@livekit/agents-plugin-deepgram 1.0.36 → 1.0.38
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/stt.cjs +25 -10
- package/dist/stt.cjs.map +1 -1
- package/dist/stt.d.ts.map +1 -1
- package/dist/stt.js +25 -10
- package/dist/stt.js.map +1 -1
- package/dist/stt_v2.cjs +19 -9
- package/dist/stt_v2.cjs.map +1 -1
- package/dist/stt_v2.d.ts.map +1 -1
- package/dist/stt_v2.js +19 -9
- package/dist/stt_v2.js.map +1 -1
- package/package.json +5 -5
- package/src/stt.ts +27 -8
- package/src/stt_v2.ts +18 -9
package/dist/stt.cjs
CHANGED
|
@@ -55,7 +55,8 @@ class STT extends import_agents.stt.STT {
|
|
|
55
55
|
constructor(opts = defaultSTTOptions) {
|
|
56
56
|
super({
|
|
57
57
|
streaming: true,
|
|
58
|
-
interimResults: opts.interimResults ?? defaultSTTOptions.interimResults
|
|
58
|
+
interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,
|
|
59
|
+
alignedTranscript: "word"
|
|
59
60
|
});
|
|
60
61
|
if (opts.apiKey === void 0 && defaultSTTOptions.apiKey === void 0) {
|
|
61
62
|
throw new Error(
|
|
@@ -276,7 +277,11 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
276
277
|
const isFinal = json["is_final"];
|
|
277
278
|
const isEndpoint = json["speech_final"];
|
|
278
279
|
this.#requestId = requestId;
|
|
279
|
-
const alternatives = liveTranscriptionToSpeechData(
|
|
280
|
+
const alternatives = liveTranscriptionToSpeechData(
|
|
281
|
+
this.#opts.language,
|
|
282
|
+
json,
|
|
283
|
+
this.startTimeOffset
|
|
284
|
+
);
|
|
280
285
|
if (alternatives[0] && alternatives[0].text) {
|
|
281
286
|
if (!this.#speaking) {
|
|
282
287
|
this.#speaking = true;
|
|
@@ -340,15 +345,25 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
340
345
|
this.queue.put(usageEvent);
|
|
341
346
|
}
|
|
342
347
|
}
|
|
343
|
-
const liveTranscriptionToSpeechData = (language, data) => {
|
|
348
|
+
const liveTranscriptionToSpeechData = (language, data, startTimeOffset = 0) => {
|
|
344
349
|
const alts = data["channel"]["alternatives"];
|
|
345
|
-
return alts.map((alt) =>
|
|
346
|
-
|
|
347
|
-
|
|
348
|
-
|
|
349
|
-
|
|
350
|
-
|
|
351
|
-
|
|
350
|
+
return alts.map((alt) => {
|
|
351
|
+
const wordsData = alt["words"] ?? [];
|
|
352
|
+
return {
|
|
353
|
+
language,
|
|
354
|
+
startTime: wordsData.length ? wordsData[0]["start"] + startTimeOffset : startTimeOffset,
|
|
355
|
+
endTime: wordsData.length ? wordsData[wordsData.length - 1]["end"] + startTimeOffset : startTimeOffset,
|
|
356
|
+
confidence: alt["confidence"],
|
|
357
|
+
text: alt["transcript"],
|
|
358
|
+
words: wordsData.map((word) => ({
|
|
359
|
+
text: word["word"] ?? "",
|
|
360
|
+
startTime: (word["start"] ?? 0) + startTimeOffset,
|
|
361
|
+
endTime: (word["end"] ?? 0) + startTimeOffset,
|
|
362
|
+
confidence: word["confidence"] ?? 0,
|
|
363
|
+
startTimeOffset
|
|
364
|
+
}))
|
|
365
|
+
};
|
|
366
|
+
});
|
|
352
367
|
};
|
|
353
368
|
// Annotate the CommonJS export names for ESM import in node:
|
|
354
369
|
0 && (module.exports = {
|
package/dist/stt.cjs.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 log,\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\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\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 dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n mipOptOut: boolean;\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 dictation: false,\n diarize: false,\n numerals: false,\n mipOptOut: false,\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 constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\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 $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${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 = { ...this.#opts, ...opts };\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(API_BASE_URL_V1);\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 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, encodeURIComponent(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(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(this.#opts.language!, json);\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): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => ({\n language,\n startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n }));\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAUO;AAEP,gBAA0B;AAC1B,mBAAkC;AAGlC,MAAM,kBAAkB;AAwBxB,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,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AAAA,EACV,WAAW;AACb;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAE9C,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;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,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;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,kBAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,qBAAO;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,gCAAkB;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,eAAe;AACzC,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,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,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,oBAAU,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,qBAAO;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,mBAAK,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,YAAQ,4BAAa,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,mBAAe,4BAAa,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,mBAAK,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,kBAAI,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,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,+BAAW;AAAA,sBACT,MAAM,kBAAI,gBAAgB;AAAA,oBAC5B,CAAC;AAAA,kBACH;AAEA,sBAAI,SAAS;AACX,+BAAW;AAAA,sBACT,MAAM,kBAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,+BAAW;AAAA,sBACT,MAAM,kBAAI,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,kBAAI,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,mBAAe,4BAAa,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,kBAAI,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,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","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 log,\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\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\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 dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n mipOptOut: boolean;\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 dictation: false,\n diarize: false,\n numerals: false,\n mipOptOut: false,\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 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 $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${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 = { ...this.#opts, ...opts };\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(API_BASE_URL_V1);\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 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, encodeURIComponent(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(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\n return {\n language,\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 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"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAUO;AAEP,gBAA0B;AAC1B,mBAAkC;AAGlC,MAAM,kBAAkB;AAwBxB,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,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AAAA,EACV,WAAW;AACb;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAE9C,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,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;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,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;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,kBAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,qBAAO;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,gCAAkB;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,eAAe;AACzC,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,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,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,oBAAU,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,qBAAO;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,mBAAK,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,YAAQ,4BAAa,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,mBAAe,4BAAa,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,mBAAK,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,kBAAI,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,kBAAI,gBAAgB;AAAA,oBAC5B,CAAC;AAAA,kBACH;AAEA,sBAAI,SAAS;AACX,+BAAW;AAAA,sBACT,MAAM,kBAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,+BAAW;AAAA,sBACT,MAAM,kBAAI,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,kBAAI,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,mBAAe,4BAAa,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,kBAAI,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;AACvB,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAE1C,WAAO;AAAA,MACL;AAAA,MACA,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,IAAI,CAAC,UAAU;AAAA,QAC9B,MAAM,KAAK,MAAM,KAAK;AAAA,QACtB,YAAY,KAAK,OAAO,KAAK,KAAK;AAAA,QAClC,UAAU,KAAK,KAAK,KAAK,KAAK;AAAA,QAC9B,YAAY,KAAK,YAAY,KAAK;AAAA,QAClC;AAAA,MACF,EAAE;AAAA,IACJ;AAAA,EACF,CAAC;AACH;","names":["stt"]}
|
package/dist/stt.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,EACtB,KAAK,WAAW,EAMhB,GAAG,EAEJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,KAAK,EAAE,YAAY,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAI3D,MAAM,WAAW,UAAU;IACzB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,QAAQ,CAAC,EAAE,YAAY,GAAG,MAAM,CAAC;IACjC,cAAc,EAAE,OAAO,CAAC;IACxB,cAAc,EAAE,OAAO,CAAC;IACxB,SAAS,EAAE,OAAO,CAAC;IACnB,KAAK,EAAE,SAAS,CAAC;IACjB,WAAW,EAAE,OAAO,CAAC;IACrB,OAAO,EAAE,OAAO,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,WAAW,EAAE,OAAO,CAAC;IACrB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,QAAQ,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC,EAAE,CAAC;IAC7B,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,eAAe,EAAE,OAAO,CAAC;IACzB,SAAS,EAAE,OAAO,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,QAAQ,EAAE,OAAO,CAAC;IAClB,SAAS,EAAE,OAAO,CAAC;CACpB;AAwBD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAG9B,KAAK,SAAkB;IACvB,OAAO,CAAC,eAAe,CAAyB;gBAEpC,IAAI,GAAE,OAAO,CAAC,UAAU,CAAqB;
|
|
1
|
+
{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,EACtB,KAAK,WAAW,EAMhB,GAAG,EAEJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,KAAK,EAAE,YAAY,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAI3D,MAAM,WAAW,UAAU;IACzB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,QAAQ,CAAC,EAAE,YAAY,GAAG,MAAM,CAAC;IACjC,cAAc,EAAE,OAAO,CAAC;IACxB,cAAc,EAAE,OAAO,CAAC;IACxB,SAAS,EAAE,OAAO,CAAC;IACnB,KAAK,EAAE,SAAS,CAAC;IACjB,WAAW,EAAE,OAAO,CAAC;IACrB,OAAO,EAAE,OAAO,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,WAAW,EAAE,OAAO,CAAC;IACrB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,QAAQ,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC,EAAE,CAAC;IAC7B,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,eAAe,EAAE,OAAO,CAAC;IACzB,SAAS,EAAE,OAAO,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,QAAQ,EAAE,OAAO,CAAC;IAClB,SAAS,EAAE,OAAO,CAAC;CACpB;AAwBD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAG9B,KAAK,SAAkB;IACvB,OAAO,CAAC,eAAe,CAAyB;gBAEpC,IAAI,GAAE,OAAO,CAAC,UAAU,CAAqB;IAwCnD,UAAU,CAAC,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IAI1D,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAIvC,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,YAAY;IAI7D,KAAK;CAGZ;AAED,qBAAa,YAAa,SAAQ,GAAG,CAAC,YAAY;;IAQhD,KAAK,SAA2B;gBAEpB,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU,EAAE,WAAW,CAAC,EAAE,iBAAiB;cAWvD,GAAG;IA0EnB,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IA8LvC,OAAO,CAAC,qBAAqB;CAU9B"}
|
package/dist/stt.js
CHANGED
|
@@ -39,7 +39,8 @@ class STT extends stt.STT {
|
|
|
39
39
|
constructor(opts = defaultSTTOptions) {
|
|
40
40
|
super({
|
|
41
41
|
streaming: true,
|
|
42
|
-
interimResults: opts.interimResults ?? defaultSTTOptions.interimResults
|
|
42
|
+
interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,
|
|
43
|
+
alignedTranscript: "word"
|
|
43
44
|
});
|
|
44
45
|
if (opts.apiKey === void 0 && defaultSTTOptions.apiKey === void 0) {
|
|
45
46
|
throw new Error(
|
|
@@ -260,7 +261,11 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
260
261
|
const isFinal = json["is_final"];
|
|
261
262
|
const isEndpoint = json["speech_final"];
|
|
262
263
|
this.#requestId = requestId;
|
|
263
|
-
const alternatives = liveTranscriptionToSpeechData(
|
|
264
|
+
const alternatives = liveTranscriptionToSpeechData(
|
|
265
|
+
this.#opts.language,
|
|
266
|
+
json,
|
|
267
|
+
this.startTimeOffset
|
|
268
|
+
);
|
|
264
269
|
if (alternatives[0] && alternatives[0].text) {
|
|
265
270
|
if (!this.#speaking) {
|
|
266
271
|
this.#speaking = true;
|
|
@@ -324,15 +329,25 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
324
329
|
this.queue.put(usageEvent);
|
|
325
330
|
}
|
|
326
331
|
}
|
|
327
|
-
const liveTranscriptionToSpeechData = (language, data) => {
|
|
332
|
+
const liveTranscriptionToSpeechData = (language, data, startTimeOffset = 0) => {
|
|
328
333
|
const alts = data["channel"]["alternatives"];
|
|
329
|
-
return alts.map((alt) =>
|
|
330
|
-
|
|
331
|
-
|
|
332
|
-
|
|
333
|
-
|
|
334
|
-
|
|
335
|
-
|
|
334
|
+
return alts.map((alt) => {
|
|
335
|
+
const wordsData = alt["words"] ?? [];
|
|
336
|
+
return {
|
|
337
|
+
language,
|
|
338
|
+
startTime: wordsData.length ? wordsData[0]["start"] + startTimeOffset : startTimeOffset,
|
|
339
|
+
endTime: wordsData.length ? wordsData[wordsData.length - 1]["end"] + startTimeOffset : startTimeOffset,
|
|
340
|
+
confidence: alt["confidence"],
|
|
341
|
+
text: alt["transcript"],
|
|
342
|
+
words: wordsData.map((word) => ({
|
|
343
|
+
text: word["word"] ?? "",
|
|
344
|
+
startTime: (word["start"] ?? 0) + startTimeOffset,
|
|
345
|
+
endTime: (word["end"] ?? 0) + startTimeOffset,
|
|
346
|
+
confidence: word["confidence"] ?? 0,
|
|
347
|
+
startTimeOffset
|
|
348
|
+
}))
|
|
349
|
+
};
|
|
350
|
+
});
|
|
336
351
|
};
|
|
337
352
|
export {
|
|
338
353
|
STT,
|
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 log,\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\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\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 dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n mipOptOut: boolean;\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 dictation: false,\n diarize: false,\n numerals: false,\n mipOptOut: false,\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 constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\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 $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${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 = { ...this.#opts, ...opts };\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(API_BASE_URL_V1);\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 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, encodeURIComponent(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(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(this.#opts.language!, json);\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): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => ({\n language,\n startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n }));\n};\n"],"mappings":"AAGA;AAAA,EAGE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,iBAAiB;AAC1B,SAAS,yBAAyB;AAGlC,MAAM,kBAAkB;AAwBxB,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,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AAAA,EACV,WAAW;AACb;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU,IAAI;AAAA,EACd,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAE9C,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;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,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;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,eAAe;AACzC,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,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,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;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,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,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,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","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 log,\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\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\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 dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n mipOptOut: boolean;\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 dictation: false,\n diarize: false,\n numerals: false,\n mipOptOut: false,\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 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 $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${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 = { ...this.#opts, ...opts };\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(API_BASE_URL_V1);\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 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, encodeURIComponent(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(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\n return {\n language,\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 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"],"mappings":"AAGA;AAAA,EAGE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,iBAAiB;AAC1B,SAAS,yBAAyB;AAGlC,MAAM,kBAAkB;AAwBxB,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,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AAAA,EACV,WAAW;AACb;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU,IAAI;AAAA,EACd,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAE9C,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,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;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,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;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,eAAe;AACzC,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,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,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;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;AACvB,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAE1C,WAAO;AAAA,MACL;AAAA,MACA,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,IAAI,CAAC,UAAU;AAAA,QAC9B,MAAM,KAAK,MAAM,KAAK;AAAA,QACtB,YAAY,KAAK,OAAO,KAAK,KAAK;AAAA,QAClC,UAAU,KAAK,KAAK,KAAK,KAAK;AAAA,QAC9B,YAAY,KAAK,YAAY,KAAK;AAAA,QAClC;AAAA,MACF,EAAE;AAAA,IACJ;AAAA,EACF,CAAC;AACH;","names":["stt"]}
|
package/dist/stt_v2.cjs
CHANGED
|
@@ -74,7 +74,8 @@ class STTv2 extends import_agents.stt.STT {
|
|
|
74
74
|
constructor(opts = {}) {
|
|
75
75
|
super({
|
|
76
76
|
streaming: true,
|
|
77
|
-
interimResults: true
|
|
77
|
+
interimResults: true,
|
|
78
|
+
alignedTranscript: "word"
|
|
78
79
|
});
|
|
79
80
|
this.#opts = { ...defaultSTTv2Options, ...opts };
|
|
80
81
|
const apiKey = opts.apiKey || process.env.DEEPGRAM_API_KEY;
|
|
@@ -297,7 +298,7 @@ class SpeechStreamv2 extends import_agents.stt.SpeechStream {
|
|
|
297
298
|
}
|
|
298
299
|
}
|
|
299
300
|
#sendTranscriptEvent(eventType, data) {
|
|
300
|
-
const alts = parseTranscription(this.#opts.language || "en", data,
|
|
301
|
+
const alts = parseTranscription(this.#opts.language || "en", data, this.startTimeOffset);
|
|
301
302
|
if (alts.length > 0) {
|
|
302
303
|
this.queue.put({
|
|
303
304
|
type: eventType,
|
|
@@ -323,7 +324,6 @@ class SpeechStreamv2 extends import_agents.stt.SpeechStream {
|
|
|
323
324
|
encoding: "linear16",
|
|
324
325
|
mip_opt_out: String(this.#opts.mipOptOut)
|
|
325
326
|
};
|
|
326
|
-
if (this.#opts.language) params.language = this.#opts.language;
|
|
327
327
|
if (this.#opts.eagerEotThreshold)
|
|
328
328
|
params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();
|
|
329
329
|
if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();
|
|
@@ -342,21 +342,31 @@ class SpeechStreamv2 extends import_agents.stt.SpeechStream {
|
|
|
342
342
|
}
|
|
343
343
|
function parseTranscription(language, data, startTimeOffset) {
|
|
344
344
|
const transcript = data.transcript;
|
|
345
|
-
const
|
|
346
|
-
if (!
