@cogitator-ai/voice 0.1.12 → 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 +46 -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 +22 -3
- package/dist/pipeline/pipeline-session.d.ts.map +1 -1
- package/dist/pipeline/pipeline-session.js +189 -61
- 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 +21 -1
- package/dist/realtime/gemini-realtime.d.ts.map +1 -1
- package/dist/realtime/gemini-realtime.js +173 -65
- package/dist/realtime/gemini-realtime.js.map +1 -1
- package/dist/realtime/openai-realtime.d.ts +12 -1
- package/dist/realtime/openai-realtime.d.ts.map +1 -1
- package/dist/realtime/openai-realtime.js +182 -53
- package/dist/realtime/openai-realtime.js.map +1 -1
- package/dist/realtime/realtime-session.d.ts +14 -0
- package/dist/realtime/realtime-session.d.ts.map +1 -1
- package/dist/realtime/realtime-session.js +70 -4
- 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 +128 -49
- 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 +75 -10
- package/dist/stt/openai-stt.js.map +1 -1
- package/dist/tools.d.ts +8 -0
- package/dist/tools.d.ts.map +1 -1
- package/dist/tools.js +12 -4
- package/dist/tools.js.map +1 -1
- package/dist/transport/index.d.ts +1 -0
- package/dist/transport/index.d.ts.map +1 -1
- package/dist/transport/ws-transport.d.ts +14 -0
- package/dist/transport/ws-transport.d.ts.map +1 -1
- package/dist/transport/ws-transport.js +119 -15
- package/dist/transport/ws-transport.js.map +1 -1
- package/dist/tts/elevenlabs-tts.d.ts +1 -0
- package/dist/tts/elevenlabs-tts.d.ts.map +1 -1
- package/dist/tts/elevenlabs-tts.js +36 -17
- package/dist/tts/elevenlabs-tts.js.map +1 -1
- package/dist/tts/openai-tts.d.ts +1 -0
- package/dist/tts/openai-tts.d.ts.map +1 -1
- package/dist/tts/openai-tts.js +23 -9
- package/dist/tts/openai-tts.js.map +1 -1
- package/dist/types.d.ts +39 -16
- package/dist/types.d.ts.map +1 -1
- package/dist/vad/silero-vad.d.ts +15 -0
- package/dist/vad/silero-vad.d.ts.map +1 -1
- package/dist/vad/silero-vad.js +109 -20
- 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 +117 -28
- package/dist/voice-agent.js.map +1 -1
- package/package.json +5 -5
package/dist/vad/silero-vad.js
CHANGED
|
@@ -1,58 +1,146 @@
|
|
|
1
|
+
const V4_STATE_SIZE = 2 * 1 * 64;
|
|
2
|
+
const V5_STATE_SIZE = 2 * 1 * 128;
|
|
3
|
+
/**
|
|
4
|
+
* Silero VAD (ONNX). Supports both the v4 (`h`/`c` state) and v5 (`state`) model
|
|
5
|
+
* signatures, detected from the model inputs. Audio of any length is accepted:
|
|
6
|
+
* samples are buffered and evaluated in fixed-size frames (512 @ 16kHz, 256 @ 8kHz).
|
|
7
|
+
*/
|
|
1
8
|
export class SileroVAD {
|
|
2
9
|
name = 'silero';
|
|
3
10
|
modelPath;
|
|
4
11
|
threshold;
|
|
5
12
|
silenceDuration;
|
|
6
13
|
sampleRate;
|
|
14
|
+
frameSize;
|
|
15
|
+
contextSize;
|
|
7
16
|
session = null;
|
|
8
17
|
ort = null;
|
|
9
|
-
|
|
10
|
-
|
|
18
|
+
version = 'v4';
|
|
19
|
+
hn = new Float32Array(V4_STATE_SIZE);
|
|
20
|
+
cn = new Float32Array(V4_STATE_SIZE);
|
|
21
|
+
stateV5 = new Float32Array(V5_STATE_SIZE);
|
|
22
|
+
context;
|
|
23
|
+
pending = new Float32Array(0);
|
|
11
24
|
state = 'idle';
|
|
12
25
|
speechSamples = 0;
|
|
13
26
|
silenceMs = 0;
|
|
27
|
+
lastSpeechDurationMs = 0;
|
|
14
28
|
constructor(config) {
|
|
29
|
+
const sampleRate = config.sampleRate ?? 16000;
|
|
30
|
+
if (sampleRate !== 16000 && sampleRate !== 8000) {
|
|
31
|
+
throw new Error(`SileroVAD: sampleRate must be 8000 or 16000, got ${sampleRate}`);
|
|
32
|
+
}
|
|
15
33
|
this.modelPath = config.modelPath;
|
|
16
34
|
this.threshold = config.threshold ?? 0.5;
|
|
17
35
|
this.silenceDuration = config.silenceDuration ?? 500;
|
|
18
|
-
this.sampleRate =
|
|
19
|
-
this.
|
|
20
|
-
this.
|
|
36
|
+
this.sampleRate = sampleRate;
|
|
37
|
+
this.frameSize = sampleRate === 16000 ? 512 : 256;
|
|
38
|
+
this.contextSize = sampleRate === 16000 ? 64 : 32;
|
|
39
|
+
this.context = new Float32Array(this.contextSize);
|
|
21
40
|
}
|
|
22
41
|
async init() {
|
|
23
42
|
if (this.session)
|
|
24
43
|
return;
|
|
25
44
|
const moduleName = 'onnxruntime-node';
|
|
45
|
+
let ort;
|
|
26
46
|
try {
|
|
27
|
-
|
|
47
|
+
ort = (await import(/* webpackIgnore: true */ moduleName));
|
|
28
48
|
}
|
|
29
49
|
catch {
|
|
30
50
|
throw new Error('onnxruntime-node is required for SileroVAD. Install it with: npm install onnxruntime-node');
|
|
31
51
|
}
|
|
32
|
-
|
|
52
|
+
