@lokutor/sdk 1.1.10 → 1.1.12
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/index.d.mts +43 -1
- package/dist/index.d.ts +43 -1
- package/dist/index.js +457 -365
- package/dist/index.mjs +457 -365
- package/package.json +1 -1
package/dist/index.js
CHANGED
|
@@ -78,287 +78,6 @@ var DEFAULT_URLS = {
|
|
|
78
78
|
TTS: "wss://api.lokutor.com/ws/tts"
|
|
79
79
|
};
|
|
80
80
|
|
|
81
|
-
// src/client.ts
|
|
82
|
-
function base64ToUint8Array(base64) {
|
|
83
|
-
const binaryString = atob(base64);
|
|
84
|
-
const bytes = new Uint8Array(binaryString.length);
|
|
85
|
-
for (let i = 0; i < binaryString.length; i++) {
|
|
86
|
-
bytes[i] = binaryString.charCodeAt(i);
|
|
87
|
-
}
|
|
88
|
-
return bytes;
|
|
89
|
-
}
|
|
90
|
-
var VoiceAgentClient = class {
|
|
91
|
-
ws = null;
|
|
92
|
-
apiKey;
|
|
93
|
-
prompt;
|
|
94
|
-
voice;
|
|
95
|
-
language;
|
|
96
|
-
// Callbacks
|
|
97
|
-
onTranscription;
|
|
98
|
-
onResponse;
|
|
99
|
-
onAudioCallback;
|
|
100
|
-
onVisemesCallback;
|
|
101
|
-
onStatus;
|
|
102
|
-
onError;
|
|
103
|
-
isConnected = false;
|
|
104
|
-
messages = [];
|
|
105
|
-
visemeListeners = [];
|
|
106
|
-
wantVisemes = false;
|
|
107
|
-
constructor(config) {
|
|
108
|
-
this.apiKey = config.apiKey;
|
|
109
|
-
this.prompt = config.prompt;
|
|
110
|
-
this.voice = config.voice || "F1" /* F1 */;
|
|
111
|
-
this.language = config.language || "en" /* ENGLISH */;
|
|
112
|
-
this.onTranscription = config.onTranscription;
|
|
113
|
-
this.onResponse = config.onResponse;
|
|
114
|
-
this.onAudioCallback = config.onAudio;
|
|
115
|
-
this.onVisemesCallback = config.onVisemes;
|
|
116
|
-
this.onStatus = config.onStatus;
|
|
117
|
-
this.onError = config.onError;
|
|
118
|
-
this.wantVisemes = config.visemes || false;
|
|
119
|
-
}
|
|
120
|
-
/**
|
|
121
|
-
* Connect to the Lokutor Voice Agent server
|
|
122
|
-
*/
|
|
123
|
-
async connect() {
|
|
124
|
-
return new Promise((resolve, reject) => {
|
|
125
|
-
try {
|
|
126
|
-
let url = DEFAULT_URLS.VOICE_AGENT;
|
|
127
|
-
if (this.apiKey) {
|
|
128
|
-
const separator = url.includes("?") ? "&" : "?";
|
|
129
|
-
url += `${separator}api_key=${this.apiKey}`;
|
|
130
|
-
}
|
|
131
|
-
console.log(`\u{1F517} Connecting to ${DEFAULT_URLS.VOICE_AGENT}...`);
|
|
132
|
-
this.ws = new WebSocket(url);
|
|
133
|
-
this.ws.binaryType = "arraybuffer";
|
|
134
|
-
this.ws.onopen = () => {
|
|
135
|
-
this.isConnected = true;
|
|
136
|
-
console.log("\u2705 Connected to voice agent!");
|
|
137
|
-
this.sendConfig();
|
|
138
|
-
resolve(true);
|
|
139
|
-
};
|
|
140
|
-
this.ws.onmessage = async (event) => {
|
|
141
|
-
if (event.data instanceof ArrayBuffer) {
|
|
142
|
-
this.handleBinaryMessage(new Uint8Array(event.data));
|
|
143
|
-
} else {
|
|
144
|
-
this.handleTextMessage(event.data.toString());
|
|
145
|
-
}
|
|
146
|
-
};
|
|
147
|
-
this.ws.onerror = (err) => {
|
|
148
|
-
console.error("\u274C WebSocket error:", err);
|
|
149
|
-
if (this.onError) this.onError(err);
|
|
150
|
-
if (!this.isConnected) reject(err);
|
|
151
|
-
};
|
|
152
|
-
this.ws.onclose = () => {
|
|
153
|
-
this.isConnected = false;
|
|
154
|
-
console.log("Disconnected");
|
|
155
|
-
};
|
|
156
|
-
} catch (err) {
|
|
157
|
-
if (this.onError) this.onError(err);
|
|
158
|
-
reject(err);
|
|
159
|
-
}
|
|
160
|
-
});
|
|
161
|
-
}
|
|
162
|
-
/**
|
|
163
|
-
* Send initial configuration to the server
|
|
164
|
-
*/
|
|
165
|
-
sendConfig() {
|
|
166
|
-
if (!this.ws || !this.isConnected) return;
|
|
167
|
-
this.ws.send(JSON.stringify({ type: "prompt", data: this.prompt }));
|
|
168
|
-
this.ws.send(JSON.stringify({ type: "voice", data: this.voice }));
|
|
169
|
-
this.ws.send(JSON.stringify({ type: "language", data: this.language }));
|
|
170
|
-
this.ws.send(JSON.stringify({ type: "visemes", data: this.wantVisemes }));
|
|
171
|
-
console.log(`\u2699\uFE0F Configured: voice=${this.voice}, language=${this.language}, visemes=${this.wantVisemes}`);
|
|
172
|
-
}
|
|
173
|
-
/**
|
|
174
|
-
* Send raw PCM audio data to the server
|
|
175
|
-
* @param audioData Int16 PCM audio buffer
|
|
176
|
-
*/
|
|
177
|
-
sendAudio(audioData) {
|
|
178
|
-
if (this.ws && this.isConnected) {
|
|
179
|
