@livekit/agents-plugin-deepgram 0.5.5 → 0.5.7
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.d.cts +2 -0
- package/dist/models.d.cts +3 -0
- package/dist/stt.cjs +13 -2
- package/dist/stt.cjs.map +1 -1
- package/dist/stt.d.cts +37 -0
- package/dist/stt.d.ts +3 -1
- package/dist/stt.d.ts.map +1 -1
- package/dist/stt.js +20 -3
- package/dist/stt.js.map +1 -1
- package/dist/stt.test.d.cts +2 -0
- package/package.json +12 -8
- package/src/stt.ts +24 -4
package/dist/index.d.cts
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
1
|
+
export type STTModels = 'nova-general' | 'nova-phonecall' | 'nova-meeting' | 'nova-2-general' | 'nova-2-meeting' | 'nova-2-phonecall' | 'nova-2-finance' | 'nova-2-conversationalai' | 'nova-2-voicemail' | 'nova-2-video' | 'nova-2-medical' | 'nova-2-drivethru' | 'nova-2-automotive' | 'nova-3-general' | 'enhanced-general' | 'enhanced-meeting' | 'enhanced-phonecall' | 'enhanced-finance' | 'base' | 'meeting' | 'phonecall' | 'finance' | 'conversationalai' | 'voicemail' | 'video' | 'whisper-tiny' | 'whisper-base' | 'whisper-small' | 'whisper-medium' | 'whisper-large';
|
|
2
|
+
export type STTLanguages = 'da' | 'de' | 'en' | 'en-AU' | 'en-GB' | 'en-IN' | 'en-NZ' | 'en-US' | 'es' | 'es-419' | 'es-LATAM' | 'fr' | 'fr-CA' | 'hi' | 'hi-Latn' | 'id' | 'it' | 'ja' | 'ko' | 'nl' | 'no' | 'pl' | 'pt' | 'pt-BR' | 'ru' | 'sv' | 'ta' | 'taq' | 'th' | 'tr' | 'uk' | 'zh' | 'zh-CN' | 'zh-TW';
|
|
3
|
+
//# sourceMappingURL=models.d.ts.map
|
package/dist/stt.cjs
CHANGED
|
@@ -84,6 +84,9 @@ class STT extends import_agents.stt.STT {
|
|
|
84
84
|
async _recognize(_) {
|
|
85
85
|
throw new Error("Recognize is not supported on Deepgram STT");
|
|
86
86
|
}
|
|
87
|
+
updateOptions(opts) {
|
|
88
|
+
this.#opts = { ...this.#opts, ...opts };
|
|
89
|
+
}
|
|
87
90
|
stream() {
|
|
88
91
|
return new SpeechStream(this, this.#opts);
|
|
89
92
|
}
|
|
@@ -93,6 +96,7 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
93
96
|
#audioEnergyFilter;
|
|
94
97
|
#logger = (0, import_agents.log)();
|
|
95
98
|
#speaking = false;
|
|
99
|
+
#resetWS = new import_agents.Future();
|
|
96
100
|
label = "deepgram.SpeechStream";
|
|
97
101
|
constructor(stt2, opts) {
|
|
98
102
|
super(stt2);
|
|
@@ -159,7 +163,12 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
159
163
|
}
|
|
160
164
|
this.closed = true;
|
|
161
165
|
}
|
|
166
|
+
updateOptions(opts) {
|
|
167
|
+
this.#opts = { ...this.#opts, ...opts };
|
|
168
|
+
this.#resetWS.resolve();
|
|
169
|
+
}
|
|
162
170
|
async #runWS(ws) {
|
|
171
|
+
this.#resetWS = new import_agents.Future();
|
|
163
172
|
let closing = false;
|
|
164
173
|
const keepalive = setInterval(() => {
|
|
165
174
|
try {
|
|
@@ -203,7 +212,7 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
203
212
|
})
|
|
204
213
|
);
|
|
205
214
|
const listenTask = async () => {
|
|
206
|
-
while (!this.closed) {
|
|
215
|
+
while (!this.closed && !closing) {
|
|
207
216
|
try {
|
|
208
217
|
await new Promise((resolve) => {
|
|
209
218
|
ws.once("message", (data) => resolve(data));
|
|
@@ -261,7 +270,9 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
261
270
|
}
|
|
262
271
|
}
|
|
263
272
|
};
|
|
264
|
-
await Promise.all([sendTask(), listenTask(), wsMonitor]);
|
|
273
|
+
await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);
|
|
274
|
+
closing = true;
|
|
275
|
+
ws.close();
|
|
265
276
|
clearInterval(keepalive);
|
|
266
277
|
}
|
|
267
278
|
}
|
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 { type AudioBuffer, AudioByteStream, AudioEnergyFilter, log, stt } from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type RawData, WebSocket } from 'ws';\nimport type { STTLanguages, STTModels } from './models.js';\n\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n}\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-2-general',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n dictation: false,\n diarize: false,\n numerals: false,\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n label = 'deepgram.STT';\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n );\n this.#opts.model = 'nova-2-general';\n }\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Recognize is not supported on Deepgram STT');\n }\n\n stream(): stt.SpeechStream {\n return new SpeechStream(this, this.#opts);\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #audioEnergyFilter: AudioEnergyFilter;\n #logger = log();\n #speaking = false;\n label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions) {\n super(stt);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n\n this.#run();\n }\n\n async #run(maxRetry = 32) {\n let retries = 0;\n let ws: WebSocket;\n while (!this.input.closed) {\n const streamURL = new URL(API_BASE_URL_V1);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n language: this.#opts.language,\n };\n Object.entries(params).forEach(([k, v]) => {\n if (v !== undefined) {\n if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n streamURL.searchParams.append(k, encodeURIComponent(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(x)));\n }\n }\n });\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (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 }\n }\n\n this.closed = true;\n }\n\n async #runWS(ws: WebSocket) {\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 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 for await (const data of this.input) {\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = stream.flush();\n } else if (\n data.sampleRate === this.