|
|
345
|
+
const wordsData = data.words || [];
|
|
346
|
+
if (!wordsData || wordsData.length === 0) {
|
|
347
347
|
return [];
|
|
348
348
|
}
|
|
349
349
|
let confidence = 0;
|
|
350
|
-
if (
|
|
351
|
-
const sum =
|
|
352
|
-
confidence = sum /
|
|
350
|
+
if (wordsData.length > 0) {
|
|
351
|
+
const sum = wordsData.reduce((acc, w) => acc + (w.confidence || 0), 0);
|
|
352
|
+
confidence = sum / wordsData.length;
|
|
353
353
|
}
|
|
354
354
|
const sd = {
|
|
355
355
|
language,
|
|
356
356
|
startTime: (data.audio_window_start || 0) + startTimeOffset,
|
|
357
357
|
endTime: (data.audio_window_end || 0) + startTimeOffset,
|
|
358
358
|
confidence,
|
|
359
|
-
text: transcript || ""
|
|
359
|
+
text: transcript || "",
|
|
360
|
+
// Note: Deepgram V2 (Flux) API does not provide word-level timing (start/end).
|
|
361
|
+
// Words only contain 'word' and 'confidence' fields, so startTime/endTime will be 0.
|
|
362
|
+
// See: https://developers.deepgram.com/docs/flux/nova-3-migration
|
|
363
|
+
words: wordsData.map((word) => ({
|
|
364
|
+
text: word.word ?? "",
|
|
365
|
+
startTime: (word.start ?? 0) + startTimeOffset,
|
|
366
|
+
endTime: (word.end ?? 0) + startTimeOffset,
|
|
367
|
+
confidence: word.confidence ?? 0,
|
|
368
|
+
startTimeOffset
|
|
369
|
+
}))
|
|
360
370
|
};
|
|
361
371
|
return [sd];
|
|
362
372
|
}
|
package/dist/stt_v2.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt_v2.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n AudioByteStream,\n Event,\n calculateAudioDurationSeconds,\n log,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport * as queryString from 'node:querystring';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { V2Models } from './models.js';\n\nconst _CLOSE_MSG = JSON.stringify({ type: 'CloseStream' });\n\n// --- Configuration ---\n\n/**\n * Configuration options for STTv2 (Deepgram Flux model).\n */\nexport interface STTv2Options {\n apiKey?: string;\n model: V2Models | string;\n sampleRate: number;\n keyterms: string[];\n endpointUrl: string;\n language?: string;\n eagerEotThreshold?: number;\n eotThreshold?: number;\n eotTimeoutMs?: number;\n mipOptOut?: boolean;\n tags?: string[];\n}\n\nconst defaultSTTv2Options: Omit<STTv2Options, 'apiKey'> = {\n model: 'flux-general-en',\n sampleRate: 16000,\n keyterms: [],\n endpointUrl: 'wss://api.deepgram.com/v2/listen',\n language: 'en',\n mipOptOut: false,\n};\n\nfunction validateTags(tags: string[]): string[] {\n for (const tag of tags) {\n if (tag.length > 128) {\n throw new Error('tag must be no more than 128 characters');\n }\n }\n return tags;\n}\n\n/**\n * Deepgram STTv2 using the Flux model for streaming speech-to-text.\n *\n * This uses Deepgram's V2 API (`/v2/listen`) which provides turn-based\n * transcription with support for preemptive generation.\n *\n * @remarks\n * Key differences from STT (V1):\n * - Uses `TurnInfo` events instead of `SpeechStarted`/`Results`\n * - Supports `eagerEotThreshold` for preemptive LLM generation\n * - Sends `PREFLIGHT_TRANSCRIPT` events when eager end-of-turn is detected\n *\n * @example\n * ```typescript\n * import { STTv2 } from '@livekit/agents-plugin-deepgram';\n *\n * const stt = new STTv2({\n * model: 'flux-general-en',\n * eagerEotThreshold: 0.5, // Enable preemptive generation\n * });\n *\n * const stream = stt.stream();\n * stream.pushFrame(audioFrame);\n *\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives?.[0]?.text);\n * }\n * }\n * ```\n */\nexport class STTv2 extends stt.STT {\n readonly label = 'deepgram.STTv2';\n #opts: STTv2Options;\n #apiKey: string;\n #logger = log();\n\n /**\n * Create a new Deepgram STTv2 instance.\n *\n * @param opts - Configuration options\n * @param opts.apiKey - Deepgram API key (defaults to `DEEPGRAM_API_KEY` env var)\n * @param opts.model - Model to use (default: `flux-general-en`)\n * @param opts.eagerEotThreshold - Threshold (0.3-0.9) for preemptive generation\n * @param opts.eotThreshold - End-of-turn detection threshold (default: 0.7)\n * @param opts.eotTimeoutMs - End-of-turn timeout in ms (default: 3000)\n * @param opts.keyterms - List of key terms to improve recognition\n * @param opts.tags - Tags for usage reporting (max 128 chars each)\n *\n * @throws Error if no API key is provided\n */\n constructor(opts: Partial<STTv2Options> = {}) {\n super({\n streaming: true,\n interimResults: true,\n });\n\n this.#opts = { ...defaultSTTv2Options, ...opts };\n\n const apiKey = opts.apiKey || process.env.DEEPGRAM_API_KEY;\n if (!apiKey) {\n throw new Error('Deepgram API key is required');\n }\n this.#apiKey = apiKey;\n\n if (this.#opts.tags) {\n this.#opts.tags = validateTags(this.#opts.tags);\n }\n }\n\n /** The model being used for transcription */\n get model(): string {\n return this.#opts.model;\n }\n\n /** The STT provider name */\n get provider(): string {\n return 'Deepgram';\n }\n\n protected async _recognize(\n _frame: AudioFrame | AudioFrame[],\n _abortSignal?: AbortSignal,\n ): Promise<stt.SpeechEvent> {\n throw new Error('V2 API does not support non-streaming recognize. Use .stream()');\n }\n\n /**\n * Create a new streaming transcription session.\n *\n * @param options - Stream options\n * @returns A SpeechStream that emits transcription events\n */\n stream(options?: { connOptions?: APIConnectOptions }): stt.SpeechStream {\n const streamOpts = { ...this.#opts, apiKey: this.#apiKey };\n return new SpeechStreamv2(this, streamOpts, options?.connOptions);\n }\n\n /**\n * Update STT options. Changes will take effect on the next stream.\n *\n * @param opts - Partial options to update\n */\n updateOptions(opts: Partial<STTv2Options>) {\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n this.#logger.debug('Updated STTv2 options');\n }\n}\n\n// --- Stream Implementation ---\n\nclass SpeechStreamv2 extends stt.SpeechStream {\n readonly label = 'deepgram.SpeechStreamv2';\n #opts: STTv2Options & { apiKey: string };\n #logger = log();\n #ws: WebSocket | null = null;\n\n #audioDurationCollector: PeriodicCollector<number>;\n #requestId = '';\n #speaking = false;\n\n // Parity: _reconnect_event - using existing Event class from @livekit/agents\n #reconnectEvent = new Event();\n\n constructor(\n sttInstance: STTv2,\n opts: STTv2Options & { apiKey: string },\n connOptions?: APIConnectOptions,\n ) {\n super(sttInstance, opts.sampleRate, connOptions);\n this.#opts = opts;\n\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.#onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n updateOptions(opts: Partial<STTv2Options>) {\n this.#logger.debug('Stream received option update', opts);\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n\n // Trigger reconnection loop\n this.#reconnectEvent.set();\n }\n\n protected async run() {\n // Outer Loop: Handles reconnections (Configuration updates)\n while (!this.closed) {\n try {\n this.#reconnectEvent.clear();\n\n const url = this.#getDeepgramUrl();\n this.#logger.debug(`Connecting to Deepgram: ${url}`);\n\n this.#ws = new WebSocket(url, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n // 1. Wait for Connection Open\n await new Promise<void>((resolve, reject) => {\n if (!this.#ws) return reject(new Error('WebSocket not initialized'));\n\n const onOpen = () => {\n this.#ws?.off('error', onError);\n resolve();\n };\n const onError = (err: Error) => {\n this.#ws?.off('open', onOpen);\n reject(err);\n };\n\n this.#ws.once('open', onOpen);\n this.#ws.once('error', onError);\n });\n\n // 2. Run Concurrent Tasks (Send & Receive)\n const sendPromise = this.#sendTask();\n const recvPromise = this.#recvTask();\n const reconnectWait = this.#reconnectEvent.wait();\n\n // 3. Race: Normal Completion vs Reconnect Signal\n const result = await Promise.race([\n Promise.all([sendPromise, recvPromise]),\n reconnectWait.then(() => 'RECONNECT'),\n ]);\n\n if (result === 'RECONNECT') {\n this.#logger.debug('Reconnecting stream due to option update...');\n // Close current socket; loop will restart and open a new one\n this.#ws.close();\n } else {\n // Normal finish (Stream ended or Error thrown)\n break;\n }\n } catch (error) {\n this.#logger.error('Deepgram stream error', { error });\n throw error; // Let Base Class handle retry logic\n } finally {\n if (this.#ws?.readyState === WebSocket.OPEN) {\n this.#ws.close();\n }\n }\n }\n this.close();\n }\n\n async #sendTask() {\n if (!this.#ws) return;\n\n // Buffer audio into 50ms chunks (Parity)\n const samples50ms = Math.floor(this.#opts.sampleRate / 20);\n const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samples50ms);\n\n let hasEnded = false;\n\n // Manual Iterator to allow racing against Reconnect Signal\n const iterator = this.input[Symbol.asyncIterator]();\n\n while (true) {\n const nextPromise = iterator.next();\n // If reconnect signal fires, abort the wait\n const abortPromise = this.#reconnectEvent.wait().then(() => ({ abort: true }) as const);\n\n const result = await Promise.race([nextPromise, abortPromise]);\n\n // Check if we need to abort (Reconnect) or if stream is done\n if ('abort' in result || result.done) {\n if (!('abort' in result) && result.done) {\n // Normal stream end\n hasEnded = true;\n } else {\n // Reconnect triggered - break loop immediately\n break;\n }\n }\n\n // If we broke above, we don't process data. If not, 'result' is IteratorResult\n if (hasEnded && !('value' in result)) {\n // Process flush below\n } else if ('value' in result) {\n const data = result.value;\n const frames: AudioFrame[] = [];\n\n if (data === SpeechStreamv2.FLUSH_SENTINEL) {\n frames.push(...audioBstream.flush());\n hasEnded = true;\n } else {\n frames.push(...audioBstream.write((data as AudioFrame).data.buffer as ArrayBuffer));\n }\n\n for (const frame of frames) {\n this.#audioDurationCollector.push(calculateAudioDurationSeconds(frame));\n\n if (this.#ws!.readyState === WebSocket.OPEN) {\n this.#ws!.send(frame.data);\n }\n\n if (hasEnded) {\n this.#audioDurationCollector.flush();\n hasEnded = false;\n }\n }\n }\n\n if (hasEnded) break;\n }\n\n // Only send CloseStream if we are exiting normally (not reconnecting)\n if (!this.#reconnectEvent.isSet && this.#ws!.readyState === WebSocket.OPEN) {\n this.#logger.debug('Sending CloseStream message to Deepgram');\n this.#ws!.send(_CLOSE_MSG);\n }\n }\n\n async #recvTask() {\n if (!this.#ws) return;\n\n return new Promise<void>((resolve) => {\n if (!this.#ws) return resolve();\n\n this.#ws.on('message', (data: Buffer, isBinary: boolean) => {\n if (isBinary) {\n this.#logger.warn('Received unexpected binary message from Deepgram');\n return;\n }\n try {\n const msg = JSON.parse(data.toString());\n this.#processStreamEvent(msg);\n } catch (error) {\n this.#logger.error('Failed to parse Deepgram message', { error });\n }\n });\n\n this.#ws.on('close', (code, reason) => {\n this.#logger.debug(`Deepgram WebSocket closed: ${code} ${reason}`);\n resolve();\n });\n\n // Errors are caught by run() listener, resolve here to clean up task\n this.#ws.on('error', () => resolve());\n });\n }\n\n #processStreamEvent(data: Record<string, unknown>) {\n if (data.request_id) {\n this.#requestId = data.request_id as string;\n }\n\n if (data.type === 'TurnInfo') {\n const eventType = data.event;\n\n if (eventType === 'StartOfTurn') {\n if (this.#speaking) return;\n\n this.#speaking = true;\n this.queue.put({\n type: stt.SpeechEventType.START_OF_SPEECH,\n requestId: this.#requestId,\n });\n\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'Update') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EagerEndOfTurn') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, data);\n } else if (eventType === 'TurnResumed') {\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EndOfTurn') {\n if (!this.#speaking) return;\n\n this.#speaking = false;\n this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, data);\n\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n }\n } else if (data.type === 'Error') {\n this.#logger.warn('deepgram sent an error', { data });\n const desc = (data.description as string) || 'unknown error from deepgram';\n throw new Error(`Deepgram API Error: ${desc}`);\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {\n // Note: start_time_offset is not yet available in the TypeScript base class\n // Using 0.0 for now - full parity would require base class changes\n const alts = parseTranscription(this.#opts.language || 'en', data, 0.0);\n\n if (alts.length > 0) {\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [alts[0]!, ...alts.slice(1)],\n });\n }\n }\n\n #onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n\n #getDeepgramUrl(): string {\n const params: Record<string, string | string[]> = {\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: 'linear16',\n mip_opt_out: String(this.#opts.mipOptOut),\n };\n\n if (this.#opts.language) params.language = this.#opts.language;\n if (this.#opts.eagerEotThreshold)\n params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();\n if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();\n if (this.#opts.eotTimeoutMs) params.eot_timeout_ms = this.#opts.eotTimeoutMs.toString();\n\n if (this.#opts.keyterms.length > 0) params.keyterm = this.#opts.keyterms;\n if (this.#opts.tags && this.#opts.tags.length > 0) params.tag = this.#opts.tags;\n\n const baseUrl = this.#opts.endpointUrl.replace(/^http/, 'ws');\n const qs = queryString.stringify(params);\n return `${baseUrl}?${qs}`;\n }\n\n override close() {\n super.close();\n this.