@livekit/agents-plugin-cartesia 1.0.43 → 1.0.45

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -15,3 +15,10 @@ information on how to use it, or browse the [API
15
15
  reference](https://docs.livekit.io/agents-js/modules/plugins_agents_plugin_cartesia.html).
16
16
  See the [repository](https://github.com/livekit/agents-js) for more information
17
17
  about the framework as a whole.
18
+
19
+ ## Troubleshooting
20
+
21
+ If you see `AggregateError` / `ETIMEDOUT` on the first Cartesia connection in Node.js (often due to IPv6/IPv4 “happy eyeballs”), you can mitigate it by:
22
+
23
+ - Setting a longer Node autoselection attempt timeout (example): `NODE_OPTIONS="--network-family-autoselection-attempt-timeout=5000"`
24
+ - Increasing the agent session TTS connect timeout via `connOptions.ttsConnOptions.timeoutMs`
package/dist/tts.cjs CHANGED
@@ -83,8 +83,8 @@ class TTS extends import_agents.tts.TTS {
83
83
  synthesize(text, connOptions, abortSignal) {
84
84
  return new ChunkedStream(this, text, this.#opts, connOptions, abortSignal);
85
85
  }
86
- stream() {
87
- return new SynthesizeStream(this, this.#opts);
86
+ stream(options) {
87
+ return new SynthesizeStream(this, this.#opts, options == null ? void 0 : options.connOptions);
88
88
  }
89
89
  }
90
90
  class ChunkedStream extends import_agents.tts.ChunkedStream {
@@ -137,22 +137,32 @@ class ChunkedStream extends import_agents.tts.ChunkedStream {
137
137
  });
138
138
  }
139
139
  this.queue.close();
140
- doneFut.resolve();
140
+ if (!doneFut.done) doneFut.resolve();
141
141
  });
142
142
  res.on("error", (err) => {
143
143
  if (err.message === "aborted") return;
144
144
  this.#logger.error({ err }, "Cartesia TTS response error");
145
+ if (!doneFut.done) doneFut.reject(err);
145
146
  });
146
147
  }
147
148
  );
148
149
  req.on("error", (err) => {
149
150
  if (err.name === "AbortError") return;
150
151
  this.#logger.error({ err }, "Cartesia TTS request error");
152
+ if (!doneFut.done) doneFut.reject(err);
153
+ });
154
+ req.on("close", () => {
155
+ if (!doneFut.done) doneFut.resolve();
151
156
  });
152
- req.on("close", () => doneFut.resolve());
153
157
  req.write(JSON.stringify(json));
154
158
  req.end();
155
- await doneFut.await;
159
+ try {
160
+ await doneFut.await;
161
+ } catch (e) {
162
+ if (this.abortSignal.aborted) return;
163
+ if (!this.queue.closed) this.queue.close();
164
+ throw toRetryableConnectionError(e);
165
+ }
156
166
  }
157
167
  }
158
168
  class SynthesizeStream extends import_agents.tts.SynthesizeStream {
@@ -162,8 +172,8 @@ class SynthesizeStream extends import_agents.tts.SynthesizeStream {
162
172
  minSentenceLength: BUFFERED_WORDS_COUNT
163
173
  }).stream();
164
174
  label = "cartesia.SynthesizeStream";
165
- constructor(tts2, opts) {
166
- super(tts2);
175
+ constructor(tts2, opts, connOptions) {
176
+ super(tts2, connOptions);
167
177
  this.#opts = opts;
168
178
  }
169
179
  updateOptions(opts) {
@@ -347,19 +357,139 @@ class SynthesizeStream extends import_agents.tts.SynthesizeStream {
347
357
  };
348
358
  const wsUrl = this.#opts.baseUrl.replace(/^http/, "ws");
349
359
  const url = `${wsUrl}/tts/websocket?api_key=${this.#opts.apiKey}&cartesia_version=${VERSION}`;
350
- const ws = new import_ws.WebSocket(url);
360
+ let ws;
351
361
  try {
352
- await new Promise((resolve, reject) => {
353
- ws.on("open", resolve);
354
- ws.on("error", (error) => reject(error));
355
- ws.on("close", (code) => reject(`WebSocket returned ${code}`));
362
+ ws = await connectCartesiaWebSocket({
363
+ url,
364
+ timeoutMs: this.connOptions.timeoutMs,
365
+ abortSignal: this.abortSignal
356
366
  });
357
367
  await Promise.all([inputTask(), sentenceStreamTask(ws), recvTask(ws)]);
358
368
  } catch (e) {
359
- throw new Error(`failed to connect to Cartesia: ${e}`);
369
+ if (this.abortSignal.aborted) {
370
+ return;
371
+ }
372
+ throw toRetryableConnectionError(e);
373
+ } finally {
374
+ if (ws && ws.readyState !== import_ws.WebSocket.CLOSED) {
375
+ safeTerminateWebSocket(ws);
376
+ }
360
377
  }
361
378
  }
362
379
  }
380
+ const asError = (e) => e instanceof Error ? e : new Error(String(e));
381
+ const transientNetworkCodes = /* @__PURE__ */ new Set([
382
+ "ETIMEDOUT",
383
+ "ECONNRESET",
384
+ "EAI_AGAIN",
385
+ "ENETUNREACH",
386
+ "ECONNREFUSED",
387
+ "EHOSTUNREACH"
388
+ ]);
389
+ const isRecord = (v) => {
390
+ return v !== null && typeof v === "object";
391
+ };
392
+ const isAggregateErrorLike = (e) => {
393
+ if (!isRecord(e)) return false;
394
+ return e.name === "AggregateError" && Array.isArray(e.errors);
395
+ };
396
+ const hasErrorCode = (e, code) => {
397
+ if (isRecord(e) && e.code === code) return true;
398
+ if (isAggregateErrorLike(e)) {
399
+ return e.errors.some((inner) => hasErrorCode(inner, code));
400
+ }
401
+ return false;
402
+ };
403
+ const hasAnyTransientCode = (e) => {
404
+ if (isRecord(e) && typeof e.code === "string") {
405
+ return transientNetworkCodes.has(e.code);
406
+ }
407
+ if (isAggregateErrorLike(e)) {
408
+ return e.errors.some((inner) => hasAnyTransientCode(inner));
409
+ }
410
+ return false;
411
+ };
412
+ const toRetryableConnectionError = (e) => {
413
+ const err = asError(e);
414
+ const isTimeout = hasErrorCode(e, "ETIMEDOUT") || typeof err.message === "string" && err.message.includes("ETIMEDOUT");
415
+ const message = isTimeout ? `Cartesia connection timed out` : `Cartesia connection failed: ${err.message || "unknown error"}`;
416
+ return isTimeout ? new import_agents.APITimeoutError({ message }) : new import_agents.APIConnectionError({ message });
417
+ };
418
+ const waitForWsOpen = async ({
419
+ ws,
420
+ timeoutMs,
421
+ abortSignal
422
+ }) => {
423
+ if (abortSignal.aborted) {
424
+ throw new Error("aborted");
425
+ }
426
+ const fut = new import_agents.Future();
427
+ let timeout;
428
+ const cleanup = () => {
429
+ if (timeout) clearTimeout(timeout);
430
+ ws.off("open", onOpen);
431
+ ws.off("error", onError);
432
+ ws.off("close", onClose);
433
+ abortSignal.removeEventListener("abort", onAbort);
434
+ };
435
+ const onOpen = () => fut.resolve();
436
+ const onError = (err) => fut.reject(asError(err));
437
+ const onClose = (code, reason) => fut.reject(
438
+ new Error(`WebSocket closed before open (code=${code}, reason=${reason.toString()})`)
439
+ );
440
+ const onAbort = () => fut.reject(new Error("aborted"));
441
+ ws.on("open", onOpen);
442
+ ws.on("error", onError);
443
+ ws.on("close", onClose);
444
+ abortSignal.addEventListener("abort", onAbort, { once: true });
445
+ if (timeoutMs > 0) {
446
+ timeout = setTimeout(() => fut.reject(new Error("connect timeout")), timeoutMs);
447
+ }
448
+ try {
449
+ await fut.await;
450
+ } finally {
451
+ cleanup();
452
+ }
453
+ };
454
+ const safeTerminateWebSocket = (ws) => {
455
+ try {
456
+ ws.on("error", () => {
457
+ });
458
+ } catch {
459
+ }
460
+ try {
461
+ if (ws.readyState === import_ws.WebSocket.CONNECTING) {
462
+ ws.close();
463
+ } else {
464
+ ws.terminate();
465
+ }
466
+ } catch {
467
+ }
468
+ };
469
+ const connectCartesiaWebSocket = async ({
470
+ url,
471
+ timeoutMs,
472
+ abortSignal
473
+ }) => {
474
+ const connectOnce = async (family) => {
475
+ const ws = new import_ws.WebSocket(url, { handshakeTimeout: timeoutMs, family });
476
+ try {
477
+ await waitForWsOpen({ ws, timeoutMs, abortSignal });
478
+ return ws;
479
+ } catch (e) {
480
+ safeTerminateWebSocket(ws);
481
+ throw e;
482
+ }
483
+ };
484
+ try {
485
+ return await connectOnce();
486
+ } catch (e) {
487
+ if (hasAnyTransientCode(e) || isAggregateErrorLike(e)) {
488
+ return await connectOnce(4);
489
+ }
490
+ throw e;
491
+ }
492
+ };
363
493
  const toCartesiaOptions = (opts, streaming = false) => {
364
494
  const voice = {};
365
495
  if (typeof opts.voice === "string") {
@@ -387,7 +517,8 @@ const toCartesiaOptions = (opts, streaming = false) => {
387
517
  encoding: opts.encoding,
388
518
  sample_rate: opts.sampleRate
389
519
  },
390
- language: opts.language
520
+ language: opts.language,
521
+ pronunciation_dict_id: opts.pronunciationDictId
391
522
  };
392
523
  if (streaming && opts.wordTimestamps !== false) {
393
524
  result.add_timestamps = true;
package/dist/tts.cjs.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n AudioByteStream,\n Future,\n type TimedString,\n createTimedString,\n log,\n shortuuid,\n stream,\n tokenize,\n tts,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { request } from 'node:https';\nimport { type RawData, WebSocket } from 'ws';\nimport {\n TTSDefaultVoiceId,\n type TTSEncoding,\n type TTSModels,\n type TTSVoiceEmotion,\n type TTSVoiceSpeed,\n} from './models.js';\nimport {\n type CartesiaServerMessage,\n cartesiaMessageSchema,\n hasWordTimestamps,\n isChunkMessage,\n isDoneMessage,\n isErrorMessage,\n} from './types.js';\n\nconst AUTHORIZATION_HEADER = 'X-API-Key';\nconst VERSION_HEADER = 'Cartesia-Version';\nconst VERSION = '2024-06-10';\nconst NUM_CHANNELS = 1;\nconst BUFFERED_WORDS_COUNT = 8;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n encoding: TTSEncoding;\n sampleRate: number;\n voice: string | number[];\n speed?: TTSVoiceSpeed | number;\n emotion?: (TTSVoiceEmotion | string)[];\n apiKey?: string;\n language: string;\n baseUrl: string;\n\n /**\n * The timeout for the next chunk to be received from the Cartesia API.\n */\n chunkTimeout: number;\n\n /**\n * Whether to add word timestamps to the output. When enabled, the TTS will return\n * timing information for each word in the transcript.\n * @defaultValue true\n */\n wordTimestamps?: boolean;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'sonic-2',\n encoding: 'pcm_s16le',\n sampleRate: 24000,\n voice: TTSDefaultVoiceId,\n apiKey: process.env.CARTESIA_API_KEY,\n language: 'en',\n baseUrl: 'https://api.cartesia.ai',\n chunkTimeout: 5000,\n wordTimestamps: true,\n};\n\nexport class TTS extends tts.TTS {\n #opts: TTSOptions;\n label = 'cartesia.TTS';\n\n constructor(opts: Partial<TTSOptions> = {}) {\n const resolvedOpts = {\n ...defaultTTSOptions,\n ...opts,\n };\n\n super(resolvedOpts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {\n streaming: true,\n alignedTranscript: resolvedOpts.wordTimestamps ?? true,\n });\n\n this.#opts = resolvedOpts;\n\n if (this.#opts.apiKey === undefined) {\n throw new Error(\n 'Cartesia API key is required, whether as an argument or as $CARTESIA_API_KEY',\n );\n }\n\n if ((this.#opts.speed || this.#opts.emotion) && this.#opts.model !== 'sonic-2-2025-03-07') {\n const logger = log();\n logger.warn(\n { model: this.#opts.model, speed: this.#opts.speed, emotion: this.#opts.emotion },\n \"speed and emotion controls are only supported for model 'sonic-2-2025-03-07', see https://docs.cartesia.ai/developer-tools/changelog for details\",\n );\n }\n }\n\n updateOptions(opts: Partial<TTSOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n if ((this.#opts.speed || this.#opts.emotion) && this.#opts.model !== 'sonic-2-2025-03-07') {\n const logger = log();\n logger.warn(\n { model: this.#opts.model, speed: this.#opts.speed, emotion: this.#opts.emotion },\n \"speed and emotion controls are only supported for model 'sonic-2-2025-03-07', see https://docs.cartesia.ai/developer-tools/changelog for details\",\n );\n }\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.#opts, connOptions, abortSignal);\n }\n\n stream(): SynthesizeStream {\n return new SynthesizeStream(this, this.#opts);\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'cartesia.ChunkedStream';\n #logger = log();\n #opts: TTSOptions;\n #text: string;\n\n constructor(\n tts: TTS,\n text: string,\n opts: TTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.#text = text;\n this.#opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.#opts.sampleRate, NUM_CHANNELS);\n const json = toCartesiaOptions(this.#opts);\n json.transcript = this.#text;\n\n const baseUrl = new URL(this.#opts.baseUrl);\n const doneFut = new Future<void>();\n\n const req = request(\n {\n hostname: baseUrl.hostname,\n port: parseInt(baseUrl.port) || (baseUrl.protocol === 'https:' ? 443 : 80),\n path: '/tts/bytes',\n method: 'POST',\n headers: {\n [AUTHORIZATION_HEADER]: this.#opts.apiKey!,\n [VERSION_HEADER]: VERSION,\n },\n signal: this.abortSignal,\n },\n (res) => {\n res.on('data', (chunk) => {\n for (const frame of bstream.write(chunk)) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n });\n res.on('close', () => {\n for (const frame of bstream.flush()) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n this.queue.close();\n doneFut.resolve();\n });\n res.on('error', (err) => {\n if (err.message === 'aborted') return;\n this.#logger.error({ err }, 'Cartesia TTS response error');\n });\n },\n );\n\n req.on('error', (err) => {\n if (err.name === 'AbortError') return;\n this.#logger.error({ err }, 'Cartesia TTS request error');\n });\n req.on('close', () => doneFut.resolve());\n req.write(JSON.stringify(json));\n req.end();\n\n await doneFut.await;\n }\n}\n\nexport class SynthesizeStream extends tts.SynthesizeStream {\n #opts: TTSOptions;\n #logger = log();\n #tokenizer = new tokenize.basic.SentenceTokenizer({\n minSentenceLength: BUFFERED_WORDS_COUNT,\n }).stream();\n label = 'cartesia.SynthesizeStream';\n\n constructor(tts: TTS, opts: TTSOptions) {\n super(tts);\n this.#opts = opts;\n }\n\n updateOptions(opts: Partial<TTSOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n if ((this.#opts.speed || this.#opts.emotion) && this.#opts.model !== 'sonic-2-2025-03-07') {\n this.#logger.warn(\n { model: this.#opts.model, speed: this.#opts.speed, emotion: this.#opts.emotion },\n \"speed and emotion controls are only supported for model 'sonic-2-2025-03-07', see https://docs.cartesia.ai/developer-tools/changelog for details\",\n );\n }\n }\n\n protected async run() {\n const requestId = shortuuid();\n let closing = false;\n // Only close WebSocket when both: 1) Cartesia returns done, AND 2) all sentences have been sent\n let sentenceStreamClosed = false;\n\n const sentenceStreamTask = async (ws: WebSocket) => {\n const packet = toCartesiaOptions(this.#opts, true);\n for await (const event of this.#tokenizer) {\n const msg = {\n ...packet,\n context_id: requestId,\n transcript: event.token + ' ',\n continue: true,\n };\n ws.send(JSON.stringify(msg));\n }\n\n const endMsg = {\n ...packet,\n context_id: requestId,\n transcript: ' ',\n continue: false,\n };\n ws.send(JSON.stringify(endMsg));\n // Mark sentence stream as closed\n sentenceStreamClosed = true;\n };\n\n const inputTask = async () => {\n for await (const data of this.input) {\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n this.#tokenizer.flush();\n continue;\n }\n this.#tokenizer.pushText(data);\n }\n this.#tokenizer.endInput();\n this.#tokenizer.close();\n };\n\n // Use event channel and set up listeners ONCE to avoid missing messages during listener re-registration\n const recvTask = async (ws: WebSocket) => {\n const bstream = new AudioByteStream(this.#opts.sampleRate, NUM_CHANNELS);\n\n // Create event channel to buffer incoming messages\n // This prevents message loss between listener re-registrations\n const eventChannel = stream.createStreamChannel<RawData>();\n\n let lastFrame: AudioFrame | undefined;\n let pendingTimedTranscripts: TimedString[] = [];\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame && !this.queue.closed) {\n // Include timedTranscripts with the audio frame\n this.queue.put({\n requestId,\n segmentId,\n frame: lastFrame,\n final,\n timedTranscripts:\n pendingTimedTranscripts.length > 0 ? pendingTimedTranscripts : undefined,\n });\n lastFrame = undefined;\n pendingTimedTranscripts = [];\n }\n };\n\n let timeout: NodeJS.Timeout | null = null;\n\n const clearTTSChunkTimeout = () => {\n if (timeout) {\n clearTimeout(timeout);\n timeout = null;\n }\n };\n\n // Set up WebSocket listeners ONCE (not in a loop)\n const onMessage = (data: RawData) => {\n void eventChannel.write(data).catch((error: unknown) => {\n this.#logger.debug({ error }, 'Failed writing Cartesia event to channel (likely closed)');\n });\n };\n\n const onClose = (code: number, reason: Buffer) => {\n if (!closing) {\n this.#logger.debug(`WebSocket closed with code ${code}: ${reason.toString()}`);\n }\n clearTTSChunkTimeout();\n void eventChannel.close();\n };\n\n const onError = (err: Error) => {\n this.#logger.error({ err }, 'Cartesia WebSocket error');\n void eventChannel.close();\n };\n\n // Attach listeners ONCE\n ws.on('message', onMessage);\n ws.on('close', onClose);\n ws.on('error', onError);\n\n try {\n // Process messages from the channel\n const reader = eventChannel.stream().getReader();\n\n while (!this.closed && !this.abortController.signal.aborted) {\n const result = await reader.read();\n if (result.done) break;\n\n const rawMsg = result.value;\n\n // Parse message with Zod schema for type safety\n let serverMsg: CartesiaServerMessage;\n try {\n const json = JSON.parse(rawMsg.toString());\n serverMsg = cartesiaMessageSchema.parse(json);\n } catch (parseErr) {\n this.