@livekit/agents-plugin-deepgram 0.4.6 → 0.5.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md ADDED
@@ -0,0 +1,17 @@
1
+ <!--
2
+ SPDX-FileCopyrightText: 2024 LiveKit, Inc.
3
+
4
+ SPDX-License-Identifier: Apache-2.0
5
+ -->
6
+ # Deepgram plugin for LiveKit Agents
7
+
8
+ The Agents Framework is designed for building realtime, programmable
9
+ participants that run on servers. Use it to create conversational, multi-modal
10
+ voice agents that can see, hear, and understand.
11
+
12
+ This package contains the Deepgram plugin, which allows for speech recognition.
13
+ Refer to the [documentation](https://docs.livekit.io/agents/overview/) for
14
+ information on how to use it, or browse the [API
15
+ reference](https://docs.livekit.io/agents-js/modules/plugins_agents_plugin_deepgram.html).
16
+ See the [repository](https://github.com/livekit/agents-js) for more information
17
+ about the framework as a whole.
package/dist/index.cjs ADDED
@@ -0,0 +1,23 @@
1
+ "use strict";
2
+ var __defProp = Object.defineProperty;
3
+ var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
4
+ var __getOwnPropNames = Object.getOwnPropertyNames;
5
+ var __hasOwnProp = Object.prototype.hasOwnProperty;
6
+ var __copyProps = (to, from, except, desc) => {
7
+ if (from && typeof from === "object" || typeof from === "function") {
8
+ for (let key of __getOwnPropNames(from))
9
+ if (!__hasOwnProp.call(to, key) && key !== except)
10
+ __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
11
+ }
12
+ return to;
13
+ };
14
+ var __reExport = (target, mod, secondTarget) => (__copyProps(target, mod, "default"), secondTarget && __copyProps(secondTarget, mod, "default"));
15
+ var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
16
+ var src_exports = {};
17
+ module.exports = __toCommonJS(src_exports);
18
+ __reExport(src_exports, require("./stt.cjs"), module.exports);
19
+ // Annotate the CommonJS export names for ESM import in node:
20
+ 0 && (module.exports = {
21
+ ...require("./stt.cjs")
22
+ });
23
+ //# sourceMappingURL=index.cjs.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport * from './stt.js';\n"],"mappings":";;;;;;;;;;;;;;;AAAA;AAAA;AAIA,wBAAc,qBAJd;","names":[]}
package/dist/index.js CHANGED
@@ -1,5 +1,2 @@
1
- // SPDX-FileCopyrightText: 2024 LiveKit, Inc.
2
- //
3
- // SPDX-License-Identifier: Apache-2.0
4
- export * from './stt.js';
1
+ export * from "./stt.js";
5
2
  //# sourceMappingURL=index.js.map
package/dist/index.js.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"file":"index.js","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA,6CAA6C;AAC7C,EAAE;AACF,sCAAsC;AAEtC,cAAc,UAAU,CAAC"}
1
+ {"version":3,"sources":["../src/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport * from './stt.js';\n"],"mappings":"AAIA,cAAc;","names":[]}
@@ -0,0 +1,17 @@
1
+ "use strict";
2
+ var __defProp = Object.defineProperty;
3
+ var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
4
+ var __getOwnPropNames = Object.getOwnPropertyNames;
5
+ var __hasOwnProp = Object.prototype.hasOwnProperty;
6
+ var __copyProps = (to, from, except, desc) => {
7
+ if (from && typeof from === "object" || typeof from === "function") {
8
+ for (let key of __getOwnPropNames(from))
9
+ if (!__hasOwnProp.call(to, key) && key !== except)
10
+ __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
11
+ }
12
+ return to;
13
+ };
14
+ var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
15
+ var models_exports = {};
16
+ module.exports = __toCommonJS(models_exports);
17
+ //# sourceMappingURL=models.cjs.map
@@ -0,0 +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":[]}
package/dist/models.js CHANGED
@@ -1,5 +1 @@
1
- // SPDX-FileCopyrightText: 2024 LiveKit, Inc.
2
- //
3
- // SPDX-License-Identifier: Apache-2.0
4
- export {};
5
1
  //# sourceMappingURL=models.js.map
@@ -1 +1 @@
1
- {"version":3,"file":"models.js","sourceRoot":"","sources":["../src/models.ts"],"names":[],"mappings":"AAAA,6CAA6C;AAC7C,EAAE;AACF,sCAAsC"}
1
+ {"version":3,"sources":[],"sourcesContent":[],"mappings":"","names":[]}
package/dist/stt.cjs ADDED
@@ -0,0 +1,274 @@
1
+ "use strict";
2
+ var __defProp = Object.defineProperty;
3
+ var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
4
+ var __getOwnPropNames = Object.getOwnPropertyNames;
5
+ var __hasOwnProp = Object.prototype.hasOwnProperty;
6
+ var __export = (target, all) => {
7
+ for (var name in all)
8
+ __defProp(target, name, { get: all[name], enumerable: true });
9
+ };
10
+ var __copyProps = (to, from, except, desc) => {
11
+ if (from && typeof from === "object" || typeof from === "function") {
12
+ for (let key of __getOwnPropNames(from))
13
+ if (!__hasOwnProp.call(to, key) && key !== except)
14
+ __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
15
+ }
16
+ return to;
17
+ };
18
+ var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
19
+ var stt_exports = {};
20
+ __export(stt_exports, {
21
+ STT: () => STT,
22
+ SpeechStream: () => SpeechStream
23
+ });
24
+ module.exports = __toCommonJS(stt_exports);
25
+ var import_agents = require("@livekit/agents");
26
+ var import_ws = require("ws");
27
+ const API_BASE_URL_V1 = "wss://api.deepgram.com/v1/listen";
28
+ const defaultSTTOptions = {
29
+ apiKey: process.env.DEEPGRAM_API_KEY,
30
+ language: "en-US",
31
+ detectLanguage: false,
32
+ interimResults: true,
33
+ punctuate: true,
34
+ model: "nova-2-general",
35
+ smartFormat: true,
36
+ noDelay: true,
37
+ endpointing: 25,
38
+ fillerWords: false,
39
+ sampleRate: 16e3,
40
+ numChannels: 1,
41
+ keywords: [],
42
+ profanityFilter: false
43
+ };
44
+ class STT extends import_agents.stt.STT {
45
+ #opts;
46
+ #logger = (0, import_agents.log)();
47
+ label = "deepgram.STT";
48
+ constructor(opts = defaultSTTOptions) {
49
+ super({
50
+ streaming: true,
51
+ interimResults: opts.interimResults ?? defaultSTTOptions.interimResults
52
+ });
53
+ if (opts.apiKey === void 0 && defaultSTTOptions.apiKey === void 0) {
54
+ throw new Error(
55
+ "Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY"
56
+ );
57
+ }
58
+ this.#opts = { ...defaultSTTOptions, ...opts };
59
+ if (this.#opts.detectLanguage) {
60
+ this.#opts.language = void 0;
61
+ } else if (this.#opts.language && !["en-US", "en"].includes(this.#opts.language) && [
62
+ "nova-2-meeting",
63
+ "nova-2-phonecall",
64
+ "nova-2-finance",
65
+ "nova-2-conversationalai",
66
+ "nova-2-voicemail",
67
+ "nova-2-video",
68
+ "nova-2-medical",
69
+ "nova-2-drivethru",
70
+ "nova-2-automotive"
71
+ ].includes(this.#opts.model)) {
72
+ this.#logger.warn(
73
+ `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`
74
+ );
75
+ this.#opts.model = "nova-2-general";
76
+ }
77
+ }
78
+ // eslint-disable-next-line @typescript-eslint/no-unused-vars
79
+ async _recognize(_) {
80
+ throw new Error("Recognize is not supported on Deepgram STT");
81
+ }
82
+ stream() {
83
+ return new SpeechStream(this, this.#opts);
84
+ }
85
+ }
86
+ class SpeechStream extends import_agents.stt.SpeechStream {
87
+ #opts;
88
+ #audioEnergyFilter;
89
+ #logger = (0, import_agents.log)();
90
+ #speaking = false;
91
+ label = "deepgram.SpeechStream";
92
+ constructor(stt2, opts) {
93
+ super(stt2);
94
