@cogitator-ai/voice 0.1.13 → 0.2.0
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 +121 -77
- package/dist/audio.d.ts +7 -0
- package/dist/audio.d.ts.map +1 -1
- package/dist/audio.js +40 -3
- package/dist/audio.js.map +1 -1
- package/dist/cogitator-runner.d.ts +34 -0
- package/dist/cogitator-runner.d.ts.map +1 -0
- package/dist/cogitator-runner.js +28 -0
- package/dist/cogitator-runner.js.map +1 -0
- package/dist/index.d.ts +6 -5
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +7 -4
- package/dist/index.js.map +1 -1
- package/dist/pipeline/index.d.ts +2 -2
- package/dist/pipeline/index.d.ts.map +1 -1
- package/dist/pipeline/index.js.map +1 -1
- package/dist/pipeline/pipeline-session.d.ts +20 -2
- package/dist/pipeline/pipeline-session.d.ts.map +1 -1
- package/dist/pipeline/pipeline-session.js +187 -63
- package/dist/pipeline/pipeline-session.js.map +1 -1
- package/dist/pipeline/voice-pipeline.d.ts +12 -7
- package/dist/pipeline/voice-pipeline.d.ts.map +1 -1
- package/dist/pipeline/voice-pipeline.js +14 -4
- package/dist/pipeline/voice-pipeline.js.map +1 -1
- package/dist/realtime/gemini-realtime.d.ts +15 -3
- package/dist/realtime/gemini-realtime.d.ts.map +1 -1
- package/dist/realtime/gemini-realtime.js +140 -94
- package/dist/realtime/gemini-realtime.js.map +1 -1
- package/dist/realtime/openai-realtime.d.ts +10 -1
- package/dist/realtime/openai-realtime.d.ts.map +1 -1
- package/dist/realtime/openai-realtime.js +140 -69
- package/dist/realtime/openai-realtime.js.map +1 -1
- package/dist/realtime/realtime-session.d.ts +10 -0
- package/dist/realtime/realtime-session.d.ts.map +1 -1
- package/dist/realtime/realtime-session.js +60 -7
- package/dist/realtime/realtime-session.js.map +1 -1
- package/dist/stt/deepgram-stt.d.ts +3 -0
- package/dist/stt/deepgram-stt.d.ts.map +1 -1
- package/dist/stt/deepgram-stt.js +104 -55
- package/dist/stt/deepgram-stt.js.map +1 -1
- package/dist/stt/openai-stt.d.ts +5 -0
- package/dist/stt/openai-stt.d.ts.map +1 -1
- package/dist/stt/openai-stt.js +70 -13
- package/dist/stt/openai-stt.js.map +1 -1
- package/dist/tools.d.ts +3 -3
- package/dist/transport/ws-transport.d.ts +6 -7
- package/dist/transport/ws-transport.d.ts.map +1 -1
- package/dist/transport/ws-transport.js +58 -20
- package/dist/transport/ws-transport.js.map +1 -1
- package/dist/tts/elevenlabs-tts.d.ts.map +1 -1
- package/dist/tts/elevenlabs-tts.js +26 -16
- package/dist/tts/elevenlabs-tts.js.map +1 -1
- package/dist/types.d.ts +38 -16
- package/dist/types.d.ts.map +1 -1
- package/dist/vad/silero-vad.d.ts +14 -1
- package/dist/vad/silero-vad.d.ts.map +1 -1
- package/dist/vad/silero-vad.js +101 -25
- package/dist/vad/silero-vad.js.map +1 -1
- package/dist/voice-agent.d.ts +14 -0
- package/dist/voice-agent.d.ts.map +1 -1
- package/dist/voice-agent.js +112 -23
- package/dist/voice-agent.js.map +1 -1
- package/package.json +5 -5
package/dist/stt/deepgram-stt.js
CHANGED
|
@@ -1,17 +1,31 @@
|
|
|
1
1
|
import { EventEmitter } from 'node:events';
|
|
2
2
|
import { WebSocket } from 'ws';
|
|
3
|
+
import { audioMimeType, detectAudioFormat } from '../audio.js';
|
|
3
4
|
const DEFAULT_MODEL = 'nova-3';
|
|
5
|
+
const DEFAULT_SAMPLE_RATE = 16000;
|
|
4
6
|
const BASE_URL = 'https://api.deepgram.com/v1/listen';
|
|
5
7
|
const WS_URL = 'wss://api.deepgram.com/v1/listen';
|
|
8
|
+
function mapWords(words) {
|
|
9
|
+
if (!words || words.length === 0)
|
|
10
|
+
return undefined;
|
|
11
|
+
return words.map((w) => ({
|
|
12
|
+
word: w.word,
|
|
13
|
+
start: w.start,
|
|
14
|
+
end: w.end,
|
|
15
|
+
confidence: w.confidence,
|
|
16
|
+
}));
|
|
17
|
+
}
|
|
6
18
|
export class DeepgramSTT {
|
|
7
19
|
name = 'deepgram';
|
|
8
20
|
apiKey;
|
|
9
21
|
model;
|
|
10
22
|
language;
|
|
23
|
+
sampleRate;
|
|
11
24
|
constructor(config) {
|
|
12
25
|
this.apiKey = config.apiKey;
|
|
13
26
|
this.model = config.model ?? DEFAULT_MODEL;
|
|
14
27
|
this.language = config.language;
|
|
28
|
+
this.sampleRate = config.sampleRate ?? DEFAULT_SAMPLE_RATE;
|
|
15
29
|
}
|
|
16
30
|
async transcribe(audio, options) {
|
|
17
31
|
const params = new URLSearchParams({
|
|
@@ -21,11 +35,22 @@ export class DeepgramSTT {
|
|
|
21
35
|
const lang = options?.language ?? this.language;
|
|
22
36
|
if (lang)
|
|
23
37
|
params.set('language', lang);
|
|
38
|
+
const format = detectAudioFormat(audio);
|
|
39
|
+
let contentType;
|
|
40
|
+
if (format) {
|
|
41
|
+
contentType = audioMimeType(format);
|
|
42
|
+
}
|
|
43
|
+
else {
|
|
44
|
+
contentType = 'application/octet-stream';
|
|
45
|
+
params.set('encoding', 'linear16');
|
|
46
|
+
params.set('sample_rate', String(this.sampleRate));
|
|
47
|
+
params.set('channels', '1');
|
|
48
|
+
}
|
|
24
49
|
const response = await fetch(`${BASE_URL}?${params}`, {
|
|
25
50
|
method: 'POST',
|
|
26
51
|
headers: {
|
|
27
52
|
Authorization: `Token ${this.apiKey}`,
|
|
28
|
-
'Content-Type':
|
|
53
|
+
'Content-Type': contentType,
|
|
29
54
|
},
|
|
30
55
|
body: new Uint8Array(audio),
|
|
31
56
|
signal: AbortSignal.timeout(30_000),
|
|
@@ -38,7 +63,10 @@ export class DeepgramSTT {
|
|
|
38
63
|
return this.mapBatchResponse(data);
|
|
39
64
|
}
|
|
40
65
|
createStream(options) {