#logger.warn({ parseErr }, 'Failed to parse Cartesia message');\n continue;\n }\n\n // Handle error messages\n if (isErrorMessage(serverMsg)) {\n this.#logger.error({ error: serverMsg.error }, 'Cartesia returned error');\n continue;\n }\n\n const segmentId = serverMsg.context_id;\n\n // Process word timestamps if present (typed via Zod schema)\n if (this.#opts.wordTimestamps !== false && hasWordTimestamps(serverMsg)) {\n const wordTimestamps = serverMsg.word_timestamps;\n for (let i = 0; i < wordTimestamps.words.length; i++) {\n const word = wordTimestamps.words[i];\n const startTime = wordTimestamps.start[i];\n const endTime = wordTimestamps.end[i];\n if (word !== undefined && startTime !== undefined && endTime !== undefined) {\n pendingTimedTranscripts.push(\n createTimedString({\n text: word + ' ', // Add space after word for consistency\n startTime,\n endTime,\n }),\n );\n }\n }\n }\n\n // Handle audio chunk messages\n if (isChunkMessage(serverMsg)) {\n const audioBuffer = Buffer.from(serverMsg.data, 'base64');\n // Extract ArrayBuffer from Buffer for AudioByteStream compatibility\n const audioData = audioBuffer.buffer.slice(\n audioBuffer.byteOffset,\n audioBuffer.byteOffset + audioBuffer.byteLength,\n );\n for (const frame of bstream.write(audioData)) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n\n // IMPORTANT: close WS if TTS chunk stream been stuck too long\n // this allows unblock the current \"broken\" TTS node so that any future TTS nodes\n // can continue to process the stream without been blocked by the stuck node\n clearTTSChunkTimeout();\n timeout = setTimeout(() => {\n // cartesia chunk timeout quite often, so we make it a debug log\n this.#logger.debug(\n `Cartesia WebSocket TTS chunk stream timeout after ${this.#opts.chunkTimeout}ms`,\n );\n ws.close();\n }, this.#opts.chunkTimeout);\n } else if (isDoneMessage(serverMsg)) {\n // This ensures all sentences have been sent before closing\n if (sentenceStreamClosed) {\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n\n if (segmentId === requestId) {\n closing = true;\n clearTTSChunkTimeout();\n ws.close();\n break; // Exit the loop\n }\n }\n // If sentenceStreamClosed is false, continue receiving - more done messages will come\n }\n }\n } catch (err) {\n // skip log error for normal websocket close\n if (err instanceof Error && !err.message.includes('WebSocket closed')) {\n if (\n err.message.includes('Queue is closed') ||\n err.message.includes('Channel is closed')\n ) {\n this.#logger.warn(\n { err },\n 'Channel closed during transcript processing (expected during disconnect)',\n );\n } else {\n this.#logger.error({ err }, 'Error in recvTask from Cartesia WebSocket');\n }\n }\n } finally {\n // IMPORTANT: Remove listeners so connection can be reused\n ws.off('message', onMessage);\n ws.off('close', onClose);\n ws.off('error', onError);\n clearTTSChunkTimeout();\n }\n };\n\n const wsUrl = this.#opts.baseUrl.replace(/^http/, 'ws');\n const url = `${wsUrl}/tts/websocket?api_key=${this.#opts.apiKey}&cartesia_version=${VERSION}`;\n const ws = new WebSocket(url);\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await Promise.all([inputTask(), sentenceStreamTask(ws), recvTask(ws)]);\n } catch (e) {\n throw new Error(`failed to connect to Cartesia: ${e}`);\n }\n }\n}\n\n/**\n * Convert TTSOptions to Cartesia API format.\n *\n * @param opts - TTS options\n * @param streaming - Whether this is for streaming (WebSocket) or non-streaming (HTTP)\n */\nconst toCartesiaOptions = (\n opts: TTSOptions,\n streaming: boolean = false,\n): { [id: string]: unknown } => {\n const voice: { [id: string]: unknown } = {};\n if (typeof opts.voice === 'string') {\n voice.mode = 'id';\n voice.id = opts.voice;\n } else {\n voice.mode = 'embedding';\n voice.embedding = opts.voice;\n }\n\n const voiceControls: { [id: string]: unknown } = {};\n if (opts.speed) {\n voiceControls.speed = opts.speed;\n }\n if (opts.emotion) {\n voiceControls.emotion = opts.emotion;\n }\n\n if (Object.keys(voiceControls).length) {\n voice.__experimental_controls = voiceControls;\n }\n\n const result: { [id: string]: unknown } = {\n model_id: opts.model,\n voice,\n output_format: {\n container: 'raw',\n encoding: opts.encoding,\n sample_rate: opts.sampleRate,\n },\n language: opts.language,\n };\n\n if (streaming && opts.wordTimestamps !== false) {\n result.add_timestamps = true;\n }\n\n return result;\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAWO;AAEP,wBAAwB;AACxB,gBAAwC;AACxC,oBAMO;AACP,mBAOO;AAEP,MAAM,uBAAuB;AAC7B,MAAM,iBAAiB;AACvB,MAAM,UAAU;AAChB,MAAM,eAAe;AACrB,MAAM,uBAAuB;AA0B7B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,OAAO;AAAA,EACP,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,SAAS;AAAA,EACT,cAAc;AAAA,EACd,gBAAgB;AAClB;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,QAAQ;AAAA,EAER,YAAY,OAA4B,CAAC,GAAG;AAC1C,UAAM,eAAe;AAAA,MACnB,GAAG;AAAA,MACH,GAAG;AAAA,IACL;AAEA,UAAM,aAAa,cAAc,kBAAkB,YAAY,cAAc;AAAA,MAC3E,WAAW;AAAA,MACX,mBAAmB,aAAa,kBAAkB;AAAA,IACpD,CAAC;AAED,SAAK,QAAQ;AAEb,QAAI,KAAK,MAAM,WAAW,QAAW;AACnC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,KAAK,MAAM,SAAS,KAAK,MAAM,YAAY,KAAK,MAAM,UAAU,sBAAsB;AACzF,YAAM,aAAS,mBAAI;AACnB,aAAO;AAAA,QACL,EAAE,OAAO,KAAK,MAAM,OAAO,OAAO,KAAK,MAAM,OAAO,SAAS,KAAK,MAAM,QAAQ;AAAA,QAChF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,SAAK,KAAK,MAAM,SAAS,KAAK,MAAM,YAAY,KAAK,MAAM,UAAU,sBAAsB;AACzF,YAAM,aAAS,mBAAI;AACnB,aAAO;AAAA,QACL,EAAE,OAAO,KAAK,MAAM,OAAO,OAAO,KAAK,MAAM,OAAO,SAAS,KAAK,MAAM,QAAQ;AAAA,QAChF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,WACE,MACA,aACA,aACmB;AACnB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,OAAO,aAAa,WAAW;AAAA,EAC3E;AAAA,EAEA,SAA2B;AACzB,WAAO,IAAI,iBAAiB,MAAM,KAAK,KAAK;AAAA,EAC9C;AACF;AAEO,MAAM,sBAAsB,kBAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACR,cAAU,mBAAI;AAAA,EACd;AAAA,EACA;AAAA,EAEA,YACEA,MACA,MACA,MACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,QAAQ;AACb,SAAK,QAAQ;AAAA,EACf;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,gBAAY,yBAAU;AAC5B,UAAM,UAAU,IAAI,8BAAgB,KAAK,MAAM,YAAY,YAAY;AACvE,UAAM,OAAO,kBAAkB,KAAK,KAAK;AACzC,SAAK,aAAa,KAAK;AAEvB,UAAM,UAAU,IAAI,IAAI,KAAK,MAAM,OAAO;AAC1C,UAAM,UAAU,IAAI,qBAAa;AAEjC,UAAM,UAAM;AAAA,MACV;AAAA,QACE,UAAU,QAAQ;AAAA,QAClB,MAAM,SAAS,QAAQ,IAAI,MAAM,QAAQ,aAAa,WAAW,MAAM;AAAA,QACvE,MAAM;AAAA,QACN,QAAQ;AAAA,QACR,SAAS;AAAA,UACP,CAAC,oBAAoB,GAAG,KAAK,MAAM;AAAA,UACnC,CAAC,cAAc,GAAG;AAAA,QACpB;AAAA,QACA,QAAQ,KAAK;AAAA,MACf;AAAA,MACA,CAAC,QAAQ;AACP,YAAI,GAAG,QAAQ,CAAC,UAAU;AACxB,qBAAW,SAAS,QAAQ,MAAM,KAAK,GAAG;AACxC,iBAAK,MAAM,IAAI;AAAA,cACb;AAAA,cACA;AAAA,cACA,OAAO;AAAA,cACP,WAAW;AAAA,YACb,CAAC;AAAA,UACH;AAAA,QACF,CAAC;AACD,YAAI,GAAG,SAAS,MAAM;AACpB,qBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,iBAAK,MAAM,IAAI;AAAA,cACb;AAAA,cACA;AAAA,cACA,OAAO;AAAA,cACP,WAAW;AAAA,YACb,CAAC;AAAA,UACH;AACA,eAAK,MAAM,MAAM;AACjB,kBAAQ,QAAQ;AAAA,QAClB,CAAC;AACD,YAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,cAAI,IAAI,YAAY,UAAW;AAC/B,eAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6BAA6B;AAAA,QAC3D,CAAC;AAAA,MACH;AAAA,IACF;AAEA,QAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,UAAI,IAAI,SAAS,aAAc;AAC/B,WAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,4BAA4B;AAAA,IAC1D,CAAC;AACD,QAAI,GAAG,SAAS,MAAM,QAAQ,QAAQ,CAAC;AACvC,QAAI,MAAM,KAAK,UAAU,IAAI,CAAC;AAC9B,QAAI,IAAI;AAER,UAAM,QAAQ;AAAA,EAChB;AACF;AAEO,MAAM,yBAAyB,kBAAI,iBAAiB;AAAA,EACzD;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,aAAa,IAAI,uBAAS,MAAM,kBAAkB;AAAA,IAChD,mBAAmB;AAAA,EACrB,CAAC,EAAE,OAAO;AAAA,EACV,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,QAAQ;AAAA,EACf;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,SAAK,KAAK,MAAM,SAAS,KAAK,MAAM,YAAY,KAAK,MAAM,UAAU,sBAAsB;AACzF,WAAK,QAAQ;AAAA,QACX,EAAE,OAAO,KAAK,MAAM,OAAO,OAAO,KAAK,MAAM,OAAO,SAAS,KAAK,MAAM,QAAQ;AAAA,QAChF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,gBAAY,yBAAU;AAC5B,QAAI,UAAU;AAEd,QAAI,uBAAuB;AAE3B,UAAM,qBAAqB,OAAOC,QAAkB;AAClD,YAAM,SAAS,kBAAkB,KAAK,OAAO,IAAI;AACjD,uBAAiB,SAAS,KAAK,YAAY;AACzC,cAAM,MAAM;AAAA,UACV,GAAG;AAAA,UACH,YAAY;AAAA,UACZ,YAAY,MAAM,QAAQ;AAAA,UAC1B,UAAU;AAAA,QACZ;AACA,QAAAA,IAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,MAC7B;AAEA,YAAM,SAAS;AAAA,QACb,GAAG;AAAA,QACH,YAAY;AAAA,QACZ,YAAY;AAAA,QACZ,UAAU;AAAA,MACZ;AACA,MAAAA,IAAG,KAAK,KAAK,UAAU,MAAM,CAAC;AAE9B,6BAAuB;AAAA,IACzB;AAEA,UAAM,YAAY,YAAY;AAC5B,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,eAAK,WAAW,MAAM;AACtB;AAAA,QACF;AACA,aAAK,WAAW,SAAS,IAAI;AAAA,MAC/B;AACA,WAAK,WAAW,SAAS;AACzB,WAAK,WAAW,MAAM;AAAA,IACxB;AAGA,UAAM,WAAW,OAAOA,QAAkB;AACxC,YAAM,UAAU,IAAI,8BAAgB,KAAK,MAAM,YAAY,YAAY;AAIvE,YAAM,eAAe,qBAAO,oBAA6B;AAEzD,UAAI;AACJ,UAAI,0BAAyC,CAAC;AAE9C,YAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,YAAI,aAAa,CAAC,KAAK,MAAM,QAAQ;AAEnC,eAAK,MAAM,IAAI;AAAA,YACb;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP;AAAA,YACA,kBACE,wBAAwB,SAAS,IAAI,0BAA0B;AAAA,UACnE,CAAC;AACD,sBAAY;AACZ,oCAA0B,CAAC;AAAA,QAC7B;AAAA,MACF;AAEA,UAAI,UAAiC;AAErC,YAAM,uBAAuB,MAAM;AACjC,YAAI,SAAS;AACX,uBAAa,OAAO;AACpB,oBAAU;AAAA,QACZ;AAAA,MACF;AAGA,YAAM,YAAY,CAAC,SAAkB;AACnC,aAAK,aAAa,MAAM,IAAI,EAAE,MAAM,CAAC,UAAmB;AACtD,eAAK,QAAQ,MAAM,EAAE,MAAM,GAAG,0DAA0D;AAAA,QAC1F,CAAC;AAAA,MACH;AAEA,YAAM,UAAU,CAAC,MAAc,WAAmB;AAChD,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,OAAO,SAAS,CAAC,EAAE;AAAA,QAC/E;AACA,6BAAqB;AACrB,aAAK,aAAa,MAAM;AAAA,MAC1B;AAEA,YAAM,UAAU,CAAC,QAAe;AAC9B,aAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,0BAA0B;AACtD,aAAK,aAAa,MAAM;AAAA,MAC1B;AAGA,MAAAA,IAAG,GAAG,WAAW,SAAS;AAC1B,MAAAA,IAAG,GAAG,SAAS,OAAO;AACtB,MAAAA,IAAG,GAAG,SAAS,OAAO;AAEtB,UAAI;AAEF,cAAM,SAAS,aAAa,OAAO,EAAE,UAAU;AAE/C,eAAO,CAAC,KAAK,UAAU,CAAC,KAAK,gBAAgB,OAAO,SAAS;AAC3D,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,OAAO,KAAM;AAEjB,gBAAM,SAAS,OAAO;AAGtB,cAAI;AACJ,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,OAAO,SAAS,CAAC;AACzC,wBAAY,mCAAsB,MAAM,IAAI;AAAA,UAC9C,SAAS,UAAU;AACjB,iBAAK,QAAQ,KAAK,EAAE,SAAS,GAAG,kCAAkC;AAClE;AAAA,UACF;AAGA,kBAAI,6BAAe,SAAS,GAAG;AAC7B,iBAAK,QAAQ,MAAM,EAAE,OAAO,UAAU,MAAM,GAAG,yBAAyB;AACxE;AAAA,UACF;AAEA,gBAAM,YAAY,UAAU;AAG5B,cAAI,KAAK,MAAM,mBAAmB,aAAS,gCAAkB,SAAS,GAAG;AACvE,kBAAM,iBAAiB,UAAU;AACjC,qBAAS,IAAI,GAAG,IAAI,eAAe,MAAM,QAAQ,KAAK;AACpD,oBAAM,OAAO,eAAe,MAAM,CAAC;AACnC,oBAAM,YAAY,eAAe,MAAM,CAAC;AACxC,oBAAM,UAAU,eAAe,IAAI,CAAC;AACpC,kBAAI,SAAS,UAAa,cAAc,UAAa,YAAY,QAAW;AAC1E,wCAAwB;AAAA,sBACtB,iCAAkB;AAAA,oBAChB,MAAM,OAAO;AAAA;AAAA,oBACb;AAAA,oBACA;AAAA,kBACF,CAAC;AAAA,gBACH;AAAA,cACF;AAAA,YACF;AAAA,UACF;AAGA,kBAAI,6BAAe,SAAS,GAAG;AAC7B,kBAAM,cAAc,OAAO,KAAK,UAAU,MAAM,QAAQ;AAExD,kBAAM,YAAY,YAAY,OAAO;AAAA,cACnC,YAAY;AAAA,cACZ,YAAY,aAAa,YAAY;AAAA,YACvC;AACA,uBAAW,SAAS,QAAQ,MAAM,SAAS,GAAG;AAC5C,4BAAc,WAAW,KAAK;AAC9B,0BAAY;AAAA,YACd;AAKA,iCAAqB;AACrB,sBAAU,WAAW,MAAM;AAEzB,mBAAK,QAAQ;AAAA,gBACX,qDAAqD,KAAK,MAAM,YAAY;AAAA,cAC9E;AACA,cAAAA,IAAG,MAAM;AAAA,YACX,GAAG,KAAK,MAAM,YAAY;AAAA,UAC5B,eAAW,4BAAc,SAAS,GAAG;AAEnC,gBAAI,sBAAsB;AACxB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,WAAW,KAAK;AAC9B,4BAAY;AAAA,cACd;AACA,4BAAc,WAAW,IAAI;AAC7B,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,cAC/C;AAEA,kBAAI,cAAc,WAAW;AAC3B,0BAAU;AACV,qCAAqB;AACrB,gBAAAA,IAAG,MAAM;AACT;AAAA,cACF;AAAA,YACF;AAAA,UAEF;AAAA,QACF;AAAA,MACF,SAAS,KAAK;AAEZ,YAAI,eAAe,SAAS,CAAC,IAAI,QAAQ,SAAS,kBAAkB,GAAG;AACrE,cACE,IAAI,QAAQ,SAAS,iBAAiB,KACtC,IAAI,QAAQ,SAAS,mBAAmB,GACxC;AACA,iBAAK,QAAQ;AAAA,cACX,EAAE,IAAI;AAAA,cACN;AAAA,YACF;AAAA,UACF,OAAO;AACL,iBAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,2CAA2C;AAAA,UACzE;AAAA,QACF;AAAA,MACF,UAAE;AAEA,QAAAA,IAAG,IAAI,WAAW,SAAS;AAC3B,QAAAA,IAAG,IAAI,SAAS,OAAO;AACvB,QAAAA,IAAG,IAAI,SAAS,OAAO;AACvB,6BAAqB;AAAA,MACvB;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,MAAM,QAAQ,QAAQ,SAAS,IAAI;AACtD,UAAM,MAAM,GAAG,KAAK,0BAA0B,KAAK,MAAM,MAAM,qBAAqB,OAAO;AAC3F,UAAM,KAAK,IAAI,oBAAU,GAAG;AAE5B,QAAI;AACF,YAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,WAAG,GAAG,QAAQ,OAAO;AACrB,WAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,WAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,MAC/D,CAAC;AAED,YAAM,QAAQ,IAAI,CAAC,UAAU,GAAG,mBAAmB,EAAE,GAAG,SAAS,EAAE,CAAC,CAAC;AAAA,IACvE,SAAS,GAAG;AACV,YAAM,IAAI,MAAM,kCAAkC,CAAC,EAAE;AAAA,IACvD;AAAA,EACF;AACF;AAQA,MAAM,oBAAoB,CACxB,MACA,YAAqB,UACS;AAC9B,QAAM,QAAmC,CAAC;AAC1C,MAAI,OAAO,KAAK,UAAU,UAAU;AAClC,UAAM,OAAO;AACb,UAAM,KAAK,KAAK;AAAA,EAClB,OAAO;AACL,UAAM,OAAO;AACb,UAAM,YAAY,KAAK;AAAA,EACzB;AAEA,QAAM,gBAA2C,CAAC;AAClD,MAAI,KAAK,OAAO;AACd,kBAAc,QAAQ,KAAK;AAAA,EAC7B;AACA,MAAI,KAAK,SAAS;AAChB,kBAAc,UAAU,KAAK;AAAA,EAC/B;AAEA,MAAI,OAAO,KAAK,aAAa,EAAE,QAAQ;AACrC,UAAM,0BAA0B;AAAA,EAClC;AAEA,QAAM,SAAoC;AAAA,IACxC,UAAU,KAAK;AAAA,IACf;AAAA,IACA,eAAe;AAAA,MACb,WAAW;AAAA,MACX,UAAU,KAAK;AAAA,MACf,aAAa,KAAK;AAAA,IACpB;AAAA,IACA,UAAU,KAAK;AAAA,EACjB;AAEA,MAAI,aAAa,KAAK,mBAAmB,OAAO;AAC9C,WAAO,iBAAiB;AAAA,EAC1B;AAEA,SAAO;AACT;","names":["tts","ws"]}
1
+ {"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n APIConnectionError,\n APITimeoutError,\n AudioByteStream,\n Future,\n type TimedString,\n createTimedString,\n log,\n shortuuid,\n stream,\n tokenize,\n tts,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { request } from 'node:https';\nimport { type RawData, WebSocket } from 'ws';\nimport {\n TTSDefaultVoiceId,\n type TTSEncoding,\n type TTSModels,\n type TTSVoiceEmotion,\n type TTSVoiceSpeed,\n} from './models.js';\nimport {\n type CartesiaServerMessage,\n cartesiaMessageSchema,\n hasWordTimestamps,\n isChunkMessage,\n isDoneMessage,\n isErrorMessage,\n} from './types.js';\n\nconst AUTHORIZATION_HEADER = 'X-API-Key';\nconst VERSION_HEADER = 'Cartesia-Version';\nconst VERSION = '2024-06-10';\nconst NUM_CHANNELS = 1;\nconst BUFFERED_WORDS_COUNT = 8;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n encoding: TTSEncoding;\n sampleRate: number;\n voice: string | number[];\n speed?: TTSVoiceSpeed | number;\n emotion?: (TTSVoiceEmotion | string)[];\n apiKey?: string;\n language: string;\n baseUrl: string;\n\n /**\n * The timeout for the next chunk to be received from the Cartesia API.\n */\n chunkTimeout: number;\n\n /**\n * Whether to add word timestamps to the output. When enabled, the TTS will return\n * timing information for each word in the transcript.\n * @defaultValue true\n */\n wordTimestamps?: boolean;\n\n pronunciationDictId?: string;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'sonic-2',\n encoding: 'pcm_s16le',\n sampleRate: 24000,\n voice: TTSDefaultVoiceId,\n apiKey: process.env.CARTESIA_API_KEY,\n language: 'en',\n baseUrl: 'https://api.cartesia.ai',\n chunkTimeout: 5000,\n wordTimestamps: true,\n};\n\nexport class TTS extends tts.TTS {\n #opts: TTSOptions;\n label = 'cartesia.TTS';\n\n constructor(opts: Partial<TTSOptions> = {}) {\n const resolvedOpts = {\n ...defaultTTSOptions,\n ...opts,\n };\n\n super(resolvedOpts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {\n streaming: true,\n alignedTranscript: resolvedOpts.wordTimestamps ?? true,\n });\n\n this.#opts = resolvedOpts;\n\n if (this.#opts.apiKey === undefined) {\n throw new Error(\n 'Cartesia API key is required, whether as an argument or as $CARTESIA_API_KEY',\n );\n }\n\n if ((this.#opts.speed || this.#opts.emotion) && this.#opts.model !== 'sonic-2-2025-03-07') {\n const logger = log();\n logger.warn(\n { model: this.#opts.model, speed: this.#opts.speed, emotion: this.#opts.emotion },\n \"speed and emotion controls are only supported for model 'sonic-2-2025-03-07', see https://docs.cartesia.ai/developer-tools/changelog for details\",\n );\n }\n }\n\n updateOptions(opts: Partial<TTSOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n if ((this.#opts.speed || this.#opts.emotion) && this.#opts.model !== 'sonic-2-2025-03-07') {\n const logger = log();\n logger.warn(\n { model: this.#opts.model, speed: this.#opts.speed, emotion: this.#opts.emotion },\n \"speed and emotion controls are only supported for model 'sonic-2-2025-03-07', see https://docs.cartesia.ai/developer-tools/changelog for details\",\n );\n }\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.#opts, connOptions, abortSignal);\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream {\n return new SynthesizeStream(this, this.#opts, options?.connOptions);\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'cartesia.ChunkedStream';\n #logger = log();\n #opts: TTSOptions;\n #text: string;\n\n constructor(\n tts: TTS,\n text: string,\n opts: TTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.#text = text;\n this.#opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.