@cogitator-ai/voice 0.1.0 → 0.1.4
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/audio.d.ts +10 -0
- package/dist/audio.d.ts.map +1 -0
- package/dist/audio.js +75 -0
- package/dist/audio.js.map +1 -0
- package/dist/index.d.ts +17 -0
- package/dist/index.d.ts.map +1 -0
- package/dist/index.js +16 -0
- package/dist/index.js.map +1 -0
- package/dist/pipeline/index.d.ts +3 -0
- package/dist/pipeline/index.d.ts.map +1 -0
- package/dist/pipeline/index.js +3 -0
- package/dist/pipeline/index.js.map +1 -0
- package/dist/pipeline/pipeline-session.d.ts +21 -0
- package/dist/pipeline/pipeline-session.d.ts.map +1 -0
- package/dist/pipeline/pipeline-session.js +131 -0
- package/dist/pipeline/pipeline-session.js.map +1 -0
- package/dist/pipeline/voice-pipeline.d.ts +13 -0
- package/dist/pipeline/voice-pipeline.d.ts.map +1 -0
- package/dist/pipeline/voice-pipeline.js +17 -0
- package/dist/pipeline/voice-pipeline.js.map +1 -0
- package/dist/realtime/gemini-realtime.d.ts +28 -0
- package/dist/realtime/gemini-realtime.d.ts.map +1 -0
- package/dist/realtime/gemini-realtime.js +153 -0
- package/dist/realtime/gemini-realtime.js.map +1 -0
- package/dist/realtime/index.d.ts +4 -0
- package/dist/realtime/index.d.ts.map +1 -0
- package/dist/realtime/index.js +4 -0
- package/dist/realtime/index.js.map +1 -0
- package/dist/realtime/openai-realtime.d.ts +27 -0
- package/dist/realtime/openai-realtime.d.ts.map +1 -0
- package/dist/realtime/openai-realtime.js +138 -0
- package/dist/realtime/openai-realtime.js.map +1 -0
- package/dist/realtime/realtime-session.d.ts +23 -0
- package/dist/realtime/realtime-session.d.ts.map +1 -0
- package/dist/realtime/realtime-session.js +47 -0
- package/dist/realtime/realtime-session.js.map +1 -0
- package/dist/stt/deepgram-stt.d.ts +17 -0
- package/dist/stt/deepgram-stt.d.ts.map +1 -0
- package/dist/stt/deepgram-stt.js +138 -0
- package/dist/stt/deepgram-stt.js.map +1 -0
- package/dist/stt/index.d.ts +3 -0
- package/dist/stt/index.d.ts.map +1 -0
- package/dist/stt/index.js +3 -0
- package/dist/stt/index.js.map +1 -0
- package/dist/stt/openai-stt.d.ts +17 -0
- package/dist/stt/openai-stt.d.ts.map +1 -0
- package/dist/stt/openai-stt.js +80 -0
- package/dist/stt/openai-stt.js.map +1 -0
- package/dist/tools.d.ts +26 -0
- package/dist/tools.d.ts.map +1 -0
- package/dist/tools.js +36 -0
- package/dist/tools.js.map +1 -0
- package/dist/transport/index.d.ts +2 -0
- package/dist/transport/index.d.ts.map +1 -0
- package/dist/transport/index.js +2 -0
- package/dist/transport/index.js.map +1 -0
- package/dist/transport/ws-transport.d.ts +37 -0
- package/dist/transport/ws-transport.d.ts.map +1 -0
- package/dist/transport/ws-transport.js +110 -0
- package/dist/transport/ws-transport.js.map +1 -0
- package/dist/tts/elevenlabs-tts.d.ts +16 -0
- package/dist/tts/elevenlabs-tts.d.ts.map +1 -0
- package/dist/tts/elevenlabs-tts.js +77 -0
- package/dist/tts/elevenlabs-tts.js.map +1 -0
- package/dist/tts/index.d.ts +3 -0
- package/dist/tts/index.d.ts.map +1 -0
- package/dist/tts/index.js +3 -0
- package/dist/tts/index.js.map +1 -0
- package/dist/tts/openai-tts.d.ts +18 -0
- package/dist/tts/openai-tts.d.ts.map +1 -0
- package/dist/tts/openai-tts.js +47 -0
- package/dist/tts/openai-tts.js.map +1 -0
- package/dist/types.d.ts +96 -0
- package/dist/types.d.ts.map +1 -0
- package/dist/types.js +2 -0
- package/dist/types.js.map +1 -0
- package/dist/vad/energy-vad.d.ts +19 -0
- package/dist/vad/energy-vad.d.ts.map +1 -0
- package/dist/vad/energy-vad.js +49 -0
- package/dist/vad/energy-vad.js.map +1 -0
- package/dist/vad/index.d.ts +3 -0
- package/dist/vad/index.d.ts.map +1 -0
- package/dist/vad/index.js +3 -0
- package/dist/vad/index.js.map +1 -0
- package/dist/vad/silero-vad.d.ts +27 -0
- package/dist/vad/silero-vad.d.ts.map +1 -0
- package/dist/vad/silero-vad.js +83 -0
- package/dist/vad/silero-vad.js.map +1 -0
- package/dist/voice-agent.d.ts +25 -0
- package/dist/voice-agent.d.ts.map +1 -0
- package/dist/voice-agent.js +139 -0
- package/dist/voice-agent.js.map +1 -0
- package/package.json +1 -1
|
@@ -0,0 +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;AAC7B,OAAO,EAAmB,SAAS,EAAE,MAAM,IAAI,CAAC;AAEhD,OAAO,KAAK,EAAE,wBAAwB,EAAE,MAAM,aAAa,CAAC;AAE5D,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;IAGlD,OAAO,CAAC,QAAQ,CAAC,EAAE;IAF/B,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;gBAES,EAAE,EAAE,SAAS;IAqB1C,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;CAG5C;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,GAAG,CAAgC;IAC3C,OAAO,CAAC,MAAM,CAA4B;IAC1C,OAAO,CAAC,UAAU,CAAS;IAC3B,OAAO,CAAC,OAAO,CAA0B;gBAE7B,MAAM,CAAC,EAAE,wBAAwB;IAMvC,MAAM,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IAOzC,cAAc,CAAC,MAAM,EAAE,IAAI,CAAC,MAAM,GAAG,IAAI;IAMzC,IAAI,IAAI,IAAI,MAAM,GAAG,SAAS,CAI7B;IAEK,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAmB5B,OAAO,CAAC,QAAQ;CA2BjB"}
|
|
@@ -0,0 +1,110 @@
|
|
|
1
|
+
import { EventEmitter } from 'node:events';
|
|
2
|
+
import http from 'node:http';
|
|
3
|
+
import { WebSocketServer, WebSocket } from 'ws';
|
|
4
|
+
import { nanoid } from 'nanoid';
|
|
5
|
+
export class VoiceClient extends EventEmitter {
|
|
6
|
+
ws;
|
|
7
|
+
id;
|
|
8
|
+
constructor(ws) {
|
|
9
|
+
super();
|
|
10
|
+
this.ws = ws;
|
|
11
|
+
this.id = nanoid();
|
|
12
|
+
ws.on('message', (data, isBinary) => {
|
|
13
|
+
if (isBinary) {
|
|
14
|
+
this.emit('audio', Buffer.from(data));
|
|
15
|
+
}