|
|
41
|
-
return new DeepgramSTTStream(this.apiKey, this.model, this.language,
|
|
66
|
+
return new DeepgramSTTStream(this.apiKey, this.model, this.language, {
|
|
67
|
+
...options,
|
|
68
|
+
sampleRate: options?.sampleRate ?? this.sampleRate,
|
|
69
|
+
});
|
|
42
70
|
}
|
|
43
71
|
mapBatchResponse(data) {
|
|
44
72
|
const alt = data.results?.channels?.[0]?.alternatives?.[0];
|
|
@@ -48,14 +76,9 @@ export class DeepgramSTT {
|
|
|
48
76
|
if (data.metadata?.duration !== undefined) {
|
|
49
77
|
result.duration = data.metadata.duration;
|
|
50
78
|
}
|
|
51
|
-
|
|
52
|
-
|
|
53
|
-
|
|
54
|
-
start: w.start,
|
|
55
|
-
end: w.end,
|
|
56
|
-
confidence: w.confidence,
|
|
57
|
-
}));
|
|
58
|
-
}
|
|
79
|
+
const words = mapWords(alt?.words);
|
|
80
|
+
if (words)
|
|
81
|
+
result.words = words;
|
|
59
82
|
return result;
|
|
60
83
|
}
|
|
61
84
|
}
|
|
@@ -64,23 +87,31 @@ class DeepgramSTTStream extends EventEmitter {
|
|
|
64
87
|
ws;
|
|
65
88
|
ready = false;
|
|
66
89
|
closed = false;
|
|
90
|
+
remoteClosed = false;
|
|
67
91
|
closePromise = null;
|
|
68
92
|
pendingChunks = [];
|
|
69
|
-
|
|
93
|
+
segments = [];
|
|
94
|
+
words = [];
|
|
95
|
+
audioEnd = 0;
|
|
70
96
|
constructor(apiKey, model, language, options) {
|
|
71
97
|
super();
|
|
72
98
|
const params = new URLSearchParams({
|
|
73
99
|
model,
|
|
74
100
|
punctuate: 'true',
|
|
75
|
-
interim_results: String(options
|
|
101
|
+
interim_results: String(options.interimResults ?? true),
|
|
102
|
+
encoding: 'linear16',
|
|
103
|
+
sample_rate: String(options.sampleRate),
|
|
104
|
+
channels: '1',
|
|
76
105
|
});
|
|
77
|
-
const lang = options
|
|
106
|
+
const lang = options.language ?? language;
|
|
78
107
|
if (lang)
|
|
79
108
|
params.set('language', lang);
|
|
80
|
-
if (options
|
|
109
|
+
if (options.endpointing !== undefined) {
|
|
81
110
|
params.set('endpointing', String(options.endpointing));
|
|
82
111
|
}
|
|
83
|
-
this.ws = new WebSocket(`${WS_URL}?${params}`,
|
|
112
|
+
this.ws = new WebSocket(`${WS_URL}?${params}`, {
|
|
113
|
+
headers: { Authorization: `Token ${apiKey}` },
|
|
114
|
+
});
|
|
84
115
|
this.ws.on('open', () => {
|
|
85
116
|
this.ready = true;
|
|
86
117
|
for (const chunk of this.pendingChunks) {
|
|
@@ -96,29 +127,56 @@ class DeepgramSTTStream extends EventEmitter {
|
|
|
96
127
|
catch {
|
|
97
128
|
return;
|
|
98
129
|
}
|
|
99
|
-
|
|
100
|
-
|
|
130
|
+
if (msg.type !== undefined && msg.type !== 'Results')
|
|
131
|
+
return;
|
|
132
|
+
const alt = msg.channel?.alternatives?.[0];
|
|
133
|
+
if (!alt)
|
|
101
134
|
return;
|
|
102
135
|
if (msg.is_final) {
|
|
103
|
-
|
|
104
|
-
|
|
105
|
-
|
|
136
|
+
if (typeof msg.start === 'number' && typeof msg.duration === 'number') {
|
|
137
|
+
this.audioEnd = Math.max(this.audioEnd, msg.start + msg.duration);
|
|
138
|
+
}
|
|
139
|
+
if (!alt.transcript)
|
|
140
|
+
return;
|
|
141
|
+
this.segments.push(alt.transcript);
|
|
142
|
+
const words = mapWords(alt.words);
|
|
143
|
+
if (words)
|
|
144
|
+
this.words.push(...words);
|
|
145
|
+
const segment = { text: alt.transcript };
|
|
146
|
+
if (words)
|
|
147
|
+
segment.words = words;
|
|
148
|
+
if (typeof msg.duration === 'number')
|
|
149
|
+
segment.duration = msg.duration;
|
|
150
|
+
this.emit('final', segment);
|
|
106
151
|
}
|
|
107
|
-
else {
|
|
152
|
+
else if (alt.transcript) {
|
|
108
153
|
this.emit('partial', alt.transcript);
|
|
109
154
|
}
|
|
110
155
|
});
|
|
111
156
|
this.ws.on('error', (err) => {
|
|
112
|
-
this.
|
|
157
|
+
if (this.listenerCount('error') > 0)
|
|
158
|
+
this.emit('error', err);
|
|
113
159
|
});
|
|
114
|
-
this.ws.on('close', () => {
|
|
160
|
+
this.ws.on('close', (code, reason) => {
|
|
115
161
|
this.ready = false;
|
|
162
|
+
if (!this.closed) {
|
|
163
|
+
this.remoteClosed = true;
|
|
164
|
+
const dropped = this.pendingChunks.length;
|
|
165
|
+
this.pendingChunks = [];
|
|
166
|
+
if (this.listenerCount('error') > 0) {
|
|
167
|
+
const detail = reason.toString() || 'no reason';
|
|
168
|
+
this.emit('error', new Error(`DeepgramSTTStream: connection closed unexpectedly (code ${code}: ${detail})` +
|
|
169
|
+
(dropped > 0 ? `, ${dropped} chunk(s) dropped` : '')));
|
|
170
|
+
}
|
|
171
|
+
}
|
|
116
172
|
});
|
|
117
173
|
}
|
|
118
174
|
write(chunk) {
|
|
119
175
|
if (this.closed) {
|
|
120
176
|
throw new Error('DeepgramSTTStream: cannot write after close');
|
|
121
177
|
}
|
|
178
|
+
if (this.remoteClosed)
|
|
179
|
+
return;
|
|
122
180
|
if (this.ready) {
|
|
123
181
|
this.ws.send(chunk);
|
|
124
182
|
}
|
|
@@ -131,52 +189,43 @@ class DeepgramSTTStream extends EventEmitter {
|
|
|
131
189
|
return this.closePromise;
|
|
132
190
|
this.closed = true;
|
|
133
191
|
this.closePromise = new Promise((resolve) => {
|
|
134
|
-
|
|
135
|
-
|
|
136
|
-
|
|
137
|
-
}
|
|
138
|
-
resolve(this.lastResult);
|
|
192
|
+
const state = this.ws.readyState;
|
|
193
|
+
if (state === WebSocket.CLOSED || state === WebSocket.CLOSING) {
|
|
194
|
+
resolve(this.buildResult());
|
|
139
195
|
return;
|
|
140
196
|
}
|
|
141
197
|
const timeout = setTimeout(() => {
|
|
142
198
|
this.ws.terminate();
|
|
143
|
-
resolve(this.
