@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
|
@@ -1,74 +1,149 @@
|
|
|
1
1
|
import { EventEmitter } from 'node:events';
|
|
2
2
|
import { WebSocket } from 'ws';
|
|
3
|
-
const DEFAULT_MODEL = 'gemini-
|
|
3
|
+
const DEFAULT_MODEL = 'gemini-3.8-live';
|
|
4
|
+
const DEFAULT_VOICE = 'Puck';
|
|
4
5
|
const BASE_URL = 'wss://generativelanguage.googleapis.com/ws/google.ai.generativelanguage.v1beta.GenerativeService.BidiGenerateContent';
|
|
6
|
+
const CONNECT_TIMEOUT_MS = 30_000;
|
|
7
|
+
const INPUT_MIME_TYPE = 'audio/pcm;rate=16000';
|
|
8
|
+
function rawDataToString(data) {
|
|
9
|
+
if (Array.isArray(data))
|
|
10
|
+
return Buffer.concat(data).toString();
|
|
11
|
+
if (data instanceof ArrayBuffer)
|
|
12
|
+
return Buffer.from(new Uint8Array(data)).toString();
|
|
13
|
+
return data.toString();
|
|
14
|
+
}
|
|
15
|
+
/**
|
|
16
|
+
* Gemini Live API adapter. Input audio is PCM16 mono at 16kHz,
|
|
17
|
+
* output audio is PCM16 mono at 24kHz.
|
|
18
|
+
*/
|
|
5
19
|
export class GeminiRealtimeAdapter extends EventEmitter {
|
|
6
20
|
config;
|
|
7
21
|
model;
|
|
8
22
|
ws = null;
|
|
23
|
+
connected = false;
|
|
24
|
+
modelTurnActive = false;
|
|
25
|
+
interrupting = false;
|
|
26
|
+
userTranscript = '';
|
|
27
|
+
assistantTranscript = '';
|
|
28
|
+
assistantText = '';
|
|
9
29
|
constructor(config) {
|
|
10
30
|
super();
|
|
11
31
|
this.config = config;
|
|
12
32
|
this.model = config.model ?? DEFAULT_MODEL;
|
|
13
33
|
}
|
|
34
|
+
get isConnected() {
|
|
35
|
+
return this.connected;
|
|
36
|
+
}
|
|
14
37
|
async connect() {
|
|
15
|
-
|
|
38
|
+
if (this.ws) {
|
|
39
|
+
throw new Error('Already connected or connecting — call close() first');
|
|
40
|
+
}
|
|
16
41
|
return new Promise((resolve, reject) => {
|
|
17
|
-
|
|
18
|
-
|
|
42
|
+
let settled = false;
|
|
43
|
+
const settle = (err) => {
|
|
44
|
+
if (settled)
|
|
45
|
+
return;
|
|
46
|
+
settled = true;
|
|
47
|
+
clearTimeout(timer);
|
|
48
|
+
if (err)
|
|
49
|
+
reject(err);
|
|
50
|
+
else
|
|
51
|
+
resolve();
|
|
52
|
+
};
|
|
53
|
+
const ws = new WebSocket(BASE_URL, {
|
|
54
|
+
headers: { 'x-goog-api-key': this.config.apiKey },
|
|
55
|
+
});
|
|
56
|
+
this.ws = ws;
|
|
57
|
+
const timer = setTimeout(() => {
|
|
58
|
+
if (this.ws === ws)
|
|
59
|
+
this.ws = null;
|
|
60
|
+
ws.removeAllListeners();
|
|
61
|
+
ws.on('error', () => { });
|
|
62
|
+
ws.terminate();
|
|
63
|
+
settle(new Error(`Connect timed out after ${CONNECT_TIMEOUT_MS}ms`));
|
|
64
|
+
}, CONNECT_TIMEOUT_MS);
|
|
65
|
+
ws.on('open', () => {
|
|
19
66
|
this.sendSetup();
|
|
20
67
|
});
|
|
21
|
-
|
|
22
|
-
const raw = typeof data === 'string' ? data : data.toString();
|
|
68
|
+
ws.on('message', (data) => {
|
|
23
69
|
let msg;
|
|
24
70
|
try {
|
|
25
|
-
msg = JSON.parse(
|
|
71
|
+
msg = JSON.parse(rawDataToString(data));
|
|
26
72
|
}
|
|
27
73
|
catch {
|
|
28
74
|
this.emit('error', new Error('Failed to parse WebSocket message'));
|
|
29
75
|
return;
|
|
30
76
|
}
|
|
77
|
+
if (msg.error) {
|
|
78
|
+
const err = new Error(msg.error.message ?? `Gemini error (code: ${msg.error.code ?? 'unknown'})`);
|
|
79
|
+
if (settled)
|
|
80
|
+
this.emit('error', err);
|
|
81
|
+
else
|
|
82
|
+
settle(err);
|
|
83
|
+
return;
|
|
84
|
+
}
|
|
31
85
|
if (msg.setupComplete) {
|
|
86
|
+
this.connected = true;
|
|
32
87
|
this.emit('connected');
|
|
33
|
-
|
|
88
|
+
settle();
|
|
34
89
|
return;
|
|
35
90
|
}
|
|
36
|
-
|
|
91
|
+
if (msg.serverContent) {
|
|
92
|
+
this.handleServerContent(msg.serverContent);