#opts.sampleRate, NUM_CHANNELS);\n const json = toCartesiaOptions(this.#opts);\n json.transcript = this.#text;\n\n const baseUrl = new URL(this.#opts.baseUrl);\n const doneFut = new Future<void>();\n\n const req = request(\n {\n hostname: baseUrl.hostname,\n port: parseInt(baseUrl.port) || (baseUrl.protocol === 'https:' ? 443 : 80),\n path: '/tts/bytes',\n method: 'POST',\n headers: {\n [AUTHORIZATION_HEADER]: this.#opts.apiKey!,\n [VERSION_HEADER]: VERSION,\n },\n signal: this.abortSignal,\n },\n (res) => {\n res.on('data', (chunk) => {\n for (const frame of bstream.write(chunk)) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n });\n res.on('close', () => {\n for (const frame of bstream.flush()) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n this.queue.close();\n if (!doneFut.done) doneFut.resolve();\n });\n res.on('error', (err) => {\n if (err.message === 'aborted') return;\n this.#logger.error({ err }, 'Cartesia TTS response error');\n if (!doneFut.done) doneFut.reject(err);\n });\n },\n );\n\n req.on('error', (err) => {\n if (err.name === 'AbortError') return;\n this.#logger.error({ err }, 'Cartesia TTS request error');\n if (!doneFut.done) doneFut.reject(err);\n });\n req.on('close', () => {\n if (!doneFut.done) doneFut.resolve();\n });\n req.write(JSON.stringify(json));\n req.end();\n\n try {\n await doneFut.await;\n } catch (e) {\n if (this.abortSignal.aborted) return;\n if (!this.queue.closed) this.queue.close();\n throw toRetryableConnectionError(e);\n }\n }\n}\n\nexport class SynthesizeStream extends tts.SynthesizeStream {\n #opts: TTSOptions;\n #logger = log();\n #tokenizer = new tokenize.basic.SentenceTokenizer({\n minSentenceLength: BUFFERED_WORDS_COUNT,\n }).stream();\n label = 'cartesia.SynthesizeStream';\n\n constructor(tts: TTS, opts: TTSOptions, connOptions?: APIConnectOptions) {\n super(tts, connOptions);\n this.#opts = opts;\n }\n\n updateOptions(opts: Partial<TTSOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n if ((this.#opts.speed || this.#opts.emotion) && this.#opts.model !== 'sonic-2-2025-03-07') {\n this.#logger.warn(\n { model: this.#opts.model, speed: this.#opts.speed, emotion: this.#opts.emotion },\n \"speed and emotion controls are only supported for model 'sonic-2-2025-03-07', see https://docs.cartesia.ai/developer-tools/changelog for details\",\n );\n }\n }\n\n protected async run() {\n const requestId = shortuuid();\n let closing = false;\n // Only close WebSocket when both: 1) Cartesia returns done, AND 2) all sentences have been sent\n let sentenceStreamClosed = false;\n\n const sentenceStreamTask = async (ws: WebSocket) => {\n const packet = toCartesiaOptions(this.#opts, true);\n for await (const event of this.#tokenizer) {\n const msg = {\n ...packet,\n context_id: requestId,\n transcript: event.token + ' ',\n continue: true,\n };\n ws.send(JSON.stringify(msg));\n }\n\n const endMsg = {\n ...packet,\n context_id: requestId,\n transcript: ' ',\n continue: false,\n };\n ws.send(JSON.stringify(endMsg));\n // Mark sentence stream as closed\n sentenceStreamClosed = true;\n };\n\n const inputTask = async () => {\n for await (const data of this.input) {\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n this.#tokenizer.flush();\n continue;\n }\n this.#tokenizer.pushText(data);\n }\n this.#tokenizer.endInput();\n this.#tokenizer.close();\n };\n\n // Use event channel and set up listeners ONCE to avoid missing messages during listener re-registration\n const recvTask = async (ws: WebSocket) => {\n const bstream = new AudioByteStream(this.#opts.sampleRate, NUM_CHANNELS);\n\n // Create event channel to buffer incoming messages\n // This prevents message loss between listener re-registrations\n const eventChannel = stream.createStreamChannel<RawData>();\n\n let lastFrame: AudioFrame | undefined;\n let pendingTimedTranscripts: TimedString[] = [];\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame && !this.queue.closed) {\n // Include timedTranscripts with the audio frame\n this.queue.put({\n requestId,\n segmentId,\n frame: lastFrame,\n final,\n timedTranscripts:\n pendingTimedTranscripts.length > 0 ? pendingTimedTranscripts : undefined,\n });\n lastFrame = undefined;\n pendingTimedTranscripts = [];\n }\n };\n\n let timeout: NodeJS.Timeout | null = null;\n\n const clearTTSChunkTimeout = () => {\n if (timeout) {\n clearTimeout(timeout);\n timeout = null;\n }\n };\n\n // Set up WebSocket listeners ONCE (not in a loop)\n const onMessage = (data: RawData) => {\n void eventChannel.write(data).catch((error: unknown) => {\n this.#logger.debug({ error }, 'Failed writing Cartesia event to channel (likely closed)');\n });\n };\n\n const onClose = (code: number, reason: Buffer) => {\n if (!closing) {\n this.#logger.debug(`WebSocket closed with code ${code}: ${reason.toString()}`);\n }\n clearTTSChunkTimeout();\n void eventChannel.close();\n };\n\n const onError = (err: Error) => {\n this.#logger.error({ err }, 'Cartesia WebSocket error');\n void eventChannel.close();\n };\n\n // Attach listeners ONCE\n ws.on('message', onMessage);\n ws.on('close', onClose);\n ws.on('error', onError);\n\n try {\n // Process messages from the channel\n const reader = eventChannel.stream().getReader();\n\n while (!this.closed && !this.abortController.signal.aborted) {\n const result = await reader.read();\n if (result.done) break;\n\n const rawMsg = result.value;\n\n // Parse message with Zod schema for type safety\n let serverMsg: CartesiaServerMessage;\n try {\n const json = JSON.parse(rawMsg.toString());\n serverMsg = cartesiaMessageSchema.parse(json);\n } catch (parseErr) {\n this.#logger.warn({ parseErr }, 'Failed to parse Cartesia message');\n continue;\n }\n\n // Handle error messages\n if (isErrorMessage(serverMsg)) {\n this.#logger.error({ error: serverMsg.error }, 'Cartesia returned error');\n continue;\n }\n\n const segmentId = serverMsg.context_id;\n\n // Process word timestamps if present (typed via Zod schema)\n if (this.#opts.wordTimestamps !== false && hasWordTimestamps(serverMsg)) {\n const wordTimestamps = serverMsg.word_timestamps;\n for (let i = 0; i < wordTimestamps.words.length; i++) {\n const word = wordTimestamps.words[i];\n const startTime = wordTimestamps.start[i];\n const endTime = wordTimestamps.end[i];\n if (word !== undefined && startTime !== undefined && endTime !== undefined) {\n pendingTimedTranscripts.push(\n createTimedString({\n text: word + ' ', // Add space after word for consistency\n startTime,\n endTime,\n }),\n );\n }\n }\n }\n\n // Handle audio chunk messages\n if (isChunkMessage(serverMsg)) {\n const audioBuffer = Buffer.from(serverMsg.data, 'base64');\n // Extract ArrayBuffer from Buffer for AudioByteStream compatibility\n const audioData = audioBuffer.buffer.slice(\n audioBuffer.byteOffset,\n audioBuffer.byteOffset + audioBuffer.byteLength,\n );\n for (const frame of bstream.write(audioData)) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n\n // IMPORTANT: close WS if TTS chunk stream been stuck too long\n // this allows unblock the current \"broken\" TTS node so that any future TTS nodes\n // can continue to process the stream without been blocked by the stuck node\n clearTTSChunkTimeout();\n timeout = setTimeout(() => {\n // cartesia chunk timeout quite often, so we make it a debug log\n this.#logger.debug(\n `Cartesia WebSocket TTS chunk stream timeout after ${this.#opts.chunkTimeout}ms`,\n );\n ws.close();\n }, this.#opts.chunkTimeout);\n } else if (isDoneMessage(serverMsg)) {\n // This ensures all sentences have been sent before closing\n if (sentenceStreamClosed) {\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n\n if (segmentId === requestId) {\n closing = true;\n clearTTSChunkTimeout();\n ws.close();\n break; // Exit the loop\n }\n }\n // If sentenceStreamClosed is false, continue receiving - more done messages will come\n }\n }\n } catch (err) {\n // skip log error for normal websocket close\n if (err instanceof Error && !err.message.includes('WebSocket closed')) {\n if (\n err.message.includes('Queue is closed') ||\n err.message.includes('Channel is closed')\n ) {\n this.#logger.warn(\n { err },\n 'Channel closed during transcript processing (expected during disconnect)',\n );\n } else {\n this.#logger.error({ err }, 'Error in recvTask from Cartesia WebSocket');\n }\n }\n } finally {\n // IMPORTANT: Remove listeners so connection can be reused\n ws.off('message', onMessage);\n ws.off('close', onClose);\n ws.off('error', onError);\n clearTTSChunkTimeout();\n }\n };\n\n const wsUrl = this.#opts.baseUrl.replace(/^http/, 'ws');\n const url = `${wsUrl}/tts/websocket?api_key=${this.#opts.apiKey}&cartesia_version=${VERSION}`;\n\n let ws: WebSocket | undefined;\n try {\n ws = await connectCartesiaWebSocket({\n url,\n timeoutMs: this.connOptions.timeoutMs,\n abortSignal: this.abortSignal,\n });\n await Promise.all([inputTask(), sentenceStreamTask(ws), recvTask(ws)]);\n } catch (e) {\n if (this.abortSignal.aborted) {\n return;\n }\n throw toRetryableConnectionError(e);\n } finally {\n // Ensure we don't leak sockets/tasks across retry attempts.\n if (ws && ws.readyState !== WebSocket.CLOSED) {\n safeTerminateWebSocket(ws);\n }\n }\n }\n}\n\nconst asError = (e: unknown): Error => (e instanceof Error ? e : new Error(String(e)));\n\nconst transientNetworkCodes = new Set([\n 'ETIMEDOUT',\n 'ECONNRESET',\n 'EAI_AGAIN',\n 'ENETUNREACH',\n 'ECONNREFUSED',\n 'EHOSTUNREACH',\n]);\n\nconst isRecord = (v: unknown): v is Record<string, unknown> => {\n return v !== null && typeof v === 'object';\n};\n\nconst isAggregateErrorLike = (e: unknown): e is { errors: unknown[]; name?: string } => {\n if (!isRecord(e)) return false;\n return e.name === 'AggregateError' && Array.isArray(e.errors);\n};\n\nconst hasErrorCode = (e: unknown, code: string): boolean => {\n if (isRecord(e) && e.code === code) return true;\n if (isAggregateErrorLike(e)) {\n return e.errors.some((inner) => hasErrorCode(inner, code));\n }\n return false;\n};\n\nconst hasAnyTransientCode = (e: unknown): boolean => {\n if (isRecord(e) && typeof e.code === 'string') {\n return transientNetworkCodes.has(e.code);\n }\n if (isAggregateErrorLike(e)) {\n return e.errors.some((inner) => hasAnyTransientCode(inner));\n }\n return false;\n};\n\nconst toRetryableConnectionError = (e: unknown): APIConnectionError => {\n const err = asError(e);\n const isTimeout =\n hasErrorCode(e, 'ETIMEDOUT') ||\n (typeof err.message === 'string' && err.message.includes('ETIMEDOUT'));\n const message = isTimeout\n ? `Cartesia connection timed out`\n : `Cartesia connection failed: ${err.message || 'unknown error'}`;\n return isTimeout ? new APITimeoutError({ message }) : new APIConnectionError({ message });\n};\n\nconst waitForWsOpen = async ({\n ws,\n timeoutMs,\n abortSignal,\n}: {\n ws: WebSocket;\n timeoutMs: number;\n abortSignal: AbortSignal;\n}) => {\n if (abortSignal.aborted) {\n throw new Error('aborted');\n }\n\n const fut = new Future<void>();\n let timeout: NodeJS.Timeout | undefined;\n\n const cleanup = () => {\n if (timeout) clearTimeout(timeout);\n ws.off('open', onOpen);\n ws.off('error', onError);\n ws.off('close', onClose);\n abortSignal.removeEventListener('abort', onAbort);\n };\n\n const onOpen = () => fut.resolve();\n const onError = (err: Error) => fut.reject(asError(err));\n const onClose = (code: number, reason: Buffer) =>\n fut.reject(\n new Error(`WebSocket closed before open (code=${code}, reason=${reason.toString()})`),\n );\n const onAbort = () => fut.reject(new Error('aborted'));\n\n ws.on('open', onOpen);\n ws.on('error', onError);\n ws.on('close', onClose);\n abortSignal.addEventListener('abort', onAbort, { once: true });\n\n if (timeoutMs > 0) {\n timeout = setTimeout(() => fut.reject(new Error('connect timeout')), timeoutMs);\n }\n\n try {\n await fut.await;\n } finally {\n cleanup();\n }\n};\n\nconst safeTerminateWebSocket = (ws: WebSocket) => {\n // `ws` can emit an 'error' event during teardown (especially if CONNECTING).\n // If there is no error listener at that moment, Node will treat it as unhandled and crash the process.\n try {\n ws.on('error', () => {});\n } catch {\n // ignore\n }\n\n try {\n // `terminate()` can throw if the socket was never established; `close()` is safer in CONNECTING.\n if (ws.readyState === WebSocket.CONNECTING) {\n ws.close();\n } else {\n ws.terminate();\n }\n } catch {\n // ignore\n }\n};\n\nconst connectCartesiaWebSocket = async ({\n url,\n timeoutMs,\n abortSignal,\n}: {\n url: string;\n timeoutMs: number;\n abortSignal: AbortSignal;\n}): Promise<WebSocket> => {\n const connectOnce = async (family?: number): Promise<WebSocket> => {\n const ws = new WebSocket(url, { handshakeTimeout: timeoutMs, family });\n try {\n await waitForWsOpen({ ws, timeoutMs, abortSignal });\n return ws;\n } catch (e) {\n safeTerminateWebSocket(ws);\n throw e;\n }\n };\n\n try {\n return await connectOnce();\n } catch (e) {\n // Mitigation for Node.js dual-stack (IPv6/IPv4) connect flakiness (\"happy eyeballs\"):\n // some environments surface `AggregateError` with nested `ETIMEDOUT` during the initial\n // WebSocket open. In that case we do a one-off retry forcing IPv4 (`family: 4`) before\n // letting the outer framework retry loop handle further attempts.\n //\n // If you still see `AggregateError`/`ETIMEDOUT`:\n // - Increase the session TTS connect timeout (`connOptions.ttsConnOptions.timeoutMs`)\n // - Or adjust Node's family autoselection behavior via `NODE_OPTIONS`, e.g.\n // `--network-family-autoselection-attempt-timeout=5000` (or disable it entirely).\n if (hasAnyTransientCode(e) || isAggregateErrorLike(e)) {\n return await connectOnce(4);\n }\n throw e;\n }\n};\n\n/**\n * Convert TTSOptions to Cartesia API format.\n *\n * @param opts - TTS options\n * @param streaming - Whether this is for streaming (WebSocket) or non-streaming (HTTP)\n */\nconst toCartesiaOptions = (\n opts: TTSOptions,\n streaming: boolean = false,\n): { [id: string]: unknown } => {\n const voice: { [id: string]: unknown } = {};\n if (typeof opts.voice === 'string') {\n voice.mode = 'id';\n voice.id = opts.voice;\n } else {\n voice.mode = 'embedding';\n voice.embedding = opts.voice;\n }\n\n const voiceControls: { [id: string]: unknown } = {};\n if (opts.speed) {\n voiceControls.speed = opts.speed;\n }\n if (opts.emotion) {\n voiceControls.emotion = opts.emotion;\n }\n\n if (Object.keys(voiceControls).length) {\n voice.__experimental_controls = voiceControls;\n }\n\n const result: { [id: string]: unknown } = {\n model_id: opts.model,\n voice,\n output_format: {\n container: 'raw',\n encoding: opts.encoding,\n sample_rate: opts.sampleRate,\n },\n language: opts.language,\n pronunciation_dict_id: opts.pronunciationDictId,\n };\n\n if (streaming && opts.wordTimestamps !