#opts.sampleRate ||\n data.channels === this.#opts.numChannels\n ) {\n frames = stream.write(data.data.buffer);\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 ws.send(frame.data.buffer);\n }\n }\n }\n\n closing = true;\n ws.send(JSON.stringify({ type: 'CloseStream' }));\n };\n\n const wsMonitor = new Promise<void>((_, reject) =>\n ws.once('close', (code, reason) => {\n if (!closing) {\n this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n reject();\n }\n }),\n );\n\n const listenTask = async () => {\n while (!this.closed) {\n try {\n await new Promise<RawData>((resolve) => {\n ws.once('message', (data) => resolve(data));\n }).then((msg) => {\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 this.queue.put({ 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 isFinal = json['is_final'];\n const isEndpoint = json['speech_final'];\n\n const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);\n\n // If, for some reason, we didn't get a SpeechStarted event but we got\n // a transcript with text, we should start speaking. It's rare but has\n // been observed.\n if (alternatives[0] && alternatives[0].text) {\n if (!this.#speaking) {\n this.#speaking = true;\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n }\n\n if (isFinal) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n } else {\n this.queue.put({\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 this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n });\n } catch (error) {\n this.#logger.child({ error }).warn('unrecoverable error, exiting');\n break;\n }\n }\n };\n\n await Promise.all([sendTask(), listenTask(), wsMonitor]);\n clearInterval(keepalive);\n }\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => ({\n language,\n startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n }));\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAA+E;AAE/E,gBAAwC;AAGxC,MAAM,kBAAkB;AAuBxB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AACZ;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,QAAQ;AAAA,EAER,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,SAA2B;AACzB,WAAO,IAAI,aAAa,MAAM,KAAK,KAAK;AAAA,EAC1C;AACF;AAEO,MAAM,qBAAqB,kBAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,YAAY;AAAA,EACZ,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,gCAAkB;AAEhD,SAAK,KAAK;AAAA,EACZ;AAAA,EAEA,MAAM,KAAK,WAAW,IAAI;AACxB,QAAI,UAAU;AACd,QAAI;AACJ,WAAO,CAAC,KAAK,MAAM,QAAQ;AACzB,YAAM,YAAY,IAAI,IAAI,eAAe;AACzC,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,UAAU,KAAK,MAAM;AAAA,MACvB;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,oBAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,WAAW,UAAU;AACvB,gBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,QACjF;AAEA,cAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,aAAK,QAAQ;AAAA,UACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,QAC3F;AACA,cAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,MAClE;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,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;AAEP,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;AAEA,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI;AACJ,YAAI,SAAS,aAAa,gBAAgB;AACxC,mBAAS,OAAO,MAAM;AAAA,QACxB,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,mBAAS,OAAO,MAAM,KAAK,KAAK,MAAM;AAAA,QACxC,OAAO;AACL,gBAAM,IAAI,MAAM,sDAAsD;AAAA,QACxE;AAEA,yBAAiB,SAAS,QAAQ;AAChC,cAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF;AAEA,gBAAU;AACV,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAAA,IACjD;AAEA,UAAM,YAAY,IAAI;AAAA,MAAc,CAAC,GAAG,WACtC,GAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,iBAAO;AAAA,QACT;AAAA,MACF,CAAC;AAAA,IACH;AAEA,UAAM,aAAa,YAAY;AAC7B,aAAO,CAAC,KAAK,QAAQ;AACnB,YAAI;AACF,gBAAM,IAAI,QAAiB,CAAC,YAAY;AACtC,eAAG,KAAK,WAAW,CAAC,SAAS,QAAQ,IAAI,CAAC;AAAA,UAC5C,CAAC,EAAE,KAAK,CAAC,QAAQ;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,qBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AAEtC,sBAAM,eAAe,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,yBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAAA,kBAC9D;AAEA,sBAAI,SAAS;AACX,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,kBAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,yBAAK,MAAM,IAAI;AAAA,sBACb,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,uBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,cAAc,CAAC;AAAA,gBAC5D;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,EAAE,MAAM,CAAC,EAAE,KAAK,8BAA8B;AACjE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,GAAG,SAAS,CAAC;AACvD,kBAAc,SAAS;AAAA,EACzB;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","names":["stt"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type AudioBuffer,\n AudioByteStream,\n AudioEnergyFilter,\n Future,\n log,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type RawData, WebSocket } from 'ws';\nimport type { STTLanguages, STTModels } from './models.js';\n\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n}\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-2-general',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n dictation: false,\n diarize: false,\n numerals: false,\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n label = 'deepgram.STT';\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n );\n this.