+ this.#opts = opts;
95
+ this.closed = false;
96
+ this.#audioEnergyFilter = new import_agents.AudioEnergyFilter();
97
+ this.#run();
98
+ }
99
+ async #run(maxRetry = 32) {
100
+ let retries = 0;
101
+ let ws;
102
+ while (!this.input.closed) {
103
+ const streamURL = new URL(API_BASE_URL_V1);
104
+ const params = {
105
+ model: this.#opts.model,
106
+ punctuate: this.#opts.punctuate,
107
+ smart_format: this.#opts.smartFormat,
108
+ no_delay: this.#opts.noDelay,
109
+ interim_results: this.#opts.interimResults,
110
+ encoding: "linear16",
111
+ vad_events: true,
112
+ sample_rate: this.#opts.sampleRate,
113
+ channels: this.#opts.numChannels,
114
+ endpointing: this.#opts.endpointing || false,
115
+ filler_words: this.#opts.fillerWords,
116
+ keywords: this.#opts.keywords.map((x) => x.join(":")),
117
+ profanity_filter: this.#opts.profanityFilter,
118
+ language: this.#opts.language
119
+ };
120
+ Object.entries(params).forEach(([k, v]) => {
121
+ if (v !== void 0) {
122
+ if (typeof v === "string" || typeof v === "number" || typeof v === "boolean") {
123
+ streamURL.searchParams.append(k, encodeURIComponent(v));
124
+ } else {
125
+ v.forEach((x) => streamURL.searchParams.append("keywords", encodeURIComponent(x)));
126
+ }
127
+ }
128
+ });
129
+ ws = new import_ws.WebSocket(streamURL, {
130
+ headers: { Authorization: `Token ${this.#opts.apiKey}` }
131
+ });
132
+ try {
133
+ await new Promise((resolve, reject) => {
134
+ ws.on("open", resolve);
135
+ ws.on("error", (error) => reject(error));
136
+ ws.on("close", (code) => reject(`WebSocket returned ${code}`));
137
+ });
138
+ await this.#runWS(ws);
139
+ } catch (e) {
140
+ if (retries >= maxRetry) {
141
+ throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);
142
+ }
143
+ const delay = Math.min(retries * 5, 10);
144
+ retries++;
145
+ this.#logger.warn(
146
+ `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`
147
+ );
148
+ await new Promise((resolve) => setTimeout(resolve, delay * 1e3));
149
+ }
150
+ }
151
+ this.closed = true;
152
+ }
153
+ async #runWS(ws) {
154
+ let closing = false;
155
+ const keepalive = setInterval(() => {
156
+ try {
157
+ ws.send(JSON.stringify({ type: "KeepAlive" }));
158
+ } catch {
159
+ clearInterval(keepalive);
160
+ return;
161
+ }
162
+ }, 5e3);
163
+ const sendTask = async () => {
164
+ const samples100Ms = Math.floor(this.#opts.sampleRate / 10);
165
+ const stream = new import_agents.AudioByteStream(
166
+ this.#opts.sampleRate,
167
+ this.#opts.numChannels,
168
+ samples100Ms
169
+ );
170
+ for await (const data of this.input) {
171
+ let frames;
172
+ if (data === SpeechStream.FLUSH_SENTINEL) {
173
+ frames = stream.flush();
174
+ } else if (data.sampleRate === this.#opts.sampleRate || data.channels === this.#opts.numChannels) {
175
+ frames = stream.write(data.data.buffer);
176
+ } else {
177
+ throw new Error(`sample rate or channel count of frame does not match`);
178
+ }
179
+ for await (const frame of frames) {
180
+ if (this.#audioEnergyFilter.pushFrame(frame)) {
181
+ ws.send(frame.data.buffer);
182
+ }
183
+ }
184
+ }
185
+ closing = true;
186
+ ws.send(JSON.stringify({ type: "CloseStream" }));
187
+ };
188
+ 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
+ while (!this.closed) {
198
+ try {
199
+ await new Promise((resolve) => {
200
+ ws.once("message", (data) => resolve(data));
201
+ }).then((msg) => {
202
+ const json = JSON.parse(msg.toString());
203
+ switch (json["type"]) {
204
+ case "SpeechStarted": {
205
+ if (this.#speaking) return;
206
+ this.#speaking = true;
207
+ this.queue.put({ type: import_agents.stt.SpeechEventType.START_OF_SPEECH });
208
+ break;
209
+ }
210
+ // see this page:
211
+ // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final
212
+ // for more information about the different types of events
213
+ case "Results": {
214
+ const isFinal = json["is_final"];
215
+ const isEndpoint = json["speech_final"];
216
+ const alternatives = liveTranscriptionToSpeechData(this.#opts.language, json);
217
+ if (alternatives[0] && alternatives[0].text) {
218
+ if (!this.#speaking) {
219
+ this.#speaking = true;
220
+ this.queue.put({ type: import_agents.stt.SpeechEventType.START_OF_SPEECH });
221
+ }
222
+ if (isFinal) {
223
+ this.queue.put({
224
+ type: import_agents.stt.SpeechEventType.FINAL_TRANSCRIPT,
225
+ alternatives: [alternatives[0], ...alternatives.slice(1)]
226
+ });
227
+ } else {
228
+ this.queue.put({
229
+ type: import_agents.stt.SpeechEventType.INTERIM_TRANSCRIPT,
230
+ alternatives: [alternatives[0], ...alternatives.slice(1)]
231
+ });
232
+ }
233
+ }
234
+ if (isEndpoint && this.#speaking) {
235
+ this.#speaking = false;
236
+ this.queue.put({ type: import_agents.stt.SpeechEventType.END_OF_SPEECH });
237
+ }
238
+ break;
239
+ }
240
+ case "Metadata": {
241
+ break;
242
+ }
243
+ default: {
244
+ this.#logger.child({ msg: json }).warn("received unexpected message from Deepgram");
245
+ break;
246
+ }
247
+ }
248
+ });
249
+ } catch (error) {
250
+ this.#logger.child({ error }).warn("unrecoverable error, exiting");
251
+ break;
252
+ }
253
+ }
254
+ };
255
+ await Promise.all([sendTask(), listenTask()]);
256
+ clearInterval(keepalive);
257
+ }
258
+ }
259
+ const liveTranscriptionToSpeechData = (language, data) => {
260
+ const alts = data["channel"]["alternatives"];
261
+ return alts.map((alt) => ({
262
+ language,
263
+ startTime: alt["words"].length ? alt["words"][0]["start"] : 0,
264
+ endTime: alt["words"].length ? alt["words"][alt["words"].length - 1]["end"] : 0,
265
+ confidence: alt["confidence"],
266
+ text: alt["transcript"]
267
+ }));
268
+ };
269
+ // Annotate the CommonJS export names for ESM import in node:
270
+ 0 && (module.exports = {
271
+ STT,
272
+ SpeechStream
273
+ });
274
+ //# sourceMappingURL=stt.cjs.map
@@ -0,0 +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"]}
package/dist/stt.d.ts CHANGED
@@ -18,12 +18,14 @@ export interface STTOptions {
18
18
  }
19
19
  export declare class STT extends stt.STT {
20
20
  #private;
21
+ label: string;
21
22
  constructor(opts?: Partial<STTOptions>);
22
- recognize(_: AudioBuffer): Promise<stt.SpeechEvent>;
23
+ _recognize(_: AudioBuffer): Promise<stt.SpeechEvent>;
23
24
  stream(): stt.SpeechStream;
24
25
  }
25
26
  export declare class SpeechStream extends stt.SpeechStream {
26
27
  #private;
27
- constructor(opts: STTOptions);
28
+ label: string;
29
+ constructor(stt: STT, opts: STTOptions);
28
30
  }
29
31
  //# 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,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;CAC1B;AAmBD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;gBAIlB,IAAI,GAAE,OAAO,CAAC,UAAU,CAAqB;IAsCnD,SAAS,CAAC,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IAIzD,MAAM,IAAI,GAAG,CAAC,YAAY;CAG3B;AAED,qBAAa,YAAa,SAAQ,GAAG,CAAC,YAAY;;gBAMpC,IAAI,EAAE,UAAU;CAyM7B"}
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;CAC1B;AAmBD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAG9B,KAAK,SAAkB;gBAEX,IAAI,GAAE,OAAO,CAAC,UAAU,CAAqB;IAsCnD,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;CAyMvC"}
package/dist/stt.js CHANGED
@@ -1,258 +1,249 @@
1
- var _a;
2
- // SPDX-FileCopyrightText: 2024 LiveKit, Inc.
