@livekit/agents 1.0.48 → 1.0.49

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.
Files changed (48) hide show
  1. package/dist/cpu.cjs +189 -0
  2. package/dist/cpu.cjs.map +1 -0
  3. package/dist/cpu.d.cts +24 -0
  4. package/dist/cpu.d.ts +24 -0
  5. package/dist/cpu.d.ts.map +1 -0
  6. package/dist/cpu.js +152 -0
  7. package/dist/cpu.js.map +1 -0
  8. package/dist/cpu.test.cjs +227 -0
  9. package/dist/cpu.test.cjs.map +1 -0
  10. package/dist/cpu.test.js +204 -0
  11. package/dist/cpu.test.js.map +1 -0
  12. package/dist/inference/llm.cjs.map +1 -1
  13. package/dist/inference/llm.d.cts +1 -1
  14. package/dist/inference/llm.d.ts +1 -1
  15. package/dist/inference/llm.d.ts.map +1 -1
  16. package/dist/inference/llm.js.map +1 -1
  17. package/dist/inference/tts.cjs.map +1 -1
  18. package/dist/inference/tts.d.cts +6 -0
  19. package/dist/inference/tts.d.ts +6 -0
  20. package/dist/inference/tts.d.ts.map +1 -1
  21. package/dist/inference/tts.js.map +1 -1
  22. package/dist/version.cjs +1 -1
  23. package/dist/version.js +1 -1
  24. package/dist/voice/agent_activity.cjs +36 -8
  25. package/dist/voice/agent_activity.cjs.map +1 -1
  26. package/dist/voice/agent_activity.d.ts.map +1 -1
  27. package/dist/voice/agent_activity.js +37 -9
  28. package/dist/voice/agent_activity.js.map +1 -1
  29. package/dist/voice/agent_session.cjs +27 -1
  30. package/dist/voice/agent_session.cjs.map +1 -1
  31. package/dist/voice/agent_session.d.cts +6 -0
  32. package/dist/voice/agent_session.d.ts +6 -0
  33. package/dist/voice/agent_session.d.ts.map +1 -1
  34. package/dist/voice/agent_session.js +27 -1
  35. package/dist/voice/agent_session.js.map +1 -1
  36. package/dist/worker.cjs +6 -29
  37. package/dist/worker.cjs.map +1 -1
  38. package/dist/worker.d.ts.map +1 -1
  39. package/dist/worker.js +6 -19
  40. package/dist/worker.js.map +1 -1
  41. package/package.json +1 -1
  42. package/src/cpu.test.ts +239 -0
  43. package/src/cpu.ts +173 -0
  44. package/src/inference/llm.ts +2 -0
  45. package/src/inference/tts.ts +8 -1
  46. package/src/voice/agent_activity.ts +58 -10
  47. package/src/voice/agent_session.ts +33 -2
  48. package/src/worker.ts +34 -50
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/inference/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { ConnectionPool } from '../connection_pool.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport { basic as tokenizeBasic } from '../tokenize/index.js';\nimport type { ChunkedStream } from '../tts/index.js';\nimport { SynthesizeStream as BaseSynthesizeStream, TTS as BaseTTS } from '../tts/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { Event, Future, Task, cancelAndWait, combineSignals, shortuuid } from '../utils.js';\nimport {\n type TtsClientEvent,\n type TtsServerEvent,\n ttsClientEventSchema,\n ttsServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken } from './utils.js';\n\nexport type CartesiaModels =\n | 'cartesia/sonic-3'\n | 'cartesia/sonic-2'\n | 'cartesia/sonic-turbo'\n | 'cartesia/sonic';\n\nexport type DeepgramTTSModels = 'deepgram/aura' | 'deepgram/aura-2';\n\nexport type ElevenlabsModels =\n | 'elevenlabs/eleven_flash_v2'\n | 'elevenlabs/eleven_flash_v2_5'\n | 'elevenlabs/eleven_turbo_v2'\n | 'elevenlabs/eleven_turbo_v2_5'\n | 'elevenlabs/eleven_multilingual_v2';\n\nexport type InworldModels =\n | 'inworld/inworld-tts-1.5-max'\n | 'inworld/inworld-tts-1.5-mini'\n | 'inworld/inworld-tts-1-max'\n | 'inworld/inworld-tts-1';\n\nexport type RimeModels = 'rime/arcana' | 'rime/mistv2';\n\nexport interface CartesiaOptions {\n /** Maximum duration of audio in seconds. */\n duration?: number;\n /** Speech speed. Default: not specified. */\n speed?: 'slow' | 'normal' | 'fast';\n}\n\nexport interface ElevenlabsOptions {\n /** Inactivity timeout in seconds. Default: 60. */\n inactivity_timeout?: number;\n /** Text normalization mode. Default: \"auto\". */\n apply_text_normalization?: 'auto' | 'off' | 'on';\n}\n\nexport interface DeepgramTTSOptions {}\n\nexport interface RimeOptions {}\n\nexport interface InworldOptions {}\n\ntype _TTSModels =\n | CartesiaModels\n | DeepgramTTSModels\n | ElevenlabsModels\n | RimeModels\n | InworldModels;\n\nexport type TTSModels =\n | CartesiaModels\n | DeepgramTTSModels\n | ElevenlabsModels\n | RimeModels\n | InworldModels\n | AnyString;\n\nexport type ModelWithVoice = `${_TTSModels}:${string}` | TTSModels;\n\nexport type TTSOptions<TModel extends TTSModels> = TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends DeepgramTTSModels\n ? DeepgramTTSOptions\n : TModel extends ElevenlabsModels\n ? ElevenlabsOptions\n : TModel extends RimeModels\n ? RimeOptions\n : TModel extends InworldModels\n ? InworldOptions\n : Record<string, unknown>;\n\n/** Parse a model string into [model, voice]. Voice is undefined if not specified. */\nexport function parseTTSModelString(model: string): [string, string | undefined] {\n const idx = model.lastIndexOf(':');\n if (idx !== -1) {\n return [model.slice(0, idx), model.slice(idx + 1)];\n }\n return [model, undefined];\n}\n\n/** A fallback model with optional extra configuration. Extra fields are passed through to the provider. */\nexport interface TTSFallbackModel {\n /** Model name (e.g. \"cartesia/sonic\", \"elevenlabs/eleven_flash_v2\", \"rime/arcana\"). */\n model: string;\n /** Voice to use for the model. */\n voice: string;\n /** Extra configuration for the model. */\n extraKwargs?: Record<string, unknown>;\n}\n\nexport type TTSFallbackModelType = TTSFallbackModel | string;\n\n/** Normalize a single or list of FallbackModelType into TTSFallbackModel[]. */\nexport function normalizeTTSFallback(\n fallback: TTSFallbackModelType | TTSFallbackModelType[],\n): TTSFallbackModel[] {\n const makeFallback = (model: TTSFallbackModelType): TTSFallbackModel => {\n if (typeof model === 'string') {\n const [name, voice] = parseTTSModelString(model);\n return { model: name, voice: voice ?? '' };\n }\n return model;\n };\n\n if (Array.isArray(fallback)) {\n return fallback.map(makeFallback);\n }\n return [makeFallback(fallback)];\n}\n\ntype TTSEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: TTSEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'https://agent-gateway.livekit.cloud/v1';\nconst NUM_CHANNELS = 1;\nconst DEFAULT_LANGUAGE = 'en';\n\nexport interface InferenceTTSOptions<TModel extends TTSModels> {\n model?: TModel;\n voice?: string;\n language?: string;\n encoding: TTSEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: TTSOptions<TModel>;\n fallback?: TTSFallbackModel[];\n connOptions?: APIConnectOptions;\n}\n\n/**\n * Livekit Cloud Inference TTS\n */\nexport class TTS<TModel extends TTSModels> extends BaseTTS {\n private opts: InferenceTTSOptions<TModel>;\n private streams: Set<SynthesizeStream<TModel>> = new Set();\n pool: ConnectionPool<WebSocket>;\n\n #logger = log();\n\n constructor(opts: {\n model: TModel;\n voice?: string;\n language?: string;\n baseURL?: string;\n encoding?: TTSEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: TTSOptions<TModel>;\n fallback?: TTSFallbackModelType | TTSFallbackModelType[];\n connOptions?: APIConnectOptions;\n }) {\n const sampleRate = opts?.sampleRate ?? DEFAULT_SAMPLE_RATE;\n super(sampleRate, 1, { streaming: true });\n\n const {\n model,\n voice,\n language = DEFAULT_LANGUAGE,\n baseURL,\n encoding = DEFAULT_ENCODING,\n apiKey,\n apiSecret,\n modelOptions = {} as TTSOptions<TModel>,\n fallback,\n connOptions,\n } = opts || {};\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n // read voice id from the model if provided: \"provider/model:voice_id\"\n let nextModel = model;\n let nextVoice = voice;\n if (typeof nextModel === 'string') {\n const idx = nextModel.lastIndexOf(':');\n if (idx !== -1) {\n const voiceFromModel = nextModel.slice(idx + 1);\n if (nextVoice && nextVoice !== voiceFromModel) {\n this.