@livekit/agents-plugin-deepgram 1.0.11 → 1.0.13
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/index.cjs +3 -1
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.cts +1 -0
- package/dist/index.d.ts +1 -0
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +1 -0
- package/dist/index.js.map +1 -1
- package/dist/models.cjs.map +1 -1
- package/dist/models.d.cts +2 -0
- package/dist/models.d.ts +2 -0
- package/dist/models.d.ts.map +1 -1
- package/dist/tts.cjs +315 -0
- package/dist/tts.cjs.map +1 -0
- package/dist/tts.d.cts +36 -0
- package/dist/tts.d.ts +36 -0
- package/dist/tts.d.ts.map +1 -0
- package/dist/tts.js +289 -0
- package/dist/tts.js.map +1 -0
- package/dist/tts.test.cjs +9 -0
- package/dist/tts.test.cjs.map +1 -0
- package/dist/tts.test.d.cts +2 -0
- package/dist/tts.test.d.ts +2 -0
- package/dist/tts.test.d.ts.map +1 -0
- package/dist/tts.test.js +8 -0
- package/dist/tts.test.js.map +1 -0
- package/package.json +5 -5
- package/src/index.ts +1 -0
- package/src/models.ts +16 -0
- package/src/tts.test.ts +11 -0
- package/src/tts.ts +352 -0
package/dist/tts.js
ADDED
|
@@ -0,0 +1,289 @@
|
|
|
1
|
+
import { AudioByteStream, shortuuid, tokenize, tts } from "@livekit/agents";
|
|
2
|
+
import { request } from "node:https";
|
|
3
|
+
import { WebSocket } from "ws";
|
|
4
|
+
const AUTHORIZATION_HEADER = "Authorization";
|
|
5
|
+
const NUM_CHANNELS = 1;
|
|
6
|
+
const MIN_SENTENCE_LENGTH = 8;
|
|
7
|
+
const defaultTTSOptions = {
|
|
8
|
+
model: "aura-asteria-en",
|
|
9
|
+
encoding: "linear16",
|
|
10
|
+
sampleRate: 24e3,
|
|
11
|
+
apiKey: process.env.DEEPGRAM_API_KEY,
|
|
12
|
+
baseUrl: "https://api.deepgram.com",
|
|
13
|
+
capabilities: {
|
|
14
|
+
streaming: true
|
|
15
|
+
},
|
|
16
|
+
sentenceTokenizer: new tokenize.basic.SentenceTokenizer({
|
|
17
|
+
minSentenceLength: MIN_SENTENCE_LENGTH
|
|
18
|
+
})
|
|
19
|
+
};
|
|
20
|
+
class TTS extends tts.TTS {
|
|
21
|
+
opts;
|
|
22
|
+
label = "deepgram.TTS";
|
|
23
|
+
constructor(opts = {}) {
|
|
24
|
+
var _a;
|
|
25
|
+
super(opts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {
|
|
26
|
+
streaming: ((_a = opts.capabilities) == null ? void 0 : _a.streaming) ?? defaultTTSOptions.capabilities.streaming
|
|
27
|
+
});
|
|
28
|
+
this.opts = {
|
|
29
|
+
...defaultTTSOptions,
|
|
30
|
+
...opts
|
|
31
|
+
};
|
|
32
|
+
if (this.opts.apiKey === void 0) {
|
|
33
|
+
throw new Error(
|
|
34
|
+
"Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY"
|
|
35
|
+
);
|
|
36
|
+
}
|
|
37
|
+
}
|
|
38
|
+
synthesize(text) {
|
|
39
|
+
return new ChunkedStream(this, text, this.opts);
|
|
40
|
+
}
|
|
41
|
+
stream() {
|
|
42
|
+
return new SynthesizeStream(this, this.opts);
|
|
43
|
+
}
|
|
44
|
+
}
|
|
45
|
+
class ChunkedStream extends tts.ChunkedStream {
|
|
46
|
+
label = "deepgram.ChunkedStream";
|
|
47
|
+
opts;
|
|
48
|
+
text;
|
|
49
|
+
constructor(tts2, text, opts) {
|
|
50
|
+
super(text, tts2);
|
|
51
|
+
this.text = text;
|
|
52
|
+
this.opts = opts;
|
|
53
|
+
}
|
|
54
|
+
async run() {
|
|
55
|
+
const requestId = shortuuid();
|
|
56
|
+
const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);
|
|
57
|
+
const json = { text: this.text };
|
|
58
|
+
const url = new URL(`${this.opts.baseUrl}/v1/speak`);
|
|
59
|
+
url.searchParams.append("sample_rate", this.opts.sampleRate.toString());
|
|
60
|
+
url.searchParams.append("model", this.opts.model);
|
|
61
|
+
url.searchParams.append("encoding", this.opts.encoding);
|
|
62
|
+
await new Promise((resolve, reject) => {
|
|
63
|
+
const req = request(
|
|
64
|
+
{
|
|
65
|
+
hostname: url.hostname,
|
|
66
|
+
port: 443,
|
|
67
|
+
path: url.pathname + url.search,
|
|
68
|
+
method: "POST",
|
|
69
|
+
headers: {
|
|
70
|
+
[AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey}`,