|
|
16
|
+
else {
|
|
17
|
+
try {
|
|
18
|
+
const parsed = JSON.parse(data.toString());
|
|
19
|
+
this.emit('message', parsed);
|
|
20
|
+
}
|
|
21
|
+
catch (err) {
|
|
22
|
+
this.emit('error', new Error(`Invalid JSON: ${err.message}`));
|
|
23
|
+
}
|
|
24
|
+
}
|
|
25
|
+
});
|
|
26
|
+
ws.on('close', () => this.emit('close'));
|
|
27
|
+
ws.on('error', (err) => this.emit('error', err));
|
|
28
|
+
}
|
|
29
|
+
sendAudio(chunk) {
|
|
30
|
+
if (this.ws.readyState === WebSocket.OPEN) {
|
|
31
|
+
this.ws.send(chunk);
|
|
32
|
+
}
|
|
33
|
+
}
|
|
34
|
+
sendMessage(msg) {
|
|
35
|
+
if (this.ws.readyState === WebSocket.OPEN) {
|
|
36
|
+
this.ws.send(JSON.stringify(msg));
|
|
37
|
+
}
|
|
38
|
+
}
|
|
39
|
+
close(code, reason) {
|
|
40
|
+
this.ws.close(code, reason);
|
|
41
|
+
}
|
|
42
|
+
}
|
|
43
|
+
export class WebSocketTransport extends EventEmitter {
|
|
44
|
+
path;
|
|
45
|
+
maxConnections;
|
|
46
|
+
wss = null;
|
|
47
|
+
server = null;
|
|
48
|
+
ownsServer = false;
|
|
49
|
+
clients = new Set();
|
|
50
|
+
constructor(config) {
|
|
51
|
+
super();
|
|
52
|
+
this.path = config?.path ?? '/voice';
|
|
53
|
+
this.maxConnections = config?.maxConnections ?? 100;
|
|
54
|
+
}
|
|
55
|
+
async listen(port) {
|
|
56
|
+
this.server = http.createServer();
|
|
57
|
+
this.ownsServer = true;
|
|
58
|
+
this.setupWss(this.server);
|
|
59
|
+
await new Promise((resolve) => this.server.listen(port, resolve));
|
|
60
|
+
}
|
|
61
|
+
attachToServer(server) {
|
|
62
|
+
this.server = server;
|
|
63
|
+
this.ownsServer = false;
|
|
64
|
+
this.setupWss(server);
|
|
65
|
+
}
|
|
66
|
+
get port() {
|
|
67
|
+
const addr = this.server?.address();
|
|
68
|
+
if (addr && typeof addr === 'object')
|
|
69
|
+
return addr.port;
|
|
70
|
+
return undefined;
|
|
71
|
+
}
|
|
72
|
+
async close() {
|
|
73
|
+
for (const client of this.clients) {
|
|
74
|
+
client.close(1001, 'server shutting down');
|
|
75
|
+
}
|
|
76
|
+
this.clients.clear();
|
|
77
|
+
if (this.wss) {
|
|
78
|
+
await new Promise((resolve) => this.wss.close(() => resolve()));
|
|
79
|
+
this.wss = null;
|
|
80
|
+
}
|
|
81
|
+
if (this.ownsServer && this.server) {
|
|
82
|
+
await new Promise((resolve, reject) => this.server.close((err) => (err ? reject(err) : resolve())));
|
|
83
|
+
this.server = null;
|
|
84
|
+
}
|
|
85
|
+
}
|
|
86
|
+
setupWss(server) {
|
|
87
|
+
this.wss = new WebSocketServer({ noServer: true });
|
|
88
|
+
server.on('upgrade', (req, socket, head) => {
|
|
89
|
+
const url = new URL(req.url || '/', `http://${req.headers.host || 'localhost'}`);
|
|
90
|
+
if (url.pathname !== this.path) {
|
|
91
|
+
socket.write('HTTP/1.1 404 Not Found\r\n\r\n');
|
|
92
|
+
socket.destroy();
|
|
93
|
+
return;
|
|
94
|
+
}
|
|
95
|
+
if (this.clients.size >= this.maxConnections) {
|
|
96
|
+
this.wss.handleUpgrade(req, socket, head, (ws) => {
|
|
97
|
+
ws.close(1013, 'max connections reached');
|
|
98
|
+
});
|
|
99
|
+
return;
|
|
100
|
+
}
|
|
101
|
+
this.wss.handleUpgrade(req, socket, head, (ws) => {
|
|
102
|
+
const client = new VoiceClient(ws);
|
|
103
|
+
this.clients.add(client);
|
|
104
|
+
client.on('close', () => this.clients.delete(client));
|
|
105
|
+
this.emit('connection', client);
|
|
106
|
+
});
|
|
107
|
+
});
|
|
108
|
+
}
|
|
109
|
+
}
|
|
110
|
+
//# sourceMappingURL=ws-transport.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"ws-transport.js","sourceRoot":"","sources":["../../src/transport/ws-transport.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,IAAI,MAAM,WAAW,CAAC;AAC7B,OAAO,EAAE,eAAe,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAChD,OAAO,EAAE,MAAM,EAAE,MAAM,QAAQ,CAAC;AAUhC,MAAM,OAAO,WAAY,SAAQ,YAA+B;IAGjC;IAFpB,EAAE,CAAS;IAEpB,YAA6B,EAAa;QACxC,KAAK,EAAE,CAAC;QADmB,OAAE,GAAF,EAAE,CAAW;QAExC,IAAI,CAAC,EAAE,GAAG,MAAM,EAAE,CAAC;QAEnB,EAAE,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,IAAqB,EAAE,QAAiB,EAAE,EAAE;YAC5D,IAAI,QAAQ,EAAE,CAAC;gBACb,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,MAAM,CAAC,IAAI,CAAC,IAAc,CAAC,CAAC,CAAC;YAClD,CAAC;iBAAM,CAAC;gBACN,IAAI,CAAC;oBACH,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,QAAQ,EAAE,CAA4B,CAAC;oBACtE,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC;gBAC/B,CAAC;gBAAC,OAAO,GAAG,EAAE,CAAC;oBACb,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,iBAAkB,GAAa,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC;gBAC3E,CAAC;YACH,CAAC;QACH,CAAC,CAAC,CAAC;QAEH,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC;QACzC,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAG,EAAE,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC,CAAC;IACnD,CAAC;IAED,SAAS,CAAC,KAAa;QACrB,IAAI,IAAI,CAAC,EAAE,CAAC,UAAU,KAAK,SAAS,CAAC,IAAI,EAAE,CAAC;YAC1C,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACtB,CAAC;IACH,CAAC;IAED,WAAW,CAAC,GAA4B;QACtC,IAAI,IAAI,CAAC,EAAE,CAAC,UAAU,KAAK,SAAS,CAAC,IAAI,EAAE,CAAC;YAC1C,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,GAAG,CAAC,CAAC,CAAC;QACpC,CAAC;IACH,CAAC;IAED,KAAK,CAAC,IAAa,EAAE,MAAe;QAClC,IAAI,CAAC,EAAE,CAAC,KAAK,CAAC,IAAI,EAAE,MAAM,CAAC,CAAC;IAC9B,CAAC;CACF;AAMD,MAAM,OAAO,kBAAmB,SAAQ,YAA6B;IAClD,IAAI,CAAS;IACb,cAAc,CAAS;IAChC,GAAG,GAA2B,IAAI,CAAC;IACnC,MAAM,GAAuB,IAAI,CAAC;IAClC,UAAU,GAAG,KAAK,CAAC;IACnB,OAAO,GAAG,IAAI,GAAG,EAAe,CAAC;IAEzC,YAAY,MAAiC;QAC3C,KAAK,EAAE,CAAC;QACR,IAAI,CAAC,IAAI,GAAG,MAAM,EAAE,IAAI,IAAI,QAAQ,CAAC;QACrC,IAAI,CAAC,cAAc,GAAG,MAAM,EAAE,cAAc,IAAI,GAAG,CAAC;IACtD,CAAC;IAED,KAAK,CAAC,MAAM,CAAC,IAAY;QACvB,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC,YAAY,EAAE,CAAC;QAClC,IAAI,CAAC,UAAU,GAAG,IAAI,CAAC;QACvB,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QAC3B,MAAM,IAAI,OAAO,CAAO,CAAC,OAAO,EAAE,EAAE,CAAC,IAAI,CAAC,MAAO,CAAC,MAAM,CAAC,IAAI,EAAE,OAAO,CAAC,CAAC,CAAC;IAC3E,CAAC;IAED,cAAc,CAAC,MAAmB;QAChC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;QACrB,IAAI,CAAC,UAAU,GAAG,KAAK,CAAC;QACxB,IAAI,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAC;IACxB,CAAC;IAED,IAAI,IAAI;QACN,MAAM,IAAI,GAAG,IAAI,CAAC,MAAM,EAAE,OAAO,EAAE,CAAC;QACpC,IAAI,IAAI,IAAI,OAAO,IAAI,KAAK,QAAQ;YAAE,OAAO,IAAI,CAAC,IAAI,CAAC;QACvD,OAAO,SAAS,CAAC;IACnB,CAAC;IAED,KAAK,CAAC,KAAK;QACT,KAAK,MAAM,MAAM,IAAI,IAAI,CAAC,OAAO,EAAE,CAAC;YAClC,MAAM,CAAC,KAAK,CAAC,IAAI,EAAE,sBAAsB,CAAC,CAAC;QAC7C,CAAC;QACD,IAAI,CAAC,OAAO,CAAC,KAAK,EAAE,CAAC;QAErB,IAAI,IAAI,CAAC,GAAG,EAAE,CAAC;YACb,MAAM,IAAI,OAAO,CAAO,CAAC,OAAO,EAAE,EAAE,CAAC,IAAI,CAAC,GAAI,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC;YACvE,IAAI,CAAC,GAAG,GAAG,IAAI,CAAC;QAClB,CAAC;QAED,IAAI,IAAI,CAAC,UAAU,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC;YACnC,MAAM,IAAI,OAAO,CAAO,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE,CAC1C,IAAI,CAAC,MAAO,CAAC,KAAK,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,CAAC,CAAC,CAC7D,CAAC;YACF,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC;QACrB,CAAC;IACH,CAAC;IAEO,QAAQ,CAAC,MAAmB;QAClC,IAAI,CAAC,GAAG,GAAG,IAAI,eAAe,CAAC,EAAE,QAAQ,EAAE,IAAI,EAAE,CAAC,CAAC;QAEnD,MAAM,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,GAAG,EAAE,MAAM,EAAE,IAAI,EAAE,EAAE;YACzC,MAAM,GAAG,GAAG,IAAI,GAAG,CAAC,GAAG,CAAC,GAAG,IAAI,GAAG,EAAE,UAAU,GAAG,CAAC,OAAO,CAAC,IAAI,IAAI,WAAW,EAAE,CAAC,CAAC;YAEjF,IAAI,GAAG,CAAC,QAAQ,KAAK,IAAI,CAAC,IAAI,EAAE,CAAC;gBAC/B,MAAM,CAAC,KAAK,CAAC,gCAAgC,CAAC,CAAC;gBAC/C,MAAM,CAAC,OAAO,EAAE,CAAC;gBACjB,OAAO;YACT,CAAC;YAED,IAAI,IAAI,CAAC,OAAO,CAAC,IAAI,IAAI,IAAI,CAAC,cAAc,EAAE,CAAC;gBAC7C,IAAI,CAAC,GAAI,CAAC,aAAa,CAAC,GAAG,EAAE,MAAM,EAAE,IAAI,EAAE,CAAC,EAAE,EAAE,EAAE;oBAChD,EAAE,CAAC,KAAK,CAAC,IAAI,EAAE,yBAAyB,CAAC,CAAC;gBAC5C,CAAC,CAAC,CAAC;gBACH,OAAO;YACT,CAAC;YAED,IAAI,CAAC,GAAI,CAAC,aAAa,CAAC,GAAG,EAAE,MAAM,EAAE,IAAI,EAAE,CAAC,EAAE,EAAE,EAAE;gBAChD,MAAM,MAAM,GAAG,IAAI,WAAW,CAAC,EAAE,CAAC,CAAC;gBACnC,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC;gBACzB,MAAM,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE,CAAC,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC;gBACtD,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,MAAM,CAAC,CAAC;YAClC,CAAC,CAAC,CAAC;QACL,CAAC,CAAC,CAAC;IACL,CAAC;CACF"}
|
|
@@ -0,0 +1,16 @@
|
|
|
1
|
+
import type { TTSProvider, TTSOptions } from '../types.js';
|
|
2
|
+
export interface ElevenLabsTTSConfig {
|
|
3
|
+
apiKey: string;
|
|
4
|
+
voiceId?: string;
|
|
5
|
+
model?: string;
|
|
6
|
+
}
|
|
7
|
+
export declare class ElevenLabsTTS implements TTSProvider {
|
|
8
|
+
readonly name = "elevenlabs";
|
|
9
|
+
private readonly apiKey;
|
|
10
|
+
private readonly voiceId;
|
|
11
|
+
private readonly model;
|
|
12
|
+
constructor(config: ElevenLabsTTSConfig);
|
|
13
|
+
synthesize(text: string, options?: TTSOptions): Promise<Buffer>;
|
|
14
|
+
streamSynthesize(text: string, options?: TTSOptions): AsyncGenerator<Buffer>;
|
|
15
|
+
}
|
|
16
|
+
//# sourceMappingURL=elevenlabs-tts.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"elevenlabs-tts.d.ts","sourceRoot":"","sources":["../../src/tts/elevenlabs-tts.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,WAAW,EAAE,UAAU,EAAoB,MAAM,aAAa,CAAC;AAE7E,MAAM,WAAW,mBAAmB;IAClC,MAAM,EAAE,MAAM,CAAC;IACf,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,KAAK,CAAC,EAAE,MAAM,CAAC;CAChB;AAiBD,qBAAa,aAAc,YAAW,WAAW;IAC/C,QAAQ,CAAC,IAAI,gBAAgB;IAE7B,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAS;IAChC,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAS;IACjC,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAS;gBAEnB,MAAM,EAAE,mBAAmB;IAMjC,UAAU,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,UAAU,GAAG,OAAO,CAAC,MAAM,CAAC;IA6B9D,gBAAgB,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,UAAU,GAAG,cAAc,CAAC,MAAM,CAAC;CA8BpF"}
|
|
@@ -0,0 +1,77 @@
|
|
|
1
|
+
const BASE_URL = 'https://api.elevenlabs.io';
|
|
2
|
+
const DEFAULT_VOICE_ID = '21m00Tcm4TlvDq8ikWAM';
|
|
3
|
+
const DEFAULT_MODEL = 'eleven_flash_v2_5';
|
|
4
|
+
function mapFormat(format) {
|
|
5
|
+
switch (format) {
|
|
6
|
+
case 'pcm16':
|
|
7
|
+
return 'pcm_16000';
|
|
8
|
+
case 'mp3':
|
|
9
|
+
return 'mp3_44100_128';
|
|
10
|
+
default:
|
|
11
|
+
return 'mp3_44100_128';
|
|
12
|
+
}
|
|
13
|
+
}
|
|
14
|
+
export class ElevenLabsTTS {
|
|
15
|
+
name = 'elevenlabs';
|
|
16
|
+
apiKey;
|
|
17
|
+
voiceId;
|
|
18
|
+
model;
|
|
19
|
+
constructor(config) {
|
|
20
|
+
this.apiKey = config.apiKey;
|
|
21
|
+
this.voiceId = config.voiceId ?? DEFAULT_VOICE_ID;
|
|
22
|
+
this.model = config.model ?? DEFAULT_MODEL;
|
|
23
|
+
}
|
|
24
|
+
async synthesize(text, options) {
|
|
25
|
+
const voiceId = options?.voice ?? this.voiceId;
|
|
26
|
+
const outputFormat = mapFormat(options?.format);
|
|
27
|
+
const url = `${BASE_URL}/v1/text-to-speech/${voiceId}?output_format=${outputFormat}`;
|
|
28
|
+
const response = await fetch(url, {
|
|
29
|
+
method: 'POST',
|
|
30
|
+
headers: {
|
|
31
|
+
'xi-api-key': this.apiKey,
|
|
32
|
+
'Content-Type': 'application/json',
|
|
33
|
+
},
|
|
34
|
+
