@livekit/agents-plugin-deepgram 1.4.9 → 1.4.11

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.
package/dist/stt.js CHANGED
@@ -6,6 +6,7 @@ import {
6
6
  createTimedString,
7
7
  getBaseLanguage,
8
8
  log,
9
+ mergeFrames,
9
10
  normalizeLanguage,
10
11
  stt,
11
12
  waitForAbort
@@ -29,6 +30,9 @@ const defaultSTTOptions = {
29
30
  keyterm: [],
30
31
  profanityFilter: false,
31
32
  redact: [],
33
+ utteranceEndMs: null,
34
+ replace: null,
35
+ search: null,
32
36
  dictation: false,
33
37
  diarize: false,
34
38
  numerals: false,
@@ -64,7 +68,59 @@ class STT extends stt.STT {
64
68
  };
65
69
  if (this.#opts.detectLanguage) {
66
70
  this.#opts.language = void 0;
67
- } else if (this.#opts.language && getBaseLanguage(this.#opts.language) !== "en" && [
71
+ } else {
72
+ this.#opts.model = this.#validateModel(this.#opts.model, this.#opts.language);
73
+ }
74
+ }
75
+ async _recognize(buffer, abortSignal) {
76
+ const frame = mergeFrames(buffer);
77
+ const listenURL = new URL(`${this.#opts.baseUrl.replace(/^ws/, "http")}/v1/listen`);
78
+ const params = {
79
+ model: this.#opts.model,
80
+ punctuate: this.#opts.punctuate,
81
+ detect_language: this.#opts.detectLanguage,
82
+ smart_format: this.#opts.smartFormat,
83
+ keywords: this.#opts.keywords.map((x) => x.join(":")),
84
+ profanity_filter: this.#opts.profanityFilter,
85
+ numerals: this.#opts.numerals,
86
+ mip_opt_out: this.#opts.mipOptOut,
87
+ language: this.#opts.language,
88
+ ...this.#opts.keyterm.length > 0 ? { keyterm: this.#opts.keyterm } : {},
89
+ ...this.#opts.redact.length > 0 ? { redact: this.#opts.redact } : {}
90
+ };
91
+ appendSearchParams(listenURL, params);
92
+ if (this.#opts.diarize) {
93
+ this.#logger.warn("speaker diarization is not supported in non-streaming mode, ignoring");
94
+ }
95
+ const resp = await fetch(listenURL, {
96
+ method: "POST",
97
+ headers: {
98
+ Authorization: `Token ${this.#opts.apiKey}`,
99
+ Accept: "application/json",
100
+ "Content-Type": "audio/wav"
101
+ },
102
+ body: new Blob([new Uint8Array(createWav(frame))], { type: "audio/wav" }),
103
+ signal: abortSignal ?? null
104
+ });
105
+ if (!resp.ok) {
106
+ throw new Error(
107
+ `Deepgram STT REST request failed with status ${resp.status}: ${await resp.text()}`
108
+ );
109
+ }
110
+ return prerecordedTranscriptionToSpeechEvent(this.#opts.language, await resp.json());
111
+ }
112
+ updateOptions(opts) {
113
+ const language = opts.language !== void 0 ? normalizeLanguage(opts.language) : this.#opts.language;
114
+ const model = opts.model !== void 0 || opts.language !== void 0 ? this.#validateModel(opts.model ?? this.#opts.model, language) : this.#opts.model;
115
+ this.#opts = {
116
+ ...this.#opts,
117
+ ...opts,
118
+ language,
119
+ model
120
+ };
121
+ }
122
+ #validateModel(model, language) {
123
+ if (language && getBaseLanguage(language) !== "en" && [
68
124
  "nova-2-meeting",
69
125
  "nova-2-phonecall",
70
126
  "nova-2-finance",
@@ -75,25 +131,20 @@ class STT extends stt.STT {
75
131
  "nova-2-drivethru",
76
132
  "nova-2-automotive",
77
133
  "nova-3-general"
78
- ].includes(this.#opts.model)) {
134
+ ].includes(model)) {
79
135
  this.#logger.warn(
80
- `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`
136
+ `${model} does not support language ${language}, falling back to nova-2-general`
81
137
  );
82
- this.#opts.model = "nova-2-general";
138
+ return "nova-2-general";
83
139
  }
84
- }
85
- // eslint-disable-next-line @typescript-eslint/no-unused-vars
86
- async _recognize(_) {
87
- throw new Error("Recognize is not supported on Deepgram STT");
88
- }
89
- updateOptions(opts) {
90
- this.#opts = {
91
- ...this.#opts,
92
- ...opts,
93
- language: opts.language !== void 0 ? normalizeLanguage(opts.language) : this.#opts.language
94
- };
140
+ return model;
95
141
  }
96
142
  stream(options) {
143
+ if (this.#opts.detectLanguage || this.#opts.language === void 0) {
144
+ throw new Error(
145
+ "language detection is not supported in streaming mode, please disable it and specify a language"
146
+ );
147
+ }
97
148
  return new SpeechStream(this, this.#opts, options == null ? void 0 : options.connOptions);
98
149
  }
99
150
  async close() {
@@ -143,19 +194,14 @@ class SpeechStream extends stt.SpeechStream {
143
194
  keywords: this.#opts.keywords.map((x) => x.join(":")),
144
195
  keyterm: this.#opts.keyterm,
145
196
  profanity_filter: this.#opts.profanityFilter,
146
- redact: this.#opts.redact,
197
+ utterance_end_ms: this.#opts.utteranceEndMs,
198
+ replace: this.#opts.replace,
199
+ search: this.#opts.search,
147
200
  language: this.#opts.language,
148
- mip_opt_out: this.#opts.mipOptOut
201
+ mip_opt_out: this.#opts.mipOptOut,
202
+ ...this.#opts.redact.length > 0 ? { redact: this.#opts.redact } : {}
149
203
  };
150
- Object.entries(params).forEach(([k, v]) => {
151
- if (v !== void 0) {
152
- if (typeof v === "string" || typeof v === "number" || typeof v === "boolean") {
153
- streamURL.searchParams.append(k, String(v));
154
- } else {
155
- v.forEach((x) => streamURL.searchParams.append(k, String(x)));
156
- }
157
- }
158
- });
204
+ appendSearchParams(streamURL, params);
159
205
  ws = new WebSocket(streamURL, {
160
206
  headers: { Authorization: `Token ${this.#opts.apiKey}` }
161
207
  });
@@ -313,6 +359,13 @@ class SpeechStream extends stt.SpeechStream {
313
359
  case "Metadata": {
314
360
  break;
315
361
  }
362
+ case "UtteranceEnd": {
363
+ if (this.#speaking) {
364
+ this.#speaking = false;
365
+ putMessage({ type: stt.SpeechEventType.END_OF_SPEECH });
366
+ }
367
+ break;
368
+ }
316
369
  default: {
317
370
  this.#logger.child({ msg: json }).warn("received unexpected message from Deepgram");
318
371
  break;
@@ -348,6 +401,41 @@ class SpeechStream extends stt.SpeechStream {
348
401
  this.queue.put(usageEvent);
349
402
  }
350
403
  }
404
+ function appendSearchParams(url, params) {
405
+ Object.entries(params).forEach(([key, value]) => {
406
+ if (value === void 0 || value === null) return;
407
+ if (Array.isArray(value)) {
408
+ value.forEach((item) => url.searchParams.append(key, String(item)));
409
+ } else if (typeof value === "object") {
410
+ Object.entries(value).forEach(([term, replacement]) => {
411
+ url.searchParams.append(key, `${term}:${replacement}`);
412
+ });
413
+ } else {
414
+ url.searchParams.append(key, String(value));
415
+ }
416
+ });
417
+ }
418
+ function createWav(frame) {
419
+ const bitsPerSample = 16;
420
+ const byteRate = frame.sampleRate * frame.channels * bitsPerSample / 8;
421
+ const blockAlign = frame.channels * bitsPerSample / 8;
422
+ const header = Buffer.alloc(44);
423
+ header.write("RIFF", 0);
424
+ header.writeUInt32LE(36 + frame.data.byteLength, 4);
425
+ header.write("WAVE", 8);
426
+ header.write("fmt ", 12);
427
+ header.writeUInt32LE(16, 16);
428
+ header.writeUInt16LE(1, 20);
429
+ header.writeUInt16LE(frame.channels, 22);
430
+ header.writeUInt32LE(frame.sampleRate, 24);
431
+ header.writeUInt32LE(byteRate, 28);
432
+ header.writeUInt16LE(blockAlign, 32);
433
+ header.writeUInt16LE(bitsPerSample, 34);
434
+ header.write("data", 36);
435
+ header.writeUInt32LE(frame.data.byteLength, 40);
436
+ const pcm = Buffer.from(frame.data.buffer, frame.data.byteOffset, frame.data.byteLength);
437
+ return Buffer.concat([header, pcm]);
438
+ }
351
439
  const liveTranscriptionToSpeechData = (language, data, startTimeOffset = 0) => {
352
440
  const alts = data["channel"]["alternatives"];
353
441
  return alts.map((alt) => {
@@ -372,6 +460,33 @@ const liveTranscriptionToSpeechData = (language, data, startTimeOffset = 0) => {
372
460
  };
373
461
  });
374
462
  };
463
+ function prerecordedTranscriptionToSpeechEvent(language, data) {
464
+ const channel = data["results"]["channels"][0];
465
+ const detectedLanguage = channel["detected_language"] ?? "";
466
+ const alternatives = channel["alternatives"].map((alt) => {
467
+ const wordsData = alt["words"] ?? [];
468
+ return {
469
+ language: normalizeLanguage(language ?? detectedLanguage),
470
+ startTime: wordsData.length ? wordsData[0]["start"] : 0,
471
+ endTime: wordsData.length ? wordsData[wordsData.length - 1]["end"] : 0,
472
+ confidence: alt["confidence"],
473
+ text: alt["transcript"],
474
+ words: wordsData.map(
475
+ (word) => createTimedString({
476
+ text: word["word"] ?? "",
477
+ startTime: word["start"] ?? 0,
478
+ endTime: word["end"] ?? 0,
479
+ confidence: word["confidence"] ?? 0
480
+ })
481
+ )
482
+ };
483
+ });
484
+ return {
485
+ type: stt.SpeechEventType.FINAL_TRANSCRIPT,
486
+ requestId: data["metadata"]["request_id"],
487
+ ...alternatives.length ? { alternatives } : {}
488
+ };
489
+ }
375
490
  export {
376
491
  STT,
377
492
  SpeechStream
package/dist/stt.js.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n type AudioBuffer,\n AudioByteStream,\n AudioEnergyFilter,\n Future,\n Task,\n createTimedString,\n getBaseLanguage,\n log,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { STTLanguages, STTModels } from './models.js';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n /**\n * Redact sensitive information from the transcription. Accepts a single value or list of\n * values. Supported values: \"pci\", \"numbers\", \"ssn\", \"true\" (redact all).\n * See https://developers.deepgram.com/docs/redaction for details.\n */\n redact: string | string[];\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n mipOptOut: boolean;\n baseUrl: string;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.DEEPGRAM_API_KEY,\n language: 'en-US',\n detectLanguage: false,\n interimResults: true,\n punctuate: true,\n model: 'nova-3',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n redact: [],\n dictation: false,\n diarize: false,\n numerals: false,\n mipOptOut: false,\n baseUrl: 'wss://api.deepgram.com',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n label = 'deepgram.STT';\n private abortController = new AbortController();\n\n get model(): string {\n return this.#opts.model;\n }\n\n get provider(): string {\n return 'Deepgram';\n }\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n alignedTranscript: 'word',\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n language: opts.language ? normalizeLanguage(opts.language) : defaultSTTOptions.language,\n };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n getBaseLanguage(this.#opts.language) !== 'en' &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n );\n this.#opts.model = 'nova-2-general';\n }\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Recognize is not supported on Deepgram STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = {\n ...this.#opts,\n ...opts,\n language:\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language,\n };\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n return new SpeechStream(this, this.#opts, options?.connOptions);\n }\n\n async close() {\n this.abortController.abort();\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #audioEnergyFilter: AudioEnergyFilter;\n #logger = log();\n #speaking = false;\n #resetWS = new Future();\n #requestId = '';\n #audioDurationCollector: PeriodicCollector<number>;\n label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(stt, opts.sampleRate, connOptions);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n let ws: WebSocket;\n\n while (!this.input.closed && !this.closed) {\n const streamURL = new URL(`${this.#opts.baseUrl}/v1/listen`);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n redact: this.#opts.redact,\n language: this.#opts.language,\n mip_opt_out: this.#opts.mipOptOut,\n };\n Object.entries(params).forEach(([k, v]) => {\n if (v !== undefined) {\n if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n streamURL.searchParams.append(k, String(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, String(x)));\n }\n }\n });\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (!this.closed && !this.input.closed) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n }\n\n const delay = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n } else {\n this.