#logger.warn(\n '`voice` is provided via both argument and model, using the one from the argument',\n { voice: nextVoice, model: nextModel },\n );\n } else {\n nextVoice = voiceFromModel;\n }\n nextModel = nextModel.slice(0, idx) as TModel;\n }\n }\n\n const normalizedFallback = fallback ? normalizeTTSFallback(fallback) : undefined;\n\n this.opts = {\n model: nextModel,\n voice: nextVoice,\n language,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n fallback: normalizedFallback,\n connOptions: connOptions ?? DEFAULT_API_CONNECT_OPTIONS,\n };\n\n // Initialize connection pool\n this.pool = new ConnectionPool<WebSocket>({\n connectCb: (timeout) => this.connectWs(timeout),\n closeCb: (ws) => this.closeWs(ws),\n maxSessionDuration: 300_000,\n markRefreshedOnGet: true,\n connectTimeout: 10_000, // 10 seconds default\n });\n }\n\n get label() {\n return 'inference.TTS';\n }\n\n static fromModelString(modelString: string): TTS<AnyString> {\n const [model, voice] = parseTTSModelString(modelString);\n return new TTS({ model, voice: voice || undefined });\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n synthesize(_: string): ChunkedStream {\n throw new Error('ChunkedStream is not implemented');\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream<TModel> {\n const { connOptions = this.opts.connOptions ?? DEFAULT_API_CONNECT_OPTIONS } = options || {};\n const stream = new SynthesizeStream(this, { ...this.opts }, connOptions);\n this.streams.add(stream);\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/tts`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const params = {\n type: 'session.create',\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n } as Record<string, unknown>;\n\n if (this.opts.voice) (params as Record<string, unknown>).voice = this.opts.voice;\n if (this.opts.model) (params as Record<string, unknown>).model = this.opts.model;\n if (this.opts.language) (params as Record<string, unknown>).language = this.opts.language;\n\n if (this.opts.fallback?.length) {\n params.fallback = {\n models: this.opts.fallback.map((m) => ({\n model: m.model,\n voice: m.voice,\n extra: m.extraKwargs ?? {},\n })),\n };\n }\n\n if (this.opts.connOptions) {\n params.connection = {\n timeout: this.opts.connOptions.timeoutMs / 1000,\n retries: this.opts.connOptions.maxRetry,\n };\n }\n\n this.#logger.debug({ url }, 'inference.TTS creating new websocket connection (pool miss)');\n const socket = await connectWs(url, headers, timeout);\n socket.send(JSON.stringify(params));\n return socket;\n }\n\n async closeWs(ws: WebSocket) {\n await ws.close();\n }\n\n prewarm(): void {\n this.pool.prewarm();\n }\n\n async close() {\n for (const stream of this.streams) {\n await stream.close();\n }\n this.streams.clear();\n await this.pool.close();\n }\n}\n\nexport class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeStream {\n private opts: InferenceTTSOptions<TModel>;\n private tts: TTS<TModel>;\n\n #logger = log();\n\n constructor(tts: TTS<TModel>, opts: InferenceTTSOptions<TModel>, connOptions: APIConnectOptions) {\n super(tts, connOptions);\n this.opts = opts;\n this.tts = tts;\n }\n\n get label() {\n return 'inference.SynthesizeStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n }\n\n protected async run(): Promise<void> {\n let closing = false;\n let lastFrame: AudioFrame | undefined;\n\n const sendTokenizerStream = new tokenizeBasic.SentenceTokenizer().stream();\n const eventChannel = createStreamChannel<TtsServerEvent>();\n const requestId = shortuuid('tts_request_');\n const inputSentEvent = new Event();\n\n // Signal for protocol-driven completion (when 'done' message is received)\n const completionFuture = new Future<void>();\n\n const resourceCleanup = async () => {\n if (closing) return;\n closing = true;\n sendTokenizerStream.close();\n // close() returns a promise; don't leak it\n await eventChannel.close();\n };\n\n const sendClientEvent = async (event: TtsClientEvent, ws: WebSocket, signal: AbortSignal) => {\n // Don't send events to a closed WebSocket or aborted controller\n if (signal.aborted || closing) return;\n\n const validatedEvent = await ttsClientEventSchema.parseAsync(event);\n if (ws.readyState !== WebSocket.OPEN) {\n this.#logger.warn('Trying to send client TTS event to a closed WebSocket');\n return;\n }\n ws.send(JSON.stringify(validatedEvent));\n };\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const createInputTask = async (signal: AbortSignal) => {\n for await (const data of this.input) {\n if (signal.aborted || closing) break;\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n sendTokenizerStream.flush();\n continue;\n }\n sendTokenizerStream.pushText(data);\n }\n // Only call endInput if the stream hasn't been closed by cleanup\n if (!closing) {\n sendTokenizerStream.endInput();\n }\n };\n\n const createSentenceStreamTask = async (ws: WebSocket, signal: AbortSignal) => {\n for await (const ev of sendTokenizerStream) {\n if (signal.aborted || closing) break;\n\n await sendClientEvent(\n {\n type: 'input_transcript',\n transcript: ev.token + ' ',\n },\n ws,\n signal,\n );\n inputSentEvent.set();\n }\n\n await sendClientEvent({ type: 'session.flush' }, ws, signal);\n // needed in case empty input is sent\n inputSentEvent.set();\n };\n\n // Handles WebSocket message routing and error handling\n // Completes based on protocol messages, NOT on ws.close()\n const createWsListenerTask = async (ws: WebSocket, signal: AbortSignal) => {\n const onMessage = (data: Buffer) => {\n try {\n const eventJson = JSON.parse(data.toString()) as Record<string, unknown>;\n const validatedEvent = ttsServerEventSchema.parse(eventJson);\n // writer.write returns a promise; avoid unhandled rejections if stream is closed\n void eventChannel.write(validatedEvent).catch((error) => {\n this.#logger.debug(\n { error },\n 'Failed writing TTS event to stream channel (likely closed)',\n );\n });\n } catch (e) {\n this.#logger.error({ error: e }, 'Error parsing WebSocket message');\n }\n };\n\n const onError = (e: Error) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n void resourceCleanup();\n try {\n // If the ws is misbehaving, hard-stop it immediately to avoid buffering.\n ws.terminate?.();\n } catch {\n // ignore\n }\n // Ensure this ws is not reused\n this.tts.pool.remove(ws);\n completionFuture.reject(e);\n };\n\n const onClose = () => {\n // WebSocket closed unexpectedly (not by us)\n if (!closing) {\n this.#logger.error('WebSocket closed unexpectedly');\n void resourceCleanup();\n // Ensure this ws is not reused\n this.tts.pool.remove(ws);\n completionFuture.reject(\n new APIStatusError({\n message: 'Gateway connection closed unexpectedly',\n options: { requestId },\n }),\n );\n }\n };\n\n const onAbort = () => {\n void resourceCleanup();\n try {\n // On interruption/abort, close the websocket immediately so the server stops streaming\n // and the ws library doesn't buffer unread frames in memory.\n ws.terminate?.();\n } catch {\n // ignore\n }\n this.tts.pool.remove(ws);\n inputSentEvent.set();\n completionFuture.resolve();\n };\n\n // Attach listeners\n ws.on('message', onMessage);\n ws.on('error', onError);\n ws.on('close', onClose);\n signal.addEventListener('abort', onAbort);\n\n try {\n // Wait for protocol-driven completion or error\n await completionFuture.await;\n } finally {\n // IMPORTANT: Remove listeners so connection can be reused\n ws.off('message', onMessage);\n ws.off('error', onError);\n ws.off('close', onClose);\n signal.removeEventListener('abort', onAbort);\n }\n };\n\n const createRecvTask = async (signal: AbortSignal) => {\n let currentSessionId: string | null = null;\n\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n await inputSentEvent.wait();\n\n while (!this.closed && !signal.aborted) {\n const result = await reader.read();\n if (signal.aborted) return;\n if (result.done) return;\n\n const serverEvent = result.value;\n switch (serverEvent.type) {\n case 'session.created':\n currentSessionId = serverEvent.session_id;\n break;\n case 'output_audio':\n const base64Data = new Int8Array(Buffer.from(serverEvent.audio, 'base64'));\n for (const frame of bstream.write(base64Data.buffer)) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n break;\n case 'done':\n for (const frame of bstream.flush()) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n sendLastFrame(currentSessionId!, true);\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n await resourceCleanup();\n completionFuture.resolve();\n return;\n case 'session.closed':\n await resourceCleanup();\n completionFuture.resolve();\n return;\n case 'error':\n this.