@livekit/agents 1.0.20 → 1.0.21
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/inference/stt.cjs +171 -125
- package/dist/inference/stt.cjs.map +1 -1
- package/dist/inference/stt.d.cts +4 -0
- package/dist/inference/stt.d.ts +4 -0
- package/dist/inference/stt.d.ts.map +1 -1
- package/dist/inference/stt.js +171 -125
- package/dist/inference/stt.js.map +1 -1
- package/dist/tts/tts.cjs +10 -1
- package/dist/tts/tts.cjs.map +1 -1
- package/dist/tts/tts.d.ts.map +1 -1
- package/dist/tts/tts.js +10 -1
- package/dist/tts/tts.js.map +1 -1
- package/dist/utils.cjs +15 -22
- package/dist/utils.cjs.map +1 -1
- package/dist/utils.d.cts +0 -1
- package/dist/utils.d.ts +0 -1
- package/dist/utils.d.ts.map +1 -1
- package/dist/utils.js +15 -21
- package/dist/utils.js.map +1 -1
- package/dist/utils.test.cjs +2 -1
- package/dist/utils.test.cjs.map +1 -1
- package/dist/utils.test.js +3 -10
- package/dist/utils.test.js.map +1 -1
- package/package.json +1 -1
- package/src/inference/stt.ts +201 -136
- package/src/tts/tts.ts +10 -1
- package/src/utils.test.ts +3 -10
- package/src/utils.ts +21 -27
package/dist/inference/stt.js
CHANGED
|
@@ -1,8 +1,8 @@
|
|
|
1
1
|
import {} from "@livekit/rtc-node";
|
|
2
|
-
import {} from "ws";
|
|
3
2
|
import { APIError, APIStatusError } from "../_exceptions.js";
|
|
4
3
|
import { AudioByteStream } from "../audio.js";
|
|
5
4
|
import { log } from "../log.js";
|
|
5
|
+
import { createStreamChannel } from "../stream/stream_channel.js";
|
|
6
6
|
import {
|
|
7
7
|
STT as BaseSTT,
|
|
8
8
|
SpeechStream as BaseSpeechStream,
|
|
@@ -80,6 +80,32 @@ class STT extends BaseSTT {
|
|
|
80
80
|
this.streams.add(stream);
|
|
81
81
|
return stream;
|
|
82
82
|
}
|
|
83
|
+
async connectWs(timeout) {
|
|
84
|
+
const params = {
|
|
85
|
+
settings: {
|
|
86
|
+
sample_rate: String(this.opts.sampleRate),
|
|
87
|
+
encoding: this.opts.encoding,
|
|
88
|
+
extra: this.opts.modelOptions
|
|
89
|
+
}
|
|
90
|
+
};
|
|
91
|
+
if (this.opts.model && this.opts.model !== "auto") {
|
|
92
|
+
params.model = this.opts.model;
|
|
93
|
+
}
|
|
94
|
+
if (this.opts.language) {
|
|
95
|
+
params.settings.language = this.opts.language;
|
|
96
|
+
}
|
|
97
|
+
let baseURL = this.opts.baseURL;
|
|
98
|
+
if (baseURL.startsWith("http://") || baseURL.startsWith("https://")) {
|
|
99
|
+
baseURL = baseURL.replace("http", "ws");
|
|
100
|
+
}
|
|
101
|
+
const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);
|
|
102
|
+
const url = `${baseURL}/stt`;
|
|
103
|
+
const headers = { Authorization: `Bearer ${token}` };
|
|
104
|
+
const socket = await connectWs(url, headers, timeout);
|
|
105
|
+
const msg = { ...params, type: "session.create" };
|
|
106
|
+
socket.send(JSON.stringify(msg));
|
|
107
|
+
return socket;
|
|
108
|
+
}
|
|
83
109
|
}
|
|
84
110
|
class SpeechStream extends BaseSpeechStream {
|
|
85
111
|
opts;
|
|
@@ -87,158 +113,178 @@ class SpeechStream extends BaseSpeechStream {
|
|
|
87
113
|
speaking = false;
|
|
88
114
|
speechDuration = 0;
|
|
89
115
|
reconnectEvent = new Event();
|
|
116
|
+
stt;
|
|
117
|
+
connOptions;
|
|
90
118
|
#logger = log();
|
|
91
119
|
constructor(sttImpl, opts, connOptions) {
|
|
92
120
|
super(sttImpl, opts.sampleRate, connOptions);
|
|
93
121
|
this.opts = opts;
|
|
122
|
+
this.stt = sttImpl;
|
|
123
|
+
this.connOptions = connOptions;
|
|
94
124
|
}
|
|
95
125
|
get label() {
|
|
96
126
|
return "inference.SpeechStream";
|
|
97
127
|
}
|
|
98
128
|
updateOptions(opts) {
|
|
99
129
|
this.opts = { ...this.opts, ...opts };
|
|
130
|
+
this.reconnectEvent.set();
|
|
100
131
|
}
|
|
101
132
|
async run() {
|
|
102
|
-
|
|
103
|
-
|
|
104
|
-
|
|
105
|
-
|
|
106
|
-
const
|
|
107
|
-
|
|
108
|
-
|
|
109
|
-
|
|
110
|
-
|
|
111
|
-
|
|
133
|
+
while (true) {
|
|
134
|
+
let ws = null;
|
|
135
|
+
let closing = false;
|
|
136
|
+
let finalReceived = false;
|
|
137
|
+
const eventChannel = createStreamChannel();
|
|
138
|
+
const resourceCleanup = () => {
|
|
139
|
+
if (closing) return;
|
|
140
|
+
closing = true;
|
|
141
|
+
eventChannel.close();
|
|
142
|
+
ws == null ? void 0 : ws.removeAllListeners();
|
|
143
|
+
ws == null ? void 0 : ws.close();
|
|
112
144
|
};
|
|
113
|
-
|
|
114
|
-
|
|
115
|
-
|
|
116
|
-
|
|
117
|
-
|
|
118
|
-
}
|
|
119
|
-
let baseURL = this.opts.baseURL;
|
|
120
|
-
if (baseURL.startsWith("http://") || baseURL.startsWith("https://")) {
|
|
121
|
-
baseURL = baseURL.replace("http", "ws");
|
|
122
|
-
}
|
|
123
|
-
const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);
|
|
124
|
-
const url = `${baseURL}/stt`;
|
|
125
|
-
const headers = { Authorization: `Bearer ${token}` };
|
|
126
|
-
const socket = await connectWs(url, headers, 1e4);
|
|
127
|
-
const msg = { ...params, type: "session.create" };
|
|
128
|
-
socket.send(JSON.stringify(msg));
|
|
129
|
-
return socket;
|
|
130
|
-
};
|
|
131
|
-
const send = async (socket, signal) => {
|
|
132
|
-
const audioStream = new AudioByteStream(
|
|
133
|
-
this.opts.sampleRate,
|
|
134
|
-
1,
|
|
135
|
-
Math.floor(this.opts.sampleRate / 20)
|
|
136
|
-
// 50ms
|
|
137
|
-
);
|
|
138
|
-
for await (const ev of this.input) {
|
|
139
|
-
if (signal.aborted) break;
|
|
140
|
-
let frames;
|
|
141
|
-
if (ev === SpeechStream.FLUSH_SENTINEL) {
|
|
142
|
-
frames = audioStream.flush();
|
|
143
|
-
} else {
|
|
144
|
-
const frame = ev;
|
|
145
|
-
frames = audioStream.write(new Int16Array(frame.data).buffer);
|
|
146
|
-
}
|
|
147
|
-
for (const frame of frames) {
|
|
148
|
-
this.speechDuration += frame.samplesPerChannel / frame.sampleRate;
|
|
149
|
-
const base64 = Buffer.from(frame.data.buffer).toString("base64");
|
|
150
|
-
const msg = { type: "input_audio", audio: base64 };
|
|
151
|
-
socket.send(JSON.stringify(msg));
|
|
152
|
-
}
|
|
153
|
-
}
|
|
154
|
-
closingWs = true;
|
|
155
|
-
socket.send(JSON.stringify({ type: "session.finalize" }));
|
|
156
|
-
};
|
|
157
|
-
const recv = async (socket, signal) => {
|
|
158
|
-
while (!this.closed && !signal.aborted) {
|
|
159
|
-
const dataPromise = new Promise((resolve, reject) => {
|
|
160
|
-
const messageHandler = (d) => {
|
|
161
|
-
resolve(d.toString());
|
|
162
|
-
removeListeners();
|
|
145
|
+
const createWsListener = async (ws2, signal) => {
|
|
146
|
+
return new Promise((resolve, reject) => {
|
|
147
|
+
const onAbort = () => {
|
|
148
|
+
resourceCleanup();
|
|
149
|
+
reject(new Error("WebSocket connection aborted"));
|
|
163
150
|
};
|
|
164
|
-
|
|
151
|
+
signal.addEventListener("abort", onAbort, { once: true });
|
|
152
|
+
ws2.on("message", (data) => {
|
|
153
|
+
const json = JSON.parse(data.toString());
|
|
154
|
+
eventChannel.write(json);
|
|
155
|
+
});
|
|
156
|
+
ws2.on("error", (e) => {
|
|
157
|
+
this.#logger.error({ error: e }, "WebSocket error");
|
|
158
|
+
resourceCleanup();
|
|
165
159
|
reject(e);
|
|
166
|
-
|
|
167
|
-
|
|
168
|
-
|
|
169
|
-
if (
|
|
170
|
-
|
|
160
|
+
});
|
|
161
|
+
ws2.on("close", (code) => {
|
|
162
|
+
resourceCleanup();
|
|
163
|
+
if (!closing) return this.#logger.error("WebSocket closed unexpectedly");
|
|
164
|
+
if (finalReceived) return resolve();
|
|
165
|
+
reject(
|
|
166
|
+
new APIStatusError({
|
|
167
|
+
message: "LiveKit STT connection closed unexpectedly",
|
|
168
|
+
options: { statusCode: code }
|
|
169
|
+
})
|
|
170
|
+
);
|
|
171
|
+
});
|
|
172
|
+
});
|
|
173
|
+
};
|
|
174
|
+
const send = async (socket, signal) => {
|
|
175
|
+
const audioStream = new AudioByteStream(
|
|
176
|
+
this.opts.sampleRate,
|
|
177
|
+
1,
|
|
178
|
+
Math.floor(this.opts.sampleRate / 20)
|
|
179
|
+
// 50ms
|
|
180
|
+
);
|
|
181
|
+
const abortPromise = new Promise((_, reject) => {
|
|
182
|
+
if (signal.aborted) {
|
|
183
|
+
return reject(new Error("Send aborted"));
|
|
184
|
+
}
|
|
185
|
+
const onAbort = () => reject(new Error("Send aborted"));
|
|
186
|
+
signal.addEventListener("abort", onAbort, { once: true });
|
|
187
|
+
});
|
|
188
|
+
const iterator = this.input[Symbol.asyncIterator]();
|
|
189
|
+
try {
|
|
190
|
+
while (true) {
|
|
191
|
+
const result = await Promise.race([iterator.next(), abortPromise]);
|
|
192
|
+
if (result.done) break;
|
|
193
|
+
const ev = result.value;
|
|
194
|
+
let frames;
|
|
195
|
+
if (ev === SpeechStream.FLUSH_SENTINEL) {
|
|
196
|
+
frames = audioStream.flush();
|
|
171
197
|
} else {
|
|
172
|
-
|
|
173
|
-
|
|
174
|
-
message: "LiveKit STT connection closed unexpectedly",
|
|
175
|
-
options: { statusCode: code }
|
|
176
|
-
})
|
|
177
|
-
);
|
|
198
|
+
const frame = ev;
|
|
199
|
+
frames = audioStream.write(new Int16Array(frame.data).buffer);
|
|
178
200
|
}
|
|
179
|
-
|
|
180
|
-
|
|
181
|
-
|
|
182
|
-
|
|
183
|
-
|
|
184
|
-
|
|
185
|
-
}
|
|
186
|
-
|
|
187
|
-
socket.
