@livekit/agents-plugin-deepgram 0.5.3 → 0.5.4
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/models.cjs.map +1 -1
- package/dist/models.d.ts +1 -1
- package/dist/models.d.ts.map +1 -1
- package/dist/stt.cjs +18 -11
- package/dist/stt.cjs.map +1 -1
- package/dist/stt.d.ts +3 -0
- package/dist/stt.d.ts.map +1 -1
- package/dist/stt.js +18 -11
- package/dist/stt.js.map +1 -1
- package/package.json +4 -4
- package/src/models.ts +1 -0
- package/src/stt.ts +20 -10
package/dist/models.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/models.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport type STTModels =\n | 'nova-general'\n | 'nova-phonecall'\n | 'nova-meeting'\n | 'nova-2-general'\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 | 'enhanced-general'\n | 'enhanced-meeting'\n | 'enhanced-phonecall'\n | 'enhanced-finance'\n | 'base'\n | 'meeting'\n | 'phonecall'\n | 'finance'\n | 'conversationalai'\n | 'voicemail'\n | 'video'\n | 'whisper-tiny'\n | 'whisper-base'\n | 'whisper-small'\n | 'whisper-medium'\n | 'whisper-large';\n\nexport type STTLanguages =\n | 'da'\n | 'de'\n | 'en'\n | 'en-AU'\n | 'en-GB'\n | 'en-IN'\n | 'en-NZ'\n | 'en-US'\n | 'es'\n | 'es-419'\n | 'es-LATAM'\n | 'fr'\n | 'fr-CA'\n | 'hi'\n | 'hi-Latn'\n | 'id'\n | 'it'\n | 'ja'\n | 'ko'\n | 'nl'\n | 'no'\n | 'pl'\n | 'pt'\n | 'pt-BR'\n | 'ru'\n | 'sv'\n | 'ta'\n | 'taq'\n | 'th'\n | 'tr'\n | 'uk'\n | 'zh'\n | 'zh-CN'\n | 'zh-TW';\n"],"mappings":";;;;;;;;;;;;;;AAAA;AAAA;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../src/models.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport type STTModels =\n | 'nova-general'\n | 'nova-phonecall'\n | 'nova-meeting'\n | 'nova-2-general'\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 | 'enhanced-general'\n | 'enhanced-meeting'\n | 'enhanced-phonecall'\n | 'enhanced-finance'\n | 'base'\n | 'meeting'\n | 'phonecall'\n | 'finance'\n | 'conversationalai'\n | 'voicemail'\n | 'video'\n | 'whisper-tiny'\n | 'whisper-base'\n | 'whisper-small'\n | 'whisper-medium'\n | 'whisper-large';\n\nexport type STTLanguages =\n | 'da'\n | 'de'\n | 'en'\n | 'en-AU'\n | 'en-GB'\n | 'en-IN'\n | 'en-NZ'\n | 'en-US'\n | 'es'\n | 'es-419'\n | 'es-LATAM'\n | 'fr'\n | 'fr-CA'\n | 'hi'\n | 'hi-Latn'\n | 'id'\n | 'it'\n | 'ja'\n | 'ko'\n | 'nl'\n | 'no'\n | 'pl'\n | 'pt'\n | 'pt-BR'\n | 'ru'\n | 'sv'\n | 'ta'\n | 'taq'\n | 'th'\n | 'tr'\n | 'uk'\n | 'zh'\n | 'zh-CN'\n | 'zh-TW';\n"],"mappings":";;;;;;;;;;;;;;AAAA;AAAA;","names":[]}
|
package/dist/models.d.ts
CHANGED
|
@@ -1,3 +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' | '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';
|
|
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
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
3
|
//# sourceMappingURL=models.d.ts.map
|
package/dist/models.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"models.d.ts","sourceRoot":"","sources":["../src/models.ts"],"names":[],"mappings":"AAIA,MAAM,MAAM,SAAS,GACjB,cAAc,GACd,gBAAgB,GAChB,cAAc,GACd,gBAAgB,GAChB,gBAAgB,GAChB,kBAAkB,GAClB,gBAAgB,GAChB,yBAAyB,GACzB,kBAAkB,GAClB,cAAc,GACd,gBAAgB,GAChB,kBAAkB,GAClB,mBAAmB,GACnB,kBAAkB,GAClB,kBAAkB,GAClB,oBAAoB,GACpB,kBAAkB,GAClB,MAAM,GACN,SAAS,GACT,WAAW,GACX,SAAS,GACT,kBAAkB,GAClB,WAAW,GACX,OAAO,GACP,cAAc,GACd,cAAc,GACd,eAAe,GACf,gBAAgB,GAChB,eAAe,CAAC;AAEpB,MAAM,MAAM,YAAY,GACpB,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,OAAO,GACP,OAAO,GACP,OAAO,GACP,OAAO,GACP,OAAO,GACP,IAAI,GACJ,QAAQ,GACR,UAAU,GACV,IAAI,GACJ,OAAO,GACP,IAAI,GACJ,SAAS,GACT,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,OAAO,GACP,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,KAAK,GACL,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,OAAO,GACP,OAAO,CAAC"}
|
|
1
|
+