#logger.error(\n { serverEvent },\n 'Received error message from LiveKit TTS WebSocket',\n );\n await resourceCleanup();\n completionFuture.reject(\n new APIError(`LiveKit TTS returned error: ${serverEvent.message}`),\n );\n return;\n default:\n this.#logger.warn('Unexpected message %s', serverEvent);\n break;\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n await this.tts.pool.withConnection(\n async (ws: WebSocket) => {\n try {\n // IMPORTANT: don't cancel the stream's controller on normal completion,\n // otherwise the pool will remove+close the ws and every run becomes a pool miss.\n const runController = new AbortController();\n const onStreamAbort = () => runController.abort(this.abortController.signal.reason);\n this.abortController.signal.addEventListener('abort', onStreamAbort, { once: true });\n\n const tasks = [\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createInputTask(combined);\n },\n undefined,\n 'inference-tts-input',\n ),\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createSentenceStreamTask(ws, combined);\n },\n undefined,\n 'inference-tts-sentence',\n ),\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createWsListenerTask(ws, combined);\n },\n undefined,\n 'inference-tts-ws-listener',\n ),\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createRecvTask(combined);\n },\n undefined,\n 'inference-tts-recv',\n ),\n ];\n\n try {\n await Promise.all(tasks.map((t) => t.result));\n } finally {\n // Mirror python finally: unblock recv and cancel all tasks.\n inputSentEvent.set();\n await resourceCleanup();\n await cancelAndWait(tasks, 5000);\n this.abortController.signal.removeEventListener('abort', onStreamAbort);\n }\n } catch (e) {\n // If aborted, don't throw - let cleanup handle it\n if (e instanceof Error && e.name === 'AbortError') {\n return;\n }\n throw e;\n }\n },\n {\n timeout: this.connOptions.timeoutMs,\n },\n );\n } catch (e) {\n // Handle connection errors\n if (e instanceof Error && e.name === 'AbortError') {\n // Abort is expected during normal shutdown\n return;\n }\n throw e;\n } finally {\n // Ensure cleanup always runs (and don't leak the promise)\n await resourceCleanup();\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,gBAA0B;AAC1B,wBAAyC;AACzC,mBAAgC;AAChC,6BAA+B;AAC/B,iBAAoB;AACpB,4BAAoC;AACpC,sBAAuC;AAEvC,iBAAyE;AACzE,mBAAoE;AACpE,mBAA8E;AAC9E,wBAKO;AACP,IAAAA,gBAA6D;AA2EtD,SAAS,oBAAoB,OAA6C;AAC/E,QAAM,MAAM,MAAM,YAAY,GAAG;AACjC,MAAI,QAAQ,IAAI;AACd,WAAO,CAAC,MAAM,MAAM,GAAG,GAAG,GAAG,MAAM,MAAM,MAAM,CAAC,CAAC;AAAA,EACnD;AACA,SAAO,CAAC,OAAO,MAAS;AAC1B;AAeO,SAAS,qBACd,UACoB;AACpB,QAAM,eAAe,CAAC,UAAkD;AACtE,QAAI,OAAO,UAAU,UAAU;AAC7B,YAAM,CAAC,MAAM,KAAK,IAAI,oBAAoB,KAAK;AAC/C,aAAO,EAAE,OAAO,MAAM,OAAO,SAAS,GAAG;AAAA,IAC3C;AACA,WAAO;AAAA,EACT;AAEA,MAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,WAAO,SAAS,IAAI,YAAY;AAAA,EAClC;AACA,SAAO,CAAC,aAAa,QAAQ,CAAC;AAChC;AAIA,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,eAAe;AACrB,MAAM,mBAAmB;AAmBlB,MAAM,YAAsC,WAAAC,IAAQ;AAAA,EACjD;AAAA,EACA,UAAyC,oBAAI,IAAI;AAAA,EACzD;AAAA,EAEA,cAAU,gBAAI;AAAA,EAEd,YAAY,MAYT;AACD,UAAM,cAAa,6BAAM,eAAc;AACvC,UAAM,YAAY,GAAG,EAAE,WAAW,KAAK,CAAC;AAExC,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA;AAAA,MACA,eAAe,CAAC;AAAA,MAChB;AAAA,MACA;AAAA,IACF,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,WAAW,QAAQ,IAAI,yBAAyB;AAClE,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAGA,QAAI,YAAY;AAChB,QAAI,YAAY;AAChB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,iBAAiB,UAAU,MAAM,MAAM,CAAC;AAC9C,YAAI,aAAa,cAAc,gBAAgB;AAC7C,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,OAAO,WAAW,OAAO,UAAU;AAAA,UACvC;AAAA,QACF,OAAO;AACL,sBAAY;AAAA,QACd;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AAEA,UAAM,qBAAqB,WAAW,qBAAqB,QAAQ,IAAI;AAEvE,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,OAAO;AAAA,MACP;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,MACA,UAAU;AAAA,MACV,aAAa,eAAe;AAAA,IAC9B;AAGA,SAAK,OAAO,IAAI,sCAA0B;AAAA,MACxC,WAAW,CAAC,YAAY,KAAK,UAAU,OAAO;AAAA,MAC9C,SAAS,CAAC,OAAO,KAAK,QAAQ,EAAE;AAAA,MAChC,oBAAoB;AAAA,MACpB,oBAAoB;AAAA,MACpB,gBAAgB;AAAA;AAAA,IAClB,CAAC;AAAA,EACH;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,UAAM,CAAC,OAAO,KAAK,IAAI,oBAAoB,WAAW;AACtD,WAAO,IAAI,IAAI,EAAE,OAAO,OAAO,SAAS,OAAU,CAAC;AAAA,EACrD;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,WAAW,GAA0B;AACnC,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AAAA,EAEA,OAAO,SAAyE;AAC9E,UAAM,EAAE,cAAc,KAAK,KAAK,eAAe,yCAA4B,IAAI,WAAW,CAAC;AAC3F,UAAM,SAAS,IAAI,iBAAiB,MAAM,EAAE,GAAG,KAAK,KAAK,GAAG,WAAW;AACvE,SAAK,QAAQ,IAAI,MAAM;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AAvRvD;AAwRI,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,UAAM,iCAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS;AAAA,MACb,MAAM;AAAA,MACN,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,MACxC,UAAU,KAAK,KAAK;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,IACnB;AAEA,QAAI,KAAK,KAAK,MAAO,CAAC,OAAmC,QAAQ,KAAK,KAAK;AAC3E,QAAI,KAAK,KAAK,MAAO,CAAC,OAAmC,QAAQ,KAAK,KAAK;AAC3E,QAAI,KAAK,KAAK,SAAU,CAAC,OAAmC,WAAW,KAAK,KAAK;AAEjF,SAAI,UAAK,KAAK,aAAV,mBAAoB,QAAQ;AAC9B,aAAO,WAAW;AAAA,QAChB,QAAQ,KAAK,KAAK,SAAS,IAAI,CAAC,OAAO;AAAA,UACrC,OAAO,EAAE;AAAA,UACT,OAAO,EAAE;AAAA,UACT,OAAO,EAAE,eAAe,CAAC;AAAA,QAC3B,EAAE;AAAA,MACJ;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,aAAa;AACzB,aAAO,aAAa;AAAA,QAClB,SAAS,KAAK,KAAK,YAAY,YAAY;AAAA,QAC3C,SAAS,KAAK,KAAK,YAAY;AAAA,MACjC;AAAA,IACF;AAEA,SAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6DAA6D;AACzF,UAAM,SAAS,UAAM,yBAAU,KAAK,SAAS,OAAO;AACpD,WAAO,KAAK,KAAK,UAAU,MAAM,CAAC;AAClC,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,QAAQ,IAAe;AAC3B,UAAM,GAAG,MAAM;AAAA,EACjB;AAAA,EAEA,UAAgB;AACd,SAAK,KAAK,QAAQ;AAAA,EACpB;AAAA,EAEA,MAAM,QAAQ;AACZ,eAAW,UAAU,KAAK,SAAS;AACjC,YAAM,OAAO,MAAM;AAAA,IACrB;AACA,SAAK,QAAQ,MAAM;AACnB,UAAM,KAAK,KAAK,MAAM;AAAA,EACxB;AACF;AAEO,MAAM,yBAAmD,WAAAC,iBAAqB;AAAA,EAC3E;AAAA,EACA;AAAA,EAER,cAAU,gBAAI;AAAA,EAEd,YAAY,KAAkB,MAAmC,aAAgC;AAC/F,UAAM,KAAK,WAAW;AACtB,SAAK,OAAO;AACZ,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAAA,EACtC;AAAA,EAEA,MAAgB,MAAqB;AACnC,QAAI,UAAU;AACd,QAAI;AAEJ,UAAM,sBAAsB,IAAI,gBAAAC,MAAc,kBAAkB,EAAE,OAAO;AACzE,UAAM,mBAAe,2CAAoC;AACzD,UAAM,gBAAY,wBAAU,cAAc;AAC1C,UAAM,iBAAiB,IAAI,mBAAM;AAGjC,UAAM,mBAAmB,IAAI,oBAAa;AAE1C,UAAM,kBAAkB,YAAY;AAClC,UAAI,QAAS;AACb,gBAAU;AACV,0BAAoB,MAAM;AAE1B,YAAM,aAAa,MAAM;AAAA,IAC3B;AAEA,UAAM,kBAAkB,OAAO,OAAuB,IAAe,WAAwB;AAE3F,UAAI,OAAO,WAAW,QAAS;AAE/B,YAAM,iBAAiB,MAAM,uCAAqB,WAAW,KAAK;AAClE,UAAI,GAAG,eAAe,oBAAU,MAAM;AACpC,aAAK,QAAQ,KAAK,uDAAuD;AACzE;AAAA,MACF;AACA,SAAG,KAAK,KAAK,UAAU,cAAc,CAAC;AAAA,IACxC;AAEA,UAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,UAAI,WAAW;AACb,aAAK,MAAM,IAAI,EAAE,WAAW,WAAW,OAAO,WAAW,MAAM,CAAC;AAChE,oBAAY;AAAA,MACd;AAAA,IACF;AAEA,UAAM,kBAAkB,OAAO,WAAwB;AACrD,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,OAAO,WAAW,QAAS;AAC/B,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,8BAAoB,MAAM;AAC1B;AAAA,QACF;AACA,4BAAoB,SAAS,IAAI;AAAA,MACnC;AAEA,UAAI,CAAC,SAAS;AACZ,4BAAoB,SAAS;AAAA,MAC/B;AAAA,IACF;AAEA,UAAM,2BAA2B,OAAO,IAAe,WAAwB;AAC7E,uBAAiB,MAAM,qBAAqB;AAC1C,YAAI,OAAO,WAAW,QAAS;AAE/B,cAAM;AAAA,UACJ;AAAA,YACE,MAAM;AAAA,YACN,YAAY,GAAG,QAAQ;AAAA,UACzB;AAAA,UACA;AAAA,UACA;AAAA,QACF;AACA,uBAAe,IAAI;AAAA,MACrB;AAEA,YAAM,gBAAgB,EAAE,MAAM,gBAAgB,GAAG,IAAI,MAAM;AAE3D,qBAAe,IAAI;AAAA,IACrB;AAIA,UAAM,uBAAuB,OAAO,IAAe,WAAwB;AACzE,YAAM,YAAY,CAAC,SAAiB;AAClC,YAAI;AACF,gBAAM,YAAY,KAAK,MAAM,KAAK,SAAS,CAAC;AAC5C,gBAAM,iBAAiB,uCAAqB,MAAM,SAAS;AAE3D,eAAK,aAAa,MAAM,cAAc,EAAE,MAAM,CAAC,UAAU;AACvD,iBAAK,QAAQ;AAAA,cACX,EAAE,MAAM;AAAA,cACR;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iCAAiC;AAAA,QACpE;AAAA,MACF;AAEA,YAAM,UAAU,CAAC,MAAa;AArcpC;AAscQ,aAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,aAAK,gBAAgB;AACrB,YAAI;AAEF,mBAAG,cAAH;AAAA,QACF,QAAQ;AAAA,QAER;AAEA,aAAK,IAAI,KAAK,OAAO,EAAE;AACvB,yBAAiB,OAAO,CAAC;AAAA,MAC3B;AAEA,YAAM,UAAU,MAAM;AAEpB,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,+BAA+B;AAClD,eAAK,gBAAgB;AAErB,eAAK,IAAI,KAAK,OAAO,EAAE;AACvB,2BAAiB;AAAA,YACf,IAAI,iCAAe;AAAA,cACjB,SAAS;AAAA,cACT,SAAS,EAAE,UAAU;AAAA,YACvB