@livekit/agents-plugin-deepgram 1.5.5 → 1.6.1
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/index.cjs +1 -1
- package/dist/index.js +1 -1
- package/dist/stt.cjs +6 -1
- package/dist/stt.cjs.map +1 -1
- package/dist/stt.d.ts.map +1 -1
- package/dist/stt.js +6 -1
- package/dist/stt.js.map +1 -1
- package/dist/stt.test.cjs +65 -0
- package/dist/stt.test.cjs.map +1 -1
- package/dist/stt.test.js +65 -0
- package/dist/stt.test.js.map +1 -1
- package/dist/stt_v2.cjs +7 -5
- package/dist/stt_v2.cjs.map +1 -1
- package/dist/stt_v2.js +7 -5
- package/dist/stt_v2.js.map +1 -1
- package/package.json +13 -10
- package/src/stt.test.ts +76 -0
- package/src/stt.ts +7 -1
- package/src/stt_v2.ts +8 -6
package/dist/index.cjs
CHANGED
package/dist/index.js
CHANGED
package/dist/stt.cjs
CHANGED
|
@@ -336,9 +336,10 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
336
336
|
}
|
|
337
337
|
const data = result.value;
|
|
338
338
|
let frames;
|
|
339
|
+
let hasEnded = false;
|
|
339
340
|
if (data === SpeechStream.FLUSH_SENTINEL) {
|
|
340
341
|
frames = stream.flush();
|
|
341
|
-
|
|
342
|
+
hasEnded = true;
|
|
342
343
|
} else if (data.sampleRate === this.#opts.sampleRate && data.channels === this.#opts.numChannels) {
|
|
343
344
|
frames = stream.write(data.data.buffer);
|
|
344
345
|
} else {
|
|
@@ -351,6 +352,10 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
351
352
|
ws.send(frame.data.buffer);
|
|
352
353
|
}
|
|
353
354
|
}
|
|
355
|
+
if (hasEnded) {
|
|
356
|
+
this.#audioDurationCollector.flush();
|
|
357
|
+
ws.send(JSON.stringify({ type: "Finalize" }));
|
|
358
|
+
}
|
|
354
359
|
}
|
|
355
360
|
} finally {
|
|
356
361
|
closing = true;
|
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 createTimedString,\n getBaseLanguage,\n log,\n mergeFrames,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { STTLanguages, STTModels } from './models.js';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels | string;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n /**\n * Redact sensitive information from the transcription. Accepts a single value or list of\n * values. Supported values: \"pci\", \"numbers\", \"ssn\", \"true\" (redact all).\n * See https://developers.deepgram.com/docs/redaction for details.\n */\n redact: string | string[];\n utteranceEndMs?: number | null;\n replace?: Record<string, string> | null;\n search?: string[] | null;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n mipOptOut: boolean;\n baseUrl: string;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.DEEPGRAM_API_KEY,\n language: 'en-US',\n detectLanguage: false,\n interimResults: true,\n punctuate: true,\n model: 'nova-3',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n redact: [],\n utteranceEndMs: null,\n replace: null,\n search: null,\n dictation: false,\n diarize: false,\n numerals: false,\n mipOptOut: false,\n baseUrl: 'wss://api.deepgram.com',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n // session keyterm propagation)\n #streams = new Set<WeakRef<SpeechStream>>();\n #userKeyterm: string[];\n #sessionKeyterms: string[] = [];\n label = 'deepgram.STT';\n private abortController = new AbortController();\n\n get model(): string {\n return this.#opts.model;\n }\n\n get provider(): string {\n return 'Deepgram';\n }\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n alignedTranscript: 'word',\n keyterms: true,\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n language: opts.language ? normalizeLanguage(opts.language) : defaultSTTOptions.language,\n };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else {\n this.#opts.model = this.#validateModel(this.#opts.model, this.#opts.language);\n }\n\n this.#userKeyterm = [...this.#opts.keyterm];\n }\n\n async _recognize(buffer: AudioBuffer, abortSignal?: AbortSignal): Promise<stt.SpeechEvent> {\n const frame = mergeFrames(buffer);\n const listenURL = new URL(`${this.#opts.baseUrl.replace(/^ws/, 'http')}/v1/listen`);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n detect_language: this.#opts.detectLanguage,\n smart_format: this.#opts.smartFormat,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n profanity_filter: this.#opts.profanityFilter,\n numerals: this.#opts.numerals,\n mip_opt_out: this.#opts.mipOptOut,\n language: this.#opts.language,\n ...(this.#opts.keyterm.length > 0 ? { keyterm: this.#opts.keyterm } : {}),\n ...(this.#opts.redact.length > 0 ? { redact: this.#opts.redact } : {}),\n };\n\n appendSearchParams(listenURL, params);\n\n if (this.#opts.diarize) {\n this.#logger.warn('speaker diarization is not supported in non-streaming mode, ignoring');\n }\n\n const resp = await fetch(listenURL, {\n method: 'POST',\n headers: {\n Authorization: `Token ${this.#opts.apiKey}`,\n Accept: 'application/json',\n 'Content-Type': 'audio/wav',\n },\n body: new Blob([new Uint8Array(createWav(frame))], { type: 'audio/wav' }),\n signal: abortSignal ?? null,\n });\n\n if (!resp.ok) {\n throw new Error(\n `Deepgram STT REST request failed with status ${resp.status}: ${await resp.text()}`,\n );\n }\n\n return prerecordedTranscriptionToSpeechEvent(this.#opts.language, await resp.json());\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n const language =\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language;\n const model =\n opts.model !== undefined || opts.language !== undefined\n ? this.#validateModel(opts.model ?? this.#opts.model, language)\n : this.#opts.model;\n\n const nextOpts = { ...opts };\n if (nextOpts.keyterm !== undefined) {\n this.#userKeyterm = [...nextOpts.keyterm];\n nextOpts.keyterm = [...new Set([...this.#userKeyterm, ...this.#sessionKeyterms])];\n }\n\n this.#opts = {\n ...this.#opts,\n ...nextOpts,\n language,\n model,\n };\n }\n\n override _updateSessionKeyterms(keyterms: string[]): void {\n if (\n keyterms.length === this.#sessionKeyterms.length &&\n keyterms.every((t, i) => t === this.#sessionKeyterms[i])\n ) {\n return;\n }\n this.#sessionKeyterms = [...keyterms];\n const merged = [...new Set([...this.#userKeyterm, ...keyterms])];\n this.#opts.keyterm = merged;\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (!stream) {\n this.#streams.delete(ref);\n continue;\n }\n if (stream._speaking) {\n // defer the reconnect to the end of the utterance so we don't cut it off\n stream._pendingKeyterm = merged;\n } else {\n stream.updateOptions({ keyterm: merged });\n }\n }\n }\n\n #validateModel(model: STTModels | string, language?: string) {\n if (\n language &&\n getBaseLanguage(language) !== 'en' &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(model)\n ) {\n this.#logger.warn(\n `${model} does not support language ${language}, falling back to nova-2-general`,\n );\n return 'nova-2-general';\n }\n\n return model;\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n // Deepgram only supports language detection for prerecorded audio, not streaming.\n if (this.#opts.detectLanguage || this.#opts.language === undefined) {\n throw new Error(\n 'language detection is not supported in streaming mode, please disable it and specify a language',\n );\n }\n const stream = new SpeechStream(this, this.#opts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\n }\n\n async close() {\n this.abortController.abort();\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #audioEnergyFilter: AudioEnergyFilter;\n #logger = log();\n #speaking = false;\n #resetWS = new Future();\n #requestId = '';\n // keyterms set while the user is speaking; applied at END_OF_SPEECH (latest wins)\n /** @internal */\n _pendingKeyterm: string[] | null = null;\n #audioDurationCollector: PeriodicCollector<number>;\n label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(stt, opts.sampleRate, connOptions);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n let ws: WebSocket;\n\n while (!this.input.closed && !this.closed) {\n const streamURL = new URL(`${this.#opts.baseUrl}/v1/listen`);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n utterance_end_ms: this.#opts.utteranceEndMs,\n replace: this.#opts.replace,\n search: this.#opts.search,\n language: this.#opts.language,\n mip_opt_out: this.#opts.mipOptOut,\n ...(this.#opts.redact.length > 0 ? { redact: this.#opts.redact } : {}),\n };\n appendSearchParams(streamURL, params);\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (!this.closed && !this.input.closed) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n }\n\n const delay = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n } else {\n this.#logger.warn(\n `Deepgram disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n );\n }\n }\n }\n\n this.closed = true;\n }\n\n /** @internal */\n get _speaking(): boolean {\n return this.#speaking;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n if (opts.keyterm !== undefined) {\n this._pendingKeyterm = null;\n }\n this.#resetWS.resolve();\n }\n\n #onEndOfSpeech() {\n if (this._pendingKeyterm !== null) {\n this.updateOptions({ keyterm: this._pendingKeyterm });\n this._pendingKeyterm = null;\n }\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 this.#onEndOfSpeech();\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n case 'UtteranceEnd': {\n if (this.#speaking) {\n this.#speaking = false;\n putMessage({ type: stt.SpeechEventType.END_OF_SPEECH });\n this.#onEndOfSpeech();\n }\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n\n if (this.closed || closing) {\n resolve();\n }\n } catch (err) {\n this.#logger.error(`STT: Error processing message: ${msg}`);\n reject(err);\n }\n });\n });\n\n await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n }, this.abortController);\n\n await Promise.race([\n this.#resetWS.await,\n Promise.all([sendTask(), listenTask.result, wsMonitor]),\n ]);\n closing = true;\n ws.close();\n clearInterval(keepalive);\n }\n\n private onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n}\n\nfunction appendSearchParams(url: URL, params: Record<string, unknown>) {\n Object.entries(params).forEach(([key, value]) => {\n if (value === undefined || value === null) return;\n\n if (Array.isArray(value)) {\n value.forEach((item) => url.searchParams.append(key, String(item)));\n } else if (typeof value === 'object') {\n Object.entries(value).forEach(([term, replacement]) => {\n url.searchParams.append(key, `${term}:${replacement}`);\n });\n } else {\n url.searchParams.append(key, String(value));\n }\n });\n}\n\nfunction createWav(frame: AudioFrame): Buffer {\n const bitsPerSample = 16;\n const byteRate = (frame.sampleRate * frame.channels * bitsPerSample) / 8;\n const blockAlign = (frame.channels * bitsPerSample) / 8;\n\n const header = Buffer.alloc(44);\n header.write('RIFF', 0);\n header.writeUInt32LE(36 + frame.data.byteLength, 4);\n header.write('WAVE', 8);\n header.write('fmt ', 12);\n header.writeUInt32LE(16, 16);\n header.writeUInt16LE(1, 20);\n header.writeUInt16LE(frame.channels, 22);\n header.writeUInt32LE(frame.sampleRate, 24);\n header.writeUInt32LE(byteRate, 28);\n header.writeUInt16LE(blockAlign, 32);\n header.writeUInt16LE(bitsPerSample, 34);\n header.write('data', 36);\n header.writeUInt32LE(frame.data.byteLength, 40);\n\n const pcm = Buffer.from(frame.data.buffer, frame.data.byteOffset, frame.data.byteLength);\n return Buffer.concat([header, pcm]);\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n startTimeOffset: number = 0,\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => {\n const wordsData: any[] = alt['words'] ?? [];\n const detectedLanguage =\n language === 'multi' ? alt['languages']?.[0] ?? language : alt['language'] ?? language;\n\n return {\n language: normalizeLanguage(detectedLanguage),\n startTime: wordsData.length ? wordsData[0]['start'] + startTimeOffset : startTimeOffset,\n endTime: wordsData.length\n ? wordsData[wordsData.length - 1]['end'] + startTimeOffset\n : startTimeOffset,\n confidence: alt['confidence'],\n text: alt['transcript'],\n words: wordsData.map((word) =>\n createTimedString({\n text: word['word'] ?? '',\n startTime: (word['start'] ?? 0) + startTimeOffset,\n endTime: (word['end'] ?? 0) + startTimeOffset,\n confidence: word['confidence'] ?? 0.0,\n startTimeOffset,\n }),\n ),\n };\n });\n};\n\nfunction prerecordedTranscriptionToSpeechEvent(\n language: STTLanguages | string | undefined,\n data: { [id: string]: any },\n): stt.SpeechEvent {\n const channel = data['results']['channels'][0];\n const detectedLanguage = channel['detected_language'] ?? '';\n const alternatives = channel['alternatives'].map((alt: { [id: string]: any }) => {\n const wordsData: any[] = alt['words'] ?? [];\n\n return {\n language: normalizeLanguage(language ?? detectedLanguage),\n startTime: wordsData.length ? wordsData[0]['start'] : 0,\n endTime: wordsData.length ? wordsData[wordsData.length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n words: wordsData.map((word) =>\n createTimedString({\n text: word['word'] ?? '',\n startTime: word['start'] ?? 0,\n endTime: word['end'] ?? 0,\n confidence: word['confidence'] ?? 0.0,\n }),\n ),\n };\n }) as stt.SpeechData[];\n\n return {\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n requestId: data['metadata']['request_id'],\n ...(alternatives.length\n ? { alternatives: alternatives as [stt.SpeechData, ...stt.SpeechData[]] }\n : {}),\n };\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAcO;AAEP,gBAA0B;AAC1B,mBAAkC;AAmClC,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,QAAQ,CAAC;AAAA,EACT,gBAAgB;AAAA,EAChB,SAAS;AAAA,EACT,QAAQ;AAAA,EACR,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AAAA,EACV,WAAW;AAAA,EACX,SAAS;AACX;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,cAAU,mBAAI;AAAA;AAAA,EAEd,WAAW,oBAAI,IAA2B;AAAA,EAC1C;AAAA,EACA,mBAA6B,CAAC;AAAA,EAC9B,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAE9C,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,MACzD,mBAAmB;AAAA,MACnB,UAAU;AAAA,IACZ,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,KAAK,eAAW,iCAAkB,KAAK,QAAQ,IAAI,kBAAkB;AAAA,IACjF;AAEA,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,OAAO;AACL,WAAK,MAAM,QAAQ,KAAK,eAAe,KAAK,MAAM,OAAO,KAAK,MAAM,QAAQ;AAAA,IAC9E;AAEA,SAAK,eAAe,CAAC,GAAG,KAAK,MAAM,OAAO;AAAA,EAC5C;AAAA,EAEA,MAAM,WAAW,QAAqB,aAAqD;AACzF,UAAM,YAAQ,2BAAY,MAAM;AAChC,UAAM,YAAY,IAAI,IAAI,GAAG,KAAK,MAAM,QAAQ,QAAQ,OAAO,MAAM,CAAC,YAAY;AAClF,UAAM,SAAS;AAAA,MACb,OAAO,KAAK,MAAM;AAAA,MAClB,WAAW,KAAK,MAAM;AAAA,MACtB,iBAAiB,KAAK,MAAM;AAAA,MAC5B,cAAc,KAAK,MAAM;AAAA,MACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,MACpD,kBAAkB,KAAK,MAAM;AAAA,MAC7B,UAAU,KAAK,MAAM;AAAA,MACrB,aAAa,KAAK,MAAM;AAAA,MACxB,UAAU,KAAK,MAAM;AAAA,MACrB,GAAI,KAAK,MAAM,QAAQ,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,IAAI,CAAC;AAAA,MACvE,GAAI,KAAK,MAAM,OAAO,SAAS,IAAI,EAAE,QAAQ,KAAK,MAAM,OAAO,IAAI,CAAC;AAAA,IACtE;AAEA,uBAAmB,WAAW,MAAM;AAEpC,QAAI,KAAK,MAAM,SAAS;AACtB,WAAK,QAAQ,KAAK,sEAAsE;AAAA,IAC1F;AAEA,UAAM,OAAO,MAAM,MAAM,WAAW;AAAA,MAClC,QAAQ;AAAA,MACR,SAAS;AAAA,QACP,eAAe,SAAS,KAAK,MAAM,MAAM;AAAA,QACzC,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA,MAAM,IAAI,KAAK,CAAC,IAAI,WAAW,UAAU,KAAK,CAAC,CAAC,GAAG,EAAE,MAAM,YAAY,CAAC;AAAA,MACxE,QAAQ,eAAe;AAAA,IACzB,CAAC;AAED,QAAI,CAAC,KAAK,IAAI;AACZ,YAAM,IAAI;AAAA,QACR,gDAAgD,KAAK,MAAM,KAAK,MAAM,KAAK,KAAK,CAAC;AAAA,MACnF;AAAA,IACF;AAEA,WAAO,sCAAsC,KAAK,MAAM,UAAU,MAAM,KAAK,KAAK,CAAC;AAAA,EACrF;AAAA,EAEA,cAAc,MAA2B;AACvC,UAAM,WACJ,KAAK,aAAa,aAAY,iCAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAC9E,UAAM,QACJ,KAAK,UAAU,UAAa,KAAK,aAAa,SAC1C,KAAK,eAAe,KAAK,SAAS,KAAK,MAAM,OAAO,QAAQ,IAC5D,KAAK,MAAM;AAEjB,UAAM,WAAW,EAAE,GAAG,KAAK;AAC3B,QAAI,SAAS,YAAY,QAAW;AAClC,WAAK,eAAe,CAAC,GAAG,SAAS,OAAO;AACxC,eAAS,UAAU,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,cAAc,GAAG,KAAK,gBAAgB,CAAC,CAAC;AAAA,IAClF;AAEA,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAES,uBAAuB,UAA0B;AACxD,QACE,SAAS,WAAW,KAAK,iBAAiB,UAC1C,SAAS,MAAM,CAAC,GAAG,MAAM,MAAM,KAAK,iBAAiB,CAAC,CAAC,GACvD;AACA;AAAA,IACF;AACA,SAAK,mBAAmB,CAAC,GAAG,QAAQ;AACpC,UAAM,SAAS,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,cAAc,GAAG,QAAQ,CAAC,CAAC;AAC/D,SAAK,MAAM,UAAU;AACrB,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,CAAC,QAAQ;AACX,aAAK,SAAS,OAAO,GAAG;AACxB;AAAA,MACF;AACA,UAAI,OAAO,WAAW;AAEpB,eAAO,kBAAkB;AAAA,MAC3B,OAAO;AACL,eAAO,cAAc,EAAE,SAAS,OAAO,CAAC;AAAA,MAC1C;AAAA,IACF;AAAA,EACF;AAAA,EAEA,eAAe,OAA2B,UAAmB;AAC3D,QACE,gBACA,+BAAgB,QAAQ,MAAM,QAC9B;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,GAChB;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,8BAA8B,QAAQ;AAAA,MAChD;AACA,aAAO;AAAA,IACT;AAEA,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,SAA6D;AAElE,QAAI,KAAK,MAAM,kBAAkB,KAAK,MAAM,aAAa,QAAW;AAClE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,UAAM,SAAS,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AACtE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;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;AAAA;AAAA,EAGb,kBAAmC;AAAA,EACnC;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,GAAG,KAAK,MAAM,OAAO,YAAY;AAC3D,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,kBAAkB,KAAK,MAAM;AAAA,QAC7B,SAAS,KAAK,MAAM;AAAA,QACpB,QAAQ,KAAK,MAAM;AAAA,QACnB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM;AAAA,QACxB,GAAI,KAAK,MAAM,OAAO,SAAS,IAAI,EAAE,QAAQ,KAAK,MAAM,OAAO,IAAI,CAAC;AAAA,MACtE;AACA,yBAAmB,WAAW,MAAM;AAEpC,WAAK,IAAI,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;AAAA,EAGA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,YAAY,QAAW;AAC9B,WAAK,kBAAkB;AAAA,IACzB;AACA,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,iBAAiB;AACf,QAAI,KAAK,oBAAoB,MAAM;AACjC,WAAK,cAAc,EAAE,SAAS,KAAK,gBAAgB,CAAC;AACpD,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;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;AACtD,uBAAK,eAAe;AAAA,gBACtB;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,KAAK,gBAAgB;AACnB,oBAAI,KAAK,WAAW;AAClB,uBAAK,YAAY;AACjB,6BAAW,EAAE,MAAM,kBAAI,gBAAgB,cAAc,CAAC;AACtD,uBAAK,eAAe;AAAA,gBACtB;AACA;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAEA,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,kCAAkC,GAAG,EAAE;AAC1D,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,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,SAAS,mBAAmB,KAAU,QAAiC;AACrE,SAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,KAAK,KAAK,MAAM;AAC/C,QAAI,UAAU,UAAa,UAAU,KAAM;AAE3C,QAAI,MAAM,QAAQ,KAAK,GAAG;AACxB,YAAM,QAAQ,CAAC,SAAS,IAAI,aAAa,OAAO,KAAK,OAAO,IAAI,CAAC,CAAC;AAAA,IACpE,WAAW,OAAO,UAAU,UAAU;AACpC,aAAO,QAAQ,KAAK,EAAE,QAAQ,CAAC,CAAC,MAAM,WAAW,MAAM;AACrD,YAAI,aAAa,OAAO,KAAK,GAAG,IAAI,IAAI,WAAW,EAAE;AAAA,MACvD,CAAC;AAAA,IACH,OAAO;AACL,UAAI,aAAa,OAAO,KAAK,OAAO,KAAK,CAAC;AAAA,IAC5C;AAAA,EACF,CAAC;AACH;AAEA,SAAS,UAAU,OAA2B;AAC5C,QAAM,gBAAgB;AACtB,QAAM,WAAY,MAAM,aAAa,MAAM,WAAW,gBAAiB;AACvE,QAAM,aAAc,MAAM,WAAW,gBAAiB;AAEtD,QAAM,SAAS,OAAO,MAAM,EAAE;AAC9B,SAAO,MAAM,QAAQ,CAAC;AACtB,SAAO,cAAc,KAAK,MAAM,KAAK,YAAY,CAAC;AAClD,SAAO,MAAM,QAAQ,CAAC;AACtB,SAAO,MAAM,QAAQ,EAAE;AACvB,SAAO,cAAc,IAAI,EAAE;AAC3B,SAAO,cAAc,GAAG,EAAE;AAC1B,SAAO,cAAc,MAAM,UAAU,EAAE;AACvC,SAAO,cAAc,MAAM,YAAY,EAAE;AACzC,SAAO,cAAc,UAAU,EAAE;AACjC,SAAO,cAAc,YAAY,EAAE;AACnC,SAAO,cAAc,eAAe,EAAE;AACtC,SAAO,MAAM,QAAQ,EAAE;AACvB,SAAO,cAAc,MAAM,KAAK,YAAY,EAAE;AAE9C,QAAM,MAAM,OAAO,KAAK,MAAM,KAAK,QAAQ,MAAM,KAAK,YAAY,MAAM,KAAK,UAAU;AACvF,SAAO,OAAO,OAAO,CAAC,QAAQ,GAAG,CAAC;AACpC;AAEA,MAAM,gCAAgC,CACpC,UACA,MACA,kBAA0B,MACL;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,QAAQ;AApnB3B;AAqnBI,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAC1C,UAAM,mBACJ,aAAa,YAAU,SAAI,WAAW,MAAf,mBAAmB,OAAM,WAAW,IAAI,UAAU,KAAK;AAEhF,WAAO;AAAA,MACL,cAAU,iCAAkB,gBAAgB;AAAA,MAC5C,WAAW,UAAU,SAAS,UAAU,CAAC,EAAE,OAAO,IAAI,kBAAkB;AAAA,MACxE,SAAS,UAAU,SACf,UAAU,UAAU,SAAS,CAAC,EAAE,KAAK,IAAI,kBACzC;AAAA,MACJ,YAAY,IAAI,YAAY;AAAA,MAC5B,MAAM,IAAI,YAAY;AAAA,MACtB,OAAO,UAAU;AAAA,QAAI,CAAC,aACpB,iCAAkB;AAAA,UAChB,MAAM,KAAK,MAAM,KAAK;AAAA,UACtB,YAAY,KAAK,OAAO,KAAK,KAAK;AAAA,UAClC,UAAU,KAAK,KAAK,KAAK,KAAK;AAAA,UAC9B,YAAY,KAAK,YAAY,KAAK;AAAA,UAClC;AAAA,QACF,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF,CAAC;AACH;AAEA,SAAS,sCACP,UACA,MACiB;AACjB,QAAM,UAAU,KAAK,SAAS,EAAE,UAAU,EAAE,CAAC;AAC7C,QAAM,mBAAmB,QAAQ,mBAAmB,KAAK;AACzD,QAAM,eAAe,QAAQ,cAAc,EAAE,IAAI,CAAC,QAA+B;AAC/E,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAE1C,WAAO;AAAA,MACL,cAAU,iCAAkB,YAAY,gBAAgB;AAAA,MACxD,WAAW,UAAU,SAAS,UAAU,CAAC,EAAE,OAAO,IAAI;AAAA,MACtD,SAAS,UAAU,SAAS,UAAU,UAAU,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,MACrE,YAAY,IAAI,YAAY;AAAA,MAC5B,MAAM,IAAI,YAAY;AAAA,MACtB,OAAO,UAAU;AAAA,QAAI,CAAC,aACpB,iCAAkB;AAAA,UAChB,MAAM,KAAK,MAAM,KAAK;AAAA,UACtB,WAAW,KAAK,OAAO,KAAK;AAAA,UAC5B,SAAS,KAAK,KAAK,KAAK;AAAA,UACxB,YAAY,KAAK,YAAY,KAAK;AAAA,QACpC,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF,CAAC;AAED,SAAO;AAAA,IACL,MAAM,kBAAI,gBAAgB;AAAA,IAC1B,WAAW,KAAK,UAAU,EAAE,YAAY;AAAA,IACxC,GAAI,aAAa,SACb,EAAE,aAAoE,IACtE,CAAC;AAAA,EACP;AACF;","names":["stt"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n type AudioBuffer,\n AudioByteStream,\n AudioEnergyFilter,\n Future,\n Task,\n createTimedString,\n getBaseLanguage,\n log,\n mergeFrames,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { STTLanguages, STTModels } from './models.js';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels | string;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n /**\n * Redact sensitive information from the transcription. Accepts a single value or list of\n * values. Supported values: \"pci\", \"numbers\", \"ssn\", \"true\" (redact all).