|
|
71
|
+
"Content-Type": "application/json"
|
|
72
|
+
}
|
|
73
|
+
},
|
|
74
|
+
(res) => {
|
|
75
|
+
if (res.statusCode !== 200) {
|
|
76
|
+
reject(
|
|
77
|
+
new Error(`Deepgram TTS HTTP request failed: ${res.statusCode} ${res.statusMessage}`)
|
|
78
|
+
);
|
|
79
|
+
return;
|
|
80
|
+
}
|
|
81
|
+
res.on("data", (chunk) => {
|
|
82
|
+
for (const frame of bstream.write(chunk)) {
|
|
83
|
+
if (!this.queue.closed) {
|
|
84
|
+
this.queue.put({
|
|
85
|
+
requestId,
|
|
86
|
+
frame,
|
|
87
|
+
final: false,
|
|
88
|
+
segmentId: requestId
|
|
89
|
+
});
|
|
90
|
+
}
|
|
91
|
+
}
|
|
92
|
+
});
|
|
93
|
+
res.on("error", (err) => {
|
|
94
|
+
reject(err);
|
|
95
|
+
});
|
|
96
|
+
res.on("close", () => {
|
|
97
|
+
for (const frame of bstream.flush()) {
|
|
98
|
+
if (!this.queue.closed) {
|
|
99
|
+
this.queue.put({
|
|
100
|
+
requestId,
|
|
101
|
+
frame,
|
|
102
|
+
final: false,
|
|
103
|
+
segmentId: requestId
|
|
104
|
+
});
|
|
105
|
+
}
|
|
106
|
+
}
|
|
107
|
+
if (!this.queue.closed) {
|
|
108
|
+
this.queue.close();
|
|
109
|
+
}
|
|
110
|
+
resolve();
|
|
111
|
+
});
|
|
112
|
+
}
|
|
113
|
+
);
|
|
114
|
+
req.on("error", (err) => {
|
|
115
|
+
reject(err);
|
|
116
|
+
});
|
|
117
|
+
req.write(JSON.stringify(json));
|
|
118
|
+
req.end();
|
|
119
|
+
});
|
|
120
|
+
}
|
|
121
|
+
}
|
|
122
|
+
class SynthesizeStream extends tts.SynthesizeStream {
|
|
123
|
+
opts;
|
|
124
|
+
tokenizer;
|
|
125
|
+
label = "deepgram.SynthesizeStream";
|
|
126
|
+
static FLUSH_MSG = JSON.stringify({ type: "Flush" });
|
|
127
|
+
static CLOSE_MSG = JSON.stringify({ type: "Close" });
|
|
128
|
+
constructor(tts2, opts) {
|
|
129
|
+
super(tts2);
|
|
130
|
+
this.opts = opts;
|
|
131
|
+
this.tokenizer = opts.sentenceTokenizer.stream();
|
|
132
|
+
}
|
|
133
|
+
async closeWebSocket(ws) {
|
|
134
|
+
try {
|
|
135
|
+
if (ws.readyState === WebSocket.OPEN) {
|
|
136
|
+
ws.send(SynthesizeStream.FLUSH_MSG);
|
|
137
|
+
ws.send(SynthesizeStream.CLOSE_MSG);
|
|
138
|
+
try {
|
|
139
|
+
await new Promise((resolve, _reject) => {
|
|
140
|
+
const timeout = setTimeout(() => {
|
|
141
|
+
resolve();
|
|
142
|
+
}, 1e3);
|
|
143
|
+
ws.once("message", () => {
|
|
144
|
+
clearTimeout(timeout);
|
|
145
|
+
resolve();
|
|
146
|
+
});
|
|
147
|
+
ws.once("close", () => {
|
|
148
|
+
clearTimeout(timeout);
|
|
149
|
+
resolve();
|
|
150
|
+
});
|
|
151
|
+
ws.once("error", () => {
|
|
152
|
+
clearTimeout(timeout);
|
|
153
|
+
resolve();
|
|
154
|
+
});
|
|
155
|
+
});
|
|
156
|
+
} catch (e) {
|
|
157
|
+
}
|
|
158
|
+
}
|
|
159
|
+
} catch (e) {
|
|
160
|
+
console.warn(`Error during WebSocket close sequence: ${e}`);
|
|
161
|
+
} finally {
|
|
162
|
+
if (ws.readyState === WebSocket.OPEN || ws.readyState === WebSocket.CONNECTING) {
|
|
163
|
+
ws.close();
|
|
164
|
+
}
|
|
165
|
+
}
|
|
166
|
+
}
|
|
167
|
+
async run() {
|
|
168
|
+
const requestId = shortuuid();
|
|
169
|
+
const segmentId = shortuuid();
|
|
170
|
+
const wsUrl = this.opts.baseUrl.replace(/^http/, "ws");
|
|
171
|
+
const url = new URL(`${wsUrl}/v1/speak`);
|
|
172
|
+
url.searchParams.append("sample_rate", this.opts.sampleRate.toString());
|
|
173
|
+
url.searchParams.append("model", this.opts.model);
|
|
174
|
+
url.searchParams.append("encoding", this.opts.encoding);
|
|
175
|
+
const ws = new WebSocket(url, {
|
|
176
|
+
headers: {
|
|
177
|
+
[AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey}`
|
|
178
|
+
}
|
|
179
|
+
});
|
|
180
|
+
await new Promise((resolve, reject) => {
|
|
181
|
+
ws.on("open", resolve);
|
|
182
|
+
ws.on("error", (error) => reject(error));
|
|
183
|
+
ws.on("close", (code) => reject(`WebSocket returned ${code}`));
|
|
184
|
+
});
|
|
185
|
+
const inputTask = async () => {
|
|
186
|
+