{"version":3,"file":"models.d.ts","sourceRoot":"","sources":["../src/models.ts"],"names":[],"mappings":"AAIA,MAAM,MAAM,SAAS,GACjB,cAAc,GACd,gBAAgB,GAChB,cAAc,GACd,gBAAgB,GAChB,gBAAgB,GAChB,kBAAkB,GAClB,gBAAgB,GAChB,yBAAyB,GACzB,kBAAkB,GAClB,cAAc,GACd,gBAAgB,GAChB,kBAAkB,GAClB,mBAAmB,GACnB,gBAAgB,GAChB,kBAAkB,GAClB,kBAAkB,GAClB,oBAAoB,GACpB,kBAAkB,GAClB,MAAM,GACN,SAAS,GACT,WAAW,GACX,SAAS,GACT,kBAAkB,GAClB,WAAW,GACX,OAAO,GACP,cAAc,GACd,cAAc,GACd,eAAe,GACf,gBAAgB,GAChB,eAAe,CAAC;AAEpB,MAAM,MAAM,YAAY,GACpB,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,OAAO,GACP,OAAO,GACP,OAAO,GACP,OAAO,GACP,OAAO,GACP,IAAI,GACJ,QAAQ,GACR,UAAU,GACV,IAAI,GACJ,OAAO,GACP,IAAI,GACJ,SAAS,GACT,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,OAAO,GACP,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,KAAK,GACL,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,OAAO,GACP,OAAO,CAAC"}
|
package/dist/stt.cjs
CHANGED
|
@@ -39,7 +39,10 @@ const defaultSTTOptions = {
|
|
|
39
39
|
sampleRate: 16e3,
|
|
40
40
|
numChannels: 1,
|
|
41
41
|
keywords: [],
|
|
42
|
-
profanityFilter: false
|
|
42
|
+
profanityFilter: false,
|
|
43
|
+
dictation: false,
|
|
44
|
+
diarize: false,
|
|
45
|
+
numerals: false
|
|
43
46
|
};
|
|
44
47
|
class STT extends import_agents.stt.STT {
|
|
45
48
|
#opts;
|
|
@@ -67,7 +70,8 @@ class STT extends import_agents.stt.STT {
|
|
|
67
70
|
"nova-2-video",
|
|
68
71
|
"nova-2-medical",
|
|
69
72
|
"nova-2-drivethru",
|
|
70
|
-
"nova-2-automotive"
|
|
73
|
+
"nova-2-automotive",
|
|
74
|
+
"nova-3-general"
|
|
71
75
|
].includes(this.#opts.model)) {
|
|
72
76
|
this.#logger.warn(
|
|
73
77
|
`${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`
|
|
@@ -105,6 +109,9 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
105
109
|
model: this.#opts.model,
|
|
106
110
|
punctuate: this.#opts.punctuate,
|
|
107
111
|
smart_format: this.#opts.smartFormat,
|
|
112
|
+
dictation: this.#opts.dictation,
|
|
113
|
+
diarize: this.#opts.diarize,
|
|
114
|
+
numerals: this.#opts.numerals,
|
|
108
115
|
no_delay: this.#opts.noDelay,
|
|
109
116
|
interim_results: this.#opts.interimResults,
|
|
110
117
|
encoding: "linear16",
|
|
@@ -185,15 +192,15 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
185
192
|
closing = true;
|
|
186
193
|
ws.send(JSON.stringify({ type: "CloseStream" }));
|
|
187
194
|
};
|
|
195
|
+
const wsMonitor = new Promise(
|
|
196
|
+
(_, reject) => ws.once("close", (code, reason) => {
|
|
197
|
+
if (!closing) {
|
|
198
|
+
this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
|
|
199
|
+
reject();
|
|
200
|
+
}
|
|
201
|
+
})
|
|
202
|
+
);
|
|
188
203
|
const listenTask = async () => {
|
|
189
|
-
new Promise(
|
|
190
|
-
(_, reject) => ws.once("close", (code, reason) => {
|
|
191
|
-
if (!closing) {
|
|
192
|
-
this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
|
|
193
|
-
reject();
|
|
194
|
-
}
|
|
195
|
-
})
|
|
196
|
-
);
|
|
197
204
|
while (!this.closed) {
|
|
198
205
|
try {
|
|
199
206
|
await new Promise((resolve) => {
|
|
@@ -252,7 +259,7 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
252
259
|
}
|
|
253
260
|
}
|
|
254
261
|
};
|
|
255
|
-
await Promise.all([sendTask(), listenTask()]);
|
|
262
|
+
await Promise.all([sendTask(), listenTask(), wsMonitor]);
|
|
256
263
|
clearInterval(keepalive);
|
|
257
264
|
}
|
|
258
265
|
}
|