#logger.warn(\n `Deepgram disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n );\n }\n }\n }\n\n this.closed = true;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n this.#resetWS.resolve();\n }\n\n async #runWS(ws: WebSocket) {\n this.#resetWS = new Future();\n let closing = false;\n\n const keepalive = setInterval(() => {\n try {\n ws.send(JSON.stringify({ type: 'KeepAlive' }));\n } catch {\n clearInterval(keepalive);\n return;\n }\n }, 5000);\n\n // gets cancelled also when sendTask is complete\n const wsMonitor = Task.from(async (controller) => {\n const closed = new Promise<void>(async (_, reject) => {\n ws.once('close', (code, reason) => {\n if (!closing) {\n this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n reject(new Error('WebSocket closed'));\n }\n });\n });\n\n await Promise.race([closed, waitForAbort(controller.signal)]);\n });\n\n const sendTask = async () => {\n const samples100Ms = Math.floor(this.#opts.sampleRate / 10);\n const stream = new AudioByteStream(\n this.#opts.sampleRate,\n this.#opts.numChannels,\n samples100Ms,\n );\n\n // waitForAbort internally sets up an abort listener on the abort signal\n // we need to put it outside loop to avoid constant re-registration of the listener\n const abortPromise = waitForAbort(this.abortSignal);\n\n try {\n while (!this.closed) {\n const result = await Promise.race([this.input.next(), abortPromise]);\n\n if (result === undefined) return; // aborted\n if (result.done) {\n break;\n }\n\n const data = result.value;\n\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = stream.flush();\n this.#audioDurationCollector.flush();\n } else if (\n data.sampleRate === this.#opts.sampleRate &&\n data.channels === this.#opts.numChannels\n ) {\n frames = stream.write(data.data.buffer as ArrayBuffer);\n } else {\n throw new Error(`sample rate or channel count of frame does not match`);\n }\n\n for await (const frame of frames) {\n if (this.#audioEnergyFilter.pushFrame(frame)) {\n const frameDuration = frame.samplesPerChannel / frame.sampleRate;\n this.#audioDurationCollector.push(frameDuration);\n ws.send(frame.data.buffer);\n }\n }\n }\n } finally {\n closing = true;\n ws.send(JSON.stringify({ type: 'CloseStream' }));\n wsMonitor.cancel();\n }\n };\n\n const listenTask = Task.from(async (controller) => {\n const putMessage = (message: stt.SpeechEvent) => {\n if (!this.queue.closed) {\n try {\n this.queue.put(message);\n } catch (e) {\n // ignore\n }\n }\n };\n\n const listenMessage = new Promise<void>((resolve, reject) => {\n ws.on('message', (msg) => {\n try {\n const json = JSON.parse(msg.toString());\n switch (json['type']) {\n case 'SpeechStarted': {\n // This is a normal case. Deepgram's SpeechStarted events\n // are not correlated with speech_final or utterance end.\n // It's possible that we receive two in a row without an endpoint\n // It's also possible we receive a transcript without a SpeechStarted event.\n if (this.#speaking) return;\n this.#speaking = true;\n putMessage({ type: stt.SpeechEventType.START_OF_SPEECH });\n break;\n }\n // see this page:\n // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final\n // for more information about the different types of events\n case 'Results': {\n const metadata = json['metadata'];\n const requestId = metadata['request_id'];\n const isFinal = json['is_final'];\n const isEndpoint = json['speech_final'];\n this.#requestId = requestId;\n\n const alternatives = liveTranscriptionToSpeechData(\n this.#opts.language!,\n json,\n this.startTimeOffset,\n );\n\n // If, for some reason, we didn't get a SpeechStarted event but we got\n // a transcript with text, we should start speaking. It's rare but has\n // been observed.\n if (alternatives[0] && alternatives[0].text) {\n if (!this.#speaking) {\n this.#speaking = true;\n putMessage({\n type: stt.SpeechEventType.START_OF_SPEECH,\n });\n }\n\n if (isFinal) {\n putMessage({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n } else {\n putMessage({\n type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n }\n }\n\n // if we receive an endpoint, only end the speech if\n // we either had a SpeechStarted event or we have a seen\n // a non-empty transcript (deepgram doesn't have a SpeechEnded event)\n if (isEndpoint && this.#speaking) {\n this.#speaking = false;\n putMessage({ type: stt.SpeechEventType.END_OF_SPEECH });\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n\n if (this.closed || closing) {\n resolve();\n }\n } catch (err) {\n this.#logger.error(`STT: Error processing message: ${msg}`);\n reject(err);\n }\n });\n });\n\n await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n }, this.abortController);\n\n await Promise.race([\n this.#resetWS.await,\n Promise.all([sendTask(), listenTask.result, wsMonitor]),\n ]);\n closing = true;\n ws.close();\n clearInterval(keepalive);\n }\n\n private onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n startTimeOffset: number = 0,\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => {\n const wordsData: any[] = alt['words'] ?? [];\n const detectedLanguage =\n language === 'multi' ? alt['languages']?.[0] ?? language : alt['language'] ?? language;\n\n return {\n language: normalizeLanguage(detectedLanguage),\n startTime: wordsData.length ? wordsData[0]['start'] + startTimeOffset : startTimeOffset,\n endTime: wordsData.length\n ? wordsData[wordsData.length - 1]['end'] + startTimeOffset\n : startTimeOffset,\n confidence: alt['confidence'],\n text: alt['transcript'],\n words: wordsData.map((word) =>\n createTimedString({\n text: word['word'] ?? '',\n startTime: (word['start'] ?? 0) + startTimeOffset,\n endTime: (word['end'] ?? 0) + startTimeOffset,\n confidence: word['confidence'] ?? 0.0,\n startTimeOffset,\n }),\n ),\n };\n });\n};\n"],"mappings":"AAGA;AAAA,EAGE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,iBAAiB;AAC1B,SAAS,yBAAyB;AAgClC,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,QAAQ,CAAC;AAAA,EACT,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AAAA,EACV,WAAW;AAAA,EACX,SAAS;AACX;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU,IAAI;AAAA,EACd,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAE9C,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,MACzD,mBAAmB;AAAA,IACrB,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,KAAK,WAAW,kBAAkB,KAAK,QAAQ,IAAI,kBAAkB;AAAA,IACjF;AAEA,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,gBAAgB,KAAK,MAAM,QAAQ,MAAM,QACzC;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,SAAY,kBAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAAA,IAChF;AAAA,EACF;AAAA,EAEA,OAAO,SAA6D;AAClE,WAAO,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AAAA,EAChE;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AACF;AAEO,MAAM,qBAAqB,IAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,UAAU,IAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,OAAO;AAAA,EACtB,aAAa;AAAA,EACb;AAAA,EACA,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,KAAK,YAAY,WAAW;AACvC,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,kBAAkB;AAChD,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,sBAAsB,QAAQ;AAAA,MACjD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AACd,QAAI;AAEJ,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,YAAM,YAAY,IAAI,IAAI,GAAG,KAAK,MAAM,OAAO,YAAY;AAC3D,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,QAAQ,KAAK,MAAM;AAAA,QACnB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM;AAAA,MAC1B;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,OAAO,CAAC,CAAC;AAAA,UAC5C,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,OAAO,CAAC,CAAC,CAAC;AAAA,UAC9D;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,UAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,CAAC,KAAK,UAAU,CAAC,KAAK,MAAM,QAAQ;AACtC,cAAI,WAAW,UAAU;AACvB,kBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,UACjF;AAEA,gBAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,eAAK,QAAQ;AAAA,YACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,UAC3F;AACA,gBAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,QAClE,OAAO;AACL,eAAK,QAAQ;AAAA,YACX,gDAAgD,CAAC,kBAAkB,KAAK,MAAM,MAAM,eAAe,KAAK,MAAM;AAAA,UAChH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,SAAK,WAAW,IAAI,OAAO;AAC3B,QAAI,UAAU;AAEd,UAAM,YAAY,YAAY,MAAM;AAClC,UAAI;AACF,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC,CAAC;AAAA,MAC/C,QAAQ;AACN,sBAAc,SAAS;AACvB;AAAA,MACF;AAAA,IACF,GAAG,GAAI;AAGP,UAAM,YAAY,KAAK,KAAK,OAAO,eAAe;AAChD,YAAM,SAAS,IAAI,QAAc,OAAO,GAAG,WAAW;AACpD,WAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,cAAI,CAAC,SAAS;AACZ,iBAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,mBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,UACtC;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,QAAQ,aAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IAC9D,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,YAAM,eAAe,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AAC1D,YAAM,SAAS,IAAI;AAAA,QACjB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX;AAAA,MACF;AAIA,YAAM,eAAe,aAAa,KAAK,WAAW;AAElD,UAAI;AACF,eAAO,CAAC,KAAK,QAAQ;AACnB,gBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,KAAK,MAAM,KAAK,GAAG,YAAY,CAAC;AAEnE,cAAI,WAAW,OAAW;AAC1B,cAAI,OAAO,MAAM;AACf;AAAA,UACF;AAEA,gBAAM,OAAO,OAAO;AAEpB,cAAI;AACJ,cAAI,SAAS,aAAa,gBAAgB;AACxC,qBAAS,OAAO,MAAM;AACtB,iBAAK,wBAAwB,MAAM;AAAA,UACrC,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,qBAAS,OAAO,MAAM,KAAK,KAAK,MAAqB;AAAA,UACvD,OAAO;AACL,kBAAM,IAAI,MAAM,sDAAsD;AAAA,UACxE;AAEA,2BAAiB,SAAS,QAAQ;AAChC,gBAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,oBAAM,gBAAgB,MAAM,oBAAoB,MAAM;AACtD,mBAAK,wBAAwB,KAAK,aAAa;AAC/C,iBAAG,KAAK,MAAM,KAAK,MAAM;AAAA,YAC3B;AAAA,UACF;AAAA,QACF;AAAA,MACF,UAAE;AACA,kBAAU;AACV,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAC/C,kBAAU,OAAO;AAAA,MACnB;AAAA,IACF;AAEA,UAAM,aAAa,KAAK,KAAK,OAAO,eAAe;AACjD,YAAM,aAAa,CAAC,YAA6B;AAC/C,YAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,cAAI;AACF,iBAAK,MAAM,IAAI,OAAO;AAAA,UACxB,SAAS,GAAG;AAAA,UAEZ;AAAA,QACF;AAAA,MACF;AAEA,YAAM,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3D,WAAG,GAAG,WAAW,CAAC,QAAQ;AACxB,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,oBAAQ,KAAK,MAAM,GAAG;AAAA,cACpB,KAAK,iBAAiB;AAKpB,oBAAI,KAAK,UAAW;AACpB,qBAAK,YAAY;AACjB,2BAAW,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AACxD;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,WAAW,KAAK,UAAU;AAChC,sBAAM,YAAY,SAAS,YAAY;AACvC,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AACtC,qBAAK,aAAa;AAElB,sBAAM,eAAe;AAAA,kBACnB,KAAK,MAAM;AAAA,kBACX;AAAA,kBACA,KAAK;AAAA,gBACP;AAKA,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,+BAAW;AAAA,sBACT,MAAM,IAAI,gBAAgB;AAAA,oBAC5B,CAAC;AAAA,kBACH;AAEA,sBAAI,SAAS;AACX,+BAAW;AAAA,sBACT,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,+BAAW;AAAA,sBACT,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH;AAAA,gBACF;AAKA,oBAAI,cAAc,KAAK,WAAW;AAChC,uBAAK,YAAY;AACjB,6BAAW,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AAAA,gBACxD;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAEA,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,kCAAkC,GAAG,EAAE;AAC1D,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,eAAe,aAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IACrE,GAAG,KAAK,eAAe;AAEvB,UAAM,QAAQ,KAAK;AAAA,MACjB,KAAK,SAAS;AAAA,MACd,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,QAAQ,SAAS,CAAC;AAAA,IACxD,CAAC;AACD,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AAAA,EAEQ,sBAAsB,UAAkB;AAC9C,UAAM,aAA8B;AAAA,MAClC,MAAM,IAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,MACA,kBAA0B,MACL;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,QAAQ;AA5c3B;AA6cI,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAC1C,UAAM,mBACJ,aAAa,YAAU,SAAI,WAAW,MAAf,mBAAmB,OAAM,WAAW,IAAI,UAAU,KAAK;AAEhF,WAAO;AAAA,MACL,UAAU,kBAAkB,gBAAgB;AAAA,MAC5C,WAAW,UAAU,SAAS,UAAU,CAAC,EAAE,OAAO,IAAI,kBAAkB;AAAA,MACxE,SAAS,UAAU,SACf,UAAU,UAAU,SAAS,CAAC,EAAE,KAAK,IAAI,kBACzC;AAAA,MACJ,YAAY,IAAI,YAAY;AAAA,MAC5B,MAAM,IAAI,YAAY;AAAA,MACtB,OAAO,UAAU;AAAA,QAAI,CAAC,SACpB,kBAAkB;AAAA,UAChB,MAAM,KAAK,MAAM,KAAK;AAAA,UACtB,YAAY,KAAK,OAAO,KAAK,KAAK;AAAA,UAClC,UAAU,KAAK,KAAK,KAAK,KAAK;AAAA,UAC9B,YAAY,KAAK,YAAY,KAAK;AAAA,UAClC;AAAA,QACF,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF,CAAC;AACH;","names":["stt"]}