|
|
188
|
-
|
|
189
|
-
|
|
190
|
-
|
|
191
|
-
|
|
192
|
-
|
|
193
|
-
const type = json.type;
|
|
194
|
-
switch (type) {
|
|
195
|
-
case "session.created":
|
|
196
|
-
case "session.finalized":
|
|
197
|
-
case "session.closed":
|
|
198
|
-
break;
|
|
199
|
-
case "interim_transcript":
|
|
200
|
-
this.processTranscript(json, false);
|
|
201
|
-
break;
|
|
202
|
-
case "final_transcript":
|
|
203
|
-
this.processTranscript(json, true);
|
|
204
|
-
break;
|
|
205
|
-
case "error":
|
|
206
|
-
this.#logger.error("received error from LiveKit STT: %o", json);
|
|
207
|
-
throw new APIError(`LiveKit STT returned error: ${JSON.stringify(json)}`);
|
|
208
|
-
default:
|
|
209
|
-
this.#logger.warn("received unexpected message from LiveKit STT: %o", json);
|
|
210
|
-
break;
|
|
201
|
+
for (const frame of frames) {
|
|
202
|
+
this.speechDuration += frame.samplesPerChannel / frame.sampleRate;
|
|
203
|
+
const base64 = Buffer.from(frame.data.buffer).toString("base64");
|
|
204
|
+
const msg = { type: "input_audio", audio: base64 };
|
|
205
|
+
socket.send(JSON.stringify(msg));
|
|
206
|
+
}
|
|
207
|
+
}
|
|
208
|
+
closing = true;
|
|
209
|
+
socket.send(JSON.stringify({ type: "session.finalize" }));
|
|
210
|
+
} catch (e) {
|
|
211
|
+
if (e.message === "Send aborted") {
|
|
212
|
+
return;
|
|
213
|
+
}
|
|
214
|
+
throw e;
|
|
211
215
|
}
|
|
212
|
-
}
|
|
213
|
-
|
|
214
|
-
|
|
216
|
+
};
|
|
217
|
+
const recv = async (signal) => {
|
|
218
|
+
const serverEventStream = eventChannel.stream();
|
|
219
|
+
const reader = serverEventStream.getReader();
|
|
220
|
+
try {
|
|
221
|
+
while (!this.closed && !signal.aborted) {
|
|
222
|
+
const result = await reader.read();
|
|
223
|
+
if (signal.aborted) return;
|
|
224
|
+
if (result.done) return;
|
|
225
|
+
const json = result.value;
|
|
226
|
+
const type = json.type;
|
|
227
|
+
switch (type) {
|
|
228
|
+
case "session.created":
|
|
229
|
+
case "session.finalized":
|
|
230
|
+
break;
|
|
231
|
+
case "session.closed":
|
|
232
|
+
finalReceived = true;
|
|
233
|
+
resourceCleanup();
|
|
234
|
+
break;
|
|
235
|
+
case "interim_transcript":
|
|
236
|
+
this.processTranscript(json, false);
|
|
237
|
+
break;
|
|
238
|
+
case "final_transcript":
|
|
239
|
+
this.processTranscript(json, true);
|
|
240
|
+
break;
|
|
241
|
+
case "error":
|
|
242
|
+
this.#logger.error({ error: json }, "Received error from LiveKit STT");
|
|
243
|
+
resourceCleanup();
|
|
244
|
+
throw new APIError(`LiveKit STT returned error: ${JSON.stringify(json)}`);
|
|
245
|
+
default:
|
|
246
|
+
this.#logger.warn(
|
|
247
|
+
{ message: json },
|
|
248
|
+
"Received unexpected message from LiveKit STT"
|
|
249
|
+
);
|
|
250
|
+
break;
|
|
251
|
+
}
|
|
252
|
+
}
|
|
253
|
+
} finally {
|
|
254
|
+
reader.releaseLock();
|
|
255
|
+
try {
|
|
256
|
+
await serverEventStream.cancel();
|
|
257
|
+
} catch (e) {
|
|
258
|
+
this.#logger.debug("Error cancelling serverEventStream (may already be cancelled):", e);
|
|
259
|
+
}
|
|
260
|
+
}
|
|
261
|
+
};
|
|
215
262
|
try {
|
|
216
|
-
ws = await
|
|
217
|
-
const
|
|
218
|
-
|
|
219
|
-
});
|
|
220
|
-
const recvTask = Task.from(
|
|
221
|
-
|
|
222
|
-
|
|
223
|
-
|
|
224
|
-
|
|
225
|
-
await Promise.race([this.reconnectEvent.wait(), waitForAbort(signal)]);
|
|
226
|
-
});
|
|
263
|
+
ws = await this.stt.connectWs(this.connOptions.timeoutMs);
|
|
264
|
+
const controller = new AbortController();
|
|
265
|
+
const sendTask = Task.from(({ signal }) => send(ws, signal), controller);
|
|
266
|
+
const wsListenerTask = Task.from(({ signal }) => createWsListener(ws, signal), controller);
|
|
267
|
+
const recvTask = Task.from(({ signal }) => recv(signal), controller);
|
|
268
|
+
const waitReconnectTask = Task.from(
|
|
269
|
+
({ signal }) => Promise.race([this.reconnectEvent.wait(), waitForAbort(signal)]),
|
|
270
|
+
controller
|
|
271
|
+
);
|
|
227
272
|
try {
|
|
228
273
|
await Promise.race([
|
|
229
|
-
Promise.all(
|
|
274
|
+
Promise.all([sendTask.result, wsListenerTask.result, recvTask.result]),
|
|
230
275
|
waitReconnectTask.result
|
|
231
276
|
]);
|
|
232
277
|
if (!waitReconnectTask.done) break;
|
|
233
278
|
this.reconnectEvent.clear();
|
|
234
279
|
} finally {
|
|
235
|
-
await cancelAndWait(
|
|
280
|
+
await cancelAndWait(
|
|
281
|
+
[sendTask, wsListenerTask, recvTask, waitReconnectTask],
|
|
282
|
+
DEFAULT_CANCEL_TIMEOUT
|
|
283
|
+
);
|
|
284
|
+
resourceCleanup();
|
|
236
285
|
}
|
|
237
286
|
} finally {
|
|
238
|
-
|
|
239
|
-
if (ws) ws.close();
|
|
240
|
-
} catch {
|
|
241
|
-
}
|
|
287
|
+
resourceCleanup();
|
|
242
288
|
}
|
|
243
289
|
}
|
|
244
290
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/inference/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame } from '@livekit/rtc-node';\nimport type { WebSocket } from 'ws';\nimport { type RawData } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport {\n STT as BaseSTT,\n SpeechStream as BaseSpeechStream,\n type SpeechData,\n type SpeechEvent,\n SpeechEventType,\n} from '../stt/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { type AudioBuffer, Event, Task, cancelAndWait, shortuuid, waitForAbort } from '../utils.js';\nimport { type AnyString, connectWs, createAccessToken } from './utils.js';\n\nexport type DeepgramModels =\n | 'deepgram'\n | 'deepgram/nova-3'\n | 'deepgram/nova-3-general'\n | 'deepgram/nova-3-medical'\n | 'deepgram/nova-2-conversationalai'\n | 'deepgram/nova-2'\n | 'deepgram/nova-2-general'\n | 'deepgram/nova-2-medical'\n | 'deepgram/nova-2-phonecall';\n\nexport type CartesiaModels = 'cartesia' | 'cartesia/ink-whisper';\n\nexport type AssemblyaiModels = 'assemblyai' | 'assemblyai/universal-streaming';\n\nexport interface CartesiaOptions {\n min_volume?: number; // default: not specified\n max_silence_duration_secs?: number; // default: not specified\n}\n\nexport interface DeepgramOptions {\n filler_words?: boolean; // default: true\n interim_results?: boolean; // default: true\n endpointing?: number; // default: 25 (ms)\n punctuate?: boolean; // default: false\n smart_format?: boolean;\n keywords?: Array<[string, number]>;\n keyterms?: string[];\n profanity_filter?: boolean;\n numerals?: boolean;\n mip_opt_out?: boolean;\n}\n\nexport interface AssemblyAIOptions {\n format_turns?: boolean; // default: false\n end_of_turn_confidence_threshold?: number; // default: 0.01\n min_end_of_turn_silence_when_confident?: number; // default: 0\n max_turn_silence?: number; // default: not specified\n keyterms_prompt?: string[]; // default: not specified\n}\n\nexport type STTLanguages =\n | 'multi'\n | 'en'\n | 'de'\n | 'es'\n | 'fr'\n | 'ja'\n | 'pt'\n | 'zh'\n | 'hi'\n | AnyString;\n\ntype _STTModels = DeepgramModels | CartesiaModels | AssemblyaiModels;\n\nexport type STTModels = _STTModels | 'auto' | AnyString;\n\nexport type ModelWithLanguage = `${_STTModels}:${STTLanguages}` | STTModels;\n\nexport type STTOptions<TModel extends STTModels> = TModel extends DeepgramModels\n ? DeepgramOptions\n : TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends AssemblyaiModels\n ? AssemblyAIOptions\n : Record<string, unknown>;\n\nexport type STTEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: STTEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'wss://agent-gateway.livekit.cloud/v1';\nconst DEFAULT_CANCEL_TIMEOUT = 5000;\n\nexport interface InferenceSTTOptions<TModel extends STTModels> {\n model?: TModel;\n language?: STTLanguages;\n encoding: STTEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: STTOptions<TModel>;\n}\n\n/**\n * Livekit Cloud Inference STT\n */\nexport class STT<TModel extends STTModels> extends BaseSTT {\n private opts: InferenceSTTOptions<TModel>;\n private streams: Set<SpeechStream<TModel>> = new Set();\n\n #logger = log();\n\n constructor(opts?: {\n model?: TModel;\n language?: STTLanguages;\n baseURL?: string;\n encoding?: STTEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: STTOptions<TModel>;\n }) {\n super({ streaming: true, interimResults: true });\n\n const {\n model,\n language,\n baseURL,\n encoding = DEFAULT_ENCODING,\n sampleRate = DEFAULT_SAMPLE_RATE,\n apiKey,\n apiSecret,\n modelOptions = {} as STTOptions<TModel>,\n } = opts || {};\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n this.opts = {\n model,\n language,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n };\n }\n\n get label(): string {\n return 'inference.STT';\n }\n\n static fromModelString(modelString: string): STT<AnyString> {\n if (modelString.includes(':')) {\n const [model, language] = modelString.split(':') as [AnyString, STTLanguages];\n return new STT({ model, language });\n }\n return new STT({ model: modelString });\n }\n\n protected async _recognize(_: AudioBuffer): Promise<SpeechEvent> {\n throw new Error('LiveKit STT does not support batch recognition, use stream() instead');\n }\n\n updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n this.opts = { ...this.opts, ...opts };\n\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n stream(options?: {\n language?: STTLanguages | string;\n connOptions?: APIConnectOptions;\n }): SpeechStream<TModel> {\n const { language, connOptions = DEFAULT_API_CONNECT_OPTIONS } = options || {};\n const streamOpts = {\n ...this.opts,\n language: language ?? this.opts.language,\n } as InferenceSTTOptions<TModel>;\n\n const stream = new SpeechStream(this, streamOpts, connOptions);\n this.streams.add(stream);\n\n return stream;\n }\n}\n\nexport class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {\n private opts: InferenceSTTOptions<TModel>;\n private requestId = shortuuid('stt_request_');\n private speaking = false;\n private speechDuration = 0;\n private reconnectEvent = new Event();\n\n #logger = log();\n\n constructor(\n sttImpl: STT<TModel>,\n opts: InferenceSTTOptions<TModel>,\n connOptions: APIConnectOptions,\n ) {\n super(sttImpl, opts.sampleRate, connOptions);\n this.opts = opts;\n }\n\n get label(): string {\n return 'inference.SpeechStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n this.opts = { ...this.opts, ...opts };\n }\n\n protected async run(): Promise<void> {\n let ws: WebSocket | null = null;\n let closingWs = false;\n\n this.reconnectEvent.set();\n\n const connect = async () => {\n const params = {\n settings: {\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n },\n } as Record<string, unknown>;\n\n if (this.opts.model && this.opts.model !