,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AAEA,YAAM,UAAU,MAAM;AAne5B;AAoeQ,aAAK,gBAAgB;AACrB,YAAI;AAGF,mBAAG,cAAH;AAAA,QACF,QAAQ;AAAA,QAER;AACA,aAAK,IAAI,KAAK,OAAO,EAAE;AACvB,uBAAe,IAAI;AACnB,yBAAiB,QAAQ;AAAA,MAC3B;AAGA,SAAG,GAAG,WAAW,SAAS;AAC1B,SAAG,GAAG,SAAS,OAAO;AACtB,SAAG,GAAG,SAAS,OAAO;AACtB,aAAO,iBAAiB,SAAS,OAAO;AAExC,UAAI;AAEF,cAAM,iBAAiB;AAAA,MACzB,UAAE;AAEA,WAAG,IAAI,WAAW,SAAS;AAC3B,WAAG,IAAI,SAAS,OAAO;AACvB,WAAG,IAAI,SAAS,OAAO;AACvB,eAAO,oBAAoB,SAAS,OAAO;AAAA,MAC7C;AAAA,IACF;AAEA,UAAM,iBAAiB,OAAO,WAAwB;AACpD,UAAI,mBAAkC;AAEtC,YAAM,UAAU,IAAI,6BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,YAAM,oBAAoB,aAAa,OAAO;AAC9C,YAAM,SAAS,kBAAkB,UAAU;AAE3C,UAAI;AACF,cAAM,eAAe,KAAK;AAE1B,eAAO,CAAC,KAAK,UAAU,CAAC,OAAO,SAAS;AACtC,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,OAAO,QAAS;AACpB,cAAI,OAAO,KAAM;AAEjB,gBAAM,cAAc,OAAO;AAC3B,kBAAQ,YAAY,MAAM;AAAA,YACxB,KAAK;AACH,iCAAmB,YAAY;AAC/B;AAAA,YACF,KAAK;AACH,oBAAM,aAAa,IAAI,UAAU,OAAO,KAAK,YAAY,OAAO,QAAQ,CAAC;AACzE,yBAAW,SAAS,QAAQ,MAAM,WAAW,MAAM,GAAG;AACpD,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA;AAAA,YACF,KAAK;AACH,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA,4BAAc,kBAAmB,IAAI;AACrC,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAC7C,oBAAM,gBAAgB;AACtB,+BAAiB,QAAQ;AACzB;AAAA,YACF,KAAK;AACH,oBAAM,gBAAgB;AACtB,+BAAiB,QAAQ;AACzB;AAAA,YACF,KAAK;AACH,mBAAK,QAAQ;AAAA,gBACX,EAAE,YAAY;AAAA,gBACd;AAAA,cACF;AACA,oBAAM,gBAAgB;AACtB,+BAAiB;AAAA,gBACf,IAAI,2BAAS,+BAA+B,YAAY,OAAO,EAAE;AAAA,cACnE;AACA;AAAA,YACF;AACE,mBAAK,QAAQ,KAAK,yBAAyB,WAAW;AACtD;AAAA,UACJ;AAAA,QACF;AAAA,MACF,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,kBAAkB,OAAO;AAAA,QACjC,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,QACxF;AAAA,MACF;AAAA,IACF;AAEA,QAAI;AACF,YAAM,KAAK,IAAI,KAAK;AAAA,QAClB,OAAO,OAAkB;AACvB,cAAI;AAGF,kBAAM,gBAAgB,IAAI,gBAAgB;AAC1C,kBAAM,gBAAgB,MAAM,cAAc,MAAM,KAAK,gBAAgB,OAAO,MAAM;AAClF,iBAAK,gBAAgB,OAAO,iBAAiB,SAAS,eAAe,EAAE,MAAM,KAAK,CAAC;AAEnF,kBAAM,QAAQ;AAAA,cACZ,kBAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,eAAW,6BAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,gBAAgB,QAAQ;AAAA,gBAChC;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,cACA,kBAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,eAAW,6BAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,yBAAyB,IAAI,QAAQ;AAAA,gBAC7C;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,cACA,kBAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,eAAW,6BAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,qBAAqB,IAAI,QAAQ;AAAA,gBACzC;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,cACA,kBAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,eAAW,6BAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,eAAe,QAAQ;AAAA,gBAC/B;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,YACF;AAEA,gBAAI;AACF,oBAAM,QAAQ,IAAI,MAAM,IAAI,CAAC,MAAM,EAAE,MAAM,CAAC;AAAA,YAC9C,UAAE;AAEA,6BAAe,IAAI;AACnB,oBAAM,gBAAgB;AACtB,wBAAM,4BAAc,OAAO,GAAI;AAC/B,mBAAK,gBAAgB,OAAO,oBAAoB,SAAS,aAAa;AAAA,YACxE;AAAA,UACF,SAAS,GAAG;AAEV,gBAAI,aAAa,SAAS,EAAE,SAAS,cAAc;AACjD;AAAA,YACF;AACA,kBAAM;AAAA,UACR;AAAA,QACF;AAAA,QACA;AAAA,UACE,SAAS,KAAK,YAAY;AAAA,QAC5B;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AAEV,UAAI,aAAa,SAAS,EAAE,SAAS,cAAc;AAEjD;AAAA,MACF;AACA,YAAM;AAAA,IACR,UAAE;AAEA,YAAM,gBAAgB;AAAA,IACxB;AAAA,EACF;AACF;","names":["import_utils","BaseTTS","BaseSynthesizeStream","tokenizeBasic"]}
1
+ {"version":3,"sources":["../../src/inference/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { ConnectionPool } from '../connection_pool.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport { basic as tokenizeBasic } from '../tokenize/index.js';\nimport type { ChunkedStream } from '../tts/index.js';\nimport { SynthesizeStream as BaseSynthesizeStream, TTS as BaseTTS } from '../tts/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { Event, Future, Task, cancelAndWait, combineSignals, shortuuid } from '../utils.js';\nimport {\n type TtsClientEvent,\n type TtsServerEvent,\n ttsClientEventSchema,\n ttsServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken } from './utils.js';\n\nexport type CartesiaModels =\n | 'cartesia/sonic-3'\n | 'cartesia/sonic-2'\n | 'cartesia/sonic-turbo'\n | 'cartesia/sonic';\n\nexport type DeepgramTTSModels = 'deepgram/aura' | 'deepgram/aura-2';\n\nexport type ElevenlabsModels =\n | 'elevenlabs/eleven_flash_v2'\n | 'elevenlabs/eleven_flash_v2_5'\n | 'elevenlabs/eleven_turbo_v2'\n | 'elevenlabs/eleven_turbo_v2_5'\n | 'elevenlabs/eleven_multilingual_v2';\n\nexport type InworldModels =\n | 'inworld/inworld-tts-1.5-max'\n | 'inworld/inworld-tts-1.5-mini'\n | 'inworld/inworld-tts-1-max'\n | 'inworld/inworld-tts-1';\n\nexport type RimeModels = 'rime/arcana' | 'rime/mistv2';\n\nexport interface CartesiaOptions {\n /** Maximum duration of audio in seconds. */\n duration?: number;\n /** Speech speed. Default: not specified. */\n speed?: 'slow' | 'normal' | 'fast';\n}\n\nexport interface ElevenlabsOptions {\n /** Inactivity timeout in seconds. Default: 60. */\n inactivity_timeout?: number;\n /** Text normalization mode. Default: \"auto\". */\n apply_text_normalization?: 'auto' | 'off' | 'on';\n}\n\nexport interface DeepgramTTSOptions {}\n\nexport interface RimeOptions {}\n\nexport interface InworldOptions {\n /** Controls how fast the voice speaks. 1.0 is normal speed, 0.5 is half, 1.5 is 1.5x. Default: 1.0. */\n speaking_rate?: number;\n /** Controls randomness in the output. Recommended between 0.6 and 1.1. Default: 1.1. */\n temperature?: number;\n /** Controls text normalization. \"ON\" expands numbers, dates, abbreviations. \"OFF\" reads text as written. Default: \"ON\". */\n text_normalization?: 'ON' | 'OFF';\n}\n\ntype _TTSModels =\n | CartesiaModels\n | DeepgramTTSModels\n | ElevenlabsModels\n | RimeModels\n | InworldModels;\n\nexport type TTSModels =\n | CartesiaModels\n | DeepgramTTSModels\n | ElevenlabsModels\n | RimeModels\n | InworldModels\n | AnyString;\n\nexport type ModelWithVoice = `${_TTSModels}:${string}` | TTSModels;\n\nexport type TTSOptions<TModel extends TTSModels> = TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends DeepgramTTSModels\n ? DeepgramTTSOptions\n : TModel extends ElevenlabsModels\n ? ElevenlabsOptions\n : TModel extends RimeModels\n ? RimeOptions\n : TModel extends InworldModels\n ? InworldOptions\n : Record<string, unknown>;\n\n/** Parse a model string into [model, voice]. Voice is undefined if not specified. */\nexport function parseTTSModelString(model: string): [string, string | undefined] {\n const idx = model.lastIndexOf(':');\n if (idx !== -1) {\n return [model.slice(0, idx), model.slice(idx + 1)];\n }\n return [model, undefined];\n}\n\n/** A fallback model with optional extra configuration. Extra fields are passed through to the provider. */\nexport interface TTSFallbackModel {\n /** Model name (e.g. \"cartesia/sonic\", \"elevenlabs/eleven_flash_v2\", \"rime/arcana\"). */\n model: string;\n /** Voice to use for the model. */\n voice: string;\n /** Extra configuration for the model. */\n extraKwargs?: Record<string, unknown>;\n}\n\nexport type TTSFallbackModelType = TTSFallbackModel | string;\n\n/** Normalize a single or list of FallbackModelType into TTSFallbackModel[]. */\nexport function normalizeTTSFallback(\n fallback: TTSFallbackModelType | TTSFallbackModelType[],\n): TTSFallbackModel[] {\n const makeFallback = (model: TTSFallbackModelType): TTSFallbackModel => {\n if (typeof model === 'string') {\n const [name, voice] = parseTTSModelString(model);\n return { model: name, voice: voice ?? '' };\n }\n return model;\n };\n\n if (Array.isArray(fallback)) {\n return fallback.map(makeFallback);\n }\n return [makeFallback(fallback)];\n}\n\ntype TTSEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: TTSEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'https://agent-gateway.livekit.cloud/v1';\nconst NUM_CHANNELS = 1;\nconst DEFAULT_LANGUAGE = 'en';\n\nexport interface InferenceTTSOptions<TModel extends TTSModels> {\n model?: TModel;\n voice?: string;\n language?: string;\n encoding: TTSEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: TTSOptions<TModel>;\n fallback?: TTSFallbackModel[];\n connOptions?: APIConnectOptions;\n}\n\n/**\n * Livekit Cloud Inference TTS\n */\nexport class TTS<TModel extends TTSModels> extends BaseTTS {\n private opts: InferenceTTSOptions<TModel>;\n private streams: Set<SynthesizeStream<TModel>> = new Set();\n pool: ConnectionPool<WebSocket>;\n\n #logger = log();\n\n constructor(opts: {\n model: TModel;\n voice?: string;\n language?: string;\n baseURL?: string;\n encoding?: TTSEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: TTSOptions<TModel>;\n fallback?: TTSFallbackModelType | TTSFallbackModelType[];\n connOptions?: APIConnectOptions;\n }) {\n const sampleRate = opts?.sampleRate ?? DEFAULT_SAMPLE_RATE;\n super(sampleRate, 1, { streaming: true });\n\n const {\n model,\n voice,\n language = DEFAULT_LANGUAGE,\n baseURL,\n encoding = DEFAULT_ENCODING,\n apiKey,\n apiSecret,\n modelOptions = {} as TTSOptions<TModel>,\n fallback,\n connOptions,\n } = opts || {};\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n // read voice id from the model if provided: \"provider/model:voice_id\"\n let nextModel = model;\n let nextVoice = voice;\n if (typeof nextModel === 'string') {\n const idx = nextModel.lastIndexOf(':');\n if (idx !