\n * See https://developers.deepgram.com/docs/redaction for details.\n */\n redact: string | string[];\n utteranceEndMs?: number | null;\n replace?: Record<string, string> | null;\n search?: string[] | null;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n mipOptOut: boolean;\n baseUrl: string;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.DEEPGRAM_API_KEY,\n language: 'en-US',\n detectLanguage: false,\n interimResults: true,\n punctuate: true,\n model: 'nova-3',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n redact: [],\n utteranceEndMs: null,\n replace: null,\n search: null,\n dictation: false,\n diarize: false,\n numerals: false,\n mipOptOut: false,\n baseUrl: 'wss://api.deepgram.com',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n // session keyterm propagation)\n #streams = new Set<WeakRef<SpeechStream>>();\n #userKeyterm: string[];\n #sessionKeyterms: string[] = [];\n label = 'deepgram.STT';\n private abortController = new AbortController();\n\n get model(): string {\n return this.#opts.model;\n }\n\n get provider(): string {\n return 'Deepgram';\n }\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n alignedTranscript: 'word',\n keyterms: true,\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n language: opts.language ? normalizeLanguage(opts.language) : defaultSTTOptions.language,\n };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else {\n this.#opts.model = this.#validateModel(this.#opts.model, this.#opts.language);\n }\n\n this.#userKeyterm = [...this.#opts.keyterm];\n }\n\n async _recognize(buffer: AudioBuffer, abortSignal?: AbortSignal): Promise<stt.SpeechEvent> {\n const frame = mergeFrames(buffer);\n const listenURL = new URL(`${this.#opts.baseUrl.replace(/^ws/, 'http')}/v1/listen`);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n detect_language: this.#opts.detectLanguage,\n smart_format: this.#opts.smartFormat,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n profanity_filter: this.#opts.profanityFilter,\n numerals: this.#opts.numerals,\n mip_opt_out: this.#opts.mipOptOut,\n language: this.#opts.language,\n ...(this.#opts.keyterm.length > 0 ? { keyterm: this.#opts.keyterm } : {}),\n ...(this.#opts.redact.length > 0 ? { redact: this.#opts.redact } : {}),\n };\n\n appendSearchParams(listenURL, params);\n\n if (this.#opts.diarize) {\n this.#logger.warn('speaker diarization is not supported in non-streaming mode, ignoring');\n }\n\n const resp = await fetch(listenURL, {\n method: 'POST',\n headers: {\n Authorization: `Token ${this.#opts.apiKey}`,\n Accept: 'application/json',\n 'Content-Type': 'audio/wav',\n },\n body: new Blob([new Uint8Array(createWav(frame))], { type: 'audio/wav' }),\n signal: abortSignal ?? null,\n });\n\n if (!resp.ok) {\n throw new Error(\n `Deepgram STT REST request failed with status ${resp.status}: ${await resp.text()}`,\n );\n }\n\n return prerecordedTranscriptionToSpeechEvent(this.#opts.language, await resp.json());\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n const language =\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language;\n const model =\n opts.model !== undefined || opts.language !== undefined\n ? this.#validateModel(opts.model ?? this.#opts.model, language)\n : this.#opts.model;\n\n const nextOpts = { ...opts };\n if (nextOpts.keyterm !== undefined) {\n this.#userKeyterm = [...nextOpts.keyterm];\n nextOpts.keyterm = [...new Set([...this.#userKeyterm, ...this.#sessionKeyterms])];\n }\n\n this.#opts = {\n ...this.#opts,\n ...nextOpts,\n language,\n model,\n };\n }\n\n override _updateSessionKeyterms(keyterms: string[]): void {\n if (\n keyterms.length === this.#sessionKeyterms.length &&\n keyterms.every((t, i) => t === this.#sessionKeyterms[i])\n ) {\n return;\n }\n this.#sessionKeyterms = [...keyterms];\n const merged = [...new Set([...this.#userKeyterm, ...keyterms])];\n this.#opts.keyterm = merged;\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (!stream) {\n this.#streams.delete(ref);\n continue;\n }\n if (stream._speaking) {\n // defer the reconnect to the end of the utterance so we don't cut it off\n stream._pendingKeyterm = merged;\n } else {\n stream.updateOptions({ keyterm: merged });\n }\n }\n }\n\n #validateModel(model: STTModels | string, language?: string) {\n if (\n language &&\n getBaseLanguage(language) !== 'en' &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(model)\n ) {\n this.#logger.warn(\n `${model} does not support language ${language}, falling back to nova-2-general`,\n );\n return 'nova-2-general';\n }\n\n return model;\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n // Deepgram only supports language detection for prerecorded audio, not streaming.\n if (this.#opts.detectLanguage || this.#opts.language === undefined) {\n throw new Error(\n 'language detection is not supported in streaming mode, please disable it and specify a language',\n );\n }\n const stream = new SpeechStream(this, this.#opts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\n }\n\n async close() {\n this.abortController.abort();\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #audioEnergyFilter: AudioEnergyFilter;\n #logger = log();\n #speaking = false;\n #resetWS = new Future();\n #requestId = '';\n // keyterms set while the user is speaking; applied at END_OF_SPEECH (latest wins)\n /** @internal */\n _pendingKeyterm: string[] | null = null;\n #audioDurationCollector: PeriodicCollector<number>;\n label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(stt, opts.sampleRate, connOptions);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n let ws: WebSocket;\n\n while (!this.input.closed && !this.closed) {\n const streamURL = new URL(`${this.#opts.baseUrl}/v1/listen`);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n utterance_end_ms: this.#opts.utteranceEndMs,\n replace: this.#opts.replace,\n search: this.#opts.search,\n language: this.#opts.language,\n mip_opt_out: this.#opts.mipOptOut,\n ...(this.#opts.redact.length > 0 ? { redact: this.#opts.redact } : {}),\n };\n appendSearchParams(streamURL, params);\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (!this.closed && !this.input.closed) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n }\n\n const delay = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n } else {\n this.#logger.warn(\n `Deepgram disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n );\n }\n }\n }\n\n this.closed = true;\n }\n\n /** @internal */\n get _speaking(): boolean {\n return this.#speaking;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n if (opts.keyterm !== undefined) {\n this._pendingKeyterm = null;\n }\n this.#resetWS.resolve();\n }\n\n #onEndOfSpeech() {\n if (this._pendingKeyterm !== null) {\n this.updateOptions({ keyterm: this._pendingKeyterm });\n this._pendingKeyterm = null;\n }\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 let hasEnded = false;\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = stream.flush();\n hasEnded = true;\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 if (hasEnded) {\n this.#audioDurationCollector.flush();\n ws.send(JSON.stringify({ type: 'Finalize' }));\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 this.#onEndOfSpeech();\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n case 'UtteranceEnd': {\n if (this.#speaking) {\n this.#speaking = false;\n putMessage({ type: stt.SpeechEventType.END_OF_SPEECH });\n this.#onEndOfSpeech();\n }\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n\n if (this.closed || closing) {\n resolve();\n }\n } catch (err) {\n this.#logger.error(`STT: Error processing message: ${msg}`);\n reject(err);\n }\n });\n });\n\n await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n }, this.abortController);\n\n await Promise.race([\n this.#resetWS.await,\n Promise.all([sendTask(), listenTask.result, wsMonitor]),\n ]);\n closing = true;\n ws.close();\n clearInterval(keepalive);\n }\n\n private onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n}\n\nfunction appendSearchParams(url: URL, params: Record<string, unknown>) {\n Object.entries(params).forEach(([key, value]) => {\n if (value === undefined || value === null) return;\n\n if (Array.isArray(value)) {\n value.forEach((item) => url.searchParams.append(key, String(item)));\n } else if (typeof value === 'object') {\n Object.entries(value).forEach(([term, replacement]) => {\n url.searchParams.append(key, `${term}:${replacement}`);\n });\n } else {\n url.searchParams.append(key, String(value));\n }\n });\n}\n\nfunction createWav(frame: AudioFrame): Buffer {\n const bitsPerSample = 16;\n const byteRate = (frame.sampleRate * frame.channels * bitsPerSample) / 8;\n const blockAlign = (frame.channels * bitsPerSample) / 8;\n\n const header = Buffer.alloc(44);\n header.write('RIFF', 0);\n header.writeUInt32LE(36 + frame.data.byteLength, 4);\n header.write('WAVE', 8);\n header.write('fmt ', 12);\n header.writeUInt32LE(16, 16);\n header.writeUInt16LE(1, 20);\n header.writeUInt16LE(frame.channels, 22);\n header.writeUInt32LE(frame.sampleRate, 24);\n header.writeUInt32LE(byteRate, 28);\n header.writeUInt16LE(blockAlign, 32);\n header.writeUInt16LE(bitsPerSample, 34);\n header.write('data', 36);\n header.writeUInt32LE(frame.data.byteLength, 40);\n\n const pcm = Buffer.from(frame.data.buffer, frame.data.byteOffset, frame.data.byteLength);\n return Buffer.concat([header, pcm]);\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n startTimeOffset: number = 0,\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => {\n const wordsData: any[] = alt['words'] ?? [];\n const detectedLanguage =\n language === 'multi' ? alt['languages']?.[0] ?? language : alt['language'] ?? language;\n\n return {\n language: normalizeLanguage(detectedLanguage),\n startTime: wordsData.length ? wordsData[0]['start'] + startTimeOffset : startTimeOffset,\n endTime: wordsData.length\n ? wordsData[wordsData.length - 1]['end'] + startTimeOffset\n : startTimeOffset,\n confidence: alt['confidence'],\n text: alt['transcript'],\n words: wordsData.map((word) =>\n createTimedString({\n text: word['word'] ?? '',\n startTime: (word['start'] ?? 0) + startTimeOffset,\n endTime: (word['end'] ?? 0) + startTimeOffset,\n confidence: word['confidence'] ?? 0.0,\n startTimeOffset,\n }),\n ),\n };\n });\n};\n\nfunction prerecordedTranscriptionToSpeechEvent(\n language: STTLanguages | string | undefined,\n data: { [id: string]: any },\n): stt.SpeechEvent {\n const channel = data['results']['channels'][0];\n const detectedLanguage = channel['detected_language'] ?? '';\n const alternatives = channel['alternatives'].map((alt: { [id: string]: any }) => {\n const wordsData: any[] = alt['words'] ?? [];\n\n return {\n language: normalizeLanguage(language ?? detectedLanguage),\n startTime: wordsData.length ? wordsData[0]['start'] : 0,\n endTime: wordsData.length ? wordsData[wordsData.length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n words: wordsData.map((word) =>\n createTimedString({\n text: word['word'] ?? '',\n startTime: word['start'] ?? 0,\n endTime: word['end'] ?? 0,\n confidence: word['confidence'] ?? 0.0,\n }),\n ),\n };\n }) as stt.SpeechData[];\n\n return {\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n requestId: data['metadata']['request_id'],\n ...(alternatives.length\n ? { alternatives: alternatives as [stt.SpeechData, ...stt.SpeechData[]] }\n : {}),\n };\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAcO;AAEP,gBAA0B;AAC1B,mBAAkC;AAmClC,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,QAAQ,CAAC;AAAA,EACT,gBAAgB;AAAA,EAChB,SAAS;AAAA,EACT,QAAQ;AAAA,EACR,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AAAA,EACV,WAAW;AAAA,EACX,SAAS;AACX;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,cAAU,mBAAI;AAAA;AAAA,EAEd,WAAW,oBAAI,IAA2B;AAAA,EAC1C;AAAA,EACA,mBAA6B,CAAC;AAAA,EAC9B,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAE9C,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,MACzD,mBAAmB;AAAA,MACnB,UAAU;AAAA,IACZ,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,KAAK,eAAW,iCAAkB,KAAK,QAAQ,IAAI,kBAAkB;AAAA,IACjF;AAEA,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,OAAO;AACL,WAAK,MAAM,QAAQ,KAAK,eAAe,KAAK,MAAM,OAAO,KAAK,MAAM,QAAQ;AAAA,IAC9E;AAEA,SAAK,eAAe,CAAC,GAAG,KAAK,MAAM,OAAO;AAAA,EAC5C;AAAA,EAEA,MAAM,WAAW,QAAqB,aAAqD;AACzF,UAAM,YAAQ,2BAAY,MAAM;AAChC,UAAM,YAAY,IAAI,IAAI,GAAG,KAAK,MAAM,QAAQ,QAAQ,OAAO,MAAM,CAAC,YAAY;AAClF,UAAM,SAAS;AAAA,MACb,OAAO,KAAK,MAAM;AAAA,MAClB,WAAW,KAAK,MAAM;AAAA,MACtB,iBAAiB,KAAK,MAAM;AAAA,MAC5B,cAAc,KAAK,MAAM;AAAA,MACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,MACpD,kBAAkB,KAAK,MAAM;AAAA,MAC7B,UAAU,KAAK,MAAM;AAAA,MACrB,aAAa,KAAK,MAAM;AAAA,MACxB,UAAU,KAAK,MAAM;AAAA,MACrB,GAAI,KAAK,MAAM,QAAQ,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,IAAI,CAAC;AAAA,MACvE,GAAI,KAAK,MAAM,OAAO,SAAS,IAAI,EAAE,QAAQ,KAAK,MAAM,OAAO,IAAI,CAAC;AAAA,IACtE;AAEA,uBAAmB,WAAW,MAAM;AAEpC,QAAI,KAAK,MAAM,SAAS;AACtB,WAAK,QAAQ,KAAK,sEAAsE;AAAA,IAC1F;AAEA,UAAM,OAAO,MAAM,MAAM,WAAW;AAAA,MAClC,QAAQ;AAAA,MACR,SAAS;AAAA,QACP,eAAe,SAAS,KAAK,MAAM,MAAM;AAAA,QACzC,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA,MAAM,IAAI,KAAK,CAAC,IAAI,WAAW,UAAU,KAAK,CAAC,CAAC,GAAG,EAAE,MAAM,YAAY,CAAC;AAAA,MACxE,QAAQ,eAAe;AAAA,IACzB,CAAC;AAED,QAAI,CAAC,KAAK,IAAI;AACZ,YAAM,IAAI;AAAA,QACR,gDAAgD,KAAK,MAAM,KAAK,MAAM,KAAK,KAAK,CAAC;AAAA,MACnF;AAAA,IACF;AAEA,WAAO,sCAAsC,KAAK,MAAM,UAAU,MAAM,KAAK,KAAK,CAAC;AAAA,EACrF;AAAA,EAEA,cAAc,MAA2B;AACvC,UAAM,WACJ,KAAK,aAAa,aAAY,iCAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAC9E,UAAM,QACJ,KAAK,UAAU,UAAa,KAAK,aAAa,SAC1C,KAAK,eAAe,KAAK,SAAS,KAAK,MAAM,OAAO,QAAQ,IAC5D,KAAK,MAAM;AAEjB,UAAM,WAAW,EAAE,GAAG,KAAK;AAC3B,QAAI,SAAS,YAAY,QAAW;AAClC,WAAK,eAAe,CAAC,GAAG,SAAS,OAAO;AACxC,eAAS,UAAU,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,cAAc,GAAG,KAAK,gBAAgB,CAAC,CAAC;AAAA,IAClF;AAEA,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAES,uBAAuB,UAA0B;AACxD,QACE,SAAS,WAAW,KAAK,iBAAiB,UAC1C,SAAS,MAAM,CAAC,GAAG,MAAM,MAAM,KAAK,iBAAiB,CAAC,CAAC,GACvD;AACA;AAAA,IACF;AACA,SAAK,mBAAmB,CAAC,GAAG,QAAQ;AACpC,UAAM,SAAS,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,cAAc,GAAG,QAAQ,CAAC,CAAC;AAC/D,SAAK,MAAM,UAAU;AACrB,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,CAAC,QAAQ;AACX,aAAK,SAAS,OAAO,GAAG;AACxB;AAAA,MACF;AACA,UAAI,OAAO,WAAW;AAEpB,eAAO,kBAAkB;AAAA,MAC3B,OAAO;AACL,eAAO,cAAc,EAAE,SAAS,OAAO,CAAC;AAAA,MAC1C;AAAA,IACF;AAAA,EACF;AAAA,EAEA,eAAe,OAA2B,UAAmB;AAC3D,QACE,gBACA,+BAAgB,QAAQ,MAAM,QAC9B;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,GAChB;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,8BAA8B,QAAQ;AAAA,MAChD;AACA,aAAO;AAAA,IACT;AAEA,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,SAA6D;AAElE,QAAI,KAAK,MAAM,kBAAkB,KAAK,MAAM,aAAa,QAAW;AAClE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,UAAM,SAAS,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AACtE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;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;AAAA;AAAA,EAGb,kBAAmC;AAAA,EACnC;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,GAAG,KAAK,MAAM,OAAO,YAAY;AAC3D,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,kBAAkB,KAAK,MAAM;AAAA,QAC7B,SAAS,KAAK,MAAM;AAAA,QACpB,QAAQ,KAAK,MAAM;AAAA,QACnB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM;AAAA,QACxB,GAAI,KAAK,MAAM,OAAO,SAAS,IAAI,EAAE,QAAQ,KAAK,MAAM,OAAO,IAAI,CAAC;AAAA,MACtE;AACA,yBAAmB,WAAW,MAAM;AAEpC,WAAK,IAAI,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;AAAA,EAGA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,YAAY,QAAW;AAC9B,WAAK,kBAAkB;AAAA,IACzB;AACA,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,iBAAiB;AACf,QAAI,KAAK,oBAAoB,MAAM;AACjC,WAAK,cAAc,EAAE,SAAS,KAAK,gBAAgB,CAAC;AACpD,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;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,WAAW;AACf,cAAI,SAAS,aAAa,gBAAgB;AACxC,qBAAS,OAAO,MAAM;AACtB,uBAAW;AAAA,UACb,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;AAEA,cAAI,UAAU;AACZ,iBAAK,wBAAwB,MAAM;AACnC,eAAG,KAAK,KAAK,UAAU,EAAE,MAAM,WAAW,CAAC,CAAC;AAAA,UAC9C;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;AACtD,uBAAK,eAAe;AAAA,gBACtB;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,KAAK,gBAAgB;AACnB,oBAAI,KAAK,WAAW;AAClB,uBAAK,YAAY;AACjB,6BAAW,EAAE,MAAM,kBAAI,gBAAgB,cAAc,CAAC;AACtD,uBAAK,eAAe;AAAA,gBACtB;AACA;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAEA,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,kCAAkC,GAAG,EAAE;AAC1D,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,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,SAAS,mBAAmB,KAAU,QAAiC;AACrE,SAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,KAAK,KAAK,MAAM;AAC/C,QAAI,UAAU,UAAa,UAAU,KAAM;AAE3C,QAAI,MAAM,QAAQ,KAAK,GAAG;AACxB,YAAM,QAAQ,CAAC,SAAS,IAAI,aAAa,OAAO,KAAK,OAAO,IAAI,CAAC,CAAC;AAAA,IACpE,WAAW,OAAO,UAAU,UAAU;AACpC,aAAO,QAAQ,KAAK,EAAE,QAAQ,CAAC,CAAC,MAAM,WAAW,MAAM;AACrD,YAAI,aAAa,OAAO,KAAK,GAAG,IAAI,IAAI,WAAW,EAAE;AAAA,MACvD,CAAC;AAAA,IACH,OAAO;AACL,UAAI,aAAa,OAAO,KAAK,OAAO,KAAK,CAAC;AAAA,IAC5C;AAAA,EACF,CAAC;AACH;AAEA,SAAS,UAAU,OAA2B;AAC5C,QAAM,gBAAgB;AACtB,QAAM,WAAY,MAAM,aAAa,MAAM,WAAW,gBAAiB;AACvE,QAAM,aAAc,MAAM,WAAW,gBAAiB;AAEtD,QAAM,SAAS,OAAO,MAAM,EAAE;AAC9B,SAAO,MAAM,QAAQ,CAAC;AACtB,SAAO,cAAc,KAAK,MAAM,KAAK,YAAY,CAAC;AAClD,SAAO,MAAM,QAAQ,CAAC;AACtB,SAAO,MAAM,QAAQ,EAAE;AACvB,SAAO,cAAc,IAAI,EAAE;AAC3B,SAAO,cAAc,GAAG,EAAE;AAC1B,SAAO,cAAc,MAAM,UAAU,EAAE;AACvC,SAAO,cAAc,MAAM,YAAY,EAAE;AACzC,SAAO,cAAc,UAAU,EAAE;AACjC,SAAO,cAAc,YAAY,EAAE;AACnC,SAAO,cAAc,eAAe,EAAE;AACtC,SAAO,MAAM,QAAQ,EAAE;AACvB,SAAO,cAAc,MAAM,KAAK,YAAY,EAAE;AAE9C,QAAM,MAAM,OAAO,KAAK,MAAM,KAAK,QAAQ,MAAM,KAAK,YAAY,MAAM,KAAK,UAAU;AACvF,SAAO,OAAO,OAAO,CAAC,QAAQ,GAAG,CAAC;AACpC;AAEA,MAAM,gCAAgC,CACpC,UACA,MACA,kBAA0B,MACL;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,QAAQ;AA1nB3B;AA2nBI,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAC1C,UAAM,mBACJ,aAAa,YAAU,SAAI,WAAW,MAAf,mBAAmB,OAAM,WAAW,IAAI,UAAU,KAAK;AAEhF,WAAO;AAAA,MACL,cAAU,iCAAkB,gBAAgB;AAAA,MAC5C,WAAW,UAAU,SAAS,UAAU,CAAC,EAAE,OAAO,IAAI,kBAAkB;AAAA,MACxE,SAAS,UAAU,SACf,UAAU,UAAU,SAAS,CAAC,EAAE,KAAK,IAAI,kBACzC;AAAA,MACJ,YAAY,IAAI,YAAY;AAAA,MAC5B,MAAM,IAAI,YAAY;AAAA,MACtB,OAAO,UAAU;AAAA,QAAI,CAAC,aACpB,iCAAkB;AAAA,UAChB,MAAM,KAAK,MAAM,KAAK;AAAA,UACtB,YAAY,KAAK,OAAO,KAAK,KAAK;AAAA,UAClC,UAAU,KAAK,KAAK,KAAK,KAAK;AAAA,UAC9B,YAAY,KAAK,YAAY,KAAK;AAAA,UAClC;AAAA,QACF,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF,CAAC;AACH;AAEA,SAAS,sCACP,UACA,MACiB;AACjB,QAAM,UAAU,KAAK,SAAS,EAAE,UAAU,EAAE,CAAC;AAC7C,QAAM,mBAAmB,QAAQ,mBAAmB,KAAK;AACzD,QAAM,eAAe,QAAQ,cAAc,EAAE,IAAI,CAAC,QAA+B;AAC/E,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAE1C,WAAO;AAAA,MACL,cAAU,iCAAkB,YAAY,gBAAgB;AAAA,MACxD,WAAW,UAAU,SAAS,UAAU,CAAC,EAAE,OAAO,IAAI;AAAA,MACtD,SAAS,UAAU,SAAS,UAAU,UAAU,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,MACrE,YAAY,IAAI,YAAY;AAAA,MAC5B,MAAM,IAAI,YAAY;AAAA,MACtB,OAAO,UAAU;AAAA,QAAI,CAAC,aACpB,iCAAkB;AAAA,UAChB,MAAM,KAAK,MAAM,KAAK;AAAA,UACtB,WAAW,KAAK,OAAO,KAAK;AAAA,UAC5B,SAAS,KAAK,KAAK,KAAK;AAAA,UACxB,YAAY,KAAK,YAAY,KAAK;AAAA,QACpC,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF,CAAC;AAED,SAAO;AAAA,IACL,MAAM,kBAAI,gBAAgB;AAAA,IAC1B,WAAW,KAAK,UAAU,EAAE,YAAY;AAAA,IACxC,GAAI,aAAa,SACb,EAAE,aAAoE,IACtE,CAAC;AAAA,EACP;AACF;","names":["stt"]}
|