== 'auto') {\n params.model = this.opts.model;\n }\n\n if (this.opts.language) {\n (params.settings as Record<string, unknown>).language = this.opts.language;\n }\n\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/stt`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const socket = await connectWs(url, headers, 10000);\n const msg = { ...params, type: 'session.create' };\n socket.send(JSON.stringify(msg));\n\n return socket;\n };\n\n const send = async (socket: WebSocket, signal: AbortSignal) => {\n const audioStream = new AudioByteStream(\n this.opts.sampleRate,\n 1,\n Math.floor(this.opts.sampleRate / 20), // 50ms\n );\n\n for await (const ev of this.input) {\n if (signal.aborted) break;\n let frames: AudioFrame[];\n\n if (ev === SpeechStream.FLUSH_SENTINEL) {\n frames = audioStream.flush();\n } else {\n const frame = ev as AudioFrame;\n frames = audioStream.write(new Int16Array(frame.data).buffer);\n }\n\n for (const frame of frames) {\n this.speechDuration += frame.samplesPerChannel / frame.sampleRate;\n const base64 = Buffer.from(frame.data.buffer).toString('base64');\n const msg = { type: 'input_audio', audio: base64 };\n socket.send(JSON.stringify(msg));\n }\n }\n\n closingWs = true;\n socket.send(JSON.stringify({ type: 'session.finalize' }));\n };\n\n const recv = async (socket: WebSocket, signal: AbortSignal) => {\n while (!this.closed && !signal.aborted) {\n const dataPromise = new Promise<string>((resolve, reject) => {\n const messageHandler = (d: RawData) => {\n resolve(d.toString());\n removeListeners();\n };\n const errorHandler = (e: Error) => {\n reject(e);\n removeListeners();\n };\n const closeHandler = (code: number) => {\n if (closingWs) {\n resolve('');\n } else {\n reject(\n new APIStatusError({\n message: 'LiveKit STT connection closed unexpectedly',\n options: { statusCode: code },\n }),\n );\n }\n removeListeners();\n };\n const removeListeners = () => {\n socket.removeListener('message', messageHandler);\n socket.removeListener('error', errorHandler);\n socket.removeListener('close', closeHandler);\n };\n socket.once('message', messageHandler);\n socket.once('error', errorHandler);\n socket.once('close', closeHandler);\n });\n\n const data = await Promise.race([dataPromise, waitForAbort(signal)]);\n\n if (!data || signal.aborted) return;\n\n const json = JSON.parse(data);\n const type = json.type as string | undefined;\n\n switch (type) {\n case 'session.created':\n case 'session.finalized':\n case 'session.closed':\n break;\n case 'interim_transcript':\n this.processTranscript(json, false);\n break;\n case 'final_transcript':\n this.processTranscript(json, true);\n break;\n case 'error':\n this.#logger.error('received error from LiveKit STT: %o', json);\n throw new APIError(`LiveKit STT returned error: ${JSON.stringify(json)}`);\n default:\n this.#logger.warn('received unexpected message from LiveKit STT: %o', json);\n break;\n }\n }\n };\n\n while (true) {\n try {\n ws = await connect();\n\n const sendTask = Task.from(async ({ signal }) => {\n await send(ws!, signal);\n });\n\n const recvTask = Task.from(async ({ signal }) => {\n await recv(ws!, signal);\n });\n\n const tasks = [sendTask, recvTask];\n const waitReconnectTask = Task.from(async ({ signal }) => {\n await Promise.race([this.reconnectEvent.wait(), waitForAbort(signal)]);\n });\n\n try {\n await Promise.race([\n Promise.all(tasks.map((task) => task.result)),\n waitReconnectTask.result,\n ]);\n\n if (!waitReconnectTask.done) break;\n this.reconnectEvent.clear();\n } finally {\n await cancelAndWait([sendTask, recvTask, waitReconnectTask], DEFAULT_CANCEL_TIMEOUT);\n }\n } finally {\n try {\n if (ws) ws.close();\n } catch {}\n }\n }\n }\n\n private processTranscript(data: Record<string, any>, isFinal: boolean) {\n const requestId = data.request_id ?? this.requestId;\n const text = data.transcript ?? '';\n const language = data.language ?? this.opts.language ?? 'en';\n\n if (!text && !isFinal) return;\n\n // We'll have a more accurate way of detecting when speech started when we have VAD\n if (!this.speaking) {\n this.speaking = true;\n this.queue.put({ type: SpeechEventType.START_OF_SPEECH });\n }\n\n const speechData: SpeechData = {\n language,\n startTime: data.start ?? 0,\n endTime: data.duration ?? 0,\n confidence: data.confidence ?? 1.0,\n text,\n };\n\n if (isFinal) {\n if (this.speechDuration > 0) {\n this.queue.put({\n type: SpeechEventType.RECOGNITION_USAGE,\n requestId,\n recognitionUsage: { audioDuration: this.speechDuration },\n });\n this.speechDuration = 0;\n }\n\n this.queue.put({\n type: SpeechEventType.FINAL_TRANSCRIPT,\n requestId,\n alternatives: [speechData],\n });\n\n if (this.speaking) {\n this.speaking = false;\n this.queue.put({ type: SpeechEventType.END_OF_SPEECH });\n }\n } else {\n this.queue.put({\n type: SpeechEventType.INTERIM_TRANSCRIPT,\n requestId,\n alternatives: [speechData],\n });\n }\n }\n}\n"],"mappings":"AAGA,eAAgC;AAEhC,eAA6B;AAC7B,SAAS,UAAU,sBAAsB;AACzC,SAAS,uBAAuB;AAChC,SAAS,WAAW;AACpB;AAAA,EACE,OAAO;AAAA,EACP,gBAAgB;AAAA,EAGhB;AAAA,OACK;AACP,SAAiC,mCAAmC;AACpE,SAA2B,OAAO,MAAM,eAAe,WAAW,oBAAoB;AACtF,SAAyB,WAAW,yBAAyB;AAuE7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,yBAAyB;AAgBxB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAqC,oBAAI,IAAI;AAAA,EAErD,UAAU,IAAI;AAAA,EAEd,YAAY,MAST;AACD,UAAM,EAAE,WAAW,MAAM,gBAAgB,KAAK,CAAC;AAE/C,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX,aAAa;AAAA,MACb;AAAA,MACA;AAAA,MACA,eAAe,CAAC;AAAA,IAClB,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,WAAW,QAAQ,IAAI,yBAAyB;AAClE,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAEA,SAAK,OAAO;AAAA,MACV;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,QAAI,YAAY,SAAS,GAAG,GAAG;AAC7B,YAAM,CAAC,OAAO,QAAQ,IAAI,YAAY,MAAM,GAAG;AAC/C,aAAO,IAAI,IAAI,EAAE,OAAO,SAAS,CAAC;AAAA,IACpC;AACA,WAAO,IAAI,IAAI,EAAE,OAAO,YAAY,CAAC;AAAA,EACvC;AAAA,EAEA,MAAgB,WAAW,GAAsC;AAC/D,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAEpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,OAAO,SAGkB;AACvB,UAAM,EAAE,UAAU,cAAc,4BAA4B,IAAI,WAAW,CAAC;AAC5E,UAAM,aAAa;AAAA,MACjB,GAAG,KAAK;AAAA,MACR,UAAU,YAAY,KAAK,KAAK;AAAA,IAClC;AAEA,UAAM,SAAS,IAAI,aAAa,MAAM,YAAY,WAAW;AAC7D,SAAK,QAAQ,IAAI,MAAM;AAEvB,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAA+C,iBAAiB;AAAA,EACnE;AAAA,EACA,YAAY,UAAU,cAAc;AAAA,EACpC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,iBAAiB,IAAI,MAAM;AAAA,EAEnC,UAAU,IAAI;AAAA,EAEd,YACE,SACA,MACA,aACA;AACA,UAAM,SAAS,KAAK,YAAY,WAAW;AAC3C,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAAA,EACtC;AAAA,EAEA,MAAgB,MAAqB;AACnC,QAAI,KAAuB;AAC3B,QAAI,YAAY;AAEhB,SAAK,eAAe,IAAI;AAExB,UAAM,UAAU,YAAY;AAC1B,YAAM,SAAS;AAAA,QACb,UAAU;AAAA,UACR,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,UACxC,UAAU,KAAK,KAAK;AAAA,UACpB,OAAO,KAAK,KAAK;AAAA,QACnB;AAAA,MACF;AAEA,UAAI,KAAK,KAAK,SAAS,KAAK,KAAK,UAAU,QAAQ;AACjD,eAAO,QAAQ,KAAK,KAAK;AAAA,MAC3B;AAEA,UAAI,KAAK,KAAK,UAAU;AACtB,QAAC,OAAO,SAAqC,WAAW,KAAK,KAAK;AAAA,MACpE;AAEA,UAAI,UAAU,KAAK,KAAK;AACxB,UAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,kBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,MACxC;AAEA,YAAM,QAAQ,MAAM,kBAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,YAAM,MAAM,GAAG,OAAO;AACtB,YAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,YAAM,SAAS,MAAM,UAAU,KAAK,SAAS,GAAK;AAClD,YAAM,MAAM,EAAE,GAAG,QAAQ,MAAM,iBAAiB;AAChD,aAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAE/B,aAAO;AAAA,IACT;AAEA,UAAM,OAAO,OAAO,QAAmB,WAAwB;AAC7D,YAAM,cAAc,IAAI;AAAA,QACtB,KAAK,KAAK;AAAA,QACV;AAAA,QACA,KAAK,MAAM,KAAK,KAAK,aAAa,EAAE;AAAA;AAAA,MACtC;AAEA,uBAAiB,MAAM,KAAK,OAAO;AACjC,YAAI,OAAO,QAAS;AACpB,YAAI;AAEJ,YAAI,OAAO,aAAa,gBAAgB;AACtC,mBAAS,YAAY,MAAM;AAAA,QAC7B,OAAO;AACL,gBAAM,QAAQ;AACd,mBAAS,YAAY,MAAM,IAAI,WAAW,MAAM,IAAI,EAAE,MAAM;AAAA,QAC9D;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,kBAAkB,MAAM,oBAAoB,MAAM;AACvD,gBAAM,SAAS,OAAO,KAAK,MAAM,KAAK,MAAM,EAAE,SAAS,QAAQ;AAC/D,gBAAM,MAAM,EAAE,MAAM,eAAe,OAAO,OAAO;AACjD,iBAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,QACjC;AAAA,MACF;AAEA,kBAAY;AACZ,aAAO,KAAK,KAAK,UAAU,EAAE,MAAM,mBAAmB,CAAC,CAAC;AAAA,IAC1D;AAEA,UAAM,OAAO,OAAO,QAAmB,WAAwB;AAC7D,aAAO,CAAC,KAAK,UAAU,CAAC,OAAO,SAAS;AACtC,cAAM,cAAc,IAAI,QAAgB,CAAC,SAAS,WAAW;AAC3D,gBAAM,iBAAiB,CAAC,MAAe;AACrC,oBAAQ,EAAE,SAAS,CAAC;AACpB,4BAAgB;AAAA,UAClB;AACA,gBAAM,eAAe,CAAC,MAAa;AACjC,mBAAO,CAAC;AACR,4BAAgB;AAAA,UAClB;AACA,gBAAM,eAAe,CAAC,SAAiB;AACrC,gBAAI,WAAW;AACb,sBAAQ,EAAE;AAAA,YACZ,OAAO;AACL;AAAA,gBACE,IAAI,eAAe;AAAA,kBACjB,SAAS;AAAA,kBACT,SAAS,EAAE,YAAY,KAAK;AAAA,gBAC9B,CAAC;AAAA,cACH;AAAA,YACF;AACA,4BAAgB;AAAA,UAClB;AACA,gBAAM,kBAAkB,MAAM;AAC5B,mBAAO,eAAe,WAAW,cAAc;AAC/C,mBAAO,eAAe,SAAS,YAAY;AAC3C,mBAAO,eAAe,SAAS,YAAY;AAAA,UAC7C;AACA,iBAAO,KAAK,WAAW,cAAc;AACrC,iBAAO,KAAK,SAAS,YAAY;AACjC,iBAAO,KAAK,SAAS,YAAY;AAAA,QACnC,CAAC;AAED,cAAM,OAAO,MAAM,QAAQ,KAAK,CAAC,aAAa,aAAa,MAAM,CAAC,CAAC;AAEnE,YAAI,CAAC,QAAQ,OAAO,QAAS;AAE7B,cAAM,OAAO,KAAK,MAAM,IAAI;AAC5B,cAAM,OAAO,KAAK;AAElB,gBAAQ,MAAM;AAAA,UACZ,KAAK;AAAA,UACL,KAAK;AAAA,UACL,KAAK;AACH;AAAA,UACF,KAAK;AACH,iBAAK,kBAAkB,MAAM,KAAK;AAClC;AAAA,UACF,KAAK;AACH,iBAAK,kBAAkB,MAAM,IAAI;AACjC;AAAA,UACF,KAAK;AACH,iBAAK,QAAQ,MAAM,uCAAuC,IAAI;AAC9D,kBAAM,IAAI,SAAS,+BAA+B,KAAK,UAAU,IAAI,CAAC,EAAE;AAAA,UAC1E;AACE,iBAAK,QAAQ,KAAK,oDAAoD,IAAI;AAC1E;AAAA,QACJ;AAAA,MACF;AAAA,IACF;AAEA,WAAO,MAAM;AACX,UAAI;AACF,aAAK,MAAM,QAAQ;AAEnB,cAAM,WAAW,KAAK,KAAK,OAAO,EAAE,OAAO,MAAM;AAC/C,gBAAM,KAAK,IAAK,MAAM;AAAA,QACxB,CAAC;AAED,cAAM,WAAW,KAAK,KAAK,OAAO,EAAE,OAAO,MAAM;AAC/C,gBAAM,KAAK,IAAK,MAAM;AAAA,QACxB,CAAC;AAED,cAAM,QAAQ,CAAC,UAAU,QAAQ;AACjC,cAAM,oBAAoB,KAAK,KAAK,OAAO,EAAE,OAAO,MAAM;AACxD,gBAAM,QAAQ,KAAK,CAAC,KAAK,eAAe,KAAK,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,QACvE,CAAC;AAED,YAAI;AACF,gBAAM,QAAQ,KAAK;AAAA,YACjB,QAAQ,IAAI,MAAM,IAAI,CAAC,SAAS,KAAK,MAAM,CAAC;AAAA,YAC5C,kBAAkB;AAAA,UACpB,CAAC;AAED,cAAI,CAAC,kBAAkB,KAAM;AAC7B,eAAK,eAAe,MAAM;AAAA,QAC5B,UAAE;AACA,gBAAM,cAAc,CAAC,UAAU,UAAU,iBAAiB,GAAG,sBAAsB;AAAA,QACrF;AAAA,MACF,UAAE;AACA,YAAI;AACF,cAAI,GAAI,IAAG,MAAM;AAAA,QACnB,QAAQ;AAAA,QAAC;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,kBAAkB,MAA2B,SAAkB;AACrE,UAAM,YAAY,KAAK,cAAc,KAAK;AAC1C,UAAM,OAAO,KAAK,cAAc;AAChC,UAAM,WAAW,KAAK,YAAY,KAAK,KAAK,YAAY;AAExD,QAAI,CAAC,QAAQ,CAAC,QAAS;AAGvB,QAAI,CAAC,KAAK,UAAU;AAClB,WAAK,WAAW;AAChB,WAAK,MAAM,IAAI,EAAE,MAAM,gBAAgB,gBAAgB,CAAC;AAAA,IAC1D;AAEA,UAAM,aAAyB;AAAA,MAC7B;AAAA,MACA,WAAW,KAAK,SAAS;AAAA,MACzB,SAAS,KAAK,YAAY;AAAA,MAC1B,YAAY,KAAK,cAAc;AAAA,MAC/B;AAAA,IACF;AAEA,QAAI,SAAS;AACX,UAAI,KAAK,iBAAiB,GAAG;AAC3B,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,gBAAgB;AAAA,UACtB;AAAA,UACA,kBAAkB,EAAE,eAAe,KAAK,eAAe;AAAA,QACzD,CAAC;AACD,aAAK,iBAAiB;AAAA,MACxB;AAEA,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,gBAAgB;AAAA,QACtB;AAAA,QACA,cAAc,CAAC,UAAU;AAAA,MAC3B,CAAC;AAED,UAAI,KAAK,UAAU;AACjB,aAAK,WAAW;AAChB,aAAK,MAAM,IAAI,EAAE,MAAM,gBAAgB,cAAc,CAAC;AAAA,MACxD;AAAA,IACF,OAAO;AACL,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,gBAAgB;AAAA,QACtB;AAAA,QACA,cAAc,CAAC,UAAU;AAAA,MAC3B,CAAC;AAAA,IACH;AAAA,EACF;AACF;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../../src/inference/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioFrame } from '@livekit/rtc-node';\nimport type { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport {\n STT as BaseSTT,\n SpeechStream as BaseSpeechStream,\n type SpeechData,\n type SpeechEvent,\n SpeechEventType,\n} from '../stt/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { type AudioBuffer, Event, Task, cancelAndWait, shortuuid, waitForAbort } from '../utils.js';\nimport { type AnyString, connectWs, createAccessToken } from './utils.js';\n\nexport type DeepgramModels =\n | 'deepgram'\n | 'deepgram/nova-3'\n | 'deepgram/nova-3-general'\n | 'deepgram/nova-3-medical'\n | 'deepgram/nova-2-conversationalai'\n | 'deepgram/nova-2'\n | 'deepgram/nova-2-general'\n | 'deepgram/nova-2-medical'\n | 'deepgram/nova-2-phonecall';\n\nexport type CartesiaModels = 'cartesia' | 'cartesia/ink-whisper';\n\nexport type AssemblyaiModels = 'assemblyai' | 'assemblyai/universal-streaming';\n\nexport interface CartesiaOptions {\n min_volume?: number; // default: not specified\n max_silence_duration_secs?: number; // default: not specified\n}\n\nexport interface DeepgramOptions {\n filler_words?: boolean; // default: true\n interim_results?: boolean; // default: true\n endpointing?: number; // default: 25 (ms)\n punctuate?: boolean; // default: false\n smart_format?: boolean;\n keywords?: Array<[string, number]>;\n keyterms?: string[];\n profanity_filter?: boolean;\n numerals?: boolean;\n mip_opt_out?: boolean;\n}\n\nexport interface AssemblyAIOptions {\n format_turns?: boolean; // default: false\n end_of_turn_confidence_threshold?: number; // default: 0.01\n min_end_of_turn_silence_when_confident?: number; // default: 0\n max_turn_silence?: number; // default: not specified\n keyterms_prompt?: string[]; // default: not specified\n}\n\nexport type STTLanguages =\n | 'multi'\n | 'en'\n | 'de'\n | 'es'\n | 'fr'\n | 'ja'\n | 'pt'\n | 'zh'\n | 'hi'\n | AnyString;\n\ntype _STTModels = DeepgramModels | CartesiaModels | AssemblyaiModels;\n\nexport type STTModels = _STTModels | 'auto' | AnyString;\n\nexport type ModelWithLanguage = `${_STTModels}:${STTLanguages}` | STTModels;\n\nexport type STTOptions<TModel extends STTModels> = TModel extends DeepgramModels\n ? DeepgramOptions\n : TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends AssemblyaiModels\n ? AssemblyAIOptions\n : Record<string, unknown>;\n\nexport type STTEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: STTEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'wss://agent-gateway.livekit.cloud/v1';\nconst DEFAULT_CANCEL_TIMEOUT = 5000;\n\nexport interface InferenceSTTOptions<TModel extends STTModels> {\n model?: TModel;\n language?: STTLanguages;\n encoding: STTEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: STTOptions<TModel>;\n}\n\n/**\n * Livekit Cloud Inference STT\n */\nexport class STT<TModel extends STTModels> extends BaseSTT {\n private opts: InferenceSTTOptions<TModel>;\n private streams: Set<SpeechStream<TModel>> = new Set();\n\n #logger = log();\n\n constructor(opts?: {\n model?: TModel;\n language?: STTLanguages;\n baseURL?: string;\n encoding?: STTEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: STTOptions<TModel>;\n }) {\n super({ streaming: true, interimResults: true });\n\n const {\n model,\n language,\n baseURL,\n encoding = DEFAULT_ENCODING,\n sampleRate = DEFAULT_SAMPLE_RATE,\n apiKey,\n apiSecret,\n modelOptions = {} as STTOptions<TModel>,\n } = opts || {};\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n this.opts = {\n model,\n language,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n };\n }\n\n get label(): string {\n return 'inference.STT';\n }\n\n static fromModelString(modelString: string): STT<AnyString> {\n if (modelString.includes(':')) {\n const [model, language] = modelString.split(':') as [AnyString, STTLanguages];\n return new STT({ model, language });\n }\n return new STT({ model: modelString });\n }\n\n protected async _recognize(_: AudioBuffer): Promise<SpeechEvent> {\n throw new Error('LiveKit STT does not support batch recognition, use stream() instead');\n }\n\n updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n this.opts = { ...this.opts, ...opts };\n\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n stream(options?: {\n language?: STTLanguages | string;\n connOptions?: APIConnectOptions;\n }): SpeechStream<TModel> {\n const { language, connOptions = DEFAULT_API_CONNECT_OPTIONS } = options || {};\n const streamOpts = {\n ...this.opts,\n language: language ?? this.opts.language,\n } as InferenceSTTOptions<TModel>;\n\n const stream = new SpeechStream(this, streamOpts, connOptions);\n this.streams.add(stream);\n\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n const params = {\n settings: {\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n },\n } as Record<string, unknown>;\n\n if (this.opts.model && this.opts.model !== 'auto') {\n params.model = this.opts.model;\n }\n\n if (this.opts.language) {\n (params.settings as Record<string, unknown>).language = this.opts.language;\n }\n\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/stt`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const socket = await connectWs(url, headers, timeout);\n const msg = { ...params, type: 'session.create' };\n socket.send(JSON.stringify(msg));\n\n return socket;\n }\n}\n\nexport class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {\n private opts: InferenceSTTOptions<TModel>;\n private requestId = shortuuid('stt_request_');\n private speaking = false;\n private speechDuration = 0;\n private reconnectEvent = new Event();\n private stt: STT<TModel>;\n private connOptions: APIConnectOptions;\n\n #logger = log();\n\n constructor(\n sttImpl: STT<TModel>,\n opts: InferenceSTTOptions<TModel>,\n connOptions: APIConnectOptions,\n ) {\n super(sttImpl, opts.sampleRate, connOptions);\n this.opts = opts;\n this.stt = sttImpl;\n this.connOptions = connOptions;\n }\n\n get label(): string {\n return 'inference.SpeechStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceSTTOptions<TModel>, 'model' | 'language'>>): void {\n this.opts = { ...this.opts, ...opts };\n this.reconnectEvent.set();\n }\n\n protected async run(): Promise<void> {\n while (true) {\n // Create fresh resources for each connection attempt\n let ws: WebSocket | null = null;\n let closing = false;\n let finalReceived = false;\n\n type SttServerEvent = Record<string, any>;\n const eventChannel = createStreamChannel<SttServerEvent>();\n\n const resourceCleanup = () => {\n if (closing) return;\n closing = true;\n eventChannel.close();\n ws?.removeAllListeners();\n ws?.close();\n };\n\n const createWsListener = async (ws: WebSocket, signal: AbortSignal) => {\n return new Promise<void>((resolve, reject) => {\n const onAbort = () => {\n resourceCleanup();\n reject(new Error('WebSocket connection aborted'));\n };\n\n signal.addEventListener('abort', onAbort, { once: true });\n\n ws.on('message', (data) => {\n const json = JSON.parse(data.toString()) as SttServerEvent;\n eventChannel.write(json);\n });\n\n ws.on('error', (e) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n resourceCleanup();\n reject(e);\n });\n\n ws.on('close', (code: number) => {\n resourceCleanup();\n\n if (!closing) return this.#logger.error('WebSocket closed unexpectedly');\n if (finalReceived) return resolve();\n\n reject(\n new APIStatusError({\n message: 'LiveKit STT connection closed unexpectedly',\n options: { statusCode: code },\n }),\n );\n });\n });\n };\n\n const send = async (socket: WebSocket, signal: AbortSignal) => {\n const audioStream = new AudioByteStream(\n this.opts.sampleRate,\n 1,\n Math.floor(this.opts.sampleRate / 20), // 50ms\n );\n\n // Create abort promise once to avoid memory leak\n const abortPromise = new Promise<never>((_, reject) => {\n if (signal.aborted) {\n return reject(new Error('Send aborted'));\n }\n const onAbort = () => reject(new Error('Send aborted'));\n signal.addEventListener('abort', onAbort, { once: true });\n });\n\n // Manual iteration to support cancellation\n const iterator = this.input[Symbol.asyncIterator]();\n try {\n while (true) {\n const result = await Promise.race([iterator.next(), abortPromise]);\n\n if (result.done) break;\n const ev = result.value;\n\n let frames: AudioFrame[];\n if (ev === SpeechStream.FLUSH_SENTINEL) {\n frames = audioStream.flush();\n } else {\n const frame = ev as AudioFrame;\n frames = audioStream.write(new Int16Array(frame.data).buffer);\n }\n\n for (const frame of frames) {\n this.speechDuration += frame.samplesPerChannel / frame.sampleRate;\n const base64 = Buffer.from(frame.data.buffer).toString('base64');\n const msg = { type: 'input_audio', audio: base64 };\n socket.send(JSON.stringify(msg));\n }\n }\n\n closing = true;\n socket.send(JSON.stringify({ type: 'session.finalize' }));\n } catch (e) {\n if ((e as Error).message === 'Send aborted') {\n // Expected abort, don't log\n return;\n }\n throw e;\n }\n };\n\n const recv = async (signal: AbortSignal) => {\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n while (!this.closed && !signal.aborted) {\n const result = await reader.read();\n if (signal.aborted) return;\n if (result.done) return;\n\n const json = result.value;\n const type = json.type as string | undefined;\n\n switch (type) {\n case 'session.created':\n case 'session.finalized':\n break;\n case 'session.closed':\n finalReceived = true;\n resourceCleanup();\n break;\n case 'interim_transcript':\n this.processTranscript(json, false);\n break;\n case 'final_transcript':\n this.processTranscript(json, true);\n break;\n case 'error':\n this.#logger.error({ error: json }, 'Received error from LiveKit STT');\n resourceCleanup();\n throw new APIError(`LiveKit STT returned error: ${JSON.stringify(json)}`);\n default:\n this.