|
|
199
|
+
resolve(this.buildResult());
|
|
144
200
|
}, CLOSE_TIMEOUT_MS);
|
|
145
201
|
this.ws.once('close', () => {
|
|
146
202
|
clearTimeout(timeout);
|
|
147
|
-
resolve(this.
|
|
203
|
+
resolve(this.buildResult());
|
|
148
204
|
});
|
|
149
|
-
if (
|
|
205
|
+
if (state === WebSocket.OPEN) {
|
|
150
206
|
this.ws.send(JSON.stringify({ type: 'CloseStream' }));
|
|
207
|
+
return;
|
|
151
208
|
}
|
|
152
|
-
|
|
153
|
-
|
|
154
|
-
|
|
155
|
-
|
|
156
|
-
|
|
157
|
-
this.
|
|
209
|
+
if (this.pendingChunks.length === 0) {
|
|
210
|
+
clearTimeout(timeout);
|
|
211
|
+
this.ws.removeAllListeners('close');
|
|
212
|
+
this.ws.on('error', () => { });
|
|
213
|
+
this.ws.terminate();
|
|
214
|
+
resolve(this.buildResult());
|
|
215
|
+
return;
|
|
158
216
|
}
|
|
217
|
+
this.ws.once('open', () => {
|
|
218
|
+
this.ws.send(JSON.stringify({ type: 'CloseStream' }));
|
|
219
|
+
});
|
|
159
220
|
});
|
|
160
221
|
return this.closePromise;
|
|
161
222
|
}
|
|
162
|
-
|
|
163
|
-
const
|
|
164
|
-
if (
|
|
165
|
-
|
|
166
|
-
|
|
167
|
-
|
|
168
|
-
};
|
|
169
|
-
if (msg.metadata?.duration !== undefined) {
|
|
170
|
-
result.duration = msg.metadata.duration;
|
|
171
|
-
}
|
|
172
|
-
if (alt.words && alt.words.length > 0) {
|
|
173
|
-
result.words = alt.words.map((w) => ({
|
|
174
|
-
word: w.word,
|
|
175
|
-
start: w.start,
|
|
176
|
-
end: w.end,
|
|
177
|
-
confidence: w.confidence,
|
|
178
|
-
}));
|
|
179
|
-
}
|
|
223
|
+
buildResult() {
|
|
224
|
+
const result = { text: this.segments.join(' ').trim() };
|
|
225
|
+
if (this.audioEnd > 0)
|
|
226
|
+
result.duration = this.audioEnd;
|
|
227
|
+
if (this.words.length > 0)
|
|
228
|
+
result.words = [...this.words];
|
|
180
229
|
return result;
|
|
181
230
|
}
|
|
182
231
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"deepgram-stt.js","sourceRoot":"","sources":["../../src/stt/deepgram-stt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;
|
|
1
|
+
{"version":3,"file":"deepgram-stt.js","sourceRoot":"","sources":["../../src/stt/deepgram-stt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAC/B,OAAO,EAAE,aAAa,EAAE,iBAAiB,EAAE,MAAM,aAAa,CAAC;AAiB/D,MAAM,aAAa,GAAG,QAAQ,CAAC;AAC/B,MAAM,mBAAmB,GAAG,KAAK,CAAC;AAClC,MAAM,QAAQ,GAAG,oCAAoC,CAAC;AACtD,MAAM,MAAM,GAAG,kCAAkC,CAAC;AAgClD,SAAS,QAAQ,CAAC,KAAiC;IACjD,IAAI,CAAC,KAAK,IAAI,KAAK,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,SAAS,CAAC;IACnD,OAAO,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;QACvB,IAAI,EAAE,CAAC,CAAC,IAAI;QACZ,KAAK,EAAE,CAAC,CAAC,KAAK;QACd,GAAG,EAAE,CAAC,CAAC,GAAG;QACV,UAAU,EAAE,CAAC,CAAC,UAAU;KACzB,CAAC,CAAC,CAAC;AACN,CAAC;AAED,MAAM,OAAO,WAAW;IACb,IAAI,GAAG,UAAU,CAAC;IAEV,MAAM,CAAS;IACf,KAAK,CAAS;IACd,QAAQ,CAAU;IAClB,UAAU,CAAS;IAEpC,YAAY,MAAyB;QACnC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,IAAI,aAAa,CAAC;QAC3C,IAAI,CAAC,QAAQ,GAAG,MAAM,CAAC,QAAQ,CAAC;QAChC,IAAI,CAAC,UAAU,GAAG,MAAM,CAAC,UAAU,IAAI,mBAAmB,CAAC;IAC7D,CAAC;IAED,KAAK,CAAC,UAAU,CAAC,KAAa,EAAE,OAAoB;QAClD,MAAM,MAAM,GAAG,IAAI,eAAe,CAAC;YACjC,KAAK,EAAE,IAAI,CAAC,KAAK;YACjB,SAAS,EAAE,MAAM;SAClB,CAAC,CAAC;QAEH,MAAM,IAAI,GAAG,OAAO,EAAE,QAAQ,IAAI,IAAI,CAAC,QAAQ,CAAC;QAChD,IAAI,IAAI;YAAE,MAAM,CAAC,GAAG,CAAC,UAAU,EAAE,IAAI,CAAC,CAAC;QAEvC,MAAM,MAAM,GAAG,iBAAiB,CAAC,KAAK,CAAC,CAAC;QACxC,IAAI,WAAmB,CAAC;QACxB,IAAI,MAAM,EAAE,CAAC;YACX,WAAW,GAAG,aAAa,CAAC,MAAM,CAAC,CAAC;QACtC,CAAC;aAAM,CAAC;YACN,WAAW,GAAG,0BAA0B,CAAC;YACzC,MAAM,CAAC,GAAG,CAAC,UAAU,EAAE,UAAU,CAAC,CAAC;YACnC,MAAM,CAAC,GAAG,CAAC,aAAa,EAAE,MAAM,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC;YACnD,MAAM,CAAC,GAAG,CAAC,UAAU,EAAE,GAAG,CAAC,CAAC;QAC9B,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,QAAQ,IAAI,MAAM,EAAE,EAAE;YACpD,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,aAAa,EAAE,SAAS,IAAI,CAAC,MAAM,EAAE;gBACrC,cAAc,EAAE,WAAW;aAC5B;YACD,IAAI,EAAE,IAAI,UAAU,CAAC,KAAK,CAAC;YAC3B,MAAM,EAAE,WAAW,CAAC,OAAO,CAAC,MAAM,CAAC;SACpC,CAAC,CAAC;QAEH,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;YACnC,MAAM,IAAI,KAAK,CAAC,sBAAsB,QAAQ,CAAC,MAAM,KAAK,IAAI,EAAE,CAAC,CAAC;QACpE,CAAC;QAED,MAAM,IAAI,GAAG,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAA0B,CAAC;QAC9D,OAAO,IAAI,CAAC,gBAAgB,CAAC,IAAI,CAAC,CAAC;IACrC,CAAC;IAED,YAAY,CAAC,OAA0B;QACrC,OAAO,IAAI,iBAAiB,CAAC,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,KAAK,EAAE,IAAI,CAAC,QAAQ,EAAE;YACnE,GAAG,OAAO;YACV,UAAU,EAAE,OAAO,EAAE,UAAU,IAAI,IAAI,CAAC,UAAU;SACnD,CAAC,CAAC;IACL,CAAC;IAEO,gBAAgB,CAAC,IAA2B;QAClD,MAAM,GAAG,GAAG,IAAI,CAAC,OAAO,EAAE,QAAQ,EAAE,CAAC,CAAC,CAAC,EAAE,YAAY,EAAE,CAAC,CAAC,CAAC,CAAC;QAE3D,MAAM,MAAM,GAAqB;YAC/B,IAAI,EAAE,GAAG,EAAE,UAAU,IAAI,EAAE;SAC5B,CAAC;QAEF,IAAI,IAAI,CAAC,QAAQ,EAAE,QAAQ,KAAK,SAAS,EAAE,CAAC;YAC1C,MAAM,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC;QAC3C,CAAC;QAED,MAAM,KAAK,GAAG,QAAQ,CAAC,GAAG,EAAE,KAAK,CAAC,CAAC;QACnC,IAAI,KAAK;YAAE,MAAM,CAAC,KAAK,GAAG,KAAK,CAAC;QAEhC,OAAO,MAAM,CAAC;IAChB,CAAC;CACF;AAED,MAAM,gBAAgB,GAAG,MAAM,CAAC;AAEhC,MAAM,iBAAkB,SAAQ,YAAY;IAClC,EAAE,CAAY;IACd,KAAK,GAAG,KAAK,CAAC;IACd,MAAM,GAAG,KAAK,CAAC;IACf,YAAY,GAAG,KAAK,CAAC;IACrB,YAAY,GAAqC,IAAI,CAAC;IACtD,aAAa,GAAa,EAAE,CAAC;IAC7B,QAAQ,GAAa,EAAE,CAAC;IACxB,KAAK,GAA2C,EAAE,CAAC;IACnD,QAAQ,GAAG,CAAC,CAAC;IAErB,YACE,MAAc,EACd,KAAa,EACb,QAA4B,EAC5B,OAAkD;QAElD,KAAK,EAAE,CAAC;QAER,MAAM,MAAM,GAAG,IAAI,eAAe,CAAC;YACjC,KAAK;YACL,SAAS,EAAE,MAAM;YACjB,eAAe,EAAE,MAAM,CAAC,OAAO,CAAC,cAAc,IAAI,IAAI,CAAC;YACvD,QAAQ,EAAE,UAAU;YACpB,WAAW,EAAE,MAAM,CAAC,OAAO,CAAC,UAAU,CAAC;YACvC,QAAQ,EAAE,GAAG;SACd,CAAC,CAAC;QAEH,MAAM,IAAI,GAAG,OAAO,CAAC,QAAQ,IAAI,QAAQ,CAAC;QAC1C,IAAI,IAAI;YAAE,MAAM,CAAC,GAAG,CAAC,UAAU,EAAE,IAAI,CAAC,CAAC;QACvC,IAAI,OAAO,CAAC,WAAW,KAAK,SAAS,EAAE,CAAC;YACtC,MAAM,CAAC,GAAG,CAAC,aAAa,EAAE,MAAM,CAAC,OAAO,CAAC,WAAW,CAAC,CAAC,CAAC;QACzD,CAAC;QAED,IAAI,CAAC,EAAE,GAAG,IAAI,SAAS,CAAC,GAAG,MAAM,IAAI,MAAM,EAAE,EAAE;YAC7C,OAAO,EAAE,EAAE,aAAa,EAAE,SAAS,MAAM,EAAE,EAAE;SAC9C,CAAC,CAAC;QAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,MAAM,EAAE,GAAG,EAAE;YACtB,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC;YAClB,KAAK,MAAM,KAAK,IAAI,IAAI,CAAC,aAAa,EAAE,CAAC;gBACvC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YACtB,CAAC;YACD,IAAI,CAAC,aAAa,GAAG,EAAE,CAAC;QAC1B,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,GAAsB,EAAE,EAAE;YAC/C,IAAI,GAA0B,CAAC;YAC/B,IAAI,CAAC;gBACH,GAAG,GAAG,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAA0B,CAAC;YACzD,CAAC;YAAC,MAAM,CAAC;gBACP,OAAO;YACT,CAAC;YACD,IAAI,GAAG,CAAC,IAAI,KAAK,SAAS,IAAI,GAAG,CAAC,IAAI,KAAK,SAAS;gBAAE,OAAO;YAE7D,MAAM,GAAG,GAAG,GAAG,CAAC,OAAO,EAAE,YAAY,EAAE,CAAC,CAAC,CAAC,CAAC;YAC3C,IAAI,CAAC,GAAG;gBAAE,OAAO;YAEjB,IAAI,GAAG,CAAC,QAAQ,EAAE,CAAC;gBACjB,IAAI,OAAO,GAAG,CAAC,KAAK,KAAK,QAAQ,IAAI,OAAO,GAAG,CAAC,QAAQ,KAAK,QAAQ,EAAE,CAAC;oBACtE,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,GAAG,GAAG,CAAC,QAAQ,CAAC,CAAC;gBACpE,CAAC;gBACD,IAAI,CAAC,GAAG,CAAC,UAAU;oBAAE,OAAO;gBAC5B,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,UAAU,CAAC,CAAC;gBACnC,MAAM,KAAK,GAAG,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;gBAClC,IAAI,KAAK;oBAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,KAAK,CAAC,CAAC;gBAErC,MAAM,OAAO,GAAqB,EAAE,IAAI,EAAE,GAAG,CAAC,UAAU,EAAE,CAAC;gBAC3D,IAAI,KAAK;oBAAE,OAAO,CAAC,KAAK,GAAG,KAAK,CAAC;gBACjC,IAAI,OAAO,GAAG,CAAC,QAAQ,KAAK,QAAQ;oBAAE,OAAO,CAAC,QAAQ,GAAG,GAAG,CAAC,QAAQ,CAAC;gBACtE,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;YAC9B,CAAC;iBAAM,IAAI,GAAG,CAAC,UAAU,EAAE,CAAC;gBAC1B,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,GAAG,CAAC,UAAU,CAAC,CAAC;YACvC,CAAC;QACH,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAU,EAAE,EAAE;YACjC,IAAI,IAAI,CAAC,aAAa,CAAC,OAAO,CAAC,GAAG,CAAC;gBAAE,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;QAC/D,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,IAAY,EAAE,MAAc,EAAE,EAAE;YACnD,IAAI,CAAC,KAAK,GAAG,KAAK,CAAC;YACnB,IAAI,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC;gBACjB,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC;gBACzB,MAAM,OAAO,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC;gBAC1C,IAAI,CAAC,aAAa,GAAG,EAAE,CAAC;gBACxB,IAAI,IAAI,CAAC,aAAa,CAAC,OAAO,CAAC,GAAG,CAAC,EAAE,CAAC;oBACpC,MAAM,MAAM,GAAG,MAAM,CAAC,QAAQ,EAAE,IAAI,WAAW,CAAC;oBAChD,IAAI,CAAC,IAAI,CACP,OAAO,EACP,IAAI,KAAK,CACP,2DAA2D,IAAI,KAAK,MAAM,GAAG;wBAC3E,CAAC,OAAO,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,OAAO,mBAAmB,CAAC,CAAC,CAAC,EAAE,CAAC,CACvD,CACF,CAAC;gBACJ,CAAC;YACH,CAAC;QACH,CAAC,CAAC,CAAC;IACL,CAAC;IAED,KAAK,CAAC,KAAa;QACjB,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC;YAChB,MAAM,IAAI,KAAK,CAAC,6CAA6C,CAAC,CAAC;QACjE,CAAC;QACD,IAAI,IAAI,CAAC,YAAY;YAAE,OAAO;QAC9B,IAAI,IAAI,CAAC,KAAK,EAAE,CAAC;YACf,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACtB,CAAC;aAAM,CAAC;YACN,IAAI,CAAC,aAAa,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACjC,CAAC;IACH,CAAC;IAED,KAAK,CAAC,KAAK;QACT,IAAI,IAAI,CAAC,YAAY;YAAE,OAAO,IAAI,CAAC,YAAY,CAAC;QAChD,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC;QAEnB,IAAI,CAAC,YAAY,GAAG,IAAI,OAAO,CAAmB,CAAC,OAAO,EAAE,EAAE;YAC5D,MAAM,KAAK,GAAG,IAAI,CAAC,EAAE,CAAC,UAAU,CAAC;YACjC,IAAI,KAAK,KAAK,SAAS,CAAC,MAAM,IAAI,KAAK,KAAK,SAAS,CAAC,OAAO,EAAE,CAAC;gBAC9D,OAAO,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC;gBAC5B,OAAO;YACT,CAAC;YAED,MAAM,OAAO,GAAG,UAAU,CAAC,GAAG,EAAE;gBAC9B,IAAI,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC;gBACpB,OAAO,