|
|
93
|
+
}
|
|
94
|
+
if (msg.toolCall?.functionCalls?.length) {
|
|
95
|
+
void this.handleToolCalls(msg.toolCall.functionCalls);
|
|
96
|
+
}
|
|
97
|
+
});
|
|
98
|
+
ws.on('error', (err) => {
|
|
99
|
+
if (settled)
|
|
100
|
+
this.emit('error', err);
|
|
101
|
+
else
|
|
102
|
+
settle(err);
|
|
37
103
|
});
|
|
38
|
-
|
|
39
|
-
|
|
40
|
-
|
|
104
|
+
ws.on('close', (code, reason) => {
|
|
105
|
+
const reasonStr = reason.toString() || 'connection closed';
|
|
106
|
+
if (this.ws === ws)
|
|
107
|
+
this.ws = null;
|
|
108
|
+
const wasConnected = this.connected;
|
|
109
|
+
this.resetTurn();
|
|
110
|
+
this.connected = false;
|
|
111
|
+
settle(new Error(`WebSocket closed before connect (code ${code}): ${reasonStr}`));
|
|
112
|
+
if (wasConnected)
|
|
113
|
+
this.emit('disconnected', code, reasonStr);
|
|
41
114
|
});
|
|
42
115
|
});
|
|
43
116
|
}
|
|
44
117
|
pushAudio(chunk) {
|
|
45
118
|
this.send({
|
|
46
119
|
realtimeInput: {
|
|
47
|
-
|
|
48
|
-
{
|
|
49
|
-
mimeType: 'audio/pcm;rate=16000',
|
|
50
|
-
data: chunk.toString('base64'),
|
|
51
|
-
},
|
|
52
|
-
],
|
|
120
|
+
audio: { mimeType: INPUT_MIME_TYPE, data: chunk.toString('base64') },
|
|
53
121
|
},
|
|
54
122
|
});
|
|
55
123
|
}
|
|
56
124
|
sendText(text) {
|
|
57
|
-
this.send({
|
|
58
|
-
clientContent: {
|
|
59
|
-
turns: [{ role: 'user', parts: [{ text }] }],
|
|
60
|
-
turnComplete: true,
|
|
61
|
-
},
|
|
62
|
-
});
|
|
125
|
+
this.send({ realtimeInput: { text } });
|
|
63
126
|
}
|
|
127
|
+
/**
|
|
128
|
+
* Gemini Live barge-in is driven by incoming user audio, not a control message.
|
|
129
|
+
* While a model turn is in progress this drops the remaining model audio until
|
|
130
|
+
* the turn completes, simulating an interruption on the consumer side.
|
|
131
|
+
*/
|
|
64
132
|
interrupt() {
|
|
65
|
-
this.
|
|
66
|
-
|
|
67
|
-
}
|
|
133
|
+
if (this.modelTurnActive) {
|
|
134
|
+
this.interrupting = true;
|
|
135
|
+
}
|
|
68
136
|
}
|
|
69
137
|
close() {
|
|
70
|
-
this.ws
|
|
138
|
+
const ws = this.ws;
|
|
139
|
+
this.connected = false;
|
|
140
|
+
this.resetTurn();
|
|
141
|
+
if (!ws)
|
|
142
|
+
return;
|
|
71
143
|
this.ws = null;
|
|
144
|
+
ws.removeAllListeners();
|
|
145
|
+
ws.on('error', () => { });
|
|
146
|
+
ws.close();
|
|
72
147
|
}
|
|
73
148
|
sendSetup() {
|
|
74
149
|
const setup = {
|
|
@@ -77,17 +152,15 @@ export class GeminiRealtimeAdapter extends EventEmitter {
|
|
|
77
152
|
responseModalities: ['AUDIO'],
|
|
78
153
|
speechConfig: {
|
|
79
154
|
voiceConfig: {
|
|
80
|
-
prebuiltVoiceConfig: {
|
|
81
|
-
voiceName: this.config.voice ?? 'Puck',
|
|
82
|
-
},
|
|
155
|
+
prebuiltVoiceConfig: { voiceName: this.config.voice ?? DEFAULT_VOICE },
|
|
83
156
|
},
|
|
84
157
|
},
|
|
85
158
|
},
|
|
159
|
+
inputAudioTranscription: {},
|
|
160
|
+
outputAudioTranscription: {},
|
|
86
161
|
};
|
|
87
162
|
if (this.config.instructions) {
|
|
88
|
-
setup.systemInstruction = {
|
|
89
|
-
parts: [{ text: this.config.instructions }],
|
|
90
|
-
};
|
|
163
|
+
setup.systemInstruction = { parts: [{ text: this.config.instructions }] };
|
|
91
164
|
}
|
|
92
165
|
if (this.config.tools?.length) {
|
|
93
166
|
setup.tools = [
|
|
@@ -102,54 +175,89 @@ export class GeminiRealtimeAdapter extends EventEmitter {
|
|
|
102
175
|
}
|
|
103
176
|
this.send({ setup });
|
|
104
177
|
}
|
|
105
|
-
|
|
106
|
-
|
|
107
|
-
|
|
178
|
+
handleServerContent(content) {
|
|
179
|
+
const inputText = content.inputTranscription?.text;
|
|
180
|
+
if (inputText) {