const session = await ort.InferenceSession.create(this.modelPath);
|
|
53
|
+
this.version = session.inputNames?.includes('state') ? 'v5' : 'v4';
|
|
54
|
+
this.ort = ort;
|
|
55
|
+
this.session = session;
|
|
33
56
|
}
|
|
34
57
|
async process(samples) {
|
|
35
58
|
if (!this.session || !this.ort) {
|
|
36
59
|
throw new Error('SileroVAD: must call init() before process()');
|
|
37
60
|
}
|
|
38
|
-
|
|
39
|
-
|
|
40
|
-
|
|
41
|
-
const
|
|
42
|
-
|
|
43
|
-
|
|
44
|
-
const
|
|
45
|
-
|
|
46
|
-
|
|
47
|
-
|
|
48
|
-
|
|
61
|
+
if (samples.length === 0) {
|
|
62
|
+
return this.state === 'speaking' ? { type: 'speech', probability: 0 } : { type: 'silence' };
|
|
63
|
+
}
|
|
64
|
+
const buffered = new Float32Array(this.pending.length + samples.length);
|
|
65
|
+
buffered.set(this.pending);
|
|
66
|
+
buffered.set(samples, this.pending.length);
|
|
67
|
+
const startState = this.state;
|
|
68
|
+
let transitioned = false;
|
|
69
|
+
let lastSpeechProbability = 0;
|
|
70
|
+
let offset = 0;
|
|
71
|
+
while (offset + this.frameSize <= buffered.length) {
|
|
72
|
+
const frame = buffered.subarray(offset, offset + this.frameSize);
|
|
73
|
+
offset += this.frameSize;
|
|
74
|
+
const probability = await this.infer(frame);
|
|
75
|
+
const event = this.evaluate(probability, frame.length);
|
|
76
|
+
if (event.type === 'speech_start' || event.type === 'speech_end') {
|
|
77
|
+
transitioned = true;
|
|
78
|
+
}
|
|
79
|
+
else if (event.type === 'speech') {
|
|
80
|
+
lastSpeechProbability = event.probability;
|
|
81
|
+
}
|
|
82
|
+
}
|
|
83
|
+
this.pending = buffered.slice(offset);
|
|
84
|
+
return this.summarize(startState, transitioned, lastSpeechProbability);
|
|
49
85
|
}
|
|
50
86
|
reset() {
|
|
51
87
|
this.state = 'idle';
|
|
52
88
|
this.speechSamples = 0;
|
|
53
89
|
this.silenceMs = 0;
|
|
54
|
-
this.hn = new Float32Array(
|
|
55
|
-
this.cn = new Float32Array(
|
|
90
|
+
this.hn = new Float32Array(V4_STATE_SIZE);
|
|
91
|
+
this.cn = new Float32Array(V4_STATE_SIZE);
|
|
92
|
+
this.stateV5 = new Float32Array(V5_STATE_SIZE);
|
|
93
|
+
this.context = new Float32Array(this.contextSize);
|
|
94
|
+
this.pending = new Float32Array(0);
|
|
95
|
+
}
|
|
96
|
+
dispose() {
|
|
97
|
+
this.session = null;
|
|
98
|
+
this.ort = null;
|
|
99
|
+
this.reset();
|
|
100
|
+
}
|
|
101
|
+
summarize(startState, transitioned, probability) {
|
|
102
|
+
if (startState === 'idle' && this.state === 'speaking') {
|
|
103
|
+
return { type: 'speech_start' };
|
|
104
|
+
}
|
|
105
|
+
if (startState === 'speaking' && this.state === 'idle') {
|
|
106
|
+
return { type: 'speech_end', duration: this.lastSpeechDurationMs };
|
|
107
|
+
}
|
|
108
|
+
if (this.state === 'speaking') {
|
|
109
|
+
return { type: 'speech', probability: transitioned ? 1 : probability };
|
|
110
|
+
}
|
|
111
|
+
return { type: 'silence' };
|
|
112
|
+
}
|
|
113
|
+
async infer(frame) {
|
|
114
|
+
const { Tensor } = this.ort;
|
|
115
|
+
const sr = new Tensor('int64', BigInt64Array.from([BigInt(this.sampleRate)]), []);
|
|
116
|
+
if (this.version === 'v5') {
|
|
117
|
+
const input = new Float32Array(this.contextSize + frame.length);
|
|
118
|
+
input.set(this.context);
|
|
119
|
+
input.set(frame, this.contextSize);
|
|
120
|
+
this.context = input.slice(input.length - this.contextSize);
|
|
121
|
+
const result = await this.session.run({
|
|
122
|
+
input: new Tensor('float32', input, [1, input.length]),
|
|
123
|
+
state: new Tensor('float32', this.stateV5, [2, 1, 128]),
|
|
124
|
+
sr,
|
|
125
|
+
});
|
|
126
|
+
const stateN = result.stateN?.data;
|
|
127
|
+
if (stateN)
|
|
128
|
+
this.stateV5 = Float32Array.from(stateN);
|
|
129
|
+
return Number(result.output?.data[0] ?? 0);
|
|
130
|
+
}
|
|
131
|
+
const result = await this.session.run({
|
|
132
|
+
input: new Tensor('float32', frame, [1, frame.length]),
|
|
133
|
+
sr,
|
|
134
|
+
h: new Tensor('float32', this.hn, [2, 1, 64]),
|
|
135
|
+
c: new Tensor('float32', this.cn, [2, 1, 64]),
|
|
136
|
+
});
|
|
137
|
+
const hn = result.hn?.data;
|
|
138
|
+
const cn = result.cn?.data;
|
|
139
|
+
if (hn)
|
|
140
|
+
this.hn = Float32Array.from(hn);
|
|
141
|
+
if (cn)
|
|
142
|
+
this.cn = Float32Array.from(cn);
|
|
143
|
+
return Number(result.output?.data[0] ?? 0);
|
|
56
144
|
}
|
|
57
145
|
evaluate(probability, chunkLength) {
|
|
58
146
|
const isSpeech = probability >= this.threshold;
|
|
@@ -74,6 +162,7 @@ export class SileroVAD {
|
|
|
74