#ws?.close();\n }\n}\n\n// --- Helpers ---\n\nfunction parseTranscription(\n language: string,\n data: Record<string, unknown>,\n startTimeOffset: number,\n): stt.SpeechData[] {\n const transcript = data.transcript as string | undefined;\n const words = (data.words as Array<Record<string, unknown>>) || [];\n\n if (!words || words.length === 0) {\n return [];\n }\n\n let confidence = 0;\n if (words.length > 0) {\n const sum = words.reduce((acc: number, w) => acc + ((w.confidence as number) || 0), 0);\n confidence = sum / words.length;\n }\n\n const sd: stt.SpeechData = {\n language: language,\n startTime: ((data.audio_window_start as number) || 0) + startTimeOffset,\n endTime: ((data.audio_window_end as number) || 0) + startTimeOffset,\n confidence: confidence,\n text: transcript || '',\n };\n\n return [sd];\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAOO;AAEP,kBAA6B;AAC7B,gBAA0B;AAC1B,mBAAkC;AAGlC,MAAM,aAAa,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC;AAqBzD,MAAM,sBAAoD;AAAA,EACxD,OAAO;AAAA,EACP,YAAY;AAAA,EACZ,UAAU,CAAC;AAAA,EACX,aAAa;AAAA,EACb,UAAU;AAAA,EACV,WAAW;AACb;AAEA,SAAS,aAAa,MAA0B;AAC9C,aAAW,OAAO,MAAM;AACtB,QAAI,IAAI,SAAS,KAAK;AACpB,YAAM,IAAI,MAAM,yCAAyC;AAAA,IAC3D;AAAA,EACF;AACA,SAAO;AACT;AAiCO,MAAM,cAAc,kBAAI,IAAI;AAAA,EACxB,QAAQ;AAAA,EACjB;AAAA,EACA;AAAA,EACA,cAAU,mBAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBd,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,IAClB,CAAC;AAED,SAAK,QAAQ,EAAE,GAAG,qBAAqB,GAAG,KAAK;AAE/C,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,8BAA8B;AAAA,IAChD;AACA,SAAK,UAAU;AAEf,QAAI,KAAK,MAAM,MAAM;AACnB,WAAK,MAAM,OAAO,aAAa,KAAK,MAAM,IAAI;AAAA,IAChD;AAAA,EACF;AAAA;AAAA,EAGA,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,MAAgB,WACd,QACA,cAC0B;AAC1B,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,OAAO,SAAiE;AACtE,UAAM,aAAa,EAAE,GAAG,KAAK,OAAO,QAAQ,KAAK,QAAQ;AACzD,WAAO,IAAI,eAAe,MAAM,YAAY,mCAAS,WAAW;AAAA,EAClE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAA6B;AACzC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AACvD,SAAK,QAAQ,MAAM,uBAAuB;AAAA,EAC5C;AACF;AAIA,MAAM,uBAAuB,kBAAI,aAAa;AAAA,EACnC,QAAQ;AAAA,EACjB;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,MAAwB;AAAA,EAExB;AAAA,EACA,aAAa;AAAA,EACb,YAAY;AAAA;AAAA,EAGZ,kBAAkB,IAAI,oBAAM;AAAA,EAE5B,YACE,aACA,MACA,aACA;AACA,UAAM,aAAa,KAAK,YAAY,WAAW;AAC/C,SAAK,QAAQ;AAEb,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,uBAAuB,QAAQ;AAAA,MAClD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,cAAc,MAA6B;AACzC,SAAK,QAAQ,MAAM,iCAAiC,IAAI;AACxD,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AAGvD,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,MAAgB,MAAM;AA5MxB;AA8MI,WAAO,CAAC,KAAK,QAAQ;AACnB,UAAI;AACF,aAAK,gBAAgB,MAAM;AAE3B,cAAM,MAAM,KAAK,gBAAgB;AACjC,aAAK,QAAQ,MAAM,2BAA2B,GAAG,EAAE;AAEnD,aAAK,MAAM,IAAI,oBAAU,KAAK;AAAA,UAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,QACzD,CAAC;AAGD,cAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,cAAI,CAAC,KAAK,IAAK,QAAO,OAAO,IAAI,MAAM,2BAA2B,CAAC;AAEnE,gBAAM,SAAS,MAAM;AA7N/B,gBAAAA;AA8NY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,SAAS;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,UAAU,CAAC,QAAe;AAjO1C,gBAAAA;AAkOY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,QAAQ;AACtB,mBAAO,GAAG;AAAA,UACZ;AAEA,eAAK,IAAI,KAAK,QAAQ,MAAM;AAC5B,eAAK,IAAI,KAAK,SAAS,OAAO;AAAA,QAChC,CAAC;AAGD,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,gBAAgB,KAAK,gBAAgB,KAAK;AAGhD,cAAM,SAAS,MAAM,QAAQ,KAAK;AAAA,UAChC,QAAQ,IAAI,CAAC,aAAa,WAAW,CAAC;AAAA,UACtC,cAAc,KAAK,MAAM,WAAW;AAAA,QACtC,CAAC;AAED,YAAI,WAAW,aAAa;AAC1B,eAAK,QAAQ,MAAM,6CAA6C;AAEhE,eAAK,IAAI,MAAM;AAAA,QACjB,OAAO;AAEL;AAAA,QACF;AAAA,MACF,SAAS,OAAO;AACd,aAAK,QAAQ,MAAM,yBAAyB,EAAE,MAAM,CAAC;AACrD,cAAM;AAAA,MACR,UAAE;AACA,cAAI,UAAK,QAAL,mBAAU,gBAAe,oBAAU,MAAM;AAC3C,eAAK,IAAI,MAAM;AAAA,QACjB;AAAA,MACF;AAAA,IACF;AACA,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAGf,UAAM,cAAc,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AACzD,UAAM,eAAe,IAAI,8BAAgB,KAAK,MAAM,YAAY,GAAG,WAAW;AAE9E,QAAI,WAAW;AAGf,UAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAElD,WAAO,MAAM;AACX,YAAM,cAAc,SAAS,KAAK;AAElC,YAAM,eAAe,KAAK,gBAAgB,KAAK,EAAE,KAAK,OAAO,EAAE,OAAO,KAAK,EAAW;AAEtF,YAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,aAAa,YAAY,CAAC;AAG7D,UAAI,WAAW,UAAU,OAAO,MAAM;AACpC,YAAI,EAAE,WAAW,WAAW,OAAO,MAAM;AAEvC,qBAAW;AAAA,QACb,OAAO;AAEL;AAAA,QACF;AAAA,MACF;AAGA,UAAI,YAAY,EAAE,WAAW,SAAS;AAAA,MAEtC,WAAW,WAAW,QAAQ;AAC5B,cAAM,OAAO,OAAO;AACpB,cAAM,SAAuB,CAAC;AAE9B,YAAI,SAAS,eAAe,gBAAgB;AAC1C,iBAAO,KAAK,GAAG,aAAa,MAAM,CAAC;AACnC,qBAAW;AAAA,QACb,OAAO;AACL,iBAAO,KAAK,GAAG,aAAa,MAAO,KAAoB,KAAK,MAAqB,CAAC;AAAA,QACpF;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,wBAAwB,SAAK,6CAA8B,KAAK,CAAC;AAEtE,cAAI,KAAK,IAAK,eAAe,oBAAU,MAAM;AAC3C,iBAAK,IAAK,KAAK,MAAM,IAAI;AAAA,UAC3B;AAEA,cAAI,UAAU;AACZ,iBAAK,wBAAwB,MAAM;AACnC,uBAAW;AAAA,UACb;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAU;AAAA,IAChB;AAGA,QAAI,CAAC,KAAK,gBAAgB,SAAS,KAAK,IAAK,eAAe,oBAAU,MAAM;AAC1E,WAAK,QAAQ,MAAM,yCAAyC;AAC5D,WAAK,IAAK,KAAK,UAAU;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAEf,WAAO,IAAI,QAAc,CAAC,YAAY;AACpC,UAAI,CAAC,KAAK,IAAK,QAAO,QAAQ;AAE9B,WAAK,IAAI,GAAG,WAAW,CAAC,MAAc,aAAsB;AAC1D,YAAI,UAAU;AACZ,eAAK,QAAQ,KAAK,kDAAkD;AACpE;AAAA,QACF;AACA,YAAI;AACF,gBAAM,MAAM,KAAK,MAAM,KAAK,SAAS,CAAC;AACtC,eAAK,oBAAoB,GAAG;AAAA,QAC9B,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,oCAAoC,EAAE,MAAM,CAAC;AAAA,QAClE;AAAA,MACF,CAAC;AAED,WAAK,IAAI,GAAG,SAAS,CAAC,MAAM,WAAW;AACrC,aAAK,QAAQ,MAAM,8BAA8B,IAAI,IAAI,MAAM,EAAE;AACjE,gBAAQ;AAAA,MACV,CAAC;AAGD,WAAK,IAAI,GAAG,SAAS,MAAM,QAAQ,CAAC;AAAA,IACtC,CAAC;AAAA,EACH;AAAA,EAEA,oBAAoB,MAA+B;AACjD,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,QAAI,KAAK,SAAS,YAAY;AAC5B,YAAM,YAAY,KAAK;AAEvB,UAAI,cAAc,eAAe;AAC/B,YAAI,KAAK,UAAW;AAEpB,aAAK,YAAY;AACjB,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,kBAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAED,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,UAAU;AACjC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,kBAAkB;AACzC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,kBAAI,gBAAgB,sBAAsB,IAAI;AAAA,MAC1E,WAAW,cAAc,eAAe;AACtC,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,aAAa;AACpC,YAAI,CAAC,KAAK,UAAW;AAErB,aAAK,YAAY;AACjB,aAAK,qBAAqB,kBAAI,gBAAgB,kBAAkB,IAAI;AAEpE,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,kBAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAAA,MACH;AAAA,IACF,WAAW,KAAK,SAAS,SAAS;AAChC,WAAK,QAAQ,KAAK,0BAA0B,EAAE,KAAK,CAAC;AACpD,YAAM,OAAQ,KAAK,eAA0B;AAC7C,YAAM,IAAI,MAAM,uBAAuB,IAAI,EAAE;AAAA,IAC/C;AAAA,EACF;AAAA,EAEA,qBAAqB,WAAgC,MAA+B;AAGlF,UAAM,OAAO,mBAAmB,KAAK,MAAM,YAAY,MAAM,MAAM,CAAG;AAEtE,QAAI,KAAK,SAAS,GAAG;AACnB,WAAK,MAAM,IAAI;AAAA,QACb,MAAM;AAAA,QACN,WAAW,KAAK;AAAA,QAChB,cAAc,CAAC,KAAK,CAAC,GAAI,GAAG,KAAK,MAAM,CAAC,CAAC;AAAA,MAC3C,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,uBAAuB,UAAkB;AACvC,UAAM,aAA8B;AAAA,MAClC,MAAM,kBAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AAAA,EAEA,kBAA0B;AACxB,UAAM,SAA4C;AAAA,MAChD,OAAO,KAAK,MAAM;AAAA,MAClB,aAAa,KAAK,MAAM,WAAW,SAAS;AAAA,MAC5C,UAAU;AAAA,MACV,aAAa,OAAO,KAAK,MAAM,SAAS;AAAA,IAC1C;AAEA,QAAI,KAAK,MAAM,SAAU,QAAO,WAAW,KAAK,MAAM;AACtD,QAAI,KAAK,MAAM;AACb,aAAO,sBAAsB,KAAK,MAAM,kBAAkB,SAAS;AACrE,QAAI,KAAK,MAAM,aAAc,QAAO,gBAAgB,KAAK,MAAM,aAAa,SAAS;AACrF,QAAI,KAAK,MAAM,aAAc,QAAO,iBAAiB,KAAK,MAAM,aAAa,SAAS;AAEtF,QAAI,KAAK,MAAM,SAAS,SAAS,EAAG,QAAO,UAAU,KAAK,MAAM;AAChE,QAAI,KAAK,MAAM,QAAQ,KAAK,MAAM,KAAK,SAAS,EAAG,QAAO,MAAM,KAAK,MAAM;AAE3E,UAAM,UAAU,KAAK,MAAM,YAAY,QAAQ,SAAS,IAAI;AAC5D,UAAM,KAAK,YAAY,UAAU,MAAM;AACvC,WAAO,GAAG,OAAO,IAAI,EAAE;AAAA,EACzB;AAAA,EAES,QAAQ;AArcnB;AAscI,UAAM,MAAM;AACZ,eAAK,QAAL,mBAAU;AAAA,EACZ;AACF;AAIA,SAAS,mBACP,UACA,MACA,iBACkB;AAClB,QAAM,aAAa,KAAK;AACxB,QAAM,QAAS,KAAK,SAA4C,CAAC;AAEjE,MAAI,CAAC,SAAS,MAAM,WAAW,GAAG;AAChC,WAAO,CAAC;AAAA,EACV;AAEA,MAAI,aAAa;AACjB,MAAI,MAAM,SAAS,GAAG;AACpB,UAAM,MAAM,MAAM,OAAO,CAAC,KAAa,MAAM,OAAQ,EAAE,cAAyB,IAAI,CAAC;AACrF,iBAAa,MAAM,MAAM;AAAA,EAC3B;AAEA,QAAM,KAAqB;AAAA,IACzB;AAAA,IACA,YAAa,KAAK,sBAAiC,KAAK;AAAA,IACxD,UAAW,KAAK,oBAA+B,KAAK;AAAA,IACpD;AAAA,IACA,MAAM,cAAc;AAAA,EACtB;AAEA,SAAO,CAAC,EAAE;AACZ;","names":["_a"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt_v2.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n AudioByteStream,\n Event,\n calculateAudioDurationSeconds,\n log,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport * as queryString from 'node:querystring';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { V2Models } from './models.js';\n\nconst _CLOSE_MSG = JSON.stringify({ type: 'CloseStream' });\n\n// --- Configuration ---\n\n/**\n * Configuration options for STTv2 (Deepgram Flux model).\n */\nexport interface STTv2Options {\n apiKey?: string;\n model: V2Models | string;\n sampleRate: number;\n keyterms: string[];\n endpointUrl: string;\n language?: string;\n eagerEotThreshold?: number;\n eotThreshold?: number;\n eotTimeoutMs?: number;\n mipOptOut?: boolean;\n tags?: string[];\n}\n\nconst defaultSTTv2Options: Omit<STTv2Options, 'apiKey'> = {\n model: 'flux-general-en',\n sampleRate: 16000,\n keyterms: [],\n endpointUrl: 'wss://api.deepgram.com/v2/listen',\n language: 'en',\n mipOptOut: false,\n};\n\nfunction validateTags(tags: string[]): string[] {\n for (const tag of tags) {\n if (tag.length > 128) {\n throw new Error('tag must be no more than 128 characters');\n }\n }\n return tags;\n}\n\n/**\n * Deepgram STTv2 using the Flux model for streaming speech-to-text.\n *\n * This uses Deepgram's V2 API (`/v2/listen`) which provides turn-based\n * transcription with support for preemptive generation.\n *\n * @remarks\n * Key differences from STT (V1):\n * - Uses `TurnInfo` events instead of `SpeechStarted`/`Results`\n * - Supports `eagerEotThreshold` for preemptive LLM generation\n * - Sends `PREFLIGHT_TRANSCRIPT` events when eager end-of-turn is detected\n *\n * @example\n * ```typescript\n * import { STTv2 } from '@livekit/agents-plugin-deepgram';\n *\n * const stt = new STTv2({\n * model: 'flux-general-en',\n * eagerEotThreshold: 0.5, // Enable preemptive generation\n * });\n *\n * const stream = stt.stream();\n * stream.pushFrame(audioFrame);\n *\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives?.[0]?.text);\n * }\n * }\n * ```\n */\nexport class STTv2 extends stt.STT {\n readonly label = 'deepgram.STTv2';\n #opts: STTv2Options;\n #apiKey: string;\n #logger = log();\n\n /**\n * Create a new Deepgram STTv2 instance.\n *\n * @param opts - Configuration options\n * @param opts.apiKey - Deepgram API key (defaults to `DEEPGRAM_API_KEY` env var)\n * @param opts.model - Model to use (default: `flux-general-en`)\n * @param opts.eagerEotThreshold - Threshold (0.3-0.9) for preemptive generation\n * @param opts.eotThreshold - End-of-turn detection threshold (default: 0.7)\n * @param opts.eotTimeoutMs - End-of-turn timeout in ms (default: 3000)\n * @param opts.keyterms - List of key terms to improve recognition\n * @param opts.tags - Tags for usage reporting (max 128 chars each)\n *\n * @throws Error if no API key is provided\n */\n constructor(opts: Partial<STTv2Options> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: 'word',\n });\n\n this.#opts = { ...defaultSTTv2Options, ...opts };\n\n const apiKey = opts.apiKey || process.env.DEEPGRAM_API_KEY;\n if (!apiKey) {\n throw new Error('Deepgram API key is required');\n }\n this.#apiKey = apiKey;\n\n if (this.#opts.tags) {\n this.#opts.tags = validateTags(this.#opts.tags);\n }\n }\n\n /** The model being used for transcription */\n get model(): string {\n return this.#opts.model;\n }\n\n /** The STT provider name */\n get provider(): string {\n return 'Deepgram';\n }\n\n protected async _recognize(\n _frame: AudioFrame | AudioFrame[],\n _abortSignal?: AbortSignal,\n ): Promise<stt.SpeechEvent> {\n throw new Error('V2 API does not support non-streaming recognize. Use .stream()');\n }\n\n /**\n * Create a new streaming transcription session.\n *\n * @param options - Stream options\n * @returns A SpeechStream that emits transcription events\n */\n stream(options?: { connOptions?: APIConnectOptions }): stt.SpeechStream {\n const streamOpts = { ...this.#opts, apiKey: this.#apiKey };\n return new SpeechStreamv2(this, streamOpts, options?.connOptions);\n }\n\n /**\n * Update STT options. Changes will take effect on the next stream.\n *\n * @param opts - Partial options to update\n */\n updateOptions(opts: Partial<STTv2Options>) {\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n this.#logger.debug('Updated STTv2 options');\n }\n}\n\n// --- Stream Implementation ---\n\nclass SpeechStreamv2 extends stt.SpeechStream {\n readonly label = 'deepgram.SpeechStreamv2';\n #opts: STTv2Options & { apiKey: string };\n #logger = log();\n #ws: WebSocket | null = null;\n\n #audioDurationCollector: PeriodicCollector<number>;\n #requestId = '';\n #speaking = false;\n\n // Parity: _reconnect_event - using existing Event class from @livekit/agents\n #reconnectEvent = new Event();\n\n constructor(\n sttInstance: STTv2,\n opts: STTv2Options & { apiKey: string },\n connOptions?: APIConnectOptions,\n ) {\n super(sttInstance, opts.sampleRate, connOptions);\n this.#opts = opts;\n\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.#onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n updateOptions(opts: Partial<STTv2Options>) {\n this.#logger.debug('Stream received option update', opts);\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n\n // Trigger reconnection loop\n this.#reconnectEvent.set();\n }\n\n protected async run() {\n // Outer Loop: Handles reconnections (Configuration updates)\n while (!this.closed) {\n try {\n this.