for await (const data of this.input) {
|
|
187
|
+
if (data === SynthesizeStream.FLUSH_SENTINEL) {
|
|
188
|
+
this.tokenizer.flush();
|
|
189
|
+
continue;
|
|
190
|
+
}
|
|
191
|
+
this.tokenizer.pushText(data);
|
|
192
|
+
}
|
|
193
|
+
this.tokenizer.endInput();
|
|
194
|
+
this.tokenizer.close();
|
|
195
|
+
};
|
|
196
|
+
const sendTask = async () => {
|
|
197
|
+
for await (const event of this.tokenizer) {
|
|
198
|
+
if (this.abortController.signal.aborted) break;
|
|
199
|
+
let text = event.token;
|
|
200
|
+
if (!text.endsWith(" ")) {
|
|
201
|
+
text += " ";
|
|
202
|
+
}
|
|
203
|
+
const message = JSON.stringify({
|
|
204
|
+
type: "Speak",
|
|
205
|
+
text
|
|
206
|
+
});
|
|
207
|
+
ws.send(message);
|
|
208
|
+
}
|
|
209
|
+
if (!this.abortController.signal.aborted) {
|
|
210
|
+
ws.send(SynthesizeStream.FLUSH_MSG);
|
|
211
|
+
}
|
|
212
|
+
};
|
|
213
|
+
const recvTask = async () => {
|
|
214
|
+
const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);
|
|
215
|
+
let finalReceived = false;
|
|
216
|
+
let timeout = null;
|
|
217
|
+
let lastFrame;
|
|
218
|
+
const sendLastFrame = (segmentId2, final) => {
|
|
219
|
+
if (lastFrame && !this.queue.closed) {
|
|
220
|
+
this.queue.put({ requestId, segmentId: segmentId2, frame: lastFrame, final });
|
|
221
|
+
lastFrame = void 0;
|
|
222
|
+
}
|
|
223
|
+
};
|
|
224
|
+
const clearMessageTimeout = () => {
|
|
225
|
+
if (timeout) {
|
|
226
|
+
clearTimeout(timeout);
|
|
227
|
+
timeout = null;
|
|
228
|
+
}
|
|
229
|
+
};
|
|
230
|
+
return new Promise((resolve, reject) => {
|
|
231
|
+
ws.on("message", (data, isBinary) => {
|
|
232
|
+
clearMessageTimeout();
|
|
233
|
+
if (!isBinary) {
|
|
234
|
+
const message = JSON.parse(data.toString());
|
|
235
|
+
if (message.type === "Flushed") {
|
|
236
|
+
finalReceived = true;
|
|
237
|
+
clearMessageTimeout();
|
|
238
|
+
for (const frame of bstream.flush()) {
|
|
239
|
+
sendLastFrame(segmentId, false);
|
|
240
|
+
lastFrame = frame;
|
|
241
|
+
}
|
|
242
|
+
sendLastFrame(segmentId, true);
|
|
243
|
+
if (!this.queue.closed) {
|
|
244
|
+
this.queue.put(SynthesizeStream.END_OF_STREAM);
|
|
245
|
+
}
|
|
246
|
+
resolve();
|
|
247
|
+
}
|
|
248
|
+
return;
|
|
249
|
+
}
|
|
250
|
+
const buffer = data instanceof Buffer ? data.buffer.slice(data.byteOffset, data.byteOffset + data.byteLength) : data;
|
|
251
|
+
for (const frame of bstream.write(buffer)) {
|
|
252
|
+
sendLastFrame(segmentId, false);
|
|
253
|
+
lastFrame = frame;
|
|
254
|
+
}
|
|
255
|
+
});
|
|
256
|
+
ws.on("close", (_code, _reason) => {
|
|
257
|
+
if (!finalReceived) {
|
|
258
|
+
for (const frame of bstream.flush()) {
|
|
259
|
+
sendLastFrame(segmentId, false);
|
|
260
|
+
lastFrame = frame;
|
|
261
|
+
}
|
|
262
|
+
sendLastFrame(segmentId, true);
|
|
263
|
+
if (!this.queue.closed) {
|
|
264
|
+
this.queue.put(SynthesizeStream.END_OF_STREAM);
|
|
265
|
+
}
|
|
266
|
+
}
|
|
267
|
+
resolve();
|
|
268
|
+
});
|
|
269
|
+
ws.on("error", (error) => {
|
|
270
|
+
clearMessageTimeout();
|
|
271
|
+
reject(error);
|
|
272
|
+
});
|
|
273
|
+
});
|
|
274
|
+
};
|
|
275
|
+
try {
|
|
276
|
+
await Promise.all([inputTask(), sendTask(), recvTask()]);
|
|
277
|
+
} catch (e) {
|
|
278
|
+
throw new Error(`failed in main task: ${e}`);
|
|
279
|
+
} finally {
|
|
280
|
+
await this.closeWebSocket(ws);
|
|
281
|
+
}
|
|
282
|
+
}
|
|
283
|
+
}
|
|
284
|
+
export {
|
|
285
|
+
ChunkedStream,
|
|
286
|
+
SynthesizeStream,
|
|
287
|
+
TTS
|
|
288
|
+
};
|
|
289
|
+
//# sourceMappingURL=tts.js.map
|
package/dist/tts.js.map
ADDED