162
|
this.silenceMs += chunkDurationMs;
|
|
75
163
|
if (this.silenceMs >= this.silenceDuration) {
|
|
76
164
|
const duration = (this.speechSamples / this.sampleRate) * 1000;
|
|
165
|
+
this.lastSpeechDurationMs = duration;
|
|
77
166
|
this.state = 'idle';
|
|
78
167
|
this.speechSamples = 0;
|
|
79
168
|
this.silenceMs = 0;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"silero-vad.js","sourceRoot":"","sources":["../../src/vad/silero-vad.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"silero-vad.js","sourceRoot":"","sources":["../../src/vad/silero-vad.ts"],"names":[],"mappings":"AA2BA,MAAM,aAAa,GAAG,CAAC,GAAG,CAAC,GAAG,EAAE,CAAC;AACjC,MAAM,aAAa,GAAG,CAAC,GAAG,CAAC,GAAG,GAAG,CAAC;AAElC;;;;GAIG;AACH,MAAM,OAAO,SAAS;IACX,IAAI,GAAG,QAAQ,CAAC;IAER,SAAS,CAAS;IAClB,SAAS,CAAS;IAClB,eAAe,CAAS;IACxB,UAAU,CAAS;IACnB,SAAS,CAAS;IAClB,WAAW,CAAS;IAE7B,OAAO,GAAuB,IAAI,CAAC;IACnC,GAAG,GAAuB,IAAI,CAAC;IAC/B,OAAO,GAAiB,IAAI,CAAC;IAE7B,EAAE,GAAG,IAAI,YAAY,CAAC,aAAa,CAAC,CAAC;IACrC,EAAE,GAAG,IAAI,YAAY,CAAC,aAAa,CAAC,CAAC;IACrC,OAAO,GAAG,IAAI,YAAY,CAAC,aAAa,CAAC,CAAC;IAC1C,OAAO,CAAe;IACtB,OAAO,GAAG,IAAI,YAAY,CAAC,CAAC,CAAC,CAAC;IAE9B,KAAK,GAAwB,MAAM,CAAC;IACpC,aAAa,GAAG,CAAC,CAAC;IAClB,SAAS,GAAG,CAAC,CAAC;IACd,oBAAoB,GAAG,CAAC,CAAC;IAEjC,YAAY,MAAuB;QACjC,MAAM,UAAU,GAAG,MAAM,CAAC,UAAU,IAAI,KAAK,CAAC;QAC9C,IAAI,UAAU,KAAK,KAAK,IAAI,UAAU,KAAK,IAAI,EAAE,CAAC;YAChD,MAAM,IAAI,KAAK,CAAC,oDAAoD,UAAU,EAAE,CAAC,CAAC;QACpF,CAAC;QACD,IAAI,CAAC,SAAS,GAAG,MAAM,CAAC,SAAS,CAAC;QAClC,IAAI,CAAC,SAAS,GAAG,MAAM,CAAC,SAAS,IAAI,GAAG,CAAC;QACzC,IAAI,CAAC,eAAe,GAAG,MAAM,CAAC,eAAe,IAAI,GAAG,CAAC;QACrD,IAAI,CAAC,UAAU,GAAG,UAAU,CAAC;QAC7B,IAAI,CAAC,SAAS,GAAG,UAAU,KAAK,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,GAAG,CAAC;QAClD,IAAI,CAAC,WAAW,GAAG,UAAU,KAAK,KAAK,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QAClD,IAAI,CAAC,OAAO,GAAG,IAAI,YAAY,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;IACpD,CAAC;IAED,KAAK,CAAC,IAAI;QACR,IAAI,IAAI,CAAC,OAAO;YAAE,OAAO;QAEzB,MAAM,UAAU,GAAG,kBAAkB,CAAC;QACtC,IAAI,GAAgB,CAAC;QACrB,IAAI,CAAC;YACH,GAAG,GAAG,CAAC,MAAM,MAAM,CAAC,yBAAyB,CAAC,UAAU,CAAC,CAAgB,CAAC;QAC5E,CAAC;QAAC,MAAM,CAAC;YACP,MAAM,IAAI,KAAK,CACb,2FAA2F,CAC5F,CAAC;QACJ,CAAC;QACD,MAAM,OAAO,GAAG,MAAM,GAAG,CAAC,gBAAgB,CAAC,MAAM,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;QAClE,IAAI,CAAC,OAAO,GAAG,OAAO,CAAC,UAAU,EAAE,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC;QACnE,IAAI,CAAC,GAAG,GAAG,GAAG,CAAC;QACf,IAAI,CAAC,OAAO,GAAG,OAAO,CAAC;IACzB,CAAC;IAED,KAAK,CAAC,OAAO,CAAC,OAAqB;QACjC,IAAI,CAAC,IAAI,CAAC,OAAO,IAAI,CAAC,IAAI,CAAC,GAAG,EAAE,CAAC;YAC/B,MAAM,IAAI,KAAK,CAAC,8CAA8C,CAAC,CAAC;QAClE,CAAC;QAED,IAAI,OAAO,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YACzB,OAAO,IAAI,CAAC,KAAK,KAAK,UAAU,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,SAAS,EAAE,CAAC;QAC9F,CAAC;QAED,MAAM,QAAQ,GAAG,IAAI,YAAY,CAAC,IAAI,CAAC,OAAO,CAAC,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;QACxE,QAAQ,CAAC,GAAG,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;QAC3B,QAAQ,CAAC,GAAG,CAAC,OAAO,EAAE,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;QAE3C,MAAM,UAAU,GAAG,IAAI,CAAC,KAAK,CAAC;QAC9B,IAAI,YAAY,GAAG,KAAK,CAAC;QACzB,IAAI,qBAAqB,GAAG,CAAC,CAAC;QAC9B,IAAI,MAAM,GAAG,CAAC,CAAC;QAEf,OAAO,MAAM,GAAG,IAAI,CAAC,SAAS,IAAI,QAAQ,CAAC,MAAM,EAAE,CAAC;YAClD,MAAM,KAAK,GAAG,QAAQ,CAAC,QAAQ,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI,CAAC,SAAS,CAAC,CAAC;YACjE,MAAM,IAAI,IAAI,CAAC,SAAS,CAAC;YACzB,MAAM,WAAW,GAAG,MAAM,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;YAC5C,MAAM,KAAK,GAAG,IAAI,CAAC,QAAQ,CAAC,WAAW,EAAE,KAAK,CAAC,MAAM,CAAC,CAAC;YACvD,IAAI,KAAK,CAAC,IAAI,KAAK,cAAc,IAAI,KAAK,CAAC,IAAI,KAAK,YAAY,EAAE,CAAC;gBACjE,YAAY,GAAG,IAAI,CAAC;YACtB,CAAC;iBAAM,IAAI,KAAK,CAAC,IAAI,KAAK,QAAQ,EAAE,CAAC;gBACnC,qBAAqB,GAAG,KAAK,CAAC,WAAW,CAAC;YAC5C,CAAC;QACH,CAAC;QAED,IAAI,CAAC,OAAO,GAAG,QAAQ,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;QAEtC,OAAO,IAAI,CAAC,SAAS,CAAC,UAAU,EAAE,YAAY,EAAE,qBAAqB,CAAC,CAAC;IACzE,CAAC;IAED,KAAK;QACH,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC;QACpB,IAAI,CAAC,aAAa,GAAG,CAAC,CAAC;QACvB,IAAI,CAAC,SAAS,GAAG,CAAC,CAAC;QACnB,IAAI,CAAC,EAAE,GAAG,IAAI,YAAY,CAAC,aAAa,CAAC,CAAC;QAC1C,IAAI,CAAC,EAAE,GAAG,IAAI,YAAY,CAAC,aAAa,CAAC,CAAC;QAC1C,IAAI,CAAC,OAAO,GAAG,IAAI,YAAY,CAAC,aAAa,CAAC,CAAC;QAC/C,IAAI,CAAC,OAAO,GAAG,IAAI,YAAY,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;QAClD,IAAI,CAAC,OAAO,GAAG,IAAI,YAAY,CAAC,CAAC,CAAC,CAAC;IACrC,CAAC;IAED,OAAO;QACL,IAAI,CAAC,OAAO,GAAG,IAAI,CAAC;QACpB,IAAI,CAAC,GAAG,GAAG,IAAI,CAAC;QAChB,IAAI,CAAC,KAAK,EAAE,CAAC;IACf,CAAC;IAEO,SAAS,CACf,UAA+B,EAC/B,YAAqB,EACrB,WAAmB;QAEnB,IAAI,UAAU,KAAK,MAAM,IAAI,IAAI,CAAC,KAAK,KAAK,UAAU,EAAE,CAAC;YACvD,OAAO,EAAE,IAAI,EAAE,cAAc,EAAE,CAAC;QAClC,CAAC;QACD,IAAI,UAAU,KAAK,UAAU,IAAI,IAAI,CAAC,KAAK,KAAK,MAAM,EAAE,CAAC;YACvD,OAAO,EAAE,IAAI,EAAE,YAAY,EAAE,QAAQ,EAAE,IAAI,CAAC,oBAAoB,EAAE,CAAC;QACrE,CAAC;QACD,IAAI,IAAI,CAAC,KAAK,KAAK,UAAU,EAAE,CAAC;YAC9B,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,YAAY,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,EAAE,CAAC;QACzE,CAAC;QACD,OAAO,EAAE,IAAI,EAAE,SAAS,EAAE,CAAC;IAC7B,CAAC;IAEO,KAAK,CAAC,KAAK,CAAC,KAAmB;QACrC,MAAM,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC,GAAI,CAAC;QAC7B,MAAM,EAAE,GAAG,IAAI,MAAM,CAAC,OAAO,EAAE,aAAa,CAAC,IAAI,CAAC,CAAC,MAAM,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;QAElF,IAAI,IAAI,CAAC,OAAO,KAAK,IAAI,EAAE,CAAC;YAC1B,MAAM,KAAK,GAAG,IAAI,YAAY,CAAC,IAAI,CAAC,WAAW,GAAG,KAAK,CAAC,MAAM,CAAC,CAAC;YAChE,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;YACxB,KAAK,CAAC,GAAG,CAAC,KAAK,EAAE,IAAI,CAAC,WAAW,CAAC,CAAC;YACnC,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC,KAAK,CAAC,KAAK,CAAC,MAAM,GAAG,IAAI,CAAC,WAAW,CAAC,CAAC;YAE5D,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,OAAQ,CAAC,GAAG,CAAC;gBACrC,KAAK,EAAE,IAAI,MAAM,CAAC,SAAS,EAAE,KAAK,EAAE,CAAC,CAAC,EAAE,KAAK,CAAC,MAAM,CAAC,CAAC;gBACtD,KAAK,EAAE,IAAI,MAAM,CAAC,SAAS,EAAE,IAAI,CAAC,OAAO,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,CAAC;gBACvD,EAAE;aACH,CAAC,CAAC;YACH,MAAM,MAAM,GAAG,MAAM,CAAC,MAAM,EAAE,IAAI,CAAC;YACnC,IAAI,MAAM;gBAAE,IAAI,CAAC,OAAO,GAAG,YAAY,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YACrD,OAAO,MAAM,CAAC,MAAM,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC;QAC7C,CAAC;QAED,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,OAAQ,CAAC,GAAG,CAAC;YACrC,KAAK,EAAE,IAAI,MAAM,CAAC,SAAS,EAAE,KAAK,EAAE,CAAC,CAAC,EAAE,KAAK,CAAC,MAAM,CAAC,CAAC;YACtD,EAAE;YACF,CAAC,EAAE,IAAI,MAAM,CAAC,SAAS,EAAE,IAAI,CAAC,EAAE,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC;YAC7C,CAAC,EAAE,IAAI,MAAM,CAAC,SAAS,EAAE,IAAI,CAAC,EAAE,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC;SAC9C,CAAC,CAAC;QACH,MAAM,EAAE,GAAG,MAAM,CAAC,EAAE,EAAE,IAAI,CAAC;QAC3B,MAAM,EAAE,GAAG,MAAM,CAAC,EAAE,EAAE,IAAI,CAAC;QAC3B,IAAI,EAAE;YAAE,IAAI,CAAC,EAAE,GAAG,YAAY,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QACxC,IAAI,EAAE;YAAE,IAAI,CAAC,EAAE,GAAG,YAAY,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QACxC,OAAO,MAAM,CAAC,MAAM,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC;IAC7C,CAAC;IAEO,QAAQ,CAAC,WAAmB,EAAE,WAAmB;QACvD,MAAM,QAAQ,GAAG,WAAW,IAAI,IAAI,CAAC,SAAS,CAAC;QAC/C,MAAM,eAAe,GAAG,CAAC,WAAW,GAAG,IAAI,CAAC,UAAU,CAAC,GAAG,IAAI,CAAC;QAE/D,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,WAAW,CAAC;gBACjC,IAAI,CAAC,SAAS,GAAG,CAAC,CAAC;gBACnB,OAAO,EAAE,IAAI,EAAE,cAAc,EAAE,CAAC;YAClC,CAAC;YACD,OAAO,EAAE,IAAI,EAAE,SAAS,EAAE,CAAC;QAC7B,CAAC;QAED,IAAI,QAAQ,EAAE,CAAC;YACb,IAAI,CAAC,aAAa,IAAI,WAAW,CAAC;YAClC,IAAI,CAAC,SAAS,GAAG,CAAC,CAAC;YACnB,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,CAAC;QACzC,CAAC;QAED,IAAI,CAAC,SAAS,IAAI,eAAe,CAAC;QAElC,IAAI,IAAI,CAAC,SAAS,IAAI,IAAI,CAAC,eAAe,EAAE,CAAC;YAC3C,MAAM,QAAQ,GAAG,CAAC,IAAI,CAAC,aAAa,GAAG,IAAI,CAAC,UAAU,CAAC,GAAG,IAAI,CAAC;YAC/D,IAAI,CAAC,oBAAoB,GAAG,QAAQ,CAAC;YACrC,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC;YACpB,IAAI,CAAC,aAAa,GAAG,CAAC,CAAC;YACvB,IAAI,CAAC,SAAS,GAAG,CAAC,CAAC;YACnB,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;CACF"}
|
package/dist/voice-agent.d.ts
CHANGED
|
@@ -1,10 +1,19 @@
|
|
|
1
1
|
import { EventEmitter } from 'node:events';
|
|
2
|
+
import type http from 'node:http';
|
|
2
3
|
import type { VoiceAgentConfig } from './types.js';
|
|
3
4
|
interface VoiceAgentEvents {
|
|
4
5
|
session_start: [sessionId: string];
|
|
5
6
|
session_end: [sessionId: string];
|
|
6
7
|
error: [error: Error];
|
|
7
8
|
}
|
|
9
|
+
/**
|
|
10
|
+
* Serves voice sessions over WebSocket.
|
|
11
|
+
*
|
|
12
|
+
* Client -> server: binary frames carry PCM16 audio; JSON text frames carry control
|
|
13
|
+
* messages `{ type: 'interrupt' }`, `{ type: 'end_of_speech' }` and `{ type: 'text', text }`.
|
|
14
|
+
* Server -> client: binary audio frames and JSON events (`transcript`, `agent_response`,
|
|
15
|
+
* `speech_start`, `speech_end`, `turn_end`).