== false) {\n result.add_timestamps = true;\n }\n\n return result;\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAaO;AAEP,wBAAwB;AACxB,gBAAwC;AACxC,oBAMO;AACP,mBAOO;AAEP,MAAM,uBAAuB;AAC7B,MAAM,iBAAiB;AACvB,MAAM,UAAU;AAChB,MAAM,eAAe;AACrB,MAAM,uBAAuB;AA4B7B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,OAAO;AAAA,EACP,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,SAAS;AAAA,EACT,cAAc;AAAA,EACd,gBAAgB;AAClB;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,QAAQ;AAAA,EAER,YAAY,OAA4B,CAAC,GAAG;AAC1C,UAAM,eAAe;AAAA,MACnB,GAAG;AAAA,MACH,GAAG;AAAA,IACL;AAEA,UAAM,aAAa,cAAc,kBAAkB,YAAY,cAAc;AAAA,MAC3E,WAAW;AAAA,MACX,mBAAmB,aAAa,kBAAkB;AAAA,IACpD,CAAC;AAED,SAAK,QAAQ;AAEb,QAAI,KAAK,MAAM,WAAW,QAAW;AACnC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,KAAK,MAAM,SAAS,KAAK,MAAM,YAAY,KAAK,MAAM,UAAU,sBAAsB;AACzF,YAAM,aAAS,mBAAI;AACnB,aAAO;AAAA,QACL,EAAE,OAAO,KAAK,MAAM,OAAO,OAAO,KAAK,MAAM,OAAO,SAAS,KAAK,MAAM,QAAQ;AAAA,QAChF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,SAAK,KAAK,MAAM,SAAS,KAAK,MAAM,YAAY,KAAK,MAAM,UAAU,sBAAsB;AACzF,YAAM,aAAS,mBAAI;AACnB,aAAO;AAAA,QACL,EAAE,OAAO,KAAK,MAAM,OAAO,OAAO,KAAK,MAAM,OAAO,SAAS,KAAK,MAAM,QAAQ;AAAA,QAChF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,WACE,MACA,aACA,aACmB;AACnB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,OAAO,aAAa,WAAW;AAAA,EAC3E;AAAA,EAEA,OAAO,SAAiE;AACtE,WAAO,IAAI,iBAAiB,MAAM,KAAK,OAAO,mCAAS,WAAW;AAAA,EACpE;AACF;AAEO,MAAM,sBAAsB,kBAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACR,cAAU,mBAAI;AAAA,EACd;AAAA,EACA;AAAA,EAEA,YACEA,MACA,MACA,MACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,QAAQ;AACb,SAAK,QAAQ;AAAA,EACf;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,gBAAY,yBAAU;AAC5B,UAAM,UAAU,IAAI,8BAAgB,KAAK,MAAM,YAAY,YAAY;AACvE,UAAM,OAAO,kBAAkB,KAAK,KAAK;AACzC,SAAK,aAAa,KAAK;AAEvB,UAAM,UAAU,IAAI,IAAI,KAAK,MAAM,OAAO;AAC1C,UAAM,UAAU,IAAI,qBAAa;AAEjC,UAAM,UAAM;AAAA,MACV;AAAA,QACE,UAAU,QAAQ;AAAA,QAClB,MAAM,SAAS,QAAQ,IAAI,MAAM,QAAQ,aAAa,WAAW,MAAM;AAAA,QACvE,MAAM;AAAA,QACN,QAAQ;AAAA,QACR,SAAS;AAAA,UACP,CAAC,oBAAoB,GAAG,KAAK,MAAM;AAAA,UACnC,CAAC,cAAc,GAAG;AAAA,QACpB;AAAA,QACA,QAAQ,KAAK;AAAA,MACf;AAAA,MACA,CAAC,QAAQ;AACP,YAAI,GAAG,QAAQ,CAAC,UAAU;AACxB,qBAAW,SAAS,QAAQ,MAAM,KAAK,GAAG;AACxC,iBAAK,MAAM,IAAI;AAAA,cACb;AAAA,cACA;AAAA,cACA,OAAO;AAAA,cACP,WAAW;AAAA,YACb,CAAC;AAAA,UACH;AAAA,QACF,CAAC;AACD,YAAI,GAAG,SAAS,MAAM;AACpB,qBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,iBAAK,MAAM,IAAI;AAAA,cACb;AAAA,cACA;AAAA,cACA,OAAO;AAAA,cACP,WAAW;AAAA,YACb,CAAC;AAAA,UACH;AACA,eAAK,MAAM,MAAM;AACjB,cAAI,CAAC,QAAQ,KAAM,SAAQ,QAAQ;AAAA,QACrC,CAAC;AACD,YAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,cAAI,IAAI,YAAY,UAAW;AAC/B,eAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6BAA6B;AACzD,cAAI,CAAC,QAAQ,KAAM,SAAQ,OAAO,GAAG;AAAA,QACvC,CAAC;AAAA,MACH;AAAA,IACF;AAEA,QAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,UAAI,IAAI,SAAS,aAAc;AAC/B,WAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,4BAA4B;AACxD,UAAI,CAAC,QAAQ,KAAM,SAAQ,OAAO,GAAG;AAAA,IACvC,CAAC;AACD,QAAI,GAAG,SAAS,MAAM;AACpB,UAAI,CAAC,QAAQ,KAAM,SAAQ,QAAQ;AAAA,IACrC,CAAC;AACD,QAAI,MAAM,KAAK,UAAU,IAAI,CAAC;AAC9B,QAAI,IAAI;AAER,QAAI;AACF,YAAM,QAAQ;AAAA,IAChB,SAAS,GAAG;AACV,UAAI,KAAK,YAAY,QAAS;AAC9B,UAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,YAAM,2BAA2B,CAAC;AAAA,IACpC;AAAA,EACF;AACF;AAEO,MAAM,yBAAyB,kBAAI,iBAAiB;AAAA,EACzD;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,aAAa,IAAI,uBAAS,MAAM,kBAAkB;AAAA,IAChD,mBAAmB;AAAA,EACrB,CAAC,EAAE,OAAO;AAAA,EACV,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,WAAW;AACtB,SAAK,QAAQ;AAAA,EACf;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,SAAK,KAAK,MAAM,SAAS,KAAK,MAAM,YAAY,KAAK,MAAM,UAAU,sBAAsB;AACzF,WAAK,QAAQ;AAAA,QACX,EAAE,OAAO,KAAK,MAAM,OAAO,OAAO,KAAK,MAAM,OAAO,SAAS,KAAK,MAAM,QAAQ;AAAA,QAChF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,gBAAY,yBAAU;AAC5B,QAAI,UAAU;AAEd,QAAI,uBAAuB;AAE3B,UAAM,qBAAqB,OAAOC,QAAkB;AAClD,YAAM,SAAS,kBAAkB,KAAK,OAAO,IAAI;AACjD,uBAAiB,SAAS,KAAK,YAAY;AACzC,cAAM,MAAM;AAAA,UACV,GAAG;AAAA,UACH,YAAY;AAAA,UACZ,YAAY,MAAM,QAAQ;AAAA,UAC1B,UAAU;AAAA,QACZ;AACA,QAAAA,IAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,MAC7B;AAEA,YAAM,SAAS;AAAA,QACb,GAAG;AAAA,QACH,YAAY;AAAA,QACZ,YAAY;AAAA,QACZ,UAAU;AAAA,MACZ;AACA,MAAAA,IAAG,KAAK,KAAK,UAAU,MAAM,CAAC;AAE9B,6BAAuB;AAAA,IACzB;AAEA,UAAM,YAAY,YAAY;AAC5B,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,eAAK,WAAW,MAAM;AACtB;AAAA,QACF;AACA,aAAK,WAAW,SAAS,IAAI;AAAA,MAC/B;AACA,WAAK,WAAW,SAAS;AACzB,WAAK,WAAW,MAAM;AAAA,IACxB;AAGA,UAAM,WAAW,OAAOA,QAAkB;AACxC,YAAM,UAAU,IAAI,8BAAgB,KAAK,MAAM,YAAY,YAAY;AAIvE,YAAM,eAAe,qBAAO,oBAA6B;AAEzD,UAAI;AACJ,UAAI,0BAAyC,CAAC;AAE9C,YAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,YAAI,aAAa,CAAC,KAAK,MAAM,QAAQ;AAEnC,eAAK,MAAM,IAAI;AAAA,YACb;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP;AAAA,YACA,kBACE,wBAAwB,SAAS,IAAI,0BAA0B;AAAA,UACnE,CAAC;AACD,sBAAY;AACZ,oCAA0B,CAAC;AAAA,QAC7B;AAAA,MACF;AAEA,UAAI,UAAiC;AAErC,YAAM,uBAAuB,MAAM;AACjC,YAAI,SAAS;AACX,uBAAa,OAAO;AACpB,oBAAU;AAAA,QACZ;AAAA,MACF;AAGA,YAAM,YAAY,CAAC,SAAkB;AACnC,aAAK,aAAa,MAAM,IAAI,EAAE,MAAM,CAAC,UAAmB;AACtD,eAAK,QAAQ,MAAM,EAAE,MAAM,GAAG,0DAA0D;AAAA,QAC1F,CAAC;AAAA,MACH;AAEA,YAAM,UAAU,CAAC,MAAc,WAAmB;AAChD,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,OAAO,SAAS,CAAC,EAAE;AAAA,QAC/E;AACA,6BAAqB;AACrB,aAAK,aAAa,MAAM;AAAA,MAC1B;AAEA,YAAM,UAAU,CAAC,QAAe;AAC9B,aAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,0BAA0B;AACtD,aAAK,aAAa,MAAM;AAAA,MAC1B;AAGA,MAAAA,IAAG,GAAG,WAAW,SAAS;AAC1B,MAAAA,IAAG,GAAG,SAAS,OAAO;AACtB,MAAAA,IAAG,GAAG,SAAS,OAAO;AAEtB,UAAI;AAEF,cAAM,SAAS,aAAa,OAAO,EAAE,UAAU;AAE/C,eAAO,CAAC,KAAK,UAAU,CAAC,KAAK,gBAAgB,OAAO,SAAS;AAC3D,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,OAAO,KAAM;AAEjB,gBAAM,SAAS,OAAO;AAGtB,cAAI;AACJ,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,OAAO,SAAS,CAAC;AACzC,wBAAY,mCAAsB,MAAM,IAAI;AAAA,UAC9C,SAAS,UAAU;AACjB,iBAAK,QAAQ,KAAK,EAAE,SAAS,GAAG,kCAAkC;AAClE;AAAA,UACF;AAGA,kBAAI,6BAAe,SAAS,GAAG;AAC7B,iBAAK,QAAQ,MAAM,EAAE,OAAO,UAAU,MAAM,GAAG,yBAAyB;AACxE;AAAA,UACF;AAEA,gBAAM,YAAY,UAAU;AAG5B,cAAI,KAAK,MAAM,mBAAmB,aAAS,gCAAkB,SAAS,GAAG;AACvE,kBAAM,iBAAiB,UAAU;AACjC,qBAAS,IAAI,GAAG,IAAI,eAAe,MAAM,QAAQ,KAAK;AACpD,oBAAM,OAAO,eAAe,MAAM,CAAC;AACnC,oBAAM,YAAY,eAAe,MAAM,CAAC;AACxC,oBAAM,UAAU,eAAe,IAAI,CAAC;AACpC,kBAAI,SAAS,UAAa,cAAc,UAAa,YAAY,QAAW;AAC1E,wCAAwB;AAAA,sBACtB,iCAAkB;AAAA,oBAChB,MAAM,OAAO;AAAA;AAAA,oBACb;AAAA,oBACA;AAAA,kBACF,CAAC;AAAA,gBACH;AAAA,cACF;AAAA,YACF;AAAA,UACF;AAGA,kBAAI,6BAAe,SAAS,GAAG;AAC7B,kBAAM,cAAc,OAAO,KAAK,UAAU,MAAM,QAAQ;AAExD,kBAAM,YAAY,YAAY,OAAO;AAAA,cACnC,YAAY;AAAA,cACZ,YAAY,aAAa,YAAY;AAAA,YACvC;AACA,uBAAW,SAAS,QAAQ,MAAM,SAAS,GAAG;AAC5C,4BAAc,WAAW,KAAK;AAC9B,0BAAY;AAAA,YACd;AAKA,iCAAqB;AACrB,sBAAU,WAAW,MAAM;AAEzB,mBAAK,QAAQ;AAAA,gBACX,qDAAqD,KAAK,MAAM,YAAY;AAAA,cAC9E;AACA,cAAAA,IAAG,MAAM;AAAA,YACX,GAAG,KAAK,MAAM,YAAY;AAAA,UAC5B,eAAW,4BAAc,SAAS,GAAG;AAEnC,gBAAI,sBAAsB;AACxB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,WAAW,KAAK;AAC9B,4BAAY;AAAA,cACd;AACA,4BAAc,WAAW,IAAI;AAC7B,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,cAC/C;AAEA,kBAAI,cAAc,WAAW;AAC3B,0BAAU;AACV,qCAAqB;AACrB,gBAAAA,IAAG,MAAM;AACT;AAAA,cACF;AAAA,YACF;AAAA,UAEF;AAAA,QACF;AAAA,MACF,SAAS,KAAK;AAEZ,YAAI,eAAe,SAAS,CAAC,IAAI,QAAQ,SAAS,kBAAkB,GAAG;AACrE,cACE,IAAI,QAAQ,SAAS,iBAAiB,KACtC,IAAI,QAAQ,SAAS,mBAAmB,GACxC;AACA,iBAAK,QAAQ;AAAA,cACX,EAAE,IAAI;AAAA,cACN;AAAA,YACF;AAAA,UACF,OAAO;AACL,iBAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,2CAA2C;AAAA,UACzE;AAAA,QACF;AAAA,MACF,UAAE;AAEA,QAAAA,IAAG,IAAI,WAAW,SAAS;AAC3B,QAAAA,IAAG,IAAI,SAAS,OAAO;AACvB,QAAAA,IAAG,IAAI,SAAS,OAAO;AACvB,6BAAqB;AAAA,MACvB;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,MAAM,QAAQ,QAAQ,SAAS,IAAI;AACtD,UAAM,MAAM,GAAG,KAAK,0BAA0B,KAAK,MAAM,MAAM,qBAAqB,OAAO;AAE3F,QAAI;AACJ,QAAI;AACF,WAAK,MAAM,yBAAyB;AAAA,QAClC;AAAA,QACA,WAAW,KAAK,YAAY;AAAA,QAC5B,aAAa,KAAK;AAAA,MACpB,CAAC;AACD,YAAM,QAAQ,IAAI,CAAC,UAAU,GAAG,mBAAmB,EAAE,GAAG,SAAS,EAAE,CAAC,CAAC;AAAA,IACvE,SAAS,GAAG;AACV,UAAI,KAAK,YAAY,SAAS;AAC5B;AAAA,MACF;AACA,YAAM,2BAA2B,CAAC;AAAA,IACpC,UAAE;AAEA,UAAI,MAAM,GAAG,eAAe,oBAAU,QAAQ;AAC5C,+BAAuB,EAAE;AAAA,MAC3B;AAAA,IACF;AAAA,EACF;AACF;AAEA,MAAM,UAAU,CAAC,MAAuB,aAAa,QAAQ,IAAI,IAAI,MAAM,OAAO,CAAC,CAAC;AAEpF,MAAM,wBAAwB,oBAAI,IAAI;AAAA,EACpC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF,CAAC;AAED,MAAM,WAAW,CAAC,MAA6C;AAC7D,SAAO,MAAM,QAAQ,OAAO,MAAM;AACpC;AAEA,MAAM,uBAAuB,CAAC,MAA0D;AACtF,MAAI,CAAC,SAAS,CAAC,EAAG,QAAO;AACzB,SAAO,EAAE,SAAS,oBAAoB,MAAM,QAAQ,EAAE,MAAM;AAC9D;AAEA,MAAM,eAAe,CAAC,GAAY,SAA0B;AAC1D,MAAI,SAAS,CAAC,KAAK,EAAE,SAAS,KAAM,QAAO;AAC3C,MAAI,qBAAqB,CAAC,GAAG;AAC3B,WAAO,EAAE,OAAO,KAAK,CAAC,UAAU,aAAa,OAAO,IAAI,CAAC;AAAA,EAC3D;AACA,SAAO;AACT;AAEA,MAAM,sBAAsB,CAAC,MAAwB;AACnD,MAAI,SAAS,CAAC,KAAK,OAAO,EAAE,SAAS,UAAU;AAC7C,WAAO,sBAAsB,IAAI,EAAE,IAAI;AAAA,EACzC;AACA,MAAI,qBAAqB,CAAC,GAAG;AAC3B,WAAO,EAAE,OAAO,KAAK,CAAC,UAAU,oBAAoB,KAAK,CAAC;AAAA,EAC5D;AACA,SAAO;AACT;AAEA,MAAM,6BAA6B,CAAC,MAAmC;AACrE,QAAM,MAAM,QAAQ,CAAC;AACrB,QAAM,YACJ,aAAa,GAAG,WAAW,KAC1B,OAAO,IAAI,YAAY,YAAY,IAAI,QAAQ,SAAS,WAAW;AACtE,QAAM,UAAU,YACZ,kCACA,+BAA+B,IAAI,WAAW,eAAe;AACjE,SAAO,YAAY,IAAI,8BAAgB,EAAE,QAAQ,CAAC,IAAI,IAAI,iCAAmB,EAAE,QAAQ,CAAC;AAC1F;AAEA,MAAM,gBAAgB,OAAO;AAAA,EAC3B;AAAA,EACA;AAAA,EACA;AACF,MAIM;AACJ,MAAI,YAAY,SAAS;AACvB,UAAM,IAAI,MAAM,SAAS;AAAA,EAC3B;AAEA,QAAM,MAAM,IAAI,qBAAa;AAC7B,MAAI;AAEJ,QAAM,UAAU,MAAM;AACpB,QAAI,QAAS,cAAa,OAAO;AACjC,OAAG,IAAI,QAAQ,MAAM;AACrB,OAAG,IAAI,SAAS,OAAO;AACvB,OAAG,IAAI,SAAS,OAAO;AACvB,gBAAY,oBAAoB,SAAS,OAAO;AAAA,EAClD;AAEA,QAAM,SAAS,MAAM,IAAI,QAAQ;AACjC,QAAM,UAAU,CAAC,QAAe,IAAI,OAAO,QAAQ,GAAG,CAAC;AACvD,QAAM,UAAU,CAAC,MAAc,WAC7B,IAAI;AAAA,IACF,IAAI,MAAM,sCAAsC,IAAI,YAAY,OAAO,SAAS,CAAC,GAAG;AAAA,EACtF;AACF,QAAM,UAAU,MAAM,IAAI,OAAO,IAAI,MAAM,SAAS,CAAC;AAErD,KAAG,GAAG,QAAQ,MAAM;AACpB,KAAG,GAAG,SAAS,OAAO;AACtB,KAAG,GAAG,SAAS,OAAO;AACtB,cAAY,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAE7D,MAAI,YAAY,GAAG;AACjB,cAAU,WAAW,MAAM,IAAI,OAAO,IAAI,MAAM,iBAAiB,CAAC,GAAG,SAAS;AAAA,EAChF;AAEA,MAAI;AACF,UAAM,IAAI;AAAA,EACZ,UAAE;AACA,YAAQ;AAAA,EACV;AACF;AAEA,MAAM,yBAAyB,CAAC,OAAkB;AAGhD,MAAI;AACF,OAAG,GAAG,SAAS,MAAM;AAAA,IAAC,CAAC;AAAA,EACzB,QAAQ;AAAA,EAER;AAEA,MAAI;AAEF,QAAI,GAAG,eAAe,oBAAU,YAAY;AAC1C,SAAG,MAAM;AAAA,IACX,OAAO;AACL,SAAG,UAAU;AAAA,IACf;AAAA,EACF,QAAQ;AAAA,EAER;AACF;AAEA,MAAM,2BAA2B,OAAO;AAAA,EACtC;AAAA,EACA;AAAA,EACA;AACF,MAI0B;AACxB,QAAM,cAAc,OAAO,WAAwC;AACjE,UAAM,KAAK,IAAI,oBAAU,KAAK,EAAE,kBAAkB,WAAW,OAAO,CAAC;AACrE,QAAI;AACF,YAAM,cAAc,EAAE,IAAI,WAAW,YAAY,CAAC;AAClD,aAAO;AAAA,IACT,SAAS,GAAG;AACV,6BAAuB,EAAE;AACzB,YAAM;AAAA,IACR;AAAA,EACF;AAEA,MAAI;AACF,WAAO,MAAM,YAAY;AAAA,EAC3B,SAAS,GAAG;AAUV,QAAI,oBAAoB,CAAC,KAAK,qBAAqB,CAAC,GAAG;AACrD,aAAO,MAAM,YAAY,CAAC;AAAA,IAC5B;AACA,UAAM;AAAA,EACR;AACF;AAQA,MAAM,oBAAoB,CACxB,MACA,YAAqB,UACS;AAC9B,QAAM,QAAmC,CAAC;AAC1C,MAAI,OAAO,KAAK,UAAU,UAAU;AAClC,UAAM,OAAO;AACb,UAAM,KAAK,KAAK;AAAA,EAClB,OAAO;AACL,UAAM,OAAO;AACb,UAAM,YAAY,KAAK;AAAA,EACzB;AAEA,QAAM,gBAA2C,CAAC;AAClD,MAAI,KAAK,OAAO;AACd,kBAAc,QAAQ,KAAK;AAAA,EAC7B;AACA,MAAI,KAAK,SAAS;AAChB,kBAAc,UAAU,KAAK;AAAA,EAC/B;AAEA,MAAI,OAAO,KAAK,aAAa,EAAE,QAAQ;AACrC,UAAM,0BAA0B;AAAA,EAClC;AAEA,QAAM,SAAoC;AAAA,IACxC,UAAU,KAAK;AAAA,IACf;AAAA,IACA,eAAe;AAAA,MACb,WAAW;AAAA,MACX,UAAU,KAAK;AAAA,MACf,aAAa,KAAK;AAAA,IACpB;AAAA,IACA,UAAU,KAAK;AAAA,IACf,uBAAuB,KAAK;AAAA,EAC9B;AAEA,MAAI,aAAa,KAAK,mBAAmB,OAAO;AAC9C,WAAO,iBAAiB;AAAA,EAC1B;AAEA,SAAO;AACT;","names":["tts","ws"]}
package/dist/tts.d.cts CHANGED
@@ -20,6 +20,7 @@ export interface TTSOptions {
20
20
  * @defaultValue true
21
21
  */
22
22
  wordTimestamps?: boolean;
23
+ pronunciationDictId?: string;
23
24
  }
24
25
  export declare class TTS extends tts.TTS {
25
26
  #private;
@@ -27,7 +28,9 @@ export declare class TTS extends tts.TTS {
27
28
  constructor(opts?: Partial<TTSOptions>);
28
29
  updateOptions(opts: Partial<TTSOptions>): void;
29
30
  synthesize(text: string, connOptions?: APIConnectOptions, abortSignal?: AbortSignal): tts.ChunkedStream;
30
- stream(): SynthesizeStream;
31
+ stream(options?: {
32
+ connOptions?: APIConnectOptions;
33
+ }): SynthesizeStream;
31
34
  }
32
35
  export declare class ChunkedStream extends tts.ChunkedStream {
33
36
  #private;
@@ -38,7 +41,7 @@ export declare class ChunkedStream extends tts.ChunkedStream {
38
41
  export declare class SynthesizeStream extends tts.SynthesizeStream {
39
42
  #private;
40
43
  label: string;
41
- constructor(tts: TTS, opts: TTSOptions);
44
+ constructor(tts: TTS, opts: TTSOptions, connOptions?: APIConnectOptions);
42
45
  updateOptions(opts: Partial<TTSOptions>): void;
43
46
  protected run(): Promise<void>;
44
47
  }
package/dist/tts.d.ts CHANGED
@@ -20,6 +20,7 @@ export interface TTSOptions {
20
20
  * @defaultValue true
21
21
  */
22
22
  wordTimestamps?: boolean;
23
+ pronunciationDictId?: string;
23
24
  }
24
25
  export declare class TTS extends tts.TTS {
25
26
  #private;
@@ -27,7 +28,9 @@ export declare class TTS extends tts.TTS {
27
28
  constructor(opts?: Partial<TTSOptions>);
28
29
  updateOptions(opts: Partial<TTSOptions>): void;
29
30
  synthesize(text: string, connOptions?: APIConnectOptions, abortSignal?: AbortSignal): tts.ChunkedStream;
30
- stream(): SynthesizeStream;
31
+ stream(options?: {
32
+ connOptions?: APIConnectOptions;
33
+ }): SynthesizeStream;
31
34
  }
32
35
  export declare class ChunkedStream extends tts.ChunkedStream {
33
36
  #private;
@@ -38,7 +41,7 @@ export declare class ChunkedStream extends tts.ChunkedStream {
38
41
  export declare class SynthesizeStream extends tts.SynthesizeStream {
39
42
  #private;
40
43
  label: string;
41
- constructor(tts: TTS, opts: TTSOptions);
44
+ constructor(tts: TTS, opts: TTSOptions, connOptions?: APIConnectOptions);
42
45
  updateOptions(opts: Partial<TTSOptions>): void;
43
46
  protected run(): Promise<void>;
44
47
  }
package/dist/tts.d.ts.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../src/tts.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,EAStB,GAAG,EACJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,EAEL,KAAK,WAAW,EAChB,KAAK,SAAS,EACd,KAAK,eAAe,EACpB,KAAK,aAAa,EACnB,MAAM,aAAa,CAAC;AAgBrB,MAAM,WAAW,UAAU;IACzB,KAAK,EAAE,SAAS,GAAG,MAAM,CAAC;IAC1B,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,KAAK,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;IACzB,KAAK,CAAC,EAAE,aAAa,GAAG,MAAM,CAAC;IAC/B,OAAO,CAAC,EAAE,CAAC,eAAe,GAAG,MAAM,CAAC,EAAE,CAAC;IACvC,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,QAAQ,EAAE,MAAM,CAAC;IACjB,OAAO,EAAE,MAAM,CAAC;IAEhB;;OAEG;IACH,YAAY,EAAE,MAAM,CAAC;IAErB;;;;OAIG;IACH,cAAc,CAAC,EAAE,OAAO,CAAC;CAC1B;AAcD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAE9B,KAAK,SAAkB;gBAEX,IAAI,GAAE,OAAO,CAAC,UAAU,CAAM;IA4B1C,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAYvC,UAAU,CACR,IAAI,EAAE,MAAM,EACZ,WAAW,CAAC,EAAE,iBAAiB,EAC/B,WAAW,CAAC,EAAE,WAAW,GACxB,GAAG,CAAC,aAAa;IAIpB,MAAM,IAAI,gBAAgB;CAG3B;AAED,qBAAa,aAAc,SAAQ,GAAG,CAAC,aAAa;;IAClD,KAAK,SAA4B;gBAM/B,GAAG,EAAE,GAAG,EACR,IAAI,EAAE,MAAM,EACZ,IAAI,EAAE,UAAU,EAChB,WAAW,CAAC,EAAE,iBAAiB,EAC/B,WAAW,CAAC,EAAE,WAAW;cAOX,GAAG;CA6DpB;AAED,qBAAa,gBAAiB,SAAQ,GAAG,CAAC,gBAAgB;;IAMxD,KAAK,SAA+B;gBAExB,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU;IAKtC,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;cAWvB,GAAG;CA2OpB"}
1
+ {"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../src/tts.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,EAWtB,GAAG,EACJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,EAEL,KAAK,WAAW,EAChB,KAAK,SAAS,EACd,KAAK,eAAe,EACpB,KAAK,aAAa,EACnB,MAAM,aAAa,CAAC;AAgBrB,MAAM,WAAW,UAAU;IACzB,KAAK,EAAE,SAAS,GAAG,MAAM,CAAC;IAC1B,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,KAAK,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;IACzB,KAAK,CAAC,EAAE,aAAa,GAAG,MAAM,CAAC;IAC/B,OAAO,CAAC,EAAE,CAAC,eAAe,GAAG,MAAM,CAAC,EAAE,CAAC;IACvC,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,QAAQ,EAAE,MAAM,CAAC;IACjB,OAAO,EAAE,MAAM,CAAC;IAEhB;;OAEG;IACH,YAAY,EAAE,MAAM,CAAC;IAErB;;;;OAIG;IACH,cAAc,CAAC,EAAE,OAAO,CAAC;IAEzB,mBAAmB,CAAC,EAAE,MAAM,CAAC;CAC9B;AAcD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAE9B,KAAK,SAAkB;gBAEX,IAAI,GAAE,OAAO,CAAC,UAAU,CAAM;IA4B1C,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAYvC,UAAU,CACR,IAAI,EAAE,MAAM,EACZ,WAAW,CAAC,EAAE,iBAAiB,EAC/B,WAAW,CAAC,EAAE,WAAW,GACxB,GAAG,CAAC,aAAa;IAIpB,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,gBAAgB;CAGxE;AAED,qBAAa,aAAc,SAAQ,GAAG,CAAC,aAAa;;IAClD,KAAK,SAA4B;gBAM/B,GAAG,EAAE,GAAG,EACR,IAAI,EAAE,MAAM,EACZ,IAAI,EAAE,UAAU,EAChB,WAAW,CAAC,EAAE,iBAAiB,EAC/B,WAAW,CAAC,EAAE,WAAW;cAOX,GAAG;CAuEpB;AAED,qBAAa,gBAAiB,SAAQ,GAAG,CAAC,gBAAgB;;IAMxD,KAAK,SAA+B;gBAExB,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU,EAAE,WAAW,CAAC,EAAE,iBAAiB;IAKvE,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;cAWvB,GAAG;CAkPpB"}
package/dist/tts.js CHANGED
@@ -1,4 +1,6 @@
1
1
  import {
2
+ APIConnectionError,
3
+ APITimeoutError,
2
4
  AudioByteStream,
3
5
  Future,
4
6
  createTimedString,
@@ -75,8 +77,8 @@ class TTS extends tts.TTS {
75
77
  synthesize(text, connOptions, abortSignal) {
76
78
  return new ChunkedStream(this, text, this.#opts, connOptions, abortSignal);
77
79
  }
78
- stream() {
79
- return new SynthesizeStream(this, this.#opts);
80
+ stream(options) {
81
+ return new SynthesizeStream(this, this.#opts, options == null ? void 0 : options.connOptions);
80
82
  }
81
83
  }
82
84
  class ChunkedStream extends tts.ChunkedStream {
@@ -129,22 +131,32 @@ class ChunkedStream extends tts.ChunkedStream {
129
131
  });
130
132
  }
131
133
  this.queue.close();
132
- doneFut.resolve();
134
+ if (!doneFut.done) doneFut.resolve();
133
135
  });
134
136
  res.on("error", (err) => {
135
137
  if (err.message === "aborted") return;
136
138
  this.#logger.error({ err }, "Cartesia TTS response error");
139
+ if (!doneFut.done) doneFut.reject(err);
137
140
  });
138
141
  }
139
142
  );
140
143
  req.on("error", (err) => {
141
144
  if (err.name === "AbortError") return;
142
145
  this.#logger.error({ err }, "Cartesia TTS request error");
146
+ if (!doneFut.done) doneFut.reject(err);
147
+ });
148
+ req.on("close", () => {
149
+ if (!doneFut.done) doneFut.resolve();
143
150
  });
144
- req.on("close", () => doneFut.resolve());
145
151
  req.write(JSON.stringify(json));
146
152
  req.end();
147
- await doneFut.await;
153
+ try {
154
+ await doneFut.await;
155
+ } catch (e) {
156
+ if (this.abortSignal.aborted) return;
157
+ if (!this.queue.closed) this.queue.close();
158
+ throw toRetryableConnectionError(e);
159
+ }
148
160
  }
149
161
  }
150
162
  class SynthesizeStream extends tts.SynthesizeStream {
@@ -154,8 +166,8 @@ class SynthesizeStream extends tts.SynthesizeStream {
154
166
  minSentenceLength: BUFFERED_WORDS_COUNT
155
167
  }).stream();
156
168
  label = "cartesia.SynthesizeStream";
157
- constructor(tts2, opts) {