1
+ {"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n type AudioBuffer,\n AudioByteStream,\n AudioEnergyFilter,\n Future,\n Task,\n createTimedString,\n getBaseLanguage,\n log,\n mergeFrames,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { STTLanguages, STTModels } from './models.js';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels | string;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n /**\n * Redact sensitive information from the transcription. Accepts a single value or list of\n * values. Supported values: \"pci\", \"numbers\", \"ssn\", \"true\" (redact all).\n * See https://developers.deepgram.com/docs/redaction for details.\n */\n redact: string | string[];\n utteranceEndMs?: number | null;\n replace?: Record<string, string> | null;\n search?: string[] | null;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n mipOptOut: boolean;\n baseUrl: string;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.DEEPGRAM_API_KEY,\n language: 'en-US',\n detectLanguage: false,\n interimResults: true,\n punctuate: true,\n model: 'nova-3',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n redact: [],\n utteranceEndMs: null,\n replace: null,\n search: null,\n dictation: false,\n diarize: false,\n numerals: false,\n mipOptOut: false,\n baseUrl: 'wss://api.deepgram.com',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n label = 'deepgram.STT';\n private abortController = new AbortController();\n\n get model(): string {\n return this.#opts.model;\n }\n\n get provider(): string {\n return 'Deepgram';\n }\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n alignedTranscript: 'word',\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n language: opts.language ? normalizeLanguage(opts.language) : defaultSTTOptions.language,\n };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else {\n this.#opts.model = this.#validateModel(this.#opts.model, this.#opts.language);\n }\n }\n\n async _recognize(buffer: AudioBuffer, abortSignal?: AbortSignal): Promise<stt.SpeechEvent> {\n const frame = mergeFrames(buffer);\n const listenURL = new URL(`${this.#opts.baseUrl.replace(/^ws/, 'http')}/v1/listen`);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n detect_language: this.#opts.detectLanguage,\n smart_format: this.#opts.smartFormat,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n profanity_filter: this.#opts.profanityFilter,\n numerals: this.#opts.numerals,\n mip_opt_out: this.#opts.mipOptOut,\n language: this.#opts.language,\n ...(this.#opts.keyterm.length > 0 ? { keyterm: this.#opts.keyterm } : {}),\n ...(this.#opts.redact.length > 0 ? { redact: this.#opts.redact } : {}),\n };\n\n appendSearchParams(listenURL, params);\n\n if (this.#opts.diarize) {\n this.#logger.warn('speaker diarization is not supported in non-streaming mode, ignoring');\n }\n\n const resp = await fetch(listenURL, {\n method: 'POST',\n headers: {\n Authorization: `Token ${this.#opts.apiKey}`,\n Accept: 'application/json',\n 'Content-Type': 'audio/wav',\n },\n body: new Blob([new Uint8Array(createWav(frame))], { type: 'audio/wav' }),\n signal: abortSignal ?? null,\n });\n\n if (!resp.ok) {\n throw new Error(\n `Deepgram STT REST request failed with status ${resp.status}: ${await resp.text()}`,\n );\n }\n\n return prerecordedTranscriptionToSpeechEvent(this.#opts.language, await resp.json());\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n const language =\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language;\n const model =\n opts.model !== undefined || opts.language !== undefined\n ? this.#validateModel(opts.model ?? this.#opts.model, language)\n : this.#opts.model;\n\n this.#opts = {\n ...this.#opts,\n ...opts,\n language,\n model,\n };\n }\n\n #validateModel(model: STTModels | string, language?: string) {\n if (\n language &&\n getBaseLanguage(language) !== 'en' &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(model)\n ) {\n this.#logger.warn(\n `${model} does not support language ${language}, falling back to nova-2-general`,\n );\n return 'nova-2-general';\n }\n\n return model;\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n // Deepgram only supports language detection for prerecorded audio, not streaming.\n if (this.#opts.detectLanguage || this.#opts.language === undefined) {\n throw new Error(\n 'language detection is not supported in streaming mode, please disable it and specify a language',\n );\n }\n return new SpeechStream(this, this.#opts, options?.connOptions);\n }\n\n async close() {\n this.abortController.abort();\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #audioEnergyFilter: AudioEnergyFilter;\n #logger = log();\n #speaking = false;\n #resetWS = new Future();\n #requestId = '';\n #audioDurationCollector: PeriodicCollector<number>;\n label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(stt, opts.sampleRate, connOptions);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n let ws: WebSocket;\n\n while (!this.input.closed && !this.closed) {\n const streamURL = new URL(`${this.#opts.baseUrl}/v1/listen`);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n utterance_end_ms: this.#opts.utteranceEndMs,\n replace: this.#opts.replace,\n search: this.#opts.search,\n language: this.#opts.language,\n mip_opt_out: this.#opts.mipOptOut,\n ...(this.#opts.redact.length > 0 ? { redact: this.#opts.redact } : {}),\n };\n appendSearchParams(streamURL, params);\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (!this.closed && !this.input.closed) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n }\n\n const delay = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n } else {\n this.#logger.warn(\n `Deepgram disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n );\n }\n }\n }\n\n this.closed = true;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n this.#resetWS.resolve();\n }\n\n async #runWS(ws: WebSocket) {\n this.#resetWS = new Future();\n let closing = false;\n\n const keepalive = setInterval(() => {\n try {\n ws.send(JSON.stringify({ type: 'KeepAlive' }));\n } catch {\n clearInterval(keepalive);\n return;\n }\n }, 5000);\n\n // gets cancelled also when sendTask is complete\n const wsMonitor = Task.from(async (controller) => {\n const closed = new Promise<void>(async (_, reject) => {\n ws.once('close', (code, reason) => {\n if (!closing) {\n this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n reject(new Error('WebSocket closed'));\n }\n });\n });\n\n await Promise.race([closed, waitForAbort(controller.signal)]);\n });\n\n const sendTask = async () => {\n const samples100Ms = Math.floor(this.#opts.sampleRate / 10);\n const stream = new AudioByteStream(\n this.#opts.sampleRate,\n this.#opts.numChannels,\n samples100Ms,\n );\n\n // waitForAbort internally sets up an abort listener on the abort signal\n // we need to put it outside loop to avoid constant re-registration of the listener\n const abortPromise = waitForAbort(this.abortSignal);\n\n try {\n while (!this.closed) {\n const result = await Promise.race([this.input.next(), abortPromise]);\n\n if (result === undefined) return; // aborted\n if (result.done) {\n break;\n }\n\n const data = result.value;\n\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = stream.flush();\n this.#audioDurationCollector.flush();\n } else if (\n data.sampleRate === this.#opts.sampleRate &&\n data.channels === this.#opts.numChannels\n ) {\n frames = stream.write(data.data.buffer as ArrayBuffer);\n } else {\n throw new Error(`sample rate or channel count of frame does not match`);\n }\n\n for await (const frame of frames) {\n if (this.#audioEnergyFilter.pushFrame(frame)) {\n const frameDuration = frame.samplesPerChannel / frame.sampleRate;\n this.#audioDurationCollector.push(frameDuration);\n ws.send(frame.data.buffer);\n }\n }\n }\n } finally {\n closing = true;\n ws.send(JSON.stringify({ type: 'CloseStream' }));\n wsMonitor.cancel();\n }\n };\n\n const listenTask = Task.from(async (controller) => {\n const putMessage = (message: stt.SpeechEvent) => {\n if (!this.queue.closed) {\n try {\n this.queue.put(message);\n } catch (e) {\n // ignore\n }\n }\n };\n\n const listenMessage = new Promise<void>((resolve, reject) => {\n ws.on('message', (msg) => {\n try {\n const json = JSON.parse(msg.toString());\n switch (json['type']) {\n case 'SpeechStarted': {\n // This is a normal case. Deepgram's SpeechStarted events\n // are not correlated with speech_final or utterance end.\n // It's possible that we receive two in a row without an endpoint\n // It's also possible we receive a transcript without a SpeechStarted event.\n if (this.#speaking) return;\n this.#speaking = true;\n putMessage({ type: stt.SpeechEventType.START_OF_SPEECH });\n break;\n }\n // see this page:\n // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final\n // for more information about the different types of events\n case 'Results': {\n const metadata = json['metadata'];\n const requestId = metadata['request_id'];\n const isFinal = json['is_final'];\n const isEndpoint = json['speech_final'];\n this.#requestId = requestId;\n\n const alternatives = liveTranscriptionToSpeechData(\n this.#opts.language!,\n json,\n this.startTimeOffset,\n );\n\n // If, for some reason, we didn't get a SpeechStarted event but we got\n // a transcript with text, we should start speaking. It's rare but has\n // been observed.\n if (alternatives[0] && alternatives[0].text) {\n if (!this.#speaking) {\n this.#speaking = true;\n putMessage({\n type: stt.SpeechEventType.START_OF_SPEECH,\n });\n }\n\n if (isFinal) {\n putMessage({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n } else {\n putMessage({\n type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n }\n }\n\n // if we receive an endpoint, only end the speech if\n // we either had a SpeechStarted event or we have a seen\n // a non-empty transcript (deepgram doesn't have a SpeechEnded event)\n if (isEndpoint && this.#speaking) {\n this.#speaking = false;\n putMessage({ type: stt.SpeechEventType.END_OF_SPEECH });\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n case 'UtteranceEnd': {\n if (this.#speaking) {\n this.#speaking = false;\n putMessage({ type: stt.SpeechEventType.END_OF_SPEECH });\n }\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n\n if (this.closed || closing) {\n resolve();\n }\n } catch (err) {\n this.#logger.error(`STT: Error processing message: ${msg}`);\n reject(err);\n }\n });\n });\n\n await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n }, this.abortController);\n\n await Promise.race([\n this.#resetWS.await,\n Promise.all([sendTask(), listenTask.result, wsMonitor]),\n ]);\n closing = true;\n ws.close();\n clearInterval(keepalive);\n }\n\n private onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n}\n\nfunction appendSearchParams(url: URL, params: Record<string, unknown>) {\n Object.entries(params).forEach(([key, value]) => {\n if (value === undefined || value === null) return;\n\n if (Array.isArray(value)) {\n value.forEach((item) => url.searchParams.append(key, String(item)));\n } else if (typeof value === 'object') {\n Object.entries(value).forEach(([term, replacement]) => {\n url.searchParams.append(key, `${term}:${replacement}`);\n });\n } else {\n url.searchParams.append(key, String(value));\n }\n });\n}\n\nfunction createWav(frame: AudioFrame): Buffer {\n const bitsPerSample = 16;\n const byteRate = (frame.sampleRate * frame.channels * bitsPerSample) / 8;\n const blockAlign = (frame.channels * bitsPerSample) / 8;\n\n const header = Buffer.alloc(44);\n header.write('RIFF', 0);\n header.writeUInt32LE(36 + frame.data.byteLength, 4);\n header.write('WAVE', 8);\n header.write('fmt ', 12);\n header.writeUInt32LE(16, 16);\n header.writeUInt16LE(1, 20);\n header.writeUInt16LE(frame.channels, 22);\n header.writeUInt32LE(frame.sampleRate, 24);\n header.writeUInt32LE(byteRate, 28);\n header.writeUInt16LE(blockAlign, 32);\n header.writeUInt16LE(bitsPerSample, 34);\n header.write('data', 36);\n header.writeUInt32LE(frame.data.byteLength, 40);\n\n const pcm = Buffer.from(frame.data.buffer, frame.data.byteOffset, frame.data.byteLength);\n return Buffer.concat([header, pcm]);\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n startTimeOffset: number = 0,\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => {\n const wordsData: any[] = alt['words'] ?? [];\n const detectedLanguage =\n language === 'multi' ? alt['languages']?.