|
|
16
|
+
*/
|
|
8
17
|
export declare class VoiceAgent extends EventEmitter<VoiceAgentEvents> {
|
|
9
18
|
private readonly config;
|
|
10
19
|
private transport;
|
|
@@ -14,11 +23,16 @@ export declare class VoiceAgent extends EventEmitter<VoiceAgentEvents> {
|
|
|
14
23
|
get port(): number | undefined;
|
|
15
24
|
get activeSessions(): number;
|
|
16
25
|
listen(port: number): Promise<void>;
|
|
26
|
+
/** Serve voice sessions on an existing HTTP server (shares its port). */
|
|
27
|
+
attach(server: http.Server): void;
|
|
17
28
|
close(): Promise<void>;
|
|
29
|
+
private createTransport;
|
|
18
30
|
private validateConfig;
|
|
19
31
|
private handleConnection;
|
|
32
|
+
private handleControlMessage;
|
|
20
33
|
private setupPipelineSession;
|
|
21
34
|
private setupRealtimeSession;
|
|
35
|
+
private reportError;
|
|
22
36
|
private cleanupSession;
|
|
23
37
|
}
|
|
24
38
|
export {};
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"voice-agent.d.ts","sourceRoot":"","sources":["../src/voice-agent.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,YAAY,CAAC;AAYnD,UAAU,gBAAgB;IACxB,aAAa,EAAE,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC;IACnC,WAAW,EAAE,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC;IACjC,KAAK,EAAE,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;CACvB;
|
|
1
|
+
{"version":3,"file":"voice-agent.d.ts","sourceRoot":"","sources":["../src/voice-agent.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAClC,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,YAAY,CAAC;AAYnD,UAAU,gBAAgB;IACxB,aAAa,EAAE,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC;IACnC,WAAW,EAAE,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC;IACjC,KAAK,EAAE,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;CACvB;AAeD;;;;;;;GAOG;AACH,qBAAa,UAAW,SAAQ,YAAY,CAAC,gBAAgB,CAAC;IAC5D,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAmB;IAC1C,OAAO,CAAC,SAAS,CAAmC;IACpD,OAAO,CAAC,QAAQ,CAAC,QAAQ,CAAmC;IAC5D,OAAO,CAAC,QAAQ,CAA8B;gBAElC,MAAM,EAAE,gBAAgB;IAepC,IAAI,IAAI,IAAI,MAAM,GAAG,SAAS,CAE7B;IAED,IAAI,cAAc,IAAI,MAAM,CAE3B;IAEK,MAAM,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IAWzC,yEAAyE;IACzE,MAAM,CAAC,MAAM,EAAE,IAAI,CAAC,MAAM,GAAG,IAAI;IAI3B,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAgB5B,OAAO,CAAC,eAAe;IAUvB,OAAO,CAAC,cAAc;IAatB,OAAO,CAAC,gBAAgB;IA2BxB,OAAO,CAAC,oBAAoB;IA2B5B,OAAO,CAAC,oBAAoB;IAqC5B,OAAO,CAAC,oBAAoB;IAsD5B,OAAO,CAAC,WAAW;YAML,cAAc;CAM7B"}
|
package/dist/voice-agent.js
CHANGED
|
@@ -2,6 +2,25 @@ import { EventEmitter } from 'node:events';
|
|
|
2
2
|
import { WebSocketTransport } from './transport/ws-transport.js';
|
|
3
3
|
import { VoicePipeline } from './pipeline/voice-pipeline.js';
|
|
4
4
|
import { RealtimeSession } from './realtime/realtime-session.js';
|
|
5
|
+
const WS_INTERNAL_ERROR = 1011;
|
|
6
|
+
const MAX_CLOSE_REASON_BYTES = 120;
|
|
7
|
+
function toError(err) {
|
|
8
|
+
return err instanceof Error ? err : new Error(String(err));
|
|
9
|
+
}
|
|
10
|
+
function closeReason(message) {
|
|
11
|
+
const buf = Buffer.from(message);
|
|
12
|
+
if (buf.length <= MAX_CLOSE_REASON_BYTES)
|
|
13
|
+
return message;
|
|
14
|
+
return buf.subarray(0, MAX_CLOSE_REASON_BYTES).toString().replace(/�+$/, '');
|
|
15
|
+
}
|
|
16
|
+
/**
|
|
17
|
+
* Serves voice sessions over WebSocket.
|
|
18
|
+
*
|
|
19
|
+
* Client -> server: binary frames carry PCM16 audio; JSON text frames carry control
|
|
20
|
+
* messages `{ type: 'interrupt' }`, `{ type: 'end_of_speech' }` and `{ type: 'text', text }`.
|
|
21
|
+
* Server -> client: binary audio frames and JSON events (`transcript`, `agent_response`,
|
|
22
|
+
* `speech_start`, `speech_end`, `turn_end`).
|
|
23
|
+
*/
|
|
5
24
|
export class VoiceAgent extends EventEmitter {
|
|
6
25
|
config;
|
|
7
26
|
transport = null;
|
|
@@ -27,19 +46,41 @@ export class VoiceAgent extends EventEmitter {
|
|
|
27
46
|
return this.sessions.size;
|
|
28
47
|
}
|
|
29
48
|
async listen(port) {
|
|
30
|
-
|
|
31
|
-
|
|
32
|
-
|
|
49
|
+
const transport = this.createTransport();
|
|
50
|
+
try {
|
|
51
|
+
await transport.listen(port);
|
|
52
|
+
}
|
|
53
|
+
catch (err) {
|
|
54
|
+
this.transport = null;
|
|
55
|
+
await transport.close().catch(() => { });
|
|
56
|
+
throw err;
|
|
57
|
+
}
|
|
58
|
+
}
|
|
59
|
+
/** Serve voice sessions on an existing HTTP server (shares its port). */
|
|
60
|
+
attach(server) {
|
|
61
|
+
this.createTransport().attachToServer(server);
|
|
33
62
|
}
|
|
34
63
|
async close() {
|
|
35
|
-
|
|
36
|
-
await this.cleanupSession(id, entry);
|
|
37
|
-
}
|
|
64
|
+
const entries = [...this.sessions.entries()];
|
|
38
65
|
this.sessions.clear();
|
|
66
|
+
await Promise.allSettled(entries.map(([, entry]) => this.cleanupSession(entry)));
|
|
67
|
+
for (const [id] of entries) {
|
|
68
|
+
this.emit('session_end', id);
|
|
69
|
+
}
|
|
39
70
|
if (this.transport) {
|
|
40
|
-
|
|
71
|
+
const transport = this.transport;
|
|
41
72
|
this.transport = null;
|
|
73
|
+
await transport.close();
|
|
74
|
+
}
|
|
75
|
+
}
|
|
76
|
+
createTransport() {
|
|
77
|
+
if (this.transport) {
|
|
78
|
+
throw new Error('VoiceAgent is already listening — call close() first');
|
|
42
79
|
}
|
|
80
|
+
const transport = new WebSocketTransport(this.config.transport);
|
|
81
|
+
transport.on('connection', (client) => this.handleConnection(client));
|
|
82
|
+
this.transport = transport;
|
|
83
|
+
return transport;
|
|
43
84
|
}
|
|
44
85
|
validateConfig(config) {
|
|
45
86
|
if (config.mode === 'pipeline') {
|
|
@@ -54,29 +95,57 @@ export class VoiceAgent extends EventEmitter {
|
|
|
54
95
|
if (!config.realtimeApiKey)
|
|
55
96
|
throw new Error('realtimeApiKey is required for realtime mode');
|
|
56
97
|
}
|
|
98
|
+
else {
|
|
99
|
+
throw new Error(`Unknown mode: ${String(config.mode)}`);
|
|
100
|
+
}
|
|
57
101
|
}
|
|
58
102
|
handleConnection(client) {
|
|
59
103
|
const sessionId = client.id;
|
|
60
104
|
const entry = { client };
|
|
61
|
-
if (this.config.mode === 'pipeline') {
|
|
62
|
-
this.setupPipelineSession(sessionId, entry);
|
|
63
|
-
}
|
|
64
|
-
else {
|
|
65
|
-
this.setupRealtimeSession(sessionId, entry);
|
|
66
|
-
}
|
|
67
105
|
this.sessions.set(sessionId, entry);
|
|
68
|
-
this.emit('session_start', sessionId);
|
|
69
106
|
client.on('close', () => {
|
|
70
|
-
|
|
71
|
-
|
|
107
|
+
if (!this.sessions.delete(sessionId))
|
|
108
|
+
return;
|
|
109
|
+
void this.cleanupSession(entry).catch((err) => this.reportError(toError(err)));
|
|
72
110
|
this.emit('session_end', sessionId);
|
|
73
111
|
});
|
|
74
112
|
client.on('error', (err) => {
|
|
75
|
-
this.