body: JSON.stringify({
|
|
35
|
+
text,
|
|
36
|
+
model_id: this.model,
|
|
37
|
+
voice_settings: {
|
|
38
|
+
stability: 0.5,
|
|
39
|
+
similarity_boost: 0.75,
|
|
40
|
+
},
|
|
41
|
+
}),
|
|
42
|
+
});
|
|
43
|
+
if (!response.ok) {
|
|
44
|
+
throw new Error(`ElevenLabs API error ${response.status}`);
|
|
45
|
+
}
|
|
46
|
+
const arrayBuffer = await response.arrayBuffer();
|
|
47
|
+
return Buffer.from(arrayBuffer);
|
|
48
|
+
}
|
|
49
|
+
async *streamSynthesize(text, options) {
|
|
50
|
+
const voiceId = options?.voice ?? this.voiceId;
|
|
51
|
+
const outputFormat = mapFormat(options?.format);
|
|
52
|
+
const url = `${BASE_URL}/v1/text-to-speech/${voiceId}/stream?output_format=${outputFormat}`;
|
|
53
|
+
const response = await fetch(url, {
|
|
54
|
+
method: 'POST',
|
|
55
|
+
headers: {
|
|
56
|
+
'xi-api-key': this.apiKey,
|
|
57
|
+
'Content-Type': 'application/json',
|
|
58
|
+
},
|
|
59
|
+
body: JSON.stringify({
|
|
60
|
+
text,
|
|
61
|
+
model_id: this.model,
|
|
62
|
+
voice_settings: {
|
|
63
|
+
stability: 0.5,
|
|
64
|
+
similarity_boost: 0.75,
|
|
65
|
+
},
|
|
66
|
+
}),
|
|
67
|
+
});
|
|
68
|
+
if (!response.ok) {
|
|
69
|
+
throw new Error(`ElevenLabs API error ${response.status}`);
|
|
70
|
+
}
|
|
71
|
+
const body = response.body;
|
|
72
|
+
for await (const chunk of body) {
|
|
73
|
+
yield Buffer.from(chunk);
|
|
74
|
+
}
|
|
75
|
+
}
|
|
76
|
+
}
|
|
77
|
+
//# sourceMappingURL=elevenlabs-tts.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"elevenlabs-tts.js","sourceRoot":"","sources":["../../src/tts/elevenlabs-tts.ts"],"names":[],"mappings":"AAQA,MAAM,QAAQ,GAAG,2BAA2B,CAAC;AAC7C,MAAM,gBAAgB,GAAG,sBAAsB,CAAC;AAChD,MAAM,aAAa,GAAG,mBAAmB,CAAC;AAE1C,SAAS,SAAS,CAAC,MAAyB;IAC1C,QAAQ,MAAM,EAAE,CAAC;QACf,KAAK,OAAO;YACV,OAAO,WAAW,CAAC;QACrB,KAAK,KAAK;YACR,OAAO,eAAe,CAAC;QACzB;YACE,OAAO,eAAe,CAAC;IAC3B,CAAC;AACH,CAAC;AAED,MAAM,OAAO,aAAa;IACf,IAAI,GAAG,YAAY,CAAC;IAEZ,MAAM,CAAS;IACf,OAAO,CAAS;IAChB,KAAK,CAAS;IAE/B,YAAY,MAA2B;QACrC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,OAAO,IAAI,gBAAgB,CAAC;QAClD,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,IAAI,aAAa,CAAC;IAC7C,CAAC;IAED,KAAK,CAAC,UAAU,CAAC,IAAY,EAAE,OAAoB;QACjD,MAAM,OAAO,GAAG,OAAO,EAAE,KAAK,IAAI,IAAI,CAAC,OAAO,CAAC;QAC/C,MAAM,YAAY,GAAG,SAAS,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC;QAChD,MAAM,GAAG,GAAG,GAAG,QAAQ,sBAAsB,OAAO,kBAAkB,YAAY,EAAE,CAAC;QAErF,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,EAAE;YAChC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,YAAY,EAAE,IAAI,CAAC,MAAM;gBACzB,cAAc,EAAE,kBAAkB;aACnC;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;gBACnB,IAAI;gBACJ,QAAQ,EAAE,IAAI,CAAC,KAAK;gBACpB,cAAc,EAAE;oBACd,SAAS,EAAE,GAAG;oBACd,gBAAgB,EAAE,IAAI;iBACvB;aACF,CAAC;SACH,CAAC,CAAC;QAEH,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,MAAM,IAAI,KAAK,CAAC,wBAAwB,QAAQ,CAAC,MAAM,EAAE,CAAC,CAAC;QAC7D,CAAC;QAED,MAAM,WAAW,GAAG,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC;QACjD,OAAO,MAAM,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;IAClC,CAAC;IAED,KAAK,CAAC,CAAC,gBAAgB,CAAC,IAAY,EAAE,OAAoB;QACxD,MAAM,OAAO,GAAG,OAAO,EAAE,KAAK,IAAI,IAAI,CAAC,OAAO,CAAC;QAC/C,MAAM,YAAY,GAAG,SAAS,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC;QAChD,MAAM,GAAG,GAAG,GAAG,QAAQ,sBAAsB,OAAO,yBAAyB,YAAY,EAAE,CAAC;QAE5F,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,EAAE;YAChC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,YAAY,EAAE,IAAI,CAAC,MAAM;gBACzB,cAAc,EAAE,kBAAkB;aACnC;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;gBACnB,IAAI;gBACJ,QAAQ,EAAE,IAAI,CAAC,KAAK;gBACpB,cAAc,EAAE;oBACd,SAAS,EAAE,GAAG;oBACd,gBAAgB,EAAE,IAAI;iBACvB;aACF,CAAC;SACH,CAAC,CAAC;QAEH,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,MAAM,IAAI,KAAK,CAAC,wBAAwB,QAAQ,CAAC,MAAM,EAAE,CAAC,CAAC;QAC7D,CAAC;QAED,MAAM,IAAI,GAAG,QAAQ,CAAC,IAA4C,CAAC;QACnE,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,IAAI,EAAE,CAAC;YAC/B,MAAM,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QAC3B,CAAC;IACH,CAAC;CACF"}
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/tts/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,SAAS,EAAE,KAAK,eAAe,EAAE,MAAM,iBAAiB,CAAC;AAClE,OAAO,EAAE,aAAa,EAAE,KAAK,mBAAmB,EAAE,MAAM,qBAAqB,CAAC"}
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/tts/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,SAAS,EAAwB,MAAM,iBAAiB,CAAC;AAClE,OAAO,EAAE,aAAa,EAA4B,MAAM,qBAAqB,CAAC"}
|
|
@@ -0,0 +1,18 @@
|
|
|
1
|
+
import type { TTSProvider, TTSOptions } from '../types.js';
|
|
2
|
+
export interface OpenAITTSConfig {
|
|
3
|
+
apiKey: string;
|
|
4
|
+
model?: string;
|
|
5
|
+
voice?: string;
|
|
6
|
+
baseURL?: string;
|
|
7
|
+
}
|
|
8
|
+
export declare class OpenAITTS implements TTSProvider {
|
|
9
|
+
readonly name = "openai";
|
|
10
|
+
private readonly client;
|
|
11
|
+
private readonly model;
|
|
12
|
+
private readonly voice;
|
|
13
|
+
constructor(config: OpenAITTSConfig);
|
|
14
|
+
synthesize(text: string, options?: TTSOptions): Promise<Buffer>;
|
|
15
|
+
streamSynthesize(text: string, options?: TTSOptions): AsyncGenerator<Buffer>;
|
|
16
|
+
private buildParams;
|
|
17
|
+
}
|
|
18
|
+