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC;YAC9B,CAAC,EAAE,gBAAgB,CAAC,CAAC;YAErB,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,EAAE;gBACzB,YAAY,CAAC,OAAO,CAAC,CAAC;gBACtB,OAAO,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC;YAC9B,CAAC,CAAC,CAAC;YAEH,IAAI,KAAK,KAAK,SAAS,CAAC,IAAI,EAAE,CAAC;gBAC7B,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,IAAI,EAAE,aAAa,EAAE,CAAC,CAAC,CAAC;gBACtD,OAAO;YACT,CAAC;YAED,IAAI,IAAI,CAAC,aAAa,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;gBACpC,YAAY,CAAC,OAAO,CAAC,CAAC;gBACtB,IAAI,CAAC,EAAE,CAAC,kBAAkB,CAAC,OAAO,CAAC,CAAC;gBACpC,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;gBAC9B,IAAI,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC;gBACpB,OAAO,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC;gBAC5B,OAAO;YACT,CAAC;YAED,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,MAAM,EAAE,GAAG,EAAE;gBACxB,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,IAAI,EAAE,aAAa,EAAE,CAAC,CAAC,CAAC;YACxD,CAAC,CAAC,CAAC;QACL,CAAC,CAAC,CAAC;QAEH,OAAO,IAAI,CAAC,YAAY,CAAC;IAC3B,CAAC;IAEO,WAAW;QACjB,MAAM,MAAM,GAAqB,EAAE,IAAI,EAAE,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC;QAC1E,IAAI,IAAI,CAAC,QAAQ,GAAG,CAAC;YAAE,MAAM,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,CAAC;QACvD,IAAI,IAAI,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC;YAAE,MAAM,CAAC,KAAK,GAAG,CAAC,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC;QAC1D,OAAO,MAAM,CAAC;IAChB,CAAC;CACF"}
|
package/dist/stt/openai-stt.d.ts
CHANGED
|
@@ -9,8 +9,13 @@ export declare class OpenAISTT implements STTProvider {
|
|
|
9
9
|
private readonly client;
|
|
10
10
|
private readonly model;
|
|
11
11
|
constructor(config: OpenAISTTConfig);
|
|
12
|
+
/**
|
|
13
|
+
* Transcribe a complete audio file (wav, mp3, ogg, flac, webm, mp4). Headerless input is
|
|
14
|
+
* treated as raw PCM16 mono at 16kHz and wrapped into a WAV container.
|
|
15
|
+
*/
|
|
12
16
|
transcribe(audio: Buffer, options?: STTOptions): Promise<TranscribeResult>;
|
|
13
17
|
createStream(options?: STTStreamOptions): STTStream;
|
|
18
|
+
private transcribeFile;
|
|
14
19
|
private mapResponse;
|
|
15
20
|
}
|
|
16
21
|
//# sourceMappingURL=openai-stt.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"openai-stt.d.ts","sourceRoot":"","sources":["../../src/stt/openai-stt.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"openai-stt.d.ts","sourceRoot":"","sources":["../../src/stt/openai-stt.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EACV,WAAW,EACX,UAAU,EACV,gBAAgB,EAChB,SAAS,EACT,gBAAgB,EACjB,MAAM,aAAa,CAAC;AAErB,MAAM,WAAW,eAAe;IAC9B,MAAM,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,OAAO,CAAC,EAAE,MAAM,CAAC;CAClB;AA0BD,qBAAa,SAAU,YAAW,WAAW;IAC3C,QAAQ,CAAC,IAAI,YAAY;IAEzB,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAS;IAChC,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAS;gBAEnB,MAAM,EAAE,eAAe;IAQnC;;;OAGG;IACG,UAAU,CAAC,KAAK,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,UAAU,GAAG,OAAO,CAAC,gBAAgB,CAAC;IAIhF,YAAY,CAAC,OAAO,CAAC,EAAE,gBAAgB,GAAG,SAAS;YAYrC,cAAc;IA0B5B,OAAO,CAAC,WAAW;CAwBpB"}
|
package/dist/stt/openai-stt.js
CHANGED
|
@@ -1,6 +1,28 @@
|
|
|
1
1
|
import { EventEmitter } from 'node:events';
|
|
2
2
|
import OpenAI from 'openai';
|
|
3
|
+
import { audioMimeType, detectAudioFormat, pcmToWav } from '../audio.js';
|
|
3
4
|
const DEFAULT_MODEL = 'gpt-4o-mini-transcribe';
|
|
5
|
+
const DEFAULT_SAMPLE_RATE = 16000;
|
|
6
|
+
function supportsVerboseJson(model) {
|
|
7
|
+
return model.startsWith('whisper');
|
|
8
|
+
}
|
|
9
|
+
function toUploadFile(audio, sampleRate) {
|
|
10
|
+
const format = detectAudioFormat(audio);
|
|
11
|
+
if (!format) {
|
|
12
|
+
return new File([new Uint8Array(pcmToWav(audio, sampleRate))], 'audio.wav', {
|
|
13
|
+
type: 'audio/wav',
|
|
14
|
+
});
|
|
15
|
+
}
|
|
16
|
+
return new File([new Uint8Array(audio)], `audio.${format}`, { type: audioMimeType(format) });
|
|
17
|
+
}
|
|
18
|
+
function hasContainerHeader(audio) {
|
|
19
|
+
const format = detectAudioFormat(audio);
|
|
20
|
+
if (format === null)
|
|
21
|
+
return false;
|
|
22
|
+
if (format === 'mp3')
|
|
23
|
+
return audio.toString('ascii', 0, 3) === 'ID3';
|
|
24
|
+
return true;
|
|
25
|
+
}
|
|
4
26
|
export class OpenAISTT {
|
|
5
27
|
name = 'openai';
|
|
6
28
|
client;
|
|
@@ -12,20 +34,47 @@ export class OpenAISTT {
|
|
|
12
34
|
});
|
|
13
35
|
this.model = config.model ?? DEFAULT_MODEL;
|
|
14
36
|
}
|
|
37
|
+
/**
|
|
38
|
+
* Transcribe a complete audio file (wav, mp3, ogg, flac, webm, mp4). Headerless input is
|
|
39
|
+
* treated as raw PCM16 mono at 16kHz and wrapped into a WAV container.