#reconnectEvent.clear();\n\n const url = this.#getDeepgramUrl();\n this.#logger.debug(`Connecting to Deepgram: ${url}`);\n\n this.#ws = new WebSocket(url, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n // 1. Wait for Connection Open\n await new Promise<void>((resolve, reject) => {\n if (!this.#ws) return reject(new Error('WebSocket not initialized'));\n\n const onOpen = () => {\n this.#ws?.off('error', onError);\n resolve();\n };\n const onError = (err: Error) => {\n this.#ws?.off('open', onOpen);\n reject(err);\n };\n\n this.#ws.once('open', onOpen);\n this.#ws.once('error', onError);\n });\n\n // 2. Run Concurrent Tasks (Send & Receive)\n const sendPromise = this.#sendTask();\n const recvPromise = this.#recvTask();\n const reconnectWait = this.#reconnectEvent.wait();\n\n // 3. Race: Normal Completion vs Reconnect Signal\n const result = await Promise.race([\n Promise.all([sendPromise, recvPromise]),\n reconnectWait.then(() => 'RECONNECT'),\n ]);\n\n if (result === 'RECONNECT') {\n this.#logger.debug('Reconnecting stream due to option update...');\n // Close current socket; loop will restart and open a new one\n this.#ws.close();\n } else {\n // Normal finish (Stream ended or Error thrown)\n break;\n }\n } catch (error) {\n this.#logger.error('Deepgram stream error', { error });\n throw error; // Let Base Class handle retry logic\n } finally {\n if (this.#ws?.readyState === WebSocket.OPEN) {\n this.#ws.close();\n }\n }\n }\n this.close();\n }\n\n async #sendTask() {\n if (!this.#ws) return;\n\n // Buffer audio into 50ms chunks (Parity)\n const samples50ms = Math.floor(this.#opts.sampleRate / 20);\n const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samples50ms);\n\n let hasEnded = false;\n\n // Manual Iterator to allow racing against Reconnect Signal\n const iterator = this.input[Symbol.asyncIterator]();\n\n while (true) {\n const nextPromise = iterator.next();\n // If reconnect signal fires, abort the wait\n const abortPromise = this.#reconnectEvent.wait().then(() => ({ abort: true }) as const);\n\n const result = await Promise.race([nextPromise, abortPromise]);\n\n // Check if we need to abort (Reconnect) or if stream is done\n if ('abort' in result || result.done) {\n if (!('abort' in result) && result.done) {\n // Normal stream end\n hasEnded = true;\n } else {\n // Reconnect triggered - break loop immediately\n break;\n }\n }\n\n // If we broke above, we don't process data. If not, 'result' is IteratorResult\n if (hasEnded && !('value' in result)) {\n // Process flush below\n } else if ('value' in result) {\n const data = result.value;\n const frames: AudioFrame[] = [];\n\n if (data === SpeechStreamv2.FLUSH_SENTINEL) {\n frames.push(...audioBstream.flush());\n hasEnded = true;\n } else {\n frames.push(...audioBstream.write((data as AudioFrame).data.buffer as ArrayBuffer));\n }\n\n for (const frame of frames) {\n this.#audioDurationCollector.push(calculateAudioDurationSeconds(frame));\n\n if (this.#ws!.readyState === WebSocket.OPEN) {\n this.#ws!.send(frame.data);\n }\n\n if (hasEnded) {\n this.#audioDurationCollector.flush();\n hasEnded = false;\n }\n }\n }\n\n if (hasEnded) break;\n }\n\n // Only send CloseStream if we are exiting normally (not reconnecting)\n if (!this.#reconnectEvent.isSet && this.#ws!.readyState === WebSocket.OPEN) {\n this.#logger.debug('Sending CloseStream message to Deepgram');\n this.#ws!.send(_CLOSE_MSG);\n }\n }\n\n async #recvTask() {\n if (!this.#ws) return;\n\n return new Promise<void>((resolve) => {\n if (!this.#ws) return resolve();\n\n this.#ws.on('message', (data: Buffer, isBinary: boolean) => {\n if (isBinary) {\n this.#logger.warn('Received unexpected binary message from Deepgram');\n return;\n }\n try {\n const msg = JSON.parse(data.toString());\n this.#processStreamEvent(msg);\n } catch (error) {\n this.#logger.error('Failed to parse Deepgram message', { error });\n }\n });\n\n this.#ws.on('close', (code, reason) => {\n this.#logger.debug(`Deepgram WebSocket closed: ${code} ${reason}`);\n resolve();\n });\n\n // Errors are caught by run() listener, resolve here to clean up task\n this.#ws.on('error', () => resolve());\n });\n }\n\n #processStreamEvent(data: Record<string, unknown>) {\n if (data.request_id) {\n this.#requestId = data.request_id as string;\n }\n\n if (data.type === 'TurnInfo') {\n const eventType = data.event;\n\n if (eventType === 'StartOfTurn') {\n if (this.#speaking) return;\n\n this.#speaking = true;\n this.queue.put({\n type: stt.SpeechEventType.START_OF_SPEECH,\n requestId: this.#requestId,\n });\n\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'Update') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EagerEndOfTurn') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, data);\n } else if (eventType === 'TurnResumed') {\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EndOfTurn') {\n if (!this.#speaking) return;\n\n this.#speaking = false;\n this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, data);\n\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n }\n } else if (data.type === 'Error') {\n this.#logger.warn('deepgram sent an error', { data });\n const desc = (data.description as string) || 'unknown error from deepgram';\n throw new Error(`Deepgram API Error: ${desc}`);\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {\n const alts = parseTranscription(this.#opts.language || 'en', data, this.startTimeOffset);\n\n if (alts.length > 0) {\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [alts[0]!, ...alts.slice(1)],\n });\n }\n }\n\n #onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n\n #getDeepgramUrl(): string {\n const params: Record<string, string | string[]> = {\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: 'linear16',\n mip_opt_out: String(this.#opts.mipOptOut),\n };\n\n // Note: v2 API does NOT include 'language' parameter\n if (this.#opts.eagerEotThreshold)\n params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();\n if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();\n if (this.#opts.eotTimeoutMs) params.eot_timeout_ms = this.#opts.eotTimeoutMs.toString();\n\n if (this.#opts.keyterms.length > 0) params.keyterm = this.#opts.keyterms;\n if (this.#opts.tags && this.#opts.tags.length > 0) params.tag = this.#opts.tags;\n\n const baseUrl = this.#opts.endpointUrl.replace(/^http/, 'ws');\n const qs = queryString.stringify(params);\n return `${baseUrl}?${qs}`;\n }\n\n override close() {\n super.close();\n this.#ws?.close();\n }\n}\n\n// --- Helpers ---\n\nfunction parseTranscription(\n language: string,\n data: Record<string, unknown>,\n startTimeOffset: number,\n): stt.SpeechData[] {\n const transcript = data.transcript as string | undefined;\n const wordsData = (data.words as Array<Record<string, unknown>>) || [];\n\n if (!wordsData || wordsData.length === 0) {\n return [];\n }\n\n let confidence = 0;\n if (wordsData.length > 0) {\n const sum = wordsData.reduce((acc: number, w) => acc + ((w.confidence as number) || 0), 0);\n confidence = sum / wordsData.length;\n }\n\n const sd: stt.SpeechData = {\n language: language,\n startTime: ((data.audio_window_start as number) || 0) + startTimeOffset,\n endTime: ((data.audio_window_end as number) || 0) + startTimeOffset,\n confidence: confidence,\n text: transcript || '',\n // Note: Deepgram V2 (Flux) API does not provide word-level timing (start/end).\n // Words only contain 'word' and 'confidence' fields, so startTime/endTime will be 0.\n // See: https://developers.deepgram.com/docs/flux/nova-3-migration\n words: wordsData.map((word) => ({\n text: (word.word as string) ?? '',\n startTime: ((word.start as number) ?? 0) + startTimeOffset,\n endTime: ((word.end as number) ?? 0) + startTimeOffset,\n confidence: (word.confidence as number) ?? 0.0,\n startTimeOffset,\n })),\n };\n\n return [sd];\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAOO;AAEP,kBAA6B;AAC7B,gBAA0B;AAC1B,mBAAkC;AAGlC,MAAM,aAAa,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC;AAqBzD,MAAM,sBAAoD;AAAA,EACxD,OAAO;AAAA,EACP,YAAY;AAAA,EACZ,UAAU,CAAC;AAAA,EACX,aAAa;AAAA,EACb,UAAU;AAAA,EACV,WAAW;AACb;AAEA,SAAS,aAAa,MAA0B;AAC9C,aAAW,OAAO,MAAM;AACtB,QAAI,IAAI,SAAS,KAAK;AACpB,YAAM,IAAI,MAAM,yCAAyC;AAAA,IAC3D;AAAA,EACF;AACA,SAAO;AACT;AAiCO,MAAM,cAAc,kBAAI,IAAI;AAAA,EACxB,QAAQ;AAAA,EACjB;AAAA,EACA;AAAA,EACA,cAAU,mBAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBd,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,IACrB,CAAC;AAED,SAAK,QAAQ,EAAE,GAAG,qBAAqB,GAAG,KAAK;AAE/C,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,8BAA8B;AAAA,IAChD;AACA,SAAK,UAAU;AAEf,QAAI,KAAK,MAAM,MAAM;AACnB,WAAK,MAAM,OAAO,aAAa,KAAK,MAAM,IAAI;AAAA,IAChD;AAAA,EACF;AAAA;AAAA,EAGA,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,MAAgB,WACd,QACA,cAC0B;AAC1B,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,OAAO,SAAiE;AACtE,UAAM,aAAa,EAAE,GAAG,KAAK,OAAO,QAAQ,KAAK,QAAQ;AACzD,WAAO,IAAI,eAAe,MAAM,YAAY,mCAAS,WAAW;AAAA,EAClE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAA6B;AACzC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AACvD,SAAK,QAAQ,MAAM,uBAAuB;AAAA,EAC5C;AACF;AAIA,MAAM,uBAAuB,kBAAI,aAAa;AAAA,EACnC,QAAQ;AAAA,EACjB;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,MAAwB;AAAA,EAExB;AAAA,EACA,aAAa;AAAA,EACb,YAAY;AAAA;AAAA,EAGZ,kBAAkB,IAAI,oBAAM;AAAA,EAE5B,YACE,aACA,MACA,aACA;AACA,UAAM,aAAa,KAAK,YAAY,WAAW;AAC/C,SAAK,QAAQ;AAEb,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,uBAAuB,QAAQ;AAAA,MAClD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,cAAc,MAA6B;AACzC,SAAK,QAAQ,MAAM,iCAAiC,IAAI;AACxD,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AAGvD,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,MAAgB,MAAM;AA7MxB;AA+MI,WAAO,CAAC,KAAK,QAAQ;AACnB,UAAI;AACF,aAAK,gBAAgB,MAAM;AAE3B,cAAM,MAAM,KAAK,gBAAgB;AACjC,aAAK,QAAQ,MAAM,2BAA2B,GAAG,EAAE;AAEnD,aAAK,MAAM,IAAI,oBAAU,KAAK;AAAA,UAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,QACzD,CAAC;AAGD,cAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,cAAI,CAAC,KAAK,IAAK,QAAO,OAAO,IAAI,MAAM,2BAA2B,CAAC;AAEnE,gBAAM,SAAS,MAAM;AA9N/B,gBAAAA;AA+NY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,SAAS;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,UAAU,CAAC,QAAe;AAlO1C,gBAAAA;AAmOY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,QAAQ;AACtB,mBAAO,GAAG;AAAA,UACZ;AAEA,eAAK,IAAI,KAAK,QAAQ,MAAM;AAC5B,eAAK,IAAI,KAAK,SAAS,OAAO;AAAA,QAChC,CAAC;AAGD,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,gBAAgB,KAAK,gBAAgB,KAAK;AAGhD,cAAM,SAAS,MAAM,QAAQ,KAAK;AAAA,UAChC,QAAQ,IAAI,CAAC,aAAa,WAAW,CAAC;AAAA,UACtC,cAAc,KAAK,MAAM,WAAW;AAAA,QACtC,CAAC;AAED,YAAI,WAAW,aAAa;AAC1B,eAAK,QAAQ,MAAM,6CAA6C;AAEhE,eAAK,IAAI,MAAM;AAAA,QACjB,OAAO;AAEL;AAAA,QACF;AAAA,MACF,SAAS,OAAO;AACd,aAAK,QAAQ,MAAM,yBAAyB,EAAE,MAAM,CAAC;AACrD,cAAM;AAAA,MACR,UAAE;AACA,cAAI,UAAK,QAAL,mBAAU,gBAAe,oBAAU,MAAM;AAC3C,eAAK,IAAI,MAAM;AAAA,QACjB;AAAA,MACF;AAAA,IACF;AACA,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAGf,UAAM,cAAc,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AACzD,UAAM,eAAe,IAAI,8BAAgB,KAAK,MAAM,YAAY,GAAG,WAAW;AAE9E,QAAI,WAAW;AAGf,UAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAElD,WAAO,MAAM;AACX,YAAM,cAAc,SAAS,KAAK;AAElC,YAAM,eAAe,KAAK,gBAAgB,KAAK,EAAE,KAAK,OAAO,EAAE,OAAO,KAAK,EAAW;AAEtF,YAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,aAAa,YAAY,CAAC;AAG7D,UAAI,WAAW,UAAU,OAAO,MAAM;AACpC,YAAI,EAAE,WAAW,WAAW,OAAO,MAAM;AAEvC,qBAAW;AAAA,QACb,OAAO;AAEL;AAAA,QACF;AAAA,MACF;AAGA,UAAI,YAAY,EAAE,WAAW,SAAS;AAAA,MAEtC,WAAW,WAAW,QAAQ;AAC5B,cAAM,OAAO,OAAO;AACpB,cAAM,SAAuB,CAAC;AAE9B,YAAI,SAAS,eAAe,gBAAgB;AAC1C,iBAAO,KAAK,GAAG,aAAa,MAAM,CAAC;AACnC,qBAAW;AAAA,QACb,OAAO;AACL,iBAAO,KAAK,GAAG,aAAa,MAAO,KAAoB,KAAK,MAAqB,CAAC;AAAA,QACpF;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,wBAAwB,SAAK,6CAA8B,KAAK,CAAC;AAEtE,cAAI,KAAK,IAAK,eAAe,oBAAU,MAAM;AAC3C,iBAAK,IAAK,KAAK,MAAM,IAAI;AAAA,UAC3B;AAEA,cAAI,UAAU;AACZ,iBAAK,wBAAwB,MAAM;AACnC,uBAAW;AAAA,UACb;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAU;AAAA,IAChB;AAGA,QAAI,CAAC,KAAK,gBAAgB,SAAS,KAAK,IAAK,eAAe,oBAAU,MAAM;AAC1E,WAAK,QAAQ,MAAM,yCAAyC;AAC5D,WAAK,IAAK,KAAK,UAAU;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAEf,WAAO,IAAI,QAAc,CAAC,YAAY;AACpC,UAAI,CAAC,KAAK,IAAK,QAAO,QAAQ;AAE9B,WAAK,IAAI,GAAG,WAAW,CAAC,MAAc,aAAsB;AAC1D,YAAI,UAAU;AACZ,eAAK,QAAQ,KAAK,kDAAkD;AACpE;AAAA,QACF;AACA,YAAI;AACF,gBAAM,MAAM,KAAK,MAAM,KAAK,SAAS,CAAC;AACtC,eAAK,oBAAoB,GAAG;AAAA,QAC9B,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,oCAAoC,EAAE,MAAM,CAAC;AAAA,QAClE;AAAA,MACF,CAAC;AAED,WAAK,IAAI,GAAG,SAAS,CAAC,MAAM,WAAW;AACrC,aAAK,QAAQ,MAAM,8BAA8B,IAAI,IAAI,MAAM,EAAE;AACjE,gBAAQ;AAAA,MACV,CAAC;AAGD,WAAK,IAAI,GAAG,SAAS,MAAM,QAAQ,CAAC;AAAA,IACtC,CAAC;AAAA,EACH;AAAA,EAEA,oBAAoB,MAA+B;AACjD,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,QAAI,KAAK,SAAS,YAAY;AAC5B,YAAM,YAAY,KAAK;AAEvB,UAAI,cAAc,eAAe;AAC/B,YAAI,KAAK,UAAW;AAEpB,aAAK,YAAY;AACjB,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,kBAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAED,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,UAAU;AACjC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,kBAAkB;AACzC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,kBAAI,gBAAgB,sBAAsB,IAAI;AAAA,MAC1E,WAAW,cAAc,eAAe;AACtC,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,aAAa;AACpC,YAAI,CAAC,KAAK,UAAW;AAErB,aAAK,YAAY;AACjB,aAAK,qBAAqB,kBAAI,gBAAgB,kBAAkB,IAAI;AAEpE,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,kBAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAAA,MACH;AAAA,IACF,WAAW,KAAK,SAAS,SAAS;AAChC,WAAK,QAAQ,KAAK,0BAA0B,EAAE,KAAK,CAAC;AACpD,YAAM,OAAQ,KAAK,eAA0B;AAC7C,YAAM,IAAI,MAAM,uBAAuB,IAAI,EAAE;AAAA,IAC/C;AAAA,EACF;AAAA,EAEA,qBAAqB,WAAgC,MAA+B;AAClF,UAAM,OAAO,mBAAmB,KAAK,MAAM,YAAY,MAAM,MAAM,KAAK,eAAe;AAEvF,QAAI,KAAK,SAAS,GAAG;AACnB,WAAK,MAAM,IAAI;AAAA,QACb,MAAM;AAAA,QACN,WAAW,KAAK;AAAA,QAChB,cAAc,CAAC,KAAK,CAAC,GAAI,GAAG,KAAK,MAAM,CAAC,CAAC;AAAA,MAC3C,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,uBAAuB,UAAkB;AACvC,UAAM,aAA8B;AAAA,MAClC,MAAM,kBAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AAAA,EAEA,kBAA0B;AACxB,UAAM,SAA4C;AAAA,MAChD,OAAO,KAAK,MAAM;AAAA,MAClB,aAAa,KAAK,MAAM,WAAW,SAAS;AAAA,MAC5C,UAAU;AAAA,MACV,aAAa,OAAO,KAAK,MAAM,SAAS;AAAA,IAC1C;AAGA,QAAI,KAAK,MAAM;AACb,aAAO,sBAAsB,KAAK,MAAM,kBAAkB,SAAS;AACrE,QAAI,KAAK,MAAM,aAAc,QAAO,gBAAgB,KAAK,MAAM,aAAa,SAAS;AACrF,QAAI,KAAK,MAAM,aAAc,QAAO,iBAAiB,KAAK,MAAM,aAAa,SAAS;AAEtF,QAAI,KAAK,MAAM,SAAS,SAAS,EAAG,QAAO,UAAU,KAAK,MAAM;AAChE,QAAI,KAAK,MAAM,QAAQ,KAAK,MAAM,KAAK,SAAS,EAAG,QAAO,MAAM,KAAK,MAAM;AAE3E,UAAM,UAAU,KAAK,MAAM,YAAY,QAAQ,SAAS,IAAI;AAC5D,UAAM,KAAK,YAAY,UAAU,MAAM;AACvC,WAAO,GAAG,OAAO,IAAI,EAAE;AAAA,EACzB;AAAA,EAES,QAAQ;AApcnB;AAqcI,UAAM,MAAM;AACZ,eAAK,QAAL,mBAAU;AAAA,EACZ;AACF;AAIA,SAAS,mBACP,UACA,MACA,iBACkB;AAClB,QAAM,aAAa,KAAK;AACxB,QAAM,YAAa,KAAK,SAA4C,CAAC;AAErE,MAAI,CAAC,aAAa,UAAU,WAAW,GAAG;AACxC,WAAO,CAAC;AAAA,EACV;AAEA,MAAI,aAAa;AACjB,MAAI,UAAU,SAAS,GAAG;AACxB,UAAM,MAAM,UAAU,OAAO,CAAC,KAAa,MAAM,OAAQ,EAAE,cAAyB,IAAI,CAAC;AACzF,iBAAa,MAAM,UAAU;AAAA,EAC/B;AAEA,QAAM,KAAqB;AAAA,IACzB;AAAA,IACA,YAAa,KAAK,sBAAiC,KAAK;AAAA,IACxD,UAAW,KAAK,oBAA+B,KAAK;AAAA,IACpD;AAAA,IACA,MAAM,cAAc;AAAA;AAAA;AAAA;AAAA,IAIpB,OAAO,UAAU,IAAI,CAAC,UAAU;AAAA,MAC9B,MAAO,KAAK,QAAmB;AAAA,MAC/B,YAAa,KAAK,SAAoB,KAAK;AAAA,MAC3C,UAAW,KAAK,OAAkB,KAAK;AAAA,MACvC,YAAa,KAAK,cAAyB;AAAA,MAC3C;AAAA,IACF,EAAE;AAAA,EACJ;AAEA,SAAO,CAAC,EAAE;AACZ;","names":["_a"]}