|
|
181
|
+
this.userTranscript += inputText;
|
|
108
182
|
}
|
|
109
|
-
|
|
110
|
-
|
|
183
|
+
const outputText = content.outputTranscription?.text;
|
|
184
|
+
const parts = content.modelTurn?.parts ?? [];
|
|
185
|
+
if (parts.length > 0 || outputText) {
|
|
186
|
+
this.modelTurnActive = true;
|
|
187
|
+
this.flushUserTranscript();
|
|
111
188
|
}
|
|
112
|
-
|
|
113
|
-
|
|
114
|
-
|
|
115
|
-
|
|
116
|
-
|
|
117
|
-
|
|
189
|
+
if (outputText) {
|
|
190
|
+
this.assistantTranscript += outputText;
|
|
191
|
+
}
|
|
192
|
+
for (const part of parts) {
|
|
193
|
+
if (part.thought)
|
|
194
|
+
continue;
|
|
195
|
+
if (part.inlineData?.data && !this.interrupting) {
|
|
118
196
|
this.emit('audio', Buffer.from(part.inlineData.data, 'base64'));
|
|
119
197
|
}
|
|
120
198
|
if (part.text) {
|
|
121
|
-
this.
|
|
199
|
+
this.assistantText += part.text;
|
|
122
200
|
}
|
|
123
201
|
}
|
|
202
|
+
if (content.interrupted) {
|
|
203
|
+
this.flushAssistantTranscript();
|
|
204
|
+
this.modelTurnActive = false;
|
|
205
|
+
this.interrupting = false;
|
|
206
|
+
this.emit('speech_start');
|
|
207
|
+
}
|
|
208
|
+
if (content.turnComplete) {
|
|
209
|
+
this.flushUserTranscript();
|
|
210
|
+
this.flushAssistantTranscript();
|
|
211
|
+
this.modelTurnActive = false;
|
|
212
|
+
this.interrupting = false;
|
|
213
|
+
this.emit('turn_end');
|
|
214
|
+
}
|
|
215
|
+
}
|
|
216
|
+
flushUserTranscript() {
|
|
217
|
+
const text = this.userTranscript.trim();
|
|
218
|
+
this.userTranscript = '';
|
|
219
|
+
if (text)
|
|
220
|
+
this.emit('transcript', text, 'user');
|
|
221
|
+
}
|
|
222
|
+
flushAssistantTranscript() {
|
|
223
|
+
const text = (this.assistantTranscript || this.assistantText).trim();
|
|
224
|
+
this.assistantTranscript = '';
|
|
225
|
+
this.assistantText = '';
|
|
226
|
+
if (text)
|
|
227
|
+
this.emit('transcript', text, 'assistant');
|
|
228
|
+
}
|
|
229
|
+
resetTurn() {
|
|
230
|
+
this.modelTurnActive = false;
|
|
231
|
+
this.interrupting = false;
|
|
232
|
+
this.userTranscript = '';
|
|
233
|
+
this.assistantTranscript = '';
|
|
234
|
+
this.assistantText = '';
|
|
124
235
|
}
|
|
125
236
|
async handleToolCalls(calls) {
|
|
126
|
-
const
|
|
127
|
-
|
|
237
|
+
const ws = this.ws;
|
|
238
|
+
const functionResponses = await Promise.all(calls.map(async (call, index) => {
|
|
239
|
+
const id = call.id ?? `${call.name}-${index}`;
|
|
240
|
+
const args = call.args ?? {};
|
|
241
|
+
this.emit('tool_call', call.name, args);
|
|
128
242
|
const tool = this.config.tools?.find((t) => t.name === call.name);
|
|
129
243
|
if (!tool) {
|
|
130
|
-
return {
|
|
131
|
-
id: call.id,
|
|
132
|
-
name: call.name,
|
|
133
|
-
response: { result: JSON.stringify({ error: `Unknown tool: ${call.name}` }) },
|
|
134
|
-
};
|
|
244
|
+
return { id, name: call.name, response: { error: `Unknown tool: ${call.name}` } };
|
|
135
245
|
}
|
|
136
|
-
let result;
|
|
137
246
|
try {
|
|
138
|
-
const output = await tool.execute(
|
|
139
|
-
|
|
247
|
+
const output = await tool.execute(args);
|
|
248
|
+
return { id, name: call.name, response: { result: output ?? null } };
|
|
140
249
|
}
|
|
141
250
|
catch (err) {
|
|
142
|
-
|
|
251
|
+
return {
|
|
252
|
+
id,
|
|
253
|
+
name: call.name,
|
|
254
|
+
response: { error: err instanceof Error ? err.message : String(err) },
|
|
255
|
+
};
|
|
143
256
|
}
|
|
144
|
-
return {
|
|
145
|
-
id: call.id,
|
|
146
|
-
name: call.name,
|
|
147
|
-
response: { result },
|
|
148
|
-
};
|
|
149
257
|
}));
|
|
150
|
-
this.
|
|
151
|
-
|
|
152
|
-
});
|
|
258
|
+
if (this.ws !== ws)
|
|
259
|
+
return;
|
|
260
|
+
this.send({ toolResponse: { functionResponses } });
|
|
153
261
|
}
|
|
154
262
|
send(data) {
|
|
155
263
|
const ws = this.ws;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"gemini-realtime.js","sourceRoot":"","sources":["../../src/realtime/gemini-realtime.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAG/B,MAAM,aAAa,GAAG,
|
|
1
|
+
{"version":3,"file":"gemini-realtime.js","sourceRoot":"","sources":["../../src/realtime/gemini-realtime.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAG/B,MAAM,aAAa,GAAG,iBAAiB,CAAC;AACxC,MAAM,aAAa,GAAG,MAAM,CAAC;AAC7B,MAAM,QAAQ,GACZ,sHAAsH,CAAC;AACzH,MAAM,kBAAkB,GAAG,MAAM,CAAC;AAClC,MAAM,eAAe,GAAG,sBAAsB,CAAC;AAwC/C,SAAS,eAAe,CAAC,IAAuB;IAC9C,IAAI,KAAK,CAAC,OAAO,CAAC,IAAI,CAAC;QAAE,OAAO,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,QAAQ,EAAE,CAAC;IAC/D,IAAI,IAAI,YAAY,WAAW;QAAE,OAAO,MAAM,CAAC,IAAI,CAAC,IAAI,UAAU,CAAC,IAAI,CAAC,CAAC,CAAC,QAAQ,EAAE,CAAC;IACrF,OAAO,IAAI,CAAC,QAAQ,EAAE,CAAC;AACzB,CAAC;AAED;;;GAGG;AACH,MAAM,OAAO,qBAAsB,SAAQ,YAAkC;IAC1D,MAAM,CAAwB;IAC9B,KAAK,CAAS;IACvB,EAAE,GAAqB,IAAI,CAAC;IAC5B,SAAS,GAAG,KAAK,CAAC;IAClB,eAAe,GAAG,KAAK,CAAC;IACxB,YAAY,GAAG,KAAK,CAAC;IACrB,cAAc,GAAG,EAAE,CAAC;IACpB,mBAAmB,GAAG,EAAE,CAAC;IACzB,aAAa,GAAG,EAAE,CAAC;IAE3B,YAAY,MAA6B;QACvC,KAAK,EAAE,CAAC;QACR,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;QACrB,IAAI,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,IAAI,aAAa,CAAC;IAC7C,CAAC;IAED,IAAI,WAAW;QACb,OAAO,IAAI,CAAC,SAAS,CAAC;IACxB,CAAC;IAED,KAAK,CAAC,OAAO;QACX,IAAI,IAAI,CAAC,EAAE,EAAE,CAAC;YACZ,MAAM,IAAI,KAAK,CAAC,sDAAsD,CAAC,CAAC;QAC1E,CAAC;QAED,OAAO,IAAI,OAAO,CAAO,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;YAC3C,IAAI,OAAO,GAAG,KAAK,CAAC;YACpB,MAAM,MAAM,GAAG,CAAC,GAAW,EAAE,EAAE;gBAC7B,IAAI,OAAO;oBAAE,OAAO;gBACpB,OAAO,GAAG,IAAI,CAAC;gBACf,YAAY,CAAC,KAAK,CAAC,CAAC;gBACpB,IAAI,GAAG;oBAAE,MAAM,CAAC,GAAG,CAAC,CAAC;;oBAChB,OAAO,EAAE,CAAC;YACjB,CAAC,CAAC;YAEF,MAAM,EAAE,GAAG,IAAI,SAAS,CAAC,QAAQ,EAAE;gBACjC,OAAO,EAAE,EAAE,gBAAgB,EAAE,IAAI,CAAC,MAAM,CAAC,MAAM,EAAE;aAClD,CAAC,CAAC;YACH,IAAI,CAAC,EAAE,GAAG,EAAE,CAAC;YAEb,MAAM,KAAK,GAAG,UAAU,CAAC,GAAG,EAAE;gBAC5B,IAAI,IAAI,CAAC,EAAE,KAAK,EAAE;oBAAE,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC;gBACnC,EAAE,CAAC,kBAAkB,EAAE,CAAC;gBACxB,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;gBACzB,EAAE,CAAC,SAAS,EAAE,CAAC;gBACf,MAAM,CAAC,IAAI,KAAK,CAAC,2BAA2B,kBAAkB,IAAI,CAAC,CAAC,CAAC;YACvE,CAAC,EAAE,kBAAkB,CAAC,CAAC;YAEvB,EAAE,CAAC,EAAE,CAAC,MAAM,EAAE,GAAG,EAAE;gBACjB,IAAI,CAAC,SAAS,EAAE,CAAC;YACnB,CAAC,CAAC,CAAC;YAEH,EAAE,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,IAAuB,EAAE,EAAE;gBAC3C,IAAI,GAAkB,CAAC;gBACvB,IAAI,CAAC;oBACH,GAAG,GAAG,IAAI,CAAC,KAAK,CAAC,eAAe,CAAC,IAAI,CAAC,CAAkB,CAAC;gBAC3D,CAAC;gBAAC,MAAM,CAAC;oBACP,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,mCAAmC,CAAC,CAAC,CAAC;oBACnE,OAAO;gBACT,CAAC;gBAED,IAAI,GAAG,CAAC,KAAK,EAAE,CAAC;oBACd,MAAM,GAAG,GAAG,IAAI,KAAK,CACnB,GAAG,CAAC,KAAK,CAAC,OAAO,IAAI,uBAAuB,GAAG,CAAC,KAAK,CAAC,IAAI,IAAI,SAAS,GAAG,CAC3E,CAAC;oBACF,IAAI,OAAO;wBAAE,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;;wBAChC,MAAM,CAAC,GAAG,CAAC,CAAC;oBACjB,OAAO;gBACT,CAAC;gBAED,IAAI,GAAG,CAAC,aAAa,EAAE,CAAC;oBACtB,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC;oBACtB,IAAI,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;oBACvB,MAAM,EAAE,CAAC;oBACT,OAAO;gBACT,CAAC;gBAED,IAAI,GAAG,CAAC,aAAa,EAAE,CAAC;oBACtB,IAAI,CAAC,mBAAmB,CAAC,GAAG,CAAC,aAAa,CAAC,CAAC;gBAC9C,CAAC;gBAED,IAAI,GAAG,CAAC,QAAQ,EAAE,aAAa,EAAE,MAAM,EAAE,CAAC;oBACxC,KAAK,IAAI,CAAC,eAAe,CAAC,GAAG,CAAC,QAAQ,CAAC,aAAa,CAAC,CAAC;gBACxD,CAAC;YACH,CAAC,CAAC,CAAC;YAEH,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAU,EAAE,EAAE;gBAC5B,IAAI,OAAO;oBAAE,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;;oBAChC,MAAM,CAAC,GAAG,CAAC,CAAC;YACnB,CAAC,CAAC,CAAC;YAEH,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,IAAY,EAAE,MAAc,EAAE,EAAE;gBAC9C,MAAM,SAAS,GAAG,MAAM,CAAC,QAAQ,EAAE,IAAI,mBAAmB,CAAC;gBAC3D,IAAI,IAAI,CAAC,EAAE,KAAK,EAAE;oBAAE,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC;gBACnC,MAAM,YAAY,GAAG,IAAI,CAAC,SAAS,CAAC;gBACpC,IAAI,CAAC,SAAS,EAAE,CAAC;gBACjB,IAAI,CAAC,SAAS,GAAG,KAAK,CAAC;gBACvB,MAAM,CAAC,IAAI,KAAK,CAAC,yCAAyC,IAAI,MAAM,SAAS,EAAE,CAAC,CAAC,CAAC;gBAClF,IAAI,YAAY;oBAAE,IAAI,CAAC,IAAI,CAAC,cAAc,EAAE,IAAI,EAAE,SAAS,CAAC,CAAC;YAC/D,CAAC,CAAC,CAAC;QACL,CAAC,CAAC,CAAC;IACL,CAAC;IAED,SAAS,CAAC,KAAa;QACrB,IAAI,CAAC,IAAI,CAAC;YACR,aAAa,EAAE;gBACb,KAAK,EAAE,EAAE,QAAQ,EAAE,eAAe,EAAE,IAAI,EAAE,KAAK,CAAC,QAAQ,CAAC,QAAQ,CAAC,EAAE;aACrE;SACF,CAAC,CAAC;IACL,CAAC;IAED,QAAQ,CAAC,IAAY;QACnB,IAAI,CAAC,IAAI,CAAC,EAAE,aAAa,EAAE,EAAE,IAAI,EAAE,EAAE,CAAC,CAAC;IACzC,CAAC;IAED;;;;OAIG;IACH,SAAS;QACP,IAAI,IAAI,CAAC,eAAe,EAAE,CAAC;YACzB,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC;QAC3B,CAAC;IACH,CAAC;IAED,KAAK;QACH,MAAM,EAAE,GAAG,IAAI,CAAC,EAAE,CAAC;QACnB,IAAI,CAAC,SAAS,GAAG,KAAK,CAAC;QACvB,IAAI,CAAC,SAAS,EAAE,CAAC;QACjB,IAAI,CAAC,EAAE;YAAE,OAAO;QAChB,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC;QACf,EAAE,CAAC,kBAAkB,EAAE,CAAC;QACxB,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;QACzB,EAAE,CAAC,KAAK,EAAE,CAAC;IACb,CAAC;IAEO,SAAS;QACf,MAAM,KAAK,GAA4B;YACrC,KAAK,EAAE,UAAU,IAAI,CAAC,KAAK,EAAE;YAC7B,gBAAgB,EAAE;gBAChB,kBAAkB,EAAE,CAAC,OAAO,CAAC;gBAC7B,YAAY,EAAE;oBACZ,WAAW,EAAE;wBACX,mBAAmB,EAAE,EAAE,SAAS,EAAE,IAAI,CAAC,MAAM,CAAC,KAAK,IAAI,aAAa,EAAE;qBACvE;iBACF;aACF;YACD,uBAAuB,EAAE,EAAE;YAC3B,wBAAwB,EAAE,EAAE;SAC7B,CAAC;QAEF,IAAI,IAAI,CAAC,MAAM,CAAC,YAAY,EAAE,CAAC;YAC7B,KAAK,CAAC,iBAAiB,GAAG,EAAE,KAAK,EAAE,CAAC,EAAE,IAAI,EAAE,IAAI,CAAC,MAAM,CAAC,YAAY,EAAE,CAAC,EAAE,CAAC;QAC5E,CAAC;QAED,IAAI,IAAI,CAAC,MAAM,CAAC,KAAK,EAAE,MAAM,EAAE,CAAC;YAC9B,KAAK,CAAC,KAAK,GAAG;gBACZ;oBACE,oBAAoB,EAAE,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;wBAClD,IAAI,EAAE,CAAC,CAAC,IAAI;wBACZ,WAAW,EAAE,CAAC,CAAC,WAAW;wBAC1B,UAAU,EAAE,CAAC,CAAC,UAAU;qBACzB,CAAC,CAAC;iBACJ;aACF,CAAC;QACJ,CAAC;QAED,IAAI,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC;IACvB,CAAC;IAEO,mBAAmB,CAAC,OAA4B;QACtD,MAAM,SAAS,GAAG,OAAO,CAAC,kBAAkB,EAAE,IAAI,CAAC;QACnD,IAAI,SAAS,EAAE,CAAC;YACd,IAAI,CAAC,cAAc,IAAI,SAAS,CAAC;QACnC,CAAC;QAED,MAAM,UAAU,GAAG,OAAO,CAAC,mBAAmB,EAAE,IAAI,CAAC;QACrD,MAAM,KAAK,GAAG,OAAO,CAAC,SAAS,EAAE,KAAK,IAAI,EAAE,CAAC;QAC7C,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC,IAAI,UAAU,EAAE,CAAC;YACnC,IAAI,CAAC,eAAe,GAAG,IAAI,CAAC;YAC5B,IAAI,CAAC,mBAAmB,EAAE,CAAC;QAC7B,CAAC;QAED,IAAI,UAAU,EAAE,CAAC;YACf,IAAI,CAAC,mBAAmB,IAAI,UAAU,CAAC;QACzC,CAAC;QAED,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE,CAAC;YACzB,IAAI,IAAI,CAAC,OAAO;gBAAE,SAAS;YAC3B,IAAI,IAAI,CAAC,UAAU,EAAE,IAAI,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,CAAC;gBAChD,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,IAAI,EAAE,QAAQ,CAAC,CAAC,CAAC;YAClE,CAAC;YACD,IAAI,IAAI,CAAC,IAAI,EAAE,CAAC;gBACd,IAAI,CAAC,aAAa,IAAI,IAAI,CAAC,IAAI,CAAC;YAClC,CAAC;QACH,CAAC;QAED,IAAI,OAAO,CAAC,WAAW,EAAE,CAAC;YACxB,IAAI,CAAC,wBAAwB,EAAE,CAAC;YAChC,IAAI,CAAC,eAAe,GAAG,KAAK,CAAC;YAC7B,IAAI,CAAC,YAAY,GAAG,KAAK,CAAC;YAC1B,IAAI,CAAC,IAAI,CAAC,cAAc,CAAC,CAAC;QAC5B,CAAC;QAED,IAAI,OAAO,CAAC,YAAY,EAAE,CAAC;YACzB,IAAI,CAAC,mBAAmB,EAAE,CAAC;YAC3B,IAAI,CAAC,wBAAwB,EAAE,CAAC;YAChC,IAAI,CAAC,eAAe,GAAG,KAAK,CAAC;YAC7B,IAAI,CAAC,YAAY,GAAG,KAAK,CAAC;YAC1B,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC;QACxB,CAAC;IACH,CAAC;IAEO,mBAAmB;QACzB,MAAM,IAAI,GAAG,IAAI,CAAC,cAAc,CAAC,IAAI,EAAE,CAAC;QACxC,IAAI,CAAC,cAAc,GAAG,EAAE,CAAC;QACzB,IAAI,IAAI;YAAE,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,IAAI,EAAE,MAAM,CAAC,CAAC;IAClD,CAAC;IAEO,wBAAwB;QAC9B,MAAM,IAAI,GAAG,CAAC,IAAI,CAAC,mBAAmB,IAAI,IAAI,CAAC,aAAa,CAAC,CAAC,IAAI,EAAE,CAAC;QACrE,IAAI,CAAC,mBAAmB,GAAG,EAAE,CAAC;QAC9B,IAAI,CAAC,aAAa,GAAG,EAAE,CAAC;QACxB,IAAI,IAAI;YAAE,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,IAAI,EAAE,WAAW,CAAC,CAAC;IACvD,CAAC;IAEO,SAAS;QACf,IAAI,CAAC,eAAe,GAAG,KAAK,CAAC;QAC7B,IAAI,CAAC,YAAY,GAAG,KAAK,CAAC;QAC1B,IAAI,CAAC,cAAc,GAAG,EAAE,CAAC;QACzB,IAAI,CAAC,mBAAmB,GAAG,EAAE,CAAC;QAC9B,IAAI,CAAC,aAAa,GAAG,EAAE,CAAC;IAC1B,CAAC;IAEO,KAAK,CAAC,eAAe,CAAC,KAA2B;QACvD,MAAM,EAAE,GAAG,IAAI,CAAC,EAAE,CAAC;QACnB,MAAM,iBAAiB,GAAG,MAAM,OAAO,CAAC,GAAG,CACzC,KAAK,CAAC,GAAG,CAAC,KAAK,EAAE,IAAI,EAAE,KAAK,EAAE,EAAE;YAC9B,MAAM,EAAE,GAAG,IAAI,CAAC,EAAE,IAAI,GAAG,IAAI,CAAC,IAAI,IAAI,KAAK,EAAE,CAAC;YAC9C,MAAM,IAAI,GAAG,IAAI,CAAC,IAAI,IAAI,EAAE,CAAC;YAC7B,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,IAAI,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC;YAExC,MAAM,IAAI,GAAG,IAAI,CAAC,MAAM,CAAC,KAAK,EAAE,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,IAAI,CAAC,IAAI,CAAC,CAAC;YAClE,IAAI,CAAC,IAAI,EAAE,CAAC;gBACV,OAAO,EAAE,EAAE,EAAE,IAAI,EAAE,IAAI,CAAC,IAAI,EAAE,QAAQ,EAAE,EAAE,KAAK,EAAE,iBAAiB,IAAI,CAAC,IAAI,EAAE,EAAE,EAAE,CAAC;YACpF,CAAC;YAED,IAAI,CAAC;gBACH,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;gBACxC,OAAO,EAAE,EAAE,EAAE,IAAI,EAAE,IAAI,CAAC,IAAI,EAAE,QAAQ,EAAE,EAAE,MAAM,EAAE,MAAM,IAAI,IAAI,EAAE,EAAE,CAAC;YACvE,CAAC;YAAC,OAAO,GAAG,EAAE,CAAC;gBACb,OAAO;oBACL,EAAE;oBACF,IAAI,EAAE,IAAI,CAAC,IAAI;oBACf,QAAQ,EAAE,EAAE,KAAK,EAAE,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE;iBACtE,CAAC;YACJ,CAAC;QACH,CAAC,CAAC,CACH,CAAC;QAEF,IAAI,IAAI,CAAC,EAAE,KAAK,EAAE;YAAE,OAAO;QAC3B,IAAI,CAAC,IAAI,CAAC,EAAE,YAAY,EAAE,EAAE,iBAAiB,EAAE,EAAE,CAAC,CAAC;IACrD,CAAC;IAEO,IAAI,CAAC,IAA6B;QACxC,MAAM,EAAE,GAAG,IAAI,CAAC,EAAE,CAAC;QACnB,IAAI,EAAE,EAAE,UAAU,KAAK,SAAS,CAAC,IAAI,EAAE,CAAC;YACtC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC;QAChC,CAAC;IACH,CAAC;CACF"}
|
|
@@ -2,17 +2,26 @@ import { EventEmitter } from 'node:events';
|
|
|
2
2
|
import type { RealtimeSessionConfig } from '../types.js';
|
|
3
3
|
interface OpenAIRealtimeEvents {
|
|
4
4
|
connected: [];
|
|
5
|
+
disconnected: [code: number, reason: string];
|
|
5
6
|
audio: [chunk: Buffer];
|
|
6
7
|
transcript: [text: string, role: 'user' | 'assistant'];
|
|
7
8
|
tool_call: [name: string, args: unknown];
|
|
8
9
|
speech_start: [];
|
|
10
|
+
turn_end: [];
|
|
9
11
|
error: [error: Error];
|
|
10
12
|
}
|
|
13
|
+
/**
|
|
14
|
+
* OpenAI Realtime API (GA interface) adapter. Audio in both directions is
|
|
15
|
+
* PCM16 mono at 24kHz.
|
|
16
|
+
*/
|
|
11
17
|
export declare class OpenAIRealtimeAdapter extends EventEmitter<OpenAIRealtimeEvents> {
|
|
12
18
|
private readonly config;
|
|
13
19
|
private readonly model;
|
|
14
20
|
private ws;
|
|
21
|
+
private connected;
|
|
22
|
+
private responseActive;
|
|
15
23
|
constructor(config: RealtimeSessionConfig);
|
|
24
|
+
get isConnected(): boolean;
|
|
16
25
|
connect(): Promise<void>;
|
|
17
26
|
pushAudio(chunk: Buffer): void;
|
|
18
27
|
sendText(text: string): void;
|
|
@@ -20,7 +29,9 @@ export declare class OpenAIRealtimeAdapter extends EventEmitter<OpenAIRealtimeEv
|
|
|
20
29
|
close(): void;
|
|
21
30
|
private sendSessionUpdate;
|
|
22
31
|
private handleMessage;
|
|
23
|
-
private
|
|
32
|
+
private handleResponseDone;
|
|
33
|
+
private handleToolCalls;
|
|
34
|
+
private executeToolCall;
|
|
24
35
|
private send;
|
|
25
36
|
}
|
|
26
37
|
export {};
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"openai-realtime.d.ts","sourceRoot":"","sources":["../../src/realtime/openai-realtime.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAE3C,OAAO,KAAK,EAAE,qBAAqB,EAAE,MAAM,aAAa,CAAC;
|
|
1
|
+
{"version":3,"file":"openai-realtime.d.ts","sourceRoot":"","sources":["../../src/realtime/openai-realtime.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAE3C,OAAO,KAAK,EAAE,qBAAqB,EAAE,MAAM,aAAa,CAAC;AASzD,UAAU,oBAAoB;IAC5B,SAAS,EAAE,EAAE,CAAC;IACd,YAAY,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC;IAC7C,KAAK,EAAE,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;IACvB,UAAU,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,GAAG,WAAW,CAAC,CAAC;IACvD,SAAS,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,OAAO,CAAC,CAAC;IACzC,YAAY,EAAE,EAAE,CAAC;IACjB,QAAQ,EAAE,EAAE,CAAC;IACb,KAAK,EAAE,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;CACvB;AAkBD;;;GAGG;AACH,qBAAa,qBAAsB,SAAQ,YAAY,CAAC,oBAAoB,CAAC;IAC3E,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAwB;IAC/C,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAS;IAC/B,OAAO,CAAC,EAAE,CAA0B;IACpC,OAAO,CAAC,SAAS,CAAS;IAC1B,OAAO,CAAC,cAAc,CAAS;gBAEnB,MAAM,EAAE,qBAAqB;IAMzC,IAAI,WAAW,IAAI,OAAO,CAEzB;IAEK,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC;IA6D9B,SAAS,CAAC,KAAK,EAAE,MAAM,GAAG,IAAI;IAO9B,QAAQ,CAAC,IAAI,EAAE,MAAM,GAAG,IAAI;IAY5B,SAAS,IAAI,IAAI;IAKjB,KAAK,IAAI,IAAI;IAWb,OAAO,CAAC,iBAAiB;IAkCzB,OAAO,CAAC,aAAa;IA8DrB,OAAO,CAAC,kBAAkB;YA2BZ,eAAe;YAcf,eAAe;IAkC7B,OAAO,CAAC,IAAI;CAMb"}
|