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 profanityFilter: 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 profanityFilter: 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 $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${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 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 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('keywords', 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 listenTask = async () => {\n 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 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()]);\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;AAmBxB,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,iBAAiB;AACnB;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,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,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,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,YAAY,mBAAmB,CAAC,CAAC,CAAC;AAAA,UACnF;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,aAAa,YAAY;AAC7B,UAAI;AAAA,QAAc,CAAC,GAAG,WACpB,GAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,cAAI,CAAC,SAAS;AACZ,iBAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,mBAAO;AAAA,UACT;AAAA,QACF,CAAC;AAAA,MACH;AAEA,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,CAAC,CAAC;AAC5C,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 { 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 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 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 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('keywords', 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;AAsBxB,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,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,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,YAAY,mBAAmB,CAAC,CAAC,CAAC;AAAA,UACnF;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"]}
|
package/dist/stt.d.ts
CHANGED
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,EAAE,KAAK,WAAW,EAA2C,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAGjG,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,eAAe,EAAE,OAAO,CAAC;
|
|
1
|
+
{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,KAAK,WAAW,EAA2C,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAGjG,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,eAAe,EAAE,OAAO,CAAC;IACzB,SAAS,EAAE,OAAO,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,QAAQ,EAAE,OAAO,CAAC;CACnB;AAsBD,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,MAAM,IAAI,GAAG,CAAC,YAAY;CAG3B;AAED,qBAAa,YAAa,SAAQ,GAAG,CAAC,YAAY;;IAKhD,KAAK,SAA2B;gBAEpB,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU;CA4MvC"}
|
package/dist/stt.js
CHANGED
|
@@ -15,7 +15,10 @@ const defaultSTTOptions = {
|
|
|
15
15
|
sampleRate: 16e3,
|
|
16
16
|
numChannels: 1,
|
|
17
17
|
keywords: [],
|
|
18
|
-
profanityFilter: false
|
|
18
|
+
profanityFilter: false,
|
|
19
|
+
dictation: false,
|
|
20
|
+
diarize: false,
|
|
21
|
+
numerals: false
|
|
19
22
|
};
|
|
20
23
|
class STT extends stt.STT {
|
|
21
24
|
#opts;
|
|
@@ -43,7 +46,8 @@ class STT extends stt.STT {
|
|
|
43
46
|
"nova-2-video",
|
|
44
47
|
"nova-2-medical",
|
|
45
48
|
"nova-2-drivethru",
|
|
46
|
-
"nova-2-automotive"
|
|
49
|
+
"nova-2-automotive",
|
|
50
|
+
"nova-3-general"
|
|
47
51
|
].includes(this.#opts.model)) {
|
|
48
52
|
this.#logger.warn(
|
|
49
53
|
`${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`
|
|
@@ -81,6 +85,9 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
81
85
|
model: this.#opts.model,
|
|
82
86
|
punctuate: this.#opts.punctuate,
|
|
83
87
|
smart_format: this.#opts.smartFormat,
|
|
88
|
+
dictation: this.#opts.dictation,
|
|
89
|
+
diarize: this.#opts.diarize,
|
|
90
|
+
numerals: this.#opts.numerals,
|
|
84
91
|
no_delay: this.#opts.noDelay,
|
|
85
92
|
interim_results: this.#opts.interimResults,
|
|
86
93
|
encoding: "linear16",
|
|
@@ -161,15 +168,15 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
161
168
|
closing = true;
|
|
162
169
|
ws.send(JSON.stringify({ type: "CloseStream" }));
|
|
163
170
|
};
|
|
171