[0] ?? language : alt['language'] ?? language;\n\n return {\n language: normalizeLanguage(detectedLanguage),\n startTime: wordsData.length ? wordsData[0]['start'] + startTimeOffset : startTimeOffset,\n endTime: wordsData.length\n ? wordsData[wordsData.length - 1]['end'] + startTimeOffset\n : startTimeOffset,\n confidence: alt['confidence'],\n text: alt['transcript'],\n words: wordsData.map((word) =>\n createTimedString({\n text: word['word'] ?? '',\n startTime: (word['start'] ?? 0) + startTimeOffset,\n endTime: (word['end'] ?? 0) + startTimeOffset,\n confidence: word['confidence'] ?? 0.0,\n startTimeOffset,\n }),\n ),\n };\n });\n};\n\nfunction prerecordedTranscriptionToSpeechEvent(\n language: STTLanguages | string | undefined,\n data: { [id: string]: any },\n): stt.SpeechEvent {\n const channel = data['results']['channels'][0];\n const detectedLanguage = channel['detected_language'] ?? '';\n const alternatives = channel['alternatives'].map((alt: { [id: string]: any }) => {\n const wordsData: any[] = alt['words'] ?? [];\n\n return {\n language: normalizeLanguage(language ?? detectedLanguage),\n startTime: wordsData.length ? wordsData[0]['start'] : 0,\n endTime: wordsData.length ? wordsData[wordsData.length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n words: wordsData.map((word) =>\n createTimedString({\n text: word['word'] ?? '',\n startTime: word['start'] ?? 0,\n endTime: word['end'] ?? 0,\n confidence: word['confidence'] ?? 0.0,\n }),\n ),\n };\n }) as stt.SpeechData[];\n\n return {\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n requestId: data['metadata']['request_id'],\n ...(alternatives.length\n ? { alternatives: alternatives as [stt.SpeechData, ...stt.SpeechData[]] }\n : {}),\n };\n}\n"],"mappings":"AAGA;AAAA,EAGE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,iBAAiB;AAC1B,SAAS,yBAAyB;AAmClC,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,QAAQ,CAAC;AAAA,EACT,gBAAgB;AAAA,EAChB,SAAS;AAAA,EACT,QAAQ;AAAA,EACR,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AAAA,EACV,WAAW;AAAA,EACX,SAAS;AACX;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU,IAAI;AAAA,EACd,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAE9C,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,MACzD,mBAAmB;AAAA,IACrB,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,KAAK,WAAW,kBAAkB,KAAK,QAAQ,IAAI,kBAAkB;AAAA,IACjF;AAEA,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,OAAO;AACL,WAAK,MAAM,QAAQ,KAAK,eAAe,KAAK,MAAM,OAAO,KAAK,MAAM,QAAQ;AAAA,IAC9E;AAAA,EACF;AAAA,EAEA,MAAM,WAAW,QAAqB,aAAqD;AACzF,UAAM,QAAQ,YAAY,MAAM;AAChC,UAAM,YAAY,IAAI,IAAI,GAAG,KAAK,MAAM,QAAQ,QAAQ,OAAO,MAAM,CAAC,YAAY;AAClF,UAAM,SAAS;AAAA,MACb,OAAO,KAAK,MAAM;AAAA,MAClB,WAAW,KAAK,MAAM;AAAA,MACtB,iBAAiB,KAAK,MAAM;AAAA,MAC5B,cAAc,KAAK,MAAM;AAAA,MACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,MACpD,kBAAkB,KAAK,MAAM;AAAA,MAC7B,UAAU,KAAK,MAAM;AAAA,MACrB,aAAa,KAAK,MAAM;AAAA,MACxB,UAAU,KAAK,MAAM;AAAA,MACrB,GAAI,KAAK,MAAM,QAAQ,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,IAAI,CAAC;AAAA,MACvE,GAAI,KAAK,MAAM,OAAO,SAAS,IAAI,EAAE,QAAQ,KAAK,MAAM,OAAO,IAAI,CAAC;AAAA,IACtE;AAEA,uBAAmB,WAAW,MAAM;AAEpC,QAAI,KAAK,MAAM,SAAS;AACtB,WAAK,QAAQ,KAAK,sEAAsE;AAAA,IAC1F;AAEA,UAAM,OAAO,MAAM,MAAM,WAAW;AAAA,MAClC,QAAQ;AAAA,MACR,SAAS;AAAA,QACP,eAAe,SAAS,KAAK,MAAM,MAAM;AAAA,QACzC,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA,MAAM,IAAI,KAAK,CAAC,IAAI,WAAW,UAAU,KAAK,CAAC,CAAC,GAAG,EAAE,MAAM,YAAY,CAAC;AAAA,MACxE,QAAQ,eAAe;AAAA,IACzB,CAAC;AAED,QAAI,CAAC,KAAK,IAAI;AACZ,YAAM,IAAI;AAAA,QACR,gDAAgD,KAAK,MAAM,KAAK,MAAM,KAAK,KAAK,CAAC;AAAA,MACnF;AAAA,IACF;AAEA,WAAO,sCAAsC,KAAK,MAAM,UAAU,MAAM,KAAK,KAAK,CAAC;AAAA,EACrF;AAAA,EAEA,cAAc,MAA2B;AACvC,UAAM,WACJ,KAAK,aAAa,SAAY,kBAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAC9E,UAAM,QACJ,KAAK,UAAU,UAAa,KAAK,aAAa,SAC1C,KAAK,eAAe,KAAK,SAAS,KAAK,MAAM,OAAO,QAAQ,IAC5D,KAAK,MAAM;AAEjB,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,eAAe,OAA2B,UAAmB;AAC3D,QACE,YACA,gBAAgB,QAAQ,MAAM,QAC9B;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,GAChB;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,8BAA8B,QAAQ;AAAA,MAChD;AACA,aAAO;AAAA,IACT;AAEA,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,SAA6D;AAElE,QAAI,KAAK,MAAM,kBAAkB,KAAK,MAAM,aAAa,QAAW;AAClE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,WAAO,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AAAA,EAChE;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AACF;AAEO,MAAM,qBAAqB,IAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,UAAU,IAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,OAAO;AAAA,EACtB,aAAa;AAAA,EACb;AAAA,EACA,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,KAAK,YAAY,WAAW;AACvC,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,kBAAkB;AAChD,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,sBAAsB,QAAQ;AAAA,MACjD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AACd,QAAI;AAEJ,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,YAAM,YAAY,IAAI,IAAI,GAAG,KAAK,MAAM,OAAO,YAAY;AAC3D,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,kBAAkB,KAAK,MAAM;AAAA,QAC7B,SAAS,KAAK,MAAM;AAAA,QACpB,QAAQ,KAAK,MAAM;AAAA,QACnB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM;AAAA,QACxB,GAAI,KAAK,MAAM,OAAO,SAAS,IAAI,EAAE,QAAQ,KAAK,MAAM,OAAO,IAAI,CAAC;AAAA,MACtE;AACA,yBAAmB,WAAW,MAAM;AAEpC,WAAK,IAAI,UAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,CAAC,KAAK,UAAU,CAAC,KAAK,MAAM,QAAQ;AACtC,cAAI,WAAW,UAAU;AACvB,kBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,UACjF;AAEA,gBAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,eAAK,QAAQ;AAAA,YACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,UAC3F;AACA,gBAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,QAClE,OAAO;AACL,eAAK,QAAQ;AAAA,YACX,gDAAgD,CAAC,kBAAkB,KAAK,MAAM,MAAM,eAAe,KAAK,MAAM;AAAA,UAChH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,SAAK,WAAW,IAAI,OAAO;AAC3B,QAAI,UAAU;AAEd,UAAM,YAAY,YAAY,MAAM;AAClC,UAAI;AACF,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC,CAAC;AAAA,MAC/C,QAAQ;AACN,sBAAc,SAAS;AACvB;AAAA,MACF;AAAA,IACF,GAAG,GAAI;AAGP,UAAM,YAAY,KAAK,KAAK,OAAO,eAAe;AAChD,YAAM,SAAS,IAAI,QAAc,OAAO,GAAG,WAAW;AACpD,WAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,cAAI,CAAC,SAAS;AACZ,iBAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,mBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,UACtC;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,QAAQ,aAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IAC9D,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,YAAM,eAAe,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AAC1D,YAAM,SAAS,IAAI;AAAA,QACjB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX;AAAA,MACF;AAIA,YAAM,eAAe,aAAa,KAAK,WAAW;AAElD,UAAI;AACF,eAAO,CAAC,KAAK,QAAQ;AACnB,gBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,KAAK,MAAM,KAAK,GAAG,YAAY,CAAC;AAEnE,cAAI,WAAW,OAAW;AAC1B,cAAI,OAAO,MAAM;AACf;AAAA,UACF;AAEA,gBAAM,OAAO,OAAO;AAEpB,cAAI;AACJ,cAAI,SAAS,aAAa,gBAAgB;AACxC,qBAAS,OAAO,MAAM;AACtB,iBAAK,wBAAwB,MAAM;AAAA,UACrC,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,qBAAS,OAAO,MAAM,KAAK,KAAK,MAAqB;AAAA,UACvD,OAAO;AACL,kBAAM,IAAI,MAAM,sDAAsD;AAAA,UACxE;AAEA,2BAAiB,SAAS,QAAQ;AAChC,gBAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,oBAAM,gBAAgB,MAAM,oBAAoB,MAAM;AACtD,mBAAK,wBAAwB,KAAK,aAAa;AAC/C,iBAAG,KAAK,MAAM,KAAK,MAAM;AAAA,YAC3B;AAAA,UACF;AAAA,QACF;AAAA,MACF,UAAE;AACA,kBAAU;AACV,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAC/C,kBAAU,OAAO;AAAA,MACnB;AAAA,IACF;AAEA,UAAM,aAAa,KAAK,KAAK,OAAO,eAAe;AACjD,YAAM,aAAa,CAAC,YAA6B;AAC/C,YAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,cAAI;AACF,iBAAK,MAAM,IAAI,OAAO;AAAA,UACxB,SAAS,GAAG;AAAA,UAEZ;AAAA,QACF;AAAA,MACF;AAEA,YAAM,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3D,WAAG,GAAG,WAAW,CAAC,QAAQ;AACxB,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,oBAAQ,KAAK,MAAM,GAAG;AAAA,cACpB,KAAK,iBAAiB;AAKpB,oBAAI,KAAK,UAAW;AACpB,qBAAK,YAAY;AACjB,2BAAW,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AACxD;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,WAAW,KAAK,UAAU;AAChC,sBAAM,YAAY,SAAS,YAAY;AACvC,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AACtC,qBAAK,aAAa;AAElB,sBAAM,eAAe;AAAA,kBACnB,KAAK,MAAM;AAAA,kBACX;AAAA,kBACA,KAAK;AAAA,gBACP;AAKA,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,+BAAW;AAAA,sBACT,MAAM,IAAI,gBAAgB;AAAA,oBAC5B,CAAC;AAAA,kBACH;AAEA,sBAAI,SAAS;AACX,+BAAW;AAAA,sBACT,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,+BAAW;AAAA,sBACT,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH;AAAA,gBACF;AAKA,oBAAI,cAAc,KAAK,WAAW;AAChC,uBAAK,YAAY;AACjB,6BAAW,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AAAA,gBACxD;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,KAAK,gBAAgB;AACnB,oBAAI,KAAK,WAAW;AAClB,uBAAK,YAAY;AACjB,6BAAW,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AAAA,gBACxD;AACA;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAEA,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,kCAAkC,GAAG,EAAE;AAC1D,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,eAAe,aAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IACrE,GAAG,KAAK,eAAe;AAEvB,UAAM,QAAQ,KAAK;AAAA,MACjB,KAAK,SAAS;AAAA,MACd,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,QAAQ,SAAS,CAAC;AAAA,IACxD,CAAC;AACD,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AAAA,EAEQ,sBAAsB,UAAkB;AAC9C,UAAM,aAA8B;AAAA,MAClC,MAAM,IAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AACF;AAEA,SAAS,mBAAmB,KAAU,QAAiC;AACrE,SAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,KAAK,KAAK,MAAM;AAC/C,QAAI,UAAU,UAAa,UAAU,KAAM;AAE3C,QAAI,MAAM,QAAQ,KAAK,GAAG;AACxB,YAAM,QAAQ,CAAC,SAAS,IAAI,aAAa,OAAO,KAAK,OAAO,IAAI,CAAC,CAAC;AAAA,IACpE,WAAW,OAAO,UAAU,UAAU;AACpC,aAAO,QAAQ,KAAK,EAAE,QAAQ,CAAC,CAAC,MAAM,WAAW,MAAM;AACrD,YAAI,aAAa,OAAO,KAAK,GAAG,IAAI,IAAI,WAAW,EAAE;AAAA,MACvD,CAAC;AAAA,IACH,OAAO;AACL,UAAI,aAAa,OAAO,KAAK,OAAO,KAAK,CAAC;AAAA,IAC5C;AAAA,EACF,CAAC;AACH;AAEA,SAAS,UAAU,OAA2B;AAC5C,QAAM,gBAAgB;AACtB,QAAM,WAAY,MAAM,aAAa,MAAM,WAAW,gBAAiB;AACvE,QAAM,aAAc,MAAM,WAAW,gBAAiB;AAEtD,QAAM,SAAS,OAAO,MAAM,EAAE;AAC9B,SAAO,MAAM,QAAQ,CAAC;AACtB,SAAO,cAAc,KAAK,MAAM,KAAK,YAAY,CAAC;AAClD,SAAO,MAAM,QAAQ,CAAC;AACtB,SAAO,MAAM,QAAQ,EAAE;AACvB,SAAO,cAAc,IAAI,EAAE;AAC3B,SAAO,cAAc,GAAG,EAAE;AAC1B,SAAO,cAAc,MAAM,UAAU,EAAE;AACvC,SAAO,cAAc,MAAM,YAAY,EAAE;AACzC,SAAO,cAAc,UAAU,EAAE;AACjC,SAAO,cAAc,YAAY,EAAE;AACnC,SAAO,cAAc,eAAe,EAAE;AACtC,SAAO,MAAM,QAAQ,EAAE;AACvB,SAAO,cAAc,MAAM,KAAK,YAAY,EAAE;AAE9C,QAAM,MAAM,OAAO,KAAK,MAAM,KAAK,QAAQ,MAAM,KAAK,YAAY,MAAM,KAAK,UAAU;AACvF,SAAO,OAAO,OAAO,CAAC,QAAQ,GAAG,CAAC;AACpC;AAEA,MAAM,gCAAgC,CACpC,UACA,MACA,kBAA0B,MACL;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,QAAQ;AAxjB3B;AAyjBI,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAC1C,UAAM,mBACJ,aAAa,YAAU,SAAI,WAAW,MAAf,mBAAmB,OAAM,WAAW,IAAI,UAAU,KAAK;AAEhF,WAAO;AAAA,MACL,UAAU,kBAAkB,gBAAgB;AAAA,MAC5C,WAAW,UAAU,SAAS,UAAU,CAAC,EAAE,OAAO,IAAI,kBAAkB;AAAA,MACxE,SAAS,UAAU,SACf,UAAU,UAAU,SAAS,CAAC,EAAE,KAAK,IAAI,kBACzC;AAAA,MACJ,YAAY,IAAI,YAAY;AAAA,MAC5B,MAAM,IAAI,YAAY;AAAA,MACtB,OAAO,UAAU;AAAA,QAAI,CAAC,SACpB,kBAAkB;AAAA,UAChB,MAAM,KAAK,MAAM,KAAK;AAAA,UACtB,YAAY,KAAK,OAAO,KAAK,KAAK;AAAA,UAClC,UAAU,KAAK,KAAK,KAAK,KAAK;AAAA,UAC9B,YAAY,KAAK,YAAY,KAAK;AAAA,UAClC;AAAA,QACF,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF,CAAC;AACH;AAEA,SAAS,sCACP,UACA,MACiB;AACjB,QAAM,UAAU,KAAK,SAAS,EAAE,UAAU,EAAE,CAAC;AAC7C,QAAM,mBAAmB,QAAQ,mBAAmB,KAAK;AACzD,QAAM,eAAe,QAAQ,cAAc,EAAE,IAAI,CAAC,QAA+B;AAC/E,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAE1C,WAAO;AAAA,MACL,UAAU,kBAAkB,YAAY,gBAAgB;AAAA,MACxD,WAAW,UAAU,SAAS,UAAU,CAAC,EAAE,OAAO,IAAI;AAAA,MACtD,SAAS,UAAU,SAAS,UAAU,UAAU,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,MACrE,YAAY,IAAI,YAAY;AAAA,MAC5B,MAAM,IAAI,YAAY;AAAA,MACtB,OAAO,UAAU;AAAA,QAAI,CAAC,SACpB,kBAAkB;AAAA,UAChB,MAAM,KAAK,MAAM,KAAK;AAAA,UACtB,WAAW,KAAK,OAAO,KAAK;AAAA,UAC5B,SAAS,KAAK,KAAK,KAAK;AAAA,UACxB,YAAY,KAAK,YAAY,KAAK;AAAA,QACpC,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF,CAAC;AAED,SAAO;AAAA,IACL,MAAM,IAAI,gBAAgB;AAAA,IAC1B,WAAW,KAAK,UAAU,EAAE,YAAY;AAAA,IACxC,GAAI,aAAa,SACb,EAAE,aAAoE,IACtE,CAAC;AAAA,EACP;AACF;","names":["stt"]}