|
|
40
|
+
*/
|
|
15
41
|
async transcribe(audio, options) {
|
|
16
|
-
|
|
17
|
-
|
|
42
|
+
return this.transcribeFile(toUploadFile(audio, DEFAULT_SAMPLE_RATE), options);
|
|
43
|
+
}
|
|
44
|
+
createStream(options) {
|
|
45
|
+
const sampleRate = options?.sampleRate ?? DEFAULT_SAMPLE_RATE;
|
|
46
|
+
return new OpenAISTTStream((audio, opts) => {
|
|
47
|
+
const file = hasContainerHeader(audio)
|
|
48
|
+
? toUploadFile(audio, sampleRate)
|
|
49
|
+
: new File([new Uint8Array(pcmToWav(audio, sampleRate))], 'audio.wav', {
|
|
50
|
+
type: 'audio/wav',
|
|
51
|
+
});
|
|
52
|
+
return this.transcribeFile(file, opts);
|
|
53
|
+
}, options);
|
|
54
|
+
}
|
|
55
|
+
async transcribeFile(file, options) {
|
|
56
|
+
const common = {
|
|
18
57
|
file,
|
|
19
58
|
model: this.model,
|
|
20
|
-
response_format: 'verbose_json',
|
|
21
|
-
timestamp_granularities: ['word'],
|
|
22
59
|
...(options?.language && { language: options.language }),
|
|
23
60
|
...(options?.prompt && { prompt: options.prompt }),
|
|
61
|
+
};
|
|
62
|
+
if (supportsVerboseJson(this.model)) {
|
|
63
|
+
const response = await this.client.audio.transcriptions.create({
|
|
64
|
+
...common,
|
|
65
|
+
response_format: 'verbose_json',
|
|
66
|
+
timestamp_granularities: ['word'],
|
|
67
|
+
});
|
|
68
|
+
return this.mapResponse(response);
|
|
69
|
+
}
|
|
70
|
+
const response = await this.client.audio.transcriptions.create({
|
|
71
|
+
...common,
|
|
72
|
+
response_format: 'json',
|
|
24
73
|
});
|
|
25
|
-
|
|
26
|
-
|
|
27
|
-
|
|
28
|
-
return
|
|
74
|
+
const result = { text: response.text };
|
|
75
|
+
if (options?.language)
|
|
76
|
+
result.language = options.language;
|
|
77
|
+
return result;
|
|
29
78
|
}
|
|
30
79
|
mapResponse(response) {
|
|
31
80
|
const result = {
|
|
@@ -51,6 +100,7 @@ export class OpenAISTT {
|
|
|
51
100
|
class OpenAISTTStream extends EventEmitter {
|
|
52
101
|
chunks = [];
|
|
53
102
|
closed = false;
|
|
103
|
+
closePromise = null;
|
|
54
104
|
transcribeFn;
|
|
55
105
|
options;
|
|
56
106
|
constructor(transcribeFn, options) {
|
|
@@ -64,13 +114,19 @@ class OpenAISTTStream extends EventEmitter {
|
|
|
64
114
|
}
|
|
65
115
|
this.chunks.push(chunk);
|
|
66
116
|
}
|
|
67
|
-
|
|
68
|
-
if (this.
|
|
69
|
-
|
|
117
|
+
close() {
|
|
118
|
+
if (!this.closePromise) {
|
|
119
|
+
this.closed = true;
|
|
120
|
+
this.closePromise = this.finish();
|
|
70
121
|
}
|
|
71
|
-
this.
|
|
122
|
+
return this.closePromise;
|
|
123
|
+
}
|
|
124
|
+
async finish() {
|
|
72
125
|
const combined = Buffer.concat(this.chunks);
|
|
73
126
|
this.chunks = [];
|
|
127
|
+
if (combined.length === 0) {
|
|
128
|
+
return { text: '' };
|
|
129
|
+
}
|
|
74
130
|
try {
|
|
75
131
|
const result = await this.transcribeFn(combined, this.options);
|
|
76
132
|
this.emit('final', result);
|
|
@@ -78,7 +134,8 @@ class OpenAISTTStream extends EventEmitter {
|
|
|
78
134
|
}
|
|
79
135
|
catch (err) {
|
|
80
136
|
const error = err instanceof Error ? err : new Error(String(err));
|
|
81
|
-
this.