|
+
const wsMonitor = new Promise(
|
|
172
|
+
(_, reject) => ws.once("close", (code, reason) => {
|
|
173
|
+
if (!closing) {
|
|
174
|
+
this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
|
|
175
|
+
reject();
|
|
176
|
+
}
|
|
177
|
+
})
|
|
178
|
+
);
|
|
164
179
|
const listenTask = async () => {
|
|
165
|
-
new Promise(
|
|
166
|
-
(_, reject) => ws.once("close", (code, reason) => {
|
|
167
|
-
if (!closing) {
|
|
168
|
-
this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
|
|
169
|
-
reject();
|
|
170
|
-
}
|
|
171
|
-
})
|
|
172
|
-
);
|
|
173
180
|
while (!this.closed) {
|
|
174
181
|
try {
|
|
175
182
|
await new Promise((resolve) => {
|
|
@@ -228,7 +235,7 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
228
235
|
}
|
|
229
236
|
}
|
|
230
237
|
};
|
|
231
|
-
await Promise.all([sendTask(), listenTask()]);
|
|
238
|
+
await Promise.all([sendTask(), listenTask(), wsMonitor]);
|
|
232
239
|
clearInterval(keepalive);
|
|
233
240
|
}
|
|
234
241
|
}
|
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 profanityFilter: 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 profanityFilter: 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 $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${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 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 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('keywords', 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 listenTask = async () => {\n 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 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()]);\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;AAmBxB,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,iBAAiB;AACnB;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,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,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,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,YAAY,mBAAmB,CAAC,CAAC,CAAC;AAAA,UACnF;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,aAAa,YAAY;AAC7B,UAAI;AAAA,QAAc,CAAC,GAAG,WACpB,GAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,cAAI,CAAC,SAAS;AACZ,iBAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,mBAAO;AAAA,UACT;AAAA,QACF,CAAC;AAAA,MACH;AAEA,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,CAAC,CAAC;AAC5C,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 { 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 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 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 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('keywords', 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;AAsBxB,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,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,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,YAAY,mBAAmB,CAAC,CAAC,CAAC;AAAA,UACnF;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"]}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@livekit/agents-plugin-deepgram",
|
|
3
|
-
"version": "0.5.
|
|
3
|
+
"version": "0.5.4",
|
|
4
4
|
"description": "Deepgram plugin for LiveKit Agents for Node.js",
|
|
5
5
|
"main": "dist/index.js",
|
|
6
6
|
"require": "dist/index.cjs",
|
|
@@ -25,7 +25,7 @@
|
|
|
25
25
|
"@livekit/agents": "^x",
|
|
26
26
|
"@livekit/agents-plugin-silero": "^x",
|
|
27
27
|
"@livekit/agents-plugins-test": "^x",
|
|
28
|
-
"@livekit/rtc-node": "^0.13.
|
|
28
|
+
"@livekit/rtc-node": "^0.13.4",
|
|
29
29
|
"@microsoft/api-extractor": "^7.35.0",
|
|
30
30
|
"@types/ws": "^8.5.10",
|
|
31
31
|
"tsup": "^8.3.5",
|
|
@@ -35,8 +35,8 @@
|
|
|
35
35
|
"ws": "^8.16.0"
|
|
36
36
|
},
|
|
37
37
|
"peerDependencies": {
|
|
38
|
-
"@livekit/rtc-node": "^0.13.