package/dist/stt.test.cjs CHANGED
@@ -4,6 +4,17 @@ var import_agents_plugins_test = require("@livekit/agents-plugins-test");
4
4
  var import_vitest = require("vitest");
5
5
  var import_stt = require("./stt.cjs");
6
6
  const hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);
7
+ (0, import_vitest.describe)("Deepgram streaming language detection", () => {
8
+ (0, import_vitest.it)("throws when starting a stream with detectLanguage enabled", () => {
9
+ const stt2 = new import_stt.STT({ apiKey: "test", detectLanguage: true });
10
+ (0, import_vitest.expect)(() => stt2.stream()).toThrow("language detection is not supported in streaming mode");
11
+ });
12
+ (0, import_vitest.it)("allows streaming with an explicit language", () => {
13
+ const stt2 = new import_stt.STT({ apiKey: "test", language: "en-US" });
14
+ const stream = stt2.stream();
15
+ stream.close();
16
+ });
17
+ });
7
18
  if (hasDeepgramApiKey) {
8
19
  (0, import_vitest.describe)("Deepgram", async () => {
9
20
  await (0, import_agents_plugins_test.stt)(new import_stt.STT(), await import_agents_plugin_silero.VAD.load(), { nonStreaming: false });
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { describe, it } from 'vitest';\nimport { STT } from './stt.js';\n\nconst hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);\n\nif (hasDeepgramApiKey) {\n describe('Deepgram', async () => {\n await stt(new STT(), await VAD.load(), { nonStreaming: false });\n });\n} else {\n describe('Deepgram', () => {\n it.skip('requires DEEPGRAM_API_KEY', () => {});\n });\n}\n"],"mappings":";AAGA,kCAAoB;AACpB,iCAAoB;AACpB,oBAA6B;AAC7B,iBAAoB;AAEpB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,gBAAgB;AAE9D,IAAI,mBAAmB;AACrB,8BAAS,YAAY,YAAY;AAC/B,cAAM,gCAAI,IAAI,eAAI,GAAG,MAAM,gCAAI,KAAK,GAAG,EAAE,cAAc,MAAM,CAAC;AAAA,EAChE,CAAC;AACH,OAAO;AACL,8BAAS,YAAY,MAAM;AACzB,qBAAG,KAAK,6BAA6B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC/C,CAAC;AACH;","names":[]}
1
+ {"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { describe, expect, it } from 'vitest';\nimport { STT } from './stt.js';\n\nconst hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);\n\ndescribe('Deepgram streaming language detection', () => {\n // Deepgram only supports language detection for prerecorded audio, so a\n // streaming session must reject it rather than silently default to English.\n // Mirrors livekit-plugins-deepgram (Python).\n it('throws when starting a stream with detectLanguage enabled', () => {\n const stt = new STT({ apiKey: 'test', detectLanguage: true });\n expect(() => stt.stream()).toThrow('language detection is not supported in streaming mode');\n });\n\n it('allows streaming with an explicit language', () => {\n const stt = new STT({ apiKey: 'test', language: 'en-US' });\n const stream = stt.stream();\n // Close immediately so the connection loop never starts (no network in unit tests).\n stream.close();\n });\n});\n\nif (hasDeepgramApiKey) {\n describe('Deepgram', async () => {\n await stt(new STT(), await VAD.load(), { nonStreaming: false });\n });\n} else {\n describe('Deepgram', () => {\n it.skip('requires DEEPGRAM_API_KEY', () => {});\n });\n}\n"],"mappings":";AAGA,kCAAoB;AACpB,iCAAoB;AACpB,oBAAqC;AACrC,iBAAoB;AAEpB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,gBAAgB;AAAA,IAE9D,wBAAS,yCAAyC,MAAM;AAItD,wBAAG,6DAA6D,MAAM;AACpE,UAAMA,OAAM,IAAI,eAAI,EAAE,QAAQ,QAAQ,gBAAgB,KAAK,CAAC;AAC5D,8BAAO,MAAMA,KAAI,OAAO,CAAC,EAAE,QAAQ,uDAAuD;AAAA,EAC5F,CAAC;AAED,wBAAG,8CAA8C,MAAM;AACrD,UAAMA,OAAM,IAAI,eAAI,EAAE,QAAQ,QAAQ,UAAU,QAAQ,CAAC;AACzD,UAAM,SAASA,KAAI,OAAO;AAE1B,WAAO,MAAM;AAAA,EACf,CAAC;AACH,CAAC;AAED,IAAI,mBAAmB;AACrB,8BAAS,YAAY,YAAY;AAC/B,cAAM,gCAAI,IAAI,eAAI,GAAG,MAAM,gCAAI,KAAK,GAAG,EAAE,cAAc,MAAM,CAAC;AAAA,EAChE,CAAC;AACH,OAAO;AACL,8BAAS,YAAY,MAAM;AACzB,qBAAG,KAAK,6BAA6B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC/C,CAAC;AACH;","names":["stt"]}
package/dist/stt.test.js CHANGED
@@ -1,8 +1,19 @@
1
1
  import { VAD } from "@livekit/agents-plugin-silero";
2
2
  import { stt } from "@livekit/agents-plugins-test";
3
- import { describe, it } from "vitest";
3
+ import { describe, expect, it } from "vitest";
4
4
  import { STT } from "./stt.js";
5
5
  const hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);
6
+ describe("Deepgram streaming language detection", () => {
7
+ it("throws when starting a stream with detectLanguage enabled", () => {
8
+ const stt2 = new STT({ apiKey: "test", detectLanguage: true });
9
+ expect(() => stt2.stream()).toThrow("language detection is not supported in streaming mode");
10
+ });
11
+ it("allows streaming with an explicit language", () => {
12
+ const stt2 = new STT({ apiKey: "test", language: "en-US" });
13
+ const stream = stt2.stream();
14
+ stream.close();
15
+ });
16
+ });
6
17
  if (hasDeepgramApiKey) {
7
18
  describe("Deepgram", async () => {
8
19
  await stt(new STT(), await VAD.load(), { nonStreaming: false });
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { describe, it } from 'vitest';\nimport { STT } from './stt.js';\n\nconst hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);\n\nif (hasDeepgramApiKey) {\n describe('Deepgram', async () => {\n await stt(new STT(), await VAD.load(), { nonStreaming: false });\n });\n} else {\n describe('Deepgram', () => {\n it.skip('requires DEEPGRAM_API_KEY', () => {});\n });\n}\n"],"mappings":"AAGA,SAAS,WAAW;AACpB,SAAS,WAAW;AACpB,SAAS,UAAU,UAAU;AAC7B,SAAS,WAAW;AAEpB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,gBAAgB;AAE9D,IAAI,mBAAmB;AACrB,WAAS,YAAY,YAAY;AAC/B,UAAM,IAAI,IAAI,IAAI,GAAG,MAAM,IAAI,KAAK,GAAG,EAAE,cAAc,MAAM,CAAC;AAAA,EAChE,CAAC;AACH,OAAO;AACL,WAAS,YAAY,MAAM;AACzB,OAAG,KAAK,6BAA6B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC/C,CAAC;AACH;","names":[]}
1
+ {"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { describe, expect, it } from 'vitest';\nimport { STT } from './stt.js';\n\nconst hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);\n\ndescribe('Deepgram streaming language detection', () => {\n // Deepgram only supports language detection for prerecorded audio, so a\n // streaming session must reject it rather than silently default to English.\n // Mirrors livekit-plugins-deepgram (Python).\n it('throws when starting a stream with detectLanguage enabled', () => {\n const stt = new STT({ apiKey: 'test', detectLanguage: true });\n expect(() => stt.stream()).toThrow('language detection is not supported in streaming mode');\n });\n\n it('allows streaming with an explicit language', () => {\n const stt = new STT({ apiKey: 'test', language: 'en-US' });\n const stream = stt.stream();\n // Close immediately so the connection loop never starts (no network in unit tests).\n stream.close();\n });\n});\n\nif (hasDeepgramApiKey) {\n describe('Deepgram', async () => {\n await stt(new STT(), await VAD.load(), { nonStreaming: false });\n });\n} else {\n describe('Deepgram', () => {\n it.skip('requires DEEPGRAM_API_KEY', () => {});\n });\n}\n"],"mappings":"AAGA,SAAS,WAAW;AACpB,SAAS,WAAW;AACpB,SAAS,UAAU,QAAQ,UAAU;AACrC,SAAS,WAAW;AAEpB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,gBAAgB;AAE9D,SAAS,yCAAyC,MAAM;AAItD,KAAG,6DAA6D,MAAM;AACpE,UAAMA,OAAM,IAAI,IAAI,EAAE,QAAQ,QAAQ,gBAAgB,KAAK,CAAC;AAC5D,WAAO,MAAMA,KAAI,OAAO,CAAC,EAAE,QAAQ,uDAAuD;AAAA,EAC5F,CAAC;AAED,KAAG,8CAA8C,MAAM;AACrD,UAAMA,OAAM,IAAI,IAAI,EAAE,QAAQ,QAAQ,UAAU,QAAQ,CAAC;AACzD,UAAM,SAASA,KAAI,OAAO;AAE1B,WAAO,MAAM;AAAA,EACf,CAAC;AACH,CAAC;AAED,IAAI,mBAAmB;AACrB,WAAS,YAAY,YAAY;AAC/B,UAAM,IAAI,IAAI,IAAI,GAAG,MAAM,IAAI,KAAK,GAAG,EAAE,cAAc,MAAM,CAAC;AAAA,EAChE,CAAC;AACH,OAAO;AACL,WAAS,YAAY,MAAM;AACzB,OAAG,KAAK,6BAA6B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC/C,CAAC;AACH;","names":["stt"]}
package/dist/tts.cjs CHANGED
@@ -30,9 +30,11 @@ const AUTHORIZATION_HEADER = "Authorization";
30
30
  const NUM_CHANNELS = 1;
31
31
  const MIN_SENTENCE_LENGTH = 8;
32
32
  const defaultTTSOptions = {
33
- model: "aura-asteria-en",
33
+ model: "aura-2-andromeda-en",
34
34
  encoding: "linear16",
35
35
  sampleRate: 24e3,
36
+ bitRate: null,
37
+ mipOptOut: false,
36
38
  apiKey: process.env.DEEPGRAM_API_KEY,
37
39
  baseUrl: "https://api.deepgram.com",
38
40
  capabilities: {
@@ -51,6 +53,9 @@ class TTS extends import_agents.tts.TTS {
51
53
  get provider() {
52
54
  return "Deepgram";
53
55
  }
56
+ get sampleRate() {
57
+ return this.opts.sampleRate;
58
+ }
54
59
  constructor(opts = {}) {
55
60
  var _a;
56
61
  super(opts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {
@@ -75,6 +80,12 @@ class TTS extends import_agents.tts.TTS {
75
80
  stream(options) {
76
81
  return new SynthesizeStream(this, this.opts, options == null ? void 0 : options.connOptions);
77
82
  }
83
+ updateOptions(opts) {
84
+ this.opts = {
85
+ ...this.opts,
86
+ ...opts
87
+ };
88
+ }
78
89
  }
79
90
  class ChunkedStream extends import_agents.tts.ChunkedStream {
80
91
  label = "deepgram.ChunkedStream";
@@ -94,6 +105,11 @@ class ChunkedStream extends import_agents.tts.ChunkedStream {
94
105
  url.searchParams.append("sample_rate", this.opts.sampleRate.toString());
95
106
  url.searchParams.append("model", this.opts.model);
96
107
  url.searchParams.append("encoding", this.opts.encoding);
108
+ url.searchParams.append("container", "none");
109
+ url.searchParams.append("mip_opt_out", String(this.opts.mipOptOut));
110
+ if (this.opts.bitRate !== void 0 && this.opts.bitRate !== null) {
111
+ url.searchParams.append("bit_rate", this.opts.bitRate.toString());
112
+ }
97
113
  if (this.opts.speed !== void 0) {
98
114
  url.searchParams.append("speed", this.opts.speed.toString());
99
115
  }
@@ -228,6 +244,10 @@ class SynthesizeStream extends import_agents.tts.SynthesizeStream {
228
244
  url.searchParams.append("sample_rate", this.opts.sampleRate.toString());
229
245
  url.searchParams.append("model", this.opts.model);
230
246
  url.searchParams.append("encoding", this.opts.encoding);
247
+ url.searchParams.append("mip_opt_out", String(this.opts.mipOptOut));
248
+ if (this.opts.bitRate !== void 0 && this.opts.bitRate !== null) {
249
+ url.searchParams.append("bit_rate", this.opts.bitRate.toString());
250
+ }
231
251
  if (this.opts.speed !== void 0) {
232
252
  url.searchParams.append("speed", this.opts.speed.toString());
233
253
  }