#logger.warn(\n { message: json },\n 'Received unexpected message from LiveKit STT',\n );\n break;\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n ws = await this.stt.connectWs(this.connOptions.timeoutMs);\n\n // Wrap tasks for proper cancellation support using Task signals\n const controller = new AbortController();\n const sendTask = Task.from(({ signal }) => send(ws!, signal), controller);\n const wsListenerTask = Task.from(({ signal }) => createWsListener(ws!, signal), controller);\n const recvTask = Task.from(({ signal }) => recv(signal), controller);\n const waitReconnectTask = Task.from(\n ({ signal }) => Promise.race([this.reconnectEvent.wait(), waitForAbort(signal)]),\n controller,\n );\n\n try {\n await Promise.race([\n Promise.all([sendTask.result, wsListenerTask.result, recvTask.result]),\n waitReconnectTask.result,\n ]);\n\n // If reconnect didn't trigger, tasks finished - exit loop\n if (!waitReconnectTask.done) break;\n\n // Reconnect triggered - clear event and continue loop\n this.reconnectEvent.clear();\n } finally {\n // Cancel all tasks to ensure cleanup\n await cancelAndWait(\n [sendTask, wsListenerTask, recvTask, waitReconnectTask],\n DEFAULT_CANCEL_TIMEOUT,\n );\n resourceCleanup();\n }\n } finally {\n // Ensure cleanup even if connectWs throws\n resourceCleanup();\n }\n }\n }\n\n private processTranscript(data: Record<string, any>, isFinal: boolean) {\n const requestId = data.request_id ?? this.requestId;\n const text = data.transcript ?? '';\n const language = data.language ?? this.opts.language ?? 'en';\n\n if (!text && !isFinal) return;\n\n // We'll have a more accurate way of detecting when speech started when we have VAD\n if (!this.speaking) {\n this.speaking = true;\n this.queue.put({ type: SpeechEventType.START_OF_SPEECH });\n }\n\n const speechData: SpeechData = {\n language,\n startTime: data.start ?? 0,\n endTime: data.duration ?? 0,\n confidence: data.confidence ?? 1.0,\n text,\n };\n\n if (isFinal) {\n if (this.speechDuration > 0) {\n this.queue.put({\n type: SpeechEventType.RECOGNITION_USAGE,\n requestId,\n recognitionUsage: { audioDuration: this.speechDuration },\n });\n this.speechDuration = 0;\n }\n\n this.queue.put({\n type: SpeechEventType.FINAL_TRANSCRIPT,\n requestId,\n alternatives: [speechData],\n });\n\n if (this.speaking) {\n this.speaking = false;\n this.queue.put({ type: SpeechEventType.END_OF_SPEECH });\n }\n } else {\n this.queue.put({\n type: SpeechEventType.INTERIM_TRANSCRIPT,\n requestId,\n alternatives: [speechData],\n });\n }\n }\n}\n"],"mappings":"AAGA,eAAgC;AAEhC,SAAS,UAAU,sBAAsB;AACzC,SAAS,uBAAuB;AAChC,SAAS,WAAW;AACpB,SAAS,2BAA2B;AACpC;AAAA,EACE,OAAO;AAAA,EACP,gBAAgB;AAAA,EAGhB;AAAA,OACK;AACP,SAAiC,mCAAmC;AACpE,SAA2B,OAAO,MAAM,eAAe,WAAW,oBAAoB;AACtF,SAAyB,WAAW,yBAAyB;AAuE7D,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,yBAAyB;AAgBxB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAqC,oBAAI,IAAI;AAAA,EAErD,UAAU,IAAI;AAAA,EAEd,YAAY,MAST;AACD,UAAM,EAAE,WAAW,MAAM,gBAAgB,KAAK,CAAC;AAE/C,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX,aAAa;AAAA,MACb;AAAA,MACA;AAAA,MACA,eAAe,CAAC;AAAA,IAClB,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,WAAW,QAAQ,IAAI,yBAAyB;AAClE,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAEA,SAAK,OAAO;AAAA,MACV;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,QAAI,YAAY,SAAS,GAAG,GAAG;AAC7B,YAAM,CAAC,OAAO,QAAQ,IAAI,YAAY,MAAM,GAAG;AAC/C,aAAO,IAAI,IAAI,EAAE,OAAO,SAAS,CAAC;AAAA,IACpC;AACA,WAAO,IAAI,IAAI,EAAE,OAAO,YAAY,CAAC;AAAA,EACvC;AAAA,EAEA,MAAgB,WAAW,GAAsC;AAC/D,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAEpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,OAAO,SAGkB;AACvB,UAAM,EAAE,UAAU,cAAc,4BAA4B,IAAI,WAAW,CAAC;AAC5E,UAAM,aAAa;AAAA,MACjB,GAAG,KAAK;AAAA,MACR,UAAU,YAAY,KAAK,KAAK;AAAA,IAClC;AAEA,UAAM,SAAS,IAAI,aAAa,MAAM,YAAY,WAAW;AAC7D,SAAK,QAAQ,IAAI,MAAM;AAEvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AACnD,UAAM,SAAS;AAAA,MACb,UAAU;AAAA,QACR,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,QACxC,UAAU,KAAK,KAAK;AAAA,QACpB,OAAO,KAAK,KAAK;AAAA,MACnB;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,SAAS,KAAK,KAAK,UAAU,QAAQ;AACjD,aAAO,QAAQ,KAAK,KAAK;AAAA,IAC3B;AAEA,QAAI,KAAK,KAAK,UAAU;AACtB,MAAC,OAAO,SAAqC,WAAW,KAAK,KAAK;AAAA,IACpE;AAEA,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,MAAM,kBAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS,MAAM,UAAU,KAAK,SAAS,OAAO;AACpD,UAAM,MAAM,EAAE,GAAG,QAAQ,MAAM,iBAAiB;AAChD,WAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAE/B,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAA+C,iBAAiB;AAAA,EACnE;AAAA,EACA,YAAY,UAAU,cAAc;AAAA,EACpC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,iBAAiB,IAAI,MAAM;AAAA,EAC3B;AAAA,EACA;AAAA,EAER,UAAU,IAAI;AAAA,EAEd,YACE,SACA,MACA,aACA;AACA,UAAM,SAAS,KAAK,YAAY,WAAW;AAC3C,SAAK,OAAO;AACZ,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAA8E;AAC1F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,SAAK,eAAe,IAAI;AAAA,EAC1B;AAAA,EAEA,MAAgB,MAAqB;AACnC,WAAO,MAAM;AAEX,UAAI,KAAuB;AAC3B,UAAI,UAAU;AACd,UAAI,gBAAgB;AAGpB,YAAM,eAAe,oBAAoC;AAEzD,YAAM,kBAAkB,MAAM;AAC5B,YAAI,QAAS;AACb,kBAAU;AACV,qBAAa,MAAM;AACnB,iCAAI;AACJ,iCAAI;AAAA,MACN;AAEA,YAAM,mBAAmB,OAAOA,KAAe,WAAwB;AACrE,eAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,gBAAM,UAAU,MAAM;AACpB,4BAAgB;AAChB,mBAAO,IAAI,MAAM,8BAA8B,CAAC;AAAA,UAClD;AAEA,iBAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAExD,UAAAA,IAAG,GAAG,WAAW,CAAC,SAAS;AACzB,kBAAM,OAAO,KAAK,MAAM,KAAK,SAAS,CAAC;AACvC,yBAAa,MAAM,IAAI;AAAA,UACzB,CAAC;AAED,UAAAA,IAAG,GAAG,SAAS,CAAC,MAAM;AACpB,iBAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,4BAAgB;AAChB,mBAAO,CAAC;AAAA,UACV,CAAC;AAED,UAAAA,IAAG,GAAG,SAAS,CAAC,SAAiB;AAC/B,4BAAgB;AAEhB,gBAAI,CAAC,QAAS,QAAO,KAAK,QAAQ,MAAM,+BAA+B;AACvE,gBAAI,cAAe,QAAO,QAAQ;AAElC;AAAA,cACE,IAAI,eAAe;AAAA,gBACjB,SAAS;AAAA,gBACT,SAAS,EAAE,YAAY,KAAK;AAAA,cAC9B,CAAC;AAAA,YACH;AAAA,UACF,CAAC;AAAA,QACH,CAAC;AAAA,MACH;AAEA,YAAM,OAAO,OAAO,QAAmB,WAAwB;AAC7D,cAAM,cAAc,IAAI;AAAA,UACtB,KAAK,KAAK;AAAA,UACV;AAAA,UACA,KAAK,MAAM,KAAK,KAAK,aAAa,EAAE;AAAA;AAAA,QACtC;AAGA,cAAM,eAAe,IAAI,QAAe,CAAC,GAAG,WAAW;AACrD,cAAI,OAAO,SAAS;AAClB,mBAAO,OAAO,IAAI,MAAM,cAAc,CAAC;AAAA,UACzC;AACA,gBAAM,UAAU,MAAM,OAAO,IAAI,MAAM,cAAc,CAAC;AACtD,iBAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAAA,QAC1D,CAAC;AAGD,cAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAClD,YAAI;AACF,iBAAO,MAAM;AACX,kBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,SAAS,KAAK,GAAG,YAAY,CAAC;AAEjE,gBAAI,OAAO,KAAM;AACjB,kBAAM,KAAK,OAAO;AAElB,gBAAI;AACJ,gBAAI,OAAO,aAAa,gBAAgB;AACtC,uBAAS,YAAY,MAAM;AAAA,YAC7B,OAAO;AACL,oBAAM,QAAQ;AACd,uBAAS,YAAY,MAAM,IAAI,WAAW,MAAM,IAAI,EAAE,MAAM;AAAA,YAC9D;AAEA,uBAAW,SAAS,QAAQ;AAC1B,mBAAK,kBAAkB,MAAM,oBAAoB,MAAM;AACvD,oBAAM,SAAS,OAAO,KAAK,MAAM,KAAK,MAAM,EAAE,SAAS,QAAQ;AAC/D,oBAAM,MAAM,EAAE,MAAM,eAAe,OAAO,OAAO;AACjD,qBAAO,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,YACjC;AAAA,UACF;AAEA,oBAAU;AACV,iBAAO,KAAK,KAAK,UAAU,EAAE,MAAM,mBAAmB,CAAC,CAAC;AAAA,QAC1D,SAAS,GAAG;AACV,cAAK,EAAY,YAAY,gBAAgB;AAE3C;AAAA,UACF;AACA,gBAAM;AAAA,QACR;AAAA,MACF;AAEA,YAAM,OAAO,OAAO,WAAwB;AAC1C,cAAM,oBAAoB,aAAa,OAAO;AAC9C,cAAM,SAAS,kBAAkB,UAAU;AAE3C,YAAI;AACF,iBAAO,CAAC,KAAK,UAAU,CAAC,OAAO,SAAS;AACtC,kBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,gBAAI,OAAO,QAAS;AACpB,gBAAI,OAAO,KAAM;AAEjB,kBAAM,OAAO,OAAO;AACpB,kBAAM,OAAO,KAAK;AAElB,oBAAQ,MAAM;AAAA,cACZ,KAAK;AAAA,cACL,KAAK;AACH;AAAA,cACF,KAAK;AACH,gCAAgB;AAChB,gCAAgB;AAChB;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,MAAM,KAAK;AAClC;AAAA,cACF,KAAK;AACH,qBAAK,kBAAkB,MAAM,IAAI;AACjC;AAAA,cACF,KAAK;AACH,qBAAK,QAAQ,MAAM,EAAE,OAAO,KAAK,GAAG,iCAAiC;AACrE,gCAAgB;AAChB,sBAAM,IAAI,SAAS,+BAA+B,KAAK,UAAU,IAAI,CAAC,EAAE;AAAA,cAC1E;AACE,qBAAK,QAAQ;AAAA,kBACX,EAAE,SAAS,KAAK;AAAA,kBAChB;AAAA,gBACF;AACA;AAAA,YACJ;AAAA,UACF;AAAA,QACF,UAAE;AACA,iBAAO,YAAY;AACnB,cAAI;AACF,kBAAM,kBAAkB,OAAO;AAAA,UACjC,SAAS,GAAG;AACV,iBAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,UACxF;AAAA,QACF;AAAA,MACF;AAEA,UAAI;AACF,aAAK,MAAM,KAAK,IAAI,UAAU,KAAK,YAAY,SAAS;AAGxD,cAAM,aAAa,IAAI,gBAAgB;AACvC,cAAM,WAAW,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,IAAK,MAAM,GAAG,UAAU;AACxE,cAAM,iBAAiB,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,iBAAiB,IAAK,MAAM,GAAG,UAAU;AAC1F,cAAM,WAAW,KAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,MAAM,GAAG,UAAU;AACnE,cAAM,oBAAoB,KAAK;AAAA,UAC7B,CAAC,EAAE,OAAO,MAAM,QAAQ,KAAK,CAAC,KAAK,eAAe,KAAK,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,UAC/E;AAAA,QACF;AAEA,YAAI;AACF,gBAAM,QAAQ,KAAK;AAAA,YACjB,QAAQ,IAAI,CAAC,SAAS,QAAQ,eAAe,QAAQ,SAAS,MAAM,CAAC;AAAA,YACrE,kBAAkB;AAAA,UACpB,CAAC;AAGD,cAAI,CAAC,kBAAkB,KAAM;AAG7B,eAAK,eAAe,MAAM;AAAA,QAC5B,UAAE;AAEA,gBAAM;AAAA,YACJ,CAAC,UAAU,gBAAgB,UAAU,iBAAiB;AAAA,YACtD;AAAA,UACF;AACA,0BAAgB;AAAA,QAClB;AAAA,MACF,UAAE;AAEA,wBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,kBAAkB,MAA2B,SAAkB;AACrE,UAAM,YAAY,KAAK,cAAc,KAAK;AAC1C,UAAM,OAAO,KAAK,cAAc;AAChC,UAAM,WAAW,KAAK,YAAY,KAAK,KAAK,YAAY;AAExD,QAAI,CAAC,QAAQ,CAAC,QAAS;AAGvB,QAAI,CAAC,KAAK,UAAU;AAClB,WAAK,WAAW;AAChB,WAAK,MAAM,IAAI,EAAE,MAAM,gBAAgB,gBAAgB,CAAC;AAAA,IAC1D;AAEA,UAAM,aAAyB;AAAA,MAC7B;AAAA,MACA,WAAW,KAAK,SAAS;AAAA,MACzB,SAAS,KAAK,YAAY;AAAA,MAC1B,YAAY,KAAK,cAAc;AAAA,MAC/B;AAAA,IACF;AAEA,QAAI,SAAS;AACX,UAAI,KAAK,iBAAiB,GAAG;AAC3B,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,gBAAgB;AAAA,UACtB;AAAA,UACA,kBAAkB,EAAE,eAAe,KAAK,eAAe;AAAA,QACzD,CAAC;AACD,aAAK,iBAAiB;AAAA,MACxB;AAEA,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,gBAAgB;AAAA,QACtB;AAAA,QACA,cAAc,CAAC,UAAU;AAAA,MAC3B,CAAC;AAED,UAAI,KAAK,UAAU;AACjB,aAAK,WAAW;AAChB,aAAK,MAAM,IAAI,EAAE,MAAM,gBAAgB,cAAc,CAAC;AAAA,MACxD;AAAA,IACF,OAAO;AACL,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,gBAAgB;AAAA,QACtB;AAAA,QACA,cAAc,CAAC,UAAU;AAAA,MAC3B,CAAC;AAAA,IACH;AAAA,EACF;AACF;","names":["ws"]}
|
package/dist/tts/tts.cjs
CHANGED
|
@@ -124,7 +124,16 @@ class SynthesizeStream {
|
|
|
124
124
|
recoverable
|
|
125
125
|
});
|
|
126
126
|
}
|
|
127
|
-
//
|
|
127
|
+
// NOTE(AJS-37): The implementation below uses an AsyncIterableQueue (`this.input`)
|
|
128
|
+
// bridged from a DeferredReadableStream (`this.deferredInputStream`) rather than
|
|
129
|
+
// consuming the stream directly.
|
|
130
|
+
//
|
|
131
|
+
// A full refactor to native Web Streams was considered but is currently deferred.
|
|
132
|
+
// The primary reason is to maintain architectural parity with the Python SDK,
|
|
133
|
+
// which is a key design goal for the project. This ensures a consistent developer
|
|
134
|
+
// experience across both platforms.
|
|
135
|
+
//
|
|
136
|
+
// For more context, see the discussion in GitHub issue # 844.