== -1) {\n const voiceFromModel = nextModel.slice(idx + 1);\n if (nextVoice && nextVoice !== voiceFromModel) {\n this.#logger.warn(\n '`voice` is provided via both argument and model, using the one from the argument',\n { voice: nextVoice, model: nextModel },\n );\n } else {\n nextVoice = voiceFromModel;\n }\n nextModel = nextModel.slice(0, idx) as TModel;\n }\n }\n\n const normalizedFallback = fallback ? normalizeTTSFallback(fallback) : undefined;\n\n this.opts = {\n model: nextModel,\n voice: nextVoice,\n language,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n fallback: normalizedFallback,\n connOptions: connOptions ?? DEFAULT_API_CONNECT_OPTIONS,\n };\n\n // Initialize connection pool\n this.pool = new ConnectionPool<WebSocket>({\n connectCb: (timeout) => this.connectWs(timeout),\n closeCb: (ws) => this.closeWs(ws),\n maxSessionDuration: 300_000,\n markRefreshedOnGet: true,\n connectTimeout: 10_000, // 10 seconds default\n });\n }\n\n get label() {\n return 'inference.TTS';\n }\n\n static fromModelString(modelString: string): TTS<AnyString> {\n const [model, voice] = parseTTSModelString(modelString);\n return new TTS({ model, voice: voice || undefined });\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n synthesize(_: string): ChunkedStream {\n throw new Error('ChunkedStream is not implemented');\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream<TModel> {\n const { connOptions = this.opts.connOptions ?? DEFAULT_API_CONNECT_OPTIONS } = options || {};\n const stream = new SynthesizeStream(this, { ...this.opts }, connOptions);\n this.streams.add(stream);\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/tts`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const params = {\n type: 'session.create',\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n } as Record<string, unknown>;\n\n if (this.opts.voice) (params as Record<string, unknown>).voice = this.opts.voice;\n if (this.opts.model) (params as Record<string, unknown>).model = this.opts.model;\n if (this.opts.language) (params as Record<string, unknown>).language = this.opts.language;\n\n if (this.opts.fallback?.length) {\n params.fallback = {\n models: this.opts.fallback.map((m) => ({\n model: m.model,\n voice: m.voice,\n extra: m.extraKwargs ?? {},\n })),\n };\n }\n\n if (this.opts.connOptions) {\n params.connection = {\n timeout: this.opts.connOptions.timeoutMs / 1000,\n retries: this.opts.connOptions.maxRetry,\n };\n }\n\n this.#logger.debug({ url }, 'inference.TTS creating new websocket connection (pool miss)');\n const socket = await connectWs(url, headers, timeout);\n socket.send(JSON.stringify(params));\n return socket;\n }\n\n async closeWs(ws: WebSocket) {\n await ws.close();\n }\n\n prewarm(): void {\n this.pool.prewarm();\n }\n\n async close() {\n for (const stream of this.streams) {\n await stream.close();\n }\n this.streams.clear();\n await this.pool.close();\n }\n}\n\nexport class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeStream {\n private opts: InferenceTTSOptions<TModel>;\n private tts: TTS<TModel>;\n\n #logger = log();\n\n constructor(tts: TTS<TModel>, opts: InferenceTTSOptions<TModel>, connOptions: APIConnectOptions) {\n super(tts, connOptions);\n this.opts = opts;\n this.tts = tts;\n }\n\n get label() {\n return 'inference.SynthesizeStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n }\n\n protected async run(): Promise<void> {\n let closing = false;\n let lastFrame: AudioFrame | undefined;\n\n const sendTokenizerStream = new tokenizeBasic.SentenceTokenizer().stream();\n const eventChannel = createStreamChannel<TtsServerEvent>();\n const requestId = shortuuid('tts_request_');\n const inputSentEvent = new Event();\n\n // Signal for protocol-driven completion (when 'done' message is received)\n const completionFuture = new Future<void>();\n\n const resourceCleanup = async () => {\n if (closing) return;\n closing = true;\n sendTokenizerStream.close();\n // close() returns a promise; don't leak it\n await eventChannel.close();\n };\n\n const sendClientEvent = async (event: TtsClientEvent, ws: WebSocket, signal: AbortSignal) => {\n // Don't send events to a closed WebSocket or aborted controller\n if (signal.aborted || closing) return;\n\n const validatedEvent = await ttsClientEventSchema.parseAsync(event);\n if (ws.readyState !== WebSocket.OPEN) {\n this.#logger.warn('Trying to send client TTS event to a closed WebSocket');\n return;\n }\n ws.send(JSON.stringify(validatedEvent));\n };\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const createInputTask = async (signal: AbortSignal) => {\n for await (const data of this.input) {\n if (signal.aborted || closing) break;\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n sendTokenizerStream.flush();\n continue;\n }\n sendTokenizerStream.pushText(data);\n }\n // Only call endInput if the stream hasn't been closed by cleanup\n if (!closing) {\n sendTokenizerStream.endInput();\n }\n };\n\n const createSentenceStreamTask = async (ws: WebSocket, signal: AbortSignal) => {\n for await (const ev of sendTokenizerStream) {\n if (signal.aborted || closing) break;\n\n await sendClientEvent(\n {\n type: 'input_transcript',\n transcript: ev.token + ' ',\n },\n ws,\n signal,\n );\n inputSentEvent.set();\n }\n\n await sendClientEvent({ type: 'session.flush' }, ws, signal);\n // needed in case empty input is sent\n inputSentEvent.set();\n };\n\n // Handles WebSocket message routing and error handling\n // Completes based on protocol messages, NOT on ws.close()\n const createWsListenerTask = async (ws: WebSocket, signal: AbortSignal) => {\n const onMessage = (data: Buffer) => {\n try {\n const eventJson = JSON.parse(data.toString()) as Record<string, unknown>;\n const validatedEvent = ttsServerEventSchema.parse(eventJson);\n // writer.write returns a promise; avoid unhandled rejections if stream is closed\n void eventChannel.write(validatedEvent).catch((error) => {\n this.#logger.debug(\n { error },\n 'Failed writing TTS event to stream channel (likely closed)',\n );\n });\n } catch (e) {\n this.#logger.error({ error: e }, 'Error parsing WebSocket message');\n }\n };\n\n const onError = (e: Error) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n void resourceCleanup();\n try {\n // If the ws is misbehaving, hard-stop it immediately to avoid buffering.\n ws.terminate?.();\n } catch {\n // ignore\n }\n // Ensure this ws is not reused\n this.tts.pool.remove(ws);\n completionFuture.reject(e);\n };\n\n const onClose = () => {\n // WebSocket closed unexpectedly (not by us)\n if (!closing) {\n this.#logger.error('WebSocket closed unexpectedly');\n void resourceCleanup();\n // Ensure this ws is not reused\n this.tts.pool.remove(ws);\n completionFuture.reject(\n new APIStatusError({\n message: 'Gateway connection closed unexpectedly',\n options: { requestId },\n }),\n );\n }\n };\n\n const onAbort = () => {\n void resourceCleanup();\n try {\n // On interruption/abort, close the websocket immediately so the server stops streaming\n // and the ws library doesn't buffer unread frames in memory.