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { AudioByteStream, shortuuid, tokenize, tts } from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { request } from 'node:https';\nimport { type RawData, WebSocket } from 'ws';\nimport type { TTSEncoding, TTSModels } from './models.js';\n\nconst AUTHORIZATION_HEADER = 'Authorization';\nconst NUM_CHANNELS = 1;\nconst MIN_SENTENCE_LENGTH = 8;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n encoding: TTSEncoding;\n sampleRate: number;\n apiKey?: string;\n baseUrl?: string;\n sentenceTokenizer: tokenize.SentenceTokenizer;\n capabilities: tts.TTSCapabilities;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'aura-asteria-en',\n encoding: 'linear16',\n sampleRate: 24000,\n apiKey: process.env.DEEPGRAM_API_KEY,\n baseUrl: 'https://api.deepgram.com',\n capabilities: {\n streaming: true,\n },\n sentenceTokenizer: new tokenize.basic.SentenceTokenizer({\n minSentenceLength: MIN_SENTENCE_LENGTH,\n }),\n};\n\nexport class TTS extends tts.TTS {\n private opts: TTSOptions;\n label = 'deepgram.TTS';\n\n constructor(opts: Partial<TTSOptions> = {}) {\n super(opts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {\n streaming: opts.capabilities?.streaming ?? defaultTTSOptions.capabilities.streaming,\n });\n\n this.opts = {\n ...defaultTTSOptions,\n ...opts,\n };\n\n if (this.opts.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n }\n\n synthesize(text: string): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.opts);\n }\n\n stream(): tts.SynthesizeStream {\n return new SynthesizeStream(this, this.opts);\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'deepgram.ChunkedStream';\n private opts: TTSOptions;\n private text: string;\n\n constructor(tts: TTS, text: string, opts: TTSOptions) {\n super(text, tts);\n this.text = text;\n this.opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const json = { text: this.text };\n const url = new URL(`${this.opts.baseUrl!}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n\n await new Promise<void>((resolve, reject) => {\n const req = request(\n {\n hostname: url.hostname,\n port: 443,\n path: url.pathname + url.search,\n method: 'POST',\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n 'Content-Type': 'application/json',\n },\n },\n (res) => {\n if (res.statusCode !== 200) {\n reject(\n new Error(`Deepgram TTS HTTP request failed: ${res.statusCode} ${res.statusMessage}`),\n );\n return;\n }\n\n res.on('data', (chunk) => {\n for (const frame of bstream.write(chunk)) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n });\n\n res.on('error', (err) => {\n reject(err);\n });\n\n res.on('close', () => {\n for (const frame of bstream.flush()) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n if (!this.queue.closed) {\n this.queue.close();\n }\n resolve();\n });\n },\n );\n\n req.on('error', (err) => {\n reject(err);\n });\n\n req.write(JSON.stringify(json));\n req.end();\n });\n }\n}\n\nexport class SynthesizeStream extends tts.SynthesizeStream {\n private opts: TTSOptions;\n private tokenizer: tokenize.SentenceStream;\n label = 'deepgram.SynthesizeStream';\n\n private static readonly FLUSH_MSG = JSON.stringify({ type: 'Flush' });\n private static readonly CLOSE_MSG = JSON.stringify({ type: 'Close' });\n\n constructor(tts: TTS, opts: TTSOptions) {\n super(tts);\n this.opts = opts;\n this.tokenizer = opts.sentenceTokenizer.stream();\n }\n\n private async closeWebSocket(ws: WebSocket): Promise<void> {\n try {\n // Send Flush and Close messages to ensure Deepgram processes all remaining audio\n // and properly terminates the session, preventing lingering TTS sessions\n if (ws.readyState === WebSocket.OPEN) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n ws.send(SynthesizeStream.CLOSE_MSG);\n\n // Wait for server acknowledgment to prevent race conditions and ensure\n // proper cleanup, avoiding 429 Too Many Requests errors from lingering sessions\n try {\n await new Promise<void>((resolve, _reject) => {\n