|
|
39
|
-
"@livekit/agents": "^0.
|
|
38
|
+
"@livekit/rtc-node": "^0.13.4",
|
|
39
|
+
"@livekit/agents": "^0.7.0x"
|
|
40
40
|
},
|
|
41
41
|
"scripts": {
|
|
42
42
|
"build": "tsup --onSuccess \"tsc --declaration --emitDeclarationOnly\"",
|
package/src/models.ts
CHANGED
package/src/stt.ts
CHANGED
|
@@ -23,6 +23,9 @@ export interface STTOptions {
|
|
|
23
23
|
numChannels: number;
|
|
24
24
|
keywords: [string, number][];
|
|
25
25
|
profanityFilter: boolean;
|
|
26
|
+
dictation: boolean;
|
|
27
|
+
diarize: boolean;
|
|
28
|
+
numerals: boolean;
|
|
26
29
|
}
|
|
27
30
|
|
|
28
31
|
const defaultSTTOptions: STTOptions = {
|
|
@@ -40,6 +43,9 @@ const defaultSTTOptions: STTOptions = {
|
|
|
40
43
|
numChannels: 1,
|
|
41
44
|
keywords: [],
|
|
42
45
|
profanityFilter: false,
|
|
46
|
+
dictation: false,
|
|
47
|
+
diarize: false,
|
|
48
|
+
numerals: false,
|
|
43
49
|
};
|
|
44
50
|
|
|
45
51
|
export class STT extends stt.STT {
|
|
@@ -75,6 +81,7 @@ export class STT extends stt.STT {
|
|
|
75
81
|
'nova-2-medical',
|
|
76
82
|
'nova-2-drivethru',
|
|
77
83
|
'nova-2-automotive',
|
|
84
|
+
'nova-3-general',
|
|
78
85
|
].includes(this.#opts.model)
|
|
79
86
|
) {
|
|
80
87
|
this.#logger.warn(
|
|
@@ -119,6 +126,9 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
119
126
|
model: this.#opts.model,
|
|
120
127
|
punctuate: this.#opts.punctuate,
|
|
121
128
|
smart_format: this.#opts.smartFormat,
|
|
129
|
+
dictation: this.#opts.dictation,
|
|
130
|
+
diarize: this.#opts.diarize,
|
|
131
|
+
numerals: this.#opts.numerals,
|
|
122
132
|
no_delay: this.#opts.noDelay,
|
|
123
133
|
interim_results: this.#opts.interimResults,
|
|
124
134
|
encoding: 'linear16',
|
|
@@ -215,16 +225,16 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
215
225
|
ws.send(JSON.stringify({ type: 'CloseStream' }));
|
|
216
226
|
};
|
|
217
227
|
|
|
218
|
-
const
|
|
219
|
-
|
|
220
|
-
|
|
221
|
-
|
|
222
|
-
|
|
223
|
-
|
|
224
|
-
|
|
225
|
-
|
|
226
|
-
);
|
|
228
|
+
const wsMonitor = new Promise<void>((_, reject) =>
|
|
229
|
+
ws.once('close', (code, reason) => {
|
|
230
|
+
if (!closing) {
|
|
231
|
+
this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
|
|
232
|
+
reject();
|
|
233
|
+
}
|
|
234
|
+
}),
|
|
235
|
+
);
|
|
227
236
|
|
|
237
|
+
const listenTask = async () => {
|
|
228
238
|
while (!this.closed) {
|
|
229
239
|
try {
|
|
230
240
|
await new Promise<RawData>((resolve) => {
|
|
@@ -299,7 +309,7 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
299
309
|
}
|
|
300
310
|
};
|
|
301
311
|
|
|
302
|
-
await Promise.all([sendTask(), listenTask()]);
|
|
312
|
+
await Promise.all([sendTask(), listenTask(), wsMonitor]);
|
|
303
313
|
clearInterval(keepalive);
|
|
304
314
|
}
|
|
305
315
|
}
|