3
- //
4
- // SPDX-License-Identifier: Apache-2.0
5
- import { AudioByteStream, AudioEnergyFilter, log, stt } from '@livekit/agents';
6
- import { WebSocket } from 'ws';
7
- const API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';
1
+ import { AudioByteStream, AudioEnergyFilter, log, stt } from "@livekit/agents";
2
+ import { WebSocket } from "ws";
3
+ const API_BASE_URL_V1 = "wss://api.deepgram.com/v1/listen";
8
4
  const defaultSTTOptions = {
9
- apiKey: process.env.DEEPGRAM_API_KEY,
10
- language: 'en-US',
11
- detectLanguage: false,
12
- interimResults: true,
13
- punctuate: true,
14
- model: 'nova-2-general',
15
- smartFormat: true,
16
- noDelay: true,
17
- endpointing: 25,
18
- fillerWords: false,
19
- sampleRate: 16000,
20
- numChannels: 1,
21
- keywords: [],
22
- profanityFilter: false,
5
+ apiKey: process.env.DEEPGRAM_API_KEY,
6
+ language: "en-US",
7
+ detectLanguage: false,
8
+ interimResults: true,
9
+ punctuate: true,
10
+ model: "nova-2-general",
11
+ smartFormat: true,
12
+ noDelay: true,
13
+ endpointing: 25,
14
+ fillerWords: false,
15
+ sampleRate: 16e3,
16
+ numChannels: 1,
17
+ keywords: [],
18
+ profanityFilter: false
23
19
  };
24
- export class STT extends stt.STT {
25
- #opts;
26
- #logger = log();
27
- constructor(opts = defaultSTTOptions) {
28
- super({
29
- streaming: true,
30
- interimResults: opts.interimResults || defaultSTTOptions.interimResults,
31
- });
32
- if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {
33
- throw new Error('Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY');
34
- }
35
- this.#opts = { ...defaultSTTOptions, ...opts };
36
- if (this.#opts.detectLanguage) {
37
- this.#opts.language = undefined;
38
- }
39
- else if (this.#opts.language &&
40
- !['en-US', 'en'].includes(this.#opts.language) &&
41
- [
42
- 'nova-2-meeting',
43
- 'nova-2-phonecall',
44
- 'nova-2-finance',
45
- 'nova-2-conversationalai',
46
- 'nova-2-voicemail',
47
- 'nova-2-video',
48
- 'nova-2-medical',
49
- 'nova-2-drivethru',
50
- 'nova-2-automotive',
51
- ].includes(this.#opts.model)) {
52
- this.#logger.warn(`${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`);
53
- this.#opts.model = 'nova-2-general';
54
- }
20
+ class STT extends stt.STT {
21
+ #opts;
22
+ #logger = log();
23
+ label = "deepgram.STT";
24
+ constructor(opts = defaultSTTOptions) {
25
+ super({
26
+ streaming: true,
27
+ interimResults: opts.interimResults ?? defaultSTTOptions.interimResults
28
+ });
29
+ if (opts.apiKey === void 0 && defaultSTTOptions.apiKey === void 0) {
30
+ throw new Error(
31
+ "Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY"
32
+ );
55
33
  }
56
- // eslint-disable-next-line @typescript-eslint/no-unused-vars
57
- async recognize(_) {
58
- throw new Error('Recognize is not supported on Deepgram STT');
59
- }
60
- stream() {
61
- return new SpeechStream(this.#opts);
34
+ this.#opts = { ...defaultSTTOptions, ...opts };
35
+ if (this.#opts.detectLanguage) {
36
+ this.#opts.language = void 0;
37
+ } else if (this.#opts.language && !["en-US", "en"].includes(this.#opts.language) && [
38
+ "nova-2-meeting",
39
+ "nova-2-phonecall",
40
+ "nova-2-finance",
41
+ "nova-2-conversationalai",
42
+ "nova-2-voicemail",
43
+ "nova-2-video",
44
+ "nova-2-medical",
45
+ "nova-2-drivethru",
46
+ "nova-2-automotive"
47
+ ].includes(this.#opts.model)) {
48
+ this.#logger.warn(
49
+ `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`
50
+ );
51
+ this.#opts.model = "nova-2-general";
62
52
  }
53
+ }
54
+ // eslint-disable-next-line @typescript-eslint/no-unused-vars
55
+ async _recognize(_) {
56
+ throw new Error("Recognize is not supported on Deepgram STT");
57
+ }
58
+ stream() {
59
+ return new SpeechStream(this, this.#opts);
60
+ }
63
61
  }
64
- export class SpeechStream extends stt.SpeechStream {
65
- #opts;
66
- #audioEnergyFilter;
67
- #logger = log();
68
- #speaking = false;
69
- constructor(opts) {
70
- super();
71
- this.#opts = opts;
72
- this.closed = false;
73
- this.#audioEnergyFilter = new AudioEnergyFilter();
74
- this.#run();
75
- }
76
- async #run(maxRetry = 32) {
77
- let retries = 0;
78
- let ws;
79
- while (!this.input.closed) {
80
- const streamURL = new URL(API_BASE_URL_V1);
81
- const params = {
82
- model: this.#opts.model,
83
- punctuate: this.#opts.punctuate,
84
- smart_format: this.#opts.smartFormat,
85
- no_delay: this.#opts.noDelay,
86
- interim_results: this.#opts.interimResults,
87
- encoding: 'linear16',
88
- vad_events: true,
89
- sample_rate: this.#opts.sampleRate,
90
- channels: this.#opts.numChannels,
91
- endpointing: this.#opts.endpointing || false,
92
- filler_words: this.#opts.fillerWords,
93
- keywords: this.#opts.keywords.map((x) => x.join(':')),
94
- profanity_filter: this.#opts.profanityFilter,
95
- language: this.#opts.language,
96
- };
97
- Object.entries(params).forEach(([k, v]) => {
98
- if (v !== undefined) {
99
- if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {
100
- streamURL.searchParams.append(k, encodeURIComponent(v));
101
- }
102
- else {
103
- v.forEach((x) => streamURL.searchParams.append('keywords', encodeURIComponent(x)));
104
- }
105
- }
106
- });
107
- ws = new WebSocket(streamURL, {
108
- headers: { Authorization: `Token ${this.#opts.apiKey}` },
109
- });
110
- try {
111
- await new Promise((resolve, reject) => {
112
- ws.on('open', resolve);