package/dist/stt.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,EAUhB,GAAG,EAEJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,KAAK,EAAE,YAAY,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAE3D,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,GAAG,MAAM,CAAC;IAC1B,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;;;;OAIG;IACH,MAAM,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;IAC1B,cAAc,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;IAC/B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,GAAG,IAAI,CAAC;IACxC,MAAM,CAAC,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC;IACzB,SAAS,EAAE,OAAO,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,QAAQ,EAAE,OAAO,CAAC;IAClB,SAAS,EAAE,OAAO,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;CACjB;AA6BD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAO9B,KAAK,SAAkB;IACvB,OAAO,CAAC,eAAe,CAAyB;IAEhD,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,IAAI,QAAQ,IAAI,MAAM,CAErB;gBAEW,IAAI,GAAE,OAAO,CAAC,UAAU,CAAqB;IA4BnD,UAAU,CAAC,MAAM,EAAE,WAAW,EAAE,WAAW,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IA2C1F,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAsB9B,sBAAsB,CAAC,QAAQ,EAAE,MAAM,EAAE,GAAG,IAAI;IAmDzD,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,YAAY;IAY7D,KAAK;CAGZ;AAED,qBAAa,YAAa,SAAQ,GAAG,CAAC,YAAY;;IAQhD,gBAAgB;IAChB,eAAe,EAAE,MAAM,EAAE,GAAG,IAAI,CAAQ;IAExC,KAAK,SAA2B;gBAEpB,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU,EAAE,WAAW,CAAC,EAAE,iBAAiB;cAWvD,GAAG;IAsEnB,gBAAgB;IAChB,IAAI,SAAS,IAAI,OAAO,CAEvB;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;
|
|
1
|
+
{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,EACtB,KAAK,WAAW,EAUhB,GAAG,EAEJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,KAAK,EAAE,YAAY,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAE3D,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,GAAG,MAAM,CAAC;IAC1B,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;;;;OAIG;IACH,MAAM,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;IAC1B,cAAc,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;IAC/B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,GAAG,IAAI,CAAC;IACxC,MAAM,CAAC,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC;IACzB,SAAS,EAAE,OAAO,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,QAAQ,EAAE,OAAO,CAAC;IAClB,SAAS,EAAE,OAAO,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;CACjB;AA6BD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAO9B,KAAK,SAAkB;IACvB,OAAO,CAAC,eAAe,CAAyB;IAEhD,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,IAAI,QAAQ,IAAI,MAAM,CAErB;gBAEW,IAAI,GAAE,OAAO,CAAC,UAAU,CAAqB;IA4BnD,UAAU,CAAC,MAAM,EAAE,WAAW,EAAE,WAAW,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IA2C1F,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAsB9B,sBAAsB,CAAC,QAAQ,EAAE,MAAM,EAAE,GAAG,IAAI;IAmDzD,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,YAAY;IAY7D,KAAK;CAGZ;AAED,qBAAa,YAAa,SAAQ,GAAG,CAAC,YAAY;;IAQhD,gBAAgB;IAChB,eAAe,EAAE,MAAM,EAAE,GAAG,IAAI,CAAQ;IAExC,KAAK,SAA2B;gBAEpB,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU,EAAE,WAAW,CAAC,EAAE,iBAAiB;cAWvD,GAAG;IAsEnB,gBAAgB;IAChB,IAAI,SAAS,IAAI,OAAO,CAEvB;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAuNvC,OAAO,CAAC,qBAAqB;CAU9B"}
|
package/dist/stt.js
CHANGED
|
@@ -324,9 +324,10 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
324
324
|
}
|
|
325
325
|
const data = result.value;
|
|
326
326
|
let frames;
|
|
327
|
+
let hasEnded = false;
|
|
327
328
|
if (data === SpeechStream.FLUSH_SENTINEL) {
|
|
328
329
|
frames = stream.flush();
|
|
329
|
-
|
|
330
|
+
hasEnded = true;
|
|
330
331
|
} else if (data.sampleRate === this.#opts.sampleRate && data.channels === this.#opts.numChannels) {
|
|
331
332
|
frames = stream.write(data.data.buffer);
|
|
332
333
|
} else {
|
|
@@ -339,6 +340,10 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
339
340
|
ws.send(frame.data.buffer);
|
|
340
341
|
}
|
|
341
342
|
}
|
|
343
|
+
if (hasEnded) {
|
|
344
|
+
this.#audioDurationCollector.flush();
|
|
345
|
+
ws.send(JSON.stringify({ type: "Finalize" }));
|
|
346
|
+
}
|
|
342
347
|
}
|
|
343
348
|
} finally {
|
|
344
349
|
closing = true;
|
package/dist/stt.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n type AudioBuffer,\n AudioByteStream,\n AudioEnergyFilter,\n Future,\n Task,\n createTimedString,\n getBaseLanguage,\n log,\n mergeFrames,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { STTLanguages, STTModels } from './models.js';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels | string;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n /**\n * Redact sensitive information from the transcription. Accepts a single value or list of\n * values. Supported values: \"pci\", \"numbers\", \"ssn\", \"true\" (redact all).\n * See https://developers.deepgram.com/docs/redaction for details.\n */\n redact: string | string[];\n utteranceEndMs?: number | null;\n replace?: Record<string, string> | null;\n search?: string[] | null;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n mipOptOut: boolean;\n baseUrl: string;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.DEEPGRAM_API_KEY,\n language: 'en-US',\n detectLanguage: false,\n interimResults: true,\n punctuate: true,\n model: 'nova-3',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n redact: [],\n utteranceEndMs: null,\n replace: null,\n search: null,\n dictation: false,\n diarize: false,\n numerals: false,\n mipOptOut: false,\n baseUrl: 'wss://api.deepgram.com',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n // session keyterm propagation)\n #streams = new Set<WeakRef<SpeechStream>>();\n #userKeyterm: string[];\n #sessionKeyterms: string[] = [];\n label = 'deepgram.STT';\n private abortController = new AbortController();\n\n get model(): string {\n return this.#opts.model;\n }\n\n get provider(): string {\n return 'Deepgram';\n }\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n alignedTranscript: 'word',\n keyterms: true,\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n language: opts.language ? normalizeLanguage(opts.language) : defaultSTTOptions.language,\n };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else {\n this.#opts.model = this.#validateModel(this.#opts.model, this.#opts.language);\n }\n\n this.#userKeyterm = [...this.#opts.keyterm];\n }\n\n async _recognize(buffer: AudioBuffer, abortSignal?: AbortSignal): Promise<stt.SpeechEvent> {\n const frame = mergeFrames(buffer);\n const listenURL = new URL(`${this.#opts.baseUrl.replace(/^ws/, 'http')}/v1/listen`);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n detect_language: this.#opts.detectLanguage,\n smart_format: this.#opts.smartFormat,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n profanity_filter: this.#opts.profanityFilter,\n numerals: this.#opts.numerals,\n mip_opt_out: this.#opts.mipOptOut,\n language: this.#opts.language,\n ...(this.#opts.keyterm.length > 0 ? { keyterm: this.#opts.keyterm } : {}),\n ...(this.#opts.redact.length > 0 ? { redact: this.#opts.redact } : {}),\n };\n\n appendSearchParams(listenURL, params);\n\n if (this.#opts.diarize) {\n this.#logger.warn('speaker diarization is not supported in non-streaming mode, ignoring');\n }\n\n const resp = await fetch(listenURL, {\n method: 'POST',\n headers: {\n Authorization: `Token ${this.#opts.apiKey}`,\n Accept: 'application/json',\n 'Content-Type': 'audio/wav',\n },\n body: new Blob([new Uint8Array(createWav(frame))], { type: 'audio/wav' }),\n signal: abortSignal ?? null,\n });\n\n if (!resp.ok) {\n throw new Error(\n `Deepgram STT REST request failed with status ${resp.status}: ${await resp.text()}`,\n );\n }\n\n return prerecordedTranscriptionToSpeechEvent(this.#opts.language, await resp.json());\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n const language =\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language;\n const model =\n opts.model !== undefined || opts.language !== undefined\n ? this.#validateModel(opts.model ?? this.#opts.model, language)\n : this.#opts.model;\n\n const nextOpts = { ...opts };\n if (nextOpts.keyterm !== undefined) {\n this.#userKeyterm = [...nextOpts.keyterm];\n nextOpts.keyterm = [...new Set([...this.#userKeyterm, ...this.#sessionKeyterms])];\n }\n\n this.#opts = {\n ...this.#opts,\n ...nextOpts,\n language,\n model,\n };\n }\n\n override _updateSessionKeyterms(keyterms: string[]): void {\n if (\n keyterms.length === this.#sessionKeyterms.length &&\n keyterms.every((t, i) => t === this.#sessionKeyterms[i])\n ) {\n return;\n }\n this.#sessionKeyterms = [...keyterms];\n const merged = [...new Set([...this.#userKeyterm, ...keyterms])];\n this.#opts.keyterm = merged;\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (!stream) {\n this.#streams.delete(ref);\n continue;\n }\n if (stream._speaking) {\n // defer the reconnect to the end of the utterance so we don't cut it off\n stream._pendingKeyterm = merged;\n } else {\n stream.updateOptions({ keyterm: merged });\n }\n }\n }\n\n #validateModel(model: STTModels | string, language?: string) {\n if (\n language &&\n getBaseLanguage(language) !== 'en' &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(model)\n ) {\n this.#logger.warn(\n `${model} does not support language ${language}, falling back to nova-2-general`,\n );\n return 'nova-2-general';\n }\n\n return model;\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n // Deepgram only supports language detection for prerecorded audio, not streaming.\n if (this.#opts.detectLanguage || this.#opts.language === undefined) {\n throw new Error(\n 'language detection is not supported in streaming mode, please disable it and specify a language',\n );\n }\n const stream = new SpeechStream(this, this.#opts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\n }\n\n async close() {\n this.abortController.abort();\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #audioEnergyFilter: AudioEnergyFilter;\n #logger = log();\n #speaking = false;\n #resetWS = new Future();\n #requestId = '';\n // keyterms set while the user is speaking; applied at END_OF_SPEECH (latest wins)\n /** @internal */\n _pendingKeyterm: string[] | null = null;\n #audioDurationCollector: PeriodicCollector<number>;\n label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(stt, opts.sampleRate, connOptions);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n let ws: WebSocket;\n\n while (!this.input.closed && !this.closed) {\n const streamURL = new URL(`${this.#opts.baseUrl}/v1/listen`);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n utterance_end_ms: this.#opts.utteranceEndMs,\n replace: this.#opts.replace,\n search: this.#opts.search,\n language: this.#opts.language,\n mip_opt_out: this.#opts.mipOptOut,\n ...(this.#opts.redact.length > 0 ? { redact: this.#opts.redact } : {}),\n };\n appendSearchParams(streamURL, params);\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (!this.closed && !this.input.closed) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n }\n\n const delay = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n } else {\n this.#logger.warn(\n `Deepgram disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n );\n }\n }\n }\n\n this.closed = true;\n }\n\n /** @internal */\n get _speaking(): boolean {\n return this.#speaking;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n if (opts.keyterm !== undefined) {\n this._pendingKeyterm = null;\n }\n this.#resetWS.resolve();\n }\n\n #onEndOfSpeech() {\n if (this._pendingKeyterm !== null) {\n this.updateOptions({ keyterm: this._pendingKeyterm });\n this._pendingKeyterm = null;\n }\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 this.#onEndOfSpeech();\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n case 'UtteranceEnd': {\n if (this.#speaking) {\n this.#speaking = false;\n putMessage({ type: stt.SpeechEventType.END_OF_SPEECH });\n this.#onEndOfSpeech();\n }\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n\n if (this.closed || closing) {\n resolve();\n }\n } catch (err) {\n this.#logger.error(`STT: Error processing message: ${msg}`);\n reject(err);\n }\n });\n });\n\n await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n }, this.abortController);\n\n await Promise.race([\n this.#resetWS.await,\n Promise.all([sendTask(), listenTask.result, wsMonitor]),\n ]);\n closing = true;\n ws.close();\n clearInterval(keepalive);\n }\n\n private onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n}\n\nfunction appendSearchParams(url: URL, params: Record<string, unknown>) {\n Object.entries(params).forEach(([key, value]) => {\n if (value === undefined || value === null) return;\n\n if (Array.isArray(value)) {\n value.forEach((item) => url.searchParams.append(key, String(item)));\n } else if (typeof value === 'object') {\n Object.entries(value).forEach(([term, replacement]) => {\n url.searchParams.append(key, `${term}:${replacement}`);\n });\n } else {\n url.searchParams.append(key, String(value));\n }\n });\n}\n\nfunction createWav(frame: AudioFrame): Buffer {\n const bitsPerSample = 16;\n const byteRate = (frame.sampleRate * frame.channels * bitsPerSample) / 8;\n const blockAlign = (frame.channels * bitsPerSample) / 8;\n\n const header = Buffer.alloc(44);\n header.write('RIFF', 0);\n header.writeUInt32LE(36 + frame.data.byteLength, 4);\n header.write('WAVE', 8);\n header.write('fmt ', 12);\n header.writeUInt32LE(16, 16);\n header.writeUInt16LE(1, 20);\n header.writeUInt16LE(frame.channels, 22);\n header.writeUInt32LE(frame.sampleRate, 24);\n header.writeUInt32LE(byteRate, 28);\n header.writeUInt16LE(blockAlign, 32);\n header.writeUInt16LE(bitsPerSample, 34);\n header.write('data', 36);\n header.writeUInt32LE(frame.data.byteLength, 40);\n\n const pcm = Buffer.from(frame.data.buffer, frame.data.byteOffset, frame.data.byteLength);\n return Buffer.concat([header, pcm]);\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n startTimeOffset: number = 0,\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => {\n const wordsData: any[] = alt['words'] ?? [];\n const detectedLanguage =\n language === 'multi' ? alt['languages']?.[0] ?? language : alt['language'] ?? language;\n\n return {\n language: normalizeLanguage(detectedLanguage),\n startTime: wordsData.length ? wordsData[0]['start'] + startTimeOffset : startTimeOffset,\n endTime: wordsData.length\n ? wordsData[wordsData.length - 1]['end'] + startTimeOffset\n : startTimeOffset,\n confidence: alt['confidence'],\n text: alt['transcript'],\n words: wordsData.map((word) =>\n createTimedString({\n text: word['word'] ?? '',\n startTime: (word['start'] ?? 0) + startTimeOffset,\n endTime: (word['end'] ?? 0) + startTimeOffset,\n confidence: word['confidence'] ?? 0.0,\n startTimeOffset,\n }),\n ),\n };\n });\n};\n\nfunction prerecordedTranscriptionToSpeechEvent(\n language: STTLanguages | string | undefined,\n data: { [id: string]: any },\n): stt.SpeechEvent {\n const channel = data['results']['channels'][0];\n const detectedLanguage = channel['detected_language'] ?? '';\n const alternatives = channel['alternatives'].map((alt: { [id: string]: any }) => {\n const wordsData: any[] = alt['words'] ?? [];\n\n return {\n language: normalizeLanguage(language ?? detectedLanguage),\n startTime: wordsData.length ? wordsData[0]['start'] : 0,\n endTime: wordsData.length ? wordsData[wordsData.length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n words: wordsData.map((word) =>\n createTimedString({\n text: word['word'] ?? '',\n startTime: word['start'] ?? 0,\n endTime: word['end'] ?? 0,\n confidence: word['confidence'] ?? 0.0,\n }),\n ),\n };\n }) as stt.SpeechData[];\n\n return {\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n requestId: data['metadata']['request_id'],\n ...(alternatives.length\n ? { alternatives: alternatives as [stt.SpeechData, ...stt.SpeechData[]] }\n : {}),\n };\n}\n"],"mappings":"AAGA;AAAA,EAGE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,iBAAiB;AAC1B,SAAS,yBAAyB;AAmClC,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,QAAQ,CAAC;AAAA,EACT,gBAAgB;AAAA,EAChB,SAAS;AAAA,EACT,QAAQ;AAAA,EACR,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AAAA,EACV,WAAW;AAAA,EACX,SAAS;AACX;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU,IAAI;AAAA;AAAA,EAEd,WAAW,oBAAI,IAA2B;AAAA,EAC1C;AAAA,EACA,mBAA6B,CAAC;AAAA,EAC9B,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAE9C,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,MACzD,mBAAmB;AAAA,MACnB,UAAU;AAAA,IACZ,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,KAAK,WAAW,kBAAkB,KAAK,QAAQ,IAAI,kBAAkB;AAAA,IACjF;AAEA,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,OAAO;AACL,WAAK,MAAM,QAAQ,KAAK,eAAe,KAAK,MAAM,OAAO,KAAK,MAAM,QAAQ;AAAA,IAC9E;AAEA,SAAK,eAAe,CAAC,GAAG,KAAK,MAAM,OAAO;AAAA,EAC5C;AAAA,EAEA,MAAM,WAAW,QAAqB,aAAqD;AACzF,UAAM,QAAQ,YAAY,MAAM;AAChC,UAAM,YAAY,IAAI,IAAI,GAAG,KAAK,MAAM,QAAQ,QAAQ,OAAO,MAAM,CAAC,YAAY;AAClF,UAAM,SAAS;AAAA,MACb,OAAO,KAAK,MAAM;AAAA,MAClB,WAAW,KAAK,MAAM;AAAA,MACtB,iBAAiB,KAAK,MAAM;AAAA,MAC5B,cAAc,KAAK,MAAM;AAAA,MACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,MACpD,kBAAkB,KAAK,MAAM;AAAA,MAC7B,UAAU,KAAK,MAAM;AAAA,MACrB,aAAa,KAAK,MAAM;AAAA,MACxB,UAAU,KAAK,MAAM;AAAA,MACrB,GAAI,KAAK,MAAM,QAAQ,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,IAAI,CAAC;AAAA,MACvE,GAAI,KAAK,MAAM,OAAO,SAAS,IAAI,EAAE,QAAQ,KAAK,MAAM,OAAO,IAAI,CAAC;AAAA,IACtE;AAEA,uBAAmB,WAAW,MAAM;AAEpC,QAAI,KAAK,MAAM,SAAS;AACtB,WAAK,QAAQ,KAAK,sEAAsE;AAAA,IAC1F;AAEA,UAAM,OAAO,MAAM,MAAM,WAAW;AAAA,MAClC,QAAQ;AAAA,MACR,SAAS;AAAA,QACP,eAAe,SAAS,KAAK,MAAM,MAAM;AAAA,QACzC,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA,MAAM,IAAI,KAAK,CAAC,IAAI,WAAW,UAAU,KAAK,CAAC,CAAC,GAAG,EAAE,MAAM,YAAY,CAAC;AAAA,MACxE,QAAQ,eAAe;AAAA,IACzB,CAAC;AAED,QAAI,CAAC,KAAK,IAAI;AACZ,YAAM,IAAI;AAAA,QACR,gDAAgD,KAAK,MAAM,KAAK,MAAM,KAAK,KAAK,CAAC;AAAA,MACnF;AAAA,IACF;AAEA,WAAO,sCAAsC,KAAK,MAAM,UAAU,MAAM,KAAK,KAAK,CAAC;AAAA,EACrF;AAAA,EAEA,cAAc,MAA2B;AACvC,UAAM,WACJ,KAAK,aAAa,SAAY,kBAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAC9E,UAAM,QACJ,KAAK,UAAU,UAAa,KAAK,aAAa,SAC1C,KAAK,eAAe,KAAK,SAAS,KAAK,MAAM,OAAO,QAAQ,IAC5D,KAAK,MAAM;AAEjB,UAAM,WAAW,EAAE,GAAG,KAAK;AAC3B,QAAI,SAAS,YAAY,QAAW;AAClC,WAAK,eAAe,CAAC,GAAG,SAAS,OAAO;AACxC,eAAS,UAAU,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,cAAc,GAAG,KAAK,gBAAgB,CAAC,CAAC;AAAA,IAClF;AAEA,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAES,uBAAuB,UAA0B;AACxD,QACE,SAAS,WAAW,KAAK,iBAAiB,UAC1C,SAAS,MAAM,CAAC,GAAG,MAAM,MAAM,KAAK,iBAAiB,CAAC,CAAC,GACvD;AACA;AAAA,IACF;AACA,SAAK,mBAAmB,CAAC,GAAG,QAAQ;AACpC,UAAM,SAAS,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,cAAc,GAAG,QAAQ,CAAC,CAAC;AAC/D,SAAK,MAAM,UAAU;AACrB,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,CAAC,QAAQ;AACX,aAAK,SAAS,OAAO,GAAG;AACxB;AAAA,MACF;AACA,UAAI,OAAO,WAAW;AAEpB,eAAO,kBAAkB;AAAA,MAC3B,OAAO;AACL,eAAO,cAAc,EAAE,SAAS,OAAO,CAAC;AAAA,MAC1C;AAAA,IACF;AAAA,EACF;AAAA,EAEA,eAAe,OAA2B,UAAmB;AAC3D,QACE,YACA,gBAAgB,QAAQ,MAAM,QAC9B;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,GAChB;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,8BAA8B,QAAQ;AAAA,MAChD;AACA,aAAO;AAAA,IACT;AAEA,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,SAA6D;AAElE,QAAI,KAAK,MAAM,kBAAkB,KAAK,MAAM,aAAa,QAAW;AAClE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,UAAM,SAAS,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AACtE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;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;AAAA;AAAA,EAGb,kBAAmC;AAAA,EACnC;AAAA,EACA,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,KAAK,YAAY,WAAW;AACvC,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,kBAAkB;AAChD,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,sBAAsB,QAAQ;AAAA,MACjD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AACd,QAAI;AAEJ,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,YAAM,YAAY,IAAI,IAAI,GAAG,KAAK,MAAM,OAAO,YAAY;AAC3D,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,kBAAkB,KAAK,MAAM;AAAA,QAC7B,SAAS,KAAK,MAAM;AAAA,QACpB,QAAQ,KAAK,MAAM;AAAA,QACnB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM;AAAA,QACxB,GAAI,KAAK,MAAM,OAAO,SAAS,IAAI,EAAE,QAAQ,KAAK,MAAM,OAAO,IAAI,CAAC;AAAA,MACtE;AACA,yBAAmB,WAAW,MAAM;AAEpC,WAAK,IAAI,UAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,CAAC,KAAK,UAAU,CAAC,KAAK,MAAM,QAAQ;AACtC,cAAI,WAAW,UAAU;AACvB,kBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,UACjF;AAEA,gBAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,eAAK,QAAQ;AAAA,YACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,UAC3F;AACA,gBAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,QAClE,OAAO;AACL,eAAK,QAAQ;AAAA,YACX,gDAAgD,CAAC,kBAAkB,KAAK,MAAM,MAAM,eAAe,KAAK,MAAM;AAAA,UAChH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA;AAAA,EAGA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,YAAY,QAAW;AAC9B,WAAK,kBAAkB;AAAA,IACzB;AACA