|
|
128
137
|
async pumpInput() {
|
|
129
138
|
const reader = this.deferredInputStream.stream.getReader();
|
|
130
139
|
try {
|
package/dist/tts/tts.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/tts/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { log } from '../log.js';\nimport type { TTSMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { AsyncIterableQueue, delay, mergeFrames, startSoon, toError } from '../utils.js';\n\n/** SynthesizedAudio is a packet of speech synthesis as returned by the TTS. */\nexport interface SynthesizedAudio {\n /** Request ID (one segment could be made up of multiple requests) */\n requestId: string;\n /** Segment ID, each segment is separated by a flush */\n segmentId: string;\n /** Synthesized audio frame */\n frame: AudioFrame;\n /** Current segment of the synthesized audio */\n deltaText?: string;\n /** Whether this is the last frame of the segment (streaming only) */\n final: boolean;\n}\n\n/**\n * Describes the capabilities of the TTS provider.\n *\n * @remarks\n * At present, only `streaming` is supplied to this interface, and the framework only supports\n * providers that do have a streaming endpoint.\n */\nexport interface TTSCapabilities {\n streaming: boolean;\n}\n\nexport interface TTSError {\n type: 'tts_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type TTSCallbacks = {\n ['metrics_collected']: (metrics: TTSMetrics) => void;\n ['error']: (error: TTSError) => void;\n};\n\n/**\n * An instance of a text-to-speech adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child TTS class, which inherits this class's methods.\n */\nexport abstract class TTS extends (EventEmitter as new () => TypedEmitter<TTSCallbacks>) {\n #capabilities: TTSCapabilities;\n #sampleRate: number;\n #numChannels: number;\n abstract label: string;\n\n constructor(sampleRate: number, numChannels: number, capabilities: TTSCapabilities) {\n super();\n this.#capabilities = capabilities;\n this.#sampleRate = sampleRate;\n this.#numChannels = numChannels;\n }\n\n /** Returns this TTS's capabilities */\n get capabilities(): TTSCapabilities {\n return this.#capabilities;\n }\n\n /** Returns the sample rate of audio frames returned by this TTS */\n get sampleRate(): number {\n return this.#sampleRate;\n }\n\n /** Returns the channel count of audio frames returned by this TTS */\n get numChannels(): number {\n return this.#numChannels;\n }\n\n /**\n * Receives text and returns synthesis in the form of a {@link ChunkedStream}\n */\n abstract synthesize(text: string): ChunkedStream;\n\n /**\n * Returns a {@link SynthesizeStream} that can be used to push text and receive audio data\n */\n abstract stream(): SynthesizeStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a text-to-speech stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * await source.captureFrame(event.frame);\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SynthesizeStream class, which inherits this class's methods.\n */\nexport abstract class SynthesizeStream\n implements AsyncIterableIterator<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>\n{\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n static readonly END_OF_STREAM = Symbol('END_OF_STREAM');\n protected input = new AsyncIterableQueue<string | typeof SynthesizeStream.FLUSH_SENTINEL>();\n protected queue = new AsyncIterableQueue<\n SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n >();\n protected output = new AsyncIterableQueue<\n SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n >();\n protected closed = false;\n abstract label: string;\n #tts: TTS;\n #metricsPendingTexts: string[] = [];\n #metricsText = '';\n #monitorMetricsTask?: Promise<void>;\n private _connOptions: APIConnectOptions;\n protected abortController = new AbortController();\n\n private deferredInputStream: DeferredReadableStream<\n string | typeof SynthesizeStream.FLUSH_SENTINEL\n >;\n private logger = log();\n\n constructor(tts: TTS, connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS) {\n this.#tts = tts;\n this._connOptions = connOptions;\n this.deferredInputStream = new DeferredReadableStream();\n this.pumpInput();\n this.abortController.signal.addEventListener('abort', () => {\n this.deferredInputStream.detachSource();\n // TODO (AJS-36) clean this up when we refactor with streams\n this.input.close();\n this.output.close();\n this.closed = true;\n });\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n startSoon(() => this.mainTask().then(() => this.queue.close()));\n }\n\n private async mainTask() {\n // TODO(brian): PR3 - Add span wrapping: tracer.startActiveSpan('tts_request', ..., { endOnExit: false })\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n // TODO(brian): PR3 - Add span for retry attempts: tracer.startActiveSpan('tts_request_run', ...)\n return await this.run();\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = this._connOptions._intervalForRetry(i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to generate TTS completion after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#tts.label, attempt: i + 1, error },\n `failed to synthesize speech, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#tts.emit('error', {\n type: 'tts_error',\n timestamp: Date.now(),\n label: this.#tts.label,\n error,\n recoverable,\n });\n }\n\n // TODO(AJS-37) Remove when refactoring TTS to use streams\n protected async pumpInput() {\n const reader = this.deferredInputStream.stream.getReader();\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done || value === SynthesizeStream.FLUSH_SENTINEL) {\n break;\n }\n this.pushText(value);\n }\n this.endInput();\n } catch (error) {\n this.logger.error(error, 'Error reading deferred input stream');\n } finally {\n reader.releaseLock();\n // Ensure output is closed when the stream ends\n if (!this.#monitorMetricsTask) {\n // No text was received, close the output directly\n this.output.close();\n }\n }\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let audioDurationMs = 0;\n let ttfb: bigint = BigInt(-1);\n let requestId = '';\n\n const emit = () => {\n if (this.#metricsPendingTexts.length) {\n const text = this.#metricsPendingTexts.shift()!;\n const duration = process.hrtime.bigint() - startTime;\n const roundedAudioDurationMs = Math.round(audioDurationMs);\n const metrics: TTSMetrics = {\n type: 'tts_metrics',\n timestamp: Date.now(),\n requestId,\n ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n charactersCount: text.length,\n audioDurationMs: roundedAudioDurationMs,\n cancelled: this.abortController.signal.aborted,\n label: this.#tts.label,\n streamed: false,\n };\n this.#tts.emit('metrics_collected', metrics);\n }\n };\n\n for await (const audio of this.queue) {\n if (this.abortController.signal.aborted) {\n break;\n }\n this.output.put(audio);\n if (audio === SynthesizeStream.END_OF_STREAM) continue;\n requestId = audio.requestId;\n if (ttfb === BigInt(-1)) {\n ttfb = process.hrtime.bigint() - startTime;\n }\n // TODO(AJS-102): use frame.durationMs once available in rtc-node\n audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n if (audio.final) {\n emit();\n }\n }\n\n if (requestId) {\n emit();\n }\n }\n\n protected abstract run(): Promise<void>;\n\n updateInputStream(text: ReadableStream<string>) {\n this.deferredInputStream.setSource(text);\n }\n\n /** Push a string of text to the TTS */\n /** @deprecated Use `updateInputStream` instead */\n pushText(text: string) {\n if (!this.#monitorMetricsTask) {\n this.#monitorMetricsTask = this.monitorMetrics();\n // Close output when metrics task completes\n this.#monitorMetricsTask.finally(() => this.output.close());\n }\n this.#metricsText += text;\n\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.put(text);\n }\n\n /** Flush the TTS, causing it to process all pending text */\n flush() {\n if (this.#metricsText) {\n this.#metricsPendingTexts.push(this.#metricsText);\n this.#metricsText = '';\n }\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.put(SynthesizeStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n this.flush();\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the TTS stream */\n close() {\n this.abortController.abort();\n }\n\n [Symbol.asyncIterator](): SynthesizeStream {\n return this;\n }\n}\n\n/**\n * An instance of a text-to-speech response, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * await source.captureFrame(event.frame);\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child ChunkedStream class, which inherits this class's methods.\n */\nexport abstract class ChunkedStream implements AsyncIterableIterator<SynthesizedAudio> {\n protected queue = new AsyncIterableQueue<SynthesizedAudio>();\n protected output = new AsyncIterableQueue<SynthesizedAudio>();\n protected closed = false;\n abstract label: string;\n #text: string;\n #tts: TTS;\n private _connOptions: APIConnectOptions;\n private logger = log();\n\n constructor(\n text: string,\n tts: TTS,\n connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n ) {\n this.#text = text;\n this.#tts = tts;\n this._connOptions = connOptions;\n\n this.monitorMetrics();\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n Promise.resolve().then(() => this.mainTask().then(() => this.queue.close()));\n }\n\n private async mainTask() {\n // TODO(brian): PR3 - Add span wrapping: tracer.startActiveSpan('tts_request', ..., { endOnExit: false })\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n // TODO(brian): PR3 - Add span for retry attempts: tracer.startActiveSpan('tts_request_run', ...)\n return await this.run();\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = this._connOptions._intervalForRetry(i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to generate TTS completion after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#tts.label, attempt: i + 1, error },\n `failed to generate TTS completion, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#tts.emit('error', {\n type: 'tts_error',\n timestamp: Date.now(),\n label: this.#tts.label,\n error,\n recoverable,\n });\n }\n\n protected abstract run(): Promise<void>;\n\n get inputText(): string {\n return this.#text;\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let audioDurationMs = 0;\n let ttfb: bigint = BigInt(-1);\n let requestId = '';\n\n for await (const audio of this.queue) {\n this.output.put(audio);\n requestId = audio.requestId;\n if (ttfb === BigInt(-1)) {\n ttfb = process.hrtime.bigint() - startTime;\n }\n audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n }\n this.output.close();\n\n const duration = process.hrtime.bigint() - startTime;\n const metrics: TTSMetrics = {\n type: 'tts_metrics',\n timestamp: Date.now(),\n requestId,\n ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n charactersCount: this.#text.length,\n audioDurationMs: Math.round(audioDurationMs),\n cancelled: false, // TODO(AJS-186): support ChunkedStream with 1.0 - add this.abortController.signal.aborted here\n label: this.#tts.label,\n streamed: false,\n };\n this.#tts.emit('metrics_collected', metrics);\n }\n\n /** Collect every frame into one in a single call */\n async collect(): Promise<AudioFrame> {\n const frames = [];\n for await (const event of this) {\n frames.push(event.frame);\n }\n return mergeFrames(frames);\n }\n\n next(): Promise<IteratorResult<SynthesizedAudio>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the TTS stream */\n close() {\n this.queue.close();\n this.output.close();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): ChunkedStream {\n return