const timeout = setTimeout(() => {\n resolve();\n }, 1000);\n\n ws.once('message', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('close', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('error', () => {\n clearTimeout(timeout);\n resolve();\n });\n });\n } catch (e) {\n // Ignore timeout or other errors during close sequence\n }\n }\n } catch (e) {\n console.warn(`Error during WebSocket close sequence: ${e}`);\n } finally {\n if (ws.readyState === WebSocket.OPEN || ws.readyState === WebSocket.CONNECTING) {\n ws.close();\n }\n }\n }\n\n protected async run() {\n const requestId = shortuuid();\n const segmentId = shortuuid();\n\n const wsUrl = this.opts.baseUrl!.replace(/^http/, 'ws');\n const url = new URL(`${wsUrl}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n\n const ws = new WebSocket(url, {\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n },\n });\n\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 const inputTask = async () => {\n for await (const data of this.input) {\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n this.tokenizer.flush();\n continue;\n }\n this.tokenizer.pushText(data);\n }\n this.tokenizer.endInput();\n this.tokenizer.close();\n };\n\n const sendTask = async () => {\n for await (const event of this.tokenizer) {\n if (this.abortController.signal.aborted) break;\n\n let text = event.token;\n if (!text.endsWith(' ')) {\n text += ' ';\n }\n\n const message = JSON.stringify({\n type: 'Speak',\n text: text,\n });\n\n ws.send(message);\n }\n\n if (!this.abortController.signal.aborted) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n }\n };\n\n const recvTask = async () => {\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n let finalReceived = false;\n let timeout: NodeJS.Timeout | null = null;\n let lastFrame: AudioFrame | undefined;\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame && !this.queue.closed) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const clearMessageTimeout = () => {\n if (timeout) {\n clearTimeout(timeout);\n timeout = null;\n }\n };\n\n return new Promise<void>((resolve, reject) => {\n ws.on('message', (data: RawData, isBinary: boolean) => {\n clearMessageTimeout();\n\n if (!isBinary) {\n const message = JSON.parse(data.toString());\n if (message.type === 'Flushed') {\n finalReceived = true;\n clearMessageTimeout();\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n resolve();\n }\n\n return;\n }\n\n const buffer =\n data instanceof Buffer\n ? data.buffer.slice(data.byteOffset, data.byteOffset + data.byteLength)\n : (data as ArrayBuffer);\n for (const frame of bstream.write(buffer as ArrayBuffer)) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n });\n\n ws.on('close', (_code, _reason) => {\n if (!finalReceived) {\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n }\n resolve();\n });\n\n ws.on('error', (error) => {\n clearMessageTimeout();\n reject(error);\n });\n });\n };\n\n try {\n await Promise.all([inputTask(), sendTask(), recvTask()]);\n } catch (e) {\n throw new Error(`failed in main task: ${e}`);\n } finally {\n await this.closeWebSocket(ws);\n }\n }\n}\n"],"mappings":"AAGA,SAAS,iBAAiB,WAAW,UAAU,WAAW;AAE1D,SAAS,eAAe;AACxB,SAAuB,iBAAiB;AAGxC,MAAM,uBAAuB;AAC7B,MAAM,eAAe;AACrB,MAAM,sBAAsB;AAY5B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,QAAQ,QAAQ,IAAI;AAAA,EACpB,SAAS;AAAA,EACT,cAAc;AAAA,IACZ,WAAW;AAAA,EACb;AAAA,EACA,mBAAmB,IAAI,SAAS,MAAM,kBAAkB;AAAA,IACtD,mBAAmB;AAAA,EACrB,CAAC;AACH;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EACvB;AAAA,EACR,QAAQ;AAAA,EAER,YAAY,OAA4B,CAAC,GAAG;AAzC9C;AA0CI,UAAM,KAAK,cAAc,kBAAkB,YAAY,cAAc;AAAA,MACnE,aAAW,UAAK,iBAAL,mBAAmB,cAAa,kBAAkB,aAAa;AAAA,IAC5E,CAAC;AAED,SAAK,OAAO;AAAA,MACV,GAAG;AAAA,MACH,GAAG;AAAA,IACL;AAEA,QAAI,KAAK,KAAK,WAAW,QAAW;AAClC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,WAAW,MAAiC;AAC1C,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,IAAI;AAAA,EAChD;AAAA,EAEA,SAA+B;AAC7B,WAAO,IAAI,iBAAiB,MAAM,KAAK,IAAI;AAAA,EAC7C;AACF;AAEO,MAAM,sBAAsB,IAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACA;AAAA,EACA;AAAA,EAER,YAAYA,MAAU,MAAc,MAAkB;AACpD,UAAM,MAAMA,IAAG;AACf,SAAK,OAAO;AACZ,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,YAAY,UAAU;AAC5B,UAAM,UAAU,IAAI,gBAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAM,OAAO,EAAE,MAAM,KAAK,KAAK;AAC/B,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,KAAK,OAAQ,WAAW;AACpD,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AAEtD,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,YAAM,MAAM;AAAA,QACV;AAAA,UACE,UAAU,IAAI;AAAA,UACd,MAAM;AAAA,UACN,MAAM,IAAI,WAAW,IAAI;AAAA,UACzB,QAAQ;AAAA,UACR,SAAS;AAAA,YACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,YAClD,gBAAgB;AAAA,UAClB;AAAA,QACF;AAAA,QACA,CAAC,QAAQ;AACP,cAAI,IAAI,eAAe,KAAK;AAC1B;AAAA,cACE,IAAI,MAAM,qCAAqC,IAAI,UAAU,IAAI,IAAI,aAAa,EAAE;AAAA,YACtF;AACA;AAAA,UACF;AAEA,cAAI,GAAG,QAAQ,CAAC,UAAU;AACxB,uBAAW,SAAS,QAAQ,MAAM,KAAK,GAAG;AACxC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AAAA,UACF,CAAC;AAED,cAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,mBAAO,GAAG;AAAA,UACZ,CAAC;AAED,cAAI,GAAG,SAAS,MAAM;AACpB,uBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AACA,gBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,mBAAK,MAAM,MAAM;AAAA,YACnB;AACA,oBAAQ;AAAA,UACV,CAAC;AAAA,QACH;AAAA,MACF;AAEA,UAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,eAAO,GAAG;AAAA,MACZ,CAAC;AAED,UAAI,MAAM,KAAK,UAAU,IAAI,CAAC;AAC9B,UAAI,IAAI;AAAA,IACV,CAAC;AAAA,EACH;AACF;AAEO,MAAM,yBAAyB,IAAI,iBAAiB;AAAA,EACjD;AAAA,EACA;AAAA,EACR,QAAQ;AAAA,EAER,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EACpE,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EAEpE,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,OAAO;AACZ,SAAK,YAAY,KAAK,kBAAkB,OAAO;AAAA,EACjD;AAAA,EAEA,MAAc,eAAe,IAA8B;AACzD,QAAI;AAGF,UAAI,GAAG,eAAe,UAAU,MAAM;AACpC,WAAG,KAAK,iBAAiB,SAAS;AAClC,WAAG,KAAK,iBAAiB,SAAS;AAIlC,YAAI;AACF,gBAAM,IAAI,QAAc,CAAC,SAAS,YAAY;AAC5C,kBAAM,UAAU,WAAW,MAAM;AAC/B,sBAAQ;AAAA,YACV,GAAG,GAAI;AAEP,eAAG,KAAK,WAAW,MAAM;AACvB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAAA,UACH,CAAC;AAAA,QACH,SAAS,GAAG;AAAA,QAEZ;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AACV,cAAQ,KAAK,0CAA0C,CAAC,EAAE;AAAA,IAC5D,UAAE;AACA,UAAI,GAAG,eAAe,UAAU,QAAQ,GAAG,eAAe,UAAU,YAAY;AAC9E,WAAG,MAAM;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,YAAY,UAAU;AAC5B,UAAM,YAAY,UAAU;AAE5B,UAAM,QAAQ,KAAK,KAAK,QAAS,QAAQ,SAAS,IAAI;AACtD,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,WAAW;AACvC,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AAEtD,UAAM,KAAK,IAAI,UAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,MACpD;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,SAAG,GAAG,QAAQ,OAAO;AACrB,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,IAC/D,CAAC;AAED,UAAM,YAAY,YAAY;AAC5B,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,eAAK,UAAU,MAAM;AACrB;AAAA,QACF;AACA,aAAK,UAAU,SAAS,IAAI;AAAA,MAC9B;AACA,WAAK,UAAU,SAAS;AACxB,WAAK,UAAU,MAAM;AAAA,IACvB;AAEA,UAAM,WAAW,YAAY;AAC3B,uBAAiB,SAAS,KAAK,WAAW;AACxC,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,YAAI,OAAO,MAAM;AACjB,YAAI,CAAC,KAAK,SAAS,GAAG,GAAG;AACvB,kBAAQ;AAAA,QACV;AAEA,cAAM,UAAU,KAAK,UAAU;AAAA,UAC7B,MAAM;AAAA,UACN;AAAA,QACF,CAAC;AAED,WAAG,KAAK,OAAO;AAAA,MACjB;AAEA,UAAI,CAAC,KAAK,gBAAgB,OAAO,SAAS;AACxC,WAAG,KAAK,iBAAiB,SAAS;AAAA,MACpC;AAAA,IACF;AAEA,UAAM,WAAW,YAAY;AAC3B,YAAM,UAAU,IAAI,gBAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAI,gBAAgB;AACpB,UAAI,UAAiC;AACrC,UAAI;AAEJ,YAAM,gBAAgB,CAACC,YAAmB,UAAmB;AAC3D,YAAI,aAAa,CAAC,KAAK,MAAM,QAAQ;