#opts.model = 'nova-2-general';\n }\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Recognize is not supported on Deepgram STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n }\n\n stream(): SpeechStream {\n return new SpeechStream(this, this.#opts);\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 label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions) {\n super(stt);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n\n this.#run();\n }\n\n async #run(maxRetry = 32) {\n let retries = 0;\n let ws: WebSocket;\n while (!this.input.closed) {\n const streamURL = new URL(API_BASE_URL_V1);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n language: this.#opts.language,\n };\n Object.entries(params).forEach(([k, v]) => {\n if (v !== undefined) {\n if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n streamURL.searchParams.append(k, encodeURIComponent(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(x)));\n }\n }\n });\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (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 }\n }\n\n this.closed = true;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n this.#resetWS.resolve();\n }\n\n async #runWS(ws: WebSocket) {\n this.#resetWS = new Future();\n let closing = false;\n\n const keepalive = setInterval(() => {\n try {\n ws.send(JSON.stringify({ type: 'KeepAlive' }));\n } catch {\n clearInterval(keepalive);\n return;\n }\n }, 5000);\n\n 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 for await (const data of this.input) {\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = stream.flush();\n } else if (\n data.sampleRate === this.#opts.sampleRate ||\n data.channels === this.#opts.numChannels\n ) {\n frames = stream.write(data.data.buffer);\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 ws.send(frame.data.buffer);\n }\n }\n }\n\n closing = true;\n ws.send(JSON.stringify({ type: 'CloseStream' }));\n };\n\n const wsMonitor = new Promise<void>((_, reject) =>\n ws.once('close', (code, reason) => {\n if (!closing) {\n this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n reject();\n }\n }),\n );\n\n const listenTask = async () => {\n while (!this.closed && !closing) {\n try {\n await new Promise<RawData>((resolve) => {\n ws.once('message', (data) => resolve(data));\n }).then((msg) => {\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 this.queue.put({ 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 isFinal = json['is_final'];\n const isEndpoint = json['speech_final'];\n\n const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);\n\n // If, for some reason, we didn't get a SpeechStarted event but we got\n // a transcript with text, we should start speaking. It's rare but has\n // been observed.\n if (alternatives[0] && alternatives[0].text) {\n if (!this.#speaking) {\n this.#speaking = true;\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n }\n\n if (isFinal) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n } else {\n this.queue.put({\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 this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n });\n } catch (error) {\n this.#logger.child({ error }).warn('unrecoverable error, exiting');\n break;\n }\n }\n };\n\n await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);\n closing = true;\n ws.close();\n clearInterval(keepalive);\n }\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => ({\n language,\n startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n }));\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAOO;AAEP,gBAAwC;AAGxC,MAAM,kBAAkB;AAuBxB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AACZ;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,QAAQ;AAAA,EAER,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;AAAA,EAEA,SAAuB;AACrB,WAAO,IAAI,aAAa,MAAM,KAAK,KAAK;AAAA,EAC1C;AACF;AAEO,MAAM,qBAAqB,kBAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,qBAAO;AAAA,EACtB,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,gCAAkB;AAEhD,SAAK,KAAK;AAAA,EACZ;AAAA,EAEA,MAAM,KAAK,WAAW,IAAI;AACxB,QAAI,UAAU;AACd,QAAI;AACJ,WAAO,CAAC,KAAK,MAAM,QAAQ;AACzB,YAAM,YAAY,IAAI,IAAI,eAAe;AACzC,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,UAAU,KAAK,MAAM;AAAA,MACvB;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,oBAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,WAAW,UAAU;AACvB,gBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,QACjF;AAEA,cAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,aAAK,QAAQ;AAAA,UACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,QAC3F;AACA,cAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,MAClE;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,SAAK,WAAW,IAAI,qBAAO;AAC3B,QAAI,UAAU;AAEd,UAAM,YAAY,YAAY,MAAM;AAClC,UAAI;AACF,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC,CAAC;AAAA,MAC/C,QAAQ;AACN,sBAAc,SAAS;AACvB;AAAA,MACF;AAAA,IACF,GAAG,GAAI;AAEP,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;AAEA,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI;AACJ,YAAI,SAAS,aAAa,gBAAgB;AACxC,mBAAS,OAAO,MAAM;AAAA,QACxB,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,mBAAS,OAAO,MAAM,KAAK,KAAK,MAAM;AAAA,QACxC,OAAO;AACL,gBAAM,IAAI,MAAM,sDAAsD;AAAA,QACxE;AAEA,yBAAiB,SAAS,QAAQ;AAChC,cAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF;AAEA,gBAAU;AACV,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAAA,IACjD;AAEA,UAAM,YAAY,IAAI;AAAA,MAAc,CAAC,GAAG,WACtC,GAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,iBAAO;AAAA,QACT;AAAA,MACF,CAAC;AAAA,IACH;AAEA,UAAM,aAAa,YAAY;AAC7B,aAAO,CAAC,KAAK,UAAU,CAAC,SAAS;AAC/B,YAAI;AACF,gBAAM,IAAI,QAAiB,CAAC,YAAY;AACtC,eAAG,KAAK,WAAW,CAAC,SAAS,QAAQ,IAAI,CAAC;AAAA,UAC5C,CAAC,EAAE,KAAK,CAAC,QAAQ;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,qBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AAEtC,sBAAM,eAAe,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,yBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAAA,kBAC9D;AAEA,sBAAI,SAAS;AACX,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,kBAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,yBAAK,MAAM,IAAI;AAAA,sBACb,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,uBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,cAAc,CAAC;AAAA,gBAC5D;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,EAAE,MAAM,CAAC,EAAE,KAAK,8BAA8B;AACjE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,CAAC,KAAK,SAAS,OAAO,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,GAAG,SAAS,CAAC,CAAC,CAAC;AAC5F,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","names":["stt"]}
|
package/dist/stt.d.cts
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
1
|
+
import { type AudioBuffer, stt } from '@livekit/agents';
|
|
2
|
+
import type { STTLanguages, STTModels } from './models.js';
|
|
3
|
+
export interface STTOptions {
|
|
4
|
+
apiKey?: string;
|
|
5
|
+
language?: STTLanguages | string;
|
|
6
|
+
detectLanguage: boolean;
|
|
7
|
+
interimResults: boolean;
|
|
8
|
+
punctuate: boolean;
|
|
9
|
+
model: STTModels;
|
|
10
|
+
smartFormat: boolean;
|
|
11
|
+
noDelay: boolean;
|
|
12
|
+
endpointing: number;
|
|
13
|
+
fillerWords: boolean;
|
|
14
|
+
sampleRate: number;
|
|
15
|
+
numChannels: number;
|
|
16
|
+
keywords: [string, number][];
|
|
17
|
+
keyterm: string[];
|
|
18
|
+
profanityFilter: boolean;
|
|
19
|
+
dictation: boolean;
|
|
20
|
+
diarize: boolean;
|
|
21
|
+
numerals: boolean;
|
|
22
|
+
}
|
|
23
|
+
export declare class STT extends stt.STT {
|
|
24
|
+
#private;
|
|
25
|
+
label: string;
|
|
26
|
+
constructor(opts?: Partial<STTOptions>);
|
|
27
|
+
_recognize(_: AudioBuffer): Promise<stt.SpeechEvent>;
|
|
28
|
+
updateOptions(opts: Partial<STTOptions>): void;
|
|
29
|
+
stream(): SpeechStream;
|
|
30
|
+
}
|
|
31
|
+
export declare class SpeechStream extends stt.SpeechStream {
|
|
32
|
+
#private;
|
|
33
|
+
label: string;
|
|
34
|
+
constructor(stt: STT, opts: STTOptions);
|
|
35
|
+
updateOptions(opts: Partial<STTOptions>): void;
|
|
36
|
+
}
|
|
37
|
+
//# sourceMappingURL=stt.d.ts.map
|
package/dist/stt.d.ts
CHANGED
|
@@ -25,11 +25,13 @@ export declare class STT extends stt.STT {
|
|
|
25
25
|
label: string;
|
|
26
26
|
constructor(opts?: Partial<STTOptions>);
|
|
27
27
|
_recognize(_: AudioBuffer): Promise<stt.SpeechEvent>;
|
|
28
|
-
|
|
28
|
+
updateOptions(opts: Partial<STTOptions>): void;
|
|
29
|
+
stream(): SpeechStream;
|
|
29
30
|
}
|
|
30
31
|
export declare class SpeechStream extends stt.SpeechStream {
|
|
31
32
|
#private;
|
|
32
33
|
label: string;
|
|
33
34
|
constructor(stt: STT, opts: STTOptions);
|
|
35
|
+
updateOptions(opts: Partial<STTOptions>): void;
|
|
34
36
|
}
|
|
35
37
|
//# sourceMappingURL=stt.d.ts.map
|
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,
|
|
1
|
+
{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,WAAW,EAKhB,GAAG,EACJ,MAAM,iBAAiB,CAAC;AAGzB,OAAO,KAAK,EAAE,YAAY,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAI3D,MAAM,WAAW,UAAU;IACzB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,QAAQ,CAAC,EAAE,YAAY,GAAG,MAAM,CAAC;IACjC,cAAc,EAAE,OAAO,CAAC;IACxB,cAAc,EAAE,OAAO,CAAC;IACxB,SAAS,EAAE,OAAO,CAAC;IACnB,KAAK,EAAE,SAAS,CAAC;IACjB,WAAW,EAAE,OAAO,CAAC;IACrB,OAAO,EAAE,OAAO,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,WAAW,EAAE,OAAO,CAAC;IACrB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,QAAQ,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC,EAAE,CAAC;IAC7B,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,eAAe,EAAE,OAAO,CAAC;IACzB,SAAS,EAAE,OAAO,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,QAAQ,EAAE,OAAO,CAAC;CACnB;AAuBD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAG9B,KAAK,SAAkB;gBAEX,IAAI,GAAE,OAAO,CAAC,UAAU,CAAqB;IAuCnD,UAAU,CAAC,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IAI1D,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAIvC,MAAM,IAAI,YAAY;CAGvB;AAED,qBAAa,YAAa,SAAQ,GAAG,CAAC,YAAY;;IAMhD,KAAK,SAA2B;gBAEpB,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU;IA0EtC,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;CA2IxC"}
|
package/dist/stt.js
CHANGED
|
@@ -1,4 +1,10 @@
|
|
|
1
|
-
import {
|
|
1
|
+
import {
|
|
2
|
+
AudioByteStream,
|
|
3
|
+
AudioEnergyFilter,
|
|
4
|
+
Future,
|
|
5
|
+
log,
|
|
6
|
+
stt
|
|
7
|
+
} from "@livekit/agents";
|
|
2
8
|
import { WebSocket } from "ws";
|
|
3
9
|
const API_BASE_URL_V1 = "wss://api.deepgram.com/v1/listen";
|
|
4
10
|
const defaultSTTOptions = {
|
|
@@ -60,6 +66,9 @@ class STT extends stt.STT {
|
|
|
60
66
|
async _recognize(_) {
|
|
61
67
|
throw new Error("Recognize is not supported on Deepgram STT");
|
|
62
68
|
}
|
|
69
|
+
updateOptions(opts) {
|
|
70
|
+
this.#opts = { ...this.#opts, ...opts };
|
|
71
|
+
}
|
|
63
72
|
stream() {
|
|
64
73
|
return new SpeechStream(this, this.#opts);
|
|
65
74
|
}
|
|
@@ -69,6 +78,7 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
69
78
|