this;\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAKA,yBAA6B;AAE7B,wBAA6C;AAC7C,iBAAoB;AAEpB,6BAAuC;AACvC,mBAAoE;AACpE,mBAA2E;AA+CpE,MAAe,YAAa,gCAAsD;AAAA,EACvF;AAAA,EACA;AAAA,EACA;AAAA,EAGA,YAAY,YAAoB,aAAqB,cAA+B;AAClF,UAAM;AACN,SAAK,gBAAgB;AACrB,SAAK,cAAc;AACnB,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,aAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,cAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAYA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAgBO,MAAe,iBAEtB;AAAA,EACE,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EAClE,OAAgB,gBAAgB,OAAO,eAAe;AAAA,EAC5C,QAAQ,IAAI,gCAAoE;AAAA,EAChF,QAAQ,IAAI,gCAEpB;AAAA,EACQ,SAAS,IAAI,gCAErB;AAAA,EACQ,SAAS;AAAA,EAEnB;AAAA,EACA,uBAAiC,CAAC;AAAA,EAClC,eAAe;AAAA,EACf;AAAA,EACQ;AAAA,EACE,kBAAkB,IAAI,gBAAgB;AAAA,EAExC;AAAA,EAGA,aAAS,gBAAI;AAAA,EAErB,YAAY,KAAU,cAAiC,0CAA6B;AAClF,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,8CAAuB;AACtD,SAAK,UAAU;AACf,SAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,WAAK,oBAAoB,aAAa;AAEtC,WAAK,MAAM,MAAM;AACjB,WAAK,OAAO,MAAM;AAClB,WAAK,SAAS;AAAA,IAChB,CAAC;AAMD,gCAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,WAAW;AAEvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AAEF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,4BAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,qCAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,6CAA6C,aAAa;AAAA,YAC5D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,sBAAM,oBAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,WAAO,sBAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,MAAgB,YAAY;AAC1B,UAAM,SAAS,KAAK,oBAAoB,OAAO,UAAU;AACzD,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,QAAQ,UAAU,iBAAiB,gBAAgB;AACrD;AAAA,QACF;AACA,aAAK,SAAS,KAAK;AAAA,MACrB;AACA,WAAK,SAAS;AAAA,IAChB,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,OAAO,qCAAqC;AAAA,IAChE,UAAE;AACA,aAAO,YAAY;AAEnB,UAAI,CAAC,KAAK,qBAAqB;AAE7B,aAAK,OAAO,MAAM;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,UAAM,OAAO,MAAM;AACjB,UAAI,KAAK,qBAAqB,QAAQ;AACpC,cAAM,OAAO,KAAK,qBAAqB,MAAM;AAC7C,cAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,cAAM,yBAAyB,KAAK,MAAM,eAAe;AACzD,cAAM,UAAsB;AAAA,UAC1B,MAAM;AAAA,UACN,WAAW,KAAK,IAAI;AAAA,UACpB;AAAA,UACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,UAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,UACzD,iBAAiB,KAAK;AAAA,UACtB,iBAAiB;AAAA,UACjB,WAAW,KAAK,gBAAgB,OAAO;AAAA,UACvC,OAAO,KAAK,KAAK;AAAA,UACjB,UAAU;AAAA,QACZ;AACA,aAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,MAC7C;AAAA,IACF;AAEA,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,KAAK,gBAAgB,OAAO,SAAS;AACvC;AAAA,MACF;AACA,WAAK,OAAO,IAAI,KAAK;AACrB,UAAI,UAAU,iBAAiB,cAAe;AAC9C,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AAEA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAC9E,UAAI,MAAM,OAAO;AACf,aAAK;AAAA,MACP;AAAA,IACF;AAEA,QAAI,WAAW;AACb,WAAK;AAAA,IACP;AAAA,EACF;AAAA,EAIA,kBAAkB,MAA8B;AAC9C,SAAK,oBAAoB,UAAU,IAAI;AAAA,EACzC;AAAA;AAAA;AAAA,EAIA,SAAS,MAAc;AACrB,QAAI,CAAC,KAAK,qBAAqB;AAC7B,WAAK,sBAAsB,KAAK,eAAe;AAE/C,WAAK,oBAAoB,QAAQ,MAAM,KAAK,OAAO,MAAM,CAAC;AAAA,IAC5D;AACA,SAAK,gBAAgB;AAErB,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,IAAI;AAAA,EACrB;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,cAAc;AACrB,WAAK,qBAAqB,KAAK,KAAK,YAAY;AAChD,WAAK,eAAe;AAAA,IACtB;AACA,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,iBAAiB,cAAc;AAAA,EAChD;AAAA;AAAA,EAGA,WAAW;AACT,SAAK,MAAM;AACX,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA0F;AACxF,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA,EAEA,CAAC,OAAO,aAAa,IAAsB;AACzC,WAAO;AAAA,EACT;AACF;AAgBO,MAAe,cAAiE;AAAA,EAC3E,QAAQ,IAAI,gCAAqC;AAAA,EACjD,SAAS,IAAI,gCAAqC;AAAA,EAClD,SAAS;AAAA,EAEnB;AAAA,EACA;AAAA,EACQ;AAAA,EACA,aAAS,gBAAI;AAAA,EAErB,YACE,MACA,KACA,cAAiC,0CACjC;AACA,SAAK,QAAQ;AACb,SAAK,OAAO;AACZ,SAAK,eAAe;AAEpB,SAAK,eAAe;AAMpB,YAAQ,QAAQ,EAAE,KAAK,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAC7E;AAAA,EAEA,MAAc,WAAW;AAEvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AAEF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,4BAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,qCAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,kDAAkD,aAAa;AAAA,YACjE;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,sBAAM,oBAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,WAAO,sBAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAIA,IAAI,YAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,qBAAiB,SAAS,KAAK,OAAO;AACpC,WAAK,OAAO,IAAI,KAAK;AACrB,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AACA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAAA,IAChF;AACA,SAAK,OAAO,MAAM;AAElB,UAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,UAAM,UAAsB;AAAA,MAC1B,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,MAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,MACzD,iBAAiB,KAAK,MAAM;AAAA,MAC5B,iBAAiB,KAAK,MAAM,eAAe;AAAA,MAC3C,WAAW;AAAA;AAAA,MACX,OAAO,KAAK,KAAK;AAAA,MACjB,UAAU;AAAA,IACZ;AACA,SAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,EAC7C;AAAA;AAAA,EAGA,MAAM,UAA+B;AACnC,UAAM,SAAS,CAAC;AAChB,qBAAiB,SAAS,MAAM;AAC9B,aAAO,KAAK,MAAM,KAAK;AAAA,IACzB;AACA,eAAO,0BAAY,MAAM;AAAA,EAC3B;AAAA,EAEA,OAAkD;AAChD,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,MAAM,MAAM;AACjB,SAAK,OAAO,MAAM;AAClB,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAmB;AACtC,WAAO;AAAA,EACT;AACF;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../../src/tts/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { log } from '../log.js';\nimport type { TTSMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { AsyncIterableQueue, delay, mergeFrames, startSoon, toError } from '../utils.js';\n\n/** SynthesizedAudio is a packet of speech synthesis as returned by the TTS. */\nexport interface SynthesizedAudio {\n /** Request ID (one segment could be made up of multiple requests) */\n requestId: string;\n /** Segment ID, each segment is separated by a flush */\n segmentId: string;\n /** Synthesized audio frame */\n frame: AudioFrame;\n /** Current segment of the synthesized audio */\n deltaText?: string;\n /** Whether this is the last frame of the segment (streaming only) */\n final: boolean;\n}\n\n/**\n * Describes the capabilities of the TTS provider.\n *\n * @remarks\n * At present, only `streaming` is supplied to this interface, and the framework only supports\n * providers that do have a streaming endpoint.\n */\nexport interface TTSCapabilities {\n streaming: boolean;\n}\n\nexport interface TTSError {\n type: 'tts_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type TTSCallbacks = {\n ['metrics_collected']: (metrics: TTSMetrics) => void;\n ['error']: (error: TTSError) => void;\n};\n\n/**\n * An instance of a text-to-speech adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child TTS class, which inherits this class's methods.\n */\nexport abstract class TTS extends (EventEmitter as new () => TypedEmitter<TTSCallbacks>) {\n #capabilities: TTSCapabilities;\n #sampleRate: number;\n #numChannels: number;\n abstract label: string;\n\n constructor(sampleRate: number, numChannels: number, capabilities: TTSCapabilities) {\n super();\n this.#capabilities = capabilities;\n this.#sampleRate = sampleRate;\n this.#numChannels = numChannels;\n }\n\n /** Returns this TTS's capabilities */\n get capabilities(): TTSCapabilities {\n return this.#capabilities;\n }\n\n /** Returns the sample rate of audio frames returned by this TTS */\n get sampleRate(): number {\n return this.#sampleRate;\n }\n\n /** Returns the channel count of audio frames returned by this TTS */\n get numChannels(): number {\n return this.#numChannels;\n }\n\n /**\n * Receives text and returns synthesis in the form of a {@link ChunkedStream}\n */\n abstract synthesize(text: string): ChunkedStream;\n\n /**\n * Returns a {@link SynthesizeStream} that can be used to push text and receive audio data\n */\n abstract stream(): SynthesizeStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a text-to-speech stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * await source.captureFrame(event.frame);\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SynthesizeStream class, which inherits this class's methods.\n */\nexport abstract class SynthesizeStream\n implements AsyncIterableIterator<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>\n{\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n static readonly END_OF_STREAM = Symbol('END_OF_STREAM');\n protected input = new AsyncIterableQueue<string | typeof SynthesizeStream.FLUSH_SENTINEL>();\n protected queue = new AsyncIterableQueue<\n SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n >();\n protected output = new AsyncIterableQueue<\n SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n >();\n protected closed = false;\n abstract label: string;\n #tts: TTS;\n #metricsPendingTexts: string[] = [];\n #metricsText = '';\n #monitorMetricsTask?: Promise<void>;\n private _connOptions: APIConnectOptions;\n protected abortController = new AbortController();\n\n private deferredInputStream: DeferredReadableStream<\n string | typeof SynthesizeStream.FLUSH_SENTINEL\n >;\n private logger = log();\n\n constructor(tts: TTS, connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS) {\n this.#tts = tts;\n this._connOptions = connOptions;\n this.deferredInputStream = new DeferredReadableStream();\n this.pumpInput();\n this.abortController.signal.addEventListener('abort', () => {\n this.deferredInputStream.detachSource();\n // TODO (AJS-36) clean this up when we refactor with streams\n this.input.close();\n this.output.close();\n this.closed = true;\n });\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n startSoon(() => this.mainTask().then(() => this.queue.close()));\n }\n\n private async mainTask() {\n // TODO(brian): PR3 - Add span wrapping: tracer.startActiveSpan('tts_request', ..., { endOnExit: false })\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n // TODO(brian): PR3 - Add span for retry attempts: tracer.startActiveSpan('tts_request_run', ...)\n return await this.run();\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = this._connOptions._intervalForRetry(i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to generate TTS completion after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#tts.label, attempt: i + 1, error },\n `failed to synthesize speech, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#tts.emit('error', {\n type: 'tts_error',\n timestamp: Date.now(),\n label: this.#tts.label,\n error,\n recoverable,\n });\n }\n\n // NOTE(AJS-37): The implementation below uses an AsyncIterableQueue (`this.input`)\n // bridged from a DeferredReadableStream (`this.deferredInputStream`) rather than\n // consuming the stream directly.\n //\n // A full refactor to native Web Streams was considered but is currently deferred.\n // The primary reason is to maintain architectural parity with the Python SDK,\n // which is a key design goal for the project. This ensures a consistent developer\n // experience across both platforms.\n //\n // For more context, see the discussion in GitHub issue # 844.\n protected async pumpInput() {\n const reader = this.deferredInputStream.stream.getReader();\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done || value === SynthesizeStream.FLUSH_SENTINEL) {\n break;\n }\n this.pushText(value);\n }\n this.endInput();\n } catch (error) {\n this.logger.error(error, 'Error reading deferred input stream');\n } finally {\n reader.releaseLock();\n // Ensure output is closed when the stream ends\n if (!this.#monitorMetricsTask) {\n // No text was received, close the output directly\n this.output.close();\n }\n }\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let audioDurationMs = 0;\n let ttfb: bigint = BigInt(-1);\n let requestId = '';\n\n const emit = () => {\n if (this.#metricsPendingTexts.length) {\n const text = this.#metricsPendingTexts.shift()!;\n const duration = process.hrtime.bigint() - startTime;\n const roundedAudioDurationMs = Math.round(audioDurationMs);\n const metrics: TTSMetrics = {\n type: 'tts_metrics',\n timestamp: Date.now(),\n requestId,\n ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n charactersCount: text.length,\n audioDurationMs: roundedAudioDurationMs,\n cancelled: this.abortController.signal.aborted,\n label: this.#tts.label,\n streamed: false,\n };\n this.#tts.emit('metrics_collected', metrics);\n }\n };\n\n for await (const audio of this.queue) {\n if (this.abortController.signal.aborted) {\n break;\n }\n this.output.put(audio);\n if (audio === SynthesizeStream.END_OF_STREAM) continue;\n requestId = audio.requestId;\n if (ttfb === BigInt(-1)) {\n ttfb = process.hrtime.bigint() - startTime;\n }\n // TODO(AJS-102): use frame.durationMs once available in rtc-node\n audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n if (audio.final) {\n emit();\n }\n }\n\n if (requestId) {\n emit();\n }\n }\n\n protected abstract run(): Promise<void>;\n\n updateInputStream(text: ReadableStream<string>) {\n this.deferredInputStream.setSource(text);\n }\n\n /** Push a string of text to the TTS */\n /** @deprecated Use `updateInputStream` instead */\n pushText(text: string) {\n if (!this.#monitorMetricsTask) {\n this.#monitorMetricsTask = this.monitorMetrics();\n // Close output when metrics task completes\n this.#monitorMetricsTask.finally(() => this.output.close());\n }\n this.#metricsText += text;\n\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.put(text);\n }\n\n /** Flush the TTS, causing it to process all pending text */\n flush() {\n if (this.#metricsText) {\n this.#metricsPendingTexts.push(this.#metricsText);\n this.#metricsText = '';\n }\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.put(SynthesizeStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n this.flush();\n if (this.input.closed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the TTS stream */\n close() {\n this.abortController.abort();\n }\n\n [Symbol.asyncIterator](): SynthesizeStream {\n return this;\n }\n}\n\n/**\n * An instance of a text-to-speech response, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * await source.captureFrame(event.frame);\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child ChunkedStream class, which inherits this class's methods.\n */\nexport abstract class ChunkedStream implements AsyncIterableIterator<SynthesizedAudio> {\n protected queue = new AsyncIterableQueue<SynthesizedAudio>();\n protected output = new AsyncIterableQueue<SynthesizedAudio>();\n protected closed = false;\n abstract label: string;\n #text: string;\n #tts: TTS;\n private _connOptions: APIConnectOptions;\n private logger = log();\n\n constructor(\n text: string,\n tts: TTS,\n connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n ) {\n this.#text = text;\n this.#tts = tts;\n this._connOptions = connOptions;\n\n this.monitorMetrics();\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n Promise.resolve().then(() => this.mainTask().then(() => this.queue.close()));\n }\n\n private async mainTask() {\n // TODO(brian): PR3 - Add span wrapping: tracer.startActiveSpan('tts_request', ..., { endOnExit: false })\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n // TODO(brian): PR3 - Add span for retry attempts: tracer.startActiveSpan('tts_request_run', ...)\n return await this.run();\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = this._connOptions._intervalForRetry(i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to generate TTS completion after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#tts.label, attempt: i + 1, error },\n `failed to generate TTS completion, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#tts.emit('error', {\n type: 'tts_error',\n timestamp: Date.now(),\n label: this.