|
|
137
|
+
if (this.listenerCount('error') > 0)
|
|
138
|
+
this.emit('error', error);
|
|
82
139
|
throw error;
|
|
83
140
|
}
|
|
84
141
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"openai-stt.js","sourceRoot":"","sources":["../../src/stt/openai-stt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,MAAM,MAAM,QAAQ,CAAC;
|
|
1
|
+
{"version":3,"file":"openai-stt.js","sourceRoot":"","sources":["../../src/stt/openai-stt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,MAAM,MAAM,QAAQ,CAAC;AAE5B,OAAO,EAAE,aAAa,EAAE,iBAAiB,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AAezE,MAAM,aAAa,GAAG,wBAAwB,CAAC;AAC/C,MAAM,mBAAmB,GAAG,KAAK,CAAC;AAElC,SAAS,mBAAmB,CAAC,KAAa;IACxC,OAAO,KAAK,CAAC,UAAU,CAAC,SAAS,CAAC,CAAC;AACrC,CAAC;AAED,SAAS,YAAY,CAAC,KAAa,EAAE,UAAkB;IACrD,MAAM,MAAM,GAAG,iBAAiB,CAAC,KAAK,CAAC,CAAC;IACxC,IAAI,CAAC,MAAM,EAAE,CAAC;QACZ,OAAO,IAAI,IAAI,CAAC,CAAC,IAAI,UAAU,CAAC,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC,CAAC,CAAC,EAAE,WAAW,EAAE;YAC1E,IAAI,EAAE,WAAW;SAClB,CAAC,CAAC;IACL,CAAC;IACD,OAAO,IAAI,IAAI,CAAC,CAAC,IAAI,UAAU,CAAC,KAAK,CAAC,CAAC,EAAE,SAAS,MAAM,EAAE,EAAE,EAAE,IAAI,EAAE,aAAa,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC;AAC/F,CAAC;AAED,SAAS,kBAAkB,CAAC,KAAa;IACvC,MAAM,MAAM,GAAG,iBAAiB,CAAC,KAAK,CAAC,CAAC;IACxC,IAAI,MAAM,KAAK,IAAI;QAAE,OAAO,KAAK,CAAC;IAClC,IAAI,MAAM,KAAK,KAAK;QAAE,OAAO,KAAK,CAAC,QAAQ,CAAC,OAAO,EAAE,CAAC,EAAE,CAAC,CAAC,KAAK,KAAK,CAAC;IACrE,OAAO,IAAI,CAAC;AACd,CAAC;AAED,MAAM,OAAO,SAAS;IACX,IAAI,GAAG,QAAQ,CAAC;IAER,MAAM,CAAS;IACf,KAAK,CAAS;IAE/B,YAAY,MAAuB;QACjC,IAAI,CAAC,MAAM,GAAG,IAAI,MAAM,CAAC;YACvB,MAAM,EAAE,MAAM,CAAC,MAAM;YACrB,GAAG,CAAC,MAAM,CAAC,OAAO,IAAI,EAAE,OAAO,EAAE,MAAM,CAAC,OAAO,EAAE,CAAC;SACnD,CAAC,CAAC;QACH,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,IAAI,aAAa,CAAC;IAC7C,CAAC;IAED;;;OAGG;IACH,KAAK,CAAC,UAAU,CAAC,KAAa,EAAE,OAAoB;QAClD,OAAO,IAAI,CAAC,cAAc,CAAC,YAAY,CAAC,KAAK,EAAE,mBAAmB,CAAC,EAAE,OAAO,CAAC,CAAC;IAChF,CAAC;IAED,YAAY,CAAC,OAA0B;QACrC,MAAM,UAAU,GAAG,OAAO,EAAE,UAAU,IAAI,mBAAmB,CAAC;QAC9D,OAAO,IAAI,eAAe,CAAC,CAAC,KAAK,EAAE,IAAI,EAAE,EAAE;YACzC,MAAM,IAAI,GAAG,kBAAkB,CAAC,KAAK,CAAC;gBACpC,CAAC,CAAC,YAAY,CAAC,KAAK,EAAE,UAAU,CAAC;gBACjC,CAAC,CAAC,IAAI,IAAI,CAAC,CAAC,IAAI,UAAU,CAAC,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC,CAAC,CAAC,EAAE,WAAW,EAAE;oBACnE,IAAI,EAAE,WAAW;iBAClB,CAAC,CAAC;YACP,OAAO,IAAI,CAAC,cAAc,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC;QACzC,CAAC,EAAE,OAAO,CAAC,CAAC;IACd,CAAC;IAEO,KAAK,CAAC,cAAc,CAAC,IAAU,EAAE,OAAoB;QAC3D,MAAM,MAAM,GAAG;YACb,IAAI;YACJ,KAAK,EAAE,IAAI,CAAC,KAAK;YACjB,GAAG,CAAC,OAAO,EAAE,QAAQ,IAAI,EAAE,QAAQ,EAAE,OAAO,CAAC,QAAQ,EAAE,CAAC;YACxD,GAAG,CAAC,OAAO,EAAE,MAAM,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,CAAC;SACnD,CAAC;QAEF,IAAI,mBAAmB,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC;YACpC,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,cAAc,CAAC,MAAM,CAAC;gBAC7D,GAAG,MAAM;gBACT,eAAe,EAAE,cAAc;gBAC/B,uBAAuB,EAAE,CAAC,MAAM,CAAC;aAClC,CAAC,CAAC;YACH,OAAO,IAAI,CAAC,WAAW,CAAC,QAAQ,CAAC,CAAC;QACpC,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,cAAc,CAAC,MAAM,CAAC;YAC7D,GAAG,MAAM;YACT,eAAe,EAAE,MAAM;SACxB,CAAC,CAAC;QACH,MAAM,MAAM,GAAqB,EAAE,IAAI,EAAE,QAAQ,CAAC,IAAI,EAAE,CAAC;QACzD,IAAI,OAAO,EAAE,QAAQ;YAAE,MAAM,CAAC,QAAQ,GAAG,OAAO,CAAC,QAAQ,CAAC;QAC1D,OAAO,MAAM,CAAC;IAChB,CAAC;IAEO,WAAW,CAAC,QAA8B;QAChD,MAAM,MAAM,GAAqB;YAC/B,IAAI,EAAE,QAAQ,CAAC,IAAI;SACpB,CAAC;QAEF,IAAI,QAAQ,CAAC,QAAQ,EAAE,CAAC;YACtB,MAAM,CAAC,QAAQ,GAAG,QAAQ,CAAC,QAAQ,CAAC;QACtC,CAAC;QAED,IAAI,QAAQ,CAAC,QAAQ,KAAK,SAAS,EAAE,CAAC;YACpC,MAAM,CAAC,QAAQ,GAAG,QAAQ,CAAC,QAAQ,CAAC;QACtC,CAAC;QAED,IAAI,QAAQ,CAAC,KAAK,IAAI,QAAQ,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YAChD,MAAM,CAAC,KAAK,GAAG,QAAQ,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;gBACxC,IAAI,EAAE,CAAC,CAAC,IAAI;gBACZ,KAAK,EAAE,CAAC,CAAC,KAAK;gBACd,GAAG,EAAE,CAAC,CAAC,GAAG;gBACV,UAAU,EAAE,CAAC;aACd,CAAC,CAAC,CAAC;QACN,CAAC;QAED,OAAO,MAAM,CAAC;IAChB,CAAC;CACF;AAID,MAAM,eAAgB,SAAQ,YAAY;IAChC,MAAM,GAAa,EAAE,CAAC;IACtB,MAAM,GAAG,KAAK,CAAC;IACf,YAAY,GAAqC,IAAI,CAAC;IAC7C,YAAY,CAAe;IAC3B,OAAO,CAAoB;IAE5C,YAAY,YAA0B,EAAE,OAA0B;QAChE,KAAK,EAAE,CAAC;QACR,IAAI,CAAC,YAAY,GAAG,YAAY,CAAC;QACjC,IAAI,CAAC,OAAO,GAAG,OAAO,CAAC;IACzB,CAAC;IAED,KAAK,CAAC,KAAa;QACjB,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC;YAChB,MAAM,IAAI,KAAK,CAAC,2CAA2C,CAAC,CAAC;QAC/D,CAAC;QACD,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;IAC1B,CAAC;IAED,KAAK;QACH,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,CAAC;YACvB,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC;YACnB,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC;QACpC,CAAC;QACD,OAAO,IAAI,CAAC,YAAY,CAAC;IAC3B,CAAC;IAEO,KAAK,CAAC,MAAM;QAClB,MAAM,QAAQ,GAAG,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QAC5C,IAAI,CAAC,MAAM,GAAG,EAAE,CAAC;QAEjB,IAAI,QAAQ,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YAC1B,OAAO,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC;QACtB,CAAC;QAED,IAAI,CAAC;YACH,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,EAAE,IAAI,CAAC,OAAO,CAAC,CAAC;YAC/D,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC;YAC3B,OAAO,MAAM,CAAC;QAChB,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,MAAM,KAAK,GAAG,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC;YAClE,IAAI,IAAI,CAAC,aAAa,CAAC,OAAO,CAAC,GAAG,CAAC;gBAAE,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC;YAC/D,MAAM,KAAK,CAAC;QACd,CAAC;IACH,CAAC;CACF"}