#audioEnergyFilter;
|
|
70
79
|
#logger = log();
|
|
71
80
|
#speaking = false;
|
|
81
|
+
#resetWS = new Future();
|
|
72
82
|
label = "deepgram.SpeechStream";
|
|
73
83
|
constructor(stt2, opts) {
|
|
74
84
|
super(stt2);
|
|
@@ -135,7 +145,12 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
135
145
|
}
|
|
136
146
|
this.closed = true;
|
|
137
147
|
}
|
|
148
|
+
updateOptions(opts) {
|
|
149
|
+
this.#opts = { ...this.#opts, ...opts };
|
|
150
|
+
this.#resetWS.resolve();
|
|
151
|
+
}
|
|
138
152
|
async #runWS(ws) {
|
|
153
|
+
this.#resetWS = new Future();
|
|
139
154
|
let closing = false;
|
|
140
155
|
const keepalive = setInterval(() => {
|
|
141
156
|
try {
|
|
@@ -179,7 +194,7 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
179
194
|
})
|
|
180
195
|
);
|
|
181
196
|
const listenTask = async () => {
|
|
182
|
-
while (!this.closed) {
|
|
197
|
+
while (!this.closed && !closing) {
|
|
183
198
|
try {
|
|
184
199
|
await new Promise((resolve) => {
|
|
185
200
|
ws.once("message", (data) => resolve(data));
|
|
@@ -237,7 +252,9 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
237
252
|
}
|
|
238
253
|
}
|
|
239
254
|
};
|
|
240
|
-
await Promise.all([sendTask(), listenTask(), wsMonitor]);
|
|
255
|
+
await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);
|
|
256
|
+
closing = true;
|
|
257
|
+
ws.close();
|
|
241
258
|
clearInterval(keepalive);
|
|
242
259
|
}
|
|
243
260
|
}
|
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 { type AudioBuffer, AudioByteStream, AudioEnergyFilter, log, stt } from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type RawData, WebSocket } from 'ws';\nimport type { STTLanguages, STTModels } from './models.js';\n\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n}\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-2-general',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n dictation: false,\n diarize: false,\n numerals: false,\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n label = 'deepgram.STT';\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n );\n this.#opts.model = 'nova-2-general';\n }\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Recognize is not supported on Deepgram STT');\n }\n\n stream(): stt.SpeechStream {\n return new SpeechStream(this, this.#opts);\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #audioEnergyFilter: AudioEnergyFilter;\n #logger = log();\n #speaking = false;\n label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions) {\n super(stt);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n\n this.#run();\n }\n\n async #run(maxRetry = 32) {\n let retries = 0;\n let ws: WebSocket;\n while (!this.input.closed) {\n const streamURL = new URL(API_BASE_URL_V1);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n language: this.#opts.language,\n };\n Object.entries(params).forEach(([k, v]) => {\n if (v !== undefined) {\n if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n streamURL.searchParams.append(k, encodeURIComponent(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(x)));\n }\n }\n });\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (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 }\n }\n\n this.closed = true;\n }\n\n async #runWS(ws: WebSocket) {\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 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 for await (const data of this.input) {\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = stream.flush();\n } else if (\n data.sampleRate === this.#opts.sampleRate ||\n data.channels === this.#opts.numChannels\n ) {\n frames = stream.write(data.data.buffer);\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 ws.send(frame.data.buffer);\n }\n }\n }\n\n closing = true;\n ws.send(JSON.stringify({ type: 'CloseStream' }));\n };\n\n const wsMonitor = new Promise<void>((_, reject) =>\n ws.once('close', (code, reason) => {\n if (!closing) {\n this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n reject();\n }\n }),\n );\n\n const listenTask = async () => {\n while (!this.closed) {\n try {\n await new Promise<RawData>((resolve) => {\n ws.once('message', (data) => resolve(data));\n }).then((msg) => {\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 this.queue.put({ 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 isFinal = json['is_final'];\n const isEndpoint = json['speech_final'];\n\n const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);\n\n // If, for some reason, we didn't get a SpeechStarted event but we got\n // a transcript with text, we should start speaking. It's rare but has\n // been observed.\n if (alternatives[0] && alternatives[0].text) {\n if (!this.#speaking) {\n this.#speaking = true;\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n }\n\n if (isFinal) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n } else {\n this.queue.put({\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 this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n });\n } catch (error) {\n this.