|
package/dist/stt_v2.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"stt_v2.d.ts","sourceRoot":"","sources":["../src/stt_v2.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,EAKtB,GAAG,EACJ,MAAM,iBAAiB,CAAC;AACzB,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAIpD,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AAM5C;;GAEG;AACH,MAAM,WAAW,YAAY;IAC3B,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,KAAK,EAAE,QAAQ,GAAG,MAAM,CAAC;IACzB,UAAU,EAAE,MAAM,CAAC;IACnB,QAAQ,EAAE,MAAM,EAAE,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,iBAAiB,CAAC,EAAE,MAAM,CAAC;IAC3B,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB,IAAI,CAAC,EAAE,MAAM,EAAE,CAAC;CACjB;AAoBD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AACH,qBAAa,KAAM,SAAQ,GAAG,CAAC,GAAG;;IAChC,QAAQ,CAAC,KAAK,oBAAoB;IAKlC;;;;;;;;;;;;;OAaG;gBACS,IAAI,GAAE,OAAO,CAAC,YAAY,CAAM;
|
|
1
|
+
{"version":3,"file":"stt_v2.d.ts","sourceRoot":"","sources":["../src/stt_v2.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,EAKtB,GAAG,EACJ,MAAM,iBAAiB,CAAC;AACzB,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAIpD,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AAM5C;;GAEG;AACH,MAAM,WAAW,YAAY;IAC3B,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,KAAK,EAAE,QAAQ,GAAG,MAAM,CAAC;IACzB,UAAU,EAAE,MAAM,CAAC;IACnB,QAAQ,EAAE,MAAM,EAAE,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,iBAAiB,CAAC,EAAE,MAAM,CAAC;IAC3B,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB,IAAI,CAAC,EAAE,MAAM,EAAE,CAAC;CACjB;AAoBD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AACH,qBAAa,KAAM,SAAQ,GAAG,CAAC,GAAG;;IAChC,QAAQ,CAAC,KAAK,oBAAoB;IAKlC;;;;;;;;;;;;;OAaG;gBACS,IAAI,GAAE,OAAO,CAAC,YAAY,CAAM;IAoB5C,6CAA6C;IAC7C,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,4BAA4B;IAC5B,IAAI,QAAQ,IAAI,MAAM,CAErB;cAEe,UAAU,CACxB,MAAM,EAAE,UAAU,GAAG,UAAU,EAAE,EACjC,YAAY,CAAC,EAAE,WAAW,GACzB,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IAI3B;;;;;OAKG;IACH,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,GAAG,CAAC,YAAY;IAKvE;;;;OAIG;IACH,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,YAAY,CAAC;CAK1C"}
|
package/dist/stt_v2.js
CHANGED
|
@@ -47,7 +47,8 @@ class STTv2 extends stt.STT {
|
|
|
47
47
|
constructor(opts = {}) {
|
|
48
48
|
super({
|
|
49
49
|
streaming: true,
|
|
50
|
-
interimResults: true
|
|
50
|
+
interimResults: true,
|
|
51
|
+
alignedTranscript: "word"
|
|
51
52
|
});
|
|
52
53
|
this.#opts = { ...defaultSTTv2Options, ...opts };
|
|
53
54
|
const apiKey = opts.apiKey || process.env.DEEPGRAM_API_KEY;
|
|
@@ -270,7 +271,7 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
270
271
|
}
|
|
271
272
|
}
|
|
272
273
|
#sendTranscriptEvent(eventType, data) {
|
|
273
|
-
const alts = parseTranscription(this.#opts.language || "en", data,
|
|
274
|
+
const alts = parseTranscription(this.#opts.language || "en", data, this.startTimeOffset);
|
|
274
275
|
if (alts.length > 0) {
|
|
275
276
|
this.queue.put({
|
|
276
277
|
type: eventType,
|
|
@@ -296,7 +297,6 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
296
297
|
encoding: "linear16",
|
|
297
298
|
mip_opt_out: String(this.#opts.mipOptOut)
|
|
298
299
|
};
|
|
299
|
-
if (this.#opts.language) params.language = this.#opts.language;
|
|
300
300
|
if (this.#opts.eagerEotThreshold)
|
|
301
301
|
params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();
|
|
302
302
|
if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();
|
|
@@ -315,21 +315,31 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
315
315
|
}
|
|
316
316
|
function parseTranscription(language, data, startTimeOffset) {
|
|
317
317
|
const transcript = data.transcript;
|
|
318
|
-
const
|
|
319
|
-
if (!
|
|
318
|
+
const wordsData = data.words || [];
|
|
319
|
+
if (!wordsData || wordsData.length === 0) {
|
|
320
320
|
return [];
|
|
321
321
|
}
|
|
322
322
|
let confidence = 0;
|
|
323
|
-
if (
|
|
324
|
-
const sum =
|
|
325
|
-
confidence = sum /
|
|
323
|
+
if (wordsData.length > 0) {
|
|
324
|
+
const sum = wordsData.reduce((acc, w) => acc + (w.confidence || 0), 0);
|
|
325
|
+
confidence = sum / wordsData.length;
|
|
326
326
|
}
|
|
327
327
|
const sd = {
|
|
328
328
|
language,
|
|
329
329
|
startTime: (data.audio_window_start || 0) + startTimeOffset,
|
|
330
330
|
endTime: (data.audio_window_end || 0) + startTimeOffset,
|
|
331
331
|
confidence,
|
|
332
|
-
text: transcript || ""
|
|
332
|
+
text: transcript || "",
|
|
333
|
+
// Note: Deepgram V2 (Flux) API does not provide word-level timing (start/end).
|
|
334
|
+
// Words only contain 'word' and 'confidence' fields, so startTime/endTime will be 0.
|
|
335
|
+
// See: https://developers.deepgram.com/docs/flux/nova-3-migration
|
|
336
|
+
words: wordsData.map((word) => ({
|
|
337
|
+
text: word.word ?? "",
|
|
338
|
+
startTime: (word.start ?? 0) + startTimeOffset,
|
|
339
|
+
endTime: (word.end ?? 0) + startTimeOffset,
|
|
340
|
+
confidence: word.confidence ?? 0,
|
|
341
|
+
startTimeOffset
|
|
342
|
+
}))
|
|
333
343
|
};
|
|
334
344
|
return [sd];
|
|
335
345
|
}
|
package/dist/stt_v2.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt_v2.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n AudioByteStream,\n Event,\n calculateAudioDurationSeconds,\n log,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport * as queryString from 'node:querystring';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { V2Models } from './models.js';\n\nconst _CLOSE_MSG = JSON.stringify({ type: 'CloseStream' });\n\n// --- Configuration ---\n\n/**\n * Configuration options for STTv2 (Deepgram Flux model).\n */\nexport interface STTv2Options {\n apiKey?: string;\n model: V2Models | string;\n sampleRate: number;\n keyterms: string[];\n endpointUrl: string;\n language?: string;\n eagerEotThreshold?: number;\n eotThreshold?: number;\n eotTimeoutMs?: number;\n mipOptOut?: boolean;\n tags?: string[];\n}\n\nconst defaultSTTv2Options: Omit<STTv2Options, 'apiKey'> = {\n model: 'flux-general-en',\n sampleRate: 16000,\n keyterms: [],\n endpointUrl: 'wss://api.deepgram.com/v2/listen',\n language: 'en',\n mipOptOut: false,\n};\n\nfunction validateTags(tags: string[]): string[] {\n for (const tag of tags) {\n if (tag.length > 128) {\n throw new Error('tag must be no more than 128 characters');\n }\n }\n return tags;\n}\n\n/**\n * Deepgram STTv2 using the Flux model for streaming speech-to-text.\n *\n * This uses Deepgram's V2 API (`/v2/listen`) which provides turn-based\n * transcription with support for preemptive generation.\n *\n * @remarks\n * Key differences from STT (V1):\n * - Uses `TurnInfo` events instead of `SpeechStarted`/`Results`\n * - Supports `eagerEotThreshold` for preemptive LLM generation\n * - Sends `PREFLIGHT_TRANSCRIPT` events when eager end-of-turn is detected\n *\n * @example\n * ```typescript\n * import { STTv2 } from '@livekit/agents-plugin-deepgram';\n *\n * const stt = new STTv2({\n * model: 'flux-general-en',\n * eagerEotThreshold: 0.5, // Enable preemptive generation\n * });\n *\n * const stream = stt.stream();\n * stream.pushFrame(audioFrame);\n *\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives?.[0]?.text);\n * }\n * }\n * ```\n */\nexport class STTv2 extends stt.STT {\n readonly label = 'deepgram.STTv2';\n #opts: STTv2Options;\n #apiKey: string;\n #logger = log();\n\n /**\n * Create a new Deepgram STTv2 instance.\n *\n * @param opts - Configuration options\n * @param opts.apiKey - Deepgram API key (defaults to `DEEPGRAM_API_KEY` env var)\n * @param opts.model - Model to use (default: `flux-general-en`)\n * @param opts.eagerEotThreshold - Threshold (0.3-0.9) for preemptive generation\n * @param opts.eotThreshold - End-of-turn detection threshold (default: 0.7)\n * @param opts.eotTimeoutMs - End-of-turn timeout in ms (default: 3000)\n * @param opts.keyterms - List of key terms to improve recognition\n * @param opts.tags - Tags for usage reporting (max 128 chars each)\n *\n * @throws Error if no API key is provided\n */\n constructor(opts: Partial<STTv2Options> = {}) {\n super({\n streaming: true,\n interimResults: true,\n });\n\n this.#opts = { ...defaultSTTv2Options, ...opts };\n\n const apiKey = opts.apiKey || process.env.DEEPGRAM_API_KEY;\n if (!apiKey) {\n throw new Error('Deepgram API key is required');\n }\n this.#apiKey = apiKey;\n\n if (this.#opts.tags) {\n this.#opts.tags = validateTags(this.#opts.tags);\n }\n }\n\n /** The model being used for transcription */\n get model(): string {\n return this.#opts.model;\n }\n\n /** The STT provider name */\n get provider(): string {\n return 'Deepgram';\n }\n\n protected async _recognize(\n _frame: AudioFrame | AudioFrame[],\n _abortSignal?: AbortSignal,\n ): Promise<stt.SpeechEvent> {\n throw new Error('V2 API does not support non-streaming recognize. Use .stream()');\n }\n\n /**\n * Create a new streaming transcription session.\n *\n * @param options - Stream options\n * @returns A SpeechStream that emits transcription events\n */\n stream(options?: { connOptions?: APIConnectOptions }): stt.SpeechStream {\n const streamOpts = { ...this.#opts, apiKey: this.#apiKey };\n return new SpeechStreamv2(this, streamOpts, options?.connOptions);\n }\n\n /**\n * Update STT options. Changes will take effect on the next stream.\n *\n * @param opts - Partial options to update\n */\n updateOptions(opts: Partial<STTv2Options>) {\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n this.#logger.debug('Updated STTv2 options');\n }\n}\n\n// --- Stream Implementation ---\n\nclass SpeechStreamv2 extends stt.SpeechStream {\n readonly label = 'deepgram.SpeechStreamv2';\n #opts: STTv2Options & { apiKey: string };\n #logger = log();\n #ws: WebSocket | null = null;\n\n #audioDurationCollector: PeriodicCollector<number>;\n #requestId = '';\n #speaking = false;\n\n // Parity: _reconnect_event - using existing Event class from @livekit/agents\n #reconnectEvent = new Event();\n\n constructor(\n sttInstance: STTv2,\n opts: STTv2Options & { apiKey: string },\n connOptions?: APIConnectOptions,\n ) {\n super(sttInstance, opts.sampleRate, connOptions);\n this.#opts = opts;\n\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.#onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n updateOptions(opts: Partial<STTv2Options>) {\n this.#logger.debug('Stream received option update', opts);\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n\n // Trigger reconnection loop\n this.#reconnectEvent.set();\n }\n\n protected async run() {\n // Outer Loop: Handles reconnections (Configuration updates)\n while (!this.closed) {\n try {\n this.#reconnectEvent.clear();\n\n const url = this.#getDeepgramUrl();\n this.#logger.debug(`Connecting to Deepgram: ${url}`);\n\n this.#ws = new WebSocket(url, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n // 1. Wait for Connection Open\n await new Promise<void>((resolve, reject) => {\n if (!this.#ws) return reject(new Error('WebSocket not initialized'));\n\n const onOpen = () => {\n this.#ws?.off('error', onError);\n resolve();\n };\n const onError = (err: Error) => {\n this.#ws?.off('open', onOpen);\n reject(err);\n };\n\n this.#ws.once('open', onOpen);\n this.#ws.once('error', onError);\n });\n\n // 2. Run Concurrent Tasks (Send & Receive)\n const sendPromise = this.#sendTask();\n const recvPromise = this.#recvTask();\n const reconnectWait = this.#reconnectEvent.wait();\n\n // 3. Race: Normal Completion vs Reconnect Signal\n const result = await Promise.race([\n Promise.all([sendPromise, recvPromise]),\n reconnectWait.then(() => 'RECONNECT'),\n ]);\n\n if (result === 'RECONNECT') {\n this.#logger.debug('Reconnecting stream due to option update...');\n // Close current socket; loop will restart and open a new one\n this.