|
|
113
|
+
this.reportError(err);
|
|
76
114
|
});
|
|
115
|
+
client.on('message', (msg) => this.handleControlMessage(entry, msg));
|
|
116
|
+
if (this.config.mode === 'pipeline') {
|
|
117
|
+
this.setupPipelineSession(entry);
|
|
118
|
+
}
|
|
119
|
+
else {
|
|
120
|
+
this.setupRealtimeSession(entry);
|
|
121
|
+
}
|
|
122
|
+
this.emit('session_start', sessionId);
|
|
123
|
+
}
|
|
124
|
+
handleControlMessage(entry, msg) {
|
|
125
|
+
switch (msg.type) {
|
|
126
|
+
case 'interrupt':
|
|
127
|
+
entry.pipelineSession?.interrupt();
|
|
128
|
+
entry.realtimeSession?.interrupt();
|
|
129
|
+
break;
|
|
130
|
+
case 'end_of_speech':
|
|
131
|
+
entry.pipelineSession?.endAudio();
|
|
132
|
+
break;
|
|
133
|
+
case 'text':
|
|
134
|
+
if (typeof msg.text !== 'string' || msg.text.trim().length === 0) {
|
|
135
|
+
this.reportError(new Error('Control message "text" requires a non-empty text field'));
|
|
136
|
+
return;
|
|
137
|
+
}
|
|
138
|
+
if (entry.pipelineSession) {
|
|
139
|
+
void entry.pipelineSession.sendText(msg.text);
|
|
140
|
+
}
|
|
141
|
+
entry.realtimeSession?.sendText(msg.text);
|
|
142
|
+
break;
|
|
143
|
+
default:
|
|
144
|
+
this.reportError(new Error(`Unknown control message type: ${String(msg.type)}`));
|
|
145
|
+
}
|
|
77
146
|
}
|
|
78
|
-
setupPipelineSession(
|
|
79
|
-
const session = this.pipeline.createSession();
|
|
147
|
+
setupPipelineSession(entry) {
|
|
148
|
+
const session = this.pipeline.createSession({ sessionId: entry.client.id });
|
|
80
149
|
entry.pipelineSession = session;
|
|
81
150
|
entry.client.on('audio', (chunk) => {
|
|
82
151
|
session.pushAudio(chunk);
|
|
@@ -96,16 +165,21 @@ export class VoiceAgent extends EventEmitter {
|
|
|
96
165
|
session.on('speech_end', () => {
|
|
97
166
|
entry.client.sendMessage({ type: 'speech_end' });
|
|
98
167
|
});
|
|
168
|
+
session.on('turn_end', () => {
|
|
169
|
+
entry.client.sendMessage({ type: 'turn_end' });
|
|
170
|
+
});
|
|
99
171
|
session.on('error', (err) => {
|
|
100
|
-
this.
|
|
172
|
+
this.reportError(err);
|
|
101
173
|
});
|
|
102
174
|
}
|
|
103
|
-
setupRealtimeSession(
|
|
175
|
+
setupRealtimeSession(entry) {
|
|
104
176
|
const session = new RealtimeSession({
|
|
105
177
|
provider: this.config.realtimeProvider,
|
|
106
178
|
apiKey: this.config.realtimeApiKey,
|
|
107
179
|
model: this.config.realtimeModel,
|
|
108
180
|
voice: this.config.voice,
|
|
181
|
+
instructions: this.config.instructions ?? this.config.agent.instructions,
|
|
182
|
+
tools: this.config.tools,
|
|
109
183
|
});
|
|
110
184
|
entry.realtimeSession = session;
|
|
111
185
|
entry.client.on('audio', (chunk) => {
|
|
@@ -120,20 +194,35 @@ export class VoiceAgent extends EventEmitter {
|
|
|
120
194
|
session.on('speech_start', () => {
|
|
121
195
|
entry.client.sendMessage({ type: 'speech_start' });
|
|
122
196
|
});
|
|
197
|
+
session.on('turn_end', () => {
|
|
198
|
+
entry.client.sendMessage({ type: 'turn_end' });
|
|
199
|
+
});
|
|
123
200
|
session.on('error', (err) => {
|
|
124
|
-
this.
|
|
201
|
+
this.reportError(err);
|
|
202
|
+
});
|
|
203
|
+
session.on('disconnected', (code, reason) => {
|
|
204
|
+
if (!this.sessions.has(entry.client.id))
|
|
205
|
+
return;
|
|
206
|
+
entry.client.close(WS_INTERNAL_ERROR, closeReason(`Realtime provider disconnected (${code}): ${reason}`));
|
|
125
207
|
});
|
|
126
|
-
|
|
127
|
-
this.
|
|
208
|
+
session.connect().catch((err) => {
|
|
209
|
+
if (!this.sessions.has(entry.client.id))
|
|
210
|
+
return;
|
|
211
|
+
const error = toError(err);
|
|
212
|
+
this.reportError(error);
|
|
213
|
+
entry.client.close(WS_INTERNAL_ERROR, closeReason(`Realtime connect failed: ${error.message}`));
|
|
128
214
|
});
|
|
129
215
|
}
|
|
130
|
-
|
|
216
|
+
reportError(err) {
|
|
217
|
+
if (this.listenerCount('error') > 0) {
|
|
218
|
+
this.emit('error', err);
|
|
219
|
+
}
|
|
220
|
+
}
|
|
221
|
+
async cleanupSession(entry) {
|
|
222
|
+
entry.realtimeSession?.close();
|
|
131
223
|
if (entry.pipelineSession) {
|
|
132
224
|
await entry.pipelineSession.close();
|
|
133
225
|
}
|
|
134
|
-
if (entry.realtimeSession) {
|
|
135
|
-
entry.realtimeSession.close();
|
|
136
|
-
}
|
|
137
226
|
}
|
|
138
227
|
}
|
|
139
228
|
//# sourceMappingURL=voice-agent.js.map
|
package/dist/voice-agent.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"voice-agent.js","sourceRoot":"","sources":["../src/voice-agent.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;
|
|
1
|
+