113
- ws.on('error', (error) => reject(error));
114
- ws.on('close', (code) => reject(`WebSocket returned ${code}`));
115
- });
116
- await this.#runWS(ws);
117
- }
118
- catch (e) {
119
- if (retries >= maxRetry) {
120
- throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);
121
- }
122
- const delay = Math.min(retries * 5, 10);
123
- retries++;
124
- this.#logger.warn(`failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`);
125
- await new Promise((resolve) => setTimeout(resolve, delay * 1000));
126
- }
62
+ class SpeechStream extends stt.SpeechStream {
63
+ #opts;
64
+ #audioEnergyFilter;
65
+ #logger = log();
66
+ #speaking = false;
67
+ label = "deepgram.SpeechStream";
68
+ constructor(stt2, opts) {
69
+ super(stt2);
70
+ this.#opts = opts;
71
+ this.closed = false;
72
+ this.#audioEnergyFilter = new AudioEnergyFilter();
73
+ this.#run();
74
+ }
75
+ async #run(maxRetry = 32) {
76
+ let retries = 0;
77
+ let ws;
78
+ while (!this.input.closed) {
79
+ const streamURL = new URL(API_BASE_URL_V1);
80
+ const params = {
81
+ model: this.#opts.model,
82
+ punctuate: this.#opts.punctuate,
83
+ smart_format: this.#opts.smartFormat,
84
+ no_delay: this.#opts.noDelay,
85
+ interim_results: this.#opts.interimResults,
86
+ encoding: "linear16",
87
+ vad_events: true,
88
+ sample_rate: this.#opts.sampleRate,
89
+ channels: this.#opts.numChannels,
90
+ endpointing: this.#opts.endpointing || false,
91
+ filler_words: this.#opts.fillerWords,
92
+ keywords: this.#opts.keywords.map((x) => x.join(":")),
93
+ profanity_filter: this.#opts.profanityFilter,
94
+ language: this.#opts.language
95
+ };
96
+ Object.entries(params).forEach(([k, v]) => {
97
+ if (v !== void 0) {
98
+ if (typeof v === "string" || typeof v === "number" || typeof v === "boolean") {
99
+ streamURL.searchParams.append(k, encodeURIComponent(v));
100
+ } else {
101
+ v.forEach((x) => streamURL.searchParams.append("keywords", encodeURIComponent(x)));
102
+ }
127
103
  }
128
- this.closed = true;
104
+ });
105
+ ws = new WebSocket(streamURL, {
106
+ headers: { Authorization: `Token ${this.#opts.apiKey}` }
107
+ });
108
+ try {
109
+ await new Promise((resolve, reject) => {
110
+ ws.on("open", resolve);
111
+ ws.on("error", (error) => reject(error));
112
+ ws.on("close", (code) => reject(`WebSocket returned ${code}`));
113
+ });
114
+ await this.#runWS(ws);
115
+ } catch (e) {
116
+ if (retries >= maxRetry) {
117
+ throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);
118
+ }
119
+ const delay = Math.min(retries * 5, 10);
120
+ retries++;
121
+ this.#logger.warn(
122
+ `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`
123
+ );
124
+ await new Promise((resolve) => setTimeout(resolve, delay * 1e3));
125
+ }
129
126
  }
130
- async #runWS(ws) {
131
- let closing = false;
132
- const keepalive = setInterval(() => {
133
- try {
134
- ws.send(JSON.stringify({ type: 'KeepAlive' }));
135
- }
136
- catch {
137
- clearInterval(keepalive);
138
- return;
139
- }
140
- }, 5000);
141
- const sendTask = async () => {
142
- const samples100Ms = Math.floor(this.#opts.sampleRate / 10);
143
- const stream = new AudioByteStream(this.#opts.sampleRate, this.#opts.numChannels, samples100Ms);
144
- for await (const data of this.input) {
145
- let frames;
146
- if (data === _a.FLUSH_SENTINEL) {
147
- frames = stream.flush();
148
- }
149
- else if (data.sampleRate === this.#opts.sampleRate ||
150
- data.channels === this.#opts.numChannels) {
151
- frames = stream.write(data.data.buffer);
152
- }
153
- else {
154
- throw new Error(`sample rate or channel count of frame does not match`);
155
- }
156
- for await (const frame of frames) {
157
- if (this.#audioEnergyFilter.pushFrame(frame)) {
158
- ws.send(frame.data.buffer);
159
- }
160
- }
161
- }
162
- closing = true;
163
- ws.send(JSON.stringify({ type: 'CloseStream' }));
164
- };
165
- const listenTask = async () => {
166
- new Promise((_, reject) => ws.once('close', (code, reason) => {
167
- if (!closing) {
168
- this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
169
- reject();
170
- }
171
- }));
172
- while (!this.closed) {
173
- try {
174
- await new Promise((resolve) => {
175
- ws.once('message', (data) => resolve(data));
176
- }).then((msg) => {
177
- const json = JSON.parse(msg.toString());
178
- switch (json['type']) {
179
- case 'SpeechStarted': {
180
- // This is a normal case. Deepgram's SpeechStarted events
181
- // are not correlated with speech_final or utterance end.
182
- // It's possible that we receive two in a row without an endpoint
183
- // It's also possible we receive a transcript without a SpeechStarted event.
184
- if (this.#speaking)
185
- return;
186
- this.#speaking = true;
187
- this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });
188
- break;
189
- }
190
- // see this page:
191
- // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final
192
- // for more information about the different types of events
193
- case 'Results': {
194
- const isFinal = json['is_final'];
195
- const isEndpoint = json['speech_final'];
196
- const alternatives = liveTranscriptionToSpeechData(this.#opts.language, json);
197
- // If, for some reason, we didn't get a SpeechStarted event but we got
198
- // a transcript with text, we should start speaking. It's rare but has
199
- // been observed.