#ws.close();\n } else {\n // Normal finish (Stream ended or Error thrown)\n break;\n }\n } catch (error) {\n this.#logger.error('Deepgram stream error', { error });\n throw error; // Let Base Class handle retry logic\n } finally {\n if (this.#ws?.readyState === WebSocket.OPEN) {\n this.#ws.close();\n }\n }\n }\n this.close();\n }\n\n async #sendTask() {\n if (!this.#ws) return;\n\n // Buffer audio into 50ms chunks (Parity)\n const samples50ms = Math.floor(this.#opts.sampleRate / 20);\n const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samples50ms);\n\n let hasEnded = false;\n\n // Manual Iterator to allow racing against Reconnect Signal\n const iterator = this.input[Symbol.asyncIterator]();\n\n while (true) {\n const nextPromise = iterator.next();\n // If reconnect signal fires, abort the wait\n const abortPromise = this.#reconnectEvent.wait().then(() => ({ abort: true }) as const);\n\n const result = await Promise.race([nextPromise, abortPromise]);\n\n // Check if we need to abort (Reconnect) or if stream is done\n if ('abort' in result || result.done) {\n if (!('abort' in result) && result.done) {\n // Normal stream end\n hasEnded = true;\n } else {\n // Reconnect triggered - break loop immediately\n break;\n }\n }\n\n // If we broke above, we don't process data. If not, 'result' is IteratorResult\n if (hasEnded && !('value' in result)) {\n // Process flush below\n } else if ('value' in result) {\n const data = result.value;\n const frames: AudioFrame[] = [];\n\n if (data === SpeechStreamv2.FLUSH_SENTINEL) {\n frames.push(...audioBstream.flush());\n hasEnded = true;\n } else {\n frames.push(...audioBstream.write((data as AudioFrame).data.buffer as ArrayBuffer));\n }\n\n for (const frame of frames) {\n this.#audioDurationCollector.push(calculateAudioDurationSeconds(frame));\n\n if (this.#ws!.readyState === WebSocket.OPEN) {\n this.#ws!.send(frame.data);\n }\n\n if (hasEnded) {\n this.#audioDurationCollector.flush();\n hasEnded = false;\n }\n }\n }\n\n if (hasEnded) break;\n }\n\n // Only send CloseStream if we are exiting normally (not reconnecting)\n if (!this.#reconnectEvent.isSet && this.#ws!.readyState === WebSocket.OPEN) {\n this.#logger.debug('Sending CloseStream message to Deepgram');\n this.#ws!.send(_CLOSE_MSG);\n }\n }\n\n async #recvTask() {\n if (!this.#ws) return;\n\n return new Promise<void>((resolve) => {\n if (!this.#ws) return resolve();\n\n this.#ws.on('message', (data: Buffer, isBinary: boolean) => {\n if (isBinary) {\n this.#logger.warn('Received unexpected binary message from Deepgram');\n return;\n }\n try {\n const msg = JSON.parse(data.toString());\n this.#processStreamEvent(msg);\n } catch (error) {\n this.#logger.error('Failed to parse Deepgram message', { error });\n }\n });\n\n this.#ws.on('close', (code, reason) => {\n this.#logger.debug(`Deepgram WebSocket closed: ${code} ${reason}`);\n resolve();\n });\n\n // Errors are caught by run() listener, resolve here to clean up task\n this.#ws.on('error', () => resolve());\n });\n }\n\n #processStreamEvent(data: Record<string, unknown>) {\n if (data.request_id) {\n this.#requestId = data.request_id as string;\n }\n\n if (data.type === 'TurnInfo') {\n const eventType = data.event;\n\n if (eventType === 'StartOfTurn') {\n if (this.#speaking) return;\n\n this.#speaking = true;\n this.queue.put({\n type: stt.SpeechEventType.START_OF_SPEECH,\n requestId: this.#requestId,\n });\n\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'Update') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EagerEndOfTurn') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, data);\n } else if (eventType === 'TurnResumed') {\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EndOfTurn') {\n if (!this.#speaking) return;\n\n this.#speaking = false;\n this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, data);\n\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n }\n } else if (data.type === 'Error') {\n this.#logger.warn('deepgram sent an error', { data });\n const desc = (data.description as string) || 'unknown error from deepgram';\n throw new Error(`Deepgram API Error: ${desc}`);\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {\n // Note: start_time_offset is not yet available in the TypeScript base class\n // Using 0.0 for now - full parity would require base class changes\n const alts = parseTranscription(this.#opts.language || 'en', data, 0.0);\n\n if (alts.length > 0) {\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [alts[0]!, ...alts.slice(1)],\n });\n }\n }\n\n #onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n\n #getDeepgramUrl(): string {\n const params: Record<string, string | string[]> = {\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: 'linear16',\n mip_opt_out: String(this.#opts.mipOptOut),\n };\n\n if (this.#opts.language) params.language = this.#opts.language;\n if (this.#opts.eagerEotThreshold)\n params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();\n if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();\n if (this.#opts.eotTimeoutMs) params.eot_timeout_ms = this.#opts.eotTimeoutMs.toString();\n\n if (this.#opts.keyterms.length > 0) params.keyterm = this.#opts.keyterms;\n if (this.#opts.tags && this.#opts.tags.length > 0) params.tag = this.#opts.tags;\n\n const baseUrl = this.#opts.endpointUrl.replace(/^http/, 'ws');\n const qs = queryString.stringify(params);\n return `${baseUrl}?${qs}`;\n }\n\n override close() {\n super.close();\n this.#ws?.close();\n }\n}\n\n// --- Helpers ---\n\nfunction parseTranscription(\n language: string,\n data: Record<string, unknown>,\n startTimeOffset: number,\n): stt.SpeechData[] {\n const transcript = data.transcript as string | undefined;\n const words = (data.words as Array<Record<string, unknown>>) || [];\n\n if (!words || words.length === 0) {\n return [];\n }\n\n let confidence = 0;\n if (words.length > 0) {\n const sum = words.reduce((acc: number, w) => acc + ((w.confidence as number) || 0), 0);\n confidence = sum / words.length;\n }\n\n const sd: stt.SpeechData = {\n language: language,\n startTime: ((data.audio_window_start as number) || 0) + startTimeOffset,\n endTime: ((data.audio_window_end as number) || 0) + startTimeOffset,\n confidence: confidence,\n text: transcript || '',\n };\n\n return [sd];\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,YAAY,iBAAiB;AAC7B,SAAS,iBAAiB;AAC1B,SAAS,yBAAyB;AAGlC,MAAM,aAAa,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC;AAqBzD,MAAM,sBAAoD;AAAA,EACxD,OAAO;AAAA,EACP,YAAY;AAAA,EACZ,UAAU,CAAC;AAAA,EACX,aAAa;AAAA,EACb,UAAU;AAAA,EACV,WAAW;AACb;AAEA,SAAS,aAAa,MAA0B;AAC9C,aAAW,OAAO,MAAM;AACtB,QAAI,IAAI,SAAS,KAAK;AACpB,YAAM,IAAI,MAAM,yCAAyC;AAAA,IAC3D;AAAA,EACF;AACA,SAAO;AACT;AAiCO,MAAM,cAAc,IAAI,IAAI;AAAA,EACxB,QAAQ;AAAA,EACjB;AAAA,EACA;AAAA,EACA,UAAU,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBd,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,IAClB,CAAC;AAED,SAAK,QAAQ,EAAE,GAAG,qBAAqB,GAAG,KAAK;AAE/C,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,8BAA8B;AAAA,IAChD;AACA,SAAK,UAAU;AAEf,QAAI,KAAK,MAAM,MAAM;AACnB,WAAK,MAAM,OAAO,aAAa,KAAK,MAAM,IAAI;AAAA,IAChD;AAAA,EACF;AAAA;AAAA,EAGA,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,MAAgB,WACd,QACA,cAC0B;AAC1B,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,OAAO,SAAiE;AACtE,UAAM,aAAa,EAAE,GAAG,KAAK,OAAO,QAAQ,KAAK,QAAQ;AACzD,WAAO,IAAI,eAAe,MAAM,YAAY,mCAAS,WAAW;AAAA,EAClE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAA6B;AACzC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AACvD,SAAK,QAAQ,MAAM,uBAAuB;AAAA,EAC5C;AACF;AAIA,MAAM,uBAAuB,IAAI,aAAa;AAAA,EACnC,QAAQ;AAAA,EACjB;AAAA,EACA,UAAU,IAAI;AAAA,EACd,MAAwB;AAAA,EAExB;AAAA,EACA,aAAa;AAAA,EACb,YAAY;AAAA;AAAA,EAGZ,kBAAkB,IAAI,MAAM;AAAA,EAE5B,YACE,aACA,MACA,aACA;AACA,UAAM,aAAa,KAAK,YAAY,WAAW;AAC/C,SAAK,QAAQ;AAEb,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,uBAAuB,QAAQ;AAAA,MAClD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,cAAc,MAA6B;AACzC,SAAK,QAAQ,MAAM,iCAAiC,IAAI;AACxD,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AAGvD,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,MAAgB,MAAM;AA5MxB;AA8MI,WAAO,CAAC,KAAK,QAAQ;AACnB,UAAI;AACF,aAAK,gBAAgB,MAAM;AAE3B,cAAM,MAAM,KAAK,gBAAgB;AACjC,aAAK,QAAQ,MAAM,2BAA2B,GAAG,EAAE;AAEnD,aAAK,MAAM,IAAI,UAAU,KAAK;AAAA,UAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,QACzD,CAAC;AAGD,cAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,cAAI,CAAC,KAAK,IAAK,QAAO,OAAO,IAAI,MAAM,2BAA2B,CAAC;AAEnE,gBAAM,SAAS,MAAM;AA7N/B,gBAAAA;AA8NY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,SAAS;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,UAAU,CAAC,QAAe;AAjO1C,gBAAAA;AAkOY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,QAAQ;AACtB,mBAAO,GAAG;AAAA,UACZ;AAEA,eAAK,IAAI,KAAK,QAAQ,MAAM;AAC5B,eAAK,IAAI,KAAK,SAAS,OAAO;AAAA,QAChC,CAAC;AAGD,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,gBAAgB,KAAK,gBAAgB,KAAK;AAGhD,cAAM,SAAS,MAAM,QAAQ,KAAK;AAAA,UAChC,QAAQ,IAAI,CAAC,aAAa,WAAW,CAAC;AAAA,UACtC,cAAc,KAAK,MAAM,WAAW;AAAA,QACtC,CAAC;AAED,YAAI,WAAW,aAAa;AAC1B,eAAK,QAAQ,MAAM,6CAA6C;AAEhE,eAAK,IAAI,MAAM;AAAA,QACjB,OAAO;AAEL;AAAA,QACF;AAAA,MACF,SAAS,OAAO;AACd,aAAK,QAAQ,MAAM,yBAAyB,EAAE,MAAM,CAAC;AACrD,cAAM;AAAA,MACR,UAAE;AACA,cAAI,UAAK,QAAL,mBAAU,gBAAe,UAAU,MAAM;AAC3C,eAAK,IAAI,MAAM;AAAA,QACjB;AAAA,MACF;AAAA,IACF;AACA,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAGf,UAAM,cAAc,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AACzD,UAAM,eAAe,IAAI,gBAAgB,KAAK,MAAM,YAAY,GAAG,WAAW;AAE9E,QAAI,WAAW;AAGf,UAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAElD,WAAO,MAAM;AACX,YAAM,cAAc,SAAS,KAAK;AAElC,YAAM,eAAe,KAAK,gBAAgB,KAAK,EAAE,KAAK,OAAO,EAAE,OAAO,KAAK,EAAW;AAEtF,YAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,aAAa,YAAY,CAAC;AAG7D,UAAI,WAAW,UAAU,OAAO,MAAM;AACpC,YAAI,EAAE,WAAW,WAAW,OAAO,MAAM;AAEvC,qBAAW;AAAA,QACb,OAAO;AAEL;AAAA,QACF;AAAA,MACF;AAGA,UAAI,YAAY,EAAE,WAAW,SAAS;AAAA,MAEtC,WAAW,WAAW,QAAQ;AAC5B,cAAM,OAAO,OAAO;AACpB,cAAM,SAAuB,CAAC;AAE9B,YAAI,SAAS,eAAe,gBAAgB;AAC1C,iBAAO,KAAK,GAAG,aAAa,MAAM,CAAC;AACnC,qBAAW;AAAA,QACb,OAAO;AACL,iBAAO,KAAK,GAAG,aAAa,MAAO,KAAoB,KAAK,MAAqB,CAAC;AAAA,QACpF;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,wBAAwB,KAAK,8BAA8B,KAAK,CAAC;AAEtE,cAAI,KAAK,IAAK,eAAe,UAAU,MAAM;AAC3C,iBAAK,IAAK,KAAK,MAAM,IAAI;AAAA,UAC3B;AAEA,cAAI,UAAU;AACZ,iBAAK,wBAAwB,MAAM;AACnC,uBAAW;AAAA,UACb;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAU;AAAA,IAChB;AAGA,QAAI,CAAC,KAAK,gBAAgB,SAAS,KAAK,IAAK,eAAe,UAAU,MAAM;AAC1E,WAAK,QAAQ,MAAM,yCAAyC;AAC5D,WAAK,IAAK,KAAK,UAAU;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAEf,WAAO,IAAI,QAAc,CAAC,YAAY;AACpC,UAAI,CAAC,KAAK,IAAK,QAAO,QAAQ;AAE9B,WAAK,IAAI,GAAG,WAAW,CAAC,MAAc,aAAsB;AAC1D,YAAI,UAAU;AACZ,eAAK,QAAQ,KAAK,kDAAkD;AACpE;AAAA,QACF;AACA,YAAI;AACF,gBAAM,MAAM,KAAK,MAAM,KAAK,SAAS,CAAC;AACtC,eAAK,oBAAoB,GAAG;AAAA,QAC9B,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,oCAAoC,EAAE,MAAM,CAAC;AAAA,QAClE;AAAA,MACF,CAAC;AAED,WAAK,IAAI,GAAG,SAAS,CAAC,MAAM,WAAW;AACrC,aAAK,QAAQ,MAAM,8BAA8B,IAAI,IAAI,MAAM,EAAE;AACjE,gBAAQ;AAAA,MACV,CAAC;AAGD,WAAK,IAAI,GAAG,SAAS,MAAM,QAAQ,CAAC;AAAA,IACtC,CAAC;AAAA,EACH;AAAA,EAEA,oBAAoB,MAA+B;AACjD,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,QAAI,KAAK,SAAS,YAAY;AAC5B,YAAM,YAAY,KAAK;AAEvB,UAAI,cAAc,eAAe;AAC/B,YAAI,KAAK,UAAW;AAEpB,aAAK,YAAY;AACjB,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,IAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAED,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,UAAU;AACjC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,kBAAkB;AACzC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,IAAI,gBAAgB,sBAAsB,IAAI;AAAA,MAC1E,WAAW,cAAc,eAAe;AACtC,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,aAAa;AACpC,YAAI,CAAC,KAAK,UAAW;AAErB,aAAK,YAAY;AACjB,aAAK,qBAAqB,IAAI,gBAAgB,kBAAkB,IAAI;AAEpE,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,IAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAAA,MACH;AAAA,IACF,WAAW,KAAK,SAAS,SAAS;AAChC,WAAK,QAAQ,KAAK,0BAA0B,EAAE,KAAK,CAAC;AACpD,YAAM,OAAQ,KAAK,eAA0B;AAC7C,YAAM,IAAI,MAAM,uBAAuB,IAAI,EAAE;AAAA,IAC/C;AAAA,EACF;AAAA,EAEA,qBAAqB,WAAgC,MAA+B;AAGlF,UAAM,OAAO,mBAAmB,KAAK,MAAM,YAAY,MAAM,MAAM,CAAG;AAEtE,QAAI,KAAK,SAAS,GAAG;AACnB,WAAK,MAAM,IAAI;AAAA,QACb,MAAM;AAAA,QACN,WAAW,KAAK;AAAA,QAChB,cAAc,CAAC,KAAK,CAAC,GAAI,GAAG,KAAK,MAAM,CAAC,CAAC;AAAA,MAC3C,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,uBAAuB,UAAkB;AACvC,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;AAAA,EAEA,kBAA0B;AACxB,UAAM,SAA4C;AAAA,MAChD,OAAO,KAAK,MAAM;AAAA,MAClB,aAAa,KAAK,MAAM,WAAW,SAAS;AAAA,MAC5C,UAAU;AAAA,MACV,aAAa,OAAO,KAAK,MAAM,SAAS;AAAA,IAC1C;AAEA,QAAI,KAAK,MAAM,SAAU,QAAO,WAAW,KAAK,MAAM;AACtD,QAAI,KAAK,MAAM;AACb,aAAO,sBAAsB,KAAK,MAAM,kBAAkB,SAAS;AACrE,QAAI,KAAK,MAAM,aAAc,QAAO,gBAAgB,KAAK,MAAM,aAAa,SAAS;AACrF,QAAI,KAAK,MAAM,aAAc,QAAO,iBAAiB,KAAK,MAAM,aAAa,SAAS;AAEtF,QAAI,KAAK,MAAM,SAAS,SAAS,EAAG,QAAO,UAAU,KAAK,MAAM;AAChE,QAAI,KAAK,MAAM,QAAQ,KAAK,MAAM,KAAK,SAAS,EAAG,QAAO,MAAM,KAAK,MAAM;AAE3E,UAAM,UAAU,KAAK,MAAM,YAAY,QAAQ,SAAS,IAAI;AAC5D,UAAM,KAAK,YAAY,UAAU,MAAM;AACvC,WAAO,GAAG,OAAO,IAAI,EAAE;AAAA,EACzB;AAAA,EAES,QAAQ;AArcnB;AAscI,UAAM,MAAM;AACZ,eAAK,QAAL,mBAAU;AAAA,EACZ;AACF;AAIA,SAAS,mBACP,UACA,MACA,iBACkB;AAClB,QAAM,aAAa,KAAK;AACxB,QAAM,QAAS,KAAK,SAA4C,CAAC;AAEjE,MAAI,CAAC,SAAS,MAAM,WAAW,GAAG;AAChC,WAAO,CAAC;AAAA,EACV;AAEA,MAAI,aAAa;AACjB,MAAI,MAAM,SAAS,GAAG;AACpB,UAAM,MAAM,MAAM,OAAO,CAAC,KAAa,MAAM,OAAQ,EAAE,cAAyB,IAAI,CAAC;AACrF,iBAAa,MAAM,MAAM;AAAA,EAC3B;AAEA,QAAM,KAAqB;AAAA,IACzB;AAAA,IACA,YAAa,KAAK,sBAAiC,KAAK;AAAA,IACxD,UAAW,KAAK,oBAA+B,KAAK;AAAA,IACpD;AAAA,IACA,MAAM,cAAc;AAAA,EACtB;AAEA,SAAO,CAAC,EAAE;AACZ;","names":["_a"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt_v2.