#logger.child({ error }).warn('unrecoverable error, exiting');\n break;\n }\n }\n };\n\n await Promise.all([sendTask(), listenTask(), wsMonitor]);\n clearInterval(keepalive);\n }\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => ({\n language,\n startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n }));\n};\n"],"mappings":"AAGA,SAA2B,iBAAiB,mBAAmB,KAAK,WAAW;AAE/E,SAAuB,iBAAiB;AAGxC,MAAM,kBAAkB;AAuBxB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AACZ;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU,IAAI;AAAA,EACd,QAAQ;AAAA,EAER,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,SAA2B;AACzB,WAAO,IAAI,aAAa,MAAM,KAAK,KAAK;AAAA,EAC1C;AACF;AAEO,MAAM,qBAAqB,IAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,UAAU,IAAI;AAAA,EACd,YAAY;AAAA,EACZ,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,kBAAkB;AAEhD,SAAK,KAAK;AAAA,EACZ;AAAA,EAEA,MAAM,KAAK,WAAW,IAAI;AACxB,QAAI,UAAU;AACd,QAAI;AACJ,WAAO,CAAC,KAAK,MAAM,QAAQ;AACzB,YAAM,YAAY,IAAI,IAAI,eAAe;AACzC,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,UAAU,KAAK,MAAM;AAAA,MACvB;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,UAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,WAAW,UAAU;AACvB,gBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,QACjF;AAEA,cAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,aAAK,QAAQ;AAAA,UACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,QAC3F;AACA,cAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,MAClE;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,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;AAEP,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;AAEA,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI;AACJ,YAAI,SAAS,aAAa,gBAAgB;AACxC,mBAAS,OAAO,MAAM;AAAA,QACxB,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,mBAAS,OAAO,MAAM,KAAK,KAAK,MAAM;AAAA,QACxC,OAAO;AACL,gBAAM,IAAI,MAAM,sDAAsD;AAAA,QACxE;AAEA,yBAAiB,SAAS,QAAQ;AAChC,cAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF;AAEA,gBAAU;AACV,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAAA,IACjD;AAEA,UAAM,YAAY,IAAI;AAAA,MAAc,CAAC,GAAG,WACtC,GAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,iBAAO;AAAA,QACT;AAAA,MACF,CAAC;AAAA,IACH;AAEA,UAAM,aAAa,YAAY;AAC7B,aAAO,CAAC,KAAK,QAAQ;AACnB,YAAI;AACF,gBAAM,IAAI,QAAiB,CAAC,YAAY;AACtC,eAAG,KAAK,WAAW,CAAC,SAAS,QAAQ,IAAI,CAAC;AAAA,UAC5C,CAAC,EAAE,KAAK,CAAC,QAAQ;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,qBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AAEtC,sBAAM,eAAe,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,yBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAAA,kBAC9D;AAEA,sBAAI,SAAS;AACX,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,yBAAK,MAAM,IAAI;AAAA,sBACb,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,uBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AAAA,gBAC5D;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,EAAE,MAAM,CAAC,EAAE,KAAK,8BAA8B;AACjE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,GAAG,SAAS,CAAC;AACvD,kBAAc,SAAS;AAAA,EACzB;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","names":["stt"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type AudioBuffer,\n AudioByteStream,\n AudioEnergyFilter,\n Future,\n log,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type RawData, WebSocket } from 'ws';\nimport type { STTLanguages, STTModels } from './models.js';\n\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n}\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-2-general',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n dictation: false,\n diarize: false,\n numerals: false,\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n label = 'deepgram.STT';\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n );\n this.#opts.model = 'nova-2-general';\n }\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Recognize is not supported on Deepgram STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n }\n\n stream(): SpeechStream {\n return new SpeechStream(this, this.#opts);\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 label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions) {\n super(stt);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n\n this.#run();\n }\n\n async #run(maxRetry = 32) {\n let retries = 0;\n let ws: WebSocket;\n while (!this.input.closed) {\n const streamURL = new URL(API_BASE_URL_V1);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n language: this.#opts.language,\n };\n Object.entries(params).forEach(([k, v]) => {\n if (v !== undefined) {\n if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n streamURL.searchParams.append(k, encodeURIComponent(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(x)));\n }\n }\n });\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (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 }\n }\n\n this.closed = true;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n this.#resetWS.resolve();\n }\n\n async #runWS(ws: WebSocket) {\n this.#resetWS = new Future();\n let closing = false;\n\n const keepalive = setInterval(() => {\n try {\n ws.send(JSON.stringify({ type: 'KeepAlive' }));\n } catch {\n clearInterval(keepalive);\n return;\n }\n }, 5000);\n\n 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 for await (const data of this.input) {\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = stream.flush();\n } else if (\n data.sampleRate === this.#opts.sampleRate ||\n data.channels === this.