,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,iBAAiB;AACf,QAAI,KAAK,oBAAoB,MAAM;AACjC,WAAK,cAAc,EAAE,SAAS,KAAK,gBAAgB,CAAC;AACpD,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;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;AACtD,uBAAK,eAAe;AAAA,gBACtB;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,KAAK,gBAAgB;AACnB,oBAAI,KAAK,WAAW;AAClB,uBAAK,YAAY;AACjB,6BAAW,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AACtD,uBAAK,eAAe;AAAA,gBACtB;AACA;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAEA,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,kCAAkC,GAAG,EAAE;AAC1D,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,eAAe,aAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IACrE,GAAG,KAAK,eAAe;AAEvB,UAAM,QAAQ,KAAK;AAAA,MACjB,KAAK,SAAS;AAAA,MACd,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,QAAQ,SAAS,CAAC;AAAA,IACxD,CAAC;AACD,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AAAA,EAEQ,sBAAsB,UAAkB;AAC9C,UAAM,aAA8B;AAAA,MAClC,MAAM,IAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AACF;AAEA,SAAS,mBAAmB,KAAU,QAAiC;AACrE,SAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,KAAK,KAAK,MAAM;AAC/C,QAAI,UAAU,UAAa,UAAU,KAAM;AAE3C,QAAI,MAAM,QAAQ,KAAK,GAAG;AACxB,YAAM,QAAQ,CAAC,SAAS,IAAI,aAAa,OAAO,KAAK,OAAO,IAAI,CAAC,CAAC;AAAA,IACpE,WAAW,OAAO,UAAU,UAAU;AACpC,aAAO,QAAQ,KAAK,EAAE,QAAQ,CAAC,CAAC,MAAM,WAAW,MAAM;AACrD,YAAI,aAAa,OAAO,KAAK,GAAG,IAAI,IAAI,WAAW,EAAE;AAAA,MACvD,CAAC;AAAA,IACH,OAAO;AACL,UAAI,aAAa,OAAO,KAAK,OAAO,KAAK,CAAC;AAAA,IAC5C;AAAA,EACF,CAAC;AACH;AAEA,SAAS,UAAU,OAA2B;AAC5C,QAAM,gBAAgB;AACtB,QAAM,WAAY,MAAM,aAAa,MAAM,WAAW,gBAAiB;AACvE,QAAM,aAAc,MAAM,WAAW,gBAAiB;AAEtD,QAAM,SAAS,OAAO,MAAM,EAAE;AAC9B,SAAO,MAAM,QAAQ,CAAC;AACtB,SAAO,cAAc,KAAK,MAAM,KAAK,YAAY,CAAC;AAClD,SAAO,MAAM,QAAQ,CAAC;AACtB,SAAO,MAAM,QAAQ,EAAE;AACvB,SAAO,cAAc,IAAI,EAAE;AAC3B,SAAO,cAAc,GAAG,EAAE;AAC1B,SAAO,cAAc,MAAM,UAAU,EAAE;AACvC,SAAO,cAAc,MAAM,YAAY,EAAE;AACzC,SAAO,cAAc,UAAU,EAAE;AACjC,SAAO,cAAc,YAAY,EAAE;AACnC,SAAO,cAAc,eAAe,EAAE;AACtC,SAAO,MAAM,QAAQ,EAAE;AACvB,SAAO,cAAc,MAAM,KAAK,YAAY,EAAE;AAE9C,QAAM,MAAM,OAAO,KAAK,MAAM,KAAK,QAAQ,MAAM,KAAK,YAAY,MAAM,KAAK,UAAU;AACvF,SAAO,OAAO,OAAO,CAAC,QAAQ,GAAG,CAAC;AACpC;AAEA,MAAM,gCAAgC,CACpC,UACA,MACA,kBAA0B,MACL;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,QAAQ;AApnB3B;AAqnBI,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAC1C,UAAM,mBACJ,aAAa,YAAU,SAAI,WAAW,MAAf,mBAAmB,OAAM,WAAW,IAAI,UAAU,KAAK;AAEhF,WAAO;AAAA,MACL,UAAU,kBAAkB,gBAAgB;AAAA,MAC5C,WAAW,UAAU,SAAS,UAAU,CAAC,EAAE,OAAO,IAAI,kBAAkB;AAAA,MACxE,SAAS,UAAU,SACf,UAAU,UAAU,SAAS,CAAC,EAAE,KAAK,IAAI,kBACzC;AAAA,MACJ,YAAY,IAAI,YAAY;AAAA,MAC5B,MAAM,IAAI,YAAY;AAAA,MACtB,OAAO,UAAU;AAAA,QAAI,CAAC,SACpB,kBAAkB;AAAA,UAChB,MAAM,KAAK,MAAM,KAAK;AAAA,UACtB,YAAY,KAAK,OAAO,KAAK,KAAK;AAAA,UAClC,UAAU,KAAK,KAAK,KAAK,KAAK;AAAA,UAC9B,YAAY,KAAK,YAAY,KAAK;AAAA,UAClC;AAAA,QACF,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF,CAAC;AACH;AAEA,SAAS,sCACP,UACA,MACiB;AACjB,QAAM,UAAU,KAAK,SAAS,EAAE,UAAU,EAAE,CAAC;AAC7C,QAAM,mBAAmB,QAAQ,mBAAmB,KAAK;AACzD,QAAM,eAAe,QAAQ,cAAc,EAAE,IAAI,CAAC,QAA+B;AAC/E,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAE1C,WAAO;AAAA,MACL,UAAU,kBAAkB,YAAY,gBAAgB;AAAA,MACxD,WAAW,UAAU,SAAS,UAAU,CAAC,EAAE,OAAO,IAAI;AAAA,MACtD,SAAS,UAAU,SAAS,UAAU,UAAU,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,MACrE,YAAY,IAAI,YAAY;AAAA,MAC5B,MAAM,IAAI,YAAY;AAAA,MACtB,OAAO,UAAU;AAAA,QAAI,CAAC,SACpB,kBAAkB;AAAA,UAChB,MAAM,KAAK,MAAM,KAAK;AAAA,UACtB,WAAW,KAAK,OAAO,KAAK;AAAA,UAC5B,SAAS,KAAK,KAAK,KAAK;AAAA,UACxB,YAAY,KAAK,YAAY,KAAK;AAAA,QACpC,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF,CAAC;AAED,SAAO;AAAA,IACL,MAAM,IAAI,gBAAgB;AAAA,IAC1B,WAAW,KAAK,UAAU,EAAE,YAAY;AAAA,IACxC,GAAI,aAAa,SACb,EAAE,aAAoE,IACtE,CAAC;AAAA,EACP;AACF;","names":["stt"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n type AudioBuffer,\n AudioByteStream,\n AudioEnergyFilter,\n Future,\n Task,\n createTimedString,\n getBaseLanguage,\n log,\n mergeFrames,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { STTLanguages, STTModels } from './models.js';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels | string;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n /**\n * Redact sensitive information from the transcription. Accepts a single value or list of\n * values. Supported values: \"pci\", \"numbers\", \"ssn\", \"true\" (redact all).\n * See https://developers.deepgram.com/docs/redaction for details.\n */\n redact: string | string[];\n utteranceEndMs?: number | null;\n replace?: Record<string, string> | null;\n search?: string[] | null;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n mipOptOut: boolean;\n baseUrl: string;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.DEEPGRAM_API_KEY,\n language: 'en-US',\n detectLanguage: false,\n interimResults: true,\n punctuate: true,\n model: 'nova-3',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n redact: [],\n utteranceEndMs: null,\n replace: null,\n search: null,\n dictation: false,\n diarize: false,\n numerals: false,\n mipOptOut: false,\n baseUrl: 'wss://api.deepgram.com',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n // session keyterm propagation)\n #streams = new Set<WeakRef<SpeechStream>>();\n #userKeyterm: string[];\n #sessionKeyterms: string[] = [];\n label = 'deepgram.STT';\n private abortController = new AbortController();\n\n get model(): string {\n return this.#opts.model;\n }\n\n get provider(): string {\n return 'Deepgram';\n }\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n alignedTranscript: 'word',\n keyterms: true,\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n language: opts.language ? normalizeLanguage(opts.language) : defaultSTTOptions.language,\n };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else {\n this.#opts.model = this.#validateModel(this.#opts.model, this.#opts.language);\n }\n\n this.#userKeyterm = [...this.#opts.keyterm];\n }\n\n async _recognize(buffer: AudioBuffer, abortSignal?: AbortSignal): Promise<stt.SpeechEvent> {\n const frame = mergeFrames(buffer);\n const listenURL = new URL(`${this.#opts.baseUrl.replace(/^ws/, 'http')}/v1/listen`);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n detect_language: this.#opts.detectLanguage,\n smart_format: this.#opts.smartFormat,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n profanity_filter: this.#opts.profanityFilter,\n numerals: this.#opts.numerals,\n mip_opt_out: this.#opts.mipOptOut,\n language: this.#opts.language,\n ...(this.#opts.keyterm.length > 0 ? { keyterm: this.#opts.keyterm } : {}),\n ...(this.#opts.redact.length > 0 ? { redact: this.#opts.redact } : {}),\n };\n\n appendSearchParams(listenURL, params);\n\n if (this.#opts.diarize) {\n this.#logger.warn('speaker diarization is not supported in non-streaming mode, ignoring');\n }\n\n const resp = await fetch(listenURL, {\n method: 'POST',\n headers: {\n Authorization: `Token ${this.#opts.apiKey}`,\n Accept: 'application/json',\n 'Content-Type': 'audio/wav',\n },\n body: new Blob([new Uint8Array(createWav(frame))], { type: 'audio/wav' }),\n signal: abortSignal ?? null,\n });\n\n if (!resp.ok) {\n throw new Error(\n `Deepgram STT REST request failed with status ${resp.status}: ${await resp.text()}`,\n );\n }\n\n return prerecordedTranscriptionToSpeechEvent(this.#opts.language, await resp.json());\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n const language =\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language;\n const model =\n opts.model !== undefined || opts.language !== undefined\n ? this.#validateModel(opts.model ?? this.#opts.model, language)\n : this.#opts.model;\n\n const nextOpts = { ...opts };\n if (nextOpts.keyterm !== undefined) {\n this.#userKeyterm = [...nextOpts.keyterm];\n nextOpts.keyterm = [...new Set([...this.#userKeyterm, ...this.#sessionKeyterms])];\n }\n\n this.#opts = {\n ...this.#opts,\n ...nextOpts,\n language,\n model,\n };\n }\n\n override _updateSessionKeyterms(keyterms: string[]): void {\n if (\n keyterms.length === this.#sessionKeyterms.length &&\n keyterms.every((t, i) => t === this.#sessionKeyterms[i])\n ) {\n return;\n }\n this.#sessionKeyterms = [...keyterms];\n const merged = [...new Set([...this.#userKeyterm, ...keyterms])];\n this.#opts.keyterm = merged;\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (!stream) {\n this.#streams.delete(ref);\n continue;\n }\n if (stream._speaking) {\n // defer the reconnect to the end of the utterance so we don't cut it off\n stream._pendingKeyterm = merged;\n } else {\n stream.updateOptions({ keyterm: merged });\n }\n }\n }\n\n #validateModel(model: STTModels | string, language?: string) {\n if (\n language &&\n getBaseLanguage(language) !== 'en' &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(model)\n ) {\n this.#logger.warn(\n `${model} does not support language ${language}, falling back to nova-2-general`,\n );\n return 'nova-2-general';\n }\n\n return model;\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n // Deepgram only supports language detection for prerecorded audio, not streaming.\n if (this.#opts.detectLanguage || this.#opts.language === undefined) {\n throw new Error(\n 'language detection is not supported in streaming mode, please disable it and specify a language',\n );\n }\n const stream = new SpeechStream(this, this.#opts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\n }\n\n async close() {\n this.abortController.abort();\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #audioEnergyFilter: AudioEnergyFilter;\n #logger = log();\n #speaking = false;\n #resetWS = new Future();\n #requestId = '';\n // keyterms set while the user is speaking; applied at END_OF_SPEECH (latest wins)\n /** @internal */\n _pendingKeyterm: string[] | null = null;\n #audioDurationCollector: PeriodicCollector<number>;\n label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(stt, opts.sampleRate, connOptions);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n let ws: WebSocket;\n\n while (!this.input.closed && !this.closed) {\n const streamURL = new URL(`${this.#opts.baseUrl}/v1/listen`);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n utterance_end_ms: this.#opts.utteranceEndMs,\n replace: this.#opts.replace,\n search: this.#opts.search,\n language: this.#opts.language,\n mip_opt_out: this.#opts.mipOptOut,\n ...(this.#opts.redact.length > 0 ? { redact: this.#opts.redact } : {}),\n };\n appendSearchParams(streamURL, params);\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (!this.closed && !this.input.closed) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n }\n\n const delay = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n } else {\n this.#logger.warn(\n `Deepgram disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n );\n }\n }\n }\n\n this.closed = true;\n }\n\n /** @internal */\n get _speaking(): boolean {\n return this.#speaking;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n if (opts.keyterm !== undefined) {\n this._pendingKeyterm = null;\n }\n this.#resetWS.resolve();\n }\n\n #onEndOfSpeech() {\n if (this._pendingKeyterm !== null) {\n this.updateOptions({ keyterm: this._pendingKeyterm });\n this._pendingKeyterm = null;\n }\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 let hasEnded = false;\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = stream.flush();\n hasEnded = true;\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 if (hasEnded) {\n this.#audioDurationCollector.flush();\n ws.send(JSON.stringify({ type: 'Finalize' }));\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 this.#onEndOfSpeech();\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n case 'UtteranceEnd': {\n if (this.#speaking) {\n this.#speaking = false;\n putMessage({ type: stt.SpeechEventType.END_OF_SPEECH });\n this.#onEndOfSpeech();\n }\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n\n if (this.closed || closing) {\n resolve();\n }\n } catch (err) {\n this.#logger.error(`STT: Error processing message: ${msg}`);\n reject(err);\n }\n });\n });\n\n await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n }, this.abortController);\n\n await Promise.race([\n this.#resetWS.await,\n Promise.all([sendTask(), listenTask.result, wsMonitor]),\n ]);\n closing = true;\n ws.close();\n clearInterval(keepalive);\n }\n\n private onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n}\n\nfunction appendSearchParams(url: URL, params: Record<string, unknown>) {\n Object.entries(params).forEach(([key, value]) => {\n if (value === undefined || value === null) return;\n\n if (Array.isArray(value)) {\n value.forEach((item) => url.searchParams.append(key, String(item)));\n } else if (typeof value === 'object') {\n Object.entries(value).forEach(([term, replacement]) => {\n url.searchParams.append(key, `${term}:${replacement}`);\n });\n } else {\n url.searchParams.append(key, String(value));\n }\n });\n}\n\nfunction createWav(frame: AudioFrame): Buffer {\n const bitsPerSample = 16;\n const byteRate = (frame.sampleRate * frame.channels * bitsPerSample) / 8;\n const blockAlign = (frame.channels * bitsPerSample) / 8;\n\n const header = Buffer.alloc(44);\n header.write('RIFF', 0);\n header.writeUInt32LE(36 + frame.data.byteLength, 4);\n header.write('WAVE', 8);\n header.write('fmt ', 12);\n header.writeUInt32LE(16, 16);\n header.writeUInt16LE(1, 20);\n header.writeUInt16LE(frame.channels, 22);\n header.writeUInt32LE(frame.sampleRate, 24);\n header.writeUInt32LE(byteRate, 28);\n header.writeUInt16LE(blockAlign, 32);\n header.writeUInt16LE(bitsPerSample, 34);\n header.write('data', 36);\n header.writeUInt32LE(frame.data.byteLength, 40);\n\n const pcm = Buffer.from(frame.data.buffer, frame.data.byteOffset, frame.data.byteLength);\n return Buffer.concat([header, pcm]);\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n startTimeOffset: number = 0,\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => {\n const wordsData: any[] = alt['words'] ?? [];\n const detectedLanguage =\n language === 'multi' ? alt['languages']?.[0] ?? language : alt['language'] ?? language;\n\n return {\n language: normalizeLanguage(detectedLanguage),\n startTime: wordsData.length ? wordsData[0]['start'] + startTimeOffset : startTimeOffset,\n endTime: wordsData.length\n ? wordsData[wordsData.length - 1]['end'] + startTimeOffset\n : startTimeOffset,\n confidence: alt['confidence'],\n text: alt['transcript'],\n words: wordsData.map((word) =>\n createTimedString({\n text: word['word'] ?? '',\n startTime: (word['start'] ?? 0) + startTimeOffset,\n endTime: (word['end'] ?? 0) + startTimeOffset,\n confidence: word['confidence'] ?? 0.0,\n startTimeOffset,\n }),\n ),\n };\n });\n};\n\nfunction prerecordedTranscriptionToSpeechEvent(\n language: STTLanguages | string | undefined,\n data: { [id: string]: any },\n): stt.SpeechEvent {\n const channel = data['results']['channels'][0];\n const detectedLanguage = channel['detected_language'] ?? '';\n const alternatives = channel['alternatives'].map((alt: { [id: string]: any }) => {\n const wordsData: any[] = alt['words'] ?? [];\n\n return {\n language: normalizeLanguage(language ?? detectedLanguage),\n startTime: wordsData.length ? wordsData[0]['start'] : 0,\n endTime: wordsData.length ? wordsData[wordsData.length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n words: wordsData.map((word) =>\n createTimedString({\n text: word['word'] ?? '',\n startTime: word['start'] ?? 0,\n endTime: word['end'] ?? 0,\n confidence: word['confidence'] ?? 0.0,\n }),\n ),\n };\n }) as stt.SpeechData[];\n\n return {\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n requestId: data['metadata']['request_id'],\n ...(alternatives.length\n ? { alternatives: alternatives as [stt.SpeechData, ...stt.SpeechData[]] }\n : {}),\n };\n}\n"],"mappings":"AAGA;AAAA,EAGE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,iBAAiB;AAC1B,SAAS,yBAAyB;AAmClC,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,QAAQ,CAAC;AAAA,EACT,gBAAgB;AAAA,EAChB,SAAS;AAAA,EACT,QAAQ;AAAA,EACR,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AAAA,EACV,WAAW;AAAA,EACX,SAAS;AACX;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU,IAAI;AAAA;AAAA,EAEd,WAAW,oBAAI,IAA2B;AAAA,EAC1C;AAAA,EACA,mBAA6B,CAAC;AAAA,EAC9B,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAE9C,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,MACzD,mBAAmB;AAAA,MACnB,UAAU;AAAA,IACZ,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,KAAK,WAAW,kBAAkB,KAAK,QAAQ,IAAI,kBAAkB;AAAA,IACjF;AAEA,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,OAAO;AACL,WAAK,MAAM,QAAQ,KAAK,eAAe,KAAK,MAAM,OAAO,KAAK,MAAM,QAAQ;AAAA,IAC9E;AAEA,SAAK,eAAe,CAAC,GAAG,KAAK,MAAM,OAAO;AAAA,EAC5C;AAAA,EAEA,MAAM,WAAW,QAAqB,aAAqD;AACzF,UAAM,QAAQ,YAAY,MAAM;AAChC,UAAM,YAAY,IAAI,IAAI,GAAG,KAAK,MAAM,QAAQ,QAAQ,OAAO,MAAM,CAAC,YAAY;AAClF,UAAM,SAAS;AAAA,MACb,OAAO,KAAK,MAAM;AAAA,MAClB,WAAW,KAAK,MAAM;AAAA,MACtB,iBAAiB,KAAK,MAAM;AAAA,MAC5B,cAAc,KAAK,MAAM;AAAA,MACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,MACpD,kBAAkB,KAAK,MAAM;AAAA,MAC7B,UAAU,KAAK,MAAM;AAAA,MACrB,aAAa,KAAK,MAAM;AAAA,MACxB,UAAU,KAAK,MAAM;AAAA,MACrB,GAAI,KAAK,MAAM,QAAQ,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,IAAI,CAAC;AAAA,MACvE,GAAI,KAAK,MAAM,OAAO,SAAS,IAAI,EAAE,QAAQ,KAAK,MAAM,OAAO,IAAI,CAAC;AAAA,IACtE;AAEA,uBAAmB,WAAW,MAAM;AAEpC,QAAI,KAAK,MAAM,SAAS;AACtB,WAAK,QAAQ,KAAK,sEAAsE;AAAA,IAC1F;AAEA,UAAM,OAAO,MAAM,MAAM,WAAW;AAAA,MAClC,QAAQ;AAAA,MACR,SAAS;AAAA,QACP,eAAe,SAAS,KAAK,MAAM,MAAM;AAAA,QACzC,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA,MAAM,IAAI,KAAK,CAAC,IAAI,WAAW,UAAU,KAAK,CAAC,CAAC,GAAG,EAAE,MAAM,YAAY,CAAC;AAAA,MACxE,QAAQ,eAAe;AAAA,IACzB,CAAC;AAED,QAAI,CAAC,KAAK,IAAI;AACZ,YAAM,IAAI;AAAA,QACR,gDAAgD,KAAK,MAAM,KAAK,MAAM,KAAK,KAAK,CAAC;AAAA,MACnF;AAAA,IACF;AAEA,WAAO,sCAAsC,KAAK,MAAM,UAAU,MAAM,KAAK,KAAK,CAAC;AAAA,EACrF;AAAA,EAEA,cAAc,MAA2B;AACvC,UAAM,WACJ,KAAK,aAAa,SAAY,kBAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAC9E,UAAM,QACJ,KAAK,UAAU,UAAa,KAAK,aAAa,SAC1C,KAAK,eAAe,KAAK,SAAS,KAAK,MAAM,OAAO,QAAQ,IAC5D,KAAK,MAAM;AAEjB,UAAM,WAAW,EAAE,GAAG,KAAK;AAC3B,QAAI,SAAS,YAAY,QAAW;AAClC,WAAK,eAAe,CAAC,GAAG,SAAS,OAAO;AACxC,eAAS,UAAU,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,cAAc,GAAG,KAAK,gBAAgB,CAAC,CAAC;AAAA,IAClF;AAEA,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAES,uBAAuB,UAA0B;AACxD,QACE,SAAS,WAAW,KAAK,iBAAiB,UAC1C,SAAS,MAAM,CAAC,GAAG,MAAM,MAAM,KAAK,iBAAiB,CAAC,CAAC,GACvD;AACA;AAAA,IACF;AACA,SAAK,mBAAmB,CAAC,GAAG,QAAQ;AACpC,UAAM,SAAS,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,cAAc,GAAG,QAAQ,CAAC,CAAC;AAC/D,SAAK,MAAM,UAAU;AACrB,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,CAAC,QAAQ;AACX,aAAK,SAAS,OAAO,GAAG;AACxB;AAAA,MACF;AACA,UAAI,OAAO,WAAW;AAEpB,eAAO,kBAAkB;AAAA,MAC3B,OAAO;AACL,eAAO,cAAc,EAAE,SAAS,OAAO,CAAC;AAAA,MAC1C;AAAA,IACF;AAAA,EACF;AAAA,EAEA,eAAe,OAA2B,UAAmB;AAC3D,QACE,YACA,gBAAgB,QAAQ,MAAM,QAC9B;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,GAChB;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,8BAA8B,QAAQ;AAAA,MAChD;AACA,aAAO;AAAA,IACT;AAEA,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,SAA6D;AAElE,QAAI,KAAK,MAAM,kBAAkB,KAAK,MAAM,aAAa,QAAW;AAClE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,UAAM,SAAS,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AACtE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;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;AAAA;AAAA,EAGb,kBAAmC;AAAA,EACnC;AAAA,EACA,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,KAAK,YAAY,WAAW;AACvC,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,kBAAkB;AAChD,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,sBAAsB,QAAQ;AAAA,MACjD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AACd,QAAI;AAEJ,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,YAAM,YAAY,IAAI,IAAI,GAAG,KAAK,MAAM,OAAO,YAAY;AAC3D,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,kBAAkB,KAAK,MAAM;AAAA,QAC7B,SAAS,KAAK,MAAM;AAAA,QACpB,QAAQ,KAAK,MAAM;AAAA,QACnB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM;AAAA,QACxB,GAAI,KAAK,MAAM,OAAO,SAAS,IAAI,EAAE,QAAQ,KAAK,MAAM,OAAO,IAAI,CAAC;AAAA,MACtE;AACA,yBAAmB,WAAW,MAAM;AAEpC,WAAK,IAAI,UAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,CAAC,KAAK,UAAU,CAAC,KAAK,MAAM,QAAQ;AACtC,cAAI,WAAW,UAAU;AACvB,kBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,UACjF;AAEA,gBAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,eAAK,QAAQ;AAAA,YACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,UAC3F;AACA,gBAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,QAClE,OAAO;AACL,eAAK,QAAQ;AAAA,YACX,gDAAgD,CAAC,kBAAkB,KAAK,MAAM,MAAM,eAAe,KAAK,MAAM;AAAA,UAChH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA;AAAA,EAGA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,YAAY,QAAW;AAC9B,WAAK,kBAAkB;AAAA,IACzB;AACA,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,iBAAiB;AACf,QAAI,KAAK,oBAAoB,MAAM;AACjC,WAAK,cAAc,EAAE,SAAS,KAAK,gBAAgB,CAAC;AACpD,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;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,WAAW;AACf,cAAI,SAAS,aAAa,gBAAgB;AACxC,qBAAS,OAAO,MAAM;AACtB,uBAAW;AAAA,UACb,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;AAEA,cAAI,UAAU;AACZ,iBAAK,wBAAwB,MAAM;AACnC,eAAG,KAAK,KAAK,UAAU,EAAE,MAAM,WAAW,CAAC,CAAC;AAAA,UAC9C;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;AACtD,uBAAK,eAAe;AAAA,gBACtB;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,KAAK,gBAAgB;AACnB,oBAAI,KAAK,WAAW;AAClB,uBAAK,YAAY;AACjB,6BAAW,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AACtD,uBAAK,eAAe;AAAA,gBACtB;AACA;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAEA,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,kCAAkC,GAAG,EAAE;AAC1D,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,eAAe,aAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IACrE,GAAG,KAAK,eAAe;AAEvB,UAAM,QAAQ,KAAK;AAAA,MACjB,KAAK,SAAS;AAAA,MACd,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,QAAQ,SAAS,CAAC;AAAA,IACxD,CAAC;AACD,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AAAA,EAEQ,sBAAsB,UAAkB;AAC9C,UAAM,aAA8B;AAAA,MAClC,MAAM,IAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AACF;AAEA,SAAS,mBAAmB,KAAU,QAAiC;AACrE,SAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,KAAK,KAAK,MAAM;AAC/C,QAAI,UAAU,UAAa,UAAU,KAAM;AAE3C,QAAI,MAAM,QAAQ,KAAK,GAAG;AACxB,YAAM,QAAQ,CAAC,SAAS,IAAI,aAAa,OAAO,KAAK,OAAO,IAAI,CAAC,CAAC;AAAA,IACpE,WAAW,OAAO,UAAU,UAAU;AACpC,aAAO,QAAQ,KAAK,EAAE,QAAQ,CAAC,CAAC,MAAM,WAAW,MAAM;AACrD,YAAI,aAAa,OAAO,KAAK,GAAG,IAAI,IAAI,WAAW,EAAE;AAAA,MACvD,CAAC;AAAA,IACH,OAAO;AACL,UAAI,aAAa,OAAO,KAAK,OAAO,KAAK,CAAC;AAAA,IAC5C;AAAA,EACF,CAAC;AACH;AAEA,SAAS,UAAU,OAA2B;AAC5C,QAAM,gBAAgB;AACtB,QAAM,WAAY,MAAM,aAAa,MAAM,WAAW,gBAAiB;AACvE,QAAM,aAAc,MAAM,WAAW,gBAAiB;AAEtD,QAAM,SAAS,OAAO,MAAM,EAAE;AAC9B,SAAO,MAAM,QAAQ,CAAC;AACtB,SAAO,cAAc,KAAK,MAAM,KAAK,YAAY,CAAC;AAClD,SAAO,MAAM,QAAQ,CAAC;AACtB,SAAO,MAAM,QAAQ,EAAE;AACvB,SAAO,cAAc,IAAI,EAAE;AAC3B,SAAO,cAAc,GAAG,EAAE;AAC1B,SAAO,cAAc,MAAM,UAAU,EAAE;AACvC,SAAO,cAAc,MAAM,YAAY,EAAE;AACzC,SAAO,cAAc,UAAU,EAAE;AACjC,SAAO,cAAc,YAAY,EAAE;AACnC,SAAO,cAAc,eAAe,EAAE;AACtC,SAAO,MAAM,QAAQ,EAAE;AACvB,SAAO,cAAc,MAAM,KAAK,YAAY,EAAE;AAE9C,QAAM,MAAM,OAAO,KAAK,MAAM,KAAK,QAAQ,MAAM,KAAK,YAAY,MAAM,KAAK,UAAU;AACvF,SAAO,OAAO,OAAO,CAAC,QAAQ,GAAG,CAAC;AACpC;AAEA,MAAM,gCAAgC,CACpC,UACA,MACA,kBAA0B,MACL;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,QAAQ;AA1nB3B;AA2nBI,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAC1C,UAAM,mBACJ,aAAa,YAAU,SAAI,WAAW,MAAf,mBAAmB,OAAM,WAAW,IAAI,UAAU,KAAK;AAEhF,WAAO;AAAA,MACL,UAAU,kBAAkB,gBAAgB;AAAA,MAC5C,WAAW,UAAU,SAAS,UAAU,CAAC,EAAE,OAAO,IAAI,kBAAkB;AAAA,MACxE,SAAS,UAAU,SACf,UAAU,UAAU,SAAS,CAAC,EAAE,KAAK,IAAI,kBACzC;AAAA,MACJ,YAAY,IAAI,YAAY;AAAA,MAC5B,MAAM,IAAI,YAAY;AAAA,MACtB,OAAO,UAAU;AAAA,QAAI,CAAC,SACpB,kBAAkB;AAAA,UAChB,MAAM,KAAK,MAAM,KAAK;AAAA,UACtB,YAAY,KAAK,OAAO,KAAK,KAAK;AAAA,UAClC,UAAU,KAAK,KAAK,KAAK,KAAK;AAAA,UAC9B,YAAY,KAAK,YAAY,KAAK;AAAA,UAClC;AAAA,QACF,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF,CAAC;AACH;AAEA,SAAS,sCACP,UACA,MACiB;AACjB,QAAM,UAAU,KAAK,SAAS,EAAE,UAAU,EAAE,CAAC;AAC7C,QAAM,mBAAmB,QAAQ,mBAAmB,KAAK;AACzD,QAAM,eAAe,QAAQ,cAAc,EAAE,IAAI,CAAC,QAA+B;AAC/E,UAAM,YAAmB,IAAI,OAAO,KAAK,CAAC;AAE1C,WAAO;AAAA,MACL,UAAU,kBAAkB,YAAY,gBAAgB;AAAA,MACxD,WAAW,UAAU,SAAS,UAAU,CAAC,EAAE,OAAO,IAAI;AAAA,MACtD,SAAS,UAAU,SAAS,UAAU,UAAU,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,MACrE,YAAY,IAAI,YAAY;AAAA,MAC5B,MAAM,IAAI,YAAY;AAAA,MACtB,OAAO,UAAU;AAAA,QAAI,CAAC,SACpB,kBAAkB;AAAA,UAChB,MAAM,KAAK,MAAM,KAAK;AAAA,UACtB,WAAW,KAAK,OAAO,KAAK;AAAA,UAC5B,SAAS,KAAK,KAAK,KAAK;AAAA,UACxB,YAAY,KAAK,YAAY,KAAK;AAAA,QACpC,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF,CAAC;AAED,SAAO;AAAA,IACL,MAAM,IAAI,gBAAgB;AAAA,IAC1B,WAAW,KAAK,UAAU,EAAE,YAAY;AAAA,IACxC,GAAI,aAAa,SACb,EAAE,aAAoE,IACtE,CAAC;AAAA,EACP;AACF;","names":["stt"]}
|
package/dist/stt.test.cjs
CHANGED
|
@@ -1,9 +1,35 @@
|
|
|
1
1
|
"use strict";
|
|
2
2
|
var import_agents_plugin_silero = require("@livekit/agents-plugin-silero");
|
|
3
3
|
var import_agents_plugins_test = require("@livekit/agents-plugins-test");
|
|
4
|
+
var import_rtc_node = require("@livekit/rtc-node");
|
|
5
|
+
var import_node_events = require("node:events");
|
|
4
6
|
var import_vitest = require("vitest");
|
|
7
|
+
var import_ws = require("ws");
|
|
5
8
|
var import_stt = require("./stt.cjs");
|
|
6
9
|
const hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);
|
|
10
|
+
function makeFrame(samplesPerChannel, sampleRate = 16e3) {
|
|
11
|
+
const data = new Int16Array(samplesPerChannel);
|
|
12
|
+
data.fill(1);
|
|
13
|
+
return new import_rtc_node.AudioFrame(data, sampleRate, 1, samplesPerChannel);
|
|
14
|
+
}
|
|
15
|
+
async function startWebSocketServer() {
|
|
16
|
+
const wss = new import_ws.WebSocketServer({ host: "127.0.0.1", port: 0 });
|
|
17
|
+
await (0, import_node_events.once)(wss, "listening");
|
|
18
|
+
const address = wss.address();
|
|
19
|
+
return { wss, baseUrl: `ws://127.0.0.1:${address.port}` };
|
|
20
|
+
}
|
|
21
|
+
async function closeWebSocketServer(wss) {
|
|
22
|
+
for (const client of wss.clients) client.close();
|
|
23
|
+
await new Promise((resolve) => wss.close(() => resolve()));
|
|
24
|
+
}
|
|
25
|
+
async function waitUntil(predicate, timeoutMs = 1e3) {
|
|
26
|
+
const startedAt = Date.now();
|
|
27
|
+
while (Date.now() - startedAt < timeoutMs) {
|
|
28
|
+
if (predicate()) return;
|
|
29
|
+
await new Promise((resolve) => setTimeout(resolve, 10));
|
|
30
|
+
}
|
|
31
|
+
throw new Error("timed out waiting for condition");
|
|
32
|
+
}
|
|
7
33
|
(0, import_vitest.describe)("Deepgram streaming language detection", () => {
|
|
8
34
|
(0, import_vitest.it)("throws when starting a stream with detectLanguage enabled", () => {
|
|
9
35
|
const stt2 = new import_stt.STT({ apiKey: "test", detectLanguage: true });
|
|
@@ -15,6 +41,45 @@ const hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);
|
|
|
15
41
|
stream.close();
|
|
16
42
|
});
|
|
17
43
|
});
|
|
44
|
+
(0, import_vitest.describe)("Deepgram streaming flush", () => {
|
|
45
|
+
(0, import_vitest.it)("finalizes the turn when no audio is left to send", async () => {
|
|
46
|
+
const { wss, baseUrl } = await startWebSocketServer();
|
|
47
|
+
const wire = [];
|
|
48
|
+
wss.on("connection", (ws) => {
|
|
49
|
+
ws.on("message", (data, isBinary) => {
|
|
50
|
+
wire.push(isBinary ? "audio" : JSON.parse(data.toString()).type);
|
|
51
|
+
});
|
|
52
|
+
});
|
|
53
|
+
const stream = new import_stt.STT({ apiKey: "test-key", baseUrl }).stream();
|
|
54
|
+
try {
|
|
55
|
+
stream.pushFrame(makeFrame(1600));
|
|
56
|
+
await waitUntil(() => wire.join(",") === "audio");
|
|
57
|
+
stream.flush();
|
|
58
|
+
await waitUntil(() => wire.join(",") === "audio,Finalize");
|
|
59
|
+
} finally {
|
|
60
|
+
stream.close();
|
|
61
|
+
await closeWebSocketServer(wss);
|
|
62
|
+
}
|
|
63
|
+
});
|
|
64
|
+
(0, import_vitest.it)("finalizes after the buffered audio", async () => {
|
|
65
|
+
const { wss, baseUrl } = await startWebSocketServer();
|
|
66
|
+
const wire = [];
|
|
67
|
+
wss.on("connection", (ws) => {
|
|
68
|
+
ws.on("message", (data, isBinary) => {
|
|
69
|
+
wire.push(isBinary ? "audio" : JSON.parse(data.toString()).type);
|
|
70
|
+
});
|
|
71
|
+
});
|
|
72
|
+
const stream = new import_stt.STT({ apiKey: "test-key", baseUrl }).stream();
|
|
73
|
+
try {
|
|
74
|
+
stream.pushFrame(makeFrame(480));
|
|
75
|
+
stream.flush();
|
|
76
|
+
await waitUntil(() => wire.join(",") === "audio,Finalize");
|
|
77
|
+
} finally {
|
|
78
|
+
stream.close();
|
|
79
|
+
await closeWebSocketServer(wss);
|
|
80
|
+
}
|
|
81
|
+
});
|
|
82
|
+
});
|
|
18
83
|
if (hasDeepgramApiKey) {
|
|
19
84
|
(0, import_vitest.describe)("Deepgram", async () => {
|
|
20
85
|
await (0, import_agents_plugins_test.stt)(new import_stt.STT(), await import_agents_plugin_silero.VAD.load(), { nonStreaming: false });
|
package/dist/stt.test.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { describe, expect, it } from 'vitest';\nimport { STT } from './stt.js';\n\nconst hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);\n\ndescribe('Deepgram streaming language detection', () => {\n // Deepgram only supports language detection for prerecorded audio, so a\n // streaming session must reject it rather than silently default to English.\n // Mirrors livekit-plugins-deepgram (Python).\n it('throws when starting a stream with detectLanguage enabled', () => {\n const stt = new STT({ apiKey: 'test', detectLanguage: true });\n expect(() => stt.stream()).toThrow('language detection is not supported in streaming mode');\n });\n\n it('allows streaming with an explicit language', () => {\n const stt = new STT({ apiKey: 'test', language: 'en-US' });\n const stream = stt.stream();\n // Close immediately so the connection loop never starts (no network in unit tests).\n stream.close();\n });\n});\n\nif (hasDeepgramApiKey) {\n describe('Deepgram', async () => {\n await stt(new STT(), await VAD.load(), { nonStreaming: false });\n });\n} else {\n describe('Deepgram', () => {\n it.skip('requires DEEPGRAM_API_KEY', () => {});\n });\n}\n"],"mappings":";AAGA,kCAAoB;AACpB,iCAAoB;AACpB,oBAAqC;AACrC,iBAAoB;AAEpB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,gBAAgB;AAAA,
|
|
1
|
+
{"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { AudioFrame } from '@livekit/rtc-node';\nimport { once } from 'node:events';\nimport type { AddressInfo } from 'node:net';\nimport { describe, expect, it } from 'vitest';\nimport { WebSocketServer } from 'ws';\nimport { STT } from './stt.js';\n\nconst hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);\n\nfunction makeFrame(samplesPerChannel: number, sampleRate = 16000): AudioFrame {\n const data = new Int16Array(samplesPerChannel);\n data.fill(1);\n return new AudioFrame(data, sampleRate, 1, samplesPerChannel);\n}\n\nasync function startWebSocketServer() {\n const wss = new WebSocketServer({ host: '127.0.0.1', port: 0 });\n await once(wss, 'listening');\n const address = wss.address() as AddressInfo;\n return { wss, baseUrl: `ws://127.0.0.1:${address.port}` };\n}\n\nasync function closeWebSocketServer(wss: WebSocketServer): Promise<void> {\n for (const client of wss.clients) client.close();\n await new Promise<void>((resolve) => wss.close(() => resolve()));\n}\n\nasync function waitUntil(predicate: () => boolean, timeoutMs = 1000): Promise<void> {\n const startedAt = Date.now();\n while (Date.now() - startedAt < timeoutMs) {\n if (predicate()) return;\n await new Promise((resolve) => setTimeout(resolve, 10));\n }\n throw new Error('timed out waiting for condition');\n}\n\ndescribe('Deepgram streaming language detection', () => {\n // Deepgram only supports language detection for prerecorded audio, so a\n // streaming session must reject it rather than silently default to English.\n // Mirrors livekit-plugins-deepgram (Python).\n it('throws when starting a stream with detectLanguage enabled', () => {\n const stt = new STT({ apiKey: 'test', detectLanguage: true });\n expect(() => stt.stream()).toThrow('language detection is not supported in streaming mode');\n });\n\n it('allows streaming with an explicit language', () => {\n const stt = new STT({ apiKey: 'test', language: 'en-US' });\n const stream = stt.stream();\n // Close immediately so the connection loop never starts (no network in unit tests).\n stream.close();\n });\n});\n\ndescribe('Deepgram streaming flush', () => {\n it('finalizes the turn when no audio is left to send', async () => {\n const { wss, baseUrl } = await startWebSocketServer();\n const wire: string[] = [];\n wss.on('connection', (ws) => {\n ws.on('message', (data, isBinary) => {\n wire.push(isBinary ? 'audio' : (JSON.parse(data.toString()) as { type: string }).type);\n });\n });\n\n const stream = new STT({ apiKey: 'test-key', baseUrl }).stream();\n try {\n // 100ms is exactly one JS repack chunk, so flush() returns no frames.\n stream.pushFrame(makeFrame(1600));\n await waitUntil(() => wire.join(',') === 'audio');\n\n stream.flush();\n await waitUntil(() => wire.join(',') === 'audio,Finalize');\n } finally {\n stream.close();\n await closeWebSocketServer(wss);\n }\n });\n\n it('finalizes after the buffered audio', async () => {\n const { wss, baseUrl } = await startWebSocketServer();\n const wire: string[] = [];\n wss.on('connection', (ws) => {\n ws.on('message', (data, isBinary) => {\n wire.push(isBinary ? 'audio' : (JSON.parse(data.toString()) as { type: string }).type);\n });\n });\n\n const stream = new STT({ apiKey: 'test-key', baseUrl }).stream();\n try {\n stream.pushFrame(makeFrame(480));\n stream.flush();\n await waitUntil(() => wire.join(',') === 'audio,Finalize');\n } finally {\n stream.close();\n await closeWebSocketServer(wss);\n }\n });\n});\n\nif (hasDeepgramApiKey) {\n describe('Deepgram', async () => {\n await stt(new STT(), await VAD.load(), { nonStreaming: false });\n });\n} else {\n describe('Deepgram', () => {\n it.skip('requires DEEPGRAM_API_KEY', () => {});\n });\n}\n"],"mappings":";AAGA,kCAAoB;AACpB,iCAAoB;AACpB,sBAA2B;AAC3B,yBAAqB;AAErB,oBAAqC;AACrC,gBAAgC;AAChC,iBAAoB;AAEpB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,gBAAgB;AAE9D,SAAS,UAAU,mBAA2B,aAAa,MAAmB;AAC5E,QAAM,OAAO,IAAI,WAAW,iBAAiB;AAC7C,OAAK,KAAK,CAAC;AACX,SAAO,IAAI,2BAAW,MAAM,YAAY,GAAG,iBAAiB;AAC9D;AAEA,eAAe,uBAAuB;AACpC,QAAM,MAAM,IAAI,0BAAgB,EAAE,MAAM,aAAa,MAAM,EAAE,CAAC;AAC9D,YAAM,yBAAK,KAAK,WAAW;AAC3B,QAAM,UAAU,IAAI,QAAQ;AAC5B,SAAO,EAAE,KAAK,SAAS,kBAAkB,QAAQ,IAAI,GAAG;AAC1D;AAEA,eAAe,qBAAqB,KAAqC;AACvE,aAAW,UAAU,IAAI,QAAS,QAAO,MAAM;AAC/C,QAAM,IAAI,QAAc,CAAC,YAAY,IAAI,MAAM,MAAM,QAAQ,CAAC,CAAC;AACjE;AAEA,eAAe,UAAU,WAA0B,YAAY,KAAqB;AAClF,QAAM,YAAY,KAAK,IAAI;AAC3B,SAAO,KAAK,IAAI,IAAI,YAAY,WAAW;AACzC,QAAI,UAAU,EAAG;AACjB,UAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,EAAE,CAAC;AAAA,EACxD;AACA,QAAM,IAAI,MAAM,iCAAiC;AACnD;AAAA,IAEA,wBAAS,yCAAyC,MAAM;AAItD,wBAAG,6DAA6D,MAAM;AACpE,UAAMA,OAAM,IAAI,eAAI,EAAE,QAAQ,QAAQ,gBAAgB,KAAK,CAAC;AAC5D,8BAAO,MAAMA,KAAI,OAAO,CAAC,EAAE,QAAQ,uDAAuD;AAAA,EAC5F,CAAC;AAED,wBAAG,8CAA8C,MAAM;AACrD,UAAMA,OAAM,IAAI,eAAI,EAAE,QAAQ,QAAQ,UAAU,QAAQ,CAAC;AACzD,UAAM,SAASA,KAAI,OAAO;AAE1B,WAAO,MAAM;AAAA,EACf,CAAC;AACH,CAAC;AAAA,IAED,wBAAS,4BAA4B,MAAM;AACzC,wBAAG,oDAAoD,YAAY;AACjE,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,UAAM,OAAiB,CAAC;AACxB,QAAI,GAAG,cAAc,CAAC,OAAO;AAC3B,SAAG,GAAG,WAAW,CAAC,MAAM,aAAa;AACnC,aAAK,KAAK,WAAW,UAAW,KAAK,MAAM,KAAK,SAAS,CAAC,EAAuB,IAAI;AAAA,MACvF,CAAC;AAAA,IACH,CAAC;AAED,UAAM,SAAS,IAAI,eAAI,EAAE,QAAQ,YAAY,QAAQ,CAAC,EAAE,OAAO;AAC/D,QAAI;AAEF,aAAO,UAAU,UAAU,IAAI,CAAC;AAChC,YAAM,UAAU,MAAM,KAAK,KAAK,GAAG,MAAM,OAAO;AAEhD,aAAO,MAAM;AACb,YAAM,UAAU,MAAM,KAAK,KAAK,GAAG,MAAM,gBAAgB;AAAA,IAC3D,UAAE;AACA,aAAO,MAAM;AACb,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AAED,wBAAG,sCAAsC,YAAY;AACnD,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,UAAM,OAAiB,CAAC;AACxB,QAAI,GAAG,cAAc,CAAC,OAAO;AAC3B,SAAG,GAAG,WAAW,CAAC,MAAM,aAAa;AACnC,aAAK,KAAK,WAAW,UAAW,KAAK,MAAM,KAAK,SAAS,CAAC,EAAuB,IAAI;AAAA,MACvF,CAAC;AAAA,IACH,CAAC;AAED,UAAM,SAAS,IAAI,eAAI,EAAE,QAAQ,YAAY,QAAQ,CAAC,EAAE,OAAO;AAC/D,QAAI;AACF,aAAO,UAAU,UAAU,GAAG,CAAC;AAC/B,aAAO,MAAM;AACb,YAAM,UAAU,MAAM,KAAK,KAAK,GAAG,MAAM,gBAAgB;AAAA,IAC3D,UAAE;AACA,aAAO,MAAM;AACb,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AACH,CAAC;AAED,IAAI,mBAAmB;AACrB,8BAAS,YAAY,YAAY;AAC/B,cAAM,gCAAI,IAAI,eAAI,GAAG,MAAM,gCAAI,KAAK,GAAG,EAAE,cAAc,MAAM,CAAC;AAAA,EAChE,CAAC;AACH,OAAO;AACL,8BAAS,YAAY,MAAM;AACzB,qBAAG,KAAK,6BAA6B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC/C,CAAC;AACH;","names":["stt"]}
|
package/dist/stt.test.js
CHANGED
|
@@ -1,8 +1,34 @@
|
|
|
1
1
|
import { VAD } from "@livekit/agents-plugin-silero";
|
|
2
2
|
import { stt } from "@livekit/agents-plugins-test";
|
|
3
|
+
import { AudioFrame } from "@livekit/rtc-node";
|
|
4
|
+
import { once } from "node:events";
|
|
3
5
|
import { describe, expect, it } from "vitest";
|
|
6
|
+
import { WebSocketServer } from "ws";
|
|
4
7
|
import { STT } from "./stt.js";
|
|
5
8
|
const hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);
|
|
9
|
+
function makeFrame(samplesPerChannel, sampleRate = 16e3) {
|
|
10
|
+
const data = new Int16Array(samplesPerChannel);
|
|
11
|
+
data.fill(1);
|
|
12
|
+
return new AudioFrame(data, sampleRate, 1, samplesPerChannel);
|
|
13
|
+
}
|
|
14
|
+
async function startWebSocketServer() {
|
|
15
|
+
const wss = new WebSocketServer({ host: "127.0.0.1", port: 0 });
|
|
16
|
+
await once(wss, "listening");
|
|
17
|
+
const address = wss.address();
|
|
18
|
+
return { wss, baseUrl: `ws://127.0.0.1:${address.port}` };
|
|
19
|
+
}
|
|
20
|
+
async function closeWebSocketServer(wss) {
|
|
21
|
+
for (const client of wss.clients) client.close();
|
|
22
|
+
await new Promise((resolve) => wss.close(() => resolve()));
|
|
23
|
+
}
|
|
24
|
+
async function waitUntil(predicate, timeoutMs = 1e3) {
|
|
25
|
+
const startedAt = Date.now();
|
|
26
|
+
while (Date.now() - startedAt < timeoutMs) {
|
|
27
|
+
if (predicate()) return;
|
|
28
|
+
await new Promise((resolve) => setTimeout(resolve, 10));
|
|
29
|
+
}
|
|
30
|
+
throw new Error("timed out waiting for condition");
|
|
31
|
+
}
|
|
6
32
|
describe("Deepgram streaming language detection", () => {
|
|
7
33
|
it("throws when starting a stream with detectLanguage enabled", () => {
|
|
8
34
|
const stt2 = new STT({ apiKey: "test", detectLanguage: true });
|
|
@@ -14,6 +40,45 @@ describe("Deepgram streaming language detection", () => {
|
|
|
14
40
|
stream.close();
|
|
15
41
|
});
|
|
16
42
|
});
|
|
43
|
+
describe("Deepgram streaming flush", () => {
|
|
44
|
+
it("finalizes the turn when no audio is left to send", async () => {
|
|
45
|
+
const { wss, baseUrl } = await startWebSocketServer();
|
|
46
|
+
const wire = [];
|
|
47
|
+
wss.on("connection", (ws) => {
|
|
48
|
+
ws.on("message", (data, isBinary) => {
|
|
49
|
+
wire.push(isBinary ? "audio" : JSON.parse(data.toString()).type);
|
|
50
|
+
});
|
|
51
|
+
});
|
|
52
|
+
const stream = new STT({ apiKey: "test-key", baseUrl }).stream();
|
|
53
|
+
try {
|
|
54
|
+
stream.pushFrame(makeFrame(1600));
|
|
55
|
+
await waitUntil(() => wire.join(",") === "audio");
|
|
56
|
+
stream.flush();
|
|
57
|
+
await waitUntil(() => wire.join(",") === "audio,Finalize");
|
|
58
|
+
} finally {
|
|
59
|
+
stream.close();
|
|
60
|
+
await closeWebSocketServer(wss);
|
|
61
|
+
}
|
|
62
|
+
});
|
|
63
|
+
it("finalizes after the buffered audio", async () => {
|
|
64
|
+
const { wss, baseUrl } = await startWebSocketServer();
|
|
65
|
+
const wire = [];
|
|
66
|
+
wss.on("connection", (ws) => {
|
|
67
|
+
ws.on("message", (data, isBinary) => {
|
|
68
|
+
wire.push(isBinary ? "audio" : JSON.parse(data.toString()).type);
|
|
69
|
+
});
|
|
70
|
+
});
|
|
71
|
+
const stream = new STT({ apiKey: "test-key", baseUrl }).stream();
|
|
72
|
+
try {
|
|
73
|
+
stream.pushFrame(makeFrame(480));
|
|
74
|
+
stream.flush();
|
|
75
|
+
await waitUntil(() => wire.join(",") === "audio,Finalize");
|
|
76
|
+
} finally {
|
|
77
|
+
stream.close();
|
|
78
|
+
await closeWebSocketServer(wss);
|
|
79
|
+
}
|
|
80
|
+
});
|
|
81
|
+
});
|
|
17
82
|
if (hasDeepgramApiKey) {
|
|
18
83
|
describe("Deepgram", async () => {
|
|
19
84
|
await stt(new STT(), await VAD.load(), { nonStreaming: false });
|