//# sourceMappingURL=openai-tts.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"openai-tts.d.ts","sourceRoot":"","sources":["../../src/tts/openai-tts.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,WAAW,EAAE,UAAU,EAAoB,MAAM,aAAa,CAAC;AAE7E,MAAM,WAAW,eAAe;IAC9B,MAAM,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,OAAO,CAAC,EAAE,MAAM,CAAC;CAClB;AAUD,qBAAa,SAAU,YAAW,WAAW;IAC3C,QAAQ,CAAC,IAAI,YAAY;IAEzB,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAS;IAChC,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAS;IAC/B,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAS;gBAEnB,MAAM,EAAE,eAAe;IAS7B,UAAU,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,UAAU,GAAG,OAAO,CAAC,MAAM,CAAC;IAM9D,gBAAgB,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,UAAU,GAAG,cAAc,CAAC,MAAM,CAAC;IAQnF,OAAO,CAAC,WAAW;CAYpB"}
|
|
@@ -0,0 +1,47 @@
|
|
|
1
|
+
import OpenAI from 'openai';
|
|
2
|
+
const DEFAULT_MODEL = 'gpt-4o-mini-tts';
|
|
3
|
+
const DEFAULT_VOICE = 'alloy';
|
|
4
|
+
function mapFormat(format) {
|
|
5
|
+
if (format === 'pcm16')
|
|
6
|
+
return 'pcm';
|
|
7
|
+
return format;
|
|
8
|
+
}
|
|
9
|
+
export class OpenAITTS {
|
|
10
|
+
name = 'openai';
|
|
11
|
+
client;
|
|
12
|
+
model;
|
|
13
|
+
voice;
|
|
14
|
+
constructor(config) {
|
|
15
|
+
this.client = new OpenAI({
|
|
16
|
+
apiKey: config.apiKey,
|
|
17
|
+
...(config.baseURL && { baseURL: config.baseURL }),
|
|
18
|
+
});
|
|
19
|
+
this.model = config.model ?? DEFAULT_MODEL;
|
|
20
|
+
this.voice = config.voice ?? DEFAULT_VOICE;
|
|
21
|
+
}
|
|
22
|
+
async synthesize(text, options) {
|
|
23
|
+
const response = await this.client.audio.speech.create(this.buildParams(text, options));
|
|
24
|
+
const arrayBuffer = await response.arrayBuffer();
|
|
25
|
+
return Buffer.from(arrayBuffer);
|
|
26
|
+
}
|
|
27
|
+
async *streamSynthesize(text, options) {
|
|
28
|
+
const response = await this.client.audio.speech.create(this.buildParams(text, options));
|
|
29
|
+
const body = response.body;
|
|
30
|
+
for await (const chunk of body) {
|
|
31
|
+
yield Buffer.from(chunk);
|
|
32
|
+
}
|
|
33
|
+
}
|
|
34
|
+
buildParams(text, options) {
|
|
35
|
+
return {
|
|
36
|
+
model: this.model,
|
|
37
|
+
voice: options?.voice ?? this.voice,
|
|
38
|
+
input: text,
|
|
39
|
+
...(options?.speed !== undefined && { speed: options.speed }),
|
|
40
|
+
...(options?.format && {
|
|
41
|
+
response_format: mapFormat(options.format),
|
|
42
|
+
}),
|
|
43
|
+
...(options?.instructions && { instructions: options.instructions }),
|
|
44
|
+
};
|
|
45
|
+
}
|
|
46
|
+
}
|
|
47
|
+
//# sourceMappingURL=openai-tts.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"openai-tts.js","sourceRoot":"","sources":["../../src/tts/openai-tts.ts"],"names":[],"mappings":"AAAA,OAAO,MAAM,MAAM,QAAQ,CAAC;AAW5B,MAAM,aAAa,GAAG,iBAAiB,CAAC;AACxC,MAAM,aAAa,GAAG,OAAO,CAAC;AAE9B,SAAS,SAAS,CAAC,MAAwB;IACzC,IAAI,MAAM,KAAK,OAAO;QAAE,OAAO,KAAK,CAAC;IACrC,OAAO,MAAM,CAAC;AAChB,CAAC;AAED,MAAM,OAAO,SAAS;IACX,IAAI,GAAG,QAAQ,CAAC;IAER,MAAM,CAAS;IACf,KAAK,CAAS;IACd,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;QAC3C,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,IAAI,aAAa,CAAC;IAC7C,CAAC;IAED,KAAK,CAAC,UAAU,CAAC,IAAY,EAAE,OAAoB;QACjD,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,WAAW,CAAC,IAAI,EAAE,OAAO,CAAC,CAAC,CAAC;QACxF,MAAM,WAAW,GAAG,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC;QACjD,OAAO,MAAM,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;IAClC,CAAC;IAED,KAAK,CAAC,CAAC,gBAAgB,CAAC,IAAY,EAAE,OAAoB;QACxD,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,WAAW,CAAC,IAAI,EAAE,OAAO,CAAC,CAAC,CAAC;QACxF,MAAM,IAAI,GAAI,QAA2D,CAAC,IAAI,CAAC;QAC/E,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,IAAI,EAAE,CAAC;YAC/B,MAAM,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QAC3B,CAAC;IACH,CAAC;IAEO,WAAW,CAAC,IAAY,EAAE,OAAoB;QACpD,OAAO;YACL,KAAK,EAAE,IAAI,CAAC,KAAK;YACjB,KAAK,EAAE,OAAO,EAAE,KAAK,IAAI,IAAI,CAAC,KAAK;YACnC,KAAK,EAAE,IAAI;YACX,GAAG,CAAC,OAAO,EAAE,KAAK,KAAK,SAAS,IAAI,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,EAAE,CAAC;YAC7D,GAAG,CAAC,OAAO,EAAE,MAAM,IAAI;gBACrB,eAAe,EAAE,SAAS,CAAC,OAAO,CAAC,MAAM,CAA0C;aACpF,CAAC;YACF,GAAG,CAAC,OAAO,EAAE,YAAY,IAAI,EAAE,YAAY,EAAE,OAAO,CAAC,YAAY,EAAE,CAAC;SACrE,CAAC;IACJ,CAAC;CACF"}
|
package/dist/types.d.ts
ADDED
|
@@ -0,0 +1,96 @@
|
|
|
1
|
+
import type { TranscribeResult, VoiceAudioFormat } from '@cogitator-ai/types';
|
|
2
|
+
export type { TranscribeResult, VoiceAudioFormat };
|
|
3
|
+
export interface STTOptions {
|
|
4
|
+
language?: string;
|
|
5
|
+
prompt?: string;
|
|
6
|
+
}
|
|
7
|
+
export interface STTStreamOptions extends STTOptions {
|
|
8
|
+
interimResults?: boolean;
|
|
9
|
+
endpointing?: number;
|
|
10
|
+
}
|
|
11
|
+
export interface STTProvider {
|
|
12
|
+
readonly name: string;
|
|
13
|
+
transcribe(audio: Buffer, options?: STTOptions): Promise<TranscribeResult>;
|
|
14
|
+
createStream(options?: STTStreamOptions): STTStream;
|
|
15
|
+
}
|
|
16
|
+
export interface STTStream {
|
|
17
|
+
write(chunk: Buffer): void;
|
|
18
|
+
close(): Promise<TranscribeResult>;
|
|
19
|
+
on(event: 'partial', cb: (text: string) => void): this;
|
|
20
|
+
on(event: 'final', cb: (result: TranscribeResult) => void): this;
|
|
21
|
+
on(event: 'error', cb: (error: Error) => void): this;
|
|
22
|
+
off(event: string, cb: (...args: unknown[]) => void): this;
|
|
23
|
+
removeAllListeners(): this;
|
|
24
|
+
}
|
|
25
|
+
export interface TTSOptions {
|
|
26
|
+
voice?: string;
|
|
27
|
+
speed?: number;
|
|
28
|
+
format?: VoiceAudioFormat;
|
|
29
|
+
instructions?: string;
|
|
30
|
+
}
|
|
31
|
+
export interface TTSProvider {
|
|
32
|
+