158
- super(tts2);
169
+ constructor(tts2, opts, connOptions) {
170
+ super(tts2, connOptions);
159
171
  this.#opts = opts;
160
172
  }
161
173
  updateOptions(opts) {
@@ -339,19 +351,139 @@ class SynthesizeStream extends tts.SynthesizeStream {
339
351
  };
340
352
  const wsUrl = this.#opts.baseUrl.replace(/^http/, "ws");
341
353
  const url = `${wsUrl}/tts/websocket?api_key=${this.#opts.apiKey}&cartesia_version=${VERSION}`;
342
- const ws = new WebSocket(url);
354
+ let ws;
343
355
  try {
344
- await new Promise((resolve, reject) => {
345
- ws.on("open", resolve);
346
- ws.on("error", (error) => reject(error));
347
- ws.on("close", (code) => reject(`WebSocket returned ${code}`));
356
+ ws = await connectCartesiaWebSocket({
357
+ url,
358
+ timeoutMs: this.connOptions.timeoutMs,
359
+ abortSignal: this.abortSignal
348
360
  });
349
361
  await Promise.all([inputTask(), sentenceStreamTask(ws), recvTask(ws)]);
350
362
  } catch (e) {
351
- throw new Error(`failed to connect to Cartesia: ${e}`);
363
+ if (this.abortSignal.aborted) {
364
+ return;
365
+ }
366
+ throw toRetryableConnectionError(e);
367
+ } finally {
368
+ if (ws && ws.readyState !== WebSocket.CLOSED) {
369
+ safeTerminateWebSocket(ws);
370
+ }
352
371
  }
353
372
  }
354
373
  }
374
+ const asError = (e) => e instanceof Error ? e : new Error(String(e));
375
+ const transientNetworkCodes = /* @__PURE__ */ new Set([
376
+ "ETIMEDOUT",
377
+ "ECONNRESET",
378
+ "EAI_AGAIN",
379
+ "ENETUNREACH",
380
+ "ECONNREFUSED",
381
+ "EHOSTUNREACH"
382
+ ]);
383
+ const isRecord = (v) => {
384
+ return v !== null && typeof v === "object";
385
+ };
386
+ const isAggregateErrorLike = (e) => {
387
+ if (!isRecord(e)) return false;
388
+ return e.name === "AggregateError" && Array.isArray(e.errors);
389
+ };
390
+ const hasErrorCode = (e, code) => {
391
+ if (isRecord(e) && e.code === code) return true;
392
+ if (isAggregateErrorLike(e)) {
393
+ return e.errors.some((inner) => hasErrorCode(inner, code));
394
+ }
395
+ return false;
396
+ };
397
+ const hasAnyTransientCode = (e) => {
398
+ if (isRecord(e) && typeof e.code === "string") {
399
+ return transientNetworkCodes.has(e.code);
400
+ }
401
+ if (isAggregateErrorLike(e)) {
402
+ return e.errors.some((inner) => hasAnyTransientCode(inner));
403
+ }
404
+ return false;
405
+ };
406
+ const toRetryableConnectionError = (e) => {
407
+ const err = asError(e);
408
+ const isTimeout = hasErrorCode(e, "ETIMEDOUT") || typeof err.message === "string" && err.message.includes("ETIMEDOUT");
409
+ const message = isTimeout ? `Cartesia connection timed out` : `Cartesia connection failed: ${err.message || "unknown error"}`;
410
+ return isTimeout ? new APITimeoutError({ message }) : new APIConnectionError({ message });
411
+ };
412
+ const waitForWsOpen = async ({
413
+ ws,
414
+ timeoutMs,
415
+ abortSignal
416
+ }) => {
417
+ if (abortSignal.aborted) {
418
+ throw new Error("aborted");
419
+ }
420
+ const fut = new Future();
421
+ let timeout;
422
+ const cleanup = () => {
423
+ if (timeout) clearTimeout(timeout);
424
+ ws.off("open", onOpen);
425
+ ws.off("error", onError);
426
+ ws.off("close", onClose);
427
+ abortSignal.removeEventListener("abort", onAbort);
428
+ };
429
+ const onOpen = () => fut.resolve();
430
+ const onError = (err) => fut.reject(asError(err));
431
+ const onClose = (code, reason) => fut.reject(
432
+ new Error(`WebSocket closed before open (code=${code}, reason=${reason.toString()})`)
433
+ );
434
+ const onAbort = () => fut.reject(new Error("aborted"));
435
+ ws.on("open", onOpen);
436
+ ws.on("error", onError);
437
+ ws.on("close", onClose);
438
+ abortSignal.addEventListener("abort", onAbort, { once: true });
439
+ if (timeoutMs > 0) {
440
+ timeout = setTimeout(() => fut.reject(new Error("connect timeout")), timeoutMs);
441
+ }
442
+ try {
443
+ await fut.await;
444
+ } finally {
445
+ cleanup();
446
+ }
447
+ };
448
+ const safeTerminateWebSocket = (ws) => {
449
+ try {
450
+ ws.on("error", () => {
451
+ });
452
+ } catch {
453
+ }
454
+ try {
455
+ if (ws.readyState === WebSocket.CONNECTING) {
456
+ ws.close();
457
+ } else {
458
+ ws.terminate();
459
+ }
460
+ } catch {
461
+ }
462
+ };
463
+ const connectCartesiaWebSocket = async ({
464
+ url,
465
+ timeoutMs,
466
+ abortSignal
467
+ }) => {
468
+ const connectOnce = async (family) => {
469
+ const ws = new WebSocket(url, { handshakeTimeout: timeoutMs, family });
470
+ try {
471
+ await waitForWsOpen({ ws, timeoutMs, abortSignal });
472
+ return ws;
473
+ } catch (e) {
474
+ safeTerminateWebSocket(ws);
475
+ throw e;
476
+ }
477
+ };
478
+ try {
479
+ return await connectOnce();
480
+ } catch (e) {
481
+ if (hasAnyTransientCode(e) || isAggregateErrorLike(e)) {
482
+ return await connectOnce(4);
483
+ }
484
+ throw e;
485
+ }
486
+ };
355
487
  const toCartesiaOptions = (opts, streaming = false) => {
356
488
  const voice = {};
357
489
  if (typeof opts.voice === "string") {
@@ -379,7 +511,8 @@ const toCartesiaOptions = (opts, streaming = false) => {
379
511
  encoding: opts.encoding,
380
512
  sample_rate: opts.sampleRate
381
513
  },
382
- language: opts.language
514
+ language: opts.language,
515
+ pronunciation_dict_id: opts.pronunciationDictId
383
516
  };
384
517
  if (streaming && opts.wordTimestamps !== false) {
385
518
  result.add_timestamps = true;
package/dist/tts.js.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n AudioByteStream,\n Future,\n type TimedString,\n createTimedString,\n log,\n shortuuid,\n stream,\n tokenize,\n tts,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { request } from 'node:https';\nimport { type RawData, WebSocket } from 'ws';\nimport {\n TTSDefaultVoiceId,\n type TTSEncoding,\n type TTSModels,\n type TTSVoiceEmotion,\n type TTSVoiceSpeed,\n} from './models.js';\nimport {\n type CartesiaServerMessage,\n cartesiaMessageSchema,\n hasWordTimestamps,\n isChunkMessage,\n isDoneMessage,\n isErrorMessage,\n} from './types.js';\n\nconst AUTHORIZATION_HEADER = 'X-API-Key';\nconst VERSION_HEADER = 'Cartesia-Version';\nconst VERSION = '2024-06-10';\nconst NUM_CHANNELS = 1;\nconst BUFFERED_WORDS_COUNT = 8;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n encoding: TTSEncoding;\n sampleRate: number;\n voice: string | number[];\n speed?: TTSVoiceSpeed | number;\n emotion?: (TTSVoiceEmotion | string)[];\n apiKey?: string;\n language: string;\n baseUrl: string;\n\n /**\n * The timeout for the next chunk to be received from the Cartesia API.\n */\n chunkTimeout: number;\n\n /**\n * Whether to add word timestamps to the output. When enabled, the TTS will return\n * timing information for each word in the transcript.\n * @defaultValue true\n */\n wordTimestamps?: boolean;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'sonic-2',\n encoding: 'pcm_s16le',\n sampleRate: 24000,\n voice: TTSDefaultVoiceId,\n apiKey: process.env.CARTESIA_API_KEY,\n language: 'en',\n baseUrl: 'https://api.cartesia.ai',\n chunkTimeout: 5000,\n wordTimestamps: true,\n};\n\nexport class TTS extends tts.TTS {\n #opts: TTSOptions;\n label = 'cartesia.TTS';\n\n constructor(opts: Partial<TTSOptions> = {}) {\n const resolvedOpts = {\n ...defaultTTSOptions,\n ...opts,\n };\n\n super(resolvedOpts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {\n streaming: true,\n alignedTranscript: resolvedOpts.wordTimestamps ?? true,\n });\n\n this.#opts = resolvedOpts;\n\n if (this.#opts.apiKey === undefined) {\n throw new Error(\n 'Cartesia API key is required, whether as an argument or as $CARTESIA_API_KEY',\n );\n }\n\n if ((this.#opts.speed || this.#opts.emotion) && this.#opts.model !== 'sonic-2-2025-03-07') {\n const logger = log();\n logger.warn(\n { model: this.#opts.model, speed: this.#opts.speed, emotion: this.#opts.emotion },\n \"speed and emotion controls are only supported for model 'sonic-2-2025-03-07', see https://docs.cartesia.ai/developer-tools/changelog for details\",\n );\n }\n }\n\n updateOptions(opts: Partial<TTSOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n if ((this.#opts.speed || this.#opts.emotion) && this.#opts.model !== 'sonic-2-2025-03-07') {\n const logger = log();\n logger.warn(\n { model: this.#opts.model, speed: this.#opts.speed, emotion: this.#opts.emotion },\n \"speed and emotion controls are only supported for model 'sonic-2-2025-03-07', see https://docs.cartesia.ai/developer-tools/changelog for details\",\n );\n }\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.#opts, connOptions, abortSignal);\n }\n\n stream(): SynthesizeStream {\n return new SynthesizeStream(this, this.#opts);\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'cartesia.ChunkedStream';\n #logger = log();\n #opts: TTSOptions;\n #text: string;\n\n constructor(\n tts: TTS,\n text: string,\n opts: TTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.#text = text;\n this.#opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.#opts.sampleRate, NUM_CHANNELS);\n const json = toCartesiaOptions(this.#opts);\n json.transcript = this.#text;\n\n const baseUrl = new URL(this.#opts.baseUrl);\n const doneFut = new Future<void>();\n\n const req = request(\n {\n hostname: baseUrl.hostname,\n port: parseInt(baseUrl.port) || (baseUrl.protocol === 'https:' ? 443 : 80),\n path: '/tts/bytes',\n method: 'POST',\n headers: {\n [AUTHORIZATION_HEADER]: this.#opts.apiKey!,\n [VERSION_HEADER]: VERSION,\n },\n signal: this.abortSignal,\n },\n (res) => {\n res.on('data', (chunk) => {\n for (const frame of bstream.write(chunk)) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n });\n res.on('close', () => {\n for (const frame of bstream.flush()) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n this.queue.close();\n doneFut.resolve();\n });\n res.on('error', (err) => {\n if (err.message === 'aborted') return;\n this.#logger.error({ err }, 'Cartesia TTS response error');\n });\n },\n );\n\n req.on('error', (err) => {\n if (err.name === 'AbortError') return;\n this.#logger.error({ err }, 'Cartesia TTS request error');\n });\n req.on('close', () => doneFut.resolve());\n req.write(JSON.stringify(json));\n req.end();\n\n await doneFut.await;\n }\n}\n\nexport class SynthesizeStream extends tts.SynthesizeStream {\n #opts: TTSOptions;\n #logger = log();\n #tokenizer = new tokenize.basic.SentenceTokenizer({\n minSentenceLength: BUFFERED_WORDS_COUNT,\n }).stream();\n label = 'cartesia.SynthesizeStream';\n\n constructor(tts: TTS, opts: TTSOptions) {\n super(tts);\n this.#opts = opts;\n }\n\n updateOptions(opts: Partial<TTSOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n if ((this.#opts.speed || this.#opts.emotion) && this.#opts.model !== 'sonic-2-2025-03-07') {\n this.#logger.warn(\n { model: this.#opts.model, speed: this.#opts.speed, emotion: this.#opts.emotion },\n \"speed and emotion controls are only supported for model 'sonic-2-2025-03-07', see https://docs.cartesia.ai/developer-tools/changelog for details\",\n );\n }\n }\n\n protected async run() {\n const requestId = shortuuid();\n let closing = false;\n // Only close WebSocket when both: 1) Cartesia returns done, AND 2) all sentences have been sent\n let sentenceStreamClosed = false;\n\n const sentenceStreamTask = async (ws: WebSocket) => {\n const packet = toCartesiaOptions(this.#opts, true);\n for await (const event of this.#tokenizer) {\n const msg = {\n ...packet,\n context_id: requestId,\n transcript: event.token + ' ',\n continue: true,\n };\n ws.send(JSON.stringify(msg));\n }\n\n const endMsg = {\n ...packet,\n context_id: requestId,\n transcript: ' ',\n continue: false,\n };\n ws.send(JSON.stringify(endMsg));\n // Mark sentence stream as closed\n sentenceStreamClosed = true;\n };\n\n const inputTask = async () => {\n for await (const data of this.input) {\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n this.#tokenizer.flush();\n continue;\n }\n this.#tokenizer.pushText(data);\n }\n this.#tokenizer.endInput();\n this.#tokenizer.close();\n };\n\n // Use event channel and set up listeners ONCE to avoid missing messages during listener re-registration\n const recvTask = async (ws: WebSocket) => {\n const bstream = new AudioByteStream(this.#opts.sampleRate, NUM_CHANNELS);\n\n // Create event channel to buffer incoming messages\n // This prevents message loss between listener re-registrations\n const eventChannel = stream.createStreamChannel<RawData>();\n\n let lastFrame: AudioFrame | undefined;\n let pendingTimedTranscripts: TimedString[] = [];\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame && !this.queue.closed) {\n // Include timedTranscripts with the audio frame\n this.queue.put({\n requestId,\n segmentId,\n frame: lastFrame,\n final,\n timedTranscripts:\n pendingTimedTranscripts.length > 0 ? pendingTimedTranscripts : undefined,\n });\n lastFrame = undefined;\n pendingTimedTranscripts = [];\n }\n };\n\n let timeout: NodeJS.Timeout | null = null;\n\n const clearTTSChunkTimeout = () => {\n if (timeout) {\n clearTimeout(timeout);\n timeout = null;\n }\n };\n\n // Set up WebSocket listeners ONCE (not in a loop)\n const onMessage = (data: RawData) => {\n void eventChannel.write(data).catch((error: unknown) => {\n this.#logger.debug({ error }, 'Failed writing Cartesia event to channel (likely closed)');\n });\n };\n\n const onClose = (code: number, reason: Buffer) => {\n if (!closing) {\n this.#logger.debug(`WebSocket closed with code ${code}: ${reason.toString()}`);\n }\n clearTTSChunkTimeout();\n void eventChannel.close();\n };\n\n const onError = (err: Error) => {\n this.#logger.error({ err }, 'Cartesia WebSocket error');\n void eventChannel.close();\n };\n\n // Attach listeners ONCE\n ws.on('message', onMessage);\n ws.on('close', onClose);\n ws.on('error', onError);\n\n try {\n // Process messages from the channel\n const reader = eventChannel.stream().getReader();\n\n while (!this.closed && !this.abortController.signal.aborted) {\n const result = await reader.read();\n if (result.done) break;\n\n const rawMsg = result.value;\n\n // Parse message with Zod schema for type safety\n let serverMsg: CartesiaServerMessage;\n try {\n const json = JSON.parse(rawMsg.toString());\n serverMsg = cartesiaMessageSchema.parse(json);\n } catch (parseErr) {\n this.#logger.warn({ parseErr }, 'Failed to parse Cartesia message');\n continue;\n }\n\n // Handle error messages\n if (isErrorMessage(serverMsg)) {\n this.#logger.error({ error: serverMsg.error }, 'Cartesia returned error');\n continue;\n }\n\n const segmentId = serverMsg.context_id;\n\n // Process word timestamps if present (typed via Zod schema)\n if (this.#opts.wordTimestamps !== false && hasWordTimestamps(serverMsg)) {\n const wordTimestamps = serverMsg.word_timestamps;\n for (let i = 0; i < wordTimestamps.words.length; i++) {\n const word = wordTimestamps.words[i];\n const startTime = wordTimestamps.start[i];\n const endTime = wordTimestamps.end[i];\n if (word !== undefined && startTime !== undefined && endTime !== undefined) {\n pendingTimedTranscripts.push(\n createTimedString({\n text: word + ' ', // Add space after word for consistency\n startTime,\n endTime,\n }),\n );\n }\n }\n }\n\n // Handle audio chunk messages\n if (isChunkMessage(serverMsg)) {\n const audioBuffer = Buffer.from(serverMsg.data, 'base64');\n // Extract ArrayBuffer from Buffer for AudioByteStream compatibility\n const audioData = audioBuffer.buffer.slice(\n audioBuffer.byteOffset,\n audioBuffer.byteOffset + audioBuffer.byteLength,\n );\n for (const frame of bstream.write(audioData)) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n\n // IMPORTANT: close WS if TTS chunk stream been stuck too long\n // this allows unblock the current \"broken\" TTS node so that any future TTS nodes\n // can continue to process the stream without been blocked by the stuck node\n clearTTSChunkTimeout();\n timeout = setTimeout(() => {\n // cartesia chunk timeout quite often, so we make it a debug log\n this.#logger.debug(\n `Cartesia WebSocket TTS chunk stream timeout after ${this.#opts.chunkTimeout}ms`,\n );\n ws.close();\n }, this.#opts.chunkTimeout);\n } else if (isDoneMessage(serverMsg)) {\n // This ensures all sentences have been sent before closing\n if (sentenceStreamClosed) {\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n\n if (segmentId === requestId) {\n closing = true;\n clearTTSChunkTimeout();\n ws.close();\n break; // Exit the loop\n }\n }\n // If sentenceStreamClosed is false, continue receiving - more done messages will come\n }\n }\n } catch (err) {\n // skip log error for normal websocket close\n if (err instanceof Error && !err.message.includes('WebSocket closed')) {\n if (\n err.message.includes('Queue is closed') ||\n err.message.includes('Channel is closed')\n ) {\n this.#logger.warn(\n { err },\n 'Channel closed during transcript processing (expected during disconnect)',\n );\n } else {\n this.