package/dist/stt.test.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { describe, expect, it } from 'vitest';\nimport { STT } from './stt.js';\n\nconst hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);\n\ndescribe('Deepgram streaming language detection', () => {\n // Deepgram only supports language detection for prerecorded audio, so a\n // streaming session must reject it rather than silently default to English.\n // Mirrors livekit-plugins-deepgram (Python).\n it('throws when starting a stream with detectLanguage enabled', () => {\n const stt = new STT({ apiKey: 'test', detectLanguage: true });\n expect(() => stt.stream()).toThrow('language detection is not supported in streaming mode');\n });\n\n it('allows streaming with an explicit language', () => {\n const stt = new STT({ apiKey: 'test', language: 'en-US' });\n const stream = stt.stream();\n // Close immediately so the connection loop never starts (no network in unit tests).\n stream.close();\n });\n});\n\nif (hasDeepgramApiKey) {\n describe('Deepgram', async () => {\n await stt(new STT(), await VAD.load(), { nonStreaming: false });\n });\n} else {\n describe('Deepgram', () => {\n it.skip('requires DEEPGRAM_API_KEY', () => {});\n });\n}\n"],"mappings":"AAGA,SAAS,WAAW;AACpB,SAAS,WAAW;AACpB,SAAS,UAAU,QAAQ,UAAU;AACrC,SAAS,WAAW;AAEpB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,gBAAgB;AAE9D,SAAS,yCAAyC,MAAM;AAItD,KAAG,6DAA6D,MAAM;AACpE,UAAMA,OAAM,IAAI,IAAI,EAAE,QAAQ,QAAQ,gBAAgB,KAAK,CAAC;AAC5D,WAAO,MAAMA,KAAI,OAAO,CAAC,EAAE,QAAQ,uDAAuD;AAAA,EAC5F,CAAC;AAED,KAAG,8CAA8C,MAAM;AACrD,UAAMA,OAAM,IAAI,IAAI,EAAE,QAAQ,QAAQ,UAAU,QAAQ,CAAC;AACzD,UAAM,SAASA,KAAI,OAAO;AAE1B,WAAO,MAAM;AAAA,EACf,CAAC;AACH,CAAC;AAED,IAAI,mBAAmB;AACrB,WAAS,YAAY,YAAY;AAC/B,UAAM,IAAI,IAAI,IAAI,GAAG,MAAM,IAAI,KAAK,GAAG,EAAE,cAAc,MAAM,CAAC;AAAA,EAChE,CAAC;AACH,OAAO;AACL,WAAS,YAAY,MAAM;AACzB,OAAG,KAAK,6BAA6B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC/C,CAAC;AACH;","names":["stt"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { AudioFrame } from '@livekit/rtc-node';\nimport { once } from 'node:events';\nimport type { AddressInfo } from 'node:net';\nimport { describe, expect, it } from 'vitest';\nimport { WebSocketServer } from 'ws';\nimport { STT } from './stt.js';\n\nconst hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);\n\nfunction makeFrame(samplesPerChannel: number, sampleRate = 16000): AudioFrame {\n const data = new Int16Array(samplesPerChannel);\n data.fill(1);\n return new AudioFrame(data, sampleRate, 1, samplesPerChannel);\n}\n\nasync function startWebSocketServer() {\n const wss = new WebSocketServer({ host: '127.0.0.1', port: 0 });\n await once(wss, 'listening');\n const address = wss.address() as AddressInfo;\n return { wss, baseUrl: `ws://127.0.0.1:${address.port}` };\n}\n\nasync function closeWebSocketServer(wss: WebSocketServer): Promise<void> {\n for (const client of wss.clients) client.close();\n await new Promise<void>((resolve) => wss.close(() => resolve()));\n}\n\nasync function waitUntil(predicate: () => boolean, timeoutMs = 1000): Promise<void> {\n const startedAt = Date.now();\n while (Date.now() - startedAt < timeoutMs) {\n if (predicate()) return;\n await new Promise((resolve) => setTimeout(resolve, 10));\n }\n throw new Error('timed out waiting for condition');\n}\n\ndescribe('Deepgram streaming language detection', () => {\n // Deepgram only supports language detection for prerecorded audio, so a\n // streaming session must reject it rather than silently default to English.\n // Mirrors livekit-plugins-deepgram (Python).\n it('throws when starting a stream with detectLanguage enabled', () => {\n const stt = new STT({ apiKey: 'test', detectLanguage: true });\n expect(() => stt.stream()).toThrow('language detection is not supported in streaming mode');\n });\n\n it('allows streaming with an explicit language', () => {\n const stt = new STT({ apiKey: 'test', language: 'en-US' });\n const stream = stt.stream();\n // Close immediately so the connection loop never starts (no network in unit tests).\n stream.close();\n });\n});\n\ndescribe('Deepgram streaming flush', () => {\n it('finalizes the turn when no audio is left to send', async () => {\n const { wss, baseUrl } = await startWebSocketServer();\n const wire: string[] = [];\n wss.on('connection', (ws) => {\n ws.on('message', (data, isBinary) => {\n wire.push(isBinary ? 'audio' : (JSON.parse(data.toString()) as { type: string }).type);\n });\n });\n\n const stream = new STT({ apiKey: 'test-key', baseUrl }).stream();\n try {\n // 100ms is exactly one JS repack chunk, so flush() returns no frames.\n stream.pushFrame(makeFrame(1600));\n await waitUntil(() => wire.join(',') === 'audio');\n\n stream.flush();\n await waitUntil(() => wire.join(',') === 'audio,Finalize');\n } finally {\n stream.close();\n await closeWebSocketServer(wss);\n }\n });\n\n it('finalizes after the buffered audio', async () => {\n const { wss, baseUrl } = await startWebSocketServer();\n const wire: string[] = [];\n wss.on('connection', (ws) => {\n ws.on('message', (data, isBinary) => {\n wire.push(isBinary ? 'audio' : (JSON.parse(data.toString()) as { type: string }).type);\n });\n });\n\n const stream = new STT({ apiKey: 'test-key', baseUrl }).stream();\n try {\n stream.pushFrame(makeFrame(480));\n stream.flush();\n await waitUntil(() => wire.join(',') === 'audio,Finalize');\n } finally {\n stream.close();\n await closeWebSocketServer(wss);\n }\n });\n});\n\nif (hasDeepgramApiKey) {\n describe('Deepgram', async () => {\n await stt(new STT(), await VAD.load(), { nonStreaming: false });\n });\n} else {\n describe('Deepgram', () => {\n it.skip('requires DEEPGRAM_API_KEY', () => {});\n });\n}\n"],"mappings":"AAGA,SAAS,WAAW;AACpB,SAAS,WAAW;AACpB,SAAS,kBAAkB;AAC3B,SAAS,YAAY;AAErB,SAAS,UAAU,QAAQ,UAAU;AACrC,SAAS,uBAAuB;AAChC,SAAS,WAAW;AAEpB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,gBAAgB;AAE9D,SAAS,UAAU,mBAA2B,aAAa,MAAmB;AAC5E,QAAM,OAAO,IAAI,WAAW,iBAAiB;AAC7C,OAAK,KAAK,CAAC;AACX,SAAO,IAAI,WAAW,MAAM,YAAY,GAAG,iBAAiB;AAC9D;AAEA,eAAe,uBAAuB;AACpC,QAAM,MAAM,IAAI,gBAAgB,EAAE,MAAM,aAAa,MAAM,EAAE,CAAC;AAC9D,QAAM,KAAK,KAAK,WAAW;AAC3B,QAAM,UAAU,IAAI,QAAQ;AAC5B,SAAO,EAAE,KAAK,SAAS,kBAAkB,QAAQ,IAAI,GAAG;AAC1D;AAEA,eAAe,qBAAqB,KAAqC;AACvE,aAAW,UAAU,IAAI,QAAS,QAAO,MAAM;AAC/C,QAAM,IAAI,QAAc,CAAC,YAAY,IAAI,MAAM,MAAM,QAAQ,CAAC,CAAC;AACjE;AAEA,eAAe,UAAU,WAA0B,YAAY,KAAqB;AAClF,QAAM,YAAY,KAAK,IAAI;AAC3B,SAAO,KAAK,IAAI,IAAI,YAAY,WAAW;AACzC,QAAI,UAAU,EAAG;AACjB,UAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,EAAE,CAAC;AAAA,EACxD;AACA,QAAM,IAAI,MAAM,iCAAiC;AACnD;AAEA,SAAS,yCAAyC,MAAM;AAItD,KAAG,6DAA6D,MAAM;AACpE,UAAMA,OAAM,IAAI,IAAI,EAAE,QAAQ,QAAQ,gBAAgB,KAAK,CAAC;AAC5D,WAAO,MAAMA,KAAI,OAAO,CAAC,EAAE,QAAQ,uDAAuD;AAAA,EAC5F,CAAC;AAED,KAAG,8CAA8C,MAAM;AACrD,UAAMA,OAAM,IAAI,IAAI,EAAE,QAAQ,QAAQ,UAAU,QAAQ,CAAC;AACzD,UAAM,SAASA,KAAI,OAAO;AAE1B,WAAO,MAAM;AAAA,EACf,CAAC;AACH,CAAC;AAED,SAAS,4BAA4B,MAAM;AACzC,KAAG,oDAAoD,YAAY;AACjE,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,UAAM,OAAiB,CAAC;AACxB,QAAI,GAAG,cAAc,CAAC,OAAO;AAC3B,SAAG,GAAG,WAAW,CAAC,MAAM,aAAa;AACnC,aAAK,KAAK,WAAW,UAAW,KAAK,MAAM,KAAK,SAAS,CAAC,EAAuB,IAAI;AAAA,MACvF,CAAC;AAAA,IACH,CAAC;AAED,UAAM,SAAS,IAAI,IAAI,EAAE,QAAQ,YAAY,QAAQ,CAAC,EAAE,OAAO;AAC/D,QAAI;AAEF,aAAO,UAAU,UAAU,IAAI,CAAC;AAChC,YAAM,UAAU,MAAM,KAAK,KAAK,GAAG,MAAM,OAAO;AAEhD,aAAO,MAAM;AACb,YAAM,UAAU,MAAM,KAAK,KAAK,GAAG,MAAM,gBAAgB;AAAA,IAC3D,UAAE;AACA,aAAO,MAAM;AACb,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AAED,KAAG,sCAAsC,YAAY;AACnD,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,UAAM,OAAiB,CAAC;AACxB,QAAI,GAAG,cAAc,CAAC,OAAO;AAC3B,SAAG,GAAG,WAAW,CAAC,MAAM,aAAa;AACnC,aAAK,KAAK,WAAW,UAAW,KAAK,MAAM,KAAK,SAAS,CAAC,EAAuB,IAAI;AAAA,MACvF,CAAC;AAAA,IACH,CAAC;AAED,UAAM,SAAS,IAAI,IAAI,EAAE,QAAQ,YAAY,QAAQ,CAAC,EAAE,OAAO;AAC/D,QAAI;AACF,aAAO,UAAU,UAAU,GAAG,CAAC;AAC/B,aAAO,MAAM;AACb,YAAM,UAAU,MAAM,KAAK,KAAK,GAAG,MAAM,gBAAgB;AAAA,IAC3D,UAAE;AACA,aAAO,MAAM;AACb,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AACH,CAAC;AAED,IAAI,mBAAmB;AACrB,WAAS,YAAY,YAAY;AAC/B,UAAM,IAAI,IAAI,IAAI,GAAG,MAAM,IAAI,KAAK,GAAG,EAAE,cAAc,MAAM,CAAC;AAAA,EAChE,CAAC;AACH,OAAO;AACL,WAAS,YAAY,MAAM;AACzB,OAAG,KAAK,6BAA6B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC/C,CAAC;AACH;","names":["stt"]}
|
package/dist/stt_v2.cjs
CHANGED
|
@@ -272,6 +272,7 @@ class SpeechStreamv2 extends import_agents.stt.SpeechStream {
|
|
|
272
272
|
const samples50ms = Math.floor(this.#opts.sampleRate / 20);
|
|
273
273
|
const audioBstream = new import_agents.AudioByteStream(this.#opts.sampleRate, 1, samples50ms);
|
|
274
274
|
let hasEnded = false;
|
|
275
|
+
let inputEnded = false;
|
|
275
276
|
const iterator = this.input[Symbol.asyncIterator]();
|
|
276
277
|
while (true) {
|
|
277
278
|
const nextPromise = iterator.next();
|
|
@@ -280,6 +281,7 @@ class SpeechStreamv2 extends import_agents.stt.SpeechStream {
|
|
|
280
281
|
if ("abort" in result || result.done) {
|
|
281
282
|
if (!("abort" in result) && result.done) {
|
|
282
283
|
hasEnded = true;
|
|
284
|
+
inputEnded = true;
|
|
283
285
|
} else {
|
|
284
286
|
break;
|
|
285
287
|
}
|
|
@@ -299,13 +301,13 @@ class SpeechStreamv2 extends import_agents.stt.SpeechStream {
|
|
|
299
301
|
if (this.#ws.readyState === import_ws.WebSocket.OPEN) {
|
|
300
302
|
this.#ws.send(frame.data);
|
|
301
303
|
}
|
|
302
|
-
if (hasEnded) {
|
|
303
|
-
this.#audioDurationCollector.flush();
|
|
304
|
-
hasEnded = false;
|
|
305
|
-
}
|
|
306
304
|
}
|
|
307
305
|
}
|
|
308
|
-
if (hasEnded)
|
|
306
|
+
if (hasEnded) {
|
|
307
|
+
this.#audioDurationCollector.flush();
|
|
308
|
+
hasEnded = false;
|
|
309
|
+
}
|
|
310
|
+
if (inputEnded) break;
|
|
309
311
|
}
|
|
310
312
|
if (!this.#reconnectEvent.isSet && this.#ws.readyState === import_ws.WebSocket.OPEN) {
|
|
311
313
|
this.#logger.debug("Sending CloseStream message to Deepgram");
|
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 createTimedString,\n log,\n normalizeLanguage,\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 * List of language hints to bias the model for improved accuracy.\n * Only usable with `flux-general-multi`.\n */\n // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 61 line\n languageHint?: 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 // session keyterm propagation)\n #streams = new Set<WeakRef<SpeechStreamv2>>();\n #userKeyterms: string[];\n #sessionKeyterms: string[] = [];\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 * @param opts.languageHint - List of language hints to bias the model for improved accuracy.\n * Only usable with `flux-general-multi`.\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 keyterms: true,\n });\n\n this.#opts = {\n ...defaultSTTv2Options,\n ...opts,\n language: opts.language ? normalizeLanguage(opts.language) : defaultSTTv2Options.language,\n };\n this.#userKeyterms = [...this.#opts.keyterms];\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 // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 134-138 lines\n if (\n this.#opts.languageHint &&\n this.#opts.languageHint.length > 0 &&\n this.#opts.model !== 'flux-general-multi'\n ) {\n this.#logger.warn(\n { model: this.#opts.model },\n '`languageHint` is only supported by `flux-general-multi` and will be ignored for this model',\n );\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 const stream = new SpeechStreamv2(this, streamOpts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\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 const nextOpts = { ...opts };\n if (nextOpts.keyterms !== undefined) {\n this.#userKeyterms = [...nextOpts.keyterms];\n nextOpts.keyterms = [...new Set([...this.#userKeyterms, ...this.#sessionKeyterms])];\n }\n this.#opts = {\n ...this.#opts,\n ...nextOpts,\n language:\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language,\n };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 244-249 lines\n if (\n this.#opts.languageHint &&\n this.#opts.languageHint.length > 0 &&\n this.#opts.model !== 'flux-general-multi'\n ) {\n this.#logger.warn(\n { model: this.#opts.model },\n '`languageHint` is only supported by `flux-general-multi` and will be ignored for this model',\n );\n }\n this.#logger.debug('Updated STTv2 options');\n }\n\n override _updateSessionKeyterms(keyterms: string[]): void {\n if (\n keyterms.length === this.#sessionKeyterms.length &&\n keyterms.every((t, i) => t === this.#sessionKeyterms[i])\n ) {\n return;\n }\n this.#sessionKeyterms = [...keyterms];\n const merged = [...new Set([...this.#userKeyterms, ...keyterms])];\n this.#opts.keyterms = merged;\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (!stream) {\n this.#streams.delete(ref);\n continue;\n }\n if (stream._speaking) {\n // defer the reconnect to the end of the utterance so we don't cut it off\n stream._pendingKeyterm = merged;\n } else {\n stream.updateOptions({ keyterms: merged });\n }\n }\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 // keyterms set while the user is speaking; applied at END_OF_SPEECH (latest wins)\n /** @internal */\n _pendingKeyterm: string[] | null = null;\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 /** @internal */\n get _speaking(): boolean {\n return this.#speaking;\n }\n\n updateOptions(opts: Partial<STTv2Options>) {\n this.#logger.debug('Stream received option update', opts);\n this.#opts = {\n ...this.#opts,\n ...opts,\n language:\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language,\n };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n if (opts.keyterms !== undefined) {\n this._pendingKeyterm = null;\n }\n\n // Trigger reconnection loop\n this.#reconnectEvent.set();\n }\n\n #onEndOfSpeech() {\n if (this._pendingKeyterm !== null) {\n this.updateOptions({ keyterms: this._pendingKeyterm });\n this._pendingKeyterm = null;\n }\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 && result.value === undefined) {\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 this.#onEndOfSpeech();\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: ${desc}`);\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {\n const alts = parseTranscription(this.#opts.language || 'en', data, this.startTimeOffset);\n\n if (alts.length > 0) {\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [alts[0]!, ...alts.slice(1)],\n });\n }\n }\n\n #onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n\n #getDeepgramUrl(): string {\n const params: Record<string, string | string[]> = {\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: 'linear16',\n mip_opt_out: String(this.#opts.mipOptOut),\n };\n\n // Note: v2 API does NOT include 'language' parameter\n if (this.#opts.eagerEotThreshold)\n params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();\n if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();\n if (this.#opts.eotTimeoutMs) params.eot_timeout_ms = this.#opts.eotTimeoutMs.toString();\n\n if (this.#opts.keyterms.length > 0) params.keyterm = this.#opts.keyterms;\n if (this.#opts.tags && this.#opts.tags.length > 0) params.tag = this.#opts.tags;\n\n // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 480-481 lines\n if (this.#opts.languageHint && this.#opts.languageHint.length > 0) {\n params.language_hint = this.#opts.languageHint;\n }\n\n const baseUrl = this.#opts.endpointUrl.replace(/^http/, 'ws');\n const qs = queryString.stringify(params);\n return `${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 // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 587-591 lines\n const detectedLanguagesRaw = Array.isArray(data.languages) ? (data.languages as string[]) : [];\n const detectedLanguages = detectedLanguagesRaw.map((lang) => normalizeLanguage(lang));\n const primaryLanguage =\n detectedLanguages.length > 0 ? detectedLanguages[0]! : normalizeLanguage(language);\n\n const sd: stt.SpeechData = {\n language: primaryLanguage,\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 // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 598 line\n sourceLanguages: detectedLanguages.length > 0 ? detectedLanguages : undefined,\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 createTimedString({\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\n return [sd];\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBASO;AAEP,kBAA6B;AAC7B,gBAA0B;AAC1B,mBAAkC;AAGlC,MAAM,aAAa,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC;AA2BzD,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,EAEd,WAAW,oBAAI,IAA6B;AAAA,EAC5C;AAAA,EACA,mBAA6B,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAkB9B,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,MACnB,UAAU;AAAA,IACZ,CAAC;AAED,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,KAAK,eAAW,iCAAkB,KAAK,QAAQ,IAAI,oBAAoB;AAAA,IACnF;AACA,SAAK,gBAAgB,CAAC,GAAG,KAAK,MAAM,QAAQ;AAE5C,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;AAGA,QACE,KAAK,MAAM,gBACX,KAAK,MAAM,aAAa,SAAS,KACjC,KAAK,MAAM,UAAU,sBACrB;AACA,WAAK,QAAQ;AAAA,QACX,EAAE,OAAO,KAAK,MAAM,MAAM;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;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,UAAM,SAAS,IAAI,eAAe,MAAM,YAAY,mCAAS,WAAW;AACxE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAA6B;AACzC,UAAM,WAAW,EAAE,GAAG,KAAK;AAC3B,QAAI,SAAS,aAAa,QAAW;AACnC,WAAK,gBAAgB,CAAC,GAAG,SAAS,QAAQ;AAC1C,eAAS,WAAW,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,eAAe,GAAG,KAAK,gBAAgB,CAAC,CAAC;AAAA,IACpF;AACA,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,aAAY,iCAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAAA,IAChF;AACA,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AAEvD,QACE,KAAK,MAAM,gBACX,KAAK,MAAM,aAAa,SAAS,KACjC,KAAK,MAAM,UAAU,sBACrB;AACA,WAAK,QAAQ;AAAA,QACX,EAAE,OAAO,KAAK,MAAM,MAAM;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;AACA,SAAK,QAAQ,MAAM,uBAAuB;AAAA,EAC5C;AAAA,EAES,uBAAuB,UAA0B;AACxD,QACE,SAAS,WAAW,KAAK,iBAAiB,UAC1C,SAAS,MAAM,CAAC,GAAG,MAAM,MAAM,KAAK,iBAAiB,CAAC,CAAC,GACvD;AACA;AAAA,IACF;AACA,SAAK,mBAAmB,CAAC,GAAG,QAAQ;AACpC,UAAM,SAAS,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,eAAe,GAAG,QAAQ,CAAC,CAAC;AAChE,SAAK,MAAM,WAAW;AACtB,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,CAAC,QAAQ;AACX,aAAK,SAAS,OAAO,GAAG;AACxB;AAAA,MACF;AACA,UAAI,OAAO,WAAW;AAEpB,eAAO,kBAAkB;AAAA,MAC3B,OAAO;AACL,eAAO,cAAc,EAAE,UAAU,OAAO,CAAC;AAAA,MAC3C;AAAA,IACF;AAAA,EACF;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;AAAA;AAAA,EAI5B,kBAAmC;AAAA,EAEnC,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;AAAA,EAGA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA6B;AACzC,SAAK,QAAQ,MAAM,iCAAiC,IAAI;AACxD,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,aAAY,iCAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAAA,IAChF;AACA,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AACvD,QAAI,KAAK,aAAa,QAAW;AAC/B,WAAK,kBAAkB;AAAA,IACzB;AAGA,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,iBAAiB;AACf,QAAI,KAAK,oBAAoB,MAAM;AACjC,WAAK,cAAc,EAAE,UAAU,KAAK,gBAAgB,CAAC;AACrD,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AArTxB;AAuTI,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;AAtU/B,gBAAAA;AAuUY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,SAAS;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,UAAU,CAAC,QAAe;AA1U1C,gBAAAA;AA2UY,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,OAAO,UAAU,QAAW;AAAA,MAE5C,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;AACD,aAAK,eAAe;AAAA,MACtB;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;AAG3E,QAAI,KAAK,MAAM,gBAAgB,KAAK,MAAM,aAAa,SAAS,GAAG;AACjE,aAAO,gBAAgB,KAAK,MAAM;AAAA,IACpC;AAEA,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;AAljBnB;AAmjBI,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;AAGA,QAAM,uBAAuB,MAAM,QAAQ,KAAK,SAAS,IAAK,KAAK,YAAyB,CAAC;AAC7F,QAAM,oBAAoB,qBAAqB,IAAI,CAAC,aAAS,iCAAkB,IAAI,CAAC;AACpF,QAAM,kBACJ,kBAAkB,SAAS,IAAI,kBAAkB,CAAC,QAAK,iCAAkB,QAAQ;AAEnF,QAAM,KAAqB;AAAA,IACzB,UAAU;AAAA,IACV,YAAa,KAAK,sBAAiC,KAAK;AAAA,IACxD,UAAW,KAAK,oBAA+B,KAAK;AAAA,IACpD;AAAA,IACA,MAAM,cAAc;AAAA;AAAA,IAEpB,iBAAiB,kBAAkB,SAAS,IAAI,oBAAoB;AAAA;AAAA;AAAA;AAAA,IAIpE,OAAO,UAAU;AAAA,MAAI,CAAC,aACpB,iCAAkB;AAAA,QAChB,MAAO,KAAK,QAAmB;AAAA,QAC/B,YAAa,KAAK,SAAoB,KAAK;AAAA,QAC3C,UAAW,KAAK,OAAkB,KAAK;AAAA,QACvC,YAAa,KAAK,cAAyB;AAAA,QAC3C;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;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 createTimedString,\n log,\n normalizeLanguage,\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 * List of language hints to bias the model for improved accuracy.\n * Only usable with `flux-general-multi`.\n */\n // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 61 line\n languageHint?: 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 // session keyterm propagation)\n #streams = new Set<WeakRef<SpeechStreamv2>>();\n #userKeyterms: string[];\n #sessionKeyterms: string[] = [];\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 * @param opts.languageHint - List of language hints to bias the model for improved accuracy.\n * Only usable with `flux-general-multi`.\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 keyterms: true,\n });\n\n this.#opts = {\n ...defaultSTTv2Options,\n ...opts,\n language: opts.language ? normalizeLanguage(opts.language) : defaultSTTv2Options.language,\n };\n this.#userKeyterms = [...this.