readonly name: string;
|
|
33
|
+
synthesize(text: string, options?: TTSOptions): Promise<Buffer>;
|
|
34
|
+
streamSynthesize(text: string, options?: TTSOptions): AsyncGenerator<Buffer>;
|
|
35
|
+
}
|
|
36
|
+
export type VADEvent = {
|
|
37
|
+
type: 'speech_start';
|
|
38
|
+
} | {
|
|
39
|
+
type: 'speech_end';
|
|
40
|
+
duration: number;
|
|
41
|
+
} | {
|
|
42
|
+
type: 'speech';
|
|
43
|
+
probability: number;
|
|
44
|
+
} | {
|
|
45
|
+
type: 'silence';
|
|
46
|
+
};
|
|
47
|
+
export interface VADProvider {
|
|
48
|
+
readonly name: string;
|
|
49
|
+
process(samples: Float32Array): VADEvent;
|
|
50
|
+
reset(): void;
|
|
51
|
+
}
|
|
52
|
+
export interface VoicePipelineConfig {
|
|
53
|
+
stt: STTProvider;
|
|
54
|
+
tts: TTSProvider;
|
|
55
|
+
vad?: VADProvider;
|
|
56
|
+
agent: {
|
|
57
|
+
run: (input: string) => Promise<{
|
|
58
|
+
content: string;
|
|
59
|
+
}>;
|
|
60
|
+
};
|
|
61
|
+
sampleRate?: number;
|
|
62
|
+
}
|
|
63
|
+
export interface RealtimeSessionConfig {
|
|
64
|
+
provider: 'openai' | 'gemini';
|
|
65
|
+
model?: string;
|
|
66
|
+
apiKey: string;
|
|
67
|
+
instructions?: string;
|
|
68
|
+
tools?: Array<{
|
|
69
|
+
name: string;
|
|
70
|
+
description: string;
|
|
71
|
+
parameters: Record<string, unknown>;
|
|
72
|
+
execute: (args: unknown) => Promise<unknown>;
|
|
73
|
+
}>;
|
|
74
|
+
voice?: string;
|
|
75
|
+
}
|
|
76
|
+
export interface WebSocketTransportConfig {
|
|
77
|
+
path?: string;
|
|
78
|
+
maxConnections?: number;
|
|
79
|
+
}
|
|
80
|
+
export interface VoiceAgentConfig {
|
|
81
|
+
agent: {
|
|
82
|
+
run: (input: string) => Promise<{
|
|
83
|
+
content: string;
|
|
84
|
+
}>;
|
|
85
|
+
};
|
|
86
|
+
mode: 'pipeline' | 'realtime';
|
|
87
|
+
stt?: STTProvider;
|
|
88
|
+
tts?: TTSProvider;
|
|
89
|
+
vad?: VADProvider;
|
|
90
|
+
realtimeProvider?: 'openai' | 'gemini';
|
|
91
|
+
realtimeApiKey?: string;
|
|
92
|
+
realtimeModel?: string;
|
|
93
|
+
voice?: string;
|
|
94
|
+
transport?: WebSocketTransportConfig;
|
|
95
|
+
}
|
|
96
|
+
//# sourceMappingURL=types.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,gBAAgB,EAAE,gBAAgB,EAAE,MAAM,qBAAqB,CAAC;AAE9E,YAAY,EAAE,gBAAgB,EAAE,gBAAgB,EAAE,CAAC;AAEnD,MAAM,WAAW,UAAU;IACzB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,MAAM,CAAC,EAAE,MAAM,CAAC;CACjB;AAED,MAAM,WAAW,gBAAiB,SAAQ,UAAU;IAClD,cAAc,CAAC,EAAE,OAAO,CAAC;IACzB,WAAW,CAAC,EAAE,MAAM,CAAC;CACtB;AAED,MAAM,WAAW,WAAW;IAC1B,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,UAAU,CAAC,KAAK,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,UAAU,GAAG,OAAO,CAAC,gBAAgB,CAAC,CAAC;IAC3E,YAAY,CAAC,OAAO,CAAC,EAAE,gBAAgB,GAAG,SAAS,CAAC;CACrD;AAED,MAAM,WAAW,SAAS;IACxB,KAAK,CAAC,KAAK,EAAE,MAAM,GAAG,IAAI,CAAC;IAC3B,KAAK,IAAI,OAAO,CAAC,gBAAgB,CAAC,CAAC;IACnC,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,EAAE,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,IAAI,GAAG,IAAI,CAAC;IACvD,EAAE,CAAC,KAAK,EAAE,OAAO,EAAE,EAAE,EAAE,CAAC,MAAM,EAAE,gBAAgB,KAAK,IAAI,GAAG,IAAI,CAAC;IACjE,EAAE,CAAC,KAAK,EAAE,OAAO,EAAE,EAAE,EAAE,CAAC,KAAK,EAAE,KAAK,KAAK,IAAI,GAAG,IAAI,CAAC;IACrD,GAAG,CAAC,KAAK,EAAE,MAAM,EAAE,EAAE,EAAE,CAAC,GAAG,IAAI,EAAE,OAAO,EAAE,KAAK,IAAI,GAAG,IAAI,CAAC;IAC3D,kBAAkB,IAAI,IAAI,CAAC;CAC5B;AAED,MAAM,WAAW,UAAU;IACzB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,MAAM,CAAC,EAAE,gBAAgB,CAAC;IAC1B,YAAY,CAAC,EAAE,MAAM,CAAC;CACvB;AAED,MAAM,WAAW,WAAW;IAC1B,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,UAAU,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,UAAU,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;IAChE,gBAAgB,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,UAAU,GAAG,cAAc,CAAC,MAAM,CAAC,CAAC;CAC9E;AAED,MAAM,MAAM,QAAQ,GAChB;IAAE,IAAI,EAAE,cAAc,CAAA;CAAE,GACxB;IAAE,IAAI,EAAE,YAAY,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAA;CAAE,GACxC;IAAE,IAAI,EAAE,QAAQ,CAAC;IAAC,WAAW,EAAE,MAAM,CAAA;CAAE,GACvC;IAAE,IAAI,EAAE,SAAS,CAAA;CAAE,CAAC;AAExB,MAAM,WAAW,WAAW;IAC1B,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,OAAO,CAAC,OAAO,EAAE,YAAY,GAAG,QAAQ,CAAC;IACzC,KAAK,IAAI,IAAI,CAAC;CACf;AAED,MAAM,WAAW,mBAAmB;IAClC,GAAG,EAAE,WAAW,CAAC;IACjB,GAAG,EAAE,WAAW,CAAC;IACjB,GAAG,CAAC,EAAE,WAAW,CAAC;IAClB,KAAK,EAAE;QAAE,GAAG,EAAE,CAAC,KAAK,EAAE,MAAM,KAAK,OAAO,CAAC;YAAE,OAAO,EAAE,MAAM,CAAA;SAAE,CAAC,CAAA;KAAE,CAAC;IAChE,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAED,MAAM,WAAW,qBAAqB;IACpC,QAAQ,EAAE,QAAQ,GAAG,QAAQ,CAAC;IAC9B,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,MAAM,EAAE,MAAM,CAAC;IACf,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,KAAK,CAAC,EAAE,KAAK,CAAC;QACZ,IAAI,EAAE,MAAM,CAAC;QACb,WAAW,EAAE,MAAM,CAAC;QACpB,UAAU,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QACpC,OAAO,EAAE,CAAC,IAAI,EAAE,OAAO,KAAK,OAAO,CAAC,OAAO,CAAC,CAAC;KAC9C,CAAC,CAAC;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;CAChB;AAED,MAAM,WAAW,wBAAwB;IACvC,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,cAAc,CAAC,EAAE,MAAM,CAAC;CACzB;AAED,MAAM,WAAW,gBAAgB;IAC/B,KAAK,EAAE;QAAE,GAAG,EAAE,CAAC,KAAK,EAAE,MAAM,KAAK,OAAO,CAAC;YAAE,OAAO,EAAE,MAAM,CAAA;SAAE,CAAC,CAAA;KAAE,CAAC;IAChE,IAAI,EAAE,UAAU,GAAG,UAAU,CAAC;IAC9B,GAAG,CAAC,EAAE,WAAW,CAAC;IAClB,GAAG,CAAC,EAAE,WAAW,CAAC;IAClB,GAAG,CAAC,EAAE,WAAW,CAAC;IAClB,gBAAgB,CAAC,EAAE,QAAQ,GAAG,QAAQ,CAAC;IACvC,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,SAAS,CAAC,EAAE,wBAAwB,CAAC;CACtC"}
|
package/dist/types.js
ADDED