#logger.error({ err }, 'Error in recvTask from Cartesia WebSocket');\n }\n }\n } finally {\n // IMPORTANT: Remove listeners so connection can be reused\n ws.off('message', onMessage);\n ws.off('close', onClose);\n ws.off('error', onError);\n clearTTSChunkTimeout();\n }\n };\n\n const wsUrl = this.#opts.baseUrl.replace(/^http/, 'ws');\n const url = `${wsUrl}/tts/websocket?api_key=${this.#opts.apiKey}&cartesia_version=${VERSION}`;\n const ws = new WebSocket(url);\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await Promise.all([inputTask(), sentenceStreamTask(ws), recvTask(ws)]);\n } catch (e) {\n throw new Error(`failed to connect to Cartesia: ${e}`);\n }\n }\n}\n\n/**\n * Convert TTSOptions to Cartesia API format.\n *\n * @param opts - TTS options\n * @param streaming - Whether this is for streaming (WebSocket) or non-streaming (HTTP)\n */\nconst toCartesiaOptions = (\n opts: TTSOptions,\n streaming: boolean = false,\n): { [id: string]: unknown } => {\n const voice: { [id: string]: unknown } = {};\n if (typeof opts.voice === 'string') {\n voice.mode = 'id';\n voice.id = opts.voice;\n } else {\n voice.mode = 'embedding';\n voice.embedding = opts.voice;\n }\n\n const voiceControls: { [id: string]: unknown } = {};\n if (opts.speed) {\n voiceControls.speed = opts.speed;\n }\n if (opts.emotion) {\n voiceControls.emotion = opts.emotion;\n }\n\n if (Object.keys(voiceControls).length) {\n voice.__experimental_controls = voiceControls;\n }\n\n const result: { [id: string]: unknown } = {\n model_id: opts.model,\n voice,\n output_format: {\n container: 'raw',\n encoding: opts.encoding,\n sample_rate: opts.sampleRate,\n },\n language: opts.language,\n };\n\n if (streaming && opts.wordTimestamps !== false) {\n result.add_timestamps = true;\n }\n\n return result;\n};\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EAEA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,eAAe;AACxB,SAAuB,iBAAiB;AACxC;AAAA,EACE;AAAA,OAKK;AACP;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,MAAM,uBAAuB;AAC7B,MAAM,iBAAiB;AACvB,MAAM,UAAU;AAChB,MAAM,eAAe;AACrB,MAAM,uBAAuB;AA0B7B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,OAAO;AAAA,EACP,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,SAAS;AAAA,EACT,cAAc;AAAA,EACd,gBAAgB;AAClB;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,QAAQ;AAAA,EAER,YAAY,OAA4B,CAAC,GAAG;AAC1C,UAAM,eAAe;AAAA,MACnB,GAAG;AAAA,MACH,GAAG;AAAA,IACL;AAEA,UAAM,aAAa,cAAc,kBAAkB,YAAY,cAAc;AAAA,MAC3E,WAAW;AAAA,MACX,mBAAmB,aAAa,kBAAkB;AAAA,IACpD,CAAC;AAED,SAAK,QAAQ;AAEb,QAAI,KAAK,MAAM,WAAW,QAAW;AACnC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,KAAK,MAAM,SAAS,KAAK,MAAM,YAAY,KAAK,MAAM,UAAU,sBAAsB;AACzF,YAAM,SAAS,IAAI;AACnB,aAAO;AAAA,QACL,EAAE,OAAO,KAAK,MAAM,OAAO,OAAO,KAAK,MAAM,OAAO,SAAS,KAAK,MAAM,QAAQ;AAAA,QAChF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,SAAK,KAAK,MAAM,SAAS,KAAK,MAAM,YAAY,KAAK,MAAM,UAAU,sBAAsB;AACzF,YAAM,SAAS,IAAI;AACnB,aAAO;AAAA,QACL,EAAE,OAAO,KAAK,MAAM,OAAO,OAAO,KAAK,MAAM,OAAO,SAAS,KAAK,MAAM,QAAQ;AAAA,QAChF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,WACE,MACA,aACA,aACmB;AACnB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,OAAO,aAAa,WAAW;AAAA,EAC3E;AAAA,EAEA,SAA2B;AACzB,WAAO,IAAI,iBAAiB,MAAM,KAAK,KAAK;AAAA,EAC9C;AACF;AAEO,MAAM,sBAAsB,IAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACR,UAAU,IAAI;AAAA,EACd;AAAA,EACA;AAAA,EAEA,YACEA,MACA,MACA,MACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,QAAQ;AACb,SAAK,QAAQ;AAAA,EACf;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,YAAY,UAAU;AAC5B,UAAM,UAAU,IAAI,gBAAgB,KAAK,MAAM,YAAY,YAAY;AACvE,UAAM,OAAO,kBAAkB,KAAK,KAAK;AACzC,SAAK,aAAa,KAAK;AAEvB,UAAM,UAAU,IAAI,IAAI,KAAK,MAAM,OAAO;AAC1C,UAAM,UAAU,IAAI,OAAa;AAEjC,UAAM,MAAM;AAAA,MACV;AAAA,QACE,UAAU,QAAQ;AAAA,QAClB,MAAM,SAAS,QAAQ,IAAI,MAAM,QAAQ,aAAa,WAAW,MAAM;AAAA,QACvE,MAAM;AAAA,QACN,QAAQ;AAAA,QACR,SAAS;AAAA,UACP,CAAC,oBAAoB,GAAG,KAAK,MAAM;AAAA,UACnC,CAAC,cAAc,GAAG;AAAA,QACpB;AAAA,QACA,QAAQ,KAAK;AAAA,MACf;AAAA,MACA,CAAC,QAAQ;AACP,YAAI,GAAG,QAAQ,CAAC,UAAU;AACxB,qBAAW,SAAS,QAAQ,MAAM,KAAK,GAAG;AACxC,iBAAK,MAAM,IAAI;AAAA,cACb;AAAA,cACA;AAAA,cACA,OAAO;AAAA,cACP,WAAW;AAAA,YACb,CAAC;AAAA,UACH;AAAA,QACF,CAAC;AACD,YAAI,GAAG,SAAS,MAAM;AACpB,qBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,iBAAK,MAAM,IAAI;AAAA,cACb;AAAA,cACA;AAAA,cACA,OAAO;AAAA,cACP,WAAW;AAAA,YACb,CAAC;AAAA,UACH;AACA,eAAK,MAAM,MAAM;AACjB,kBAAQ,QAAQ;AAAA,QAClB,CAAC;AACD,YAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,cAAI,IAAI,YAAY,UAAW;AAC/B,eAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6BAA6B;AAAA,QAC3D,CAAC;AAAA,MACH;AAAA,IACF;AAEA,QAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,UAAI,IAAI,SAAS,aAAc;AAC/B,WAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,4BAA4B;AAAA,IAC1D,CAAC;AACD,QAAI,GAAG,SAAS,MAAM,QAAQ,QAAQ,CAAC;AACvC,QAAI,MAAM,KAAK,UAAU,IAAI,CAAC;AAC9B,QAAI,IAAI;AAER,UAAM,QAAQ;AAAA,EAChB;AACF;AAEO,MAAM,yBAAyB,IAAI,iBAAiB;AAAA,EACzD;AAAA,EACA,UAAU,IAAI;AAAA,EACd,aAAa,IAAI,SAAS,MAAM,kBAAkB;AAAA,IAChD,mBAAmB;AAAA,EACrB,CAAC,EAAE,OAAO;AAAA,EACV,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,QAAQ;AAAA,EACf;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,SAAK,KAAK,MAAM,SAAS,KAAK,MAAM,YAAY,KAAK,MAAM,UAAU,sBAAsB;AACzF,WAAK,QAAQ;AAAA,QACX,EAAE,OAAO,KAAK,MAAM,OAAO,OAAO,KAAK,MAAM,OAAO,SAAS,KAAK,MAAM,QAAQ;AAAA,QAChF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,YAAY,UAAU;AAC5B,QAAI,UAAU;AAEd,QAAI,uBAAuB;AAE3B,UAAM,qBAAqB,OAAOC,QAAkB;AAClD,YAAM,SAAS,kBAAkB,KAAK,OAAO,IAAI;AACjD,uBAAiB,SAAS,KAAK,YAAY;AACzC,cAAM,MAAM;AAAA,UACV,GAAG;AAAA,UACH,YAAY;AAAA,UACZ,YAAY,MAAM,QAAQ;AAAA,UAC1B,UAAU;AAAA,QACZ;AACA,QAAAA,IAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,MAC7B;AAEA,YAAM,SAAS;AAAA,QACb,GAAG;AAAA,QACH,YAAY;AAAA,QACZ,YAAY;AAAA,QACZ,UAAU;AAAA,MACZ;AACA,MAAAA,IAAG,KAAK,KAAK,UAAU,MAAM,CAAC;AAE9B,6BAAuB;AAAA,IACzB;AAEA,UAAM,YAAY,YAAY;AAC5B,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,eAAK,WAAW,MAAM;AACtB;AAAA,QACF;AACA,aAAK,WAAW,SAAS,IAAI;AAAA,MAC/B;AACA,WAAK,WAAW,SAAS;AACzB,WAAK,WAAW,MAAM;AAAA,IACxB;AAGA,UAAM,WAAW,OAAOA,QAAkB;AACxC,YAAM,UAAU,IAAI,gBAAgB,KAAK,MAAM,YAAY,YAAY;AAIvE,YAAM,eAAe,OAAO,oBAA6B;AAEzD,UAAI;AACJ,UAAI,0BAAyC,CAAC;AAE9C,YAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,YAAI,aAAa,CAAC,KAAK,MAAM,QAAQ;AAEnC,eAAK,MAAM,IAAI;AAAA,YACb;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP;AAAA,YACA,kBACE,wBAAwB,SAAS,IAAI,0BAA0B;AAAA,UACnE,CAAC;AACD,sBAAY;AACZ,oCAA0B,CAAC;AAAA,QAC7B;AAAA,MACF;AAEA,UAAI,UAAiC;AAErC,YAAM,uBAAuB,MAAM;AACjC,YAAI,SAAS;AACX,uBAAa,OAAO;AACpB,oBAAU;AAAA,QACZ;AAAA,MACF;AAGA,YAAM,YAAY,CAAC,SAAkB;AACnC,aAAK,aAAa,MAAM,IAAI,EAAE,MAAM,CAAC,UAAmB;AACtD,eAAK,QAAQ,MAAM,EAAE,MAAM,GAAG,0DAA0D;AAAA,QAC1F,CAAC;AAAA,MACH;AAEA,YAAM,UAAU,CAAC,MAAc,WAAmB;AAChD,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,OAAO,SAAS,CAAC,EAAE;AAAA,QAC/E;AACA,6BAAqB;AACrB,aAAK,aAAa,MAAM;AAAA,MAC1B;AAEA,YAAM,UAAU,CAAC,QAAe;AAC9B,aAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,0BAA0B;AACtD,aAAK,aAAa,MAAM;AAAA,MAC1B;AAGA,MAAAA,IAAG,GAAG,WAAW,SAAS;AAC1B,MAAAA,IAAG,GAAG,SAAS,OAAO;AACtB,MAAAA,IAAG,GAAG,SAAS,OAAO;AAEtB,UAAI;AAEF,cAAM,SAAS,aAAa,OAAO,EAAE,UAAU;AAE/C,eAAO,CAAC,KAAK,UAAU,CAAC,KAAK,gBAAgB,OAAO,SAAS;AAC3D,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,OAAO,KAAM;AAEjB,gBAAM,SAAS,OAAO;AAGtB,cAAI;AACJ,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,OAAO,SAAS,CAAC;AACzC,wBAAY,sBAAsB,MAAM,IAAI;AAAA,UAC9C,SAAS,UAAU;AACjB,iBAAK,QAAQ,KAAK,EAAE,SAAS,GAAG,kCAAkC;AAClE;AAAA,UACF;AAGA,cAAI,eAAe,SAAS,GAAG;AAC7B,iBAAK,QAAQ,MAAM,EAAE,OAAO,UAAU,MAAM,GAAG,yBAAyB;AACxE;AAAA,UACF;AAEA,gBAAM,YAAY,UAAU;AAG5B,cAAI,KAAK,MAAM,mBAAmB,SAAS,kBAAkB,SAAS,GAAG;AACvE,kBAAM,iBAAiB,UAAU;AACjC,qBAAS,IAAI,GAAG,IAAI,eAAe,MAAM,QAAQ,KAAK;AACpD,oBAAM,OAAO,eAAe,MAAM,CAAC;AACnC,oBAAM,YAAY,eAAe,MAAM,CAAC;AACxC,oBAAM,UAAU,eAAe,IAAI,CAAC;AACpC,kBAAI,SAAS,UAAa,cAAc,UAAa,YAAY,QAAW;AAC1E,wCAAwB;AAAA,kBACtB,kBAAkB;AAAA,oBAChB,MAAM,OAAO;AAAA;AAAA,oBACb;AAAA,oBACA;AAAA,kBACF,CAAC;AAAA,gBACH;AAAA,cACF;AAAA,YACF;AAAA,UACF;AAGA,cAAI,eAAe,SAAS,GAAG;AAC7B,kBAAM,cAAc,OAAO,KAAK,UAAU,MAAM,QAAQ;AAExD,kBAAM,YAAY,YAAY,OAAO;AAAA,cACnC,YAAY;AAAA,cACZ,YAAY,aAAa,YAAY;AAAA,YACvC;AACA,uBAAW,SAAS,QAAQ,MAAM,SAAS,GAAG;AAC5C,4BAAc,WAAW,KAAK;AAC9B,0BAAY;AAAA,YACd;AAKA,iCAAqB;AACrB,sBAAU,WAAW,MAAM;AAEzB,mBAAK,QAAQ;AAAA,gBACX,qDAAqD,KAAK,MAAM,YAAY;AAAA,cAC9E;AACA,cAAAA,IAAG,MAAM;AAAA,YACX,GAAG,KAAK,MAAM,YAAY;AAAA,UAC5B,WAAW,cAAc,SAAS,GAAG;AAEnC,gBAAI,sBAAsB;AACxB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,WAAW,KAAK;AAC9B,4BAAY;AAAA,cACd;AACA,4BAAc,WAAW,IAAI;AAC7B,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,cAC/C;AAEA,kBAAI,cAAc,WAAW;AAC3B,0BAAU;AACV,qCAAqB;AACrB,gBAAAA,IAAG,MAAM;AACT;AAAA,cACF;AAAA,YACF;AAAA,UAEF;AAAA,QACF;AAAA,MACF,SAAS,KAAK;AAEZ,YAAI,eAAe,SAAS,CAAC,IAAI,QAAQ,SAAS,kBAAkB,GAAG;AACrE,cACE,IAAI,QAAQ,SAAS,iBAAiB,KACtC,IAAI,QAAQ,SAAS,mBAAmB,GACxC;AACA,iBAAK,QAAQ;AAAA,cACX,EAAE,IAAI;AAAA,cACN;AAAA,YACF;AAAA,UACF,OAAO;AACL,iBAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,2CAA2C;AAAA,UACzE;AAAA,QACF;AAAA,MACF,UAAE;AAEA,QAAAA,IAAG,IAAI,WAAW,SAAS;AAC3B,QAAAA,IAAG,IAAI,SAAS,OAAO;AACvB,QAAAA,IAAG,IAAI,SAAS,OAAO;AACvB,6BAAqB;AAAA,MACvB;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,MAAM,QAAQ,QAAQ,SAAS,IAAI;AACtD,UAAM,MAAM,GAAG,KAAK,0BAA0B,KAAK,MAAM,MAAM,qBAAqB,OAAO;AAC3F,UAAM,KAAK,IAAI,UAAU,GAAG;AAE5B,QAAI;AACF,YAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,WAAG,GAAG,QAAQ,OAAO;AACrB,WAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,WAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,MAC/D,CAAC;AAED,YAAM,QAAQ,IAAI,CAAC,UAAU,GAAG,mBAAmB,EAAE,GAAG,SAAS,EAAE,CAAC,CAAC;AAAA,IACvE,SAAS,GAAG;AACV,YAAM,IAAI,MAAM,kCAAkC,CAAC,EAAE;AAAA,IACvD;AAAA,EACF;AACF;AAQA,MAAM,oBAAoB,CACxB,MACA,YAAqB,UACS;AAC9B,QAAM,QAAmC,CAAC;AAC1C,MAAI,OAAO,KAAK,UAAU,UAAU;AAClC,UAAM,OAAO;AACb,UAAM,KAAK,KAAK;AAAA,EAClB,OAAO;AACL,UAAM,OAAO;AACb,UAAM,YAAY,KAAK;AAAA,EACzB;AAEA,QAAM,gBAA2C,CAAC;AAClD,MAAI,KAAK,OAAO;AACd,kBAAc,QAAQ,KAAK;AAAA,EAC7B;AACA,MAAI,KAAK,SAAS;AAChB,kBAAc,UAAU,KAAK;AAAA,EAC/B;AAEA,MAAI,OAAO,KAAK,aAAa,EAAE,QAAQ;AACrC,UAAM,0BAA0B;AAAA,EAClC;AAEA,QAAM,SAAoC;AAAA,IACxC,UAAU,KAAK;AAAA,IACf;AAAA,IACA,eAAe;AAAA,MACb,WAAW;AAAA,MACX,UAAU,KAAK;AAAA,MACf,aAAa,KAAK;AAAA,IACpB;AAAA,IACA,UAAU,KAAK;AAAA,EACjB;AAEA,MAAI,aAAa,KAAK,mBAAmB,OAAO;AAC9C,WAAO,iBAAiB;AAAA,EAC1B;AAEA,SAAO;AACT;","names":["tts","ws"]}
1
+ {"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n APIConnectionError,\n APITimeoutError,\n AudioByteStream,\n Future,\n type TimedString,\n createTimedString,\n log,\n shortuuid,\n stream,\n tokenize,\n tts,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { request } from 'node:https';\nimport { type RawData, WebSocket } from 'ws';\nimport {\n TTSDefaultVoiceId,\n type TTSEncoding,\n type TTSModels,\n type TTSVoiceEmotion,\n type TTSVoiceSpeed,\n} from './models.js';\nimport {\n type CartesiaServerMessage,\n cartesiaMessageSchema,\n hasWordTimestamps,\n isChunkMessage,\n isDoneMessage,\n isErrorMessage,\n} from './types.js';\n\nconst AUTHORIZATION_HEADER = 'X-API-Key';\nconst VERSION_HEADER = 'Cartesia-Version';\nconst VERSION = '2024-06-10';\nconst NUM_CHANNELS = 1;\nconst BUFFERED_WORDS_COUNT = 8;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n encoding: TTSEncoding;\n sampleRate: number;\n voice: string | number[];\n speed?: TTSVoiceSpeed | number;\n emotion?: (TTSVoiceEmotion | string)[];\n apiKey?: string;\n language: string;\n baseUrl: string;\n\n /**\n * The timeout for the next chunk to be received from the Cartesia API.\n */\n chunkTimeout: number;\n\n /**\n * Whether to add word timestamps to the output. When enabled, the TTS will return\n * timing information for each word in the transcript.\n * @defaultValue true\n */\n wordTimestamps?: boolean;\n\n pronunciationDictId?: string;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'sonic-2',\n encoding: 'pcm_s16le',\n sampleRate: 24000,\n voice: TTSDefaultVoiceId,\n apiKey: process.env.CARTESIA_API_KEY,\n language: 'en',\n baseUrl: 'https://api.cartesia.ai',\n chunkTimeout: 5000,\n wordTimestamps: true,\n};\n\nexport class TTS extends tts.TTS {\n #opts: TTSOptions;\n label = 'cartesia.TTS';\n\n constructor(opts: Partial<TTSOptions> = {}) {\n const resolvedOpts = {\n ...defaultTTSOptions,\n ...opts,\n };\n\n super(resolvedOpts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {\n streaming: true,\n alignedTranscript: resolvedOpts.wordTimestamps ?? true,\n });\n\n this.#opts = resolvedOpts;\n\n if (this.#opts.apiKey === undefined) {\n throw new Error(\n 'Cartesia API key is required, whether as an argument or as $CARTESIA_API_KEY',\n );\n }\n\n if ((this.#opts.speed || this.#opts.emotion) && this.#opts.model !== 'sonic-2-2025-03-07') {\n const logger = log();\n logger.warn(\n { model: this.#opts.model, speed: this.#opts.speed, emotion: this.#opts.emotion },\n \"speed and emotion controls are only supported for model 'sonic-2-2025-03-07', see https://docs.cartesia.ai/developer-tools/changelog for details\",\n );\n }\n }\n\n updateOptions(opts: Partial<TTSOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n if ((this.#opts.speed || this.#opts.emotion) && this.#opts.model !== 'sonic-2-2025-03-07') {\n const logger = log();\n logger.warn(\n { model: this.#opts.model, speed: this.#opts.speed, emotion: this.#opts.emotion },\n \"speed and emotion controls are only supported for model 'sonic-2-2025-03-07', see https://docs.cartesia.ai/developer-tools/changelog for details\",\n );\n }\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.#opts, connOptions, abortSignal);\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream {\n return new SynthesizeStream(this, this.#opts, options?.connOptions);\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'cartesia.ChunkedStream';\n #logger = log();\n #opts: TTSOptions;\n #text: string;\n\n constructor(\n tts: TTS,\n text: string,\n opts: TTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.#text = text;\n this.#opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.#opts.sampleRate, NUM_CHANNELS);\n const json = toCartesiaOptions(this.#opts);\n json.transcript = this.#text;\n\n const baseUrl = new URL(this.#opts.baseUrl);\n const doneFut = new Future<void>();\n\n const req = request(\n {\n hostname: baseUrl.hostname,\n port: parseInt(baseUrl.port) || (baseUrl.protocol === 'https:' ? 443 : 80),\n path: '/tts/bytes',\n method: 'POST',\n headers: {\n [AUTHORIZATION_HEADER]: this.#opts.apiKey!,\n [VERSION_HEADER]: VERSION,\n },\n signal: this.abortSignal,\n },\n (res) => {\n res.on('data', (chunk) => {\n for (const frame of bstream.write(chunk)) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n });\n res.on('close', () => {\n for (const frame of bstream.flush()) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n this.queue.close();\n if (!doneFut.done) doneFut.resolve();\n });\n res.on('error', (err) => {\n if (err.message === 'aborted') return;\n this.#logger.error({ err }, 'Cartesia TTS response error');\n if (!doneFut.done) doneFut.reject(err);\n });\n },\n );\n\n req.on('error', (err) => {\n if (err.name === 'AbortError') return;\n this.#logger.error({ err }, 'Cartesia TTS request error');\n if (!doneFut.done) doneFut.reject(err);\n });\n req.on('close', () => {\n if (!doneFut.done) doneFut.resolve();\n });\n req.write(JSON.stringify(json));\n req.end();\n\n try {\n await doneFut.await;\n } catch (e) {\n if (this.abortSignal.aborted) return;\n if (!this.queue.closed) this.queue.close();\n throw toRetryableConnectionError(e);\n }\n }\n}\n\nexport class SynthesizeStream extends tts.SynthesizeStream {\n #opts: TTSOptions;\n #logger = log();\n #tokenizer = new tokenize.basic.SentenceTokenizer({\n minSentenceLength: BUFFERED_WORDS_COUNT,\n }).stream();\n label = 'cartesia.SynthesizeStream';\n\n constructor(tts: TTS, opts: TTSOptions, connOptions?: APIConnectOptions) {\n super(tts, connOptions);\n this.#opts = opts;\n }\n\n updateOptions(opts: Partial<TTSOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n if ((this.#opts.speed || this.#opts.emotion) && this.#opts.model !== 'sonic-2-2025-03-07') {\n this.#logger.warn(\n { model: this.#opts.model, speed: this.#opts.speed, emotion: this.#opts.emotion },\n \"speed and emotion controls are only supported for model 'sonic-2-2025-03-07', see https://docs.cartesia.ai/developer-tools/changelog for details\",\n );\n }\n }\n\n protected async run() {\n const requestId = shortuuid();\n let closing = false;\n // Only close WebSocket when both: 1) Cartesia returns done, AND 2) all sentences have been sent\n let sentenceStreamClosed = false;\n\n const sentenceStreamTask = async (ws: WebSocket) => {\n const packet = toCartesiaOptions(this.#opts, true);\n for await (const event of this.