200
- if (alternatives[0] && alternatives[0].text) {
201
- if (!this.#speaking) {
202
- this.#speaking = true;
203
- this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });
204
- }
205
- if (isFinal) {
206
- this.queue.put({
207
- type: stt.SpeechEventType.FINAL_TRANSCRIPT,
208
- alternatives: [alternatives[0], ...alternatives.splice(0)],
209
- });
210
- }
211
- else {
212
- this.queue.put({
213
- type: stt.SpeechEventType.INTERIM_TRANSCRIPT,
214
- alternatives: [alternatives[0], ...alternatives.splice(0)],
215
- });
216
- }
217
- }
218
- // if we receive an endpoint, only end the speech if
219
- // we either had a SpeechStarted event or we have a seen
220
- // a non-empty transcript (deepgram doesn't have a SpeechEnded event)
221
- if (isEndpoint && this.#speaking) {
222
- this.#speaking = false;
223
- this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });
224
- }
225
- break;
226
- }
227
- case 'Metadata': {
228
- break;
229
- }
230
- default: {
231
- this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');
232
- break;
233
- }
234
- }
127
+ this.closed = true;
128
+ }
129
+ async #runWS(ws) {
130
+ let closing = false;
131
+ const keepalive = setInterval(() => {
132
+ try {
133
+ ws.send(JSON.stringify({ type: "KeepAlive" }));
134
+ } catch {
135
+ clearInterval(keepalive);
136
+ return;
137
+ }
138
+ }, 5e3);
139
+ const sendTask = async () => {
140
+ const samples100Ms = Math.floor(this.#opts.sampleRate / 10);
141
+ const stream = new AudioByteStream(
142
+ this.#opts.sampleRate,
143
+ this.#opts.numChannels,
144
+ samples100Ms
145
+ );
146
+ for await (const data of this.input) {
147
+ let frames;
148
+ if (data === SpeechStream.FLUSH_SENTINEL) {
149
+ frames = stream.flush();
150
+ } else if (data.sampleRate === this.#opts.sampleRate || data.channels === this.#opts.numChannels) {
151
+ frames = stream.write(data.data.buffer);
152
+ } else {
153
+ throw new Error(`sample rate or channel count of frame does not match`);
154
+ }
155
+ for await (const frame of frames) {
156
+ if (this.#audioEnergyFilter.pushFrame(frame)) {
157
+ ws.send(frame.data.buffer);
158
+ }
159
+ }
160
+ }
161
+ closing = true;
162
+ ws.send(JSON.stringify({ type: "CloseStream" }));
163
+ };
164
+ 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
+ while (!this.closed) {
174
+ try {
175
+ await new Promise((resolve) => {
176
+ ws.once("message", (data) => resolve(data));
177
+ }).then((msg) => {
178
+ const json = JSON.parse(msg.toString());
179
+ switch (json["type"]) {
180
+ case "SpeechStarted": {
181
+ if (this.#speaking) return;
182
+ this.#speaking = true;
183
+ this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });
184
+ break;
185
+ }
186
+ // see this page:
187
+ // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final
188
+ // for more information about the different types of events
189
+ case "Results": {
190
+ const isFinal = json["is_final"];
191
+ const isEndpoint = json["speech_final"];
192
+ const alternatives = liveTranscriptionToSpeechData(this.#opts.language, json);
193
+ if (alternatives[0] && alternatives[0].text) {
194
+ if (!this.#speaking) {
195
+ this.#speaking = true;
196
+ this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });
197
+ }
198
+ if (isFinal) {
199
+ this.queue.put({
200
+ type: stt.SpeechEventType.FINAL_TRANSCRIPT,
201
+ alternatives: [alternatives[0], ...alternatives.slice(1)]
235
202
  });
203
+ } else {
204
+ this.queue.put({
205
+ type: stt.SpeechEventType.INTERIM_TRANSCRIPT,
206
+ alternatives: [alternatives[0], ...alternatives.slice(1)]
207
+ });
208
+ }
236
209
  }
237
- catch (error) {
238
- this.#logger.child({ error }).warn('unrecoverable error, exiting');
239
- break;
210
+ if (isEndpoint && this.#speaking) {
211
+ this.#speaking = false;
212
+ this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });
240
213
  }
214
+ break;
215
+ }
216
+ case "Metadata": {
217
+ break;
218
+ }
219
+ default: {
220
+ this.#logger.child({ msg: json }).warn("received unexpected message from Deepgram");
221
+ break;
222
+ }
241
223
  }
242
- };
243
- await Promise.all([sendTask(), listenTask()]);
244
- clearInterval(keepalive);
245
- }
224
+ });
225
+ } catch (error) {
226
+ this.#logger.child({ error }).warn("unrecoverable error, exiting");
227
+ break;
228
+ }
229
+ }
230
+ };
231
+ await Promise.all([sendTask(), listenTask()]);
232
+ clearInterval(keepalive);
233
+ }
246
234
  }
247
- _a = SpeechStream;
248
235
  const liveTranscriptionToSpeechData = (language, data) => {
249
- const alts = data['channel']['alternatives'];
250
- return alts.map((alt) => ({
251
- language,
252
- startTime: alt['words'].length ? alt['words'][0]['start'] : 0,
253
- endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,
254
- confidence: alt['confidence'],
255
- text: alt['transcript'],
256
- }));
236
+ const alts = data["channel"]["alternatives"];
237
+ return alts.map((alt) => ({
238
+ language,
239
+ startTime: alt["words"].length ? alt["words"][0]["start"] : 0,
240
+ endTime: alt["words"].length ? alt["words"][alt["words"].length - 1]["end"] : 0,
241
+ confidence: alt["confidence"],
242
+ text: alt["transcript"]
243
+ }));
244
+ };
245
+ export {
246
+ STT,
247
+ SpeechStream
257
248
  };
258
249
  //# sourceMappingURL=stt.js.map