#tts.label,\n error,\n recoverable,\n });\n }\n\n protected abstract run(): Promise<void>;\n\n get inputText(): string {\n return this.#text;\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let audioDurationMs = 0;\n let ttfb: bigint = BigInt(-1);\n let requestId = '';\n\n for await (const audio of this.queue) {\n this.output.put(audio);\n requestId = audio.requestId;\n if (ttfb === BigInt(-1)) {\n ttfb = process.hrtime.bigint() - startTime;\n }\n audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n }\n this.output.close();\n\n const duration = process.hrtime.bigint() - startTime;\n const metrics: TTSMetrics = {\n type: 'tts_metrics',\n timestamp: Date.now(),\n requestId,\n ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n charactersCount: this.#text.length,\n audioDurationMs: Math.round(audioDurationMs),\n cancelled: false, // TODO(AJS-186): support ChunkedStream with 1.0 - add this.abortController.signal.aborted here\n label: this.#tts.label,\n streamed: false,\n };\n this.#tts.emit('metrics_collected', metrics);\n }\n\n /** Collect every frame into one in a single call */\n async collect(): Promise<AudioFrame> {\n const frames = [];\n for await (const event of this) {\n frames.push(event.frame);\n }\n return mergeFrames(frames);\n }\n\n next(): Promise<IteratorResult<SynthesizedAudio>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the TTS stream */\n close() {\n this.queue.close();\n this.output.close();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): ChunkedStream {\n return this;\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAKA,yBAA6B;AAE7B,wBAA6C;AAC7C,iBAAoB;AAEpB,6BAAuC;AACvC,mBAAoE;AACpE,mBAA2E;AA+CpE,MAAe,YAAa,gCAAsD;AAAA,EACvF;AAAA,EACA;AAAA,EACA;AAAA,EAGA,YAAY,YAAoB,aAAqB,cAA+B;AAClF,UAAM;AACN,SAAK,gBAAgB;AACrB,SAAK,cAAc;AACnB,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,aAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,cAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAYA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAgBO,MAAe,iBAEtB;AAAA,EACE,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EAClE,OAAgB,gBAAgB,OAAO,eAAe;AAAA,EAC5C,QAAQ,IAAI,gCAAoE;AAAA,EAChF,QAAQ,IAAI,gCAEpB;AAAA,EACQ,SAAS,IAAI,gCAErB;AAAA,EACQ,SAAS;AAAA,EAEnB;AAAA,EACA,uBAAiC,CAAC;AAAA,EAClC,eAAe;AAAA,EACf;AAAA,EACQ;AAAA,EACE,kBAAkB,IAAI,gBAAgB;AAAA,EAExC;AAAA,EAGA,aAAS,gBAAI;AAAA,EAErB,YAAY,KAAU,cAAiC,0CAA6B;AAClF,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,8CAAuB;AACtD,SAAK,UAAU;AACf,SAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,WAAK,oBAAoB,aAAa;AAEtC,WAAK,MAAM,MAAM;AACjB,WAAK,OAAO,MAAM;AAClB,WAAK,SAAS;AAAA,IAChB,CAAC;AAMD,gCAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,WAAW;AAEvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AAEF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,4BAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,qCAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,6CAA6C,aAAa;AAAA,YAC5D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,sBAAM,oBAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,WAAO,sBAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,MAAgB,YAAY;AAC1B,UAAM,SAAS,KAAK,oBAAoB,OAAO,UAAU;AACzD,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,QAAQ,UAAU,iBAAiB,gBAAgB;AACrD;AAAA,QACF;AACA,aAAK,SAAS,KAAK;AAAA,MACrB;AACA,WAAK,SAAS;AAAA,IAChB,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,OAAO,qCAAqC;AAAA,IAChE,UAAE;AACA,aAAO,YAAY;AAEnB,UAAI,CAAC,KAAK,qBAAqB;AAE7B,aAAK,OAAO,MAAM;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,UAAM,OAAO,MAAM;AACjB,UAAI,KAAK,qBAAqB,QAAQ;AACpC,cAAM,OAAO,KAAK,qBAAqB,MAAM;AAC7C,cAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,cAAM,yBAAyB,KAAK,MAAM,eAAe;AACzD,cAAM,UAAsB;AAAA,UAC1B,MAAM;AAAA,UACN,WAAW,KAAK,IAAI;AAAA,UACpB;AAAA,UACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,UAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,UACzD,iBAAiB,KAAK;AAAA,UACtB,iBAAiB;AAAA,UACjB,WAAW,KAAK,gBAAgB,OAAO;AAAA,UACvC,OAAO,KAAK,KAAK;AAAA,UACjB,UAAU;AAAA,QACZ;AACA,aAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,MAC7C;AAAA,IACF;AAEA,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,KAAK,gBAAgB,OAAO,SAAS;AACvC;AAAA,MACF;AACA,WAAK,OAAO,IAAI,KAAK;AACrB,UAAI,UAAU,iBAAiB,cAAe;AAC9C,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AAEA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAC9E,UAAI,MAAM,OAAO;AACf,aAAK;AAAA,MACP;AAAA,IACF;AAEA,QAAI,WAAW;AACb,WAAK;AAAA,IACP;AAAA,EACF;AAAA,EAIA,kBAAkB,MAA8B;AAC9C,SAAK,oBAAoB,UAAU,IAAI;AAAA,EACzC;AAAA;AAAA;AAAA,EAIA,SAAS,MAAc;AACrB,QAAI,CAAC,KAAK,qBAAqB;AAC7B,WAAK,sBAAsB,KAAK,eAAe;AAE/C,WAAK,oBAAoB,QAAQ,MAAM,KAAK,OAAO,MAAM,CAAC;AAAA,IAC5D;AACA,SAAK,gBAAgB;AAErB,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,IAAI;AAAA,EACrB;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,cAAc;AACrB,WAAK,qBAAqB,KAAK,KAAK,YAAY;AAChD,WAAK,eAAe;AAAA,IACtB;AACA,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,iBAAiB,cAAc;AAAA,EAChD;AAAA;AAAA,EAGA,WAAW;AACT,SAAK,MAAM;AACX,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA0F;AACxF,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA,EAEA,CAAC,OAAO,aAAa,IAAsB;AACzC,WAAO;AAAA,EACT;AACF;AAgBO,MAAe,cAAiE;AAAA,EAC3E,QAAQ,IAAI,gCAAqC;AAAA,EACjD,SAAS,IAAI,gCAAqC;AAAA,EAClD,SAAS;AAAA,EAEnB;AAAA,EACA;AAAA,EACQ;AAAA,EACA,aAAS,gBAAI;AAAA,EAErB,YACE,MACA,KACA,cAAiC,0CACjC;AACA,SAAK,QAAQ;AACb,SAAK,OAAO;AACZ,SAAK,eAAe;AAEpB,SAAK,eAAe;AAMpB,YAAQ,QAAQ,EAAE,KAAK,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAC7E;AAAA,EAEA,MAAc,WAAW;AAEvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AAEF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,4BAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,qCAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,kDAAkD,aAAa;AAAA,YACjE;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,sBAAM,oBAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,WAAO,sBAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAIA,IAAI,YAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,qBAAiB,SAAS,KAAK,OAAO;AACpC,WAAK,OAAO,IAAI,KAAK;AACrB,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AACA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAAA,IAChF;AACA,SAAK,OAAO,MAAM;AAElB,UAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,UAAM,UAAsB;AAAA,MAC1B,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,MAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,MACzD,iBAAiB,KAAK,MAAM;AAAA,MAC5B,iBAAiB,KAAK,MAAM,eAAe;AAAA,MAC3C,WAAW;AAAA;AAAA,MACX,OAAO,KAAK,KAAK;AAAA,MACjB,UAAU;AAAA,IACZ;AACA,SAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,EAC7C;AAAA;AAAA,EAGA,MAAM,UAA+B;AACnC,UAAM,SAAS,CAAC;AAChB,qBAAiB,SAAS,MAAM;AAC9B,aAAO,KAAK,MAAM,KAAK;AAAA,IACzB;AACA,eAAO,0BAAY,MAAM;AAAA,EAC3B;AAAA,EAEA,OAAkD;AAChD,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,MAAM,MAAM;AACjB,SAAK,OAAO,MAAM;AAClB,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAmB;AACtC,WAAO;AAAA,EACT;AACF;","names":[]}
|
package/dist/tts/tts.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../../src/tts/tts.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AACpD,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAEhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AAGtD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,oBAAoB,CAAC;AAErD,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AAClF,OAAO,EAAE,kBAAkB,EAA0C,MAAM,aAAa,CAAC;AAEzF,+EAA+E;AAC/E,MAAM,WAAW,gBAAgB;IAC/B,qEAAqE;IACrE,SAAS,EAAE,MAAM,CAAC;IAClB,uDAAuD;IACvD,SAAS,EAAE,MAAM,CAAC;IAClB,8BAA8B;IAC9B,KAAK,EAAE,UAAU,CAAC;IAClB,+CAA+C;IAC/C,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,qEAAqE;IACrE,KAAK,EAAE,OAAO,CAAC;CAChB;AAED;;;;;;GAMG;AACH,MAAM,WAAW,eAAe;IAC9B,SAAS,EAAE,OAAO,CAAC;CACpB;AAED,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,WAAW,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;IAClB,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,KAAK,CAAC;IACb,WAAW,EAAE,OAAO,CAAC;CACtB;AAED,MAAM,MAAM,YAAY,GAAG;IACzB,CAAC,mBAAmB,CAAC,EAAE,CAAC,OAAO,EAAE,UAAU,KAAK,IAAI,CAAC;IACrD,CAAC,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,CAAC;CACtC,CAAC;kCAS2D,aAAa,YAAY,CAAC;AAPvF;;;;;;GAMG;AACH,8BAAsB,GAAI,SAAQ,QAAsD;;IAItF,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;gBAEX,UAAU,EAAE,MAAM,EAAE,WAAW,EAAE,MAAM,EAAE,YAAY,EAAE,eAAe;IAOlF,sCAAsC;IACtC,IAAI,YAAY,IAAI,eAAe,CAElC;IAED,mEAAmE;IACnE,IAAI,UAAU,IAAI,MAAM,CAEvB;IAED,qEAAqE;IACrE,IAAI,WAAW,IAAI,MAAM,CAExB;IAED;;OAEG;IACH,QAAQ,CAAC,UAAU,CAAC,IAAI,EAAE,MAAM,GAAG,aAAa;IAEhD;;OAEG;IACH,QAAQ,CAAC,MAAM,IAAI,gBAAgB;IAE7B,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;CAG7B;AAED;;;;;;;;;;;;;GAaG;AACH,8BAAsB,gBACpB,YAAW,qBAAqB,CAAC,gBAAgB,GAAG,OAAO,gBAAgB,CAAC,aAAa,CAAC;;IAE1F,SAAS,CAAC,MAAM,CAAC,QAAQ,CAAC,cAAc,gBAA4B;IACpE,MAAM,CAAC,QAAQ,CAAC,aAAa,gBAA2B;IACxD,SAAS,CAAC,KAAK,sEAA6E;IAC5F,SAAS,CAAC,KAAK,+EAEX;IACJ,SAAS,CAAC,MAAM,+EAEZ;IACJ,SAAS,CAAC,MAAM,UAAS;IACzB,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IAKvB,OAAO,CAAC,YAAY,CAAoB;IACxC,SAAS,CAAC,eAAe,kBAAyB;IAElD,OAAO,CAAC,mBAAmB,CAEzB;IACF,OAAO,CAAC,MAAM,CAAS;gBAEX,GAAG,EAAE,GAAG,EAAE,WAAW,GAAE,iBAA+C;YAoBpE,QAAQ;IAuCtB,OAAO,CAAC,SAAS;
|
|
1
|
+
{"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../../src/tts/tts.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AACpD,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAEhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AAGtD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,oBAAoB,CAAC;AAErD,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AAClF,OAAO,EAAE,kBAAkB,EAA0C,MAAM,aAAa,CAAC;AAEzF,+EAA+E;AAC/E,MAAM,WAAW,gBAAgB;IAC/B,qEAAqE;IACrE,SAAS,EAAE,MAAM,CAAC;IAClB,uDAAuD;IACvD,SAAS,EAAE,MAAM,CAAC;IAClB,8BAA8B;IAC9B,KAAK,EAAE,UAAU,CAAC;IAClB,+CAA+C;IAC/C,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,qEAAqE;IACrE,KAAK,EAAE,OAAO,CAAC;CAChB;AAED;;;;;;GAMG;AACH,MAAM,WAAW,eAAe;IAC9B,SAAS,EAAE,OAAO,CAAC;CACpB;AAED,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,WAAW,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;IAClB,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,KAAK,CAAC;IACb,WAAW,EAAE,OAAO,CAAC;CACtB;AAED,MAAM,MAAM,YAAY,GAAG;IACzB,CAAC,mBAAmB,CAAC,EAAE,CAAC,OAAO,EAAE,UAAU,KAAK,IAAI,CAAC;IACrD,CAAC,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,CAAC;CACtC,CAAC;kCAS2D,aAAa,YAAY,CAAC;AAPvF;;;;;;GAMG;AACH,8BAAsB,GAAI,SAAQ,QAAsD;;IAItF,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;gBAEX,UAAU,EAAE,MAAM,EAAE,WAAW,EAAE,MAAM,EAAE,YAAY,EAAE,eAAe;IAOlF,sCAAsC;IACtC,IAAI,YAAY,IAAI,eAAe,CAElC;IAED,mEAAmE;IACnE,IAAI,UAAU,IAAI,MAAM,CAEvB;IAED,qEAAqE;IACrE,IAAI,WAAW,IAAI,MAAM,CAExB;IAED;;OAEG;IACH,QAAQ,CAAC,UAAU,CAAC,IAAI,EAAE,MAAM,GAAG,aAAa;IAEhD;;OAEG;IACH,QAAQ,CAAC,MAAM,IAAI,gBAAgB;IAE7B,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;CAG7B;AAED;;;;;;;;;;;;;GAaG;AACH,8BAAsB,gBACpB,YAAW,qBAAqB,CAAC,gBAAgB,GAAG,OAAO,gBAAgB,CAAC,aAAa,CAAC;;IAE1F,SAAS,CAAC,MAAM,CAAC,QAAQ,CAAC,cAAc,gBAA4B;IACpE,MAAM,CAAC,QAAQ,CAAC,aAAa,gBAA2B;IACxD,SAAS,CAAC,KAAK,sEAA6E;IAC5F,SAAS,CAAC,KAAK,+EAEX;IACJ,SAAS,CAAC,MAAM,+EAEZ;IACJ,SAAS,CAAC,MAAM,UAAS;IACzB,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IAKvB,OAAO,CAAC,YAAY,CAAoB;IACxC,SAAS,CAAC,eAAe,kBAAyB;IAElD,OAAO,CAAC,mBAAmB,CAEzB;IACF,OAAO,CAAC,MAAM,CAAS;gBAEX,GAAG,EAAE,GAAG,EAAE,WAAW,GAAE,iBAA+C;YAoBpE,QAAQ;IAuCtB,OAAO,CAAC,SAAS;cAoBD,SAAS;cAuBT,cAAc;IAiD9B,SAAS,CAAC,QAAQ,CAAC,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IAEvC,iBAAiB,CAAC,IAAI,EAAE,cAAc,CAAC,MAAM,CAAC;IAI9C,uCAAuC;IACvC,kDAAkD;IAClD,QAAQ,CAAC,IAAI,EAAE,MAAM;IAiBrB,4DAA4D;IAC5D,KAAK;IAcL,2DAA2D;IAC3D,QAAQ;IAWR,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,gBAAgB,GAAG,OAAO,gBAAgB,CAAC,aAAa,CAAC,CAAC;IAIzF,wDAAwD;IACxD,KAAK;IAIL,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,gBAAgB;CAG3C;AAED;;;;;;;;;;;;;GAaG;AACH,8BAAsB,aAAc,YAAW,qBAAqB,CAAC,gBAAgB,CAAC;;IACpF,SAAS,CAAC,KAAK,uCAA8C;IAC7D,SAAS,CAAC,MAAM,uCAA8C;IAC9D,SAAS,CAAC,MAAM,UAAS;IACzB,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IAGvB,OAAO,CAAC,YAAY,CAAoB;IACxC,OAAO,CAAC,MAAM,CAAS;gBAGrB,IAAI,EAAE,MAAM,EACZ,GAAG,EAAE,GAAG,EACR,WAAW,GAAE,iBAA+C;YAehD,QAAQ;IAuCtB,OAAO,CAAC,SAAS;IAUjB,SAAS,CAAC,QAAQ,CAAC,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IAEvC,IAAI,SAAS,IAAI,MAAM,CAEtB;cAEe,cAAc;IAgC9B,oDAAoD;IAC9C,OAAO,IAAI,OAAO,CAAC,UAAU,CAAC;IAQpC,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,gBAAgB,CAAC,CAAC;IAIjD,wDAAwD;IACxD,KAAK;IAML,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,aAAa;CAGxC"}
|
package/dist/tts/tts.js
CHANGED
|
@@ -99,7 +99,16 @@ class SynthesizeStream {
|
|
|
99
99
|
recoverable
|
|
100
100
|
});
|
|
101
101
|
}
|
|
102
|
-
//
|
|
102
|
+
// NOTE(AJS-37): The implementation below uses an AsyncIterableQueue (`this.input`)
|
|
103
|
+
// bridged from a DeferredReadableStream (`this.deferredInputStream`) rather than
|
|
104
|
+
// consuming the stream directly.
|
|
105
|
+
//
|
|
106
|
+
// A full refactor to native Web Streams was considered but is currently deferred.
|
|
107
|
+
// The primary reason is to maintain architectural parity with the Python SDK,
|
|
108
|
+
// which is a key design goal for the project. This ensures a consistent developer
|
|
109
|
+
// experience across both platforms.
|
|
110
|
+
//
|
|
111
|
+
// For more context, see the discussion in GitHub issue # 844.
|
|
103
112
|
async pumpInput() {
|
|
104
113
|
const reader = this.deferredInputStream.stream.getReader();
|
|
105
114
|
try {
|