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n AudioByteStream,\n Event,\n calculateAudioDurationSeconds,\n log,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport * as queryString from 'node:querystring';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { V2Models } from './models.js';\n\nconst _CLOSE_MSG = JSON.stringify({ type: 'CloseStream' });\n\n// --- Configuration ---\n\n/**\n * Configuration options for STTv2 (Deepgram Flux model).\n */\nexport interface STTv2Options {\n apiKey?: string;\n model: V2Models | string;\n sampleRate: number;\n keyterms: string[];\n endpointUrl: string;\n language?: string;\n eagerEotThreshold?: number;\n eotThreshold?: number;\n eotTimeoutMs?: number;\n mipOptOut?: boolean;\n tags?: string[];\n}\n\nconst defaultSTTv2Options: Omit<STTv2Options, 'apiKey'> = {\n model: 'flux-general-en',\n sampleRate: 16000,\n keyterms: [],\n endpointUrl: 'wss://api.deepgram.com/v2/listen',\n language: 'en',\n mipOptOut: false,\n};\n\nfunction validateTags(tags: string[]): string[] {\n for (const tag of tags) {\n if (tag.length > 128) {\n throw new Error('tag must be no more than 128 characters');\n }\n }\n return tags;\n}\n\n/**\n * Deepgram STTv2 using the Flux model for streaming speech-to-text.\n *\n * This uses Deepgram's V2 API (`/v2/listen`) which provides turn-based\n * transcription with support for preemptive generation.\n *\n * @remarks\n * Key differences from STT (V1):\n * - Uses `TurnInfo` events instead of `SpeechStarted`/`Results`\n * - Supports `eagerEotThreshold` for preemptive LLM generation\n * - Sends `PREFLIGHT_TRANSCRIPT` events when eager end-of-turn is detected\n *\n * @example\n * ```typescript\n * import { STTv2 } from '@livekit/agents-plugin-deepgram';\n *\n * const stt = new STTv2({\n * model: 'flux-general-en',\n * eagerEotThreshold: 0.5, // Enable preemptive generation\n * });\n *\n * const stream = stt.stream();\n * stream.pushFrame(audioFrame);\n *\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives?.[0]?.text);\n * }\n * }\n * ```\n */\nexport class STTv2 extends stt.STT {\n readonly label = 'deepgram.STTv2';\n #opts: STTv2Options;\n #apiKey: string;\n #logger = log();\n\n /**\n * Create a new Deepgram STTv2 instance.\n *\n * @param opts - Configuration options\n * @param opts.apiKey - Deepgram API key (defaults to `DEEPGRAM_API_KEY` env var)\n * @param opts.model - Model to use (default: `flux-general-en`)\n * @param opts.eagerEotThreshold - Threshold (0.3-0.9) for preemptive generation\n * @param opts.eotThreshold - End-of-turn detection threshold (default: 0.7)\n * @param opts.eotTimeoutMs - End-of-turn timeout in ms (default: 3000)\n * @param opts.keyterms - List of key terms to improve recognition\n * @param opts.tags - Tags for usage reporting (max 128 chars each)\n *\n * @throws Error if no API key is provided\n */\n constructor(opts: Partial<STTv2Options> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: 'word',\n });\n\n this.#opts = { ...defaultSTTv2Options, ...opts };\n\n const apiKey = opts.apiKey || process.env.DEEPGRAM_API_KEY;\n if (!apiKey) {\n throw new Error('Deepgram API key is required');\n }\n this.#apiKey = apiKey;\n\n if (this.#opts.tags) {\n this.#opts.tags = validateTags(this.#opts.tags);\n }\n }\n\n /** The model being used for transcription */\n get model(): string {\n return this.#opts.model;\n }\n\n /** The STT provider name */\n get provider(): string {\n return 'Deepgram';\n }\n\n protected async _recognize(\n _frame: AudioFrame | AudioFrame[],\n _abortSignal?: AbortSignal,\n ): Promise<stt.SpeechEvent> {\n throw new Error('V2 API does not support non-streaming recognize. Use .stream()');\n }\n\n /**\n * Create a new streaming transcription session.\n *\n * @param options - Stream options\n * @returns A SpeechStream that emits transcription events\n */\n stream(options?: { connOptions?: APIConnectOptions }): stt.SpeechStream {\n const streamOpts = { ...this.#opts, apiKey: this.#apiKey };\n return new SpeechStreamv2(this, streamOpts, options?.connOptions);\n }\n\n /**\n * Update STT options. Changes will take effect on the next stream.\n *\n * @param opts - Partial options to update\n */\n updateOptions(opts: Partial<STTv2Options>) {\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n this.#logger.debug('Updated STTv2 options');\n }\n}\n\n// --- Stream Implementation ---\n\nclass SpeechStreamv2 extends stt.SpeechStream {\n readonly label = 'deepgram.SpeechStreamv2';\n #opts: STTv2Options & { apiKey: string };\n #logger = log();\n #ws: WebSocket | null = null;\n\n #audioDurationCollector: PeriodicCollector<number>;\n #requestId = '';\n #speaking = false;\n\n // Parity: _reconnect_event - using existing Event class from @livekit/agents\n #reconnectEvent = new Event();\n\n constructor(\n sttInstance: STTv2,\n opts: STTv2Options & { apiKey: string },\n connOptions?: APIConnectOptions,\n ) {\n super(sttInstance, opts.sampleRate, connOptions);\n this.#opts = opts;\n\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.#onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n updateOptions(opts: Partial<STTv2Options>) {\n this.#logger.debug('Stream received option update', opts);\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n\n // Trigger reconnection loop\n this.#reconnectEvent.set();\n }\n\n protected async run() {\n // Outer Loop: Handles reconnections (Configuration updates)\n while (!this.closed) {\n try {\n this.#reconnectEvent.clear();\n\n const url = this.#getDeepgramUrl();\n this.#logger.debug(`Connecting to Deepgram: ${url}`);\n\n this.#ws = new WebSocket(url, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n // 1. Wait for Connection Open\n await new Promise<void>((resolve, reject) => {\n if (!this.#ws) return reject(new Error('WebSocket not initialized'));\n\n const onOpen = () => {\n this.#ws?.off('error', onError);\n resolve();\n };\n const onError = (err: Error) => {\n this.#ws?.off('open', onOpen);\n reject(err);\n };\n\n this.#ws.once('open', onOpen);\n this.#ws.once('error', onError);\n });\n\n // 2. Run Concurrent Tasks (Send & Receive)\n const sendPromise = this.#sendTask();\n const recvPromise = this.#recvTask();\n const reconnectWait = this.#reconnectEvent.wait();\n\n // 3. Race: Normal Completion vs Reconnect Signal\n const result = await Promise.race([\n Promise.all([sendPromise, recvPromise]),\n reconnectWait.then(() => 'RECONNECT'),\n ]);\n\n if (result === 'RECONNECT') {\n this.#logger.debug('Reconnecting stream due to option update...');\n // Close current socket; loop will restart and open a new one\n this.#ws.close();\n } else {\n // Normal finish (Stream ended or Error thrown)\n break;\n }\n } catch (error) {\n this.#logger.error('Deepgram stream error', { error });\n throw error; // Let Base Class handle retry logic\n } finally {\n if (this.#ws?.readyState === WebSocket.OPEN) {\n this.#ws.close();\n }\n }\n }\n this.close();\n }\n\n async #sendTask() {\n if (!this.#ws) return;\n\n // Buffer audio into 50ms chunks (Parity)\n const samples50ms = Math.floor(this.#opts.sampleRate / 20);\n const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samples50ms);\n\n let hasEnded = false;\n\n // Manual Iterator to allow racing against Reconnect Signal\n const iterator = this.input[Symbol.asyncIterator]();\n\n while (true) {\n const nextPromise = iterator.next();\n // If reconnect signal fires, abort the wait\n const abortPromise = this.#reconnectEvent.wait().then(() => ({ abort: true }) as const);\n\n const result = await Promise.race([nextPromise, abortPromise]);\n\n // Check if we need to abort (Reconnect) or if stream is done\n if ('abort' in result || result.done) {\n if (!('abort' in result) && result.done) {\n // Normal stream end\n hasEnded = true;\n } else {\n // Reconnect triggered - break loop immediately\n break;\n }\n }\n\n // If we broke above, we don't process data. If not, 'result' is IteratorResult\n if (hasEnded && !('value' in result)) {\n // Process flush below\n } else if ('value' in result) {\n const data = result.value;\n const frames: AudioFrame[] = [];\n\n if (data === SpeechStreamv2.FLUSH_SENTINEL) {\n frames.push(...audioBstream.flush());\n hasEnded = true;\n } else {\n frames.push(...audioBstream.write((data as AudioFrame).data.buffer as ArrayBuffer));\n }\n\n for (const frame of frames) {\n this.#audioDurationCollector.push(calculateAudioDurationSeconds(frame));\n\n if (this.#ws!.readyState === WebSocket.OPEN) {\n this.#ws!.send(frame.data);\n }\n\n if (hasEnded) {\n this.#audioDurationCollector.flush();\n hasEnded = false;\n }\n }\n }\n\n if (hasEnded) break;\n }\n\n // Only send CloseStream if we are exiting normally (not reconnecting)\n if (!this.#reconnectEvent.isSet && this.#ws!.readyState === WebSocket.OPEN) {\n this.#logger.debug('Sending CloseStream message to Deepgram');\n this.#ws!.send(_CLOSE_MSG);\n }\n }\n\n async #recvTask() {\n if (!this.#ws) return;\n\n return new Promise<void>((resolve) => {\n if (!this.#ws) return resolve();\n\n this.#ws.on('message', (data: Buffer, isBinary: boolean) => {\n if (isBinary) {\n this.#logger.warn('Received unexpected binary message from Deepgram');\n return;\n }\n try {\n const msg = JSON.parse(data.toString());\n this.#processStreamEvent(msg);\n } catch (error) {\n this.#logger.error('Failed to parse Deepgram message', { error });\n }\n });\n\n this.#ws.on('close', (code, reason) => {\n this.#logger.debug(`Deepgram WebSocket closed: ${code} ${reason}`);\n resolve();\n });\n\n // Errors are caught by run() listener, resolve here to clean up task\n this.#ws.on('error', () => resolve());\n });\n }\n\n #processStreamEvent(data: Record<string, unknown>) {\n if (data.request_id) {\n this.#requestId = data.request_id as string;\n }\n\n if (data.type === 'TurnInfo') {\n const eventType = data.event;\n\n if (eventType === 'StartOfTurn') {\n if (this.#speaking) return;\n\n this.#speaking = true;\n this.queue.put({\n type: stt.SpeechEventType.START_OF_SPEECH,\n requestId: this.#requestId,\n });\n\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'Update') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EagerEndOfTurn') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, data);\n } else if (eventType === 'TurnResumed') {\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EndOfTurn') {\n if (!this.#speaking) return;\n\n this.#speaking = false;\n this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, data);\n\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n }\n } else if (data.type === 'Error') {\n this.#logger.warn('deepgram sent an error', { data });\n const desc = (data.description as string) || 'unknown error from deepgram';\n throw new Error(`Deepgram API Error: ${desc}`);\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {\n const alts = parseTranscription(this.#opts.language || 'en', data, this.startTimeOffset);\n\n if (alts.length > 0) {\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [alts[0]!, ...alts.slice(1)],\n });\n }\n }\n\n #onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n\n #getDeepgramUrl(): string {\n const params: Record<string, string | string[]> = {\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: 'linear16',\n mip_opt_out: String(this.#opts.mipOptOut),\n };\n\n // Note: v2 API does NOT include 'language' parameter\n if (this.#opts.eagerEotThreshold)\n params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();\n if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();\n if (this.#opts.eotTimeoutMs) params.eot_timeout_ms = this.#opts.eotTimeoutMs.toString();\n\n if (this.#opts.keyterms.length > 0) params.keyterm = this.#opts.keyterms;\n if (this.#opts.tags && this.#opts.tags.length > 0) params.tag = this.#opts.tags;\n\n const baseUrl = this.#opts.endpointUrl.replace(/^http/, 'ws');\n const qs = queryString.stringify(params);\n return `${baseUrl}?${qs}`;\n }\n\n override close() {\n super.close();\n this.#ws?.close();\n }\n}\n\n// --- Helpers ---\n\nfunction parseTranscription(\n language: string,\n data: Record<string, unknown>,\n startTimeOffset: number,\n): stt.SpeechData[] {\n const transcript = data.transcript as string | undefined;\n const wordsData = (data.words as Array<Record<string, unknown>>) || [];\n\n if (!wordsData || wordsData.length === 0) {\n return [];\n }\n\n let confidence = 0;\n if (wordsData.length > 0) {\n const sum = wordsData.reduce((acc: number, w) => acc + ((w.confidence as number) || 0), 0);\n confidence = sum / wordsData.length;\n }\n\n const sd: stt.SpeechData = {\n language: language,\n startTime: ((data.audio_window_start as number) || 0) + startTimeOffset,\n endTime: ((data.audio_window_end as number) || 0) + startTimeOffset,\n confidence: confidence,\n text: transcript || '',\n // Note: Deepgram V2 (Flux) API does not provide word-level timing (start/end).\n // Words only contain 'word' and 'confidence' fields, so startTime/endTime will be 0.\n // See: https://developers.deepgram.com/docs/flux/nova-3-migration\n words: wordsData.map((word) => ({\n text: (word.word as string) ?? '',\n startTime: ((word.start as number) ?? 0) + startTimeOffset,\n endTime: ((word.end as number) ?? 0) + startTimeOffset,\n confidence: (word.confidence as number) ?? 