package/dist/stt.js.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"file":"stt.js","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":";AAAA,6CAA6C;AAC7C,EAAE;AACF,sCAAsC;AACtC,OAAO,EAAoB,eAAe,EAAE,iBAAiB,EAAE,GAAG,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAEjG,OAAO,EAAgB,SAAS,EAAE,MAAM,IAAI,CAAC;AAG7C,MAAM,eAAe,GAAG,kCAAkC,CAAC;AAmB3D,MAAM,iBAAiB,GAAe;IACpC,MAAM,EAAE,OAAO,CAAC,GAAG,CAAC,gBAAgB;IACpC,QAAQ,EAAE,OAAO;IACjB,cAAc,EAAE,KAAK;IACrB,cAAc,EAAE,IAAI;IACpB,SAAS,EAAE,IAAI;IACf,KAAK,EAAE,gBAAgB;IACvB,WAAW,EAAE,IAAI;IACjB,OAAO,EAAE,IAAI;IACb,WAAW,EAAE,EAAE;IACf,WAAW,EAAE,KAAK;IAClB,UAAU,EAAE,KAAK;IACjB,WAAW,EAAE,CAAC;IACd,QAAQ,EAAE,EAAE;IACZ,eAAe,EAAE,KAAK;CACvB,CAAC;AAEF,MAAM,OAAO,GAAI,SAAQ,GAAG,CAAC,GAAG;IAC9B,KAAK,CAAa;IAClB,OAAO,GAAG,GAAG,EAAE,CAAC;IAEhB,YAAY,OAA4B,iBAAiB;QACvD,KAAK,CAAC;YACJ,SAAS,EAAE,IAAI;YACf,cAAc,EAAE,IAAI,CAAC,cAAc,IAAI,iBAAiB,CAAC,cAAc;SACxE,CAAC,CAAC;QACH,IAAI,IAAI,CAAC,MAAM,KAAK,SAAS,IAAI,iBAAiB,CAAC,MAAM,KAAK,SAAS,EAAE,CAAC;YACxE,MAAM,IAAI,KAAK,CACb,8EAA8E,CAC/E,CAAC;QACJ,CAAC;QAED,IAAI,CAAC,KAAK,GAAG,EAAE,GAAG,iBAAiB,EAAE,GAAG,IAAI,EAAE,CAAC;QAE/C,IAAI,IAAI,CAAC,KAAK,CAAC,cAAc,EAAE,CAAC;YAC9B,IAAI,CAAC,KAAK,CAAC,QAAQ,GAAG,SAAS,CAAC;QAClC,CAAC;aAAM,IACL,IAAI,CAAC,KAAK,CAAC,QAAQ;YACnB,CAAC,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC,QAAQ,CAAC,IAAI,CAAC,KAAK,CAAC,QAAQ,CAAC;YAC9C;gBACE,gBAAgB;gBAChB,kBAAkB;gBAClB,gBAAgB;gBAChB,yBAAyB;gBACzB,kBAAkB;gBAClB,cAAc;gBACd,gBAAgB;gBAChB,kBAAkB;gBAClB,mBAAmB;aACpB,CAAC,QAAQ,CAAC,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,EAC5B,CAAC;YACD,IAAI,CAAC,OAAO,CAAC,IAAI,CACf,GAAG,IAAI,CAAC,KAAK,CAAC,KAAK,8BAA8B,IAAI,CAAC,KAAK,CAAC,QAAQ,kCAAkC,CACvG,CAAC;YACF,IAAI,CAAC,KAAK,CAAC,KAAK,GAAG,gBAAgB,CAAC;QACtC,CAAC;IACH,CAAC;IAED,6DAA6D;IAC7D,KAAK,CAAC,SAAS,CAAC,CAAc;QAC5B,MAAM,IAAI,KAAK,CAAC,4CAA4C,CAAC,CAAC;IAChE,CAAC;IAED,MAAM;QACJ,OAAO,IAAI,YAAY,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;IACtC,CAAC;CACF;AAED,MAAM,OAAO,YAAa,SAAQ,GAAG,CAAC,YAAY;IAChD,KAAK,CAAa;IAClB,kBAAkB,CAAoB;IACtC,OAAO,GAAG,GAAG,EAAE,CAAC;IAChB,SAAS,GAAG,KAAK,CAAC;IAElB,YAAY,IAAgB;QAC1B,KAAK,EAAE,CAAC;QACR,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC;QAClB,IAAI,CAAC,MAAM,GAAG,KAAK,CAAC;QACpB,IAAI,CAAC,kBAAkB,GAAG,IAAI,iBAAiB,EAAE,CAAC;QAElD,IAAI,CAAC,IAAI,EAAE,CAAC;IACd,CAAC;IAED,KAAK,CAAC,IAAI,CAAC,QAAQ,GAAG,EAAE;QACtB,IAAI,OAAO,GAAG,CAAC,CAAC;QAChB,IAAI,EAAa,CAAC;QAClB,OAAO,CAAC,IAAI,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC;YAC1B,MAAM,SAAS,GAAG,IAAI,GAAG,CAAC,eAAe,CAAC,CAAC;YAC3C,MAAM,MAAM,GAAG;gBACb,KAAK,EAAE,IAAI,CAAC,KAAK,CAAC,KAAK;gBACvB,SAAS,EAAE,IAAI,CAAC,KAAK,CAAC,SAAS;gBAC/B,YAAY,EAAE,IAAI,CAAC,KAAK,CAAC,WAAW;gBACpC,QAAQ,EAAE,IAAI,CAAC,KAAK,CAAC,OAAO;gBAC5B,eAAe,EAAE,IAAI,CAAC,KAAK,CAAC,cAAc;gBAC1C,QAAQ,EAAE,UAAU;gBACpB,UAAU,EAAE,IAAI;gBAChB,WAAW,EAAE,IAAI,CAAC,KAAK,CAAC,UAAU;gBAClC,QAAQ,EAAE,IAAI,CAAC,KAAK,CAAC,WAAW;gBAChC,WAAW,EAAE,IAAI,CAAC,KAAK,CAAC,WAAW,IAAI,KAAK;gBAC5C,YAAY,EAAE,IAAI,CAAC,KAAK,CAAC,WAAW;gBACpC,QAAQ,EAAE,IAAI,CAAC,KAAK,CAAC,QAAQ,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;gBACrD,gBAAgB,EAAE,IAAI,CAAC,KAAK,CAAC,eAAe;gBAC5C,QAAQ,EAAE,IAAI,CAAC,KAAK,CAAC,QAAQ;aAC9B,CAAC;YACF,MAAM,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,EAAE,EAAE;gBACxC,IAAI,CAAC,KAAK,SAAS,EAAE,CAAC;oBACpB,IAAI,OAAO,CAAC,KAAK,QAAQ,IAAI,OAAO,CAAC,KAAK,QAAQ,IAAI,OAAO,CAAC,KAAK,SAAS,EAAE,CAAC;wBAC7E,SAAS,CAAC,YAAY,CAAC,MAAM,CAAC,CAAC,EAAE,kBAAkB,CAAC,CAAC,CAAC,CAAC,CAAC;oBAC1D,CAAC;yBAAM,CAAC;wBACN,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,SAAS,CAAC,YAAY,CAAC,MAAM,CAAC,UAAU,EAAE,kBAAkB,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;oBACrF,CAAC;gBACH,CAAC;YACH,CAAC,CAAC,CAAC;YAEH,EAAE,GAAG,IAAI,SAAS,CAAC,SAAS,EAAE;gBAC5B,OAAO,EAAE,EAAE,aAAa,EAAE,SAAS,IAAI,CAAC,KAAK,CAAC,MAAM,EAAE,EAAE;aACzD,CAAC,CAAC;YAEH,IAAI,CAAC;gBACH,MAAM,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;oBACpC,EAAE,CAAC,EAAE,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;oBACvB,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,KAAK,EAAE,EAAE,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC;oBACzC,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,IAAI,EAAE,EAAE,CAAC,MAAM,CAAC,sBAAsB,IAAI,EAAE,CAAC,CAAC,CAAC;gBACjE,CAAC,CAAC,CAAC;gBAEH,MAAM,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC;YACxB,CAAC;YAAC,OAAO,CAAC,EAAE,CAAC;gBACX,IAAI,OAAO,IAAI,QAAQ,EAAE,CAAC;oBACxB,MAAM,IAAI,KAAK,CAAC,uCAAuC,OAAO,cAAc,CAAC,EAAE,CAAC,CAAC;gBACnF,CAAC;gBAED,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,OAAO,GAAG,CAAC,EAAE,EAAE,CAAC,CAAC;gBACxC,OAAO,EAAE,CAAC;gBAEV,IAAI,CAAC,OAAO,CAAC,IAAI,CACf,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ,GAAG,CAC7F,CAAC;gBACF,MAAM,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,UAAU,CAAC,OAAO,EAAE,KAAK,GAAG,IAAI,CAAC,CAAC,CAAC;YACpE,CAAC;QACH,CAAC;QAED,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC;IACrB,CAAC;IAED,KAAK,CAAC,MAAM,CAAC,EAAa;QACxB,IAAI,OAAO,GAAG,KAAK,CAAC;QAEpB,MAAM,SAAS,GAAG,WAAW,CAAC,GAAG,EAAE;YACjC,IAAI,CAAC;gBACH,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,IAAI,EAAE,WAAW,EAAE,CAAC,CAAC,CAAC;YACjD,CAAC;YAAC,MAAM,CAAC;gBACP,aAAa,CAAC,SAAS,CAAC,CAAC;gBACzB,OAAO;YACT,CAAC;QACH,CAAC,EAAE,IAAI,CAAC,CAAC;QAET,MAAM,QAAQ,GAAG,KAAK,IAAI,EAAE;YAC1B,MAAM,YAAY,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,UAAU,GAAG,EAAE,CAAC,CAAC;YAC5D,MAAM,MAAM,GAAG,IAAI,eAAe,CAChC,IAAI,CAAC,KAAK,CAAC,UAAU,EACrB,IAAI,CAAC,KAAK,CAAC,WAAW,EACtB,YAAY,CACb,CAAC;YAEF,IAAI,KAAK,EAAE,MAAM,IAAI,IAAI,IAAI,CAAC,KAAK,EAAE,CAAC;gBACpC,IAAI,MAAoB,CAAC;gBACzB,IAAI,IAAI,KAAK,EAAY,CAAC,cAAc,EAAE,CAAC;oBACzC,MAAM,GAAG,MAAM,CAAC,KAAK,EAAE,CAAC;gBAC1B,CAAC;qBAAM,IACL,IAAI,CAAC,UAAU,KAAK,IAAI,CAAC,KAAK,CAAC,UAAU;oBACzC,IAAI,CAAC,QAAQ,KAAK,IAAI,CAAC,KAAK,CAAC,WAAW,EACxC,CAAC;oBACD,