\n ws.terminate?.();\n } catch {\n // ignore\n }\n this.tts.pool.remove(ws);\n inputSentEvent.set();\n completionFuture.resolve();\n };\n\n // Attach listeners\n ws.on('message', onMessage);\n ws.on('error', onError);\n ws.on('close', onClose);\n signal.addEventListener('abort', onAbort);\n\n try {\n // Wait for protocol-driven completion or error\n await completionFuture.await;\n } finally {\n // IMPORTANT: Remove listeners so connection can be reused\n ws.off('message', onMessage);\n ws.off('error', onError);\n ws.off('close', onClose);\n signal.removeEventListener('abort', onAbort);\n }\n };\n\n const createRecvTask = async (signal: AbortSignal) => {\n let currentSessionId: string | null = null;\n\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n await inputSentEvent.wait();\n\n while (!this.closed && !signal.aborted) {\n const result = await reader.read();\n if (signal.aborted) return;\n if (result.done) return;\n\n const serverEvent = result.value;\n switch (serverEvent.type) {\n case 'session.created':\n currentSessionId = serverEvent.session_id;\n break;\n case 'output_audio':\n const base64Data = new Int8Array(Buffer.from(serverEvent.audio, 'base64'));\n for (const frame of bstream.write(base64Data.buffer)) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n break;\n case 'done':\n for (const frame of bstream.flush()) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n sendLastFrame(currentSessionId!, true);\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n await resourceCleanup();\n completionFuture.resolve();\n return;\n case 'session.closed':\n await resourceCleanup();\n completionFuture.resolve();\n return;\n case 'error':\n this.#logger.error(\n { serverEvent },\n 'Received error message from LiveKit TTS WebSocket',\n );\n await resourceCleanup();\n completionFuture.reject(\n new APIError(`LiveKit TTS returned error: ${serverEvent.message}`),\n );\n return;\n default:\n this.#logger.warn('Unexpected message %s', serverEvent);\n break;\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n await this.tts.pool.withConnection(\n async (ws: WebSocket) => {\n try {\n // IMPORTANT: don't cancel the stream's controller on normal completion,\n // otherwise the pool will remove+close the ws and every run becomes a pool miss.\n const runController = new AbortController();\n const onStreamAbort = () => runController.abort(this.abortController.signal.reason);\n this.abortController.signal.addEventListener('abort', onStreamAbort, { once: true });\n\n const tasks = [\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createInputTask(combined);\n },\n undefined,\n 'inference-tts-input',\n ),\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createSentenceStreamTask(ws, combined);\n },\n undefined,\n 'inference-tts-sentence',\n ),\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createWsListenerTask(ws, combined);\n },\n undefined,\n 'inference-tts-ws-listener',\n ),\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createRecvTask(combined);\n },\n undefined,\n 'inference-tts-recv',\n ),\n ];\n\n try {\n await Promise.all(tasks.map((t) => t.result));\n } finally {\n // Mirror python finally: unblock recv and cancel all tasks.\n inputSentEvent.set();\n await resourceCleanup();\n await cancelAndWait(tasks, 5000);\n this.abortController.signal.removeEventListener('abort', onStreamAbort);\n }\n } catch (e) {\n // If aborted, don't throw - let cleanup handle it\n if (e instanceof Error && e.name === 'AbortError') {\n return;\n }\n throw e;\n }\n },\n {\n timeout: this.connOptions.timeoutMs,\n },\n );\n } catch (e) {\n // Handle connection errors\n if (e instanceof Error && e.name === 'AbortError') {\n // Abort is expected during normal shutdown\n return;\n }\n throw e;\n } finally {\n // Ensure cleanup always runs (and don't leak the promise)\n await resourceCleanup();\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,gBAA0B;AAC1B,wBAAyC;AACzC,mBAAgC;AAChC,6BAA+B;AAC/B,iBAAoB;AACpB,4BAAoC;AACpC,sBAAuC;AAEvC,iBAAyE;AACzE,mBAAoE;AACpE,mBAA8E;AAC9E,wBAKO;AACP,IAAAA,gBAA6D;AAkFtD,SAAS,oBAAoB,OAA6C;AAC/E,QAAM,MAAM,MAAM,YAAY,GAAG;AACjC,MAAI,QAAQ,IAAI;AACd,WAAO,CAAC,MAAM,MAAM,GAAG,GAAG,GAAG,MAAM,MAAM,MAAM,CAAC,CAAC;AAAA,EACnD;AACA,SAAO,CAAC,OAAO,MAAS;AAC1B;AAeO,SAAS,qBACd,UACoB;AACpB,QAAM,eAAe,CAAC,UAAkD;AACtE,QAAI,OAAO,UAAU,UAAU;AAC7B,YAAM,CAAC,MAAM,KAAK,IAAI,oBAAoB,KAAK;AAC/C,aAAO,EAAE,OAAO,MAAM,OAAO,SAAS,GAAG;AAAA,IAC3C;AACA,WAAO;AAAA,EACT;AAEA,MAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,WAAO,SAAS,IAAI,YAAY;AAAA,EAClC;AACA,SAAO,CAAC,aAAa,QAAQ,CAAC;AAChC;AAIA,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,eAAe;AACrB,MAAM,mBAAmB;AAmBlB,MAAM,YAAsC,WAAAC,IAAQ;AAAA,EACjD;AAAA,EACA,UAAyC,oBAAI,IAAI;AAAA,EACzD;AAAA,EAEA,cAAU,gBAAI;AAAA,EAEd,YAAY,MAYT;AACD,UAAM,cAAa,6BAAM,eAAc;AACvC,UAAM,YAAY,GAAG,EAAE,WAAW,KAAK,CAAC;AAExC,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA;AAAA,MACA,eAAe,CAAC;AAAA,MAChB;AAAA,MACA;AAAA,IACF,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,WAAW,QAAQ,IAAI,yBAAyB;AAClE,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAGA,QAAI,YAAY;AAChB,QAAI,YAAY;AAChB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,iBAAiB,UAAU,MAAM,MAAM,CAAC;AAC9C,YAAI,aAAa,cAAc,gBAAgB;AAC7C,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,OAAO,WAAW,OAAO,UAAU;AAAA,UACvC;AAAA,QACF,OAAO;AACL,sBAAY;AAAA,QACd;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AAEA,UAAM,qBAAqB,WAAW,qBAAqB,QAAQ,IAAI;AAEvE,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,OAAO;AAAA,MACP;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,MACA,UAAU;AAAA,MACV,aAAa,eAAe;AAAA,IAC9B;AAGA,SAAK,OAAO,IAAI,sCAA0B;AAAA,MACxC,WAAW,CAAC,YAAY,KAAK,UAAU,OAAO;AAAA,MAC9C,SAAS,CAAC,OAAO,KAAK,QAAQ,EAAE;AAAA,MAChC,oBAAoB;AAAA,MACpB,oBAAoB;AAAA,MACpB,gBAAgB;AAAA;AAAA,IAClB,CAAC;AAAA,EACH;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,UAAM,CAAC,OAAO,KAAK,IAAI,oBAAoB,WAAW;AACtD,WAAO,IAAI,IAAI,EAAE,OAAO,OAAO,SAAS,OAAU,CAAC;AAAA,EACrD;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,WAAW,GAA0B;AACnC,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AAAA,EAEA,OAAO,SAAyE;AAC9E,UAAM,EAAE,cAAc,KAAK,KAAK,eAAe,yCAA4B,IAAI,WAAW,CAAC;AAC3F,UAAM,SAAS,IAAI,iBAAiB,MAAM,EAAE,GAAG,KAAK,KAAK,GAAG,WAAW;AACvE,SAAK,QAAQ,IAAI,MAAM;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AA9RvD;AA+RI,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,UAAM,iCAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS;AAAA,MACb,MAAM;AAAA,MACN,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,MACxC,UAAU,KAAK,KAAK;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,IACnB;AAEA,QAAI,KAAK,KAAK,MAAO,CAAC,OAAmC,QAAQ,KAAK,KAAK;AAC3E,QAAI,KAAK,KAAK,MAAO,CAAC,OAAmC,QAAQ,KAAK,KAAK;AAC3E,QAAI,KAAK,KAAK,SAAU,CAAC,OAAmC,WAAW,KAAK,KAAK;AAEjF,SAAI,UAAK,KAAK,aAAV,mBAAoB,QAAQ;AAC9B,aAAO,WAAW;AAAA,QAChB,QAAQ,KAAK,KAAK,SAAS,IAAI,CAAC,OAAO;AAAA,UACrC,OAAO,EAAE;AAAA,UACT,OAAO,EAAE;AAAA,UACT,OAAO,EAAE,eAAe,CAAC;AAAA,QAC3B,EAAE;AAAA,MACJ;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,aAAa;AACzB,aAAO,aAAa;AAAA,QAClB,SAAS,KAAK,KAAK,YAAY,YAAY;AAAA,QAC3C,SAAS,KAAK,KAAK,YAAY;AAAA,MACjC;AAAA,IACF;AAEA,SAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6DAA6D;AACzF,UAAM,SAAS,UAAM,yBAAU,KAAK,SAAS,OAAO;AACpD,WAAO,KAAK,KAAK,UAAU,MAAM,CAAC;AAClC,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,QAAQ,IAAe;AAC3B,UAAM,GAAG,MAAM;AAAA,EACjB;AAAA,EAEA,UAAgB;AACd,SAAK,KAAK,QAAQ;AAAA,EACpB;AAAA,EAEA,MAAM,QAAQ;AACZ,eAAW,UAAU,KAAK,SAAS;AACjC,YAAM,OAAO,MAAM;AAAA,IACrB;AACA,SAAK,QAAQ,MAAM;AACnB,UAAM,KAAK,KAAK,MAAM;AAAA,EACxB;AACF;AAEO,MAAM,yBAAmD,WAAAC,iBAAqB;AAAA,EAC3E;AAAA,EACA;AAAA,EAER,cAAU,gBAAI;AAAA,EAEd,YAAY,KAAkB,MAAmC,aAAgC;AAC/F,UAAM,KAAK,WAAW;AACtB,SAAK,OAAO;AACZ,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAAA,EACtC;AAAA,EAEA,MAAgB,MAAqB;AACnC,QAAI,UAAU;AACd,QAAI;AAEJ,UAAM,sBAAsB,IAAI,gBAAAC,MAAc,kBAAkB,EAAE,OAAO;AACzE,UAAM,mBAAe,2CAAoC;AACzD,UAAM,gBAAY,wBAAU,cAAc;AAC1C,UAAM,iBAAiB,IAAI,mBAAM;AAGjC,UAAM,mBAAmB,IAAI,oBAAa;AAE1C,UAAM,kBAAkB,YAAY;AAClC,UAAI,QAAS;AACb,gBAAU;AACV,0BAAoB,MAAM;AAE1B,YAAM,aAAa,MAAM;AAAA,IAC3B;AAEA,UAAM,kBAAkB,OAAO,OAAuB,IAAe,WAAwB;AAE3F,UAAI,OAAO,WAAW,QAAS;AAE/B,YAAM,iBAAiB,MAAM,uCAAqB,WAAW,KAAK;AAClE,UAAI,GAAG,eAAe,oBAAU,MAAM;AACpC,aAAK,QAAQ,KAAK,uDAAuD;AACzE;AAAA,MACF;AACA,SAAG,KAAK,KAAK,UAAU,cAAc,CAAC;AAAA,IACxC;AAEA,UAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,UAAI,WAAW;AACb,aAAK,MAAM,IAAI,EAAE,WAAW,WAAW,OAAO,WAAW,MAAM,CAAC;AAChE,oBAAY;AAAA,MACd;AAAA,IACF;AAEA,UAAM,kBAAkB,OAAO,WAAwB;AACrD,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,OAAO,WAAW,QAAS;AAC/B,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,8BAAoB,MAAM;AAC1B;AAAA,QACF;AACA,4BAAoB,SAAS,IAAI;AAAA,MACnC;AAEA,UAAI,CAAC,SAAS;AACZ,4BAAoB,SAAS;AAAA,MAC/B;AAAA,IACF;AAEA,UAAM,2BAA2B,OAAO,IAAe,WAAwB;AAC7E,uBAAiB,MAAM,qBAAqB;AAC1C,YAAI,OAAO,WAAW,QAAS;AAE/B,cAAM;AAAA,UACJ;AAAA,YACE,MAAM;AAAA,YACN,YAAY,GAAG,QAAQ;AAAA,UACzB;AAAA,UACA;AAAA,UACA;AAAA,QACF;AACA,uBAAe,IAAI;AAAA,MACrB;AAEA,YAAM,gBAAgB,EAAE,MAAM,gBAAgB,GAAG,IAAI,MAAM;AAE3D,qBAAe,IAAI;AAAA,IACrB;AAIA,UAAM,uBAAuB,OAAO,IAAe,WAAwB;AACzE,YAAM,YAAY,CAAC,SAAiB;AAClC,YAAI;AACF,gBAAM,YAAY,KAAK,MAAM,KAAK,SAAS,CAAC;AAC5C,gBAAM,iBAAiB,uCAAqB,MAAM,SAAS;AAE3D,eAAK,aAAa,MAAM,cAAc,EAAE,MAAM,CAAC,UAAU;AACvD,iBAAK,QAAQ;AAAA,cACX,EAAE,MAAM;AAAA,cACR;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iCAAiC;AAAA,QACpE;AAAA,MACF;AAEA,YAAM,UAAU,CAAC,MAAa;AA5cpC;AA6cQ,aAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,aAAK,gBAAgB;AACrB,YAAI;AAEF,mBAAG,cAAH;AAAA,QACF,QAAQ;AAAA,QAER;AAEA,aAAK,IAAI,KAAK,OAAO,EAAE;AACvB,yBAAiB,OAAO,CAAC;AAAA,MAC3B;AAEA,YAAM,UAAU,MAAM;AAEpB,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,+BAA+B;AAClD,eAAK,gBAAgB;AAErB,eAAK,IAAI,KAAK,OAAO,EAAE;AACvB,2BAAiB;AAAA,YACf,IAAI,iCAAe;AAAA,cACjB,SAAS;AAAA,cACT,SAAS,EAAE,UAAU;AAAA,YACvB,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AAEA,YAAM,UAAU,MAAM;AA1e5B;AA2eQ,aAAK,gBAAgB;AACrB,YAAI;AAGF,mBAAG,cAAH;AAAA,QACF,QAAQ;AAAA,QAER;AACA,aAAK,IAAI,KAAK,OAAO,EAAE;AACvB,uBAAe,IAAI;AACnB,yBAAiB,QAAQ;AAAA,MAC3B;AAGA,SAAG,GAAG,WAAW,SAAS;AAC1B,SAAG,GAAG,SAAS,OAAO;AACtB,SAAG,GAAG,SAAS,OAAO;AACtB,aAAO,iBAAiB,SAAS,OAAO;AAExC,UAAI;AAEF,cAAM,iBAAiB;AAAA,MACzB,UAAE;AAEA,WAAG,IAAI,WAAW,SAAS;AAC3B,WAAG,IAAI,SAAS,OAAO;AACvB,WAAG,IAAI,SAAS,OAAO;AACvB,eAAO,oBAAoB,SAAS,OAAO;AAAA,MAC7C;AAAA,IACF;AAEA,UAAM,iBAAiB,OAAO,WAAwB;AACpD,UAAI,mBAAkC;AAEtC,YAAM,UAAU,IAAI,6BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,YAAM,oBAAoB,aAAa,OAAO;AAC9C,YAAM,SAAS,kBAAkB,UAAU;AAE3C,UAAI;AACF,cAAM,eAAe,KAAK;AAE1B,eAAO,CAAC,KAAK,UAAU,CAAC,OAAO,SAAS;AACtC,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,OAAO,QAAS;AACpB,cAAI,OAAO,KAAM;AAEjB,gBAAM,cAAc,OAAO;AAC3B,kBAAQ,YAAY,MAAM;AAAA,YACxB,KAAK;AACH,iCAAmB,YAAY;AAC/B;AAAA,YACF,KAAK;AACH,oBAAM,aAAa,IAAI,UAAU,OAAO,KAAK,YAAY,OAAO,QAAQ,CAAC;AACzE,yBAAW,SAAS,QAAQ,MAAM,WAAW,MAAM,GAAG;AACpD,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA;AAAA,YACF,KAAK;AACH,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA,4BAAc,kBAAmB,IAAI;AACrC,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAC7C,oBAAM,gBAAgB;AACtB,+BAAiB,QAAQ;AACzB;AAAA,YACF,KAAK;AACH,oBAAM,gBAAgB;AACtB,+BAAiB,QAAQ;AACzB;AAAA,YACF,KAAK;AACH,mBAAK,QAAQ;AAAA,gBACX,EAAE,YAAY;AAAA,gBACd;AAAA,cACF;AACA,oBAAM,gBAAgB;AACtB,+BAAiB;AAAA,gBACf,IAAI,2BAAS,+BAA+B,YAAY,OAAO,EAAE;AAAA,cACnE;AACA;AAAA,YACF;AACE,mBAAK,QAAQ,KAAK,yBAAyB,WAAW;AACtD;AAAA,UACJ;AAAA,QACF;AAAA,MACF,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,kBAAkB,OAAO;AAAA,QACjC,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,QACxF;AAAA,MACF;AAAA,IACF;AAEA,QAAI;AACF,YAAM,KAAK,IAAI,KAAK;AAAA,QAClB,OAAO,OAAkB;AACvB,cAAI;AAGF,kBAAM,gBAAgB,IAAI,gBAAgB;AAC1C,kBAAM,gBAAgB,MAAM,cAAc,MAAM,KAAK,gBAAgB,OAAO,MAAM;AAClF,iBAAK,gBAAgB,OAAO,iBAAiB,SAAS,eAAe,EAAE,MAAM,KAAK,CAAC;AAEnF,kBAAM,QAAQ;AAAA,cACZ,kBAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,eAAW,6BAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,gBAAgB,QAAQ;AAAA,gBAChC;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,cACA,kBAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,eAAW,6BAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,yBAAyB,IAAI,QAAQ;AAAA,gBAC7C;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,cACA,kBAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,eAAW,6BAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,qBAAqB,IAAI,QAAQ;AAAA,gBACzC;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,cACA,kBAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,eAAW,6BAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,eAAe,QAAQ;AAAA,gBAC/B;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,YACF;AAEA,gBAAI;AACF,oBAAM,QAAQ,IAAI,MAAM,IAAI,CAAC,MAAM,EAAE,MAAM,CAAC;AAAA,YAC9C,UAAE;AAEA,6BAAe,IAAI;AACnB,oBAAM,gBAAgB;AACtB,wBAAM,4BAAc,OAAO,GAAI;AAC/B,mBAAK,gBAAgB,OAAO,oBAAoB,SAAS,aAAa;AAAA,YACxE;AAAA,UACF,SAAS,GAAG;AAEV,gBAAI,aAAa,SAAS,EAAE,SAAS,cAAc;AACjD;AAAA,YACF;AACA,kBAAM;AAAA,UACR;AAAA,QACF;AAAA,QACA;AAAA,UACE,SAAS,KAAK,YAAY;AAAA,QAC5B;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AAEV,UAAI,aAAa,SAAS,EAAE,SAAS,cAAc;AAEjD;AAAA,MACF;AACA,YAAM;AAAA,IACR,UAAE;AAEA,YAAM,gBAAgB;AAAA,IACxB;AAAA,EACF;AACF;","names":["import_utils","BaseTTS","BaseSynthesizeStream","tokenizeBasic"]}
@@ -26,6 +26,12 @@ export interface DeepgramTTSOptions {
26
26
  export interface RimeOptions {
27
27
  }
28
28
  export interface InworldOptions {
29
+ /** Controls how fast the voice speaks. 1.0 is normal speed, 0.5 is half, 1.5 is 1.5x. Default: 1.0. */
30
+ speaking_rate?: number;
31
+ /** Controls randomness in the output. Recommended between 0.6 and 1.1. Default: 1.1. */
32
+ temperature?: number;
33
+ /** Controls text normalization. "ON" expands numbers, dates, abbreviations. "OFF" reads text as written. Default: "ON". */
34
+ text_normalization?: 'ON' | 'OFF';
29
35
  }
30
36
  type _TTSModels = CartesiaModels | DeepgramTTSModels | ElevenlabsModels | RimeModels | InworldModels;
31
37
  export type TTSModels = CartesiaModels | DeepgramTTSModels | ElevenlabsModels | RimeModels | InworldModels | AnyString;
@@ -26,6 +26,12 @@ export interface DeepgramTTSOptions {
26
26
  export interface RimeOptions {
27
27
  }
28
28
  export interface InworldOptions {
29
+ /** Controls how fast the voice speaks. 1.0 is normal speed, 0.5 is half, 1.5 is 1.5x. Default: 1.0. */
30
+ speaking_rate?: number;
31
+ /** Controls randomness in the output. Recommended between 0.6 and 1.1. Default: 1.1. */
32
+ temperature?: number;
33
+ /** Controls text normalization. "ON" expands numbers, dates, abbreviations. "OFF" reads text as written. Default: "ON". */
34
+ text_normalization?: 'ON' | 'OFF';
29
35
  }
30
36
  type _TTSModels = CartesiaModels | DeepgramTTSModels | ElevenlabsModels | RimeModels | InworldModels;
31
37
  export type TTSModels = CartesiaModels | DeepgramTTSModels | ElevenlabsModels | RimeModels | InworldModels | AnyString;
@@ -1 +1 @@
1
- {"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../../src/inference/tts.ts"],"names":[],"mappings":"AAIA,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAG/B,OAAO,EAAE,cAAc,EAAE,MAAM,uBAAuB,CAAC;AAIvD,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AACrD,OAAO,EAAE,gBAAgB,IAAI,oBAAoB,EAAE,GAAG,IAAI,OAAO,EAAE,MAAM,iBAAiB,CAAC;AAC3F,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AAQlF,OAAO,EAAE,KAAK,SAAS,EAAgC,MAAM,YAAY,CAAC;AAE1E,MAAM,MAAM,cAAc,GACtB,kBAAkB,GAClB,kBAAkB,GAClB,sBAAsB,GACtB,gBAAgB,CAAC;AAErB,MAAM,MAAM,iBAAiB,GAAG,eAAe,GAAG,iBAAiB,CAAC;AAEpE,MAAM,MAAM,gBAAgB,GACxB,4BAA4B,GAC5B,8BAA8B,GAC9B,4BAA4B,GAC5B,8BAA8B,GAC9B,mCAAmC,CAAC;AAExC,MAAM,MAAM,aAAa,GACrB,6BAA6B,GAC7B,8BAA8B,GAC9B,2BAA2B,GAC3B,uBAAuB,CAAC;AAE5B,MAAM,MAAM,UAAU,GAAG,aAAa,GAAG,aAAa,CAAC;AAEvD,MAAM,WAAW,eAAe;IAC9B,4CAA4C;IAC5C,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,4CAA4C;IAC5C,KAAK,CAAC,EAAE,MAAM,GAAG,QAAQ,GAAG,MAAM,CAAC;CACpC;AAED,MAAM,WAAW,iBAAiB;IAChC,kDAAkD;IAClD,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,gDAAgD;IAChD,wBAAwB,CAAC,EAAE,MAAM,GAAG,KAAK,GAAG,IAAI,CAAC;CAClD;AAED,MAAM,WAAW,kBAAkB;CAAG;AAEtC,MAAM,WAAW,WAAW;CAAG;AAE/B,MAAM,WAAW,cAAc;CAAG;AAElC,KAAK,UAAU,GACX,cAAc,GACd,iBAAiB,GACjB,gBAAgB,GAChB,UAAU,GACV,aAAa,CAAC;AAElB,MAAM,MAAM,SAAS,GACjB,cAAc,GACd,iBAAiB,GACjB,gBAAgB,GAChB,UAAU,GACV,aAAa,GACb,SAAS,CAAC;AAEd,MAAM,MAAM,cAAc,GAAG,GAAG,UAAU,IAAI,MAAM,EAAE,GAAG,SAAS,CAAC;AAEnE,MAAM,MAAM,UAAU,CAAC,MAAM,SAAS,SAAS,IAAI,MAAM,SAAS,cAAc,GAC5E,eAAe,GACf,MAAM,SAAS,iBAAiB,GAC9