{"version":3,"file":"voice-agent.js","sourceRoot":"","sources":["../src/voice-agent.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAG3C,OAAO,EAAE,kBAAkB,EAAe,MAAM,6BAA6B,CAAC;AAC9E,OAAO,EAAE,aAAa,EAAE,MAAM,8BAA8B,CAAC;AAE7D,OAAO,EAAE,eAAe,EAAE,MAAM,gCAAgC,CAAC;AAcjE,MAAM,iBAAiB,GAAG,IAAI,CAAC;AAC/B,MAAM,sBAAsB,GAAG,GAAG,CAAC;AAEnC,SAAS,OAAO,CAAC,GAAY;IAC3B,OAAO,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC;AAC7D,CAAC;AAED,SAAS,WAAW,CAAC,OAAe;IAClC,MAAM,GAAG,GAAG,MAAM,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;IACjC,IAAI,GAAG,CAAC,MAAM,IAAI,sBAAsB;QAAE,OAAO,OAAO,CAAC;IACzD,OAAO,GAAG,CAAC,QAAQ,CAAC,CAAC,EAAE,sBAAsB,CAAC,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC;AAC/E,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,OAAO,UAAW,SAAQ,YAA8B;IAC3C,MAAM,CAAmB;IAClC,SAAS,GAA8B,IAAI,CAAC;IACnC,QAAQ,GAAG,IAAI,GAAG,EAAwB,CAAC;IACpD,QAAQ,GAAyB,IAAI,CAAC;IAE9C,YAAY,MAAwB;QAClC,KAAK,EAAE,CAAC;QACR,IAAI,CAAC,cAAc,CAAC,MAAM,CAAC,CAAC;QAC5B,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;QAErB,IAAI,MAAM,CAAC,IAAI,KAAK,UAAU,EAAE,CAAC;YAC/B,IAAI,CAAC,QAAQ,GAAG,IAAI,aAAa,CAAC;gBAChC,GAAG,EAAE,MAAM,CAAC,GAAI;gBAChB,GAAG,EAAE,MAAM,CAAC,GAAI;gBAChB,GAAG,EAAE,MAAM,CAAC,GAAG;gBACf,KAAK,EAAE,MAAM,CAAC,KAAK;aACpB,CAAC,CAAC;QACL,CAAC;IACH,CAAC;IAED,IAAI,IAAI;QACN,OAAO,IAAI,CAAC,SAAS,EAAE,IAAI,CAAC;IAC9B,CAAC;IAED,IAAI,cAAc;QAChB,OAAO,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC;IAC5B,CAAC;IAED,KAAK,CAAC,MAAM,CAAC,IAAY;QACvB,MAAM,SAAS,GAAG,IAAI,CAAC,eAAe,EAAE,CAAC;QACzC,IAAI,CAAC;YACH,MAAM,SAAS,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC;QAC/B,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC;YACtB,MAAM,SAAS,CAAC,KAAK,EAAE,CAAC,KAAK,CAAC,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;YACxC,MAAM,GAAG,CAAC;QACZ,CAAC;IACH,CAAC;IAED,yEAAyE;IACzE,MAAM,CAAC,MAAmB;QACxB,IAAI,CAAC,eAAe,EAAE,CAAC,cAAc,CAAC,MAAM,CAAC,CAAC;IAChD,CAAC;IAED,KAAK,CAAC,KAAK;QACT,MAAM,OAAO,GAAG,CAAC,GAAG,IAAI,CAAC,QAAQ,CAAC,OAAO,EAAE,CAAC,CAAC;QAC7C,IAAI,CAAC,QAAQ,CAAC,KAAK,EAAE,CAAC;QAEtB,MAAM,OAAO,CAAC,UAAU,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,KAAK,CAAC,EAAE,EAAE,CAAC,IAAI,CAAC,cAAc,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QACjF,KAAK,MAAM,CAAC,EAAE,CAAC,IAAI,OAAO,EAAE,CAAC;YAC3B,IAAI,CAAC,IAAI,CAAC,aAAa,EAAE,EAAE,CAAC,CAAC;QAC/B,CAAC;QAED,IAAI,IAAI,CAAC,SAAS,EAAE,CAAC;YACnB,MAAM,SAAS,GAAG,IAAI,CAAC,SAAS,CAAC;YACjC,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC;YACtB,MAAM,SAAS,CAAC,KAAK,EAAE,CAAC;QAC1B,CAAC;IACH,CAAC;IAEO,eAAe;QACrB,IAAI,IAAI,CAAC,SAAS,EAAE,CAAC;YACnB,MAAM,IAAI,KAAK,CAAC,sDAAsD,CAAC,CAAC;QAC1E,CAAC;QACD,MAAM,SAAS,GAAG,IAAI,kBAAkB,CAAC,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC;QAChE,SAAS,CAAC,EAAE,CAAC,YAAY,EAAE,CAAC,MAAM,EAAE,EAAE,CAAC,IAAI,CAAC,gBAAgB,CAAC,MAAM,CAAC,CAAC,CAAC;QACtE,IAAI,CAAC,SAAS,GAAG,SAAS,CAAC;QAC3B,OAAO,SAAS,CAAC;IACnB,CAAC;IAEO,cAAc,CAAC,MAAwB;QAC7C,IAAI,MAAM,CAAC,IAAI,KAAK,UAAU,EAAE,CAAC;YAC/B,IAAI,CAAC,MAAM,CAAC,GAAG;gBAAE,MAAM,IAAI,KAAK,CAAC,mCAAmC,CAAC,CAAC;YACtE,IAAI,CAAC,MAAM,CAAC,GAAG;gBAAE,MAAM,IAAI,KAAK,CAAC,mCAAmC,CAAC,CAAC;QACxE,CAAC;aAAM,IAAI,MAAM,CAAC,IAAI,KAAK,UAAU,EAAE,CAAC;YACtC,IAAI,CAAC,MAAM,CAAC,gBAAgB;gBAC1B,MAAM,IAAI,KAAK,CAAC,gDAAgD,CAAC,CAAC;YACpE,IAAI,CAAC,MAAM,CAAC,cAAc;gBAAE,MAAM,IAAI,KAAK,CAAC,8CAA8C,CAAC,CAAC;QAC9F,CAAC;aAAM,CAAC;YACN,MAAM,IAAI,KAAK,CAAC,iBAAiB,MAAM,CAAE,MAA4B,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QACjF,CAAC;IACH,CAAC;IAEO,gBAAgB,CAAC,MAAmB;QAC1C,MAAM,SAAS,GAAG,MAAM,CAAC,EAAE,CAAC;QAC5B,MAAM,KAAK,GAAiB,EAAE,MAAM,EAAE,CAAC;QAEvC,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,SAAS,EAAE,KAAK,CAAC,CAAC;QAEpC,MAAM,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE;YACtB,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,MAAM,CAAC,SAAS,CAAC;gBAAE,OAAO;YAC7C,KAAK,IAAI,CAAC,cAAc,CAAC,KAAK,CAAC,CAAC,KAAK,CAAC,CAAC,GAAY,EAAE,EAAE,CAAC,IAAI,CAAC,WAAW,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YACxF,IAAI,CAAC,IAAI,CAAC,aAAa,EAAE,SAAS,CAAC,CAAC;QACtC,CAAC,CAAC,CAAC;QAEH,MAAM,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAG,EAAE,EAAE;YACzB,IAAI,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC;QACxB,CAAC,CAAC,CAAC;QAEH,MAAM,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,GAAG,EAAE,EAAE,CAAC,IAAI,CAAC,oBAAoB,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC,CAAC;QAErE,IAAI,IAAI,CAAC,MAAM,CAAC,IAAI,KAAK,UAAU,EAAE,CAAC;YACpC,IAAI,CAAC,oBAAoB,CAAC,KAAK,CAAC,CAAC;QACnC,CAAC;aAAM,CAAC;YACN,IAAI,CAAC,oBAAoB,CAAC,KAAK,CAAC,CAAC;QACnC,CAAC;QAED,IAAI,CAAC,IAAI,CAAC,eAAe,EAAE,SAAS,CAAC,CAAC;IACxC,CAAC;IAEO,oBAAoB,CAAC,KAAmB,EAAE,GAA4B;QAC5E,QAAQ,GAAG,CAAC,IAAI,EAAE,CAAC;YACjB,KAAK,WAAW;gBACd,KAAK,CAAC,eAAe,EAAE,SAAS,EAAE,CAAC;gBACnC,KAAK,CAAC,eAAe,EAAE,SAAS,EAAE,CAAC;gBACnC,MAAM;YAER,KAAK,eAAe;gBAClB,KAAK,CAAC,eAAe,EAAE,QAAQ,EAAE,CAAC;gBAClC,MAAM;YAER,KAAK,MAAM;gBACT,IAAI,OAAO,GAAG,CAAC,IAAI,KAAK,QAAQ,IAAI,GAAG,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;oBACjE,IAAI,CAAC,WAAW,CAAC,IAAI,KAAK,CAAC,wDAAwD,CAAC,CAAC,CAAC;oBACtF,OAAO;gBACT,CAAC;gBACD,IAAI,KAAK,CAAC,eAAe,EAAE,CAAC;oBAC1B,KAAK,KAAK,CAAC,eAAe,CAAC,QAAQ,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;gBAChD,CAAC;gBACD,KAAK,CAAC,eAAe,EAAE,QAAQ,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;gBAC1C,MAAM;YAER;gBACE,IAAI,CAAC,WAAW,CAAC,IAAI,KAAK,CAAC,iCAAiC,MAAM,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC;QACrF,CAAC;IACH,CAAC;IAEO,oBAAoB,CAAC,KAAmB;QAC9C,MAAM,OAAO,GAAG,IAAI,CAAC,QAAS,CAAC,aAAa,CAAC,EAAE,SAAS,EAAE,KAAK,CAAC,MAAM,CAAC,EAAE,EAAE,CAAC,CAAC;QAC7E,KAAK,CAAC,eAAe,GAAG,OAAO,CAAC;QAEhC,KAAK,CAAC,MAAM,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,KAAK,EAAE,EAAE;YACjC,OAAO,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC;QAC3B,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,KAAa,EAAE,EAAE;YACpC,KAAK,CAAC,MAAM,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC;QAChC,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,YAAY,EAAE,CAAC,IAAY,EAAE,OAAgB,EAAE,EAAE;YAC1D,KAAK,CAAC,MAAM,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,YAAY,EAAE,IAAI,EAAE,OAAO,EAAE,CAAC,CAAC;QAClE,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,gBAAgB,EAAE,CAAC,IAAY,EAAE,EAAE;YAC5C,KAAK,CAAC,MAAM,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,gBAAgB,EAAE,IAAI,EAAE,CAAC,CAAC;QAC7D,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,cAAc,EAAE,GAAG,EAAE;YAC9B,KAAK,CAAC,MAAM,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,cAAc,EAAE,CAAC,CAAC;QACrD,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,YAAY,EAAE,GAAG,EAAE;YAC5B,KAAK,CAAC,MAAM,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,YAAY,EAAE,CAAC,CAAC;QACnD,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,UAAU,EAAE,GAAG,EAAE;YAC1B,KAAK,CAAC,MAAM,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,UAAU,EAAE,CAAC,CAAC;QACjD,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAU,EAAE,EAAE;YACjC,IAAI,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC;QACxB,CAAC,CAAC,CAAC;IACL,CAAC;IAEO,oBAAoB,CAAC,KAAmB;QAC9C,MAAM,OAAO,GAAG,IAAI,eAAe,CAAC;YAClC,QAAQ,EAAE,IAAI,CAAC,MAAM,CAAC,gBAAiB;YACvC,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC,cAAe;YACnC,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,aAAa;YAChC,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,KAAK;YACxB,YAAY,EAAE,IAAI,CAAC,MAAM,CAAC,YAAY,IAAI,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,YAAY;YACxE,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,KAAK;SACzB,CAAC,CAAC;QACH,KAAK,CAAC,eAAe,GAAG,OAAO,CAAC;QAEhC,KAAK,CAAC,MAAM,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,KAAK,EAAE,EAAE;YACjC,OAAO,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC;QAC3B,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,KAAa,EAAE,EAAE;YACpC,KAAK,CAAC,MAAM,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC;QAChC,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,YAAY,EAAE,CAAC,IAAY,EAAE,IAA0B,EAAE,EAAE;YACpE,KAAK,CAAC,MAAM,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,YAAY,EAAE,IAAI,EAAE,IAAI,EAAE,CAAC,CAAC;QAC/D,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,cAAc,EAAE,GAAG,EAAE;YAC9B,KAAK,CAAC,MAAM,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,cAAc,EAAE,CAAC,CAAC;QACrD,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,UAAU,EAAE,GAAG,EAAE;YAC1B,KAAK,CAAC,MAAM,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,UAAU,EAAE,CAAC,CAAC;QACjD,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAU,EAAE,EAAE;YACjC,IAAI,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC;QACxB,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,EAAE,CAAC,cAAc,EAAE,CAAC,IAAY,EAAE,MAAc,EAAE,EAAE;YAC1D,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,MAAM,CAAC,EAAE,CAAC;gBAAE,OAAO;YAChD,KAAK,CAAC,MAAM,CAAC,KAAK,CAChB,iBAAiB,EACjB,WAAW,CAAC,mCAAmC,IAAI,MAAM,MAAM,EAAE,CAAC,CACnE,CAAC;QACJ,CAAC,CAAC,CAAC;QAEH,OAAO,CAAC,OAAO,EAAE,CAAC,KAAK,CAAC,CAAC,GAAY,EAAE,EAAE;YACvC,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,MAAM,CAAC,EAAE,CAAC;gBAAE,OAAO;YAChD,MAAM,KAAK,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC;YAC3B,IAAI,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC;YACxB,KAAK,CAAC,MAAM,CAAC,KAAK,CAChB,iBAAiB,EACjB,WAAW,CAAC,4BAA4B,KAAK,CAAC,OAAO,EAAE,CAAC,CACzD,CAAC;QACJ,CAAC,CAAC,CAAC;IACL,CAAC;IAEO,WAAW,CAAC,GAAU;QAC5B,IAAI,IAAI,CAAC,aAAa,CAAC,OAAO,CAAC,GAAG,CAAC,EAAE,CAAC;YACpC,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;QAC1B,CAAC;IACH,CAAC;IAEO,KAAK,CAAC,cAAc,CAAC,KAAmB;QAC9C,KAAK,CAAC,eAAe,EAAE,KAAK,EAAE,CAAC;QAC/B,IAAI,KAAK,CAAC,eAAe,EAAE,CAAC;YAC1B,MAAM,KAAK,CAAC,eAAe,CAAC,KAAK,EAAE,CAAC;QACtC,CAAC;IACH,CAAC;CACF"}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@cogitator-ai/voice",
|
|
3
|
-
"version": "0.
|
|
3
|
+
"version": "0.2.0",
|
|
4
4
|
"description": "Voice/Realtime agent capabilities for Cogitator",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"main": "./dist/index.js",
|
|
@@ -18,11 +18,11 @@
|
|
|
18
18
|
"nanoid": "^5.0.4",
|
|
19
19
|
"ws": "^8.18.0",
|
|
20
20
|
"zod": "^4.3.6",
|
|
21
|
-
"@cogitator-ai/types": "0.
|
|
21
|
+
"@cogitator-ai/types": "0.23.0"
|
|
22
22
|
},
|
|
23
23
|
"peerDependencies": {
|
|
24
|
-
"
|
|
25
|
-
"
|
|
24
|
+
"@cogitator-ai/core": ">=0.20.0",
|
|
25
|
+
"openai": "^6.0.0"
|
|
26
26
|
},
|
|
27
27
|
"peerDependenciesMeta": {
|
|
28
28
|
"@cogitator-ai/core": {
|
|
@@ -48,7 +48,7 @@
|
|
|
48
48
|
},
|
|
49
49
|
"license": "MIT",
|
|
50
50
|
"scripts": {
|
|
51
|
-
"build": "tsc",
|
|
51
|
+
"build": "tsc && tsx ../../scripts/fix-esm-imports.ts .",
|
|
52
52
|
"dev": "tsc --watch",
|
|
53
53
|
"clean": "rm -rf dist",
|
|
54
54
|
"typecheck": "tsc --noEmit",
|