#opts.keyterms];\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 // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 134-138 lines\n if (\n this.#opts.languageHint &&\n this.#opts.languageHint.length > 0 &&\n this.#opts.model !== 'flux-general-multi'\n ) {\n this.#logger.warn(\n { model: this.#opts.model },\n '`languageHint` is only supported by `flux-general-multi` and will be ignored for this model',\n );\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 const stream = new SpeechStreamv2(this, streamOpts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\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 const nextOpts = { ...opts };\n if (nextOpts.keyterms !== undefined) {\n this.#userKeyterms = [...nextOpts.keyterms];\n nextOpts.keyterms = [...new Set([...this.#userKeyterms, ...this.#sessionKeyterms])];\n }\n this.#opts = {\n ...this.#opts,\n ...nextOpts,\n language:\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language,\n };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 244-249 lines\n if (\n this.#opts.languageHint &&\n this.#opts.languageHint.length > 0 &&\n this.#opts.model !== 'flux-general-multi'\n ) {\n this.#logger.warn(\n { model: this.#opts.model },\n '`languageHint` is only supported by `flux-general-multi` and will be ignored for this model',\n );\n }\n this.#logger.debug('Updated STTv2 options');\n }\n\n override _updateSessionKeyterms(keyterms: string[]): void {\n if (\n keyterms.length === this.#sessionKeyterms.length &&\n keyterms.every((t, i) => t === this.#sessionKeyterms[i])\n ) {\n return;\n }\n this.#sessionKeyterms = [...keyterms];\n const merged = [...new Set([...this.#userKeyterms, ...keyterms])];\n this.#opts.keyterms = merged;\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (!stream) {\n this.#streams.delete(ref);\n continue;\n }\n if (stream._speaking) {\n // defer the reconnect to the end of the utterance so we don't cut it off\n stream._pendingKeyterm = merged;\n } else {\n stream.updateOptions({ keyterms: merged });\n }\n }\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 // keyterms set while the user is speaking; applied at END_OF_SPEECH (latest wins)\n /** @internal */\n _pendingKeyterm: string[] | null = null;\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 /** @internal */\n get _speaking(): boolean {\n return this.#speaking;\n }\n\n updateOptions(opts: Partial<STTv2Options>) {\n this.#logger.debug('Stream received option update', opts);\n this.#opts = {\n ...this.#opts,\n ...opts,\n language:\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language,\n };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n if (opts.keyterms !== undefined) {\n this._pendingKeyterm = null;\n }\n\n // Trigger reconnection loop\n this.#reconnectEvent.set();\n }\n\n #onEndOfSpeech() {\n if (this._pendingKeyterm !== null) {\n this.updateOptions({ keyterms: this._pendingKeyterm });\n this._pendingKeyterm = null;\n }\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 let inputEnded = 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 inputEnded = 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 && result.value === undefined) {\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 }\n\n if (hasEnded) {\n this.#audioDurationCollector.flush();\n hasEnded = false;\n }\n\n if (inputEnded) 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 this.#onEndOfSpeech();\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: ${desc}`);\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {\n const alts = parseTranscription(this.#opts.language || 'en', data, this.startTimeOffset);\n\n if (alts.length > 0) {\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [alts[0]!, ...alts.slice(1)],\n });\n }\n }\n\n #onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n\n #getDeepgramUrl(): string {\n const params: Record<string, string | string[]> = {\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: 'linear16',\n mip_opt_out: String(this.#opts.mipOptOut),\n };\n\n // Note: v2 API does NOT include 'language' parameter\n if (this.#opts.eagerEotThreshold)\n params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();\n if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();\n if (this.#opts.eotTimeoutMs) params.eot_timeout_ms = this.#opts.eotTimeoutMs.toString();\n\n if (this.#opts.keyterms.length > 0) params.keyterm = this.#opts.keyterms;\n if (this.#opts.tags && this.#opts.tags.length > 0) params.tag = this.#opts.tags;\n\n // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 480-481 lines\n if (this.#opts.languageHint && this.#opts.languageHint.length > 0) {\n params.language_hint = this.#opts.languageHint;\n }\n\n const baseUrl = this.#opts.endpointUrl.replace(/^http/, 'ws');\n const qs = queryString.stringify(params);\n return `${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 // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 587-591 lines\n const detectedLanguagesRaw = Array.isArray(data.languages) ? (data.languages as string[]) : [];\n const detectedLanguages = detectedLanguagesRaw.map((lang) => normalizeLanguage(lang));\n const primaryLanguage =\n detectedLanguages.length > 0 ? detectedLanguages[0]! : normalizeLanguage(language);\n\n const sd: stt.SpeechData = {\n language: primaryLanguage,\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 // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 598 line\n sourceLanguages: detectedLanguages.length > 0 ? detectedLanguages : undefined,\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 createTimedString({\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\n return [sd];\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBASO;AAEP,kBAA6B;AAC7B,gBAA0B;AAC1B,mBAAkC;AAGlC,MAAM,aAAa,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC;AA2BzD,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,EAEd,WAAW,oBAAI,IAA6B;AAAA,EAC5C;AAAA,EACA,mBAA6B,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAkB9B,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,MACnB,UAAU;AAAA,IACZ,CAAC;AAED,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,KAAK,eAAW,iCAAkB,KAAK,QAAQ,IAAI,oBAAoB;AAAA,IACnF;AACA,SAAK,gBAAgB,CAAC,GAAG,KAAK,MAAM,QAAQ;AAE5C,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;AAGA,QACE,KAAK,MAAM,gBACX,KAAK,MAAM,aAAa,SAAS,KACjC,KAAK,MAAM,UAAU,sBACrB;AACA,WAAK,QAAQ;AAAA,QACX,EAAE,OAAO,KAAK,MAAM,MAAM;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;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,UAAM,SAAS,IAAI,eAAe,MAAM,YAAY,mCAAS,WAAW;AACxE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAA6B;AACzC,UAAM,WAAW,EAAE,GAAG,KAAK;AAC3B,QAAI,SAAS,aAAa,QAAW;AACnC,WAAK,gBAAgB,CAAC,GAAG,SAAS,QAAQ;AAC1C,eAAS,WAAW,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,eAAe,GAAG,KAAK,gBAAgB,CAAC,CAAC;AAAA,IACpF;AACA,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,aAAY,iCAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAAA,IAChF;AACA,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AAEvD,QACE,KAAK,MAAM,gBACX,KAAK,MAAM,aAAa,SAAS,KACjC,KAAK,MAAM,UAAU,sBACrB;AACA,WAAK,QAAQ;AAAA,QACX,EAAE,OAAO,KAAK,MAAM,MAAM;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;AACA,SAAK,QAAQ,MAAM,uBAAuB;AAAA,EAC5C;AAAA,EAES,uBAAuB,UAA0B;AACxD,QACE,SAAS,WAAW,KAAK,iBAAiB,UAC1C,SAAS,MAAM,CAAC,GAAG,MAAM,MAAM,KAAK,iBAAiB,CAAC,CAAC,GACvD;AACA;AAAA,IACF;AACA,SAAK,mBAAmB,CAAC,GAAG,QAAQ;AACpC,UAAM,SAAS,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,eAAe,GAAG,QAAQ,CAAC,CAAC;AAChE,SAAK,MAAM,WAAW;AACtB,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,CAAC,QAAQ;AACX,aAAK,SAAS,OAAO,GAAG;AACxB;AAAA,MACF;AACA,UAAI,OAAO,WAAW;AAEpB,eAAO,kBAAkB;AAAA,MAC3B,OAAO;AACL,eAAO,cAAc,EAAE,UAAU,OAAO,CAAC;AAAA,MAC3C;AAAA,IACF;AAAA,EACF;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;AAAA;AAAA,EAI5B,kBAAmC;AAAA,EAEnC,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;AAAA,EAGA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA6B;AACzC,SAAK,QAAQ,MAAM,iCAAiC,IAAI;AACxD,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,aAAY,iCAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAAA,IAChF;AACA,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AACvD,QAAI,KAAK,aAAa,QAAW;AAC/B,WAAK,kBAAkB;AAAA,IACzB;AAGA,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,iBAAiB;AACf,QAAI,KAAK,oBAAoB,MAAM;AACjC,WAAK,cAAc,EAAE,UAAU,KAAK,gBAAgB,CAAC;AACrD,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AArTxB;AAuTI,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;AAtU/B,gBAAAA;AAuUY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,SAAS;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,UAAU,CAAC,QAAe;AA1U1C,gBAAAA;AA2UY,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;AACf,QAAI,aAAa;AAGjB,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;AACX,uBAAa;AAAA,QACf,OAAO;AAEL;AAAA,QACF;AAAA,MACF;AAGA,UAAI,YAAY,OAAO,UAAU,QAAW;AAAA,MAE5C,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;AAAA,QACF;AAAA,MACF;AAEA,UAAI,UAAU;AACZ,aAAK,wBAAwB,MAAM;AACnC,mBAAW;AAAA,MACb;AAEA,UAAI,WAAY;AAAA,IAClB;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;AACD,aAAK,eAAe;AAAA,MACtB;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;AAG3E,QAAI,KAAK,MAAM,gBAAgB,KAAK,MAAM,aAAa,SAAS,GAAG;AACjE,aAAO,gBAAgB,KAAK,MAAM;AAAA,IACpC;AAEA,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;AApjBnB;AAqjBI,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;AAGA,QAAM,uBAAuB,MAAM,QAAQ,KAAK,SAAS,IAAK,KAAK,YAAyB,CAAC;AAC7F,QAAM,oBAAoB,qBAAqB,IAAI,CAAC,aAAS,iCAAkB,IAAI,CAAC;AACpF,QAAM,kBACJ,kBAAkB,SAAS,IAAI,kBAAkB,CAAC,QAAK,iCAAkB,QAAQ;AAEnF,QAAM,KAAqB;AAAA,IACzB,UAAU;AAAA,IACV,YAAa,KAAK,sBAAiC,KAAK;AAAA,IACxD,UAAW,KAAK,oBAA+B,KAAK;AAAA,IACpD;AAAA,IACA,MAAM,cAAc;AAAA;AAAA,IAEpB,iBAAiB,kBAAkB,SAAS,IAAI,oBAAoB;AAAA;AAAA;AAAA;AAAA,IAIpE,OAAO,UAAU;AAAA,MAAI,CAAC,aACpB,iCAAkB;AAAA,QAChB,MAAO,KAAK,QAAmB;AAAA,QAC/B,YAAa,KAAK,SAAoB,KAAK;AAAA,QAC3C,UAAW,KAAK,OAAkB,KAAK;AAAA,QACvC,YAAa,KAAK,cAAyB;AAAA,QAC3C;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AAEA,SAAO,CAAC,EAAE;AACZ;","names":["_a"]}
|
package/dist/stt_v2.js
CHANGED
|
@@ -247,6 +247,7 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
247
247
|
const samples50ms = Math.floor(this.#opts.sampleRate / 20);
|
|
248
248
|
const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samples50ms);
|
|
249
249
|
let hasEnded = false;
|
|
250
|
+
let inputEnded = false;
|
|
250
251
|
const iterator = this.input[Symbol.asyncIterator]();
|
|
251
252
|
while (true) {
|
|
252
253
|
const nextPromise = iterator.next();
|
|
@@ -255,6 +256,7 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
255
256
|
if ("abort" in result || result.done) {
|
|
256
257
|
if (!("abort" in result) && result.done) {
|
|
257
258
|
hasEnded = true;
|
|
259
|
+
inputEnded = true;
|
|
258
260
|
} else {
|
|
259
261
|
break;
|
|
260
262
|
}
|
|
@@ -274,13 +276,13 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
274
276
|
if (this.#ws.readyState === WebSocket.OPEN) {
|
|
275
277
|
this.#ws.send(frame.data);
|
|
276
278
|
}
|
|
277
|
-
if (hasEnded) {
|
|
278
|
-
this.#audioDurationCollector.flush();
|
|
279
|
-
hasEnded = false;
|
|
280
|
-
}
|
|
281
279
|
}
|
|
282
280
|
}
|
|
283
|
-
if (hasEnded)
|
|
281
|
+
if (hasEnded) {
|
|
282
|
+
this.#audioDurationCollector.flush();
|
|
283
|
+
hasEnded = false;
|
|
284
|
+
}
|
|
285
|
+
if (inputEnded) break;
|
|
284
286
|
}
|
|
285
287
|
if (!this.#reconnectEvent.isSet && this.#ws.readyState === WebSocket.OPEN) {
|
|
286
288
|
this.#logger.debug("Sending CloseStream message to Deepgram");
|
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 createTimedString,\n log,\n normalizeLanguage,\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 * List of language hints to bias the model for improved accuracy.\n * Only usable with `flux-general-multi`.\n */\n // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 61 line\n languageHint?: 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 // session keyterm propagation)\n #streams = new Set<WeakRef<SpeechStreamv2>>();\n #userKeyterms: string[];\n #sessionKeyterms: string[] = [];\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 * @param opts.languageHint - List of language hints to bias the model for improved accuracy.\n * Only usable with `flux-general-multi`.\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 keyterms: true,\n });\n\n this.#opts = {\n ...defaultSTTv2Options,\n ...opts,\n language: opts.language ? normalizeLanguage(opts.language) : defaultSTTv2Options.language,\n };\n this.#userKeyterms = [...this.#opts.keyterms];\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 // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 134-138 lines\n if (\n this.#opts.languageHint &&\n this.#opts.languageHint.length > 0 &&\n this.#opts.model !== 'flux-general-multi'\n ) {\n this.#logger.warn(\n { model: this.#opts.model },\n '`languageHint` is only supported by `flux-general-multi` and will be ignored for this model',\n );\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 const stream = new SpeechStreamv2(this, streamOpts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\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 const nextOpts = { ...opts };\n if (nextOpts.keyterms !== undefined) {\n this.#userKeyterms = [...nextOpts.keyterms];\n nextOpts.keyterms = [...new Set([...this.#userKeyterms, ...this.#sessionKeyterms])];\n }\n this.#opts = {\n ...this.#opts,\n ...nextOpts,\n language:\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language,\n };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 244-249 lines\n if (\n this.#opts.languageHint &&\n this.#opts.languageHint.length > 0 &&\n this.#opts.model !== 'flux-general-multi'\n ) {\n this.#logger.warn(\n { model: this.#opts.model },\n '`languageHint` is only supported by `flux-general-multi` and will be ignored for this model',\n );\n }\n this.#logger.debug('Updated STTv2 options');\n }\n\n override _updateSessionKeyterms(keyterms: string[]): void {\n if (\n keyterms.length === this.#sessionKeyterms.length &&\n keyterms.every((t, i) => t === this.#sessionKeyterms[i])\n ) {\n return;\n }\n this.#sessionKeyterms = [...keyterms];\n const merged = [...new Set([...this.#userKeyterms, ...keyterms])];\n this.#opts.keyterms = merged;\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (!stream) {\n this.#streams.delete(ref);\n continue;\n }\n if (stream._speaking) {\n // defer the reconnect to the end of the utterance so we don't cut it off\n stream._pendingKeyterm = merged;\n } else {\n stream.updateOptions({ keyterms: merged });\n }\n }\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 // keyterms set while the user is speaking; applied at END_OF_SPEECH (latest wins)\n /** @internal */\n _pendingKeyterm: string[] | null = null;\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 /** @internal */\n get _speaking(): boolean {\n return this.#speaking;\n }\n\n updateOptions(opts: Partial<STTv2Options>) {\n this.#logger.debug('Stream received option update', opts);\n this.#opts = {\n ...this.#opts,\n ...opts,\n language:\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language,\n };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n if (opts.keyterms !== undefined) {\n this._pendingKeyterm = null;\n }\n\n // Trigger reconnection loop\n this.#reconnectEvent.set();\n }\n\n #onEndOfSpeech() {\n if (this._pendingKeyterm !== null) {\n this.updateOptions({ keyterms: this._pendingKeyterm });\n this._pendingKeyterm = null;\n }\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 && result.value === undefined) {\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 this.#onEndOfSpeech();\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: ${desc}`);\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {\n const alts = parseTranscription(this.#opts.language || 'en', data, this.startTimeOffset);\n\n if (alts.length > 0) {\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [alts[0]!, ...alts.slice(1)],\n });\n }\n }\n\n #onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n\n #getDeepgramUrl(): string {\n const params: Record<string, string | string[]> = {\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: 'linear16',\n mip_opt_out: String(this.#opts.mipOptOut),\n };\n\n // Note: v2 API does NOT include 'language' parameter\n if (this.#opts.eagerEotThreshold)\n params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();\n if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();\n if (this.#opts.eotTimeoutMs) params.eot_timeout_ms = this.#opts.eotTimeoutMs.toString();\n\n if (this.#opts.keyterms.length > 0) params.keyterm = this.#opts.keyterms;\n if (this.#opts.tags && this.#opts.tags.length > 0) params.tag = this.#opts.tags;\n\n // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 480-481 lines\n if (this.#opts.languageHint && this.#opts.languageHint.length > 0) {\n params.language_hint = this.#opts.languageHint;\n }\n\n const baseUrl = this.#opts.endpointUrl.replace(/^http/, 'ws');\n const qs = queryString.stringify(params);\n return `${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 // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 587-591 lines\n const detectedLanguagesRaw = Array.isArray(data.languages) ? (data.languages as string[]) : [];\n const detectedLanguages = detectedLanguagesRaw.map((lang) => normalizeLanguage(lang));\n const primaryLanguage =\n detectedLanguages.length > 0 ? detectedLanguages[0]! : normalizeLanguage(language);\n\n const sd: stt.SpeechData = {\n language: primaryLanguage,\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 // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 598 line\n sourceLanguages: detectedLanguages.length > 0 ? detectedLanguages : undefined,\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 createTimedString({\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\n return [sd];\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;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;AA2BzD,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,EAEd,WAAW,oBAAI,IAA6B;AAAA,EAC5C;AAAA,EACA,mBAA6B,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAkB9B,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,MACnB,UAAU;AAAA,IACZ,CAAC;AAED,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,KAAK,WAAW,kBAAkB,KAAK,QAAQ,IAAI,oBAAoB;AAAA,IACnF;AACA,SAAK,gBAAgB,CAAC,GAAG,KAAK,MAAM,QAAQ;AAE5C,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;AAGA,QACE,KAAK,MAAM,gBACX,KAAK,MAAM,aAAa,SAAS,KACjC,KAAK,MAAM,UAAU,sBACrB;AACA,WAAK,QAAQ;AAAA,QACX,EAAE,OAAO,KAAK,MAAM,MAAM;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;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,UAAM,SAAS,IAAI,eAAe,MAAM,YAAY,mCAAS,WAAW;AACxE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAA6B;AACzC,UAAM,WAAW,EAAE,GAAG,KAAK;AAC3B,QAAI,SAAS,aAAa,QAAW;AACnC,WAAK,gBAAgB,CAAC,GAAG,SAAS,QAAQ;AAC1C,eAAS,WAAW,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,eAAe,GAAG,KAAK,gBAAgB,CAAC,CAAC;AAAA,IACpF;AACA,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,SAAY,kBAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAAA,IAChF;AACA,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AAEvD,QACE,KAAK,MAAM,gBACX,KAAK,MAAM,aAAa,SAAS,KACjC,KAAK,MAAM,UAAU,sBACrB;AACA,WAAK,QAAQ;AAAA,QACX,EAAE,OAAO,KAAK,MAAM,MAAM;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;AACA,SAAK,QAAQ,MAAM,uBAAuB;AAAA,EAC5C;AAAA,EAES,uBAAuB,UAA0B;AACxD,QACE,SAAS,WAAW,KAAK,iBAAiB,UAC1C,SAAS,MAAM,CAAC,GAAG,MAAM,MAAM,KAAK,iBAAiB,CAAC,CAAC,GACvD;AACA;AAAA,IACF;AACA,SAAK,mBAAmB,CAAC,GAAG,QAAQ;AACpC,UAAM,SAAS,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,eAAe,GAAG,QAAQ,CAAC,CAAC;AAChE,SAAK,MAAM,WAAW;AACtB,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,CAAC,QAAQ;AACX,aAAK,SAAS,OAAO,GAAG;AACxB;AAAA,MACF;AACA,UAAI,OAAO,WAAW;AAEpB,eAAO,kBAAkB;AAAA,MAC3B,OAAO;AACL,eAAO,cAAc,EAAE,UAAU,OAAO,CAAC;AAAA,MAC3C;AAAA,IACF;AAAA,EACF;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;AAAA;AAAA,EAI5B,kBAAmC;AAAA,EAEnC,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;AAAA,EAGA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA6B;AACzC,SAAK,QAAQ,MAAM,iCAAiC,IAAI;AACxD,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,SAAY,kBAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAAA,IAChF;AACA,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AACvD,QAAI,KAAK,aAAa,QAAW;AAC/B,WAAK,kBAAkB;AAAA,IACzB;AAGA,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,iBAAiB;AACf,QAAI,KAAK,oBAAoB,MAAM;AACjC,WAAK,cAAc,EAAE,UAAU,KAAK,gBAAgB,CAAC;AACrD,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AArTxB;AAuTI,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;AAtU/B,gBAAAA;AAuUY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,SAAS;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,UAAU,CAAC,QAAe;AA1U1C,gBAAAA;AA2UY,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,OAAO,UAAU,QAAW;AAAA,MAE5C,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;AACD,aAAK,eAAe;AAAA,MACtB;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;AAG3E,QAAI,KAAK,MAAM,gBAAgB,KAAK,MAAM,aAAa,SAAS,GAAG;AACjE,aAAO,gBAAgB,KAAK,MAAM;AAAA,IACpC;AAEA,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;AAljBnB;AAmjBI,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;AAGA,QAAM,uBAAuB,MAAM,QAAQ,KAAK,SAAS,IAAK,KAAK,YAAyB,CAAC;AAC7F,QAAM,oBAAoB,qBAAqB,IAAI,CAAC,SAAS,kBAAkB,IAAI,CAAC;AACpF,QAAM,kBACJ,kBAAkB,SAAS,IAAI,kBAAkB,CAAC,IAAK,kBAAkB,QAAQ;AAEnF,QAAM,KAAqB;AAAA,IACzB,UAAU;AAAA,IACV,YAAa,KAAK,sBAAiC,KAAK;AAAA,IACxD,UAAW,KAAK,oBAA+B,KAAK;AAAA,IACpD;AAAA,IACA,MAAM,cAAc;AAAA;AAAA,IAEpB,iBAAiB,kBAAkB,SAAS,IAAI,oBAAoB;AAAA;AAAA;AAAA;AAAA,IAIpE,OAAO,UAAU;AAAA,MAAI,CAAC,SACpB,kBAAkB;AAAA,QAChB,MAAO,KAAK,QAAmB;AAAA,QAC/B,YAAa,KAAK,SAAoB,KAAK;AAAA,QAC3C,UAAW,KAAK,OAAkB,KAAK;AAAA,QACvC,YAAa,KAAK,cAAyB;AAAA,QAC3C;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;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 createTimedString,\n log,\n normalizeLanguage,\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 * List of language hints to bias the model for improved accuracy.