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"types.js","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":""}
|
|
@@ -0,0 +1,19 @@
|
|
|
1
|
+
import type { VADEvent, VADProvider } from '../types.js';
|
|
2
|
+
export interface EnergyVADConfig {
|
|
3
|
+
threshold?: number;
|
|
4
|
+
silenceDuration?: number;
|
|
5
|
+
sampleRate?: number;
|
|
6
|
+
}
|
|
7
|
+
export declare class EnergyVAD implements VADProvider {
|
|
8
|
+
readonly name = "energy";
|
|
9
|
+
private readonly threshold;
|
|
10
|
+
private readonly silenceDuration;
|
|
11
|
+
private readonly sampleRate;
|
|
12
|
+
private state;
|
|
13
|
+
private speechSamples;
|
|
14
|
+
private silenceSamples;
|
|
15
|
+
constructor(config?: EnergyVADConfig);
|
|
16
|
+
process(samples: Float32Array): VADEvent;
|
|
17
|
+
reset(): void;
|
|
18
|
+
}
|
|
19
|
+
//# sourceMappingURL=energy-vad.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"energy-vad.d.ts","sourceRoot":"","sources":["../../src/vad/energy-vad.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,QAAQ,EAAE,WAAW,EAAE,MAAM,aAAa,CAAC;AAGzD,MAAM,WAAW,eAAe;IAC9B,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAED,qBAAa,SAAU,YAAW,WAAW;IAC3C,QAAQ,CAAC,IAAI,YAAY;IAEzB,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAS;IACnC,OAAO,CAAC,QAAQ,CAAC,eAAe,CAAS;IACzC,OAAO,CAAC,QAAQ,CAAC,UAAU,CAAS;IAEpC,OAAO,CAAC,KAAK,CAA+B;IAC5C,OAAO,CAAC,aAAa,CAAK;IAC1B,OAAO,CAAC,cAAc,CAAK;gBAEf,MAAM,GAAE,eAAoB;IAMxC,OAAO,CAAC,OAAO,EAAE,YAAY,GAAG,QAAQ;IAmCxC,KAAK,IAAI,IAAI;CAKd"}
|
|
@@ -0,0 +1,49 @@
|
|
|
1
|
+
import { calculateRMS } from '../audio.js';
|
|
2
|
+
export class EnergyVAD {
|
|
3
|
+
name = 'energy';
|
|
4
|
+
threshold;
|
|
5
|
+
silenceDuration;
|
|
6
|
+
sampleRate;
|
|
7
|
+
state = 'idle';
|
|
8
|
+
speechSamples = 0;
|
|
9
|
+
silenceSamples = 0;
|
|
10
|
+
constructor(config = {}) {
|
|
11
|
+
this.threshold = config.threshold ?? 0.01;
|
|
12
|
+
this.silenceDuration = config.silenceDuration ?? 500;
|
|
13
|
+
this.sampleRate = config.sampleRate ?? 16000;
|
|
14
|
+
}
|
|
15
|
+
process(samples) {
|
|
16
|
+
const rms = calculateRMS(samples);
|
|
17
|
+
const isSpeech = rms > this.threshold;
|
|
18
|
+
const chunkDurationMs = (samples.length / this.sampleRate) * 1000;
|
|
19
|
+
if (this.state === 'idle') {
|
|
20
|
+
if (isSpeech) {
|
|
21
|
+
this.state = 'speaking';
|
|
22
|
+
this.speechSamples = samples.length;
|
|
23
|
+
this.silenceSamples = 0;
|
|
24
|
+
return { type: 'speech_start' };
|
|
25
|
+
}
|
|
26
|
+
return { type: 'silence' };
|
|
27
|
+
}
|
|
28
|
+
this.speechSamples += samples.length;
|
|
29
|
+
if (isSpeech) {
|
|
30
|
+
this.silenceSamples = 0;
|
|
31
|
+
return { type: 'speech', probability: Math.min(1, rms / 0.5) };
|
|
32
|
+
}
|
|
33
|
+
this.silenceSamples += chunkDurationMs;
|
|
34
|
+
if (this.silenceSamples >= this.silenceDuration) {
|
|
35
|
+
const duration = (this.speechSamples / this.sampleRate) * 1000;
|
|
36
|
+
this.state = 'idle';
|
|
37
|
+
this.speechSamples = 0;
|
|
38
|
+
this.silenceSamples = 0;
|
|
39
|
+
return { type: 'speech_end', duration };
|
|
40
|
+
}
|
|
41
|
+
return { type: 'speech', probability: 0 };
|
|
42
|
+
}
|
|
43
|
+
reset() {
|
|
44
|
+
this.state = 'idle';
|
|
45
|
+
this.silenceSamples = 0;
|
|
46
|
+
this.speechSamples = 0;
|
|
47
|
+
}
|
|
48
|
+
}
|
|
49
|
+
//# sourceMappingURL=energy-vad.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"energy-vad.js","sourceRoot":"","sources":["../../src/vad/energy-vad.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAQ3C,MAAM,OAAO,SAAS;IACX,IAAI,GAAG,QAAQ,CAAC;IAER,SAAS,CAAS;IAClB,eAAe,CAAS;IACxB,UAAU,CAAS;IAE5B,KAAK,GAAwB,MAAM,CAAC;IACpC,aAAa,GAAG,CAAC,CAAC;IAClB,cAAc,GAAG,CAAC,CAAC;IAE3B,YAAY,SAA0B,EAAE;QACtC,IAAI,CAAC,SAAS,GAAG,MAAM,CAAC,SAAS,IAAI,IAAI,CAAC;QAC1C,IAAI,CAAC,eAAe,GAAG,MAAM,CAAC,eAAe,IAAI,GAAG,CAAC;QACrD,IAAI,CAAC,UAAU,GAAG,MAAM,CAAC,UAAU,IAAI,KAAK,CAAC;IAC/C,CAAC;IAED,OAAO,CAAC,OAAqB;QAC3B,MAAM,GAAG,GAAG,YAAY,CAAC,OAAO,CAAC,CAAC;QAClC,MAAM,QAAQ,GAAG,GAAG,GAAG,IAAI,CAAC,SAAS,CAAC;QACtC,MAAM,eAAe,GAAG,CAAC,OAAO,CAAC,MAAM,GAAG,IAAI,CAAC,UAAU,CAAC,GAAG,IAAI,CAAC;QAElE,IAAI,IAAI,CAAC,KAAK,KAAK,MAAM,EAAE,CAAC;YAC1B,IAAI,QAAQ,EAAE,CAAC;gBACb,IAAI,CAAC,KAAK,GAAG,UAAU,CAAC;gBACxB,IAAI,CAAC,aAAa,GAAG,OAAO,CAAC,MAAM,CAAC;gBACpC,IAAI,CAAC,cAAc,GAAG,CAAC,CAAC;gBACxB,OAAO,EAAE,IAAI,EAAE,cAAc,EAAE,CAAC;YAClC,CAAC;YACD,OAAO,EAAE,IAAI,EAAE,SAAS,EAAE,CAAC;QAC7B,CAAC;QAED,IAAI,CAAC,aAAa,IAAI,OAAO,CAAC,MAAM,CAAC;QAErC,IAAI,QAAQ,EAAE,CAAC;YACb,IAAI,CAAC,cAAc,GAAG,CAAC,CAAC;YACxB,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,GAAG,GAAG,GAAG,CAAC,EAAE,CAAC;QACjE,CAAC;QAED,IAAI,CAAC,cAAc,IAAI,eAAe,CAAC;QAEvC,IAAI,IAAI,CAAC,cAAc,IAAI,IAAI,CAAC,eAAe,EAAE,CAAC;YAChD,MAAM,QAAQ,GAAG,CAAC,IAAI,CAAC,aAAa,GAAG,IAAI,CAAC,UAAU,CAAC,GAAG,IAAI,CAAC;YAC/D,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC;YACpB,IAAI,CAAC,aAAa,GAAG,CAAC,CAAC;YACvB,IAAI,CAAC,cAAc,GAAG,CAAC,CAAC;YACxB,OAAO,EAAE,IAAI,EAAE,YAAY,EAAE,QAAQ,EAAE,CAAC;QAC1C,CAAC;QAED,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,CAAC,EAAE,CAAC;IAC5C,CAAC;IAED,KAAK;QACH,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC;QACpB,IAAI,CAAC,cAAc,GAAG,CAAC,CAAC;QACxB,IAAI,CAAC,aAAa,GAAG,CAAC,CAAC;IACzB,CAAC;CACF"}
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/vad/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,SAAS,EAAE,KAAK,eAAe,EAAE,MAAM,iBAAiB,CAAC;AAClE,OAAO,EAAE,SAAS,EAAE,KAAK,eAAe,EAAE,MAAM,iBAAiB,CAAC"}