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;gBAC1C,CAAC;qBAAM,CAAC;oBACN,MAAM,IAAI,KAAK,CAAC,sDAAsD,CAAC,CAAC;gBAC1E,CAAC;gBAED,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;oBACjC,IAAI,IAAI,CAAC,kBAAkB,CAAC,SAAS,CAAC,KAAK,CAAC,EAAE,CAAC;wBAC7C,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;oBAC7B,CAAC;gBACH,CAAC;YACH,CAAC;YAED,OAAO,GAAG,IAAI,CAAC;YACf,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,IAAI,EAAE,aAAa,EAAE,CAAC,CAAC,CAAC;QACnD,CAAC,CAAC;QAEF,MAAM,UAAU,GAAG,KAAK,IAAI,EAAE;YAC5B,IAAI,OAAO,CAAO,CAAC,CAAC,EAAE,MAAM,EAAE,EAAE,CAC9B,EAAE,CAAC,IAAI,CAAC,OAAO,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,EAAE;gBAChC,IAAI,CAAC,OAAO,EAAE,CAAC;oBACb,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,8BAA8B,IAAI,KAAK,MAAM,EAAE,CAAC,CAAC;oBACpE,MAAM,EAAE,CAAC;gBACX,CAAC;YACH,CAAC,CAAC,CACH,CAAC;YAEF,OAAO,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC;gBACpB,IAAI,CAAC;oBACH,MAAM,IAAI,OAAO,CAAU,CAAC,OAAO,EAAE,EAAE;wBACrC,EAAE,CAAC,IAAI,CAAC,SAAS,EAAE,CAAC,IAAI,EAAE,EAAE,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC;oBAC9C,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,GAAG,EAAE,EAAE;wBACd,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,QAAQ,EAAE,CAAC,CAAC;wBACxC,QAAQ,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC;4BACrB,KAAK,eAAe,CAAC,CAAC,CAAC;gCACrB,yDAAyD;gCACzD,yDAAyD;gCACzD,iEAAiE;gCACjE,4EAA4E;gCAC5E,IAAI,IAAI,CAAC,SAAS;oCAAE,OAAO;gCAC3B,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC;gCACtB,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,EAAE,IAAI,EAAE,GAAG,CAAC,eAAe,CAAC,eAAe,EAAE,CAAC,CAAC;gCAC9D,MAAM;4BACR,CAAC;4BACD,iBAAiB;4BACjB,6GAA6G;4BAC7G,2DAA2D;4BAC3D,KAAK,SAAS,CAAC,CAAC,CAAC;gCACf,MAAM,OAAO,GAAG,IAAI,CAAC,UAAU,CAAC,CAAC;gCACjC,MAAM,UAAU,GAAG,IAAI,CAAC,cAAc,CAAC,CAAC;gCAExC,MAAM,YAAY,GAAG,6BAA6B,CAAC,IAAI,CAAC,KAAK,CAAC,QAAS,EAAE,IAAI,CAAC,CAAC;gCAE/E,sEAAsE;gCACtE,sEAAsE;gCACtE,iBAAiB;gCACjB,IAAI,YAAY,CAAC,CAAC,CAAC,IAAI,YAAY,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;oCAC5C,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,CAAC;wCACpB,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC;wCACtB,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,EAAE,IAAI,EAAE,GAAG,CAAC,eAAe,CAAC,eAAe,EAAE,CAAC,CAAC;oCAChE,CAAC;oCAED,IAAI,OAAO,EAAE,CAAC;wCACZ,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC;4CACb,IAAI,EAAE,GAAG,CAAC,eAAe,CAAC,gBAAgB;4CAC1C,YAAY,EAAE,CAAC,YAAY,CAAC,CAAC,CAAC,EAAE,GAAG,YAAY,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC;yCAC3D,CAAC,CAAC;oCACL,CAAC;yCAAM,CAAC;wCACN,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC;4CACb,IAAI,EAAE,GAAG,CAAC,eAAe,CAAC,kBAAkB;4CAC5C,YAAY,EAAE,CAAC,YAAY,CAAC,CAAC,CAAC,EAAE,GAAG,YAAY,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC;yCAC3D,CAAC,CAAC;oCACL,CAAC;gCACH,CAAC;gCAED,oDAAoD;gCACpD,wDAAwD;gCACxD,qEAAqE;gCACrE,IAAI,UAAU,IAAI,IAAI,CAAC,SAAS,EAAE,CAAC;oCACjC,IAAI,CAAC,SAAS,GAAG,KAAK,CAAC;oCACvB,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,EAAE,IAAI,EAAE,GAAG,CAAC,eAAe,CAAC,aAAa,EAAE,CAAC,CAAC;gCAC9D,CAAC;gCAED,MAAM;4BACR,CAAC;4BACD,KAAK,UAAU,CAAC,CAAC,CAAC;gCAChB,MAAM;4BACR,CAAC;4BACD,OAAO,CAAC,CAAC,CAAC;gCACR,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,EAAE,GAAG,EAAE,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,2CAA2C,CAAC,CAAC;gCACpF,MAAM;4BACR,CAAC;wBACH,CAAC;oBACH,CAAC,CAAC,CAAC;gBACL,CAAC;gBAAC,OAAO,KAAK,EAAE,CAAC;oBACf,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC,IAAI,CAAC,8BAA8B,CAAC,CAAC;oBACnE,MAAM;gBACR,CAAC;YACH,CAAC;QACH,CAAC,CAAC;QAEF,MAAM,OAAO,CAAC,GAAG,CAAC,CAAC,QAAQ,EAAE,EAAE,UAAU,EAAE,CAAC,CAAC,CAAC;QAC9C,aAAa,CAAC,SAAS,CAAC,CAAC;IAC3B,CAAC;CACF;;AAED,MAAM,6BAA6B,GAAG,CACpC,QAA+B,EAC/B,IAA2B,EACT,EAAE;IACpB,MAAM,IAAI,GAAU,IAAI,CAAC,SAAS,CAAC,CAAC,cAAc,CAAC,CAAC;IAEpD,OAAO,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,CAAC;QACxB,QAAQ;QACR,SAAS,EAAE,GAAG,CAAC,OAAO,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC;QAC7D,OAAO,EAAE,GAAG,CAAC,OAAO,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC;QAC/E,UAAU,EAAE,GAAG,CAAC,YAAY,CAAC;QAC7B,IAAI,EAAE,GAAG,CAAC,YAAY,CAAC;KACxB,CAAC,CAAC,CAAC;AACN,CAAC,CAAC"}
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"]}
@@ -0,0 +1,11 @@
1
+ "use strict";
2
+ var import_agents = require("@livekit/agents");
3
+ var import_agents_plugin_silero = require("@livekit/agents-plugin-silero");
4
+ var import_agents_plugins_test = require("@livekit/agents-plugins-test");
5
+ var import_vitest = require("vitest");
6
+ var import_stt = require("./stt.cjs");
7
+ (0, import_vitest.describe)("Deepgram", async () => {
8
+ (0, import_agents.initializeLogger)({ pretty: false });
9
+ await (0, import_agents_plugins_test.stt)(new import_stt.STT(), await import_agents_plugin_silero.VAD.load(), { nonStreaming: false });
10
+ });
11
+ //# sourceMappingURL=stt.test.cjs.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { initializeLogger } from '@livekit/agents';\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { describe } from 'vitest';\nimport { STT } from './stt.js';\n\ndescribe('Deepgram', async () => {\n initializeLogger({ pretty: false });\n await stt(new STT(), await VAD.load(), { nonStreaming: false });\n});\n"],"mappings":";AAGA,oBAAiC;AACjC,kCAAoB;AACpB,iCAAoB;AACpB,oBAAyB;AACzB,iBAAoB;AAAA,IAEpB,wBAAS,YAAY,YAAY;AAC/B,sCAAiB,EAAE,QAAQ,MAAM,CAAC;AAClC,YAAM,gCAAI,IAAI,eAAI,GAAG,MAAM,gCAAI,KAAK,GAAG,EAAE,cAAc,MAAM,CAAC;AAChE,CAAC;","names":[]}
@@ -0,0 +1,2 @@
1
+ export {};
2