|
package/dist/tools.d.ts
CHANGED
|
@@ -15,12 +15,12 @@ declare const SpeakParamsSchema: z.ZodObject<{
|
|
|
15
15
|
text: z.ZodString;
|
|
16
16
|
voice: z.ZodOptional<z.ZodString>;
|
|
17
17
|
format: z.ZodOptional<z.ZodEnum<{
|
|
18
|
-
|
|
18
|
+
wav: "wav";
|
|
19
19
|
mp3: "mp3";
|
|
20
|
+
flac: "flac";
|
|
21
|
+
pcm16: "pcm16";
|
|
20
22
|
opus: "opus";
|
|
21
23
|
aac: "aac";
|
|
22
|
-
flac: "flac";
|
|
23
|
-
wav: "wav";
|
|
24
24
|
}>>;
|
|
25
25
|
}, z.core.$strip>;
|
|
26
26
|
type SpeakParams = z.infer<typeof SpeakParamsSchema>;
|
|
@@ -2,12 +2,8 @@ import { EventEmitter } from 'node:events';
|
|
|
2
2
|
import http from 'node:http';
|
|
3
3
|
import { WebSocket } from 'ws';
|
|
4
4
|
import type { WebSocketTransportConfig } from '../types.js';
|
|
5
|
-
|
|
6
|
-
|
|
7
|
-
code: number;
|
|
8
|
-
message: string;
|
|
9
|
-
};
|
|
10
|
-
}
|
|
5
|
+
/** @deprecated Use `WebSocketTransportConfig`. */
|
|
6
|
+
export type WebSocketTransportOptions = WebSocketTransportConfig;
|
|
11
7
|
interface VoiceClientEvents {
|
|
12
8
|
audio: [chunk: Buffer];
|
|
13
9
|
message: [msg: Record<string, unknown>];
|
|
@@ -25,6 +21,7 @@ export declare class VoiceClient extends EventEmitter<VoiceClientEvents> {
|
|
|
25
21
|
sendAudio(chunk: Buffer): void;
|
|
26
22
|
sendMessage(msg: Record<string, unknown>): void;
|
|
27
23
|
close(code?: number, reason?: string): void;
|
|
24
|
+
terminate(): void;
|
|
28
25
|
}
|
|
29
26
|
interface TransportEvents {
|
|
30
27
|
connection: [client: VoiceClient];
|
|
@@ -39,12 +36,14 @@ export declare class WebSocketTransport extends EventEmitter<TransportEvents> {
|
|
|
39
36
|
private clients;
|
|
40
37
|
private upgradeHandler;
|
|
41
38
|
private pingInterval;
|
|
42
|
-
|
|
39
|
+
private pendingUpgrades;
|
|
40
|
+
constructor(config?: WebSocketTransportConfig);
|
|
43
41
|
listen(port: number): Promise<void>;
|
|
44
42
|
attachToServer(server: http.Server): void;
|
|
45
43
|
get port(): number | undefined;
|
|
46
44
|
close(): Promise<void>;
|
|
47
45
|
private setupWss;
|
|
46
|
+
private handleUpgrade;
|
|
48
47
|
private startKeepalive;
|
|
49
48
|
private stopKeepalive;
|
|
50
49
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"ws-transport.d.ts","sourceRoot":"","sources":["../../src/transport/ws-transport.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,IAAI,MAAM,WAAW,CAAC;AAE7B,OAAO,EAAmB,SAAS,EAAE,MAAM,IAAI,CAAC;AAEhD,OAAO,KAAK,EAAE,wBAAwB,EAAE,MAAM,aAAa,CAAC;AAE5D,
|
|
1
|
+
{"version":3,"file":"ws-transport.d.ts","sourceRoot":"","sources":["../../src/transport/ws-transport.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,IAAI,MAAM,WAAW,CAAC;AAE7B,OAAO,EAAmB,SAAS,EAAE,MAAM,IAAI,CAAC;AAEhD,OAAO,KAAK,EAAE,wBAAwB,EAAE,MAAM,aAAa,CAAC;AAE5D,kDAAkD;AAClD,MAAM,MAAM,yBAAyB,GAAG,wBAAwB,CAAC;AAWjE,UAAU,iBAAiB;IACzB,KAAK,EAAE,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;IACvB,OAAO,EAAE,CAAC,GAAG,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC;IACxC,KAAK,EAAE,EAAE,CAAC;IACV,KAAK,EAAE,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;CACvB;AAED,qBAAa,WAAY,SAAQ,YAAY,CAAC,iBAAiB,CAAC;IAIlD,OAAO,CAAC,QAAQ,CAAC,EAAE;IAH/B,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;IACpB,OAAO,CAAC,KAAK,CAAQ;gBAEQ,EAAE,EAAE,SAAS;IA+B1C,IAAI,OAAO,IAAI,OAAO,CAErB;IAED,QAAQ,IAAI,IAAI;IAIhB,IAAI,IAAI,IAAI;IAMZ,SAAS,CAAC,KAAK,EAAE,MAAM,GAAG,IAAI;IAM9B,WAAW,CAAC,GAAG,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,IAAI;IAM/C,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,IAAI;IAI3C,SAAS,IAAI,IAAI;CAGlB;AAED,UAAU,eAAe;IACvB,UAAU,EAAE,CAAC,MAAM,EAAE,WAAW,CAAC,CAAC;CACnC;AAED,qBAAa,kBAAmB,SAAQ,YAAY,CAAC,eAAe,CAAC;IACnE,OAAO,CAAC,QAAQ,CAAC,IAAI,CAAS;IAC9B,OAAO,CAAC,QAAQ,CAAC,cAAc,CAAS;IACxC,OAAO,CAAC,QAAQ,CAAC,YAAY,CAAC,CAA2C;IACzE,OAAO,CAAC,GAAG,CAAgC;IAC3C,OAAO,CAAC,MAAM,CAA4B;IAC1C,OAAO,CAAC,UAAU,CAAS;IAC3B,OAAO,CAAC,OAAO,CAA0B;IACzC,OAAO,CAAC,cAAc,CAEN;IAChB,OAAO,CAAC,YAAY,CAA+C;IACnE,OAAO,CAAC,eAAe,CAAK;gBAEhB,MAAM,CAAC,EAAE,wBAAwB;IAOvC,MAAM,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IAczC,cAAc,CAAC,MAAM,EAAE,IAAI,CAAC,MAAM,GAAG,IAAI;IAOzC,IAAI,IAAI,IAAI,MAAM,GAAG,SAAS,CAI7B;IAEK,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IA2B5B,OAAO,CAAC,QAAQ;YA0BF,aAAa;IAgD3B,OAAO,CAAC,cAAc;IAetB,OAAO,CAAC,aAAa;CAMtB"}
|