\n * Only usable with `flux-general-multi`.\n */\n // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 61 line\n languageHint?: 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 // session keyterm propagation)\n #streams = new Set<WeakRef<SpeechStreamv2>>();\n #userKeyterms: string[];\n #sessionKeyterms: string[] = [];\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 * @param opts.languageHint - List of language hints to bias the model for improved accuracy.\n * Only usable with `flux-general-multi`.\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 keyterms: true,\n });\n\n this.#opts = {\n ...defaultSTTv2Options,\n ...opts,\n language: opts.language ? normalizeLanguage(opts.language) : defaultSTTv2Options.language,\n };\n this.#userKeyterms = [...this.#opts.keyterms];\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 // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 134-138 lines\n if (\n this.#opts.languageHint &&\n this.#opts.languageHint.length > 0 &&\n this.#opts.model !== 'flux-general-multi'\n ) {\n this.#logger.warn(\n { model: this.#opts.model },\n '`languageHint` is only supported by `flux-general-multi` and will be ignored for this model',\n );\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 const stream = new SpeechStreamv2(this, streamOpts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\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 const nextOpts = { ...opts };\n if (nextOpts.keyterms !== undefined) {\n this.#userKeyterms = [...nextOpts.keyterms];\n nextOpts.keyterms = [...new Set([...this.#userKeyterms, ...this.#sessionKeyterms])];\n }\n this.#opts = {\n ...this.#opts,\n ...nextOpts,\n language:\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language,\n };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 244-249 lines\n if (\n this.#opts.languageHint &&\n this.#opts.languageHint.length > 0 &&\n this.#opts.model !== 'flux-general-multi'\n ) {\n this.#logger.warn(\n { model: this.#opts.model },\n '`languageHint` is only supported by `flux-general-multi` and will be ignored for this model',\n );\n }\n this.#logger.debug('Updated STTv2 options');\n }\n\n override _updateSessionKeyterms(keyterms: string[]): void {\n if (\n keyterms.length === this.#sessionKeyterms.length &&\n keyterms.every((t, i) => t === this.#sessionKeyterms[i])\n ) {\n return;\n }\n this.#sessionKeyterms = [...keyterms];\n const merged = [...new Set([...this.#userKeyterms, ...keyterms])];\n this.#opts.keyterms = merged;\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (!stream) {\n this.#streams.delete(ref);\n continue;\n }\n if (stream._speaking) {\n // defer the reconnect to the end of the utterance so we don't cut it off\n stream._pendingKeyterm = merged;\n } else {\n stream.updateOptions({ keyterms: merged });\n }\n }\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 // keyterms set while the user is speaking; applied at END_OF_SPEECH (latest wins)\n /** @internal */\n _pendingKeyterm: string[] | null = null;\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 /** @internal */\n get _speaking(): boolean {\n return this.#speaking;\n }\n\n updateOptions(opts: Partial<STTv2Options>) {\n this.#logger.debug('Stream received option update', opts);\n this.#opts = {\n ...this.#opts,\n ...opts,\n language:\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language,\n };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n if (opts.keyterms !== undefined) {\n this._pendingKeyterm = null;\n }\n\n // Trigger reconnection loop\n this.#reconnectEvent.set();\n }\n\n #onEndOfSpeech() {\n if (this._pendingKeyterm !== null) {\n this.updateOptions({ keyterms: this._pendingKeyterm });\n this._pendingKeyterm = null;\n }\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 let inputEnded = 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 inputEnded = 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 && result.value === undefined) {\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 }\n\n if (hasEnded) {\n this.#audioDurationCollector.flush();\n hasEnded = false;\n }\n\n if (inputEnded) 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 this.#onEndOfSpeech();\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: ${desc}`);\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {\n const alts = parseTranscription(this.#opts.language || 'en', data, this.startTimeOffset);\n\n if (alts.length > 0) {\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [alts[0]!, ...alts.slice(1)],\n });\n }\n }\n\n #onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n\n #getDeepgramUrl(): string {\n const params: Record<string, string | string[]> = {\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: 'linear16',\n mip_opt_out: String(this.#opts.mipOptOut),\n };\n\n // Note: v2 API does NOT include 'language' parameter\n if (this.#opts.eagerEotThreshold)\n params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();\n if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();\n if (this.#opts.eotTimeoutMs) params.eot_timeout_ms = this.#opts.eotTimeoutMs.toString();\n\n if (this.#opts.keyterms.length > 0) params.keyterm = this.#opts.keyterms;\n if (this.#opts.tags && this.#opts.tags.length > 0) params.tag = this.#opts.tags;\n\n // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 480-481 lines\n if (this.#opts.languageHint && this.#opts.languageHint.length > 0) {\n params.language_hint = this.#opts.languageHint;\n }\n\n const baseUrl = this.#opts.endpointUrl.replace(/^http/, 'ws');\n const qs = queryString.stringify(params);\n return `${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 // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 587-591 lines\n const detectedLanguagesRaw = Array.isArray(data.languages) ? (data.languages as string[]) : [];\n const detectedLanguages = detectedLanguagesRaw.map((lang) => normalizeLanguage(lang));\n const primaryLanguage =\n detectedLanguages.length > 0 ? detectedLanguages[0]! : normalizeLanguage(language);\n\n const sd: stt.SpeechData = {\n language: primaryLanguage,\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 // Ref: python livekit-plugins/livekit-plugins-deepgram/livekit/plugins/deepgram/stt_v2.py - 598 line\n sourceLanguages: detectedLanguages.length > 0 ? detectedLanguages : undefined,\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 createTimedString({\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\n return [sd];\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;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;AA2BzD,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,EAEd,WAAW,oBAAI,IAA6B;AAAA,EAC5C;AAAA,EACA,mBAA6B,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAkB9B,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,MACnB,UAAU;AAAA,IACZ,CAAC;AAED,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,KAAK,WAAW,kBAAkB,KAAK,QAAQ,IAAI,oBAAoB;AAAA,IACnF;AACA,SAAK,gBAAgB,CAAC,GAAG,KAAK,MAAM,QAAQ;AAE5C,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;AAGA,QACE,KAAK,MAAM,gBACX,KAAK,MAAM,aAAa,SAAS,KACjC,KAAK,MAAM,UAAU,sBACrB;AACA,WAAK,QAAQ;AAAA,QACX,EAAE,OAAO,KAAK,MAAM,MAAM;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;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,UAAM,SAAS,IAAI,eAAe,MAAM,YAAY,mCAAS,WAAW;AACxE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAA6B;AACzC,UAAM,WAAW,EAAE,GAAG,KAAK;AAC3B,QAAI,SAAS,aAAa,QAAW;AACnC,WAAK,gBAAgB,CAAC,GAAG,SAAS,QAAQ;AAC1C,eAAS,WAAW,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,eAAe,GAAG,KAAK,gBAAgB,CAAC,CAAC;AAAA,IACpF;AACA,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,SAAY,kBAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAAA,IAChF;AACA,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AAEvD,QACE,KAAK,MAAM,gBACX,KAAK,MAAM,aAAa,SAAS,KACjC,KAAK,MAAM,UAAU,sBACrB;AACA,WAAK,QAAQ;AAAA,QACX,EAAE,OAAO,KAAK,MAAM,MAAM;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;AACA,SAAK,QAAQ,MAAM,uBAAuB;AAAA,EAC5C;AAAA,EAES,uBAAuB,UAA0B;AACxD,QACE,SAAS,WAAW,KAAK,iBAAiB,UAC1C,SAAS,MAAM,CAAC,GAAG,MAAM,MAAM,KAAK,iBAAiB,CAAC,CAAC,GACvD;AACA;AAAA,IACF;AACA,SAAK,mBAAmB,CAAC,GAAG,QAAQ;AACpC,UAAM,SAAS,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,eAAe,GAAG,QAAQ,CAAC,CAAC;AAChE,SAAK,MAAM,WAAW;AACtB,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,CAAC,QAAQ;AACX,aAAK,SAAS,OAAO,GAAG;AACxB;AAAA,MACF;AACA,UAAI,OAAO,WAAW;AAEpB,eAAO,kBAAkB;AAAA,MAC3B,OAAO;AACL,eAAO,cAAc,EAAE,UAAU,OAAO,CAAC;AAAA,MAC3C;AAAA,IACF;AAAA,EACF;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;AAAA;AAAA,EAI5B,kBAAmC;AAAA,EAEnC,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;AAAA,EAGA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA6B;AACzC,SAAK,QAAQ,MAAM,iCAAiC,IAAI;AACxD,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,SAAY,kBAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAAA,IAChF;AACA,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AACvD,QAAI,KAAK,aAAa,QAAW;AAC/B,WAAK,kBAAkB;AAAA,IACzB;AAGA,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,iBAAiB;AACf,QAAI,KAAK,oBAAoB,MAAM;AACjC,WAAK,cAAc,EAAE,UAAU,KAAK,gBAAgB,CAAC;AACrD,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AArTxB;AAuTI,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;AAtU/B,gBAAAA;AAuUY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,SAAS;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,UAAU,CAAC,QAAe;AA1U1C,gBAAAA;AA2UY,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;AACf,QAAI,aAAa;AAGjB,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;AACX,uBAAa;AAAA,QACf,OAAO;AAEL;AAAA,QACF;AAAA,MACF;AAGA,UAAI,YAAY,OAAO,UAAU,QAAW;AAAA,MAE5C,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;AAAA,QACF;AAAA,MACF;AAEA,UAAI,UAAU;AACZ,aAAK,wBAAwB,MAAM;AACnC,mBAAW;AAAA,MACb;AAEA,UAAI,WAAY;AAAA,IAClB;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;AACD,aAAK,eAAe;AAAA,MACtB;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;AAG3E,QAAI,KAAK,MAAM,gBAAgB,KAAK,MAAM,aAAa,SAAS,GAAG;AACjE,aAAO,gBAAgB,KAAK,MAAM;AAAA,IACpC;AAEA,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;AApjBnB;AAqjBI,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;AAGA,QAAM,uBAAuB,MAAM,QAAQ,KAAK,SAAS,IAAK,KAAK,YAAyB,CAAC;AAC7F,QAAM,oBAAoB,qBAAqB,IAAI,CAAC,SAAS,kBAAkB,IAAI,CAAC;AACpF,QAAM,kBACJ,kBAAkB,SAAS,IAAI,kBAAkB,CAAC,IAAK,kBAAkB,QAAQ;AAEnF,QAAM,KAAqB;AAAA,IACzB,UAAU;AAAA,IACV,YAAa,KAAK,sBAAiC,KAAK;AAAA,IACxD,UAAW,KAAK,oBAA+B,KAAK;AAAA,IACpD;AAAA,IACA,MAAM,cAAc;AAAA;AAAA,IAEpB,iBAAiB,kBAAkB,SAAS,IAAI,oBAAoB;AAAA;AAAA;AAAA;AAAA,IAIpE,OAAO,UAAU;AAAA,MAAI,CAAC,SACpB,kBAAkB;AAAA,QAChB,MAAO,KAAK,QAAmB;AAAA,QAC/B,YAAa,KAAK,SAAoB,KAAK;AAAA,QAC3C,UAAW,KAAK,OAAkB,KAAK;AAAA,QACvC,YAAa,KAAK,cAAyB;AAAA,QAC3C;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;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.
|
|
3
|
+
"version": "1.6.1",
|
|
4
4
|
"description": "Deepgram plugin for LiveKit Agents for Node.js",
|
|
5
5
|
"main": "dist/index.js",
|
|
6
6
|
"require": "dist/index.cjs",
|
|
@@ -17,7 +17,10 @@
|
|
|
17
17
|
},
|
|
18
18
|
"author": "LiveKit",
|
|
19
19
|
"type": "module",
|
|
20
|
-
"repository":
|
|
20
|
+
"repository": {
|
|
21
|
+
"type": "git",
|
|
22
|
+
"url": "git@github.com:livekit/agents-js.git"
|
|
23
|
+
},
|
|
21
24
|
"license": "Apache-2.0",
|
|
22
25
|
"files": [
|
|
23
26
|
"dist",
|
|
@@ -25,22 +28,22 @@
|
|
|
25
28
|
"README.md"
|
|
26
29
|
],
|
|
27
30
|
"devDependencies": {
|
|
28
|
-
"@livekit/rtc-node": "^0.13.
|
|
31
|
+
"@livekit/rtc-node": "^0.13.33",
|
|
29
32
|
"@microsoft/api-extractor": "^7.35.0",
|
|
30
33
|
"@types/ws": "^8.5.10",
|
|
31
34
|
"tsup": "^8.3.5",
|
|
32
35
|
"typescript": "^5.0.0",
|
|
33
|
-
"@livekit/agents": "1.
|
|
34
|
-
"@livekit/agents-plugin-openai": "^1.
|
|
35
|
-
"@livekit/agents-plugin-silero": "1.
|
|
36
|
-
"@livekit/agents-plugins-test": "1.
|
|
36
|
+
"@livekit/agents": "1.6.1",
|
|
37
|
+
"@livekit/agents-plugin-openai": "^1.6.1",
|
|
38
|
+
"@livekit/agents-plugin-silero": "1.6.1",
|
|
39
|
+
"@livekit/agents-plugins-test": "1.6.1"
|
|
37
40
|
},
|
|
38
41
|
"dependencies": {
|
|
39
|
-
"ws": "^8.
|
|
42
|
+
"ws": "^8.21.0"
|
|
40
43
|
},
|
|
41
44
|
"peerDependencies": {
|
|
42
|
-
"@livekit/rtc-node": "^0.13.
|
|
43
|
-
"@livekit/agents": "1.
|
|
45
|
+
"@livekit/rtc-node": "^0.13.33",
|
|
46
|
+
"@livekit/agents": "1.6.1"
|
|
44
47
|
},
|
|
45
48
|
"scripts": {
|
|
46
49
|
"build": "tsup --onSuccess \"pnpm build:types\"",
|
package/src/stt.test.ts
CHANGED
|
@@ -3,11 +3,42 @@
|
|
|
3
3
|
// SPDX-License-Identifier: Apache-2.0
|
|
4
4
|
import { VAD } from '@livekit/agents-plugin-silero';
|
|
5
5
|
import { stt } from '@livekit/agents-plugins-test';
|
|
6
|
+
import { AudioFrame } from '@livekit/rtc-node';
|
|
7
|
+
import { once } from 'node:events';
|
|
8
|
+
import type { AddressInfo } from 'node:net';
|
|
6
9
|
import { describe, expect, it } from 'vitest';
|
|
10
|
+
import { WebSocketServer } from 'ws';
|
|
7
11
|
import { STT } from './stt.js';
|
|
8
12
|
|
|
9
13
|
const hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);
|
|
10
14
|
|
|
15
|
+
function makeFrame(samplesPerChannel: number, sampleRate = 16000): AudioFrame {
|
|
16
|
+
const data = new Int16Array(samplesPerChannel);
|
|
17
|
+
data.fill(1);
|
|
18
|
+
return new AudioFrame(data, sampleRate, 1, samplesPerChannel);
|
|
19
|
+
}
|
|
20
|
+
|
|
21
|
+
async function startWebSocketServer() {
|
|
22
|
+
const wss = new WebSocketServer({ host: '127.0.0.1', port: 0 });
|
|
23
|
+
await once(wss, 'listening');
|
|
24
|
+
const address = wss.address() as AddressInfo;
|
|
25
|
+
return { wss, baseUrl: `ws://127.0.0.1:${address.port}` };
|
|
26
|
+
}
|
|
27
|
+
|
|
28
|
+
async function closeWebSocketServer(wss: WebSocketServer): Promise<void> {
|
|
29
|
+
for (const client of wss.clients) client.close();
|
|
30
|
+
await new Promise<void>((resolve) => wss.close(() => resolve()));
|
|
31
|
+
}
|
|
32
|
+
|
|
33
|
+
async function waitUntil(predicate: () => boolean, timeoutMs = 1000): Promise<void> {
|
|
34
|
+
const startedAt = Date.now();
|
|
35
|
+
while (Date.now() - startedAt < timeoutMs) {
|
|
36
|
+
if (predicate()) return;
|
|
37
|
+
await new Promise((resolve) => setTimeout(resolve, 10));
|
|
38
|
+
}
|
|
39
|
+
throw new Error('timed out waiting for condition');
|
|
40
|
+
}
|
|
41
|
+
|
|
11
42
|
describe('Deepgram streaming language detection', () => {
|
|
12
43
|
// Deepgram only supports language detection for prerecorded audio, so a
|
|
13
44
|
// streaming session must reject it rather than silently default to English.
|
|
@@ -25,6 +56,51 @@ describe('Deepgram streaming language detection', () => {
|
|
|
25
56
|
});
|
|
26
57
|
});
|
|
27
58
|
|
|
59
|
+
describe('Deepgram streaming flush', () => {
|
|
60
|
+
it('finalizes the turn when no audio is left to send', async () => {
|
|
61
|
+
const { wss, baseUrl } = await startWebSocketServer();
|
|
62
|
+
const wire: string[] = [];
|
|
63
|
+
wss.on('connection', (ws) => {
|
|
64
|
+
ws.on('message', (data, isBinary) => {
|
|
65
|
+
wire.push(isBinary ? 'audio' : (JSON.parse(data.toString()) as { type: string }).type);
|
|
66
|
+
});
|
|
67
|
+
});
|
|
68
|
+
|
|
69
|
+
const stream = new STT({ apiKey: 'test-key', baseUrl }).stream();
|
|
70
|
+
try {
|
|
71
|
+
// 100ms is exactly one JS repack chunk, so flush() returns no frames.
|
|
72
|
+
stream.pushFrame(makeFrame(1600));
|
|
73
|
+
await waitUntil(() => wire.join(',') === 'audio');
|
|
74
|
+
|
|
75
|
+
stream.flush();
|
|
76
|
+
await waitUntil(() => wire.join(',') === 'audio,Finalize');
|
|
77
|
+
} finally {
|
|
78
|
+
stream.close();
|
|
79
|
+
await closeWebSocketServer(wss);
|
|
80
|
+
}
|
|
81
|
+
});
|
|
82
|
+
|
|
83
|
+
it('finalizes after the buffered audio', async () => {
|
|
84
|
+
const { wss, baseUrl } = await startWebSocketServer();
|
|
85
|
+
const wire: string[] = [];
|
|
86
|
+
wss.on('connection', (ws) => {
|
|
87
|
+
ws.on('message', (data, isBinary) => {
|
|
88
|
+
wire.push(isBinary ? 'audio' : (JSON.parse(data.toString()) as { type: string }).type);
|
|
89
|
+
});
|
|
90
|
+
});
|
|
91
|
+
|
|
92
|
+
const stream = new STT({ apiKey: 'test-key', baseUrl }).stream();
|
|
93
|
+
try {
|
|
94
|
+
stream.pushFrame(makeFrame(480));
|
|
95
|
+
stream.flush();
|
|
96
|
+
await waitUntil(() => wire.join(',') === 'audio,Finalize');
|
|
97
|
+
} finally {
|
|
98
|
+
stream.close();
|
|
99
|
+
await closeWebSocketServer(wss);
|
|
100
|
+
}
|
|
101
|
+
});
|
|
102
|
+
});
|
|
103
|
+
|
|
28
104
|
if (hasDeepgramApiKey) {
|
|
29
105
|
describe('Deepgram', async () => {
|
|
30
106
|
await stt(new STT(), await VAD.load(), { nonStreaming: false });
|
package/src/stt.ts
CHANGED
|
@@ -424,9 +424,10 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
424
424
|
const data = result.value;
|
|
425
425
|
|
|
426
426
|
let frames: AudioFrame[];
|
|
427
|
+
let hasEnded = false;
|
|
427
428
|
if (data === SpeechStream.FLUSH_SENTINEL) {
|
|
428
429
|
frames = stream.flush();
|
|
429
|
-
|
|
430
|
+
hasEnded = true;
|
|
430
431
|
} else if (
|
|
431
432
|
data.sampleRate === this.#opts.sampleRate &&
|
|
432
433
|
data.channels === this.#opts.numChannels
|
|
@@ -443,6 +444,11 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
443
444
|
ws.send(frame.data.buffer);
|
|
444
445
|
}
|
|
445
446
|
}
|
|
447
|
+
|
|
448
|
+
if (hasEnded) {
|
|
449
|
+
this.#audioDurationCollector.flush();
|
|
450
|
+
ws.send(JSON.stringify({ type: 'Finalize' }));
|
|
451
|
+
}
|
|
446
452
|
}
|
|
447
453
|
} finally {
|
|
448
454
|
closing = true;
|
package/src/stt_v2.ts
CHANGED
|
@@ -376,6 +376,7 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
376
376
|
const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samples50ms);
|
|
377
377
|
|
|
378
378
|
let hasEnded = false;
|
|
379
|
+
let inputEnded = false;
|
|
379
380
|
|
|
380
381
|
// Manual Iterator to allow racing against Reconnect Signal
|
|
381
382
|
const iterator = this.input[Symbol.asyncIterator]();
|
|
@@ -392,6 +393,7 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
392
393
|
if (!('abort' in result) && result.done) {
|
|
393
394
|
// Normal stream end
|
|
394
395
|
hasEnded = true;
|
|
396
|
+
inputEnded = true;
|
|
395
397
|
} else {
|
|
396
398
|
// Reconnect triggered - break loop immediately
|
|
397
399
|
break;
|
|
@@ -418,15 +420,15 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
418
420
|
if (this.#ws!.readyState === WebSocket.OPEN) {
|
|
419
421
|
this.#ws!.send(frame.data);
|
|
420
422
|
}
|
|
421
|
-
|
|
422
|
-
if (hasEnded) {
|
|
423
|
-
this.#audioDurationCollector.flush();
|
|
424
|
-
hasEnded = false;
|
|
425
|
-
}
|
|
426
423
|
}
|
|
427
424
|
}
|
|
428
425
|
|
|
429
|
-
if (hasEnded)
|
|
426
|
+
if (hasEnded) {
|
|
427
|
+
this.#audioDurationCollector.flush();
|
|
428
|
+
hasEnded = false;
|
|
429
|
+
}
|
|
430
|
+
|
|
431
|
+
if (inputEnded) break;
|
|
430
432
|
}
|
|
431
433
|
|
|
432
434
|
// Only send CloseStream if we are exiting normally (not reconnecting)
|