AACnC,eAAK,MAAM,IAAI,EAAE,WAAW,WAAAA,YAAW,OAAO,WAAW,MAAM,CAAC;AAChE,sBAAY;AAAA,QACd;AAAA,MACF;AAEA,YAAM,sBAAsB,MAAM;AAChC,YAAI,SAAS;AACX,uBAAa,OAAO;AACpB,oBAAU;AAAA,QACZ;AAAA,MACF;AAEA,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,WAAG,GAAG,WAAW,CAAC,MAAe,aAAsB;AACrD,8BAAoB;AAEpB,cAAI,CAAC,UAAU;AACb,kBAAM,UAAU,KAAK,MAAM,KAAK,SAAS,CAAC;AAC1C,gBAAI,QAAQ,SAAS,WAAW;AAC9B,8BAAgB;AAChB,kCAAoB;AACpB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,WAAW,KAAK;AAC9B,4BAAY;AAAA,cACd;AACA,4BAAc,WAAW,IAAI;AAE7B,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,cAC/C;AACA,sBAAQ;AAAA,YACV;AAEA;AAAA,UACF;AAEA,gBAAM,SACJ,gBAAgB,SACZ,KAAK,OAAO,MAAM,KAAK,YAAY,KAAK,aAAa,KAAK,UAAU,IACnE;AACP,qBAAW,SAAS,QAAQ,MAAM,MAAqB,GAAG;AACxD,0BAAc,WAAW,KAAK;AAC9B,wBAAY;AAAA,UACd;AAAA,QACF,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,OAAO,YAAY;AACjC,cAAI,CAAC,eAAe;AAClB,uBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,4BAAc,WAAW,KAAK;AAC9B,0BAAY;AAAA,YACd;AACA,0BAAc,WAAW,IAAI;AAE7B,gBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,YAC/C;AAAA,UACF;AACA,kBAAQ;AAAA,QACV,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,UAAU;AACxB,8BAAoB;AACpB,iBAAO,KAAK;AAAA,QACd,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,UAAU,GAAG,SAAS,GAAG,SAAS,CAAC,CAAC;AAAA,IACzD,SAAS,GAAG;AACV,YAAM,IAAI,MAAM,wBAAwB,CAAC,EAAE;AAAA,IAC7C,UAAE;AACA,YAAM,KAAK,eAAe,EAAE;AAAA,IAC9B;AAAA,EACF;AACF;","names":["tts","segmentId"]}
|
|
@@ -0,0 +1,9 @@
|
|
|
1
|
+
"use strict";
|
|
2
|
+
var import_agents_plugin_deepgram = require("@livekit/agents-plugin-deepgram");
|
|
3
|
+
var import_agents_plugins_test = require("@livekit/agents-plugins-test");
|
|
4
|
+
var import_vitest = require("vitest");
|
|
5
|
+
var import_tts = require("./tts.cjs");
|
|
6
|
+
(0, import_vitest.describe)("Deepgram", async () => {
|
|
7
|
+
await (0, import_agents_plugins_test.tts)(new import_tts.TTS(), new import_agents_plugin_deepgram.STT());
|
|
8
|
+
});
|
|
9
|
+
//# sourceMappingURL=tts.test.cjs.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"sources":["../src/tts.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { STT } from '@livekit/agents-plugin-deepgram';\nimport { tts } from '@livekit/agents-plugins-test';\nimport { describe } from 'vitest';\nimport { TTS } from './tts.js';\n\ndescribe('Deepgram', async () => {\n await tts(new TTS(), new STT());\n});\n"],"mappings":";AAGA,oCAAoB;AACpB,iCAAoB;AACpB,oBAAyB;AACzB,iBAAoB;AAAA,IAEpB,wBAAS,YAAY,YAAY;AAC/B,YAAM,gCAAI,IAAI,eAAI,GAAG,IAAI,kCAAI,CAAC;AAChC,CAAC;","names":[]}
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"tts.test.d.ts","sourceRoot":"","sources":["../src/tts.test.ts"],"names":[],"mappings":""}
|
package/dist/tts.test.js
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
1
|
+
import { STT } from "@livekit/agents-plugin-deepgram";
|
|
2
|
+
import { tts } from "@livekit/agents-plugins-test";
|
|
3
|
+
import { describe } from "vitest";
|
|
4
|
+
import { TTS } from "./tts.js";
|
|
5
|
+
describe("Deepgram", async () => {
|
|
6
|
+
await tts(new TTS(), new STT());
|
|
7
|
+
});
|
|
8
|
+
//# sourceMappingURL=tts.test.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"sources":["../src/tts.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { STT } from '@livekit/agents-plugin-deepgram';\nimport { tts } from '@livekit/agents-plugins-test';\nimport { describe } from 'vitest';\nimport { TTS } from './tts.js';\n\ndescribe('Deepgram', async () => {\n await tts(new TTS(), new STT());\n});\n"],"mappings":"AAGA,SAAS,WAAW;AACpB,SAAS,WAAW;AACpB,SAAS,gBAAgB;AACzB,SAAS,WAAW;AAEpB,SAAS,YAAY,YAAY;AAC/B,QAAM,IAAI,IAAI,IAAI,GAAG,IAAI,IAAI,CAAC;AAChC,CAAC;","names":[]}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@livekit/agents-plugin-deepgram",
|
|
3
|
-
"version": "1.0.