+ //# sourceMappingURL=stt.test.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"stt.test.d.ts","sourceRoot":"","sources":["../src/stt.test.ts"],"names":[],"mappings":""}
@@ -0,0 +1,10 @@
1
+ import { initializeLogger } from "@livekit/agents";
2
+ import { VAD } from "@livekit/agents-plugin-silero";
3
+ import { stt } from "@livekit/agents-plugins-test";
4
+ import { describe } from "vitest";
5
+ import { STT } from "./stt.js";
6
+ describe("Deepgram", async () => {
7
+ initializeLogger({ pretty: false });
8
+ await stt(new STT(), await VAD.load(), { nonStreaming: false });
9
+ });
10
+ //# sourceMappingURL=stt.test.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { initializeLogger } from '@livekit/agents';\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { describe } from 'vitest';\nimport { STT } from './stt.js';\n\ndescribe('Deepgram', async () => {\n initializeLogger({ pretty: false });\n await stt(new STT(), await VAD.load(), { nonStreaming: false });\n});\n"],"mappings":"AAGA,SAAS,wBAAwB;AACjC,SAAS,WAAW;AACpB,SAAS,WAAW;AACpB,SAAS,gBAAgB;AACzB,SAAS,WAAW;AAEpB,SAAS,YAAY,YAAY;AAC/B,mBAAiB,EAAE,QAAQ,MAAM,CAAC;AAClC,QAAM,IAAI,IAAI,IAAI,GAAG,MAAM,IAAI,KAAK,GAAG,EAAE,cAAc,MAAM,CAAC;AAChE,CAAC;","names":[]}
package/package.json CHANGED
@@ -1,33 +1,45 @@
1
1
  {
2
2
  "name": "@livekit/agents-plugin-deepgram",
3
- "version": "0.4.6",
3
+ "version": "0.5.1",
4
4
  "description": "Deepgram plugin for LiveKit Agents for Node.js",
5
5
  "main": "dist/index.js",
6
+ "require": "dist/index.cjs",
6
7
  "types": "dist/index.d.ts",
8
+ "exports": {
9
+ ".": {
10
+ "types": "./dist/index.d.ts",
11
+ "import": "./dist/index.js",
12
+ "require": "./dist/index.cjs"
13
+ }
14
+ },
7
15
  "author": "LiveKit",
8
16
  "type": "module",
9
17
  "repository": "git@github.com:livekit/agents-js.git",
10
18
  "license": "Apache-2.0",
11
19
  "files": [
12
20
  "dist",
13
- "src"
21
+ "src",
22
+ "README.md"
14
23
  ],
15
24
  "devDependencies": {
25
+ "@livekit/agents": "^x",
26
+ "@livekit/agents-plugin-silero": "^x",
27
+ "@livekit/agents-plugins-test": "^x",
28
+ "@livekit/rtc-node": "^0.12.1",
16
29
  "@microsoft/api-extractor": "^7.35.0",
17
- "@livekit/rtc-node": "^0.11.1",
18
30
  "@types/ws": "^8.5.10",
19
- "typescript": "^5.0.0",
20
- "@livekit/agents": "^0.4.6"
31
+ "tsup": "^8.3.5",
32
+ "typescript": "^5.0.0"
21
33
  },
22
34
  "dependencies": {
23
35
  "ws": "^8.16.0"
24
36
  },
25
37
  "peerDependencies": {
26
- "@livekit/rtc-node": "^0.11.1",
27
- "@livekit/agents": "^0.4.6"
38
+ "@livekit/rtc-node": "^0.12.1",
39
+ "@livekit/agents": "^0.6.0x"
28
40
  },
29
41
  "scripts": {
30
- "build": "tsc",
42
+ "build": "tsup --onSuccess \"tsc --declaration --emitDeclarationOnly\"",
31
43
  "clean": "rm -rf dist",
32
44
  "clean:build": "pnpm clean && pnpm build",
33
45
  "lint": "eslint -f unix \"src/**/*.{ts,js}\"",
@@ -0,0 +1,13 @@
1
+ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.
2
+ //
3
+ // SPDX-License-Identifier: Apache-2.0
4
+ import { initializeLogger } from '@livekit/agents';
5
+ import { VAD } from '@livekit/agents-plugin-silero';
6
+ import { stt } from '@livekit/agents-plugins-test';
7
+ import { describe } from 'vitest';
8
+ import { STT } from './stt.js';
9
+
10
+ describe('Deepgram', async () => {
11
+ initializeLogger({ pretty: false });
12
+ await stt(new STT(), await VAD.load(), { nonStreaming: false });
13
+ });
package/src/stt.ts CHANGED
@@ -45,11 +45,12 @@ const defaultSTTOptions: STTOptions = {
45
45
  export class STT extends stt.STT {
46
46
  #opts: STTOptions;
47
47
  #logger = log();
48
+ label = 'deepgram.STT';
48
49
 
49
50
  constructor(opts: Partial<STTOptions> = defaultSTTOptions) {
50
51
  super({
51
52
  streaming: true,
52
- interimResults: opts.interimResults || defaultSTTOptions.interimResults,
53
+ interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,
53
54
  });
54
55
  if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {
55
56
  throw new Error(
@@ -84,12 +85,12 @@ export class STT extends stt.STT {
84
85
  }
85
86
 
86
87
  // eslint-disable-next-line @typescript-eslint/no-unused-vars
87
- async recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {
88
+ async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {
88
89
  throw new Error('Recognize is not supported on Deepgram STT');
89
90
  }
90
91
 
91
92
  stream(): stt.SpeechStream {
92
- return new SpeechStream(this.#opts);
93
+ return new SpeechStream(this, this.#opts);
93
94
  }
94
95
  }
95
96
 
@@ -98,9 +99,10 @@ export class SpeechStream extends stt.SpeechStream {
98
99
  #audioEnergyFilter: AudioEnergyFilter;
99
100
  #logger = log();
100
101
  #speaking = false;
102
+ label = 'deepgram.SpeechStream';
101
103
 
102
- constructor(opts: STTOptions) {
103
- super();
104
+ constructor(stt: STT, opts: STTOptions) {
105
+ super(stt);
104
106
  this.#opts = opts;
105
107
  this.closed = false;
106
108
  this.#audioEnergyFilter = new AudioEnergyFilter();
@@ -261,12 +263,12 @@ export class SpeechStream extends stt.SpeechStream {
261
263
  if (isFinal) {
262
264
  this.queue.put({
263
265
  type: stt.SpeechEventType.FINAL_TRANSCRIPT,
264
- alternatives: [alternatives[0], ...alternatives.splice(0)],
266
+ alternatives: [alternatives[0], ...alternatives.slice(1)],
265
267
  });
266
268
  } else {
267
269
  this.queue.put({
268
270
  type: stt.SpeechEventType.INTERIM_TRANSCRIPT,
269
- alternatives: [alternatives[0], ...alternatives.splice(0)],
271
+ alternatives: [alternatives[0], ...alternatives.slice(1)],
270
272
  });
271
273
  }
272
274
  }