#opts.numChannels\n ) {\n frames = stream.write(data.data.buffer);\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 ws.send(frame.data.buffer);\n }\n }\n }\n\n closing = true;\n ws.send(JSON.stringify({ type: 'CloseStream' }));\n };\n\n const wsMonitor = new Promise<void>((_, reject) =>\n ws.once('close', (code, reason) => {\n if (!closing) {\n this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n reject();\n }\n }),\n );\n\n const listenTask = async () => {\n while (!this.closed && !closing) {\n try {\n await new Promise<RawData>((resolve) => {\n ws.once('message', (data) => resolve(data));\n }).then((msg) => {\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 this.queue.put({ 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 isFinal = json['is_final'];\n const isEndpoint = json['speech_final'];\n\n const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);\n\n // If, for some reason, we didn't get a SpeechStarted event but we got\n // a transcript with text, we should start speaking. It's rare but has\n // been observed.\n if (alternatives[0] && alternatives[0].text) {\n if (!this.#speaking) {\n this.#speaking = true;\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n }\n\n if (isFinal) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n } else {\n this.queue.put({\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 this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n });\n } catch (error) {\n this.#logger.child({ error }).warn('unrecoverable error, exiting');\n break;\n }\n }\n };\n\n await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);\n closing = true;\n ws.close();\n clearInterval(keepalive);\n }\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => ({\n language,\n startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n }));\n};\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAuB,iBAAiB;AAGxC,MAAM,kBAAkB;AAuBxB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AACZ;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU,IAAI;AAAA,EACd,QAAQ;AAAA,EAER,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;AAAA,EAEA,SAAuB;AACrB,WAAO,IAAI,aAAa,MAAM,KAAK,KAAK;AAAA,EAC1C;AACF;AAEO,MAAM,qBAAqB,IAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,UAAU,IAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,OAAO;AAAA,EACtB,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,kBAAkB;AAEhD,SAAK,KAAK;AAAA,EACZ;AAAA,EAEA,MAAM,KAAK,WAAW,IAAI;AACxB,QAAI,UAAU;AACd,QAAI;AACJ,WAAO,CAAC,KAAK,MAAM,QAAQ;AACzB,YAAM,YAAY,IAAI,IAAI,eAAe;AACzC,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,UAAU,KAAK,MAAM;AAAA,MACvB;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,UAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,WAAW,UAAU;AACvB,gBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,QACjF;AAEA,cAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,aAAK,QAAQ;AAAA,UACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,QAC3F;AACA,cAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,MAClE;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,SAAK,WAAW,IAAI,OAAO;AAC3B,QAAI,UAAU;AAEd,UAAM,YAAY,YAAY,MAAM;AAClC,UAAI;AACF,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC,CAAC;AAAA,MAC/C,QAAQ;AACN,sBAAc,SAAS;AACvB;AAAA,MACF;AAAA,IACF,GAAG,GAAI;AAEP,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;AAEA,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI;AACJ,YAAI,SAAS,aAAa,gBAAgB;AACxC,mBAAS,OAAO,MAAM;AAAA,QACxB,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,mBAAS,OAAO,MAAM,KAAK,KAAK,MAAM;AAAA,QACxC,OAAO;AACL,gBAAM,IAAI,MAAM,sDAAsD;AAAA,QACxE;AAEA,yBAAiB,SAAS,QAAQ;AAChC,cAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF;AAEA,gBAAU;AACV,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAAA,IACjD;AAEA,UAAM,YAAY,IAAI;AAAA,MAAc,CAAC,GAAG,WACtC,GAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,iBAAO;AAAA,QACT;AAAA,MACF,CAAC;AAAA,IACH;AAEA,UAAM,aAAa,YAAY;AAC7B,aAAO,CAAC,KAAK,UAAU,CAAC,SAAS;AAC/B,YAAI;AACF,gBAAM,IAAI,QAAiB,CAAC,YAAY;AACtC,eAAG,KAAK,WAAW,CAAC,SAAS,QAAQ,IAAI,CAAC;AAAA,UAC5C,CAAC,EAAE,KAAK,CAAC,QAAQ;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,qBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AAEtC,sBAAM,eAAe,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,yBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAAA,kBAC9D;AAEA,sBAAI,SAAS;AACX,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,yBAAK,MAAM,IAAI;AAAA,sBACb,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,uBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AAAA,gBAC5D;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,EAAE,MAAM,CAAC,EAAE,KAAK,8BAA8B;AACjE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,CAAC,KAAK,SAAS,OAAO,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,GAAG,SAAS,CAAC,CAAC,CAAC;AAC5F,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","names":["stt"]}
|
package/package.json
CHANGED
|
@@ -1,15 +1,18 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@livekit/agents-plugin-deepgram",
|
|
3
|
-
"version": "0.5.
|
|
3
|
+
"version": "0.5.7",
|
|
4
4
|
"description": "Deepgram plugin for LiveKit Agents for Node.js",
|
|
5
5
|
"main": "dist/index.js",
|
|
6
6
|
"require": "dist/index.cjs",
|
|
7
7
|
"types": "dist/index.d.ts",
|
|
8
8
|
"exports": {
|
|
9
|
-
"
|
|
9
|
+
"import": {
|
|
10
10
|
"types": "./dist/index.d.ts",
|
|
11
|
-
"
|
|
12
|
-
|
|
11
|
+
"default": "./dist/index.js"
|
|
12
|
+
},
|
|
13
|
+
"require": {
|
|
14
|
+
"types": "./dist/index.d.cts",
|
|
15
|
+
"default": "./dist/index.cjs"
|
|
13
16
|
}
|
|
14
17
|
},
|
|
15
18
|
"author": "LiveKit",
|
|
@@ -25,7 +28,7 @@
|
|
|
25
28
|
"@livekit/agents": "^x",
|
|
26
29
|
"@livekit/agents-plugin-silero": "^x",
|
|
27
30
|
"@livekit/agents-plugins-test": "^x",
|
|
28
|
-
"@livekit/rtc-node": "^0.13.