B,kBAAkB,GAClB,MAAM,SAAS,gBAAgB,GAC7B,iBAAiB,GACjB,MAAM,SAAS,UAAU,GACvB,WAAW,GACX,MAAM,SAAS,aAAa,GAC1B,cAAc,GACd,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAEpC,qFAAqF;AACrF,wBAAgB,mBAAmB,CAAC,KAAK,EAAE,MAAM,GAAG,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,CAAC,CAM/E;AAED,2GAA2G;AAC3G,MAAM,WAAW,gBAAgB;IAC/B,uFAAuF;IACvF,KAAK,EAAE,MAAM,CAAC;IACd,kCAAkC;IAClC,KAAK,EAAE,MAAM,CAAC;IACd,yCAAyC;IACzC,WAAW,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACvC;AAED,MAAM,MAAM,oBAAoB,GAAG,gBAAgB,GAAG,MAAM,CAAC;AAE7D,+EAA+E;AAC/E,wBAAgB,oBAAoB,CAClC,QAAQ,EAAE,oBAAoB,GAAG,oBAAoB,EAAE,GACtD,gBAAgB,EAAE,CAapB;AAED,KAAK,WAAW,GAAG,WAAW,CAAC;AAQ/B,MAAM,WAAW,mBAAmB,CAAC,MAAM,SAAS,SAAS;IAC3D,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,YAAY,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;IACjC,QAAQ,CAAC,EAAE,gBAAgB,EAAE,CAAC;IAC9B,WAAW,CAAC,EAAE,iBAAiB,CAAC;CACjC;AAED;;GAEG;AACH,qBAAa,GAAG,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,OAAO;;IACxD,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,OAAO,CAA4C;IAC3D,IAAI,EAAE,cAAc,CAAC,SAAS,CAAC,CAAC;gBAIpB,IAAI,EAAE;QAChB,KAAK,EAAE,MAAM,CAAC;QACd,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,QAAQ,CAAC,EAAE,MAAM,CAAC;QAClB,OAAO,CAAC,EAAE,MAAM,CAAC;QACjB,QAAQ,CAAC,EAAE,WAAW,CAAC;QACvB,UAAU,CAAC,EAAE,MAAM,CAAC;QACpB,MAAM,CAAC,EAAE,MAAM,CAAC;QAChB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;QAClC,QAAQ,CAAC,EAAE,oBAAoB,GAAG,oBAAoB,EAAE,CAAC;QACzD,WAAW,CAAC,EAAE,iBAAiB,CAAC;KACjC;IA0ED,IAAI,KAAK,WAER;IAED,MAAM,CAAC,eAAe,CAAC,WAAW,EAAE,MAAM,GAAG,GAAG,CAAC,SAAS,CAAC;IAK3D,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,OAAO,GAAG,UAAU,CAAC,CAAC;IAO9F,UAAU,CAAC,CAAC,EAAE,MAAM,GAAG,aAAa;IAIpC,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,gBAAgB,CAAC,MAAM,CAAC;IAOzE,SAAS,CAAC,OAAO,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC;IA4C9C,OAAO,CAAC,EAAE,EAAE,SAAS;IAI3B,OAAO,IAAI,IAAI;IAIT,KAAK;CAOZ;AAED,qBAAa,gBAAgB,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,oBAAoB;;IAClF,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,GAAG,CAAc;gBAIb,GAAG,EAAE,GAAG,CAAC,MAAM,CAAC,EAAE,IAAI,EAAE,mBAAmB,CAAC,MAAM,CAAC,EAAE,WAAW,EAAE,iBAAiB;IAM/F,IAAI,KAAK,WAER;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,OAAO,GAAG,UAAU,CAAC,CAAC;cAI9E,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;CA2SrC"}
1
+ {"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../../src/inference/tts.ts"],"names":[],"mappings":"AAIA,OAAO,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AAG/B,OAAO,EAAE,cAAc,EAAE,MAAM,uBAAuB,CAAC;AAIvD,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AACrD,OAAO,EAAE,gBAAgB,IAAI,oBAAoB,EAAE,GAAG,IAAI,OAAO,EAAE,MAAM,iBAAiB,CAAC;AAC3F,OAAO,EAAE,KAAK,iBAAiB,EAA+B,MAAM,aAAa,CAAC;AAQlF,OAAO,EAAE,KAAK,SAAS,EAAgC,MAAM,YAAY,CAAC;AAE1E,MAAM,MAAM,cAAc,GACtB,kBAAkB,GAClB,kBAAkB,GAClB,sBAAsB,GACtB,gBAAgB,CAAC;AAErB,MAAM,MAAM,iBAAiB,GAAG,eAAe,GAAG,iBAAiB,CAAC;AAEpE,MAAM,MAAM,gBAAgB,GACxB,4BAA4B,GAC5B,8BAA8B,GAC9B,4BAA4B,GAC5B,8BAA8B,GAC9B,mCAAmC,CAAC;AAExC,MAAM,MAAM,aAAa,GACrB,6BAA6B,GAC7B,8BAA8B,GAC9B,2BAA2B,GAC3B,uBAAuB,CAAC;AAE5B,MAAM,MAAM,UAAU,GAAG,aAAa,GAAG,aAAa,CAAC;AAEvD,MAAM,WAAW,eAAe;IAC9B,4CAA4C;IAC5C,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,4CAA4C;IAC5C,KAAK,CAAC,EAAE,MAAM,GAAG,QAAQ,GAAG,MAAM,CAAC;CACpC;AAED,MAAM,WAAW,iBAAiB;IAChC,kDAAkD;IAClD,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,gDAAgD;IAChD,wBAAwB,CAAC,EAAE,MAAM,GAAG,KAAK,GAAG,IAAI,CAAC;CAClD;AAED,MAAM,WAAW,kBAAkB;CAAG;AAEtC,MAAM,WAAW,WAAW;CAAG;AAE/B,MAAM,WAAW,cAAc;IAC7B,uGAAuG;IACvG,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,wFAAwF;IACxF,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,2HAA2H;IAC3H,kBAAkB,CAAC,EAAE,IAAI,GAAG,KAAK,CAAC;CACnC;AAED,KAAK,UAAU,GACX,cAAc,GACd,iBAAiB,GACjB,gBAAgB,GAChB,UAAU,GACV,aAAa,CAAC;AAElB,MAAM,MAAM,SAAS,GACjB,cAAc,GACd,iBAAiB,GACjB,gBAAgB,GAChB,UAAU,GACV,aAAa,GACb,SAAS,CAAC;AAEd,MAAM,MAAM,cAAc,GAAG,GAAG,UAAU,IAAI,MAAM,EAAE,GAAG,SAAS,CAAC;AAEnE,MAAM,MAAM,UAAU,CAAC,MAAM,SAAS,SAAS,IAAI,MAAM,SAAS,cAAc,GAC5E,eAAe,GACf,MAAM,SAAS,iBAAiB,GAC9B,kBAAkB,GAClB,MAAM,SAAS,gBAAgB,GAC7B,iBAAiB,GACjB,MAAM,SAAS,UAAU,GACvB,WAAW,GACX,MAAM,SAAS,aAAa,GAC1B,cAAc,GACd,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAEpC,qFAAqF;AACrF,wBAAgB,mBAAmB,CAAC,KAAK,EAAE,MAAM,GAAG,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,CAAC,CAM/E;AAED,2GAA2G;AAC3G,MAAM,WAAW,gBAAgB;IAC/B,uFAAuF;IACvF,KAAK,EAAE,MAAM,CAAC;IACd,kCAAkC;IAClC,KAAK,EAAE,MAAM,CAAC;IACd,yCAAyC;IACzC,WAAW,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACvC;AAED,MAAM,MAAM,oBAAoB,GAAG,gBAAgB,GAAG,MAAM,CAAC;AAE7D,+EAA+E;AAC/E,wBAAgB,oBAAoB,CAClC,QAAQ,EAAE,oBAAoB,GAAG,oBAAoB,EAAE,GACtD,gBAAgB,EAAE,CAapB;AAED,KAAK,WAAW,GAAG,WAAW,CAAC;AAQ/B,MAAM,WAAW,mBAAmB,CAAC,MAAM,SAAS,SAAS;IAC3D,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,YAAY,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;IACjC,QAAQ,CAAC,EAAE,gBAAgB,EAAE,CAAC;IAC9B,WAAW,CAAC,EAAE,iBAAiB,CAAC;CACjC;AAED;;GAEG;AACH,qBAAa,GAAG,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,OAAO;;IACxD,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,OAAO,CAA4C;IAC3D,IAAI,EAAE,cAAc,CAAC,SAAS,CAAC,CAAC;gBAIpB,IAAI,EAAE;QAChB,KAAK,EAAE,MAAM,CAAC;QACd,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,QAAQ,CAAC,EAAE,MAAM,CAAC;QAClB,OAAO,CAAC,EAAE,MAAM,CAAC;QACjB,QAAQ,CAAC,EAAE,WAAW,CAAC;QACvB,UAAU,CAAC,EAAE,MAAM,CAAC;QACpB,MAAM,CAAC,EAAE,MAAM,CAAC;QAChB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC;QAClC,QAAQ,CAAC,EAAE,oBAAoB,GAAG,oBAAoB,EAAE,CAAC;QACzD,WAAW,CAAC,EAAE,iBAAiB,CAAC;KACjC;IA0ED,IAAI,KAAK,WAER;IAED,MAAM,CAAC,eAAe,CAAC,WAAW,EAAE,MAAM,GAAG,GAAG,CAAC,SAAS,CAAC;IAK3D,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,OAAO,GAAG,UAAU,CAAC,CAAC;IAO9F,UAAU,CAAC,CAAC,EAAE,MAAM,GAAG,aAAa;IAIpC,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,gBAAgB,CAAC,MAAM,CAAC;IAOzE,SAAS,CAAC,OAAO,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC;IA4C9C,OAAO,CAAC,EAAE,EAAE,SAAS;IAI3B,OAAO,IAAI,IAAI;IAIT,KAAK;CAOZ;AAED,qBAAa,gBAAgB,CAAC,MAAM,SAAS,SAAS,CAAE,SAAQ,oBAAoB;;IAClF,OAAO,CAAC,IAAI,CAA8B;IAC1C,OAAO,CAAC,GAAG,CAAc;gBAIb,GAAG,EAAE,GAAG,CAAC,MAAM,CAAC,EAAE,IAAI,EAAE,mBAAmB,CAAC,MAAM,CAAC,EAAE,WAAW,EAAE,iBAAiB;IAM/F,IAAI,KAAK,WAER;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,EAAE,OAAO,GAAG,OAAO,GAAG,UAAU,CAAC,CAAC;cAI9E,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;CA2SrC"}
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/inference/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { ConnectionPool } from '../connection_pool.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport { basic as tokenizeBasic } from '../tokenize/index.js';\nimport type { ChunkedStream } from '../tts/index.js';\nimport { SynthesizeStream as BaseSynthesizeStream, TTS as BaseTTS } from '../tts/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { Event, Future, Task, cancelAndWait, combineSignals, shortuuid } from '../utils.js';\nimport {\n type TtsClientEvent,\n type TtsServerEvent,\n ttsClientEventSchema,\n ttsServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken } from './utils.js';\n\nexport type CartesiaModels =\n | 'cartesia/sonic-3'\n | 'cartesia/sonic-2'\n | 'cartesia/sonic-turbo'\n | 'cartesia/sonic';\n\nexport type DeepgramTTSModels = 'deepgram/aura' | 'deepgram/aura-2';\n\nexport type ElevenlabsModels =\n | 'elevenlabs/eleven_flash_v2'\n | 'elevenlabs/eleven_flash_v2_5'\n | 'elevenlabs/eleven_turbo_v2'\n | 'elevenlabs/eleven_turbo_v2_5'\n | 'elevenlabs/eleven_multilingual_v2';\n\nexport type InworldModels =\n | 'inworld/inworld-tts-1.5-max'\n | 'inworld/inworld-tts-1.5-mini'\n | 'inworld/inworld-tts-1-max'\n | 'inworld/inworld-tts-1';\n\nexport type RimeModels = 'rime/arcana' | 'rime/mistv2';\n\nexport interface CartesiaOptions {\n /** Maximum duration of audio in seconds. */\n duration?: number;\n /** Speech speed. Default: not specified. */\n speed?: 'slow' | 'normal' | 'fast';\n}\n\nexport interface ElevenlabsOptions {\n /** Inactivity timeout in seconds. Default: 60. */\n inactivity_timeout?: number;\n /** Text normalization mode. Default: \"auto\". */\n apply_text_normalization?: 'auto' | 'off' | 'on';\n}\n\nexport interface DeepgramTTSOptions {}\n\nexