@livekit/agents 1.0.3 → 1.0.4
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/index.cjs +2 -5
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.cts +2 -3
- package/dist/index.d.ts +2 -3
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +1 -3
- package/dist/index.js.map +1 -1
- package/dist/tokenize/basic/hyphenator.cjs.map +1 -1
- package/dist/tokenize/basic/hyphenator.js.map +1 -1
- package/dist/utils.cjs +77 -0
- package/dist/utils.cjs.map +1 -1
- package/dist/utils.d.cts +21 -0
- package/dist/utils.d.ts +21 -0
- package/dist/utils.d.ts.map +1 -1
- package/dist/utils.js +76 -1
- package/dist/utils.js.map +1 -1
- package/dist/voice/agent_activity.cjs +112 -71
- package/dist/voice/agent_activity.cjs.map +1 -1
- package/dist/voice/agent_activity.d.ts.map +1 -1
- package/dist/voice/agent_activity.js +112 -71
- package/dist/voice/agent_activity.js.map +1 -1
- package/dist/voice/avatar/datastream_io.cjs +204 -0
- package/dist/voice/avatar/datastream_io.cjs.map +1 -0
- package/dist/voice/avatar/datastream_io.d.cts +37 -0
- package/dist/voice/avatar/datastream_io.d.ts +37 -0
- package/dist/voice/avatar/datastream_io.d.ts.map +1 -0
- package/dist/voice/avatar/datastream_io.js +188 -0
- package/dist/voice/avatar/datastream_io.js.map +1 -0
- package/dist/{multimodal → voice/avatar}/index.cjs +4 -4
- package/dist/voice/avatar/index.cjs.map +1 -0
- package/dist/voice/avatar/index.d.cts +2 -0
- package/dist/voice/avatar/index.d.ts +2 -0
- package/dist/voice/avatar/index.d.ts.map +1 -0
- package/dist/voice/avatar/index.js +2 -0
- package/dist/voice/avatar/index.js.map +1 -0
- package/dist/voice/index.cjs +2 -0
- package/dist/voice/index.cjs.map +1 -1
- package/dist/voice/index.d.cts +1 -0
- package/dist/voice/index.d.ts +1 -0
- package/dist/voice/index.d.ts.map +1 -1
- package/dist/voice/index.js +1 -0
- package/dist/voice/index.js.map +1 -1
- package/dist/voice/io.cjs.map +1 -1
- package/dist/voice/io.d.cts +1 -1
- package/dist/voice/io.d.ts +1 -1
- package/dist/voice/io.d.ts.map +1 -1
- package/dist/voice/io.js.map +1 -1
- package/dist/voice/room_io/_input.cjs +2 -1
- package/dist/voice/room_io/_input.cjs.map +1 -1
- package/dist/voice/room_io/_input.d.ts.map +1 -1
- package/dist/voice/room_io/_input.js +2 -1
- package/dist/voice/room_io/_input.js.map +1 -1
- package/dist/voice/run_context.cjs +13 -0
- package/dist/voice/run_context.cjs.map +1 -1
- package/dist/voice/run_context.d.cts +10 -0
- package/dist/voice/run_context.d.ts +10 -0
- package/dist/voice/run_context.d.ts.map +1 -1
- package/dist/voice/run_context.js +13 -0
- package/dist/voice/run_context.js.map +1 -1
- package/dist/voice/speech_handle.cjs +152 -30
- package/dist/voice/speech_handle.cjs.map +1 -1
- package/dist/voice/speech_handle.d.cts +67 -16
- package/dist/voice/speech_handle.d.ts +67 -16
- package/dist/voice/speech_handle.d.ts.map +1 -1
- package/dist/voice/speech_handle.js +153 -31
- package/dist/voice/speech_handle.js.map +1 -1
- package/dist/worker.cjs +4 -1
- package/dist/worker.cjs.map +1 -1
- package/dist/worker.d.ts.map +1 -1
- package/dist/worker.js +4 -1
- package/dist/worker.js.map +1 -1
- package/package.json +2 -2
- package/src/index.ts +2 -3
- package/src/tokenize/basic/hyphenator.ts +1 -1
- package/src/utils.ts +121 -1
- package/src/voice/agent_activity.ts +128 -78
- package/src/voice/avatar/datastream_io.ts +247 -0
- package/src/voice/avatar/index.ts +4 -0
- package/src/voice/index.ts +2 -0
- package/src/voice/io.ts +1 -1
- package/src/voice/room_io/_input.ts +8 -3
- package/src/voice/run_context.ts +16 -2
- package/src/voice/speech_handle.ts +183 -38
- package/src/worker.ts +5 -1
- package/dist/multimodal/agent_playout.cjs +0 -233
- package/dist/multimodal/agent_playout.cjs.map +0 -1
- package/dist/multimodal/agent_playout.d.cts +0 -34
- package/dist/multimodal/agent_playout.d.ts +0 -34
- package/dist/multimodal/agent_playout.d.ts.map +0 -1
- package/dist/multimodal/agent_playout.js +0 -207
- package/dist/multimodal/agent_playout.js.map +0 -1
- package/dist/multimodal/index.cjs.map +0 -1
- package/dist/multimodal/index.d.cts +0 -2
- package/dist/multimodal/index.d.ts +0 -2
- package/dist/multimodal/index.d.ts.map +0 -1
- package/dist/multimodal/index.js +0 -2
- package/dist/multimodal/index.js.map +0 -1
- package/src/multimodal/agent_playout.ts +0 -266
- package/src/multimodal/index.ts +0 -4
|
@@ -1,266 +0,0 @@
|
|
|
1
|
-
// SPDX-FileCopyrightText: 2024 LiveKit, Inc.
|
|
2
|
-
//
|
|
3
|
-
// SPDX-License-Identifier: Apache-2.0
|
|
4
|
-
import type { AudioFrame } from '@livekit/rtc-node';
|
|
5
|
-
import { type AudioSource } from '@livekit/rtc-node';
|
|
6
|
-
import { EventEmitter } from 'node:events';
|
|
7
|
-
import { AudioByteStream } from '../audio.js';
|
|
8
|
-
import type { TextAudioSynchronizer } from '../transcription.js';
|
|
9
|
-
import { type AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';
|
|
10
|
-
|
|
11
|
-
export const proto = {};
|
|
12
|
-
|
|
13
|
-
export class PlayoutHandle extends EventEmitter {
|
|
14
|
-
#audioSource: AudioSource;
|
|
15
|
-
#sampleRate: number;
|
|
16
|
-
#itemId: string;
|
|
17
|
-
#contentIndex: number;
|
|
18
|
-
/** @internal */
|
|
19
|
-
synchronizer: TextAudioSynchronizer;
|
|
20
|
-
/** @internal */
|
|
21
|
-
doneFut: Future;
|
|
22
|
-
/** @internal */
|
|
23
|
-
intFut: Future;
|
|
24
|
-
/** @internal */
|
|
25
|
-
#interrupted: boolean;
|
|
26
|
-
/** @internal */
|
|
27
|
-
pushedDuration: number;
|
|
28
|
-
/** @internal */
|
|
29
|
-
totalPlayedTime: number | undefined; // Set when playout is done
|
|
30
|
-
|
|
31
|
-
constructor(
|
|
32
|
-
audioSource: AudioSource,
|
|
33
|
-
sampleRate: number,
|
|
34
|
-
itemId: string,
|
|
35
|
-
contentIndex: number,
|
|
36
|
-
synchronizer: TextAudioSynchronizer,
|
|
37
|
-
) {
|
|
38
|
-
super();
|
|
39
|
-
this.#audioSource = audioSource;
|
|
40
|
-
this.#sampleRate = sampleRate;
|
|
41
|
-
this.#itemId = itemId;
|
|
42
|
-
this.#contentIndex = contentIndex;
|
|
43
|
-
this.synchronizer = synchronizer;
|
|
44
|
-
this.doneFut = new Future();
|
|
45
|
-
this.intFut = new Future();
|
|
46
|
-
this.#interrupted = false;
|
|
47
|
-
this.pushedDuration = 0;
|
|
48
|
-
this.totalPlayedTime = undefined;
|
|
49
|
-
}
|
|
50
|
-
|
|
51
|
-
get itemId(): string {
|
|
52
|
-
return this.#itemId;
|
|
53
|
-
}
|
|
54
|
-
|
|
55
|
-
get audioSamples(): number {
|
|
56
|
-
if (this.totalPlayedTime !== undefined) {
|
|
57
|
-
return Math.floor(this.totalPlayedTime * this.#sampleRate);
|
|
58
|
-
}
|
|
59
|
-
|
|
60
|
-
return Math.max(
|
|
61
|
-
0,
|
|
62
|
-
Math.floor(
|
|
63
|
-
(this.pushedDuration - this.#audioSource.queuedDuration) * (this.#sampleRate / 1000),
|
|
64
|
-
),
|
|
65
|
-
);
|
|
66
|
-
}
|
|
67
|
-
|
|
68
|
-
get textChars(): number {
|
|
69
|
-
return this.synchronizer.playedText.length;
|
|
70
|
-
}
|
|
71
|
-
|
|
72
|
-
get contentIndex(): number {
|
|
73
|
-
return this.#contentIndex;
|
|
74
|
-
}
|
|
75
|
-
|
|
76
|
-
get interrupted(): boolean {
|
|
77
|
-
return this.#interrupted;
|
|
78
|
-
}
|
|
79
|
-
|
|
80
|
-
get done(): boolean {
|
|
81
|
-
return this.doneFut.done || this.#interrupted;
|
|
82
|
-
}
|
|
83
|
-
|
|
84
|
-
interrupt() {
|
|
85
|
-
if (this.doneFut.done) return;
|
|
86
|
-
this.intFut.resolve();
|
|
87
|
-
this.#interrupted = true;
|
|
88
|
-
}
|
|
89
|
-
}
|
|
90
|
-
|
|
91
|
-
export class AgentPlayout extends EventEmitter {
|
|
92
|
-
#audioSource: AudioSource;
|
|
93
|
-
#playoutTask: CancellablePromise<void> | null;
|
|
94
|
-
#sampleRate: number;
|
|
95
|
-
#numChannels: number;
|
|
96
|
-
#inFrameSize: number;
|
|
97
|
-
#outFrameSize: number;
|
|
98
|
-
constructor(
|
|
99
|
-
audioSource: AudioSource,
|
|
100
|
-
sampleRate: number,
|
|
101
|
-
numChannels: number,
|
|
102
|
-
inFrameSize: number,
|
|
103
|
-
outFrameSize: number,
|
|
104
|
-
) {
|
|
105
|
-
super();
|
|
106
|
-
this.#audioSource = audioSource;
|
|
107
|
-
this.#playoutTask = null;
|
|
108
|
-
this.#sampleRate = sampleRate;
|
|
109
|
-
this.#numChannels = numChannels;
|
|
110
|
-
this.#inFrameSize = inFrameSize;
|
|
111
|
-
this.#outFrameSize = outFrameSize;
|
|
112
|
-
}
|
|
113
|
-
|
|
114
|
-
play(
|
|
115
|
-
itemId: string,
|
|
116
|
-
contentIndex: number,
|
|
117
|
-
synchronizer: TextAudioSynchronizer,
|
|
118
|
-
textStream: AsyncIterableQueue<string>,
|
|
119
|
-
audioStream: AsyncIterableQueue<AudioFrame>,
|
|
120
|
-
): PlayoutHandle {
|
|
121
|
-
const handle = new PlayoutHandle(
|
|
122
|
-
this.#audioSource,
|
|
123
|
-
this.#sampleRate,
|
|
124
|
-
itemId,
|
|
125
|
-
contentIndex,
|
|
126
|
-
synchronizer,
|
|
127
|
-
);
|
|
128
|
-
this.#playoutTask = this.#makePlayoutTask(this.#playoutTask, handle, textStream, audioStream);
|
|
129
|
-
return handle;
|
|
130
|
-
}
|
|
131
|
-
|
|
132
|
-
#makePlayoutTask(
|
|
133
|
-
oldTask: CancellablePromise<void> | null,
|
|
134
|
-
handle: PlayoutHandle,
|
|
135
|
-
textStream: AsyncIterableQueue<string>,
|
|
136
|
-
audioStream: AsyncIterableQueue<AudioFrame>,
|
|
137
|
-
): CancellablePromise<void> {
|
|
138
|
-
return new CancellablePromise<void>((resolve, reject, onCancel) => {
|
|
139
|
-
let cancelled = false;
|
|
140
|
-
onCancel(() => {
|
|
141
|
-
cancelled = true;
|
|
142
|
-
});
|
|
143
|
-
|
|
144
|
-
(async () => {
|
|
145
|
-
try {
|
|
146
|
-
if (oldTask) {
|
|
147
|
-
await gracefullyCancel(oldTask);
|
|
148
|
-
}
|
|
149
|
-
|
|
150
|
-
let firstFrame = true;
|
|
151
|
-
|
|
152
|
-
const readText = () =>
|
|
153
|
-
new CancellablePromise<void>((resolveText, rejectText, onCancelText) => {
|
|
154
|
-
let cancelledText = false;
|
|
155
|
-
onCancelText(() => {
|
|
156
|
-
cancelledText = true;
|
|
157
|
-
});
|
|
158
|
-
|
|
159
|
-
(async () => {
|
|
160
|
-
try {
|
|
161
|
-
for await (const text of textStream) {
|
|
162
|
-
if (cancelledText || cancelled) {
|
|
163
|
-
break;
|
|
164
|
-
}
|
|
165
|
-
handle.synchronizer.pushText(text);
|
|
166
|
-
}
|
|
167
|
-
if (!cancelled) {
|
|
168
|
-
handle.synchronizer.markTextSegmentEnd();
|
|
169
|
-
}
|
|
170
|
-
resolveText();
|
|
171
|
-
} catch (error) {
|
|
172
|
-
rejectText(error);
|
|
173
|
-
}
|
|
174
|
-
})();
|
|
175
|
-
});
|
|
176
|
-
|
|
177
|
-
const capture = () =>
|
|
178
|
-
new CancellablePromise<void>((resolveCapture, rejectCapture, onCancelCapture) => {
|
|
179
|
-
let cancelledCapture = false;
|
|
180
|
-
onCancelCapture(() => {
|
|
181
|
-
cancelledCapture = true;
|
|
182
|
-
});
|
|
183
|
-
|
|
184
|
-
(async () => {
|
|
185
|
-
try {
|
|
186
|
-
const samplesPerChannel = this.#outFrameSize;
|
|
187
|
-
const bstream = new AudioByteStream(
|
|
188
|
-
this.#sampleRate,
|
|
189
|
-
this.#numChannels,
|
|
190
|
-
samplesPerChannel,
|
|
191
|
-
);
|
|
192
|
-
|
|
193
|
-
for await (const frame of audioStream) {
|
|
194
|
-
if (cancelledCapture || cancelled) {
|
|
195
|
-
break;
|
|
196
|
-
}
|
|
197
|
-
if (firstFrame) {
|
|
198
|
-
handle.synchronizer.segmentPlayoutStarted();
|
|
199
|
-
this.emit('playout_started');
|
|
200
|
-
firstFrame = false;
|
|
201
|
-
}
|
|
202
|
-
|
|
203
|
-
handle.synchronizer.pushAudio(frame);
|
|
204
|
-
|
|
205
|
-
for (const f of bstream.write(frame.data.buffer)) {
|
|
206
|
-
handle.pushedDuration += (f.samplesPerChannel / f.sampleRate) * 1000;
|
|
207
|
-
await this.#audioSource.captureFrame(f);
|
|
208
|
-
}
|
|
209
|
-
}
|
|
210
|
-
|
|
211
|
-
if (!cancelledCapture && !cancelled) {
|
|
212
|
-
for (const f of bstream.flush()) {
|
|
213
|
-
handle.pushedDuration += (f.samplesPerChannel / f.sampleRate) * 1000;
|
|
214
|
-
await this.#audioSource.captureFrame(f);
|
|
215
|
-
}
|
|
216
|
-
|
|
217
|
-
handle.synchronizer.markAudioSegmentEnd();
|
|
218
|
-
|
|
219
|
-
await this.#audioSource.waitForPlayout();
|
|
220
|
-
}
|
|
221
|
-
|
|
222
|
-
resolveCapture();
|
|
223
|
-
} catch (error) {
|
|
224
|
-
rejectCapture(error);
|
|
225
|
-
}
|
|
226
|
-
})();
|
|
227
|
-
});
|
|
228
|
-
|
|
229
|
-
const readTextTask = readText();
|
|
230
|
-
const captureTask = capture();
|
|
231
|
-
|
|
232
|
-
try {
|
|
233
|
-
await Promise.race([captureTask, handle.intFut.await]);
|
|
234
|
-
} finally {
|
|
235
|
-
if (!captureTask.isCancelled) {
|
|
236
|
-
await gracefullyCancel(captureTask);
|
|
237
|
-
}
|
|
238
|
-
|
|
239
|
-
if (!readTextTask.isCancelled) {
|
|
240
|
-
await gracefullyCancel(readTextTask);
|
|
241
|
-
}
|
|
242
|
-
|
|
243
|
-
handle.totalPlayedTime = handle.pushedDuration - this.#audioSource.queuedDuration;
|
|
244
|
-
|
|
245
|
-
if (handle.interrupted || captureTask.error) {
|
|
246
|
-
this.#audioSource.clearQueue(); // make sure to remove any queued frames
|
|
247
|
-
}
|
|
248
|
-
|
|
249
|
-
if (!firstFrame) {
|
|
250
|
-
this.emit('playout_stopped', handle.interrupted);
|
|
251
|
-
}
|
|
252
|
-
|
|
253
|
-
handle.doneFut.resolve();
|
|
254
|
-
|
|
255
|
-
const isInterrupted = handle.interrupted || !!captureTask.error;
|
|
256
|
-
await handle.synchronizer.close(isInterrupted);
|
|
257
|
-
}
|
|
258
|
-
|
|
259
|
-
resolve();
|
|
260
|
-
} catch (error) {
|
|
261
|
-
reject(error);
|
|
262
|
-
}
|
|
263
|
-
})();
|
|
264
|
-
});
|
|
265
|
-
}
|
|
266
|
-
}
|
package/src/multimodal/index.ts
DELETED