|
|
31
|
+
"@livekit/rtc-node": "^0.13.11",
|
|
29
32
|
"@microsoft/api-extractor": "^7.35.0",
|
|
30
33
|
"@types/ws": "^8.5.10",
|
|
31
34
|
"tsup": "^8.3.5",
|
|
@@ -35,11 +38,12 @@
|
|
|
35
38
|
"ws": "^8.16.0"
|
|
36
39
|
},
|
|
37
40
|
"peerDependencies": {
|
|
38
|
-
"@livekit/rtc-node": "^0.13.
|
|
39
|
-
"@livekit/agents": "^0.7.
|
|
41
|
+
"@livekit/rtc-node": "^0.13.11",
|
|
42
|
+
"@livekit/agents": "^0.7.7x"
|
|
40
43
|
},
|
|
41
44
|
"scripts": {
|
|
42
|
-
"build": "tsup --onSuccess \"
|
|
45
|
+
"build": "tsup --onSuccess \"pnpm build:types\"",
|
|
46
|
+
"build:types": "tsc --declaration --emitDeclarationOnly && node ../../scripts/copyDeclarationOutput.js",
|
|
43
47
|
"clean": "rm -rf dist",
|
|
44
48
|
"clean:build": "pnpm clean && pnpm build",
|
|
45
49
|
"lint": "eslint -f unix \"src/**/*.{ts,js}\"",
|
package/src/stt.ts
CHANGED
|
@@ -1,7 +1,14 @@
|
|
|
1
1
|
// SPDX-FileCopyrightText: 2024 LiveKit, Inc.
|
|
2
2
|
//
|
|
3
3
|
// SPDX-License-Identifier: Apache-2.0
|
|
4
|
-
import {
|
|
4
|
+
import {
|
|
5
|
+
type AudioBuffer,
|
|
6
|
+
AudioByteStream,
|
|
7
|
+
AudioEnergyFilter,
|
|
8
|
+
Future,
|
|
9
|
+
log,
|
|
10
|
+
stt,
|
|
11
|
+
} from '@livekit/agents';
|
|
5
12
|
import type { AudioFrame } from '@livekit/rtc-node';
|
|
6
13
|
import { type RawData, WebSocket } from 'ws';
|
|
7
14
|
import type { STTLanguages, STTModels } from './models.js';
|
|
@@ -98,7 +105,11 @@ export class STT extends stt.STT {
|
|
|
98
105
|
throw new Error('Recognize is not supported on Deepgram STT');
|
|
99
106
|
}
|
|
100
107
|
|
|
101
|
-
|
|
108
|
+
updateOptions(opts: Partial<STTOptions>) {
|
|
109
|
+
this.#opts = { ...this.#opts, ...opts };
|
|
110
|
+
}
|
|
111
|
+
|
|
112
|
+
stream(): SpeechStream {
|
|
102
113
|
return new SpeechStream(this, this.#opts);
|
|
103
114
|
}
|
|
104
115
|
}
|
|
@@ -108,6 +119,7 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
108
119
|
#audioEnergyFilter: AudioEnergyFilter;
|
|
109
120
|
#logger = log();
|
|
110
121
|
#speaking = false;
|
|
122
|
+
#resetWS = new Future();
|
|
111
123
|
label = 'deepgram.SpeechStream';
|
|
112
124
|
|
|
113
125
|
constructor(stt: STT, opts: STTOptions) {
|
|
@@ -184,7 +196,13 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
184
196
|
this.closed = true;
|
|
185
197
|
}
|
|
186
198
|
|
|
199
|
+
updateOptions(opts: Partial<STTOptions>) {
|
|
200
|
+
this.#opts = { ...this.#opts, ...opts };
|
|
201
|
+
this.#resetWS.resolve();
|
|
202
|
+
}
|
|
203
|
+
|
|
187
204
|
async #runWS(ws: WebSocket) {
|
|
205
|
+
this.#resetWS = new Future();
|
|
188
206
|
let closing = false;
|
|
189
207
|
|
|
190
208
|
const keepalive = setInterval(() => {
|
|
@@ -238,7 +256,7 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
238
256
|
);
|
|
239
257
|
|
|
240
258
|
const listenTask = async () => {
|
|
241
|
-
while (!this.closed) {
|
|
259
|
+
while (!this.closed && !closing) {
|
|
242
260
|
try {
|
|
243
261
|
await new Promise<RawData>((resolve) => {
|
|
244
262
|
ws.once('message', (data) => resolve(data));
|
|
@@ -312,7 +330,9 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
312
330
|
}
|
|
313
331
|
};
|
|
314
332
|
|
|
315
|
-
await Promise.all([sendTask(), listenTask(), wsMonitor]);
|
|
333
|
+
await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);
|
|
334
|
+
closing = true;
|
|
335
|
+
ws.close();
|
|
316
336
|
clearInterval(keepalive);
|
|
317
337
|
}
|
|
318
338
|
}
|