,8FAA8F;IACxF,SAAS,CAAC,KAAK,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;IAezD,SAAS,CAAC,QAAQ,CAAC,UAAU,CAAC,KAAK,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;IAEvE;;;OAGG;IACH,QAAQ,CAAC,MAAM,IAAI,YAAY;IAEzB,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;CAG7B;AAED;;;;;;;;;;;;;;;GAeG;AACH,8BAAsB,YAAa,YAAW,qBAAqB,CAAC,WAAW,CAAC;;IAC9E,SAAS,CAAC,MAAM,CAAC,QAAQ,CAAC,cAAc,gBAA4B;IACpE,SAAS,CAAC,KAAK,sEAA6E;IAC5F,SAAS,CAAC,MAAM,kCAAyC;IACzD,SAAS,CAAC,KAAK,kCAAyC;IACxD,SAAS,CAAC,gBAAgB,CAAC,EAAE,MAAM,CAAC;IACpC,SAAS,CAAC,SAAS,CAAC,EAAE,cAAc,CAAC;IACrC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,SAAS,CAAC,MAAM,UAAS;IAEzB,OAAO,CAAC,mBAAmB,CAAqC;IAChE,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,YAAY,CAAoB;gBAGtC,GAAG,EAAE,GAAG,EACR,UAAU,CAAC,EAAE,MAAM,EACnB,iBAAiB,GAAE,iBAA+C;YAgBtD,QAAQ;IAqCtB,OAAO,CAAC,SAAS;cAUD,SAAS;cAkBT,cAAc;IA+B9B,SAAS,CAAC,QAAQ,CAAC,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IAEvC,iBAAiB,CAAC,WAAW,EAAE,cAAc,CAAC,UAAU,CAAC;IAIzD,iBAAiB;IAIjB,qCAAqC;IACrC,SAAS,CAAC,KAAK,EAAE,UAAU;IAwB3B,4DAA4D;IAC5D,KAAK;IAUL,2DAA2D;IAC3D,QAAQ;IAUR,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,WAAW,CAAC,CAAC;IAI5C,wDAAwD;IACxD,KAAK;IAOL,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,YAAY;CAGvC"}
package/dist/stt/stt.js CHANGED
@@ -124,7 +124,18 @@ class SpeechStream {
124
124
  }
125
125
  async monitorMetrics() {
126
126
  for await (const event of this.queue) {
127
- this.output.put(event);
127
+ if (!this.output.closed) {
128
+ try {
129
+ this.output.put(event);
130
+ } catch (e) {
131
+ if (e instanceof Error && e.message.includes("Queue is closed")) {
132
+ this.logger.warn(
133
+ { err: e },
134
+ "Queue closed during transcript processing (expected during disconnect)"
135
+ );
136
+ }
137
+ }
138
+ }
128
139
  if (event.type !== 4 /* RECOGNITION_USAGE */) continue;
129
140
  const metrics = {
130
141
  type: "stt_metrics",
@@ -137,7 +148,9 @@ class SpeechStream {
137
148
  };
138
149
  this.#stt.emit("metrics_collected", metrics);
139
150
  }
140
- this.output.close();
151
+ if (!this.output.closed) {
152
+ this.output.close();
153
+ }
141
154
  }
142
155
  updateInputStream(audioStream) {
143
156
  this.deferredInputStream.setSource(audioStream);
@@ -192,9 +205,9 @@ class SpeechStream {
192
205
  }
193
206
  /** Close both the input and output of the STT stream */
194
207
  close() {
195
- this.input.close();
196
- this.queue.close();
197
- this.output.close();
208
+ if (!this.input.closed) this.input.close();
209
+ if (!this.queue.closed) this.queue.close();
210
+ if (!this.output.closed) this.output.close();
198
211
  this.closed = true;
199
212
  }
200
213
  [Symbol.asyncIterator]() {
@@ -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 } 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): Promise<SpeechEvent> {\n const startTime = process.hrtime.bigint();\n const event = await this._recognize(frame);\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 protected abstract _recognize(frame: AudioBuffer): Promise<SpeechEvent>;\n\n /**\n * Returns a {@link SpeechStream} that can be used to push audio frames and receive\n * transcriptions\n */\n abstract stream(): 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 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().then(() => 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 = this._connOptions._intervalForRetry(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 this.output.put(event);\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 this.output.close();\n }\n\n protected abstract run(): Promise<void>;\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 this.input.close();\n this.queue.close();\n this.output.close();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): SpeechStream {\n return this;\n }\n}\n"],"mappings":"AAGA,SAA0B,sBAAsB;AAEhD,SAAS,oBAAoB;AAE7B,SAAS,oBAAoB,gBAAgB;AAC7C,SAAS,qCAAqC;AAC9C,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAiC,mCAAmC;AAEpE,SAAS,oBAAoB,OAAO,WAAW,eAAe;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,aAAsD;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,OAA0C;AACxD,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,UAAM,QAAQ,MAAM,KAAK,WAAW,KAAK;AACzC,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,MAAM,8BAA8B,KAAK,IAAI,GAAI;AAAA,MACvE,UAAU;AAAA,IACZ,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EASA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAkBO,MAAe,aAA2D;AAAA,EAC/E,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,mBAAoE;AAAA,EAChF,SAAS,IAAI,mBAAgC;AAAA,EAC7C,QAAQ,IAAI,mBAAgC;AAAA,EAC5C;AAAA,EACA;AAAA,EAEA,SAAS;AAAA,EACnB;AAAA,EACQ;AAAA,EACA,SAAS,IAAI;AAAA,EACb;AAAA,EAER,YACE,KACA,YACA,oBAAuC,6BACvC;AACA,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,uBAAmC;AAClE,SAAK,mBAAmB;AACxB,SAAK,eAAe;AACpB,SAAK,UAAU;AAMf,cAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;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,UAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,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,mBAAmB;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,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,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,WAAK,OAAO,IAAI,KAAK;AACrB,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,SAAK,OAAO,MAAM;AAAA,EACpB;AAAA,EAIA,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,eAAe,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,SAAK,MAAM,MAAM;AACjB,SAAK,MAAM,MAAM;AACjB,SAAK,OAAO,MAAM;AAClB,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 } 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): Promise<SpeechEvent> {\n const startTime = process.hrtime.bigint();\n const event = await this._recognize(frame);\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 protected abstract _recognize(frame: AudioBuffer): Promise<SpeechEvent>;\n\n /**\n * Returns a {@link SpeechStream} that can be used to push audio frames and receive\n * transcriptions\n */\n abstract stream(): 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 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().then(() => 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 = this._connOptions._intervalForRetry(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 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 this.closed = true;\n }\n\n [Symbol.asyncIterator](): SpeechStream {\n return this;\n }\n}\n"],"mappings":"AAGA,SAA0B,sBAAsB;AAEhD,SAAS,oBAAoB;AAE7B,SAAS,oBAAoB,gBAAgB;AAC7C,SAAS,qCAAqC;AAC9C,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAiC,mCAAmC;AAEpE,SAAS,oBAAoB,OAAO,WAAW,eAAe;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,aAAsD;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,OAA0C;AACxD,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,UAAM,QAAQ,MAAM,KAAK,WAAW,KAAK;AACzC,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,MAAM,8BAA8B,KAAK,IAAI,GAAI;AAAA,MACvE,UAAU;AAAA,IACZ,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EASA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAkBO,MAAe,aAA2D;AAAA,EAC/E,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,mBAAoE;AAAA,EAChF,SAAS,IAAI,mBAAgC;AAAA,EAC7C,QAAQ,IAAI,mBAAgC;AAAA,EAC5C;AAAA,EACA;AAAA,EAEA,SAAS;AAAA,EACnB;AAAA,EACQ;AAAA,EACA,SAAS,IAAI;AAAA,EACb;AAAA,EAER,YACE,KACA,YACA,oBAAuC,6BACvC;AACA,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,uBAAmC;AAClE,SAAK,mBAAmB;AACxB,SAAK,eAAe;AACpB,SAAK,UAAU;AAMf,cAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;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,UAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,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,mBAAmB;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,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,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,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,eAAe,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,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAkB;AACrC,WAAO;AAAA,EACT;AACF;","names":["SpeechEventType","frame"]}
package/dist/tts/tts.cjs CHANGED
@@ -124,7 +124,16 @@ class SynthesizeStream {
124
124
  recoverable
125
125
  });
126
126
  }
127
- // TODO(AJS-37) Remove when refactoring TTS to use streams
127
+ // NOTE(AJS-37): The implementation below uses an AsyncIterableQueue (`this.input`)
128
+ // bridged from a DeferredReadableStream (`this.deferredInputStream`) rather than
129
+ // consuming the stream directly.
130
+ //
131
+ // A full refactor to native Web Streams was considered but is currently deferred.
132
+ // The primary reason is to maintain architectural parity with the Python SDK,
133
+ // which is a key design goal for the project. This ensures a consistent developer
134
+ // experience across both platforms.
135
+ //
136
+ // For more context, see the discussion in GitHub issue # 844.
128
137
  async pumpInput() {
129
138
  const reader = this.deferredInputStream.stream.getReader();
130
139
  try {
@@ -200,11 +209,8 @@ class SynthesizeStream {
200
209
  this.#monitorMetricsTask.finally(() => this.output.close());
201
210
  }
202
211
  this.#metricsText += text;
203
- if (this.input.closed) {
204
- throw new Error("Input is closed");
205
- }
206
- if (this.closed) {
207
- throw new Error("Stream is closed");
212
+ if (this.input.closed || this.closed) {
213
+ return;
208
214
  }
209
215
  this.input.put(text);
210
216
  }
@@ -214,22 +220,16 @@ class SynthesizeStream {
214
220
  this.#metricsPendingTexts.push(this.#metricsText);
215
221
  this.#metricsText = "";
216
222
  }
217
- if (this.input.closed) {
218
- throw new Error("Input is closed");
219
- }
220
- if (this.closed) {
221
- throw new Error("Stream is closed");
223
+ if (this.input.closed || this.closed) {
224
+ return;
222
225
  }
223
226
  this.input.put(SynthesizeStream.FLUSH_SENTINEL);
224
227
  }
225
228
  /** Mark the input as ended and forbid additional pushes */
226
229
  endInput() {
227
230
  this.flush();
228
- if (this.input.closed) {
229
- throw new Error("Input is closed");
230
- }
231
- if (this.closed) {
232
- throw new Error("Stream is closed");
231
+ if (this.input.closed || this.closed) {
232
+ return;
233
233
  }
234
234
  this.input.close();
235
235
  }