0.0,\n startTimeOffset,\n })),\n };\n\n return [sd];\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,YAAY,iBAAiB;AAC7B,SAAS,iBAAiB;AAC1B,SAAS,yBAAyB;AAGlC,MAAM,aAAa,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC;AAqBzD,MAAM,sBAAoD;AAAA,EACxD,OAAO;AAAA,EACP,YAAY;AAAA,EACZ,UAAU,CAAC;AAAA,EACX,aAAa;AAAA,EACb,UAAU;AAAA,EACV,WAAW;AACb;AAEA,SAAS,aAAa,MAA0B;AAC9C,aAAW,OAAO,MAAM;AACtB,QAAI,IAAI,SAAS,KAAK;AACpB,YAAM,IAAI,MAAM,yCAAyC;AAAA,IAC3D;AAAA,EACF;AACA,SAAO;AACT;AAiCO,MAAM,cAAc,IAAI,IAAI;AAAA,EACxB,QAAQ;AAAA,EACjB;AAAA,EACA;AAAA,EACA,UAAU,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBd,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,IACrB,CAAC;AAED,SAAK,QAAQ,EAAE,GAAG,qBAAqB,GAAG,KAAK;AAE/C,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,8BAA8B;AAAA,IAChD;AACA,SAAK,UAAU;AAEf,QAAI,KAAK,MAAM,MAAM;AACnB,WAAK,MAAM,OAAO,aAAa,KAAK,MAAM,IAAI;AAAA,IAChD;AAAA,EACF;AAAA;AAAA,EAGA,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,MAAgB,WACd,QACA,cAC0B;AAC1B,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,OAAO,SAAiE;AACtE,UAAM,aAAa,EAAE,GAAG,KAAK,OAAO,QAAQ,KAAK,QAAQ;AACzD,WAAO,IAAI,eAAe,MAAM,YAAY,mCAAS,WAAW;AAAA,EAClE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAA6B;AACzC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AACvD,SAAK,QAAQ,MAAM,uBAAuB;AAAA,EAC5C;AACF;AAIA,MAAM,uBAAuB,IAAI,aAAa;AAAA,EACnC,QAAQ;AAAA,EACjB;AAAA,EACA,UAAU,IAAI;AAAA,EACd,MAAwB;AAAA,EAExB;AAAA,EACA,aAAa;AAAA,EACb,YAAY;AAAA;AAAA,EAGZ,kBAAkB,IAAI,MAAM;AAAA,EAE5B,YACE,aACA,MACA,aACA;AACA,UAAM,aAAa,KAAK,YAAY,WAAW;AAC/C,SAAK,QAAQ;AAEb,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,uBAAuB,QAAQ;AAAA,MAClD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,cAAc,MAA6B;AACzC,SAAK,QAAQ,MAAM,iCAAiC,IAAI;AACxD,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AAGvD,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,MAAgB,MAAM;AA7MxB;AA+MI,WAAO,CAAC,KAAK,QAAQ;AACnB,UAAI;AACF,aAAK,gBAAgB,MAAM;AAE3B,cAAM,MAAM,KAAK,gBAAgB;AACjC,aAAK,QAAQ,MAAM,2BAA2B,GAAG,EAAE;AAEnD,aAAK,MAAM,IAAI,UAAU,KAAK;AAAA,UAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,QACzD,CAAC;AAGD,cAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,cAAI,CAAC,KAAK,IAAK,QAAO,OAAO,IAAI,MAAM,2BAA2B,CAAC;AAEnE,gBAAM,SAAS,MAAM;AA9N/B,gBAAAA;AA+NY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,SAAS;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,UAAU,CAAC,QAAe;AAlO1C,gBAAAA;AAmOY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,QAAQ;AACtB,mBAAO,GAAG;AAAA,UACZ;AAEA,eAAK,IAAI,KAAK,QAAQ,MAAM;AAC5B,eAAK,IAAI,KAAK,SAAS,OAAO;AAAA,QAChC,CAAC;AAGD,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,gBAAgB,KAAK,gBAAgB,KAAK;AAGhD,cAAM,SAAS,MAAM,QAAQ,KAAK;AAAA,UAChC,QAAQ,IAAI,CAAC,aAAa,WAAW,CAAC;AAAA,UACtC,cAAc,KAAK,MAAM,WAAW;AAAA,QACtC,CAAC;AAED,YAAI,WAAW,aAAa;AAC1B,eAAK,QAAQ,MAAM,6CAA6C;AAEhE,eAAK,IAAI,MAAM;AAAA,QACjB,OAAO;AAEL;AAAA,QACF;AAAA,MACF,SAAS,OAAO;AACd,aAAK,QAAQ,MAAM,yBAAyB,EAAE,MAAM,CAAC;AACrD,cAAM;AAAA,MACR,UAAE;AACA,cAAI,UAAK,QAAL,mBAAU,gBAAe,UAAU,MAAM;AAC3C,eAAK,IAAI,MAAM;AAAA,QACjB;AAAA,MACF;AAAA,IACF;AACA,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAGf,UAAM,cAAc,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AACzD,UAAM,eAAe,IAAI,gBAAgB,KAAK,MAAM,YAAY,GAAG,WAAW;AAE9E,QAAI,WAAW;AAGf,UAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAElD,WAAO,MAAM;AACX,YAAM,cAAc,SAAS,KAAK;AAElC,YAAM,eAAe,KAAK,gBAAgB,KAAK,EAAE,KAAK,OAAO,EAAE,OAAO,KAAK,EAAW;AAEtF,YAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,aAAa,YAAY,CAAC;AAG7D,UAAI,WAAW,UAAU,OAAO,MAAM;AACpC,YAAI,EAAE,WAAW,WAAW,OAAO,MAAM;AAEvC,qBAAW;AAAA,QACb,OAAO;AAEL;AAAA,QACF;AAAA,MACF;AAGA,UAAI,YAAY,EAAE,WAAW,SAAS;AAAA,MAEtC,WAAW,WAAW,QAAQ;AAC5B,cAAM,OAAO,OAAO;AACpB,cAAM,SAAuB,CAAC;AAE9B,YAAI,SAAS,eAAe,gBAAgB;AAC1C,iBAAO,KAAK,GAAG,aAAa,MAAM,CAAC;AACnC,qBAAW;AAAA,QACb,OAAO;AACL,iBAAO,KAAK,GAAG,aAAa,MAAO,KAAoB,KAAK,MAAqB,CAAC;AAAA,QACpF;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,wBAAwB,KAAK,8BAA8B,KAAK,CAAC;AAEtE,cAAI,KAAK,IAAK,eAAe,UAAU,MAAM;AAC3C,iBAAK,IAAK,KAAK,MAAM,IAAI;AAAA,UAC3B;AAEA,cAAI,UAAU;AACZ,iBAAK,wBAAwB,MAAM;AACnC,uBAAW;AAAA,UACb;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAU;AAAA,IAChB;AAGA,QAAI,CAAC,KAAK,gBAAgB,SAAS,KAAK,IAAK,eAAe,UAAU,MAAM;AAC1E,WAAK,QAAQ,MAAM,yCAAyC;AAC5D,WAAK,IAAK,KAAK,UAAU;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAEf,WAAO,IAAI,QAAc,CAAC,YAAY;AACpC,UAAI,CAAC,KAAK,IAAK,QAAO,QAAQ;AAE9B,WAAK,IAAI,GAAG,WAAW,CAAC,MAAc,aAAsB;AAC1D,YAAI,UAAU;AACZ,eAAK,QAAQ,KAAK,kDAAkD;AACpE;AAAA,QACF;AACA,YAAI;AACF,gBAAM,MAAM,KAAK,MAAM,KAAK,SAAS,CAAC;AACtC,eAAK,oBAAoB,GAAG;AAAA,QAC9B,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,oCAAoC,EAAE,MAAM,CAAC;AAAA,QAClE;AAAA,MACF,CAAC;AAED,WAAK,IAAI,GAAG,SAAS,CAAC,MAAM,WAAW;AACrC,aAAK,QAAQ,MAAM,8BAA8B,IAAI,IAAI,MAAM,EAAE;AACjE,gBAAQ;AAAA,MACV,CAAC;AAGD,WAAK,IAAI,GAAG,SAAS,MAAM,QAAQ,CAAC;AAAA,IACtC,CAAC;AAAA,EACH;AAAA,EAEA,oBAAoB,MAA+B;AACjD,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,QAAI,KAAK,SAAS,YAAY;AAC5B,YAAM,YAAY,KAAK;AAEvB,UAAI,cAAc,eAAe;AAC/B,YAAI,KAAK,UAAW;AAEpB,aAAK,YAAY;AACjB,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,IAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAED,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,UAAU;AACjC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,kBAAkB;AACzC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,IAAI,gBAAgB,sBAAsB,IAAI;AAAA,MAC1E,WAAW,cAAc,eAAe;AACtC,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,aAAa;AACpC,YAAI,CAAC,KAAK,UAAW;AAErB,aAAK,YAAY;AACjB,aAAK,qBAAqB,IAAI,gBAAgB,kBAAkB,IAAI;AAEpE,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,IAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAAA,MACH;AAAA,IACF,WAAW,KAAK,SAAS,SAAS;AAChC,WAAK,QAAQ,KAAK,0BAA0B,EAAE,KAAK,CAAC;AACpD,YAAM,OAAQ,KAAK,eAA0B;AAC7C,YAAM,IAAI,MAAM,uBAAuB,IAAI,EAAE;AAAA,IAC/C;AAAA,EACF;AAAA,EAEA,qBAAqB,WAAgC,MAA+B;AAClF,UAAM,OAAO,mBAAmB,KAAK,MAAM,YAAY,MAAM,MAAM,KAAK,eAAe;AAEvF,QAAI,KAAK,SAAS,GAAG;AACnB,WAAK,MAAM,IAAI;AAAA,QACb,MAAM;AAAA,QACN,WAAW,KAAK;AAAA,QAChB,cAAc,CAAC,KAAK,CAAC,GAAI,GAAG,KAAK,MAAM,CAAC,CAAC;AAAA,MAC3C,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,uBAAuB,UAAkB;AACvC,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;AAAA,EAEA,kBAA0B;AACxB,UAAM,SAA4C;AAAA,MAChD,OAAO,KAAK,MAAM;AAAA,MAClB,aAAa,KAAK,MAAM,WAAW,SAAS;AAAA,MAC5C,UAAU;AAAA,MACV,aAAa,OAAO,KAAK,MAAM,SAAS;AAAA,IAC1C;AAGA,QAAI,KAAK,MAAM;AACb,aAAO,sBAAsB,KAAK,MAAM,kBAAkB,SAAS;AACrE,QAAI,KAAK,MAAM,aAAc,QAAO,gBAAgB,KAAK,MAAM,aAAa,SAAS;AACrF,QAAI,KAAK,MAAM,aAAc,QAAO,iBAAiB,KAAK,MAAM,aAAa,SAAS;AAEtF,QAAI,KAAK,MAAM,SAAS,SAAS,EAAG,QAAO,UAAU,KAAK,MAAM;AAChE,QAAI,KAAK,MAAM,QAAQ,KAAK,MAAM,KAAK,SAAS,EAAG,QAAO,MAAM,KAAK,MAAM;AAE3E,UAAM,UAAU,KAAK,MAAM,YAAY,QAAQ,SAAS,IAAI;AAC5D,UAAM,KAAK,YAAY,UAAU,MAAM;AACvC,WAAO,GAAG,OAAO,IAAI,EAAE;AAAA,EACzB;AAAA,EAES,QAAQ;AApcnB;AAqcI,UAAM,MAAM;AACZ,eAAK,QAAL,mBAAU;AAAA,EACZ;AACF;AAIA,SAAS,mBACP,UACA,MACA,iBACkB;AAClB,QAAM,aAAa,KAAK;AACxB,QAAM,YAAa,KAAK,SAA4C,CAAC;AAErE,MAAI,CAAC,aAAa,UAAU,WAAW,GAAG;AACxC,WAAO,CAAC;AAAA,EACV;AAEA,MAAI,aAAa;AACjB,MAAI,UAAU,SAAS,GAAG;AACxB,UAAM,MAAM,UAAU,OAAO,CAAC,KAAa,MAAM,OAAQ,EAAE,cAAyB,IAAI,CAAC;AACzF,iBAAa,MAAM,UAAU;AAAA,EAC/B;AAEA,QAAM,KAAqB;AAAA,IACzB;AAAA,IACA,YAAa,KAAK,sBAAiC,KAAK;AAAA,IACxD,UAAW,KAAK,oBAA+B,KAAK;AAAA,IACpD;AAAA,IACA,MAAM,cAAc;AAAA;AAAA;AAAA;AAAA,IAIpB,OAAO,UAAU,IAAI,CAAC,UAAU;AAAA,MAC9B,MAAO,KAAK,QAAmB;AAAA,MAC/B,YAAa,KAAK,SAAoB,KAAK;AAAA,MAC3C,UAAW,KAAK,OAAkB,KAAK;AAAA,MACvC,YAAa,KAAK,cAAyB;AAAA,MAC3C;AAAA,IACF,EAAE;AAAA,EACJ;AAEA,SAAO,CAAC,EAAE;AACZ;","names":["_a"]}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@livekit/agents-plugin-deepgram",
|
|
3
|
-
"version": "1.0.
|
|
3
|
+
"version": "1.0.38",
|
|
4
4
|
"description": "Deepgram plugin for LiveKit Agents for Node.js",
|
|
5
5
|
"main": "dist/index.js",
|
|
6
6
|
"require": "dist/index.cjs",
|
|
@@ -30,16 +30,16 @@
|
|
|
30
30
|
"@types/ws": "^8.5.10",
|
|
31
31
|
"tsup": "^8.3.5",
|
|
32
32
|
"typescript": "^5.0.0",
|
|
33
|
-
"@livekit/agents": "1.0.
|
|
34
|
-
"@livekit/agents-plugin-silero": "1.0.
|
|
35
|
-
"@livekit/agents-plugins-test": "1.0.
|
|
33
|
+
"@livekit/agents": "1.0.38",
|
|
34
|
+
"@livekit/agents-plugin-silero": "1.0.38",
|
|
35
|
+
"@livekit/agents-plugins-test": "1.0.38"
|
|
36
36
|
},
|
|
37
37
|
"dependencies": {
|
|
38
38
|
"ws": "^8.16.0"
|
|
39
39
|
},
|
|
40
40
|
"peerDependencies": {
|
|
41
41
|
"@livekit/rtc-node": "^0.13.24",
|
|
42
|
-
"@livekit/agents": "1.0.
|
|
42
|
+
"@livekit/agents": "1.0.38"
|
|
43
43
|
},
|
|
44
44
|
"scripts": {
|
|
45
45
|
"build": "tsup --onSuccess \"pnpm build:types\"",
|
package/src/stt.ts
CHANGED
|
@@ -73,6 +73,7 @@ export class STT extends stt.STT {
|
|
|
73
73
|
super({
|
|
74
74
|
streaming: true,
|
|
75
75
|
interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,
|
|
76
|
+
alignedTranscript: 'word',
|
|
76
77
|
});
|
|
77
78
|
if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {
|
|
78
79
|
throw new Error(
|
|
@@ -339,7 +340,11 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
339
340
|
const isEndpoint = json['speech_final'];
|
|
340
341
|
this.#requestId = requestId;
|
|
341
342
|
|
|
342
|
-
const alternatives = liveTranscriptionToSpeechData(
|
|
343
|
+
const alternatives = liveTranscriptionToSpeechData(
|
|
344
|
+
this.#opts.language!,
|
|
345
|
+
json,
|
|
346
|
+
this.startTimeOffset,
|
|
347
|
+
);
|
|
343
348
|
|
|
344
349
|
// If, for some reason, we didn't get a SpeechStarted event but we got
|
|
345
350
|
// a transcript with text, we should start speaking. It's rare but has
|
|
@@ -421,14 +426,28 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
421
426
|
const liveTranscriptionToSpeechData = (
|
|
422
427
|
language: STTLanguages | string,
|
|
423
428
|
data: { [id: string]: any },
|
|
429
|
+
startTimeOffset: number = 0,
|
|
424
430
|
): stt.SpeechData[] => {
|
|
425
431
|
const alts: any[] = data['channel']['alternatives'];
|
|
426
432
|
|
|
427
|
-
return alts.map((alt) =>
|
|
428
|
-
|
|
429
|
-
|
|
430
|
-
|
|
431
|
-
|
|
432
|
-
|
|
433
|
-
|
|
433
|
+
return alts.map((alt) => {
|
|
434
|
+
const wordsData: any[] = alt['words'] ?? [];
|
|
435
|
+
|
|
436
|
+
return {
|
|
437
|
+
language,
|
|
438
|
+
startTime: wordsData.length ? wordsData[0]['start'] + startTimeOffset : startTimeOffset,
|
|
439
|
+
endTime: wordsData.length
|
|
440
|
+
? wordsData[wordsData.length - 1]['end'] + startTimeOffset
|
|
441
|
+
: startTimeOffset,
|
|
442
|
+
confidence: alt['confidence'],
|
|
443
|
+
text: alt['transcript'],
|
|
444
|
+
words: wordsData.map((word) => ({
|
|
445
|
+
text: word['word'] ?? '',
|
|
446
|
+
startTime: (word['start'] ?? 0) + startTimeOffset,
|
|
447
|
+
endTime: (word['end'] ?? 0) + startTimeOffset,
|
|
448
|
+
confidence: word['confidence'] ?? 0.0,
|
|
449
|
+
startTimeOffset,
|
|
450
|
+
})),
|
|
451
|
+
};
|
|
452
|
+
});
|
|
434
453
|
};
|
package/src/stt_v2.ts
CHANGED
|
@@ -109,6 +109,7 @@ export class STTv2 extends stt.STT {
|
|
|
109
109
|
super({
|
|
110
110
|
streaming: true,
|
|
111
111
|
interimResults: true,
|
|
112
|
+
alignedTranscript: 'word',
|
|
112
113
|
});
|
|
113
114
|
|
|
114
115
|
this.#opts = { ...defaultSTTv2Options, ...opts };
|
|
@@ -405,9 +406,7 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
405
406
|
}
|
|
406
407
|
|
|
407
408
|
#sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {
|
|
408
|
-
|
|
409
|
-
// Using 0.0 for now - full parity would require base class changes
|
|
410
|
-
const alts = parseTranscription(this.#opts.language || 'en', data, 0.0);
|
|
409
|
+
const alts = parseTranscription(this.#opts.language || 'en', data, this.startTimeOffset);
|
|
411
410
|
|
|
412
411
|
if (alts.length > 0) {
|
|
413
412
|
this.queue.put({
|
|
@@ -437,7 +436,7 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
437
436
|
mip_opt_out: String(this.#opts.mipOptOut),
|
|
438
437
|
};
|
|
439
438
|
|
|
440
|
-
|
|
439
|
+
// Note: v2 API does NOT include 'language' parameter
|
|
441
440
|
if (this.#opts.eagerEotThreshold)
|
|
442
441
|
params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();
|
|
443
442
|
if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();
|
|
@@ -465,16 +464,16 @@ function parseTranscription(
|
|
|
465
464
|
startTimeOffset: number,
|
|
466
465
|
): stt.SpeechData[] {
|
|
467
466
|
const transcript = data.transcript as string | undefined;
|
|
468
|
-
const
|
|
467
|
+
const wordsData = (data.words as Array<Record<string, unknown>>) || [];
|
|
469
468
|
|
|
470
|
-
if (!
|
|
469
|
+
if (!wordsData || wordsData.length === 0) {
|
|
471
470
|
return [];
|
|
472
471
|
}
|
|
473
472
|
|
|
474
473
|
let confidence = 0;
|
|
475
|
-
if (
|
|
476
|
-
const sum =
|
|
477
|
-
confidence = sum /
|
|
474
|
+
if (wordsData.length > 0) {
|
|
475
|
+
const sum = wordsData.reduce((acc: number, w) => acc + ((w.confidence as number) || 0), 0);
|
|
476
|
+
confidence = sum / wordsData.length;
|
|
478
477
|
}
|
|
479
478
|
|
|
480
479
|
const sd: stt.SpeechData = {
|
|
@@ -483,6 +482,16 @@ function parseTranscription(
|
|
|
483
482
|
endTime: ((data.audio_window_end as number) || 0) + startTimeOffset,
|
|
484
483
|
confidence: confidence,
|
|
485
484
|
text: transcript || '',
|
|
485
|
+
// Note: Deepgram V2 (Flux) API does not provide word-level timing (start/end).
|
|
486
|
+
// Words only contain 'word' and 'confidence' fields, so startTime/endTime will be 0.
|
|
487
|
+
// See: https://developers.deepgram.com/docs/flux/nova-3-migration
|
|
488
|
+
words: wordsData.map((word) => ({
|
|
489
|
+
text: (word.word as string) ?? '',
|
|
490
|
+
startTime: ((word.start as number) ?? 0) + startTimeOffset,
|
|
491
|
+
endTime: ((word.end as number) ?? 0) + startTimeOffset,
|
|
492
|
+
confidence: (word.confidence as number) ?? 0.0,
|
|
493
|
+
startTimeOffset,
|
|
494
|
+
})),
|
|
486
495
|
};
|
|
487
496
|
|
|
488
497
|
return [sd];
|