package/dist/tts.cjs.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n APIConnectionError,\n APIError,\n APIStatusError,\n APITimeoutError,\n AudioByteStream,\n log,\n shortuuid,\n tokenize,\n tts,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { request } from 'node:https';\nimport { type RawData, WebSocket } from 'ws';\nimport type { TTSEncoding, TTSModels } from './models.js';\n\nconst AUTHORIZATION_HEADER = 'Authorization';\nconst NUM_CHANNELS = 1;\nconst MIN_SENTENCE_LENGTH = 8;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n encoding: TTSEncoding;\n sampleRate: number;\n speed?: number;\n apiKey?: string;\n baseUrl?: string;\n sentenceTokenizer: tokenize.SentenceTokenizer;\n capabilities: tts.TTSCapabilities;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'aura-asteria-en',\n encoding: 'linear16',\n sampleRate: 24000,\n apiKey: process.env.DEEPGRAM_API_KEY,\n baseUrl: 'https://api.deepgram.com',\n capabilities: {\n streaming: true,\n },\n sentenceTokenizer: new tokenize.basic.SentenceTokenizer({\n minSentenceLength: MIN_SENTENCE_LENGTH,\n }),\n};\n\nexport class TTS extends tts.TTS {\n private opts: TTSOptions;\n label = 'deepgram.TTS';\n\n get model(): string {\n return this.opts.model;\n }\n\n get provider(): string {\n return 'Deepgram';\n }\n\n constructor(opts: Partial<TTSOptions> = {}) {\n super(opts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {\n streaming: opts.capabilities?.streaming ?? defaultTTSOptions.capabilities.streaming,\n });\n\n this.opts = {\n ...defaultTTSOptions,\n ...opts,\n };\n\n if (this.opts.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n if (this.opts.speed !== undefined && (this.opts.speed < 0.7 || this.opts.speed > 1.5)) {\n throw new Error(`Deepgram TTS speed must be between 0.7 and 1.5, got ${this.opts.speed}`);\n }\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.opts, connOptions, abortSignal);\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): tts.SynthesizeStream {\n return new SynthesizeStream(this, this.opts, options?.connOptions);\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'deepgram.ChunkedStream';\n #logger = log();\n private opts: TTSOptions;\n private text: string;\n\n constructor(\n tts: TTS,\n text: string,\n opts: TTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.text = text;\n this.opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const json = { text: this.text };\n const url = new URL(`${this.opts.baseUrl!}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n if (this.opts.speed !== undefined) {\n url.searchParams.append('speed', this.opts.speed.toString());\n }\n\n await new Promise<void>((resolve, reject) => {\n let settled = false;\n const settle = (fn: () => void) => {\n if (!settled) {\n settled = true;\n fn();\n }\n };\n\n const req = request(\n {\n hostname: url.hostname,\n port: 443,\n path: url.pathname + url.search,\n method: 'POST',\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n 'Content-Type': 'application/json',\n },\n signal: this.abortSignal,\n },\n (res) => {\n if (res.statusCode !== 200) {\n settle(() =>\n reject(\n new Error(\n `Deepgram TTS HTTP request failed: ${res.statusCode} ${res.statusMessage}`,\n ),\n ),\n );\n return;\n }\n\n res.on('data', (chunk) => {\n for (const frame of bstream.write(chunk)) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n });\n\n res.on('error', (err) => {\n if (err.message === 'aborted') return;\n this.#logger.error({ err }, 'Deepgram TTS response error');\n settle(() => reject(err));\n });\n\n res.on('close', () => {\n for (const frame of bstream.flush()) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n if (!this.queue.closed) {\n this.queue.close();\n }\n settle(() => resolve());\n });\n },\n );\n\n req.on('error', (err) => {\n if (err.name === 'AbortError') return;\n this.#logger.error({ err }, 'Deepgram TTS request error');\n settle(() => reject(err));\n });\n\n req.on('close', () => settle(() => resolve()));\n req.write(JSON.stringify(json));\n req.end();\n });\n }\n}\n\nexport class SynthesizeStream extends tts.SynthesizeStream {\n private opts: TTSOptions;\n private tokenizer: tokenize.SentenceStream;\n #logger = log();\n label = 'deepgram.SynthesizeStream';\n\n private static readonly FLUSH_MSG = JSON.stringify({ type: 'Flush' });\n private static readonly CLOSE_MSG = JSON.stringify({ type: 'Close' });\n\n constructor(tts: TTS, opts: TTSOptions, connOptions?: APIConnectOptions) {\n super(tts, connOptions);\n this.opts = opts;\n this.tokenizer = opts.sentenceTokenizer.stream();\n }\n\n private async closeWebSocket(ws: WebSocket): Promise<void> {\n try {\n // Send Flush and Close messages to ensure Deepgram processes all remaining audio\n // and properly terminates the session, preventing lingering TTS sessions\n if (ws.readyState === WebSocket.OPEN) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n ws.send(SynthesizeStream.CLOSE_MSG);\n\n // Wait for server acknowledgment to prevent race conditions and ensure\n // proper cleanup, avoiding 429 Too Many Requests errors from lingering sessions\n try {\n await new Promise<void>((resolve, _reject) => {\n const timeout = setTimeout(() => {\n resolve();\n }, 1000);\n\n ws.once('message', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('close', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('error', () => {\n clearTimeout(timeout);\n resolve();\n });\n });\n } catch (e) {\n // Ignore timeout or other errors during close sequence\n }\n }\n } catch (e) {\n console.warn(`Error during WebSocket close sequence: ${e}`);\n } finally {\n if (ws.readyState === WebSocket.OPEN || ws.readyState === WebSocket.CONNECTING) {\n ws.close();\n }\n }\n }\n\n protected async run() {\n const requestId = shortuuid();\n const segmentId = shortuuid();\n\n const wsUrl = this.opts.baseUrl!.replace(/^http/, 'ws');\n const url = new URL(`${wsUrl}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n if (this.opts.speed !== undefined) {\n url.searchParams.append('speed', this.opts.speed.toString());\n }\n\n const ws = new WebSocket(url, {\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n },\n });\n\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n const inputTask = async () => {\n for await (const data of this.input) {\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n this.tokenizer.flush();\n continue;\n }\n this.tokenizer.pushText(data);\n }\n this.tokenizer.endInput();\n this.tokenizer.close();\n };\n\n let markInputSent: () => void = () => {};\n const inputSent = new Promise<void>((resolve) => {\n markInputSent = resolve;\n });\n\n const sendTask = async () => {\n try {\n for await (const event of this.tokenizer) {\n if (this.abortController.signal.aborted) break;\n\n let text = event.token;\n if (!text.endsWith(' ')) {\n text += ' ';\n }\n\n const message = JSON.stringify({\n type: 'Speak',\n text: text,\n });\n\n ws.send(message);\n markInputSent();\n }\n\n if (!this.abortController.signal.aborted) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n markInputSent();\n }\n } finally {\n markInputSent();\n }\n };\n\n const recvTask = async () => {\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n let finalReceived = false;\n let timeout: NodeJS.Timeout | null = null;\n let lastFrame: AudioFrame | undefined;\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame && !this.queue.closed) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const clearMessageTimeout = () => {\n if (timeout) {\n clearTimeout(timeout);\n timeout = null;\n }\n };\n\n const resetMessageTimeout = (reject: (reason?: unknown) => void) => {\n clearMessageTimeout();\n timeout = setTimeout(() => {\n reject(new APITimeoutError({ message: 'Deepgram TTS recv idle timeout' }));\n }, this.connOptions.timeoutMs);\n };\n\n await inputSent;\n if (this.abortController.signal.aborted) return;\n\n return new Promise<void>((resolve, reject) => {\n resetMessageTimeout(reject);\n\n ws.on('message', (data: RawData, isBinary: boolean) => {\n clearMessageTimeout();\n\n if (!isBinary) {\n const message = JSON.parse(data.toString());\n if (message.type === 'Flushed') {\n finalReceived = true;\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n resolve();\n return;\n } else if (message.type === 'Warning') {\n this.#logger.warn(`Deepgram warning: ${message.warn_msg}`);\n } else if (message.type === 'Error' || message.type === 'error') {\n reject(new APIError('Deepgram TTS returned error', { body: message }));\n return;\n } else if (message.type !== 'Metadata') {\n this.#logger.warn({ message }, 'Unknown Deepgram message type');\n }\n\n resetMessageTimeout(reject);\n return;\n }\n\n const buffer =\n data instanceof Buffer\n ? data.buffer.slice(data.byteOffset, data.byteOffset + data.byteLength)\n : (data as ArrayBuffer);\n for (const frame of bstream.write(buffer as ArrayBuffer)) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n resetMessageTimeout(reject);\n });\n\n ws.on('close', (code, reason) => {\n clearMessageTimeout();\n if (!finalReceived) {\n reject(\n new APIStatusError({\n message: 'Deepgram websocket connection closed unexpectedly',\n options: {\n statusCode: code || -1,\n body: { reason: reason.toString() },\n },\n }),\n );\n return;\n }\n resolve();\n });\n\n ws.on('error', (error) => {\n clearMessageTimeout();\n reject(error);\n });\n });\n };\n\n try {\n await Promise.all([inputTask(), sendTask(), recvTask()]);\n } catch (e) {\n if (this.abortController.signal.aborted) return;\n if (e instanceof APIError) throw e;\n throw new APIConnectionError({\n message: `Deepgram TTS WebSocket failed: ${(e as Error).message ?? 'unknown error'}`,\n });\n } finally {\n await this.closeWebSocket(ws);\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAWO;AAEP,wBAAwB;AACxB,gBAAwC;AAGxC,MAAM,uBAAuB;AAC7B,MAAM,eAAe;AACrB,MAAM,sBAAsB;AAa5B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,QAAQ,QAAQ,IAAI;AAAA,EACpB,SAAS;AAAA,EACT,cAAc;AAAA,IACZ,WAAW;AAAA,EACb;AAAA,EACA,mBAAmB,IAAI,uBAAS,MAAM,kBAAkB;AAAA,IACtD,mBAAmB;AAAA,EACrB,CAAC;AACH;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EACvB;AAAA,EACR,QAAQ;AAAA,EAER,IAAI,QAAgB;AAClB,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AA7D9C;AA8DI,UAAM,KAAK,cAAc,kBAAkB,YAAY,cAAc;AAAA,MACnE,aAAW,UAAK,iBAAL,mBAAmB,cAAa,kBAAkB,aAAa;AAAA,IAC5E,CAAC;AAED,SAAK,OAAO;AAAA,MACV,GAAG;AAAA,MACH,GAAG;AAAA,IACL;AAEA,QAAI,KAAK,KAAK,WAAW,QAAW;AAClC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,UAAU,WAAc,KAAK,KAAK,QAAQ,OAAO,KAAK,KAAK,QAAQ,MAAM;AACrF,YAAM,IAAI,MAAM,uDAAuD,KAAK,KAAK,KAAK,EAAE;AAAA,IAC1F;AAAA,EACF;AAAA,EAEA,WACE,MACA,aACA,aACmB;AACnB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,MAAM,aAAa,WAAW;AAAA,EAC1E;AAAA,EAEA,OAAO,SAAqE;AAC1E,WAAO,IAAI,iBAAiB,MAAM,KAAK,MAAM,mCAAS,WAAW;AAAA,EACnE;AACF;AAEO,MAAM,sBAAsB,kBAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACR,cAAU,mBAAI;AAAA,EACN;AAAA,EACA;AAAA,EAER,YACEA,MACA,MACA,MACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,OAAO;AACZ,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,gBAAY,yBAAU;AAC5B,UAAM,UAAU,IAAI,8BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAM,OAAO,EAAE,MAAM,KAAK,KAAK;AAC/B,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,KAAK,OAAQ,WAAW;AACpD,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AACtD,QAAI,KAAK,KAAK,UAAU,QAAW;AACjC,UAAI,aAAa,OAAO,SAAS,KAAK,KAAK,MAAM,SAAS,CAAC;AAAA,IAC7D;AAEA,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,UAAI,UAAU;AACd,YAAM,SAAS,CAAC,OAAmB;AACjC,YAAI,CAAC,SAAS;AACZ,oBAAU;AACV,aAAG;AAAA,QACL;AAAA,MACF;AAEA,YAAM,UAAM;AAAA,QACV;AAAA,UACE,UAAU,IAAI;AAAA,UACd,MAAM;AAAA,UACN,MAAM,IAAI,WAAW,IAAI;AAAA,UACzB,QAAQ;AAAA,UACR,SAAS;AAAA,YACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,YAClD,gBAAgB;AAAA,UAClB;AAAA,UACA,QAAQ,KAAK;AAAA,QACf;AAAA,QACA,CAAC,QAAQ;AACP,cAAI,IAAI,eAAe,KAAK;AAC1B;AAAA,cAAO,MACL;AAAA,gBACE,IAAI;AAAA,kBACF,qCAAqC,IAAI,UAAU,IAAI,IAAI,aAAa;AAAA,gBAC1E;AAAA,cACF;AAAA,YACF;AACA;AAAA,UACF;AAEA,cAAI,GAAG,QAAQ,CAAC,UAAU;AACxB,uBAAW,SAAS,QAAQ,MAAM,KAAK,GAAG;AACxC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AAAA,UACF,CAAC;AAED,cAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,gBAAI,IAAI,YAAY,UAAW;AAC/B,iBAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6BAA6B;AACzD,mBAAO,MAAM,OAAO,GAAG,CAAC;AAAA,UAC1B,CAAC;AAED,cAAI,GAAG,SAAS,MAAM;AACpB,uBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AACA,gBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,mBAAK,MAAM,MAAM;AAAA,YACnB;AACA,mBAAO,MAAM,QAAQ,CAAC;AAAA,UACxB,CAAC;AAAA,QACH;AAAA,MACF;AAEA,UAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,YAAI,IAAI,SAAS,aAAc;AAC/B,aAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,4BAA4B;AACxD,eAAO,MAAM,OAAO,GAAG,CAAC;AAAA,MAC1B,CAAC;AAED,UAAI,GAAG,SAAS,MAAM,OAAO,MAAM,QAAQ,CAAC,CAAC;AAC7C,UAAI,MAAM,KAAK,UAAU,IAAI,CAAC;AAC9B,UAAI,IAAI;AAAA,IACV,CAAC;AAAA,EACH;AACF;AAEO,MAAM,yBAAyB,kBAAI,iBAAiB;AAAA,EACjD;AAAA,EACA;AAAA,EACR,cAAU,mBAAI;AAAA,EACd,QAAQ;AAAA,EAER,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EACpE,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EAEpE,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,WAAW;AACtB,SAAK,OAAO;AACZ,SAAK,YAAY,KAAK,kBAAkB,OAAO;AAAA,EACjD;AAAA,EAEA,MAAc,eAAe,IAA8B;AACzD,QAAI;AAGF,UAAI,GAAG,eAAe,oBAAU,MAAM;AACpC,WAAG,KAAK,iBAAiB,SAAS;AAClC,WAAG,KAAK,iBAAiB,SAAS;AAIlC,YAAI;AACF,gBAAM,IAAI,QAAc,CAAC,SAAS,YAAY;AAC5C,kBAAM,UAAU,WAAW,MAAM;AAC/B,sBAAQ;AAAA,YACV,GAAG,GAAI;AAEP,eAAG,KAAK,WAAW,MAAM;AACvB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAAA,UACH,CAAC;AAAA,QACH,SAAS,GAAG;AAAA,QAEZ;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AACV,cAAQ,KAAK,0CAA0C,CAAC,EAAE;AAAA,IAC5D,UAAE;AACA,UAAI,GAAG,eAAe,oBAAU,QAAQ,GAAG,eAAe,oBAAU,YAAY;AAC9E,WAAG,MAAM;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,gBAAY,yBAAU;AAC5B,UAAM,gBAAY,yBAAU;AAE5B,UAAM,QAAQ,KAAK,KAAK,QAAS,QAAQ,SAAS,IAAI;AACtD,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,WAAW;AACvC,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AACtD,QAAI,KAAK,KAAK,UAAU,QAAW;AACjC,UAAI,aAAa,OAAO,SAAS,KAAK,KAAK,MAAM,SAAS,CAAC;AAAA,IAC7D;AAEA,UAAM,KAAK,IAAI,oBAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,MACpD;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,SAAG,GAAG,QAAQ,OAAO;AACrB,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,IAC/D,CAAC;AAED,UAAM,YAAY,YAAY;AAC5B,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,eAAK,UAAU,MAAM;AACrB;AAAA,QACF;AACA,aAAK,UAAU,SAAS,IAAI;AAAA,MAC9B;AACA,WAAK,UAAU,SAAS;AACxB,WAAK,UAAU,MAAM;AAAA,IACvB;AAEA,QAAI,gBAA4B,MAAM;AAAA,IAAC;AACvC,UAAM,YAAY,IAAI,QAAc,CAAC,YAAY;AAC/C,sBAAgB;AAAA,IAClB,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,UAAI;AACF,yBAAiB,SAAS,KAAK,WAAW;AACxC,cAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,cAAI,OAAO,MAAM;AACjB,cAAI,CAAC,KAAK,SAAS,GAAG,GAAG;AACvB,oBAAQ;AAAA,UACV;AAEA,gBAAM,UAAU,KAAK,UAAU;AAAA,YAC7B,MAAM;AAAA,YACN;AAAA,UACF,CAAC;AAED,aAAG,KAAK,OAAO;AACf,wBAAc;AAAA,QAChB;AAEA,YAAI,CAAC,KAAK,gBAAgB,OAAO,SAAS;AACxC,aAAG,KAAK,iBAAiB,SAAS;AAClC,wBAAc;AAAA,QAChB;AAAA,MACF,UAAE;AACA,sBAAc;AAAA,MAChB;AAAA,IACF;AAEA,UAAM,WAAW,YAAY;AAC3B,YAAM,UAAU,IAAI,8BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAI,gBAAgB;AACpB,UAAI,UAAiC;AACrC,UAAI;AAEJ,YAAM,gBAAgB,CAACC,YAAmB,UAAmB;AAC3D,YAAI,aAAa,CAAC,KAAK,MAAM,QAAQ;AACnC,eAAK,MAAM,IAAI,EAAE,WAAW,WAAAA,YAAW,OAAO,WAAW,MAAM,CAAC;AAChE,sBAAY;AAAA,QACd;AAAA,MACF;AAEA,YAAM,sBAAsB,MAAM;AAChC,YAAI,SAAS;AACX,uBAAa,OAAO;AACpB,oBAAU;AAAA,QACZ;AAAA,MACF;AAEA,YAAM,sBAAsB,CAAC,WAAuC;AAClE,4BAAoB;AACpB,kBAAU,WAAW,MAAM;AACzB,iBAAO,IAAI,8BAAgB,EAAE,SAAS,iCAAiC,CAAC,CAAC;AAAA,QAC3E,GAAG,KAAK,YAAY,SAAS;AAAA,MAC/B;AAEA,YAAM;AACN,UAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,4BAAoB,MAAM;AAE1B,WAAG,GAAG,WAAW,CAAC,MAAe,aAAsB;AACrD,8BAAoB;AAEpB,cAAI,CAAC,UAAU;AACb,kBAAM,UAAU,KAAK,MAAM,KAAK,SAAS,CAAC;AAC1C,gBAAI,QAAQ,SAAS,WAAW;AAC9B,8BAAgB;AAChB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,WAAW,KAAK;AAC9B,4BAAY;AAAA,cACd;AACA,4BAAc,WAAW,IAAI;AAE7B,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,cAC/C;AACA,sBAAQ;AACR;AAAA,YACF,WAAW,QAAQ,SAAS,WAAW;AACrC,mBAAK,QAAQ,KAAK,qBAAqB,QAAQ,QAAQ,EAAE;AAAA,YAC3D,WAAW,QAAQ,SAAS,WAAW,QAAQ,SAAS,SAAS;AAC/D,qBAAO,IAAI,uBAAS,+BAA+B,EAAE,MAAM,QAAQ,CAAC,CAAC;AACrE;AAAA,YACF,WAAW,QAAQ,SAAS,YAAY;AACtC,mBAAK,QAAQ,KAAK,EAAE,QAAQ,GAAG,+BAA+B;AAAA,YAChE;AAEA,gCAAoB,MAAM;AAC1B;AAAA,UACF;AAEA,gBAAM,SACJ,gBAAgB,SACZ,KAAK,OAAO,MAAM,KAAK,YAAY,KAAK,aAAa,KAAK,UAAU,IACnE;AACP,qBAAW,SAAS,QAAQ,MAAM,MAAqB,GAAG;AACxD,0BAAc,WAAW,KAAK;AAC9B,wBAAY;AAAA,UACd;AACA,8BAAoB,MAAM;AAAA,QAC5B,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,MAAM,WAAW;AAC/B,8BAAoB;AACpB,cAAI,CAAC,eAAe;AAClB;AAAA,cACE,IAAI,6BAAe;AAAA,gBACjB,SAAS;AAAA,gBACT,SAAS;AAAA,kBACP,YAAY,QAAQ;AAAA,kBACpB,MAAM,EAAE,QAAQ,OAAO,SAAS,EAAE;AAAA,gBACpC;AAAA,cACF,CAAC;AAAA,YACH;AACA;AAAA,UACF;AACA,kBAAQ;AAAA,QACV,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,UAAU;AACxB,8BAAoB;AACpB,iBAAO,KAAK;AAAA,QACd,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,UAAU,GAAG,SAAS,GAAG,SAAS,CAAC,CAAC;AAAA,IACzD,SAAS,GAAG;AACV,UAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,UAAI,aAAa,uBAAU,OAAM;AACjC,YAAM,IAAI,iCAAmB;AAAA,QAC3B,SAAS,kCAAmC,EAAY,WAAW,eAAe;AAAA,MACpF,CAAC;AAAA,IACH,UAAE;AACA,YAAM,KAAK,eAAe,EAAE;AAAA,IAC9B;AAAA,EACF;AACF;","names":["tts","segmentId"]}
1
+ {"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n APIConnectionError,\n APIError,\n APIStatusError,\n APITimeoutError,\n AudioByteStream,\n log,\n shortuuid,\n tokenize,\n tts,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { request } from 'node:https';\nimport { type RawData, WebSocket } from 'ws';\nimport type { TTSEncoding, TTSModels } from './models.js';\n\nconst AUTHORIZATION_HEADER = 'Authorization';\nconst NUM_CHANNELS = 1;\nconst MIN_SENTENCE_LENGTH = 8;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n encoding: TTSEncoding;\n sampleRate: number;\n bitRate?: number | null;\n speed?: number;\n mipOptOut: boolean;\n apiKey?: string;\n baseUrl?: string;\n sentenceTokenizer: tokenize.SentenceTokenizer;\n capabilities: tts.TTSCapabilities;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'aura-2-andromeda-en',\n encoding: 'linear16',\n sampleRate: 24000,\n bitRate: null,\n mipOptOut: false,\n apiKey: process.env.DEEPGRAM_API_KEY,\n baseUrl: 'https://api.deepgram.com',\n capabilities: {\n streaming: true,\n },\n sentenceTokenizer: new tokenize.basic.SentenceTokenizer({\n minSentenceLength: MIN_SENTENCE_LENGTH,\n }),\n};\n\nexport class TTS extends tts.TTS {\n private opts: TTSOptions;\n label = 'deepgram.TTS';\n\n get model(): string {\n return this.opts.model;\n }\n\n get provider(): string {\n return 'Deepgram';\n }\n\n override get sampleRate(): number {\n return this.opts.sampleRate;\n }\n\n constructor(opts: Partial<TTSOptions> = {}) {\n super(opts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {\n streaming: opts.capabilities?.streaming ?? defaultTTSOptions.capabilities.streaming,\n });\n\n this.opts = {\n ...defaultTTSOptions,\n ...opts,\n };\n\n if (this.opts.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n if (this.opts.speed !== undefined && (this.opts.speed < 0.7 || this.opts.speed > 1.5)) {\n throw new Error(`Deepgram TTS speed must be between 0.7 and 1.5, got ${this.opts.speed}`);\n }\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.opts, connOptions, abortSignal);\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): tts.SynthesizeStream {\n return new SynthesizeStream(this, this.opts, options?.connOptions);\n }\n\n updateOptions(\n opts: Partial<\n Pick<TTSOptions, 'model' | 'encoding' | 'sampleRate' | 'bitRate' | 'speed' | 'mipOptOut'>\n >,\n ) {\n this.opts = {\n ...this.opts,\n ...opts,\n };\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'deepgram.ChunkedStream';\n #logger = log();\n private opts: TTSOptions;\n private text: string;\n\n constructor(\n tts: TTS,\n text: string,\n opts: TTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.text = text;\n this.opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const json = { text: this.text };\n const url = new URL(`${this.opts.baseUrl!}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n url.searchParams.append('container', 'none');\n url.searchParams.append('mip_opt_out', String(this.opts.mipOptOut));\n if (this.opts.bitRate !== undefined && this.opts.bitRate !== null) {\n url.searchParams.append('bit_rate', this.opts.bitRate.toString());\n }\n if (this.opts.speed !== undefined) {\n url.searchParams.append('speed', this.opts.speed.toString());\n }\n\n await new Promise<void>((resolve, reject) => {\n let settled = false;\n const settle = (fn: () => void) => {\n if (!settled) {\n settled = true;\n fn();\n }\n };\n\n const req = request(\n {\n hostname: url.hostname,\n port: 443,\n path: url.pathname + url.search,\n method: 'POST',\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n 'Content-Type': 'application/json',\n },\n signal: this.abortSignal,\n },\n (res) => {\n if (res.statusCode !== 200) {\n settle(() =>\n reject(\n new Error(\n `Deepgram TTS HTTP request failed: ${res.statusCode} ${res.statusMessage}`,\n ),\n ),\n );\n return;\n }\n\n res.on('data', (chunk) => {\n for (const frame of bstream.write(chunk)) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n });\n\n res.on('error', (err) => {\n if (err.message === 'aborted') return;\n this.#logger.error({ err }, 'Deepgram TTS response error');\n settle(() => reject(err));\n });\n\n res.on('close', () => {\n for (const frame of bstream.flush()) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n if (!this.queue.closed) {\n this.queue.close();\n }\n settle(() => resolve());\n });\n },\n );\n\n req.on('error', (err) => {\n if (err.name === 'AbortError') return;\n this.#logger.error({ err }, 'Deepgram TTS request error');\n settle(() => reject(err));\n });\n\n req.on('close', () => settle(() => resolve()));\n req.write(JSON.stringify(json));\n req.end();\n });\n }\n}\n\nexport class SynthesizeStream extends tts.SynthesizeStream {\n private opts: TTSOptions;\n private tokenizer: tokenize.SentenceStream;\n #logger = log();\n label = 'deepgram.SynthesizeStream';\n\n private static readonly FLUSH_MSG = JSON.stringify({ type: 'Flush' });\n private static readonly CLOSE_MSG = JSON.stringify({ type: 'Close' });\n\n constructor(tts: TTS, opts: TTSOptions, connOptions?: APIConnectOptions) {\n super(tts, connOptions);\n this.opts = opts;\n this.tokenizer = opts.sentenceTokenizer.stream();\n }\n\n private async closeWebSocket(ws: WebSocket): Promise<void> {\n try {\n // Send Flush and Close messages to ensure Deepgram processes all remaining audio\n // and properly terminates the session, preventing lingering TTS sessions\n if (ws.readyState === WebSocket.OPEN) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n ws.send(SynthesizeStream.CLOSE_MSG);\n\n // Wait for server acknowledgment to prevent race conditions and ensure\n // proper cleanup, avoiding 429 Too Many Requests errors from lingering sessions\n try {\n await new Promise<void>((resolve, _reject) => {\n const timeout = setTimeout(() => {\n resolve();\n }, 1000);\n\n ws.once('message', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('close', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('error', () => {\n clearTimeout(timeout);\n resolve();\n });\n });\n } catch (e) {\n // Ignore timeout or other errors during close sequence\n }\n }\n } catch (e) {\n console.warn(`Error during WebSocket close sequence: ${e}`);\n } finally {\n if (ws.readyState === WebSocket.OPEN || ws.readyState === WebSocket.CONNECTING) {\n ws.close();\n }\n }\n }\n\n protected async run() {\n const requestId = shortuuid();\n const segmentId = shortuuid();\n\n const wsUrl = this.opts.baseUrl!.replace(/^http/, 'ws');\n const url = new URL(`${wsUrl}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n url.searchParams.append('mip_opt_out', String(this.opts.mipOptOut));\n if (this.opts.bitRate !== undefined && this.opts.bitRate !== null) {\n url.searchParams.append('bit_rate', this.opts.bitRate.toString());\n }\n if (this.opts.speed !