#tokenizer) {\n const msg = {\n ...packet,\n context_id: requestId,\n transcript: event.token + ' ',\n continue: true,\n };\n ws.send(JSON.stringify(msg));\n }\n\n const endMsg = {\n ...packet,\n context_id: requestId,\n transcript: ' ',\n continue: false,\n };\n ws.send(JSON.stringify(endMsg));\n // Mark sentence stream as closed\n sentenceStreamClosed = true;\n };\n\n const inputTask = async () => {\n for await (const data of this.input) {\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n this.#tokenizer.flush();\n continue;\n }\n this.#tokenizer.pushText(data);\n }\n this.#tokenizer.endInput();\n this.#tokenizer.close();\n };\n\n // Use event channel and set up listeners ONCE to avoid missing messages during listener re-registration\n const recvTask = async (ws: WebSocket) => {\n const bstream = new AudioByteStream(this.#opts.sampleRate, NUM_CHANNELS);\n\n // Create event channel to buffer incoming messages\n // This prevents message loss between listener re-registrations\n const eventChannel = stream.createStreamChannel<RawData>();\n\n let lastFrame: AudioFrame | undefined;\n let pendingTimedTranscripts: TimedString[] = [];\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame && !this.queue.closed) {\n // Include timedTranscripts with the audio frame\n this.queue.put({\n requestId,\n segmentId,\n frame: lastFrame,\n final,\n timedTranscripts:\n pendingTimedTranscripts.length > 0 ? pendingTimedTranscripts : undefined,\n });\n lastFrame = undefined;\n pendingTimedTranscripts = [];\n }\n };\n\n let timeout: NodeJS.Timeout | null = null;\n\n const clearTTSChunkTimeout = () => {\n if (timeout) {\n clearTimeout(timeout);\n timeout = null;\n }\n };\n\n // Set up WebSocket listeners ONCE (not in a loop)\n const onMessage = (data: RawData) => {\n void eventChannel.write(data).catch((error: unknown) => {\n this.#logger.debug({ error }, 'Failed writing Cartesia event to channel (likely closed)');\n });\n };\n\n const onClose = (code: number, reason: Buffer) => {\n if (!closing) {\n this.#logger.debug(`WebSocket closed with code ${code}: ${reason.toString()}`);\n }\n clearTTSChunkTimeout();\n void eventChannel.close();\n };\n\n const onError = (err: Error) => {\n this.#logger.error({ err }, 'Cartesia WebSocket error');\n void eventChannel.close();\n };\n\n // Attach listeners ONCE\n ws.on('message', onMessage);\n ws.on('close', onClose);\n ws.on('error', onError);\n\n try {\n // Process messages from the channel\n const reader = eventChannel.stream().getReader();\n\n while (!this.closed && !this.abortController.signal.aborted) {\n const result = await reader.read();\n if (result.done) break;\n\n const rawMsg = result.value;\n\n // Parse message with Zod schema for type safety\n let serverMsg: CartesiaServerMessage;\n try {\n const json = JSON.parse(rawMsg.toString());\n serverMsg = cartesiaMessageSchema.parse(json);\n } catch (parseErr) {\n this.#logger.warn({ parseErr }, 'Failed to parse Cartesia message');\n continue;\n }\n\n // Handle error messages\n if (isErrorMessage(serverMsg)) {\n this.#logger.error({ error: serverMsg.error }, 'Cartesia returned error');\n continue;\n }\n\n const segmentId = serverMsg.context_id;\n\n // Process word timestamps if present (typed via Zod schema)\n if (this.#opts.wordTimestamps !== false && hasWordTimestamps(serverMsg)) {\n const wordTimestamps = serverMsg.word_timestamps;\n for (let i = 0; i < wordTimestamps.words.length; i++) {\n const word = wordTimestamps.words[i];\n const startTime = wordTimestamps.start[i];\n const endTime = wordTimestamps.end[i];\n if (word !== undefined && startTime !== undefined && endTime !== undefined) {\n pendingTimedTranscripts.push(\n createTimedString({\n text: word + ' ', // Add space after word for consistency\n startTime,\n endTime,\n }),\n );\n }\n }\n }\n\n // Handle audio chunk messages\n if (isChunkMessage(serverMsg)) {\n const audioBuffer = Buffer.from(serverMsg.data, 'base64');\n // Extract ArrayBuffer from Buffer for AudioByteStream compatibility\n const audioData = audioBuffer.buffer.slice(\n audioBuffer.byteOffset,\n audioBuffer.byteOffset + audioBuffer.byteLength,\n );\n for (const frame of bstream.write(audioData)) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n\n // IMPORTANT: close WS if TTS chunk stream been stuck too long\n // this allows unblock the current \"broken\" TTS node so that any future TTS nodes\n // can continue to process the stream without been blocked by the stuck node\n clearTTSChunkTimeout();\n timeout = setTimeout(() => {\n // cartesia chunk timeout quite often, so we make it a debug log\n this.#logger.debug(\n `Cartesia WebSocket TTS chunk stream timeout after ${this.#opts.chunkTimeout}ms`,\n );\n ws.close();\n }, this.#opts.chunkTimeout);\n } else if (isDoneMessage(serverMsg)) {\n // This ensures all sentences have been sent before closing\n if (sentenceStreamClosed) {\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n\n if (segmentId === requestId) {\n closing = true;\n clearTTSChunkTimeout();\n ws.close();\n break; // Exit the loop\n }\n }\n // If sentenceStreamClosed is false, continue receiving - more done messages will come\n }\n }\n } catch (err) {\n // skip log error for normal websocket close\n if (err instanceof Error && !err.message.includes('WebSocket closed')) {\n if (\n err.message.includes('Queue is closed') ||\n err.message.includes('Channel is closed')\n ) {\n this.#logger.warn(\n { err },\n 'Channel closed during transcript processing (expected during disconnect)',\n );\n } else {\n this.#logger.error({ err }, 'Error in recvTask from Cartesia WebSocket');\n }\n }\n } finally {\n // IMPORTANT: Remove listeners so connection can be reused\n ws.off('message', onMessage);\n ws.off('close', onClose);\n ws.off('error', onError);\n clearTTSChunkTimeout();\n }\n };\n\n const wsUrl = this.#opts.baseUrl.replace(/^http/, 'ws');\n const url = `${wsUrl}/tts/websocket?api_key=${this.#opts.apiKey}&cartesia_version=${VERSION}`;\n\n let ws: WebSocket | undefined;\n try {\n ws = await connectCartesiaWebSocket({\n url,\n timeoutMs: this.connOptions.timeoutMs,\n abortSignal: this.abortSignal,\n });\n await Promise.all([inputTask(), sentenceStreamTask(ws), recvTask(ws)]);\n } catch (e) {\n if (this.abortSignal.aborted) {\n return;\n }\n throw toRetryableConnectionError(e);\n } finally {\n // Ensure we don't leak sockets/tasks across retry attempts.\n if (ws && ws.readyState !== WebSocket.CLOSED) {\n safeTerminateWebSocket(ws);\n }\n }\n }\n}\n\nconst asError = (e: unknown): Error => (e instanceof Error ? e : new Error(String(e)));\n\nconst transientNetworkCodes = new Set([\n 'ETIMEDOUT',\n 'ECONNRESET',\n 'EAI_AGAIN',\n 'ENETUNREACH',\n 'ECONNREFUSED',\n 'EHOSTUNREACH',\n]);\n\nconst isRecord = (v: unknown): v is Record<string, unknown> => {\n return v !== null && typeof v === 'object';\n};\n\nconst isAggregateErrorLike = (e: unknown): e is { errors: unknown[]; name?: string } => {\n if (!isRecord(e)) return false;\n return e.name === 'AggregateError' && Array.isArray(e.errors);\n};\n\nconst hasErrorCode = (e: unknown, code: string): boolean => {\n if (isRecord(e) && e.code === code) return true;\n if (isAggregateErrorLike(e)) {\n return e.errors.some((inner) => hasErrorCode(inner, code));\n }\n return false;\n};\n\nconst hasAnyTransientCode = (e: unknown): boolean => {\n if (isRecord(e) && typeof e.code === 'string') {\n return transientNetworkCodes.has(e.code);\n }\n if (isAggregateErrorLike(e)) {\n return e.errors.some((inner) => hasAnyTransientCode(inner));\n }\n return false;\n};\n\nconst toRetryableConnectionError = (e: unknown): APIConnectionError => {\n const err = asError(e);\n const isTimeout =\n hasErrorCode(e, 'ETIMEDOUT') ||\n (typeof err.message === 'string' && err.message.includes('ETIMEDOUT'));\n const message = isTimeout\n ? `Cartesia connection timed out`\n : `Cartesia connection failed: ${err.message || 'unknown error'}`;\n return isTimeout ? new APITimeoutError({ message }) : new APIConnectionError({ message });\n};\n\nconst waitForWsOpen = async ({\n ws,\n timeoutMs,\n abortSignal,\n}: {\n ws: WebSocket;\n timeoutMs: number;\n abortSignal: AbortSignal;\n}) => {\n if (abortSignal.aborted) {\n throw new Error('aborted');\n }\n\n const fut = new Future<void>();\n let timeout: NodeJS.Timeout | undefined;\n\n const cleanup = () => {\n if (timeout) clearTimeout(timeout);\n ws.off('open', onOpen);\n ws.off('error', onError);\n ws.off('close', onClose);\n abortSignal.removeEventListener('abort', onAbort);\n };\n\n const onOpen = () => fut.resolve();\n const onError = (err: Error) => fut.reject(asError(err));\n const onClose = (code: number, reason: Buffer) =>\n fut.reject(\n new Error(`WebSocket closed before open (code=${code}, reason=${reason.toString()})`),\n );\n const onAbort = () => fut.reject(new Error('aborted'));\n\n ws.on('open', onOpen);\n ws.on('error', onError);\n ws.on('close', onClose);\n abortSignal.addEventListener('abort', onAbort, { once: true });\n\n if (timeoutMs > 0) {\n timeout = setTimeout(() => fut.reject(new Error('connect timeout')), timeoutMs);\n }\n\n try {\n await fut.await;\n } finally {\n cleanup();\n }\n};\n\nconst safeTerminateWebSocket = (ws: WebSocket) => {\n // `ws` can emit an 'error' event during teardown (especially if CONNECTING).\n // If there is no error listener at that moment, Node will treat it as unhandled and crash the process.\n try {\n ws.on('error', () => {});\n } catch {\n // ignore\n }\n\n try {\n // `terminate()` can throw if the socket was never established; `close()` is safer in CONNECTING.\n if (ws.readyState === WebSocket.CONNECTING) {\n ws.close();\n } else {\n ws.terminate();\n }\n } catch {\n // ignore\n }\n};\n\nconst connectCartesiaWebSocket = async ({\n url,\n timeoutMs,\n abortSignal,\n}: {\n url: string;\n timeoutMs: number;\n abortSignal: AbortSignal;\n}): Promise<WebSocket> => {\n const connectOnce = async (family?: number): Promise<WebSocket> => {\n const ws = new WebSocket(url, { handshakeTimeout: timeoutMs, family });\n try {\n await waitForWsOpen({ ws, timeoutMs, abortSignal });\n return ws;\n } catch (e) {\n safeTerminateWebSocket(ws);\n throw e;\n }\n };\n\n try {\n return await connectOnce();\n } catch (e) {\n // Mitigation for Node.js dual-stack (IPv6/IPv4) connect flakiness (\"happy eyeballs\"):\n // some environments surface `AggregateError` with nested `ETIMEDOUT` during the initial\n // WebSocket open. In that case we do a one-off retry forcing IPv4 (`family: 4`) before\n // letting the outer framework retry loop handle further attempts.\n //\n // If you still see `AggregateError`/`ETIMEDOUT`:\n // - Increase the session TTS connect timeout (`connOptions.ttsConnOptions.timeoutMs`)\n // - Or adjust Node's family autoselection behavior via `NODE_OPTIONS`, e.g.\n // `--network-family-autoselection-attempt-timeout=5000` (or disable it entirely).\n if (hasAnyTransientCode(e) || isAggregateErrorLike(e)) {\n return await connectOnce(4);\n }\n throw e;\n }\n};\n\n/**\n * Convert TTSOptions to Cartesia API format.\n *\n * @param opts - TTS options\n * @param streaming - Whether this is for streaming (WebSocket) or non-streaming (HTTP)\n */\nconst toCartesiaOptions = (\n opts: TTSOptions,\n streaming: boolean = false,\n): { [id: string]: unknown } => {\n const voice: { [id: string]: unknown } = {};\n if (typeof opts.voice === 'string') {\n voice.mode = 'id';\n voice.id = opts.voice;\n } else {\n voice.mode = 'embedding';\n voice.embedding = opts.voice;\n }\n\n const voiceControls: { [id: string]: unknown } = {};\n if (opts.speed) {\n voiceControls.speed = opts.speed;\n }\n if (opts.emotion) {\n voiceControls.emotion = opts.emotion;\n }\n\n if (Object.keys(voiceControls).length) {\n voice.__experimental_controls = voiceControls;\n }\n\n const result: { [id: string]: unknown } = {\n model_id: opts.model,\n voice,\n output_format: {\n container: 'raw',\n encoding: opts.encoding,\n sample_rate: opts.sampleRate,\n },\n language: opts.language,\n pronunciation_dict_id: opts.pronunciationDictId,\n };\n\n if (streaming && opts.wordTimestamps !== false) {\n result.add_timestamps = true;\n }\n\n return result;\n};\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,eAAe;AACxB,SAAuB,iBAAiB;AACxC;AAAA,EACE;AAAA,OAKK;AACP;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,MAAM,uBAAuB;AAC7B,MAAM,iBAAiB;AACvB,MAAM,UAAU;AAChB,MAAM,eAAe;AACrB,MAAM,uBAAuB;AA4B7B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,OAAO;AAAA,EACP,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,SAAS;AAAA,EACT,cAAc;AAAA,EACd,gBAAgB;AAClB;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,QAAQ;AAAA,EAER,YAAY,OAA4B,CAAC,GAAG;AAC1C,UAAM,eAAe;AAAA,MACnB,GAAG;AAAA,MACH,GAAG;AAAA,IACL;AAEA,UAAM,aAAa,cAAc,kBAAkB,YAAY,cAAc;AAAA,MAC3E,WAAW;AAAA,MACX,mBAAmB,aAAa,kBAAkB;AAAA,IACpD,CAAC;AAED,SAAK,QAAQ;AAEb,QAAI,KAAK,MAAM,WAAW,QAAW;AACnC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,KAAK,MAAM,SAAS,KAAK,MAAM,YAAY,KAAK,MAAM,UAAU,sBAAsB;AACzF,YAAM,SAAS,IAAI;AACnB,aAAO;AAAA,QACL,EAAE,OAAO,KAAK,MAAM,OAAO,OAAO,KAAK,MAAM,OAAO,SAAS,KAAK,MAAM,QAAQ;AAAA,QAChF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,SAAK,KAAK,MAAM,SAAS,KAAK,MAAM,YAAY,KAAK,MAAM,UAAU,sBAAsB;AACzF,YAAM,SAAS,IAAI;AACnB,aAAO;AAAA,QACL,EAAE,OAAO,KAAK,MAAM,OAAO,OAAO,KAAK,MAAM,OAAO,SAAS,KAAK,MAAM,QAAQ;AAAA,QAChF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,WACE,MACA,aACA,aACmB;AACnB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,OAAO,aAAa,WAAW;AAAA,EAC3E;AAAA,EAEA,OAAO,SAAiE;AACtE,WAAO,IAAI,iBAAiB,MAAM,KAAK,OAAO,mCAAS,WAAW;AAAA,EACpE;AACF;AAEO,MAAM,sBAAsB,IAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACR,UAAU,IAAI;AAAA,EACd;AAAA,EACA;AAAA,EAEA,YACEA,MACA,MACA,MACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,QAAQ;AACb,SAAK,QAAQ;AAAA,EACf;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,YAAY,UAAU;AAC5B,UAAM,UAAU,IAAI,gBAAgB,KAAK,MAAM,YAAY,YAAY;AACvE,UAAM,OAAO,kBAAkB,KAAK,KAAK;AACzC,SAAK,aAAa,KAAK;AAEvB,UAAM,UAAU,IAAI,IAAI,KAAK,MAAM,OAAO;AAC1C,UAAM,UAAU,IAAI,OAAa;AAEjC,UAAM,MAAM;AAAA,MACV;AAAA,QACE,UAAU,QAAQ;AAAA,QAClB,MAAM,SAAS,QAAQ,IAAI,MAAM,QAAQ,aAAa,WAAW,MAAM;AAAA,QACvE,MAAM;AAAA,QACN,QAAQ;AAAA,QACR,SAAS;AAAA,UACP,CAAC,oBAAoB,GAAG,KAAK,MAAM;AAAA,UACnC,CAAC,cAAc,GAAG;AAAA,QACpB;AAAA,QACA,QAAQ,KAAK;AAAA,MACf;AAAA,MACA,CAAC,QAAQ;AACP,YAAI,GAAG,QAAQ,CAAC,UAAU;AACxB,qBAAW,SAAS,QAAQ,MAAM,KAAK,GAAG;AACxC,iBAAK,MAAM,IAAI;AAAA,cACb;AAAA,cACA;AAAA,cACA,OAAO;AAAA,cACP,WAAW;AAAA,YACb,CAAC;AAAA,UACH;AAAA,QACF,CAAC;AACD,YAAI,GAAG,SAAS,MAAM;AACpB,qBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,iBAAK,MAAM,IAAI;AAAA,cACb;AAAA,cACA;AAAA,cACA,OAAO;AAAA,cACP,WAAW;AAAA,YACb,CAAC;AAAA,UACH;AACA,eAAK,MAAM,MAAM;AACjB,cAAI,CAAC,QAAQ,KAAM,SAAQ,QAAQ;AAAA,QACrC,CAAC;AACD,YAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,cAAI,IAAI,YAAY,UAAW;AAC/B,eAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6BAA6B;AACzD,cAAI,CAAC,QAAQ,KAAM,SAAQ,OAAO,GAAG;AAAA,QACvC,CAAC;AAAA,MACH;AAAA,IACF;AAEA,QAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,UAAI,IAAI,SAAS,aAAc;AAC/B,WAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,4BAA4B;AACxD,UAAI,CAAC,QAAQ,KAAM,SAAQ,OAAO,GAAG;AAAA,IACvC,CAAC;AACD,QAAI,GAAG,SAAS,MAAM;AACpB,UAAI,CAAC,QAAQ,KAAM,SAAQ,QAAQ;AAAA,IACrC,CAAC;AACD,QAAI,MAAM,KAAK,UAAU,IAAI,CAAC;AAC9B,QAAI,IAAI;AAER,QAAI;AACF,YAAM,QAAQ;AAAA,IAChB,SAAS,GAAG;AACV,UAAI,KAAK,YAAY,QAAS;AAC9B,UAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,YAAM,2BAA2B,CAAC;AAAA,IACpC;AAAA,EACF;AACF;AAEO,MAAM,yBAAyB,IAAI,iBAAiB;AAAA,EACzD;AAAA,EACA,UAAU,IAAI;AAAA,EACd,aAAa,IAAI,SAAS,MAAM,kBAAkB;AAAA,IAChD,mBAAmB;AAAA,EACrB,CAAC,EAAE,OAAO;AAAA,EACV,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,WAAW;AACtB,SAAK,QAAQ;AAAA,EACf;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,SAAK,KAAK,MAAM,SAAS,KAAK,MAAM,YAAY,KAAK,MAAM,UAAU,sBAAsB;AACzF,WAAK,QAAQ;AAAA,QACX,EAAE,OAAO,KAAK,MAAM,OAAO,OAAO,KAAK,MAAM,OAAO,SAAS,KAAK,MAAM,QAAQ;AAAA,QAChF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,YAAY,UAAU;AAC5B,QAAI,UAAU;AAEd,QAAI,uBAAuB;AAE3B,UAAM,qBAAqB,OAAOC,QAAkB;AAClD,YAAM,SAAS,kBAAkB,KAAK,OAAO,IAAI;AACjD,uBAAiB,SAAS,KAAK,YAAY;AACzC,cAAM,MAAM;AAAA,UACV,GAAG;AAAA,UACH,YAAY;AAAA,UACZ,YAAY,MAAM,QAAQ;AAAA,UAC1B,UAAU;AAAA,QACZ;AACA,QAAAA,IAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,MAC7B;AAEA,YAAM,SAAS;AAAA,QACb,GAAG;AAAA,QACH,YAAY;AAAA,QACZ,YAAY;AAAA,QACZ,UAAU;AAAA,MACZ;AACA,MAAAA,IAAG,KAAK,KAAK,UAAU,MAAM,CAAC;AAE9B,6BAAuB;AAAA,IACzB;AAEA,UAAM,YAAY,YAAY;AAC5B,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,eAAK,WAAW,MAAM;AACtB;AAAA,QACF;AACA,aAAK,WAAW,SAAS,IAAI;AAAA,MAC/B;AACA,WAAK,WAAW,SAAS;AACzB,WAAK,WAAW,MAAM;AAAA,IACxB;AAGA,UAAM,WAAW,OAAOA,QAAkB;AACxC,YAAM,UAAU,IAAI,gBAAgB,KAAK,MAAM,YAAY,YAAY;AAIvE,YAAM,eAAe,OAAO,oBAA6B;AAEzD,UAAI;AACJ,UAAI,0BAAyC,CAAC;AAE9C,YAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,YAAI,aAAa,CAAC,KAAK,MAAM,QAAQ;AAEnC,eAAK,MAAM,IAAI;AAAA,YACb;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP;AAAA,YACA,kBACE,wBAAwB,SAAS,IAAI,0BAA0B;AAAA,UACnE,CAAC;AACD,sBAAY;AACZ,oCAA0B,CAAC;AAAA,QAC7B;AAAA,MACF;AAEA,UAAI,UAAiC;AAErC,YAAM,uBAAuB,MAAM;AACjC,YAAI,SAAS;AACX,uBAAa,OAAO;AACpB,oBAAU;AAAA,QACZ;AAAA,MACF;AAGA,YAAM,YAAY,CAAC,SAAkB;AACnC,aAAK,aAAa,MAAM,IAAI,EAAE,MAAM,CAAC,UAAmB;AACtD,eAAK,QAAQ,MAAM,EAAE,MAAM,GAAG,0DAA0D;AAAA,QAC1F,CAAC;AAAA,MACH;AAEA,YAAM,UAAU,CAAC,MAAc,WAAmB;AAChD,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,OAAO,SAAS,CAAC,EAAE;AAAA,QAC/E;AACA,6BAAqB;AACrB,aAAK,aAAa,MAAM;AAAA,MAC1B;AAEA,YAAM,UAAU,CAAC,QAAe;AAC9B,aAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,0BAA0B;AACtD,aAAK,aAAa,MAAM;AAAA,MAC1B;AAGA,MAAAA,IAAG,GAAG,WAAW,SAAS;AAC1B,MAAAA,IAAG,GAAG,SAAS,OAAO;AACtB,MAAAA,IAAG,GAAG,SAAS,OAAO;AAEtB,UAAI;AAEF,cAAM,SAAS,aAAa,OAAO,EAAE,UAAU;AAE/C,eAAO,CAAC,KAAK,UAAU,CAAC,KAAK,gBAAgB,OAAO,SAAS;AAC3D,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,OAAO,KAAM;AAEjB,gBAAM,SAAS,OAAO;AAGtB,cAAI;AACJ,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,OAAO,SAAS,CAAC;AACzC,wBAAY,sBAAsB,MAAM,IAAI;AAAA,UAC9C,SAAS,UAAU;AACjB,iBAAK,QAAQ,KAAK,EAAE,SAAS,GAAG,kCAAkC;AAClE;AAAA,UACF;AAGA,cAAI,eAAe,SAAS,GAAG;AAC7B,iBAAK,QAAQ,MAAM,EAAE,OAAO,UAAU,MAAM,GAAG,yBAAyB;AACxE;AAAA,UACF;AAEA,gBAAM,YAAY,UAAU;AAG5B,cAAI,KAAK,MAAM,mBAAmB,SAAS,kBAAkB,SAAS,GAAG;AACvE,kBAAM,iBAAiB,UAAU;AACjC,qBAAS,IAAI,GAAG,IAAI,eAAe,MAAM,QAAQ,KAAK;AACpD,oBAAM,OAAO,eAAe,MAAM,CAAC;AACnC,oBAAM,YAAY,eAAe,MAAM,CAAC;AACxC,oBAAM,UAAU,eAAe,IAAI,CAAC;AACpC,kBAAI,SAAS,UAAa,cAAc,UAAa,YAAY,QAAW;AAC1E,wCAAwB;AAAA,kBACtB,kBAAkB;AAAA,oBAChB,MAAM,OAAO;AAAA;AAAA,oBACb;AAAA,oBACA;AAAA,kBACF,CAAC;AAAA,gBACH;AAAA,cACF;AAAA,YACF;AAAA,UACF;AAGA,cAAI,eAAe,SAAS,GAAG;AAC7B,kBAAM,cAAc,OAAO,KAAK,UAAU,MAAM,QAAQ;AAExD,kBAAM,YAAY,YAAY,OAAO;AAAA,cACnC,YAAY;AAAA,cACZ,YAAY,aAAa,YAAY;AAAA,YACvC;AACA,uBAAW,SAAS,QAAQ,MAAM,SAAS,GAAG;AAC5C,4BAAc,WAAW,KAAK;AAC9B,0BAAY;AAAA,YACd;AAKA,iCAAqB;AACrB,sBAAU,WAAW,MAAM;AAEzB,mBAAK,QAAQ;AAAA,gBACX,qDAAqD,KAAK,MAAM,YAAY;AAAA,cAC9E;AACA,cAAAA,IAAG,MAAM;AAAA,YACX,GAAG,KAAK,MAAM,YAAY;AAAA,UAC5B,WAAW,cAAc,SAAS,GAAG;AAEnC,gBAAI,sBAAsB;AACxB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,WAAW,KAAK;AAC9B,4BAAY;AAAA,cACd;AACA,4BAAc,WAAW,IAAI;AAC7B,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,cAC/C;AAEA,kBAAI,cAAc,WAAW;AAC3B,0BAAU;AACV,qCAAqB;AACrB,gBAAAA,IAAG,MAAM;AACT;AAAA,cACF;AAAA,YACF;AAAA,UAEF;AAAA,QACF;AAAA,MACF,SAAS,KAAK;AAEZ,YAAI,eAAe,SAAS,CAAC,IAAI,QAAQ,SAAS,kBAAkB,GAAG;AACrE,cACE,IAAI,QAAQ,SAAS,iBAAiB,KACtC,IAAI,QAAQ,SAAS,mBAAmB,GACxC;AACA,iBAAK,QAAQ;AAAA,cACX,EAAE,IAAI;AAAA,cACN;AAAA,YACF;AAAA,UACF,OAAO;AACL,iBAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,2CAA2C;AAAA,UACzE;AAAA,QACF;AAAA,MACF,UAAE;AAEA,QAAAA,IAAG,IAAI,WAAW,SAAS;AAC3B,QAAAA,IAAG,IAAI,SAAS,OAAO;AACvB,QAAAA,IAAG,IAAI,SAAS,OAAO;AACvB,6BAAqB;AAAA,MACvB;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,MAAM,QAAQ,QAAQ,SAAS,IAAI;AACtD,UAAM,MAAM,GAAG,KAAK,0BAA0B,KAAK,MAAM,MAAM,qBAAqB,OAAO;AAE3F,QAAI;AACJ,QAAI;AACF,WAAK,MAAM,yBAAyB;AAAA,QAClC;AAAA,QACA,WAAW,KAAK,YAAY;AAAA,QAC5B,aAAa,KAAK;AAAA,MACpB,CAAC;AACD,YAAM,QAAQ,IAAI,CAAC,UAAU,GAAG,mBAAmB,EAAE,GAAG,SAAS,EAAE,CAAC,CAAC;AAAA,IACvE,SAAS,GAAG;AACV,UAAI,KAAK,YAAY,SAAS;AAC5B;AAAA,MACF;AACA,YAAM,2BAA2B,CAAC;AAAA,IACpC,UAAE;AAEA,UAAI,MAAM,GAAG,eAAe,UAAU,QAAQ;AAC5C,+BAAuB,EAAE;AAAA,MAC3B;AAAA,IACF;AAAA,EACF;AACF;AAEA,MAAM,UAAU,CAAC,MAAuB,aAAa,QAAQ,IAAI,IAAI,MAAM,OAAO,CAAC,CAAC;AAEpF,MAAM,wBAAwB,oBAAI,IAAI;AAAA,EACpC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF,CAAC;AAED,MAAM,WAAW,CAAC,MAA6C;AAC7D,SAAO,MAAM,QAAQ,OAAO,MAAM;AACpC;AAEA,MAAM,uBAAuB,CAAC,MAA0D;AACtF,MAAI,CAAC,SAAS,CAAC,EAAG,QAAO;AACzB,SAAO,EAAE,SAAS,oBAAoB,MAAM,QAAQ,EAAE,MAAM;AAC9D;AAEA,MAAM,eAAe,CAAC,GAAY,SAA0B;AAC1D,MAAI,SAAS,CAAC,KAAK,EAAE,SAAS,KAAM,QAAO;AAC3C,MAAI,qBAAqB,CAAC,GAAG;AAC3B,WAAO,EAAE,OAAO,KAAK,CAAC,UAAU,aAAa,OAAO,IAAI,CAAC;AAAA,EAC3D;AACA,SAAO;AACT;AAEA,MAAM,sBAAsB,CAAC,MAAwB;AACnD,MAAI,SAAS,CAAC,KAAK,OAAO,EAAE,SAAS,UAAU;AAC7C,WAAO,sBAAsB,IAAI,EAAE,IAAI;AAAA,EACzC;AACA,MAAI,qBAAqB,CAAC,GAAG;AAC3B,WAAO,EAAE,OAAO,KAAK,CAAC,UAAU,oBAAoB,KAAK,CAAC;AAAA,EAC5D;AACA,SAAO;AACT;AAEA,MAAM,6BAA6B,CAAC,MAAmC;AACrE,QAAM,MAAM,QAAQ,CAAC;AACrB,QAAM,YACJ,aAAa,GAAG,WAAW,KAC1B,OAAO,IAAI,YAAY,YAAY,IAAI,QAAQ,SAAS,WAAW;AACtE,QAAM,UAAU,YACZ,kCACA,+BAA+B,IAAI,WAAW,eAAe;AACjE,SAAO,YAAY,IAAI,gBAAgB,EAAE,QAAQ,CAAC,IAAI,IAAI,mBAAmB,EAAE,QAAQ,CAAC;AAC1F;AAEA,MAAM,gBAAgB,OAAO;AAAA,EAC3B;AAAA,EACA;AAAA,EACA;AACF,MAIM;AACJ,MAAI,YAAY,SAAS;AACvB,UAAM,IAAI,MAAM,SAAS;AAAA,EAC3B;AAEA,QAAM,MAAM,IAAI,OAAa;AAC7B,MAAI;AAEJ,QAAM,UAAU,MAAM;AACpB,QAAI,QAAS,cAAa,OAAO;AACjC,OAAG,IAAI,QAAQ,MAAM;AACrB,OAAG,IAAI,SAAS,OAAO;AACvB,OAAG,IAAI,SAAS,OAAO;AACvB,gBAAY,oBAAoB,SAAS,OAAO;AAAA,EAClD;AAEA,QAAM,SAAS,MAAM,IAAI,QAAQ;AACjC,QAAM,UAAU,CAAC,QAAe,IAAI,OAAO,QAAQ,GAAG,CAAC;AACvD,QAAM,UAAU,CAAC,MAAc,WAC7B,IAAI;AAAA,IACF,IAAI,MAAM,sCAAsC,IAAI,YAAY,OAAO,SAAS,CAAC,GAAG;AAAA,EACtF;AACF,QAAM,UAAU,MAAM,IAAI,OAAO,IAAI,MAAM,SAAS,CAAC;AAErD,KAAG,GAAG,QAAQ,MAAM;AACpB,KAAG,GAAG,SAAS,OAAO;AACtB,KAAG,GAAG,SAAS,OAAO;AACtB,cAAY,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAE7D,MAAI,YAAY,GAAG;AACjB,cAAU,WAAW,MAAM,IAAI,OAAO,IAAI,MAAM,iBAAiB,CAAC,GAAG,SAAS;AAAA,EAChF;AAEA,MAAI;AACF,UAAM,IAAI;AAAA,EACZ,UAAE;AACA,YAAQ;AAAA,EACV;AACF;AAEA,MAAM,yBAAyB,CAAC,OAAkB;AAGhD,MAAI;AACF,OAAG,GAAG,SAAS,MAAM;AAAA,IAAC,CAAC;AAAA,EACzB,QAAQ;AAAA,EAER;AAEA,MAAI;AAEF,QAAI,GAAG,eAAe,UAAU,YAAY;AAC1C,SAAG,MAAM;AAAA,IACX,OAAO;AACL,SAAG,UAAU;AAAA,IACf;AAAA,EACF,QAAQ;AAAA,EAER;AACF;AAEA,MAAM,2BAA2B,OAAO;AAAA,EACtC;AAAA,EACA;AAAA,EACA;AACF,MAI0B;AACxB,QAAM,cAAc,OAAO,WAAwC;AACjE,UAAM,KAAK,IAAI,UAAU,KAAK,EAAE,kBAAkB,WAAW,OAAO,CAAC;AACrE,QAAI;AACF,YAAM,cAAc,EAAE,IAAI,WAAW,YAAY,CAAC;AAClD,aAAO;AAAA,IACT,SAAS,GAAG;AACV,6BAAuB,EAAE;AACzB,YAAM;AAAA,IACR;AAAA,EACF;AAEA,MAAI;AACF,WAAO,MAAM,YAAY;AAAA,EAC3B,SAAS,GAAG;AAUV,QAAI,oBAAoB,CAAC,KAAK,qBAAqB,CAAC,GAAG;AACrD,aAAO,MAAM,YAAY,CAAC;AAAA,IAC5B;AACA,UAAM;AAAA,EACR;AACF;AAQA,MAAM,oBAAoB,CACxB,MACA,YAAqB,UACS;AAC9B,QAAM,QAAmC,CAAC;AAC1C,MAAI,OAAO,KAAK,UAAU,UAAU;AAClC,UAAM,OAAO;AACb,UAAM,KAAK,KAAK;AAAA,EAClB,OAAO;AACL,UAAM,OAAO;AACb,UAAM,YAAY,KAAK;AAAA,EACzB;AAEA,QAAM,gBAA2C,CAAC;AAClD,MAAI,KAAK,OAAO;AACd,kBAAc,QAAQ,KAAK;AAAA,EAC7B;AACA,MAAI,KAAK,SAAS;AAChB,kBAAc,UAAU,KAAK;AAAA,EAC/B;AAEA,MAAI,OAAO,KAAK,aAAa,EAAE,QAAQ;AACrC,UAAM,0BAA0B;AAAA,EAClC;AAEA,QAAM,SAAoC;AAAA,IACxC,UAAU,KAAK;AAAA,IACf;AAAA,IACA,eAAe;AAAA,MACb,WAAW;AAAA,MACX,UAAU,KAAK;AAAA,MACf,aAAa,KAAK;AAAA,IACpB;AAAA,IACA,UAAU,KAAK;AAAA,IACf,uBAAuB,KAAK;AAAA,EAC9B;AAEA,MAAI,aAAa,KAAK,mBAAmB,OAAO;AAC9C,WAAO,iBAAiB;AAAA,EAC1B;AAEA,SAAO;AACT;","names":["tts","ws"]}
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@livekit/agents-plugin-cartesia",
3
- "version": "1.0.43",
3
+ "version": "1.0.45",
4
4
  "description": "Cartesia plugin for LiveKit Node Agents",
5
5
  "main": "dist/index.js",
6
6
  "require": "dist/index.cjs",
@@ -30,9 +30,9 @@
30
30
  "@types/ws": "^8.5.10",
31
31
  "tsup": "^8.3.5",
32
32
  "typescript": "^5.0.0",
33
- "@livekit/agents": "1.0.43",
34
- "@livekit/agents-plugin-openai": "1.0.43",
35
- "@livekit/agents-plugins-test": "1.0.43"
33
+ "@livekit/agents": "1.0.45",
34
+ "@livekit/agents-plugin-openai": "1.0.45",
35
+ "@livekit/agents-plugins-test": "1.0.45"
36
36
  },
37
37
  "dependencies": {
38
38
  "ws": "^8.16.0"
@@ -40,7 +40,7 @@
40
40
  "peerDependencies": {
41
41
  "@livekit/rtc-node": "^0.13.24",
42
42
  "zod": "^3.25.76 || ^4.1.8",
43
- "@livekit/agents": "1.0.43"
43
+ "@livekit/agents": "1.0.45"
44
44
  },
45
45
  "scripts": {
46
46
  "build": "tsup --onSuccess \"pnpm build:types\"",
package/src/tts.ts CHANGED
@@ -3,6 +3,8 @@
3
3
  // SPDX-License-Identifier: Apache-2.0
4
4
  import {
5
5
  type APIConnectOptions,
6
+ APIConnectionError,
7
+ APITimeoutError,
6
8
  AudioByteStream,
7
9
  Future,
8
10
  type TimedString,
@@ -60,6 +62,8 @@ export interface TTSOptions {
60
62
  * @defaultValue true
61
63
  */
62
64
  wordTimestamps?: boolean;
65
+
66
+ pronunciationDictId?: string;
63
67
  }
64
68
 
65
69
  const defaultTTSOptions: TTSOptions = {
@@ -126,8 +130,8 @@ export class TTS extends tts.TTS {
126
130
  return new ChunkedStream(this, text, this.#opts, connOptions, abortSignal);
127
131
  }
128
132
 
129
- stream(): SynthesizeStream {
130
- return new SynthesizeStream(this, this.#opts);
133
+ stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream {
134
+ return new SynthesizeStream(this, this.#opts, options?.connOptions);
131
135
  }
132
136
  }
133
137
 
@@ -191,11 +195,12 @@ export class ChunkedStream extends tts.ChunkedStream {
191
195
  });
192
196
  }
193
197
  this.queue.close();
194
- doneFut.resolve();
198
+ if (!doneFut.done) doneFut.resolve();
195
199
  });
196
200
  res.on('error', (err) => {
197
201
  if (err.message === 'aborted') return;
198
202
  this.#logger.error({ err }, 'Cartesia TTS response error');
203
+ if (!doneFut.done) doneFut.reject(err);
199
204
  });
200
205
  },
201
206
  );
@@ -203,12 +208,21 @@ export class ChunkedStream extends tts.ChunkedStream {
203
208
  req.on('error', (err) => {
204
209
  if (err.name === 'AbortError') return;
205
210
  this.#logger.error({ err }, 'Cartesia TTS request error');
211
+ if (!doneFut.done) doneFut.reject(err);
212
+ });
213
+ req.on('close', () => {
214
+ if (!doneFut.done) doneFut.resolve();
206
215
  });
207
- req.on('close', () => doneFut.resolve());
208
216
  req.write(JSON.stringify(json));
209
217
  req.end();
210
218
 
211
- await doneFut.await;
219
+ try {
220
+ await doneFut.await;
221
+ } catch (e) {
222
+ if (this.abortSignal.aborted) return;
223
+ if (!this.queue.closed) this.queue.close();
224
+ throw toRetryableConnectionError(e);
225
+ }
212
226
  }
213
227
  }
214
228
 
@@ -220,8 +234,8 @@ export class SynthesizeStream extends tts.SynthesizeStream {
220
234
  }).stream();
221
235
  label = 'cartesia.SynthesizeStream';
222
236
 
223
- constructor(tts: TTS, opts: TTSOptions) {
224
- super(tts);
237
+ constructor(tts: TTS, opts: TTSOptions, connOptions?: APIConnectOptions) {
238
+ super(tts, connOptions);
225
239
  this.#opts = opts;
226
240
  }
227
241
 
@@ -457,22 +471,186 @@ export class SynthesizeStream extends tts.SynthesizeStream {
457
471
 
458
472
  const wsUrl = this.#opts.baseUrl.replace(/^http/, 'ws');
459
473
  const url = `${wsUrl}/tts/websocket?api_key=${this.#opts.apiKey}&cartesia_version=${VERSION}`;
460
- const ws = new WebSocket(url);
461
474
 
475
+ let ws: WebSocket | undefined;
462
476
  try {
463
- await new Promise((resolve, reject) => {
464
- ws.on('open', resolve);
465
- ws.on('error', (error) => reject(error));
466
- ws.on('close', (code) => reject(`WebSocket returned ${code}`));
477
+ ws = await connectCartesiaWebSocket({
478
+ url,
479
+ timeoutMs: this.connOptions.timeoutMs,
480
+ abortSignal: this.abortSignal,
467
481
  });
468
-
469
482
  await Promise.all([inputTask(), sentenceStreamTask(ws), recvTask(ws)]);
470
483
  } catch (e) {
471
- throw new Error(`failed to connect to Cartesia: ${e}`);
484
+ if (this.abortSignal.aborted) {
485
+ return;
486
+ }
487
+ throw toRetryableConnectionError(e);
488
+ } finally {
489
+ // Ensure we don't leak sockets/tasks across retry attempts.
490
+ if (ws && ws.readyState !== WebSocket.CLOSED) {
491
+ safeTerminateWebSocket(ws);
492
+ }
472
493
  }
473
494
  }
474
495
  }
475
496
 
497
+ const asError = (e: unknown): Error => (e instanceof Error ? e : new Error(String(e)));
498
+
499
+ const transientNetworkCodes = new Set([
500
+ 'ETIMEDOUT',
501
+ 'ECONNRESET',
502
+ 'EAI_AGAIN',
503
+ 'ENETUNREACH',
504
+ 'ECONNREFUSED',
505
+ 'EHOSTUNREACH',
506
+ ]);
507
+
508
+ const isRecord = (v: unknown): v is Record<string, unknown> => {
509
+ return v !== null && typeof v === 'object';
510
+ };
511
+
512
+ const isAggregateErrorLike = (e: unknown): e is { errors: unknown[]; name?: string } => {
513
+ if (!isRecord(e)) return false;
514
+ return e.name === 'AggregateError' && Array.isArray(e.errors);
515
+ };
516
+
517
+ const hasErrorCode = (e: unknown, code: string): boolean => {
518
+ if (isRecord(e) && e.code === code) return true;
519
+ if (isAggregateErrorLike(e)) {
520
+ return e.errors.some((inner) => hasErrorCode(inner, code));
521
+ }
522
+ return false;
523
+ };
524
+
525
+ const hasAnyTransientCode = (e: unknown): boolean => {
526
+ if (isRecord(e) && typeof e.code === 'string') {
527
+ return transientNetworkCodes.has(e.code);
528
+ }
529
+ if (isAggregateErrorLike(e)) {
530
+ return e.errors.some((inner) => hasAnyTransientCode(inner));
531
+ }
532
+ return false;
533
+ };
534
+
535
+ const toRetryableConnectionError = (e: unknown): APIConnectionError => {
536
+ const err = asError(e);
537
+ const isTimeout =
538
+ hasErrorCode(e, 'ETIMEDOUT') ||
539
+ (typeof err.message === 'string' && err.message.includes('ETIMEDOUT'));
540
+ const message = isTimeout
541
+ ? `Cartesia connection timed out`
542
+ : `Cartesia connection failed: ${err.message || 'unknown error'}`;
543
+ return isTimeout ? new APITimeoutError({ message }) : new APIConnectionError({ message });
544
+ };
545
+
546
+ const waitForWsOpen = async ({
547
+ ws,
548
+ timeoutMs,
549
+ abortSignal,
550
+ }: {
551
+ ws: WebSocket;
552
+ timeoutMs: number;
553
+ abortSignal: AbortSignal;
554
+ }) => {
555
+ if (abortSignal.aborted) {
556
+ throw new Error('aborted');
557
+ }
558
+
559
+ const fut = new Future<void>();
560
+ let timeout: NodeJS.Timeout | undefined;
561
+
562
+ const cleanup = () => {
563
+ if (timeout) clearTimeout(timeout);
564
+ ws.off('open', onOpen);
565
+ ws.off('error', onError);
566
+ ws.off('close', onClose);
567
+ abortSignal.removeEventListener('abort', onAbort);
568
+ };
569
+
570
+ const onOpen = () => fut.resolve();
571
+ const onError = (err: Error) => fut.reject(asError(err));
572
+ const onClose = (code: number, reason: Buffer) =>
573
+ fut.reject(
574
+ new Error(`WebSocket closed before open (code=${code}, reason=${reason.toString()})`),
575
+ );
576
+ const onAbort = () => fut.reject(new Error('aborted'));
577
+
578
+ ws.on('open', onOpen);
579
+ ws.on('error', onError);
580
+ ws.on('close', onClose);
581
+ abortSignal.addEventListener('abort', onAbort, { once: true });
582
+
583
+ if (timeoutMs > 0) {
584
+ timeout = setTimeout(() => fut.reject(new Error('connect timeout')), timeoutMs);
585
+ }
586
+
587
+ try {
588
+ await fut.await;
589
+ } finally {
590
+ cleanup();
591
+ }
592
+ };
593
+
594
+ const safeTerminateWebSocket = (ws: WebSocket) => {
595
+ // `ws` can emit an 'error' event during teardown (especially if CONNECTING).
596
+ // If there is no error listener at that moment, Node will treat it as unhandled and crash the process.
597
+ try {
598
+ ws.on('error', () => {});
599
+ } catch {
600
+ // ignore
601
+ }
602
+
603
+ try {
604
+ // `terminate()` can throw if the socket was never established; `close()` is safer in CONNECTING.
605
+ if (ws.readyState === WebSocket.CONNECTING) {
606
+ ws.close();
607
+ } else {
608
+ ws.terminate();
609
+ }
610
+ } catch {
611
+ // ignore
612
+ }
613
+ };
614
+
615
+ const connectCartesiaWebSocket = async ({
616
+ url,
617
+ timeoutMs,
618
+ abortSignal,
619
+ }: {
620
+ url: string;
621
+ timeoutMs: number;
622
+ abortSignal: AbortSignal;
623
+ }): Promise<WebSocket> => {
624
+ const connectOnce = async (family?: number): Promise<WebSocket> => {
625
+ const ws = new WebSocket(url, { handshakeTimeout: timeoutMs, family });
626
+ try {
627
+ await waitForWsOpen({ ws, timeoutMs, abortSignal });
628
+ return ws;
629
+ } catch (e) {
630
+ safeTerminateWebSocket(ws);
631
+ throw e;
632
+ }
633
+ };
634
+
635
+ try {
636
+ return await connectOnce();
637
+ } catch (e) {
638
+ // Mitigation for Node.js dual-stack (IPv6/IPv4) connect flakiness ("happy eyeballs"):
639
+ // some environments surface `AggregateError` with nested `ETIMEDOUT` during the initial
640
+ // WebSocket open. In that case we do a one-off retry forcing IPv4 (`family: 4`) before
641
+ // letting the outer framework retry loop handle further attempts.
642
+ //
643
+ // If you still see `AggregateError`/`ETIMEDOUT`:
644
+ // - Increase the session TTS connect timeout (`connOptions.ttsConnOptions.timeoutMs`)
645
+ // - Or adjust Node's family autoselection behavior via `NODE_OPTIONS`, e.g.
646
+ // `--network-family-autoselection-attempt-timeout=5000` (or disable it entirely).
647
+ if (hasAnyTransientCode(e) || isAggregateErrorLike(e)) {
648
+ return await connectOnce(4);
649
+ }
650
+ throw e;
651
+ }
652
+ };
653
+
476
654
  /**
477
655
  * Convert TTSOptions to Cartesia API format.
478
656
  *
@@ -513,6 +691,7 @@ const toCartesiaOptions = (
513
691
  sample_rate: opts.sampleRate,
514
692
  },
515
693
  language: opts.language,
694
+ pronunciation_dict_id: opts.pronunciationDictId,
516
695
  };
517
696
 
518
697
  if (streaming && opts.wordTimestamps !== false) {