|
|
3
|
+
"version": "1.0.13",
|
|
4
4
|
"description": "Deepgram plugin for LiveKit Agents for Node.js",
|
|
5
5
|
"main": "dist/index.js",
|
|
6
6
|
"require": "dist/index.cjs",
|
|
@@ -30,16 +30,16 @@
|
|
|
30
30
|
"@types/ws": "^8.5.10",
|
|
31
31
|
"tsup": "^8.3.5",
|
|
32
32
|
"typescript": "^5.0.0",
|
|
33
|
-
"@livekit/agents": "1.0.
|
|
34
|
-
"@livekit/agents-plugin-silero": "1.0.
|
|
35
|
-
"@livekit/agents-plugins-test": "1.0.
|
|
33
|
+
"@livekit/agents": "1.0.13",
|
|
34
|
+
"@livekit/agents-plugin-silero": "1.0.13",
|
|
35
|
+
"@livekit/agents-plugins-test": "1.0.13"
|
|
36
36
|
},
|
|
37
37
|
"dependencies": {
|
|
38
38
|
"ws": "^8.16.0"
|
|
39
39
|
},
|
|
40
40
|
"peerDependencies": {
|
|
41
41
|
"@livekit/rtc-node": "^0.13.12",
|
|
42
|
-
"@livekit/agents": "1.0.
|
|
42
|
+
"@livekit/agents": "1.0.13"
|
|
43
43
|
},
|
|
44
44
|
"scripts": {
|
|
45
45
|
"build": "tsup --onSuccess \"pnpm build:types\"",
|
package/src/index.ts
CHANGED
package/src/models.ts
CHANGED
|
@@ -2,6 +2,22 @@
|
|
|
2
2
|
//
|
|
3
3
|
// SPDX-License-Identifier: Apache-2.0
|
|
4
4
|
|
|
5
|
+
export type TTSModels =
|
|
6
|
+
| 'aura-asteria-en'
|
|
7
|
+
| 'aura-luna-en'
|
|
8
|
+
| 'aura-stella-en'
|
|
9
|
+
| 'aura-athena-en'
|
|
10
|
+
| 'aura-hera-en'
|
|
11
|
+
| 'aura-orion-en'
|
|
12
|
+
| 'aura-arcas-en'
|
|
13
|
+
| 'aura-perseus-en'
|
|
14
|
+
| 'aura-angus-en'
|
|
15
|
+
| 'aura-orpheus-en'
|
|
16
|
+
| 'aura-helios-en'
|
|
17
|
+
| 'aura-zeus-en';
|
|
18
|
+
|
|
19
|
+
export type TTSEncoding = 'linear16' | 'mulaw' | 'alaw' | 'mp3' | 'opus' | 'flac' | 'aac';
|
|
20
|
+
|
|
5
21
|
export type STTModels =
|
|
6
22
|
| 'nova-general'
|
|
7
23
|
| 'nova-phonecall'
|
package/src/tts.test.ts
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
1
|
+
// SPDX-FileCopyrightText: 2024 LiveKit, Inc.
|
|
2
|
+
//
|
|
3
|
+
// SPDX-License-Identifier: Apache-2.0
|
|
4
|
+
import { STT } from '@livekit/agents-plugin-deepgram';
|
|
5
|
+
import { tts } from '@livekit/agents-plugins-test';
|
|
6
|
+
import { describe } from 'vitest';
|
|
7
|
+
import { TTS } from './tts.js';
|
|
8
|
+
|
|
9
|
+
describe('Deepgram', async () => {
|
|
10
|
+
await tts(new TTS(), new STT());
|
|
11
|
+
});
|