== undefined) {\n url.searchParams.append('speed', this.opts.speed.toString());\n }\n\n const ws = new WebSocket(url, {\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n },\n });\n\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n const inputTask = async () => {\n for await (const data of this.input) {\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n this.tokenizer.flush();\n continue;\n }\n this.tokenizer.pushText(data);\n }\n this.tokenizer.endInput();\n this.tokenizer.close();\n };\n\n let markInputSent: () => void = () => {};\n const inputSent = new Promise<void>((resolve) => {\n markInputSent = resolve;\n });\n\n const sendTask = async () => {\n try {\n for await (const event of this.tokenizer) {\n if (this.abortController.signal.aborted) break;\n\n let text = event.token;\n if (!text.endsWith(' ')) {\n text += ' ';\n }\n\n const message = JSON.stringify({\n type: 'Speak',\n text: text,\n });\n\n ws.send(message);\n markInputSent();\n }\n\n if (!this.abortController.signal.aborted) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n markInputSent();\n }\n } finally {\n markInputSent();\n }\n };\n\n const recvTask = async () => {\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n let finalReceived = false;\n let timeout: NodeJS.Timeout | null = null;\n let lastFrame: AudioFrame | undefined;\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame && !this.queue.closed) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const clearMessageTimeout = () => {\n if (timeout) {\n clearTimeout(timeout);\n timeout = null;\n }\n };\n\n const resetMessageTimeout = (reject: (reason?: unknown) => void) => {\n clearMessageTimeout();\n timeout = setTimeout(() => {\n reject(new APITimeoutError({ message: 'Deepgram TTS recv idle timeout' }));\n }, this.connOptions.timeoutMs);\n };\n\n await inputSent;\n if (this.abortController.signal.aborted) return;\n\n return new Promise<void>((resolve, reject) => {\n resetMessageTimeout(reject);\n\n ws.on('message', (data: RawData, isBinary: boolean) => {\n clearMessageTimeout();\n\n if (!isBinary) {\n const message = JSON.parse(data.toString());\n if (message.type === 'Flushed') {\n finalReceived = true;\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n resolve();\n return;\n } else if (message.type === 'Warning') {\n this.#logger.warn(`Deepgram warning: ${message.warn_msg}`);\n } else if (message.type === 'Error' || message.type === 'error') {\n reject(new APIError('Deepgram TTS returned error', { body: message }));\n return;\n } else if (message.type !== 'Metadata') {\n this.#logger.warn({ message }, 'Unknown Deepgram message type');\n }\n\n resetMessageTimeout(reject);\n return;\n }\n\n const buffer =\n data instanceof Buffer\n ? data.buffer.slice(data.byteOffset, data.byteOffset + data.byteLength)\n : (data as ArrayBuffer);\n for (const frame of bstream.write(buffer as ArrayBuffer)) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n resetMessageTimeout(reject);\n });\n\n ws.on('close', (code, reason) => {\n clearMessageTimeout();\n if (!finalReceived) {\n reject(\n new APIStatusError({\n message: 'Deepgram websocket connection closed unexpectedly',\n options: {\n statusCode: code || -1,\n body: { reason: reason.toString() },\n },\n }),\n );\n return;\n }\n resolve();\n });\n\n ws.on('error', (error) => {\n clearMessageTimeout();\n reject(error);\n });\n });\n };\n\n try {\n await Promise.all([inputTask(), sendTask(), recvTask()]);\n } catch (e) {\n if (this.abortController.signal.aborted) return;\n if (e instanceof APIError) throw e;\n throw new APIConnectionError({\n message: `Deepgram TTS WebSocket failed: ${(e as Error).message ?? 'unknown error'}`,\n });\n } finally {\n await this.closeWebSocket(ws);\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAWO;AAEP,wBAAwB;AACxB,gBAAwC;AAGxC,MAAM,uBAAuB;AAC7B,MAAM,eAAe;AACrB,MAAM,sBAAsB;AAe5B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,SAAS;AAAA,EACT,WAAW;AAAA,EACX,QAAQ,QAAQ,IAAI;AAAA,EACpB,SAAS;AAAA,EACT,cAAc;AAAA,IACZ,WAAW;AAAA,EACb;AAAA,EACA,mBAAmB,IAAI,uBAAS,MAAM,kBAAkB;AAAA,IACtD,mBAAmB;AAAA,EACrB,CAAC;AACH;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EACvB;AAAA,EACR,QAAQ;AAAA,EAER,IAAI,QAAgB;AAClB,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,IAAa,aAAqB;AAChC,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AArE9C;AAsEI,UAAM,KAAK,cAAc,kBAAkB,YAAY,cAAc;AAAA,MACnE,aAAW,UAAK,iBAAL,mBAAmB,cAAa,kBAAkB,aAAa;AAAA,IAC5E,CAAC;AAED,SAAK,OAAO;AAAA,MACV,GAAG;AAAA,MACH,GAAG;AAAA,IACL;AAEA,QAAI,KAAK,KAAK,WAAW,QAAW;AAClC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,UAAU,WAAc,KAAK,KAAK,QAAQ,OAAO,KAAK,KAAK,QAAQ,MAAM;AACrF,YAAM,IAAI,MAAM,uDAAuD,KAAK,KAAK,KAAK,EAAE;AAAA,IAC1F;AAAA,EACF;AAAA,EAEA,WACE,MACA,aACA,aACmB;AACnB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,MAAM,aAAa,WAAW;AAAA,EAC1E;AAAA,EAEA,OAAO,SAAqE;AAC1E,WAAO,IAAI,iBAAiB,MAAM,KAAK,MAAM,mCAAS,WAAW;AAAA,EACnE;AAAA,EAEA,cACE,MAGA;AACA,SAAK,OAAO;AAAA,MACV,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AACF;AAEO,MAAM,sBAAsB,kBAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACR,cAAU,mBAAI;AAAA,EACN;AAAA,EACA;AAAA,EAER,YACEA,MACA,MACA,MACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,OAAO;AACZ,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,gBAAY,yBAAU;AAC5B,UAAM,UAAU,IAAI,8BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAM,OAAO,EAAE,MAAM,KAAK,KAAK;AAC/B,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,KAAK,OAAQ,WAAW;AACpD,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AACtD,QAAI,aAAa,OAAO,aAAa,MAAM;AAC3C,QAAI,aAAa,OAAO,eAAe,OAAO,KAAK,KAAK,SAAS,CAAC;AAClE,QAAI,KAAK,KAAK,YAAY,UAAa,KAAK,KAAK,YAAY,MAAM;AACjE,UAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ,SAAS,CAAC;AAAA,IAClE;AACA,QAAI,KAAK,KAAK,UAAU,QAAW;AACjC,UAAI,aAAa,OAAO,SAAS,KAAK,KAAK,MAAM,SAAS,CAAC;AAAA,IAC7D;AAEA,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,UAAI,UAAU;AACd,YAAM,SAAS,CAAC,OAAmB;AACjC,YAAI,CAAC,SAAS;AACZ,oBAAU;AACV,aAAG;AAAA,QACL;AAAA,MACF;AAEA,YAAM,UAAM;AAAA,QACV;AAAA,UACE,UAAU,IAAI;AAAA,UACd,MAAM;AAAA,UACN,MAAM,IAAI,WAAW,IAAI;AAAA,UACzB,QAAQ;AAAA,UACR,SAAS;AAAA,YACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,YAClD,gBAAgB;AAAA,UAClB;AAAA,UACA,QAAQ,KAAK;AAAA,QACf;AAAA,QACA,CAAC,QAAQ;AACP,cAAI,IAAI,eAAe,KAAK;AAC1B;AAAA,cAAO,MACL;AAAA,gBACE,IAAI;AAAA,kBACF,qCAAqC,IAAI,UAAU,IAAI,IAAI,aAAa;AAAA,gBAC1E;AAAA,cACF;AAAA,YACF;AACA;AAAA,UACF;AAEA,cAAI,GAAG,QAAQ,CAAC,UAAU;AACxB,uBAAW,SAAS,QAAQ,MAAM,KAAK,GAAG;AACxC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AAAA,UACF,CAAC;AAED,cAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,gBAAI,IAAI,YAAY,UAAW;AAC/B,iBAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6BAA6B;AACzD,mBAAO,MAAM,OAAO,GAAG,CAAC;AAAA,UAC1B,CAAC;AAED,cAAI,GAAG,SAAS,MAAM;AACpB,uBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AACA,gBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,mBAAK,MAAM,MAAM;AAAA,YACnB;AACA,mBAAO,MAAM,QAAQ,CAAC;AAAA,UACxB,CAAC;AAAA,QACH;AAAA,MACF;AAEA,UAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,YAAI,IAAI,SAAS,aAAc;AAC/B,aAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,4BAA4B;AACxD,eAAO,MAAM,OAAO,GAAG,CAAC;AAAA,MAC1B,CAAC;AAED,UAAI,GAAG,SAAS,MAAM,OAAO,MAAM,QAAQ,CAAC,CAAC;AAC7C,UAAI,MAAM,KAAK,UAAU,IAAI,CAAC;AAC9B,UAAI,IAAI;AAAA,IACV,CAAC;AAAA,EACH;AACF;AAEO,MAAM,yBAAyB,kBAAI,iBAAiB;AAAA,EACjD;AAAA,EACA;AAAA,EACR,cAAU,mBAAI;AAAA,EACd,QAAQ;AAAA,EAER,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EACpE,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EAEpE,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,WAAW;AACtB,SAAK,OAAO;AACZ,SAAK,YAAY,KAAK,kBAAkB,OAAO;AAAA,EACjD;AAAA,EAEA,MAAc,eAAe,IAA8B;AACzD,QAAI;AAGF,UAAI,GAAG,eAAe,oBAAU,MAAM;AACpC,WAAG,KAAK,iBAAiB,SAAS;AAClC,WAAG,KAAK,iBAAiB,SAAS;AAIlC,YAAI;AACF,gBAAM,IAAI,QAAc,CAAC,SAAS,YAAY;AAC5C,kBAAM,UAAU,WAAW,MAAM;AAC/B,sBAAQ;AAAA,YACV,GAAG,GAAI;AAEP,eAAG,KAAK,WAAW,MAAM;AACvB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAAA,UACH,CAAC;AAAA,QACH,SAAS,GAAG;AAAA,QAEZ;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AACV,cAAQ,KAAK,0CAA0C,CAAC,EAAE;AAAA,IAC5D,UAAE;AACA,UAAI,GAAG,eAAe,oBAAU,QAAQ,GAAG,eAAe,oBAAU,YAAY;AAC9E,WAAG,MAAM;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,gBAAY,yBAAU;AAC5B,UAAM,gBAAY,yBAAU;AAE5B,UAAM,QAAQ,KAAK,KAAK,QAAS,QAAQ,SAAS,IAAI;AACtD,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,WAAW;AACvC,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AACtD,QAAI,aAAa,OAAO,eAAe,OAAO,KAAK,KAAK,SAAS,CAAC;AAClE,QAAI,KAAK,KAAK,YAAY,UAAa,KAAK,KAAK,YAAY,MAAM;AACjE,UAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ,SAAS,CAAC;AAAA,IAClE;AACA,QAAI,KAAK,KAAK,UAAU,QAAW;AACjC,UAAI,aAAa,OAAO,SAAS,KAAK,KAAK,MAAM,SAAS,CAAC;AAAA,IAC7D;AAEA,UAAM,KAAK,IAAI,oBAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,MACpD;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,SAAG,GAAG,QAAQ,OAAO;AACrB,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,IAC/D,CAAC;AAED,UAAM,YAAY,YAAY;AAC5B,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,eAAK,UAAU,MAAM;AACrB;AAAA,QACF;AACA,aAAK,UAAU,SAAS,IAAI;AAAA,MAC9B;AACA,WAAK,UAAU,SAAS;AACxB,WAAK,UAAU,MAAM;AAAA,IACvB;AAEA,QAAI,gBAA4B,MAAM;AAAA,IAAC;AACvC,UAAM,YAAY,IAAI,QAAc,CAAC,YAAY;AAC/C,sBAAgB;AAAA,IAClB,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,UAAI;AACF,yBAAiB,SAAS,KAAK,WAAW;AACxC,cAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,cAAI,OAAO,MAAM;AACjB,cAAI,CAAC,KAAK,SAAS,GAAG,GAAG;AACvB,oBAAQ;AAAA,UACV;AAEA,gBAAM,UAAU,KAAK,UAAU;AAAA,YAC7B,MAAM;AAAA,YACN;AAAA,UACF,CAAC;AAED,aAAG,KAAK,OAAO;AACf,wBAAc;AAAA,QAChB;AAEA,YAAI,CAAC,KAAK,gBAAgB,OAAO,SAAS;AACxC,aAAG,KAAK,iBAAiB,SAAS;AAClC,wBAAc;AAAA,QAChB;AAAA,MACF,UAAE;AACA,sBAAc;AAAA,MAChB;AAAA,IACF;AAEA,UAAM,WAAW,YAAY;AAC3B,YAAM,UAAU,IAAI,8BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAI,gBAAgB;AACpB,UAAI,UAAiC;AACrC,UAAI;AAEJ,YAAM,gBAAgB,CAACC,YAAmB,UAAmB;AAC3D,YAAI,aAAa,CAAC,KAAK,MAAM,QAAQ;AACnC,eAAK,MAAM,IAAI,EAAE,WAAW,WAAAA,YAAW,OAAO,WAAW,MAAM,CAAC;AAChE,sBAAY;AAAA,QACd;AAAA,MACF;AAEA,YAAM,sBAAsB,MAAM;AAChC,YAAI,SAAS;AACX,uBAAa,OAAO;AACpB,oBAAU;AAAA,QACZ;AAAA,MACF;AAEA,YAAM,sBAAsB,CAAC,WAAuC;AAClE,4BAAoB;AACpB,kBAAU,WAAW,MAAM;AACzB,iBAAO,IAAI,8BAAgB,EAAE,SAAS,iCAAiC,CAAC,CAAC;AAAA,QAC3E,GAAG,KAAK,YAAY,SAAS;AAAA,MAC/B;AAEA,YAAM;AACN,UAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,4BAAoB,MAAM;AAE1B,WAAG,GAAG,WAAW,CAAC,MAAe,aAAsB;AACrD,8BAAoB;AAEpB,cAAI,CAAC,UAAU;AACb,kBAAM,UAAU,KAAK,MAAM,KAAK,SAAS,CAAC;AAC1C,gBAAI,QAAQ,SAAS,WAAW;AAC9B,8BAAgB;AAChB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,WAAW,KAAK;AAC9B,4BAAY;AAAA,cACd;AACA,4BAAc,WAAW,IAAI;AAE7B,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,cAC/C;AACA,sBAAQ;AACR;AAAA,YACF,WAAW,QAAQ,SAAS,WAAW;AACrC,mBAAK,QAAQ,KAAK,qBAAqB,QAAQ,QAAQ,EAAE;AAAA,YAC3D,WAAW,QAAQ,SAAS,WAAW,QAAQ,SAAS,SAAS;AAC/D,qBAAO,IAAI,uBAAS,+BAA+B,EAAE,MAAM,QAAQ,CAAC,CAAC;AACrE;AAAA,YACF,WAAW,QAAQ,SAAS,YAAY;AACtC,mBAAK,QAAQ,KAAK,EAAE,QAAQ,GAAG,+BAA+B;AAAA,YAChE;AAEA,gCAAoB,MAAM;AAC1B;AAAA,UACF;AAEA,gBAAM,SACJ,gBAAgB,SACZ,KAAK,OAAO,MAAM,KAAK,YAAY,KAAK,aAAa,KAAK,UAAU,IACnE;AACP,qBAAW,SAAS,QAAQ,MAAM,MAAqB,GAAG;AACxD,0BAAc,WAAW,KAAK;AAC9B,wBAAY;AAAA,UACd;AACA,8BAAoB,MAAM;AAAA,QAC5B,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,MAAM,WAAW;AAC/B,8BAAoB;AACpB,cAAI,CAAC,eAAe;AAClB;AAAA,cACE,IAAI,6BAAe;AAAA,gBACjB,SAAS;AAAA,gBACT,SAAS;AAAA,kBACP,YAAY,QAAQ;AAAA,kBACpB,MAAM,EAAE,QAAQ,OAAO,SAAS,EAAE;AAAA,gBACpC;AAAA,cACF,CAAC;AAAA,YACH;AACA;AAAA,UACF;AACA,kBAAQ;AAAA,QACV,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,UAAU;AACxB,8BAAoB;AACpB,iBAAO,KAAK;AAAA,QACd,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,UAAU,GAAG,SAAS,GAAG,SAAS,CAAC,CAAC;AAAA,IACzD,SAAS,GAAG;AACV,UAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,UAAI,aAAa,uBAAU,OAAM;AACjC,YAAM,IAAI,iCAAmB;AAAA,QAC3B,SAAS,kCAAmC,EAAY,WAAW,eAAe;AAAA,MACpF,CAAC;AAAA,IACH,UAAE;AACA,YAAM,KAAK,eAAe,EAAE;AAAA,IAC9B;AAAA,EACF;AACF;","names":["tts","segmentId"]}
package/dist/tts.d.cts CHANGED
@@ -4,7 +4,9 @@ export interface TTSOptions {
4
4
  model: TTSModels | string;
5
5
  encoding: TTSEncoding;
6
6
  sampleRate: number;
7
+ bitRate?: number | null;
7
8
  speed?: number;
9
+ mipOptOut: boolean;
8
10
  apiKey?: string;
9
11
  baseUrl?: string;
10
12
  sentenceTokenizer: tokenize.SentenceTokenizer;
@@ -15,11 +17,13 @@ export declare class TTS extends tts.TTS {
15
17
  label: string;
16
18
  get model(): string;
17
19
  get provider(): string;
20
+ get sampleRate(): number;
18
21
  constructor(opts?: Partial<TTSOptions>);
19
22
  synthesize(text: string, connOptions?: APIConnectOptions, abortSignal?: AbortSignal): tts.ChunkedStream;
20
23
  stream(options?: {
21
24
  connOptions?: APIConnectOptions;
22
25
  }): tts.SynthesizeStream;
26
+ updateOptions(opts: Partial<Pick<TTSOptions, 'model' | 'encoding' | 'sampleRate' | 'bitRate' | 'speed' | 'mipOptOut'>>): void;
23
27
  }
24
28
  export declare class ChunkedStream extends tts.ChunkedStream {
25
29
  #private;