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/vad/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,SAAS,EAAwB,MAAM,iBAAiB,CAAC;AAClE,OAAO,EAAE,SAAS,EAAwB,MAAM,iBAAiB,CAAC"}
|
|
@@ -0,0 +1,27 @@
|
|
|
1
|
+
import type { VADEvent, VADProvider } from '../types.js';
|
|
2
|
+
export interface SileroVADConfig {
|
|
3
|
+
modelPath: string;
|
|
4
|
+
threshold?: number;
|
|
5
|
+
silenceDuration?: number;
|
|
6
|
+
sampleRate?: number;
|
|
7
|
+
}
|
|
8
|
+
export declare class SileroVAD implements VADProvider {
|
|
9
|
+
readonly name = "silero";
|
|
10
|
+
private readonly modelPath;
|
|
11
|
+
private readonly threshold;
|
|
12
|
+
private readonly silenceDuration;
|
|
13
|
+
private readonly sampleRate;
|
|
14
|
+
private session;
|
|
15
|
+
private ort;
|
|
16
|
+
private hn;
|
|
17
|
+
private cn;
|
|
18
|
+
private state;
|
|
19
|
+
private speechSamples;
|
|
20
|
+
private silenceSamples;
|
|
21
|
+
constructor(config: SileroVADConfig);
|
|
22
|
+
init(): Promise<void>;
|
|
23
|
+
process(samples: Float32Array): VADEvent;
|
|
24
|
+
reset(): void;
|
|
25
|
+
private evaluate;
|
|
26
|
+
}
|
|
27
|
+
//# sourceMappingURL=silero-vad.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"silero-vad.d.ts","sourceRoot":"","sources":["../../src/vad/silero-vad.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,QAAQ,EAAE,WAAW,EAAE,MAAM,aAAa,CAAC;AAEzD,MAAM,WAAW,eAAe;IAC9B,SAAS,EAAE,MAAM,CAAC;IAClB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAmBD,qBAAa,SAAU,YAAW,WAAW;IAC3C,QAAQ,CAAC,IAAI,YAAY;IAEzB,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAS;IACnC,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAS;IACnC,OAAO,CAAC,QAAQ,CAAC,eAAe,CAAS;IACzC,OAAO,CAAC,QAAQ,CAAC,UAAU,CAAS;IAEpC,OAAO,CAAC,OAAO,CAA4B;IAC3C,OAAO,CAAC,GAAG,CAA4B;IAEvC,OAAO,CAAC,EAAE,CAAe;IACzB,OAAO,CAAC,EAAE,CAAe;IAEzB,OAAO,CAAC,KAAK,CAA+B;IAC5C,OAAO,CAAC,aAAa,CAAK;IAC1B,OAAO,CAAC,cAAc,CAAK;gBAEf,MAAM,EAAE,eAAe;IAU7B,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC;IAY3B,OAAO,CAAC,OAAO,EAAE,YAAY,GAAG,QAAQ;IAqBxC,KAAK,IAAI,IAAI;IAQb,OAAO,CAAC,QAAQ;CAiCjB"}
|
|
@@ -0,0 +1,83 @@
|
|
|
1
|
+
export class SileroVAD {
|
|
2
|
+
name = 'silero';
|
|
3
|
+
modelPath;
|
|
4
|
+
threshold;
|
|
5
|
+
silenceDuration;
|
|
6
|
+
sampleRate;
|
|
7
|
+
session = null;
|
|
8
|
+
ort = null;
|
|
9
|
+
hn;
|
|
10
|
+
cn;
|
|
11
|
+
state = 'idle';
|
|
12
|
+
speechSamples = 0;
|
|
13
|
+
silenceSamples = 0;
|
|
14
|
+
constructor(config) {
|
|
15
|
+
this.modelPath = config.modelPath;
|
|
16
|
+
this.threshold = config.threshold ?? 0.5;
|
|
17
|
+
this.silenceDuration = config.silenceDuration ?? 500;
|
|
18
|
+
this.sampleRate = config.sampleRate ?? 16000;
|
|
19
|
+
this.hn = new Float32Array(2 * 1 * 64);
|
|
20
|
+
this.cn = new Float32Array(2 * 1 * 64);
|
|
21
|
+
}
|
|
22
|
+
async init() {
|
|
23
|
+
const moduleName = 'onnxruntime-node';
|
|
24
|
+
try {
|
|
25
|
+
this.ort = (await import(/* webpackIgnore: true */ moduleName));
|
|
26
|
+
}
|
|
27
|
+
catch {
|
|
28
|
+
throw new Error('onnxruntime-node is required for SileroVAD. Install it with: npm install onnxruntime-node');
|
|
29
|
+
}
|
|
30
|
+
this.session = await this.ort.InferenceSession.create(this.modelPath);
|
|
31
|
+
}
|
|
32
|
+
process(samples) {
|
|
33
|
+
if (!this.session || !this.ort) {
|
|
34
|
+
throw new Error('SileroVAD: must call init() before process()');
|
|
35
|
+
}
|
|
36
|
+
const { Tensor } = this.ort;
|
|
37
|
+
const input = new Tensor('float32', samples, [1, samples.length]);
|
|
38
|
+
const sr = new Tensor('int64', BigInt64Array.from([BigInt(this.sampleRate)]), [1]);
|
|
39
|
+
const h = new Tensor('float32', this.hn, [2, 1, 64]);
|
|
40
|
+
const c = new Tensor('float32', this.cn, [2, 1, 64]);
|
|
41
|
+
const feeds = { input, sr, h, c };
|
|
42
|
+
const result = this.session.run(feeds);
|
|
43
|
+
const probability = result.output.data[0];
|
|
44
|
+
this.hn = new Float32Array(result.hn.data);
|
|
45
|
+
this.cn = new Float32Array(result.cn.data);
|
|
46
|
+
return this.evaluate(probability, samples.length);
|
|
47
|
+
}
|
|
48
|
+
reset() {
|
|
49
|
+
this.state = 'idle';
|
|
50
|
+
this.speechSamples = 0;
|
|
51
|
+
this.silenceSamples = 0;
|
|
52
|
+
this.hn = new Float32Array(2 * 1 * 64);
|
|
53
|
+
this.cn = new Float32Array(2 * 1 * 64);
|
|
54
|
+
}
|
|
55
|
+
evaluate(probability, chunkLength) {
|
|
56
|
+
const isSpeech = probability >= this.threshold;
|
|
57
|
+
const chunkDurationMs = (chunkLength / this.sampleRate) * 1000;
|
|
58
|
+
if (this.state === 'idle') {
|
|
59
|
+
if (isSpeech) {
|
|
60
|
+
this.state = 'speaking';
|
|
61
|
+
this.speechSamples = chunkLength;
|
|
62
|
+
this.silenceSamples = 0;
|
|
63
|
+
return { type: 'speech_start' };
|
|
64
|
+
}
|
|
65
|
+
return { type: 'silence' };
|
|
66
|
+
}
|
|
67
|
+
this.speechSamples += chunkLength;
|
|
68
|
+
if (isSpeech) {
|
|
69
|
+
this.silenceSamples = 0;
|
|
70
|
+
return { type: 'speech', probability };
|
|
71
|
+
}
|
|
72
|
+
this.silenceSamples += chunkDurationMs;
|
|
73
|
+
if (this.silenceSamples >= this.silenceDuration) {
|
|
74
|
+
const duration = (this.speechSamples / this.sampleRate) * 1000;
|
|
75
|
+
this.state = 'idle';
|
|
76
|
+
this.speechSamples = 0;
|
|
77
|
+
this.silenceSamples = 0;
|
|
78
|
+
return { type: 'speech_end', duration };
|
|
79
|
+
}
|
|
80
|
+
return { type: 'speech', probability: 0 };
|
|
81
|
+
}
|
|
82
|
+
}
|
|
83
|
+
//# sourceMappingURL=silero-vad.js.map
|