-
this.ws.send(audioData);
|
|
180
|
-
}
|
|
181
|
-
}
|
|
182
|
-
/**
|
|
183
|
-
* Handle incoming binary data (audio response)
|
|
184
|
-
*/
|
|
185
|
-
handleBinaryMessage(data) {
|
|
186
|
-
this.emit("audio", data);
|
|
187
|
-
}
|
|
188
|
-
/**
|
|
189
|
-
* Handle incoming text messages (metadata/transcriptions)
|
|
190
|
-
*/
|
|
191
|
-
handleTextMessage(text) {
|
|
192
|
-
try {
|
|
193
|
-
const msg = JSON.parse(text);
|
|
194
|
-
switch (msg.type) {
|
|
195
|
-
case "audio":
|
|
196
|
-
if (msg.data) {
|
|
197
|
-
const buffer = base64ToUint8Array(msg.data);
|
|
198
|
-
this.handleBinaryMessage(buffer);
|
|
199
|
-
}
|
|
200
|
-
break;
|
|
201
|
-
case "transcript":
|
|
202
|
-
const role = msg.role === "user" ? "user" : "agent";
|
|
203
|
-
this.messages.push({
|
|
204
|
-
role,
|
|
205
|
-
text: msg.data,
|
|
206
|
-
timestamp: Date.now()
|
|
207
|
-
});
|
|
208
|
-
if (msg.role === "user") {
|
|
209
|
-
if (this.onTranscription) this.onTranscription(msg.data);
|
|
210
|
-
console.log(`\u{1F4AC} You: ${msg.data}`);
|
|
211
|
-
} else {
|
|
212
|
-
if (this.onResponse) this.onResponse(msg.data);
|
|
213
|
-
console.log(`\u{1F916} Agent: ${msg.data}`);
|
|
214
|
-
}
|
|
215
|
-
break;
|
|
216
|
-
case "status":
|
|
217
|
-
if (this.onStatus) this.onStatus(msg.data);
|
|
218
|
-
const icons = {
|
|
219
|
-
"interrupted": "\u26A1",
|
|
220
|
-
"thinking": "\u{1F9E0}",
|
|
221
|
-
"speaking": "\u{1F50A}",
|
|
222
|
-
"listening": "\u{1F442}"
|
|
223
|
-
};
|
|
224
|
-
console.log(`${icons[msg.data] || ""} Status: ${msg.data}`);
|
|
225
|
-
break;
|
|
226
|
-
case "visemes":
|
|
227
|
-
if (Array.isArray(msg.data) && msg.data.length > 0) {
|
|
228
|
-
this.emit("visemes", msg.data);
|
|
229
|
-
}
|
|
230
|
-
break;
|
|
231
|
-
case "error":
|
|
232
|
-
if (this.onError) this.onError(msg.data);
|
|
233
|
-
console.error(`\u274C Server error: ${msg.data}`);
|
|
234
|
-
break;
|
|
235
|
-
}
|
|
236
|
-
} catch (e) {
|
|
237
|
-
}
|
|
238
|
-
}
|
|
239
|
-
audioListeners = [];
|
|
240
|
-
emit(event, data) {
|
|
241
|
-
if (event === "audio") {
|
|
242
|
-
if (this.onAudioCallback) this.onAudioCallback(data);
|
|
243
|
-
this.audioListeners.forEach((l) => l(data));
|
|
244
|
-
} else if (event === "visemes") {
|
|
245
|
-
if (this.onVisemesCallback) this.onVisemesCallback(data);
|
|
246
|
-
this.visemeListeners.forEach((l) => l(data));
|
|
247
|
-
}
|
|
248
|
-
}
|
|
249
|
-
onAudio(callback) {
|
|
250
|
-
this.audioListeners.push(callback);
|
|
251
|
-
}
|
|
252
|
-
onVisemes(callback) {
|
|
253
|
-
this.visemeListeners.push(callback);
|
|
254
|
-
}
|
|
255
|
-
/**
|
|
256
|
-
* Disconnect from the server
|
|
257
|
-
*/
|
|
258
|
-
disconnect() {
|
|
259
|
-
if (this.ws) {
|
|
260
|
-
this.ws.close();
|
|
261
|
-
this.ws = null;
|
|
262
|
-
}
|
|
263
|
-
}
|
|
264
|
-
/**
|
|
265
|
-
* Update the system prompt mid-conversation
|
|
266
|
-
*/
|
|
267
|
-
updatePrompt(newPrompt) {
|
|
268
|
-
this.prompt = newPrompt;
|
|
269
|
-
if (this.ws && this.isConnected) {
|
|
270
|
-
try {
|
|
271
|
-
this.ws.send(JSON.stringify({ type: "prompt", data: newPrompt }));
|
|
272
|
-
console.log(`\u2699\uFE0F Updated prompt: ${newPrompt.substring(0, 50)}...`);
|
|
273
|
-
} catch (error) {
|
|
274
|
-
console.error("Error updating prompt:", error);
|
|
275
|
-
}
|
|
276
|
-
} else {
|
|
277
|
-
console.warn("Not connected - prompt will be updated on next connection");
|
|
278
|
-
}
|
|
279
|
-
}
|
|
280
|
-
/**
|
|
281
|
-
* Get full conversation transcript
|
|
282
|
-
*/
|
|
283
|
-
getTranscript() {
|
|
284
|
-
return this.messages.slice();
|
|
285
|
-
}
|
|
286
|
-
/**
|
|
287
|
-
* Get conversation as formatted text
|
|
288
|
-
*/
|
|
289
|
-
getTranscriptText() {
|
|
290
|
-
return this.messages.map((msg) => `${msg.role === "user" ? "You" : "Agent"}: ${msg.text}`).join("\n");
|
|
291
|
-
}
|
|
292
|
-
};
|
|
293
|
-
var TTSClient = class {
|
|
294
|
-
apiKey;
|
|
295
|
-
constructor(config) {
|
|
296
|
-
this.apiKey = config.apiKey;
|
|
297
|
-
}
|
|
298
|
-
/**
|
|
299
|
-
* Synthesize text to speech
|
|
300
|
-
*
|
|
301
|
-
* This opens a temporary WebSocket connection, sends the request,
|
|
302
|
-
* and streams back the audio.
|
|
303
|
-
*/
|
|
304
|
-
synthesize(options) {
|
|
305
|
-
return new Promise((resolve, reject) => {
|
|
306
|
-
try {
|
|
307
|
-
let url = DEFAULT_URLS.TTS;
|
|
308
|
-
if (this.apiKey) {
|
|
309
|
-
const separator = url.includes("?") ? "&" : "?";
|
|
310
|
-
url += `${separator}api_key=${this.apiKey}`;
|
|
311
|
-
}
|
|
312
|
-
const ws = new WebSocket(url);
|
|
313
|
-
ws.binaryType = "arraybuffer";
|
|
314
|
-
ws.onopen = () => {
|
|
315
|
-
const req = {
|
|
316
|
-
text: options.text,
|
|
317
|
-
voice: options.voice || "F1" /* F1 */,
|
|
318
|
-
lang: options.language || "en" /* ENGLISH */,
|
|
319
|
-
speed: options.speed || 1.05,
|
|
320
|
-
steps: options.steps || 24,
|
|
321
|
-
visemes: options.visemes || false
|
|
322
|
-
};
|
|
323
|
-
ws.send(JSON.stringify(req));
|
|
324
|
-
};
|
|
325
|
-
ws.onmessage = async (event) => {
|
|
326
|
-
if (event.data instanceof ArrayBuffer) {
|
|
327
|
-
if (options.onAudio) options.onAudio(new Uint8Array(event.data));
|
|
328
|
-
} else {
|
|
329
|
-
try {
|
|
330
|
-
const msg = JSON.parse(event.data.toString());
|
|
331
|
-
if (Array.isArray(msg) && options.onVisemes) {
|
|
332
|
-
options.onVisemes(msg);
|
|
333
|
-
}
|
|
334
|
-
} catch (e) {
|
|
335
|
-
}
|
|
336
|
-
}
|
|
337
|
-
};
|
|
338
|
-
ws.onerror = (err) => {
|
|
339
|
-
if (options.onError) options.onError(err);
|
|
340
|
-
reject(err);
|
|
341
|
-
};
|
|
342
|
-
ws.onclose = () => {
|
|
343
|
-
resolve();
|
|
344
|
-
};
|
|
345
|
-
} catch (err) {
|
|
346
|
-
if (options.onError) options.onError(err);
|
|
347
|
-
reject(err);
|
|
348
|
-
}
|
|
349
|
-
});
|
|
350
|
-
}
|
|
351
|
-
};
|
|
352
|
-
async function simpleConversation(config) {
|
|
353
|
-
const client = new VoiceAgentClient(config);
|
|
354
|
-
await client.connect();
|
|
355
|
-
return client;
|
|
356
|
-
}
|
|
357
|
-
async function simpleTTS(options) {
|
|
358
|
-
const client = new TTSClient({ apiKey: options.apiKey });
|
|
359
|
-
return client.synthesize(options);
|
|
360
|
-
}
|
|
361
|
-
|
|
362
81
|
// src/audio-utils.ts
|
|
363
82
|
function pcm16ToFloat32(int16Data) {
|
|
364
83
|
const float32 = new Float32Array(int16Data.length);
|
|
@@ -498,6 +217,7 @@ var BrowserAudioManager = class {
|
|
|
498
217
|
scriptProcessor = null;
|
|
499
218
|
analyserNode = null;
|
|
500
219
|
mediaStream = null;
|
|
220
|
+
resampler = null;
|
|
501
221
|
// Playback scheduling
|
|
502
222
|
nextPlaybackTime = 0;
|
|
503
223
|
activeSources = [];
|
|
@@ -575,6 +295,12 @@ var BrowserAudioManager = class {
|
|
|
575
295
|
if (this.analyserNode) {
|
|
576
296
|
this.mediaStreamAudioSourceNode.connect(this.analyserNode);
|
|
577
297
|
}
|
|
298
|
+
const hardwareRate = this.audioContext.sampleRate;
|
|
299
|
+
if (hardwareRate !== this.inputSampleRate) {
|
|
300
|
+
this.resampler = new StreamResampler(hardwareRate, this.inputSampleRate);
|
|
301
|
+
} else {
|
|
302
|
+
this.resampler = null;
|
|
303
|
+
}
|
|
578
304
|
this.scriptProcessor.onaudioprocess = (event) => {
|
|
579
305
|
this._processAudioInput(event);
|
|
580
306
|
};
|
|
@@ -597,15 +323,11 @@ var BrowserAudioManager = class {
|
|
|
597
323
|
for (let i = 0; i < outputBuffer.getChannelData(0).length; i++) {
|
|
598
324
|
outputBuffer.getChannelData(0)[i] = 0;
|
|
599
325
|
}
|
|
600
|
-
const hardwareRate = this.audioContext.sampleRate;
|
|
601
326
|
let processedData = new Float32Array(inputData);
|
|
602
|
-
if (
|
|
603
|
-
processedData =
|
|
604
|
-
processedData,
|
|
605
|
-
hardwareRate,
|
|
606
|
-
this.inputSampleRate
|
|
607
|
-
);
|
|
327
|
+
if (this.resampler) {
|
|
328
|
+
processedData = this.resampler.process(processedData);
|
|
608
329
|
}
|
|
330
|
+
if (processedData.length === 0) return;
|
|
609
331
|
const int16Data = float32ToPcm16(processedData);
|
|
610
332
|
const uint8Data = new Uint8Array(
|
|
611
333
|
int16Data.buffer,
|
|
@@ -657,117 +379,487 @@ var BrowserAudioManager = class {
|
|
|
657
379
|
this._schedulePlayback(audioBuffer);
|
|
658
380
|
}
|
|
659
381
|
/**
|
|
660
|
-
* Internal method to schedule and play audio with sample-accurate timing
|
|
382
|
+
* Internal method to schedule and play audio with sample-accurate timing
|
|
383
|
+
*/
|
|
384
|
+
_schedulePlayback(audioBuffer) {
|
|
385
|
+
if (!this.audioContext) return;
|
|
386
|
+
const currentTime = this.audioContext.currentTime;
|
|
387
|
+
const duration = audioBuffer.length / this.outputSampleRate;
|
|
388
|
+
const startTime = Math.max(
|
|
389
|
+
currentTime + 0.01,
|
|
390
|
+
// Minimum 10ms delay
|
|
391
|
+
this.nextPlaybackTime
|
|
392
|
+
);
|
|
393
|
+
this.nextPlaybackTime = startTime + duration;
|
|
394
|
+
const source = this.audioContext.createBufferSource();
|
|
395
|
+
source.buffer = audioBuffer;
|
|
396
|
+
source.connect(this.audioContext.destination);
|
|
397
|
+
if (this.analyserNode) {
|
|
398
|
+
source.connect(this.analyserNode);
|
|
399
|
+
}
|
|
400
|
+
source.start(startTime);
|
|
401
|
+
this.activeSources.push(source);
|
|
402
|
+
source.onended = () => {
|
|
403
|
+
const index = this.activeSources.indexOf(source);
|
|
404
|
+
if (index > -1) {
|
|
405
|
+
this.activeSources.splice(index, 1);
|
|
406
|
+
}
|
|
407
|
+
};
|
|
408
|
+
}
|
|
409
|
+
/**
|
|
410
|
+
* Stop all currently playing audio and clear the queue
|
|
411
|
+
*/
|
|
412
|
+
stopPlayback() {
|
|
413
|
+
this.activeSources.forEach((source) => {
|
|
414
|
+
try {
|
|
415
|
+
source.stop();
|
|
416
|
+
} catch (e) {
|
|
417
|
+
}
|
|
418
|
+
});
|
|
419
|
+
this.activeSources = [];
|
|
420
|
+
this.playbackQueue = [];
|
|
421
|
+
this.nextPlaybackTime = this.audioContext?.currentTime ?? 0;
|
|
422
|
+
console.log("\u{1F507} Playback stopped");
|
|
423
|
+
}
|
|
424
|
+
/**
|
|
425
|
+
* Toggle mute state
|
|
426
|
+
*/
|
|
427
|
+
setMuted(muted) {
|
|
428
|
+
this.isMuted = muted;
|
|
429
|
+
}
|
|
430
|
+
/**
|
|
431
|
+
* Get current mute state
|
|
432
|
+
*/
|
|
433
|
+
isMicMuted() {
|
|
434
|
+
return this.isMuted;
|
|
435
|
+
}
|
|
436
|
+
/**
|
|
437
|
+
* Get current amplitude from analyser (for visualization)
|
|
438
|
+
* Returns value between 0 and 1
|
|
439
|
+
*/
|
|
440
|
+
getAmplitude() {
|
|
441
|
+
if (!this.analyserNode) return 0;
|
|
442
|
+
const dataArray = new Uint8Array(this.analyserNode.frequencyBinCount);
|
|
443
|
+
this.analyserNode.getByteTimeDomainData(dataArray);
|
|
444
|
+
const rms = calculateRMS(dataArray);
|
|
445
|
+
return Math.min(rms * 10, 1);
|
|
446
|
+
}
|
|
447
|
+
/**
|
|
448
|
+
* Get frequency data from analyser for visualization
|
|
449
|
+
*/
|
|
450
|
+
getFrequencyData() {
|
|
451
|
+
if (!this.analyserNode) {
|
|
452
|
+
return new Uint8Array(0);
|
|
453
|
+
}
|
|
454
|
+
const dataArray = new Uint8Array(this.analyserNode.frequencyBinCount);
|
|
455
|
+
this.analyserNode.getByteFrequencyData(dataArray);
|
|
456
|
+
return dataArray;
|
|
457
|
+
}
|
|
458
|
+
/**
|
|
459
|
+
* Get time-domain data from analyser for waveform visualization
|
|
460
|
+
*/
|
|
461
|
+
getWaveformData() {
|
|
462
|
+
if (!this.analyserNode) {
|
|
463
|
+
return new Uint8Array(0);
|
|
464
|
+
}
|
|
465
|
+
const dataArray = new Uint8Array(this.analyserNode.frequencyBinCount);
|
|
466
|
+
this.analyserNode.getByteTimeDomainData(dataArray);
|
|
467
|
+
return dataArray;
|
|
468
|
+
}
|
|
469
|
+
/**
|
|
470
|
+
* Cleanup and close AudioContext
|
|
471
|
+
*/
|
|
472
|
+
cleanup() {
|
|
473
|
+
this.stopMicrophone();
|
|
474
|
+
this.stopPlayback();
|
|
475
|
+
if (this.analyserNode) {
|
|
476
|
+
this.analyserNode.disconnect();
|
|
477
|
+
this.analyserNode = null;
|
|
478
|
+
}
|
|
479
|
+
}
|
|
480
|
+
/**
|
|
481
|
+
* Get current audio context state
|
|
482
|
+
*/
|
|
483
|
+
getState() {
|
|
484
|
+
return this.audioContext?.state ?? null;
|
|
485
|
+
}
|
|
486
|
+
/**
|
|
487
|
+
* Check if microphone is currently listening
|
|
488
|
+
*/
|
|
489
|
+
isRecording() {
|
|
490
|
+
return this.isListening;
|
|
491
|
+
}
|
|
492
|
+
};
|
|
493
|
+
|
|
494
|
+
// src/client.ts
|
|
495
|
+
function base64ToUint8Array(base64) {
|
|
496
|
+
const binaryString = atob(base64);
|
|
497
|
+
const bytes = new Uint8Array(binaryString.length);
|
|
498
|
+
for (let i = 0; i < binaryString.length; i++) {
|
|
499
|
+
bytes[i] = binaryString.charCodeAt(i);
|
|
500
|
+
}
|
|
501
|
+
return bytes;
|
|
502
|
+
}
|
|
503
|
+
var VoiceAgentClient = class {
|
|
504
|
+
ws = null;
|
|
505
|
+
apiKey;
|
|
506
|
+
prompt;
|
|
507
|
+
voice;
|
|
508
|
+
language;
|
|
509
|
+
tools = [];
|
|
510
|
+
// Callbacks
|
|
511
|
+
onTranscription;
|
|
512
|
+
onResponse;
|
|
513
|
+
onAudioCallback;
|
|
514
|
+
onVisemesCallback;
|
|
515
|
+
onStatus;
|
|
516
|
+
onError;
|
|
517
|
+
isConnected = false;
|
|
518
|
+
messages = [];
|
|
519
|
+
visemeListeners = [];
|
|
520
|
+
wantVisemes = false;
|
|
521
|
+
audioManager = null;
|
|
522
|
+
enableAudio = false;
|
|
523
|
+
currentGeneration = 0;
|
|
524
|
+
// Connection resilience
|
|
525
|
+
isUserDisconnect = false;
|
|
526
|
+
reconnecting = false;
|
|
527
|
+
reconnectAttempts = 0;
|
|
528
|
+
maxReconnectAttempts = 5;
|
|
529
|
+
constructor(config) {
|
|
530
|
+
this.apiKey = config.apiKey;
|
|
531
|
+
this.prompt = config.prompt;
|
|
532
|
+
this.voice = config.voice || "F1" /* F1 */;
|
|
533
|
+
this.language = config.language || "en" /* ENGLISH */;
|
|
534
|
+
this.onTranscription = config.onTranscription;
|
|
535
|
+
this.onResponse = config.onResponse;
|
|
536
|
+
this.onAudioCallback = config.onAudio;
|
|
537
|
+
this.onVisemesCallback = config.onVisemes;
|
|
538
|
+
this.onStatus = config.onStatus;
|
|
539
|
+
this.onError = config.onError;
|
|
540
|
+
this.wantVisemes = config.visemes || false;
|
|
541
|
+
this.enableAudio = config.enableAudio ?? false;
|
|
542
|
+
this.tools = config.tools || [];
|
|
543
|
+
}
|
|
544
|
+
/**
|
|
545
|
+
* Connect to the Lokutor Voice Agent server
|
|
546
|
+
*/
|
|
547
|
+
async connect() {
|
|
548
|
+
this.isUserDisconnect = false;
|
|
549
|
+
if (this.enableAudio) {
|
|
550
|
+
if (!this.audioManager) {
|
|
551
|
+
this.audioManager = new BrowserAudioManager();
|
|
552
|
+
}
|
|
553
|
+
await this.audioManager.init();
|
|
554
|
+
}
|
|
555
|
+
return new Promise((resolve, reject) => {
|
|
556
|
+
try {
|
|
557
|
+
let url = DEFAULT_URLS.VOICE_AGENT;
|
|
558
|
+
if (this.apiKey) {
|
|
559
|
+
const separator = url.includes("?") ? "&" : "?";
|
|
560
|
+
url += `${separator}api_key=${this.apiKey}`;
|
|
561
|
+
}
|
|
562
|
+
console.log(`\u{1F517} Connecting to ${DEFAULT_URLS.VOICE_AGENT}...`);
|
|
563
|
+
this.ws = new WebSocket(url);
|
|
564
|
+
this.ws.binaryType = "arraybuffer";
|
|
565
|
+
this.ws.onopen = async () => {
|
|
566
|
+
this.isConnected = true;
|
|
567
|
+
this.reconnectAttempts = 0;
|
|
568
|
+
this.reconnecting = false;
|
|
569
|
+
console.log("\u2705 Connected to voice agent!");
|
|
570
|
+
this.sendConfig();
|
|
571
|
+
if (this.audioManager) {
|
|
572
|
+
await this.audioManager.startMicrophone((data) => {
|
|
573
|
+
if (this.isConnected) {
|
|
574
|
+
this.sendAudio(data);
|
|
575
|
+
}
|
|
576
|
+
});
|
|
577
|
+
}
|
|
578
|
+
resolve(true);
|
|
579
|
+
};
|
|
580
|
+
this.ws.onmessage = async (event) => {
|
|
581
|
+
if (event.data instanceof ArrayBuffer) {
|
|
582
|
+
this.handleBinaryMessage(new Uint8Array(event.data));
|
|
583
|
+
} else {
|
|
584
|
+
this.handleTextMessage(event.data.toString());
|
|
585
|
+
}
|
|
586
|
+
};
|
|
587
|
+
this.ws.onerror = (err) => {
|
|
588
|
+
console.error("\u274C WebSocket error:", err);
|
|
589
|
+
if (this.onError) this.onError(err);
|
|
590
|
+
if (!this.isConnected) reject(err);
|
|
591
|
+
};
|
|
592
|
+
this.ws.onclose = () => {
|
|
593
|
+
this.isConnected = false;
|
|
594
|
+
if (!this.isUserDisconnect && this.reconnectAttempts < this.maxReconnectAttempts) {
|
|
595
|
+
this.reconnecting = true;
|
|
596
|
+
this.reconnectAttempts++;
|
|
597
|
+
const backoffDelay = Math.min(1e3 * Math.pow(2, this.reconnectAttempts), 1e4);
|
|
598
|
+
console.warn(`Connection lost. Reconnecting in ${backoffDelay}ms (attempt ${this.reconnectAttempts}/${this.maxReconnectAttempts})`);
|
|
599
|
+
if (this.onStatus) this.onStatus("reconnecting");
|
|
600
|
+
setTimeout(() => {
|
|
601
|
+
this.connect().catch((e) => console.error("Reconnect failed", e));
|
|
602
|
+
}, backoffDelay);
|
|
603
|
+
} else {
|
|
604
|
+
console.log("Disconnected");
|
|
605
|
+
if (this.onStatus) this.onStatus("disconnected");
|
|
606
|
+
}
|
|
607
|
+
};
|
|
608
|
+
} catch (err) {
|
|
609
|
+
if (this.onError) this.onError(err);
|
|
610
|
+
reject(err);
|
|
611
|
+
}
|
|
612
|
+
});
|
|
613
|
+
}
|
|
614
|
+
/**
|
|
615
|
+
* Send initial configuration to the server
|
|
661
616
|
*/
|
|
662
|
-
|
|
663
|
-
if (!this.
|
|
664
|
-
|
|
665
|
-
|
|
666
|
-
|
|
667
|
-
|
|
668
|
-
|
|
669
|
-
this.
|
|
670
|
-
);
|
|
671
|
-
this.nextPlaybackTime = startTime + duration;
|
|
672
|
-
const source = this.audioContext.createBufferSource();
|
|
673
|
-
source.buffer = audioBuffer;
|
|
674
|
-
source.connect(this.audioContext.destination);
|
|
675
|
-
if (this.analyserNode) {
|
|
676
|
-
source.connect(this.analyserNode);
|
|
617
|
+
sendConfig() {
|
|
618
|
+
if (!this.ws || !this.isConnected) return;
|
|
619
|
+
this.ws.send(JSON.stringify({ type: "prompt", data: this.prompt }));
|
|
620
|
+
this.ws.send(JSON.stringify({ type: "voice", data: this.voice }));
|
|
621
|
+
this.ws.send(JSON.stringify({ type: "language", data: this.language }));
|
|
622
|
+
this.ws.send(JSON.stringify({ type: "visemes", data: this.wantVisemes }));
|
|
623
|
+
if (this.tools && this.tools.length > 0) {
|
|
624
|
+
this.ws.send(JSON.stringify({ type: "tools", data: this.tools }));
|
|
677
625
|
}
|
|
678
|
-
|
|
679
|
-
this.activeSources.push(source);
|
|
680
|
-
source.onended = () => {
|
|
681
|
-
const index = this.activeSources.indexOf(source);
|
|
682
|
-
if (index > -1) {
|
|
683
|
-
this.activeSources.splice(index, 1);
|
|
684
|
-
}
|
|
685
|
-
};
|
|
626
|
+
console.log(`\u2699\uFE0F Configured: voice=${this.voice}, language=${this.language}, visemes=${this.wantVisemes}, tools=${this.tools.length}`);
|
|
686
627
|
}
|
|
687
628
|
/**
|
|
688
|
-
*
|
|
629
|
+
* Send raw PCM audio data to the server
|
|
630
|
+
* @param audioData Int16 PCM audio buffer
|
|
689
631
|
*/
|
|
690
|
-
|
|
691
|
-
this.
|
|
692
|
-
|
|
693
|
-
|
|
694
|
-
} catch (e) {
|
|
695
|
-
}
|
|
696
|
-
});
|
|
697
|
-
this.activeSources = [];
|
|
698
|
-
this.playbackQueue = [];
|
|
699
|
-
this.nextPlaybackTime = this.audioContext?.currentTime ?? 0;
|
|
700
|
-
console.log("\u{1F507} Playback stopped");
|
|
632
|
+
sendAudio(audioData) {
|
|
633
|
+
if (this.ws && this.ws.readyState === WebSocket.OPEN && this.isConnected) {
|
|
634
|
+
this.ws.send(audioData);
|
|
635
|
+
}
|
|
701
636
|
}
|
|
702
637
|
/**
|
|
703
|
-
*
|
|
638
|
+
* Handle incoming binary data (audio response)
|
|
704
639
|
*/
|
|
705
|
-
|
|
706
|
-
this.
|
|
640
|
+
handleBinaryMessage(data, generation) {
|
|
641
|
+
if (generation !== void 0 && generation < this.currentGeneration) {
|
|
642
|
+
console.log(`\u{1F5D1}\uFE0F Discarding ghost audio (Gen ${generation} < ${this.currentGeneration})`);
|
|
643
|
+
return;
|
|
644
|
+
}
|
|
645
|
+
if (this.audioManager) {
|
|
646
|
+
this.audioManager.playAudio(data);
|
|
647
|
+
}
|
|
648
|
+
this.emit("audio", data);
|
|
707
649
|
}
|
|
708
650
|
/**
|
|
709
|
-
*
|
|
651
|
+
* Handle incoming text messages (metadata/transcriptions)
|
|
710
652
|
*/
|
|
711
|
-
|
|
712
|
-
|
|
653
|
+
handleTextMessage(text) {
|
|
654
|
+
try {
|
|
655
|
+
const msg = JSON.parse(text);
|
|
656
|
+
switch (msg.type) {
|
|
657
|
+
case "audio":
|
|
658
|
+
if (msg.data) {
|
|
659
|
+
const buffer = base64ToUint8Array(msg.data);
|
|
660
|
+
this.handleBinaryMessage(buffer, msg.generation);
|
|
661
|
+
}
|
|
662
|
+
break;
|
|
663
|
+
case "transcript":
|
|
664
|
+
const role = msg.role === "user" ? "user" : "agent";
|
|
665
|
+
this.messages.push({
|
|
666
|
+
role,
|
|
667
|
+
text: msg.data,
|
|
668
|
+
timestamp: Date.now()
|
|
669
|
+
});
|
|
670
|
+
if (msg.role === "user") {
|
|
671
|
+
if (this.onTranscription) this.onTranscription(msg.data);
|
|
672
|
+
console.log(`\u{1F4AC} You: ${msg.data}`);
|
|
673
|
+
} else {
|
|
674
|
+
if (this.onResponse) this.onResponse(msg.data);
|
|
675
|
+
console.log(`\u{1F916} Agent: ${msg.data}`);
|
|
676
|
+
}
|
|
677
|
+
break;
|
|
678
|
+
case "status":
|
|
679
|
+
if (msg.data === "thinking") {
|
|
680
|
+
const newGen = msg.generation || 0;
|
|
681
|
+
if (newGen > this.currentGeneration) {
|
|
682
|
+
console.log(`\u{1F9E0} New thought (Gen ${newGen}) - Clearing audio queue`);
|
|
683
|
+
this.currentGeneration = newGen;
|
|
684
|
+
if (this.audioManager) this.audioManager.stopPlayback();
|
|
685
|
+
}
|
|
686
|
+
}
|
|
687
|
+
if (msg.data === "interrupted" && this.audioManager) {
|
|
688
|
+
this.audioManager.stopPlayback();
|
|
689
|
+
}
|
|
690
|
+
if (this.onStatus) this.onStatus(msg.data);
|
|
691
|
+
const icons = {
|
|
692
|
+
"interrupted": "\u26A1",
|
|
693
|
+
"thinking": "\u{1F9E0}",
|
|
694
|
+
"speaking": "\u{1F50A}",
|
|
695
|
+
"listening": "\u{1F442}"
|
|
696
|
+
};
|
|
697
|
+
console.log(`${icons[msg.data] || ""} Status: ${msg.data}`);
|
|
698
|
+
break;
|
|
699
|
+
case "visemes":
|
|
700
|
+
if (Array.isArray(msg.data) && msg.data.length > 0) {
|
|
701
|
+
this.emit("visemes", msg.data);
|
|
702
|
+
}
|
|
703
|
+
break;
|
|
704
|
+
case "error":
|
|
705
|
+
if (this.onError) this.onError(msg.data);
|
|
706
|
+
console.error(`\u274C Server error: ${msg.data}`);
|
|
707
|
+
break;
|
|
708
|
+
case "tool_call":
|
|
709
|
+
console.log(`\u{1F6E0}\uFE0F Tool Call: ${msg.name}(${msg.arguments})`);
|
|
710
|
+
break;
|
|
711
|
+
}
|
|
712
|
+
} catch (e) {
|
|
713
|
+
}
|
|
714
|
+
}
|
|
715
|
+
audioListeners = [];
|
|
716
|
+
emit(event, data) {
|
|
717
|
+
if (event === "audio") {
|
|
718
|
+
if (this.onAudioCallback) this.onAudioCallback(data);
|
|
719
|
+
this.audioListeners.forEach((l) => l(data));
|
|
720
|
+
} else if (event === "visemes") {
|
|
721
|
+
if (this.onVisemesCallback) this.onVisemesCallback(data);
|
|
722
|
+
this.visemeListeners.forEach((l) => l(data));
|
|
723
|
+
}
|
|
724
|
+
}
|
|
725
|
+
onAudio(callback) {
|
|
726
|
+
this.audioListeners.push(callback);
|
|
727
|
+
}
|
|
728
|
+
onVisemes(callback) {
|
|
729
|
+
this.visemeListeners.push(callback);
|
|
713
730
|
}
|
|
714
731
|
/**
|
|
715
|
-
*
|
|
716
|
-
* Returns value between 0 and 1
|
|
732
|
+
* Disconnect from the server
|
|
717
733
|
*/
|
|
718
|
-
|
|
719
|
-
|
|
720
|
-
|
|
721
|
-
|
|
722
|
-
|
|
723
|
-
|
|
734
|
+
disconnect() {
|
|
735
|
+
this.isUserDisconnect = true;
|
|
736
|
+
if (this.ws) {
|
|
737
|
+
this.ws.close();
|
|
738
|
+
this.ws = null;
|
|
739
|
+
}
|
|
740
|
+
if (this.audioManager) {
|
|
741
|
+
this.audioManager.cleanup();
|
|
742
|
+
}
|
|
743
|
+
this.isConnected = false;
|
|
724
744
|
}
|
|
725
745
|
/**
|
|
726
|
-
*
|
|
746
|
+
* Toggles the microphone mute state (if managed by client)
|
|
747
|
+
* returns the new mute state
|
|
727
748
|
*/
|
|
728
|
-
|
|
729
|
-
if (
|
|
730
|
-
|
|
749
|
+
toggleMute() {
|
|
750
|
+
if (this.audioManager) {
|
|
751
|
+
const isMuted = this.audioManager.isMicMuted();
|
|
752
|
+
this.audioManager.setMuted(!isMuted);
|
|
753
|
+
return !isMuted;
|
|
731
754
|
}
|
|
732
|
-
|
|
733
|
-
this.analyserNode.getByteFrequencyData(dataArray);
|
|
734
|
-
return dataArray;
|
|
755
|
+
return false;
|
|
735
756
|
}
|
|
736
757
|
/**
|
|
737
|
-
*
|
|
758
|
+
* Gets the microphone volume amplitude 0-1 (if managed by client)
|
|
738
759
|
*/
|
|
739
|
-
|
|
740
|
-
if (
|
|
741
|
-
return
|
|
760
|
+
getAmplitude() {
|
|
761
|
+
if (this.audioManager) {
|
|
762
|
+
return this.audioManager.getAmplitude();
|
|
742
763
|
}
|
|
743
|
-
|
|
744
|
-
this.analyserNode.getByteTimeDomainData(dataArray);
|
|
745
|
-
return dataArray;
|
|
764
|
+
return 0;
|
|
746
765
|
}
|
|
747
766
|
/**
|
|
748
|
-
*
|
|
767
|
+
* Update the system prompt mid-conversation
|
|
749
768
|
*/
|
|
750
|
-
|
|
751
|
-
this.
|
|
752
|
-
this.
|
|
753
|
-
|
|
754
|
-
|
|
755
|
-
|
|
769
|
+
updatePrompt(newPrompt) {
|
|
770
|
+
this.prompt = newPrompt;
|
|
771
|
+
if (this.ws && this.isConnected) {
|
|
772
|
+
try {
|
|
773
|
+
this.ws.send(JSON.stringify({ type: "prompt", data: newPrompt }));
|
|
774
|
+
console.log(`\u2699\uFE0F Updated prompt: ${newPrompt.substring(0, 50)}...`);
|
|
775
|
+
} catch (error) {
|
|
776
|
+
console.error("Error updating prompt:", error);
|
|
777
|
+
}
|
|
778
|
+
} else {
|
|
779
|
+
console.warn("Not connected - prompt will be updated on next connection");
|
|
756
780
|
}
|
|
757
781
|
}
|
|
758
782
|
/**
|
|
759
|
-
* Get
|
|
783
|
+
* Get full conversation transcript
|
|
760
784
|
*/
|
|
761
|
-
|
|
762
|
-
return this.
|
|
785
|
+
getTranscript() {
|
|
786
|
+
return this.messages.slice();
|
|
763
787
|
}
|
|
764
788
|
/**
|
|
765
|
-
*
|
|
789
|
+
* Get conversation as formatted text
|
|
766
790
|
*/
|
|
767
|
-
|
|
768
|
-
return this.
|
|
791
|
+
getTranscriptText() {
|
|
792
|
+
return this.messages.map((msg) => `${msg.role === "user" ? "You" : "Agent"}: ${msg.text}`).join("\n");
|
|
793
|
+
}
|
|
794
|
+
};
|
|
795
|
+
var TTSClient = class {
|
|
796
|
+
apiKey;
|
|
797
|
+
constructor(config) {
|
|
798
|
+
this.apiKey = config.apiKey;
|
|
799
|
+
}
|
|
800
|
+
/**
|
|
801
|
+
* Synthesize text to speech
|
|
802
|
+
*
|
|
803
|
+
* This opens a temporary WebSocket connection, sends the request,
|
|
804
|
+
* and streams back the audio.
|
|
805
|
+
*/
|
|
806
|
+
synthesize(options) {
|
|
807
|
+
return new Promise((resolve, reject) => {
|
|
808
|
+
try {
|
|
809
|
+
let url = DEFAULT_URLS.TTS;
|
|
810
|
+
if (this.apiKey) {
|
|
811
|
+
const separator = url.includes("?") ? "&" : "?";
|
|
812
|
+
url += `${separator}api_key=${this.apiKey}`;
|
|
813
|
+
}
|
|
814
|
+
const ws = new WebSocket(url);
|
|
815
|
+
ws.binaryType = "arraybuffer";
|
|
816
|
+
ws.onopen = () => {
|
|
817
|
+
const req = {
|
|
818
|
+
text: options.text,
|
|
819
|
+
voice: options.voice || "F1" /* F1 */,
|
|
820
|
+
lang: options.language || "en" /* ENGLISH */,
|
|
821
|
+
speed: options.speed || 1.05,
|
|
822
|
+
steps: options.steps || 24,
|
|
823
|
+
visemes: options.visemes || false
|
|
824
|
+
};
|
|
825
|
+
ws.send(JSON.stringify(req));
|
|
826
|
+
};
|
|
827
|
+
ws.onmessage = async (event) => {
|
|
828
|
+
if (event.data instanceof ArrayBuffer) {
|
|
829
|
+
if (options.onAudio) options.onAudio(new Uint8Array(event.data));
|
|
830
|
+
} else {
|
|
831
|
+
try {
|
|
832
|
+
const msg = JSON.parse(event.data.toString());
|
|
833
|
+
if (Array.isArray(msg) && options.onVisemes) {
|
|
834
|
+
options.onVisemes(msg);
|
|
835
|
+
}
|
|
836
|
+
} catch (e) {
|
|
837
|
+
}
|
|
838
|
+
}
|
|
839
|
+
};
|
|
840
|
+
ws.onerror = (err) => {
|
|
841
|
+
if (options.onError) options.onError(err);
|
|
842
|
+
reject(err);
|
|
843
|
+
};
|
|
844
|
+
ws.onclose = () => {
|
|
845
|
+
resolve();
|
|
846
|
+
};
|
|
847
|
+
} catch (err) {
|
|
848
|
+
if (options.onError) options.onError(err);
|
|
849
|
+
reject(err);
|
|
850
|
+
}
|
|
851
|
+
});
|
|
769
852
|
}
|
|
770
853
|
};
|
|
854
|
+
async function simpleConversation(config) {
|
|
855
|
+
const client = new VoiceAgentClient(config);
|
|
856
|
+
await client.connect();
|
|
857
|
+
return client;
|
|
858
|
+
}
|
|
859
|
+
async function simpleTTS(options) {
|
|
860
|
+
const client = new TTSClient({ apiKey: options.apiKey });
|
|
861
|
+
return client.synthesize(options);
|
|
862
|
+
}
|
|
771
863
|
// Annotate the CommonJS export names for ESM import in node:
|
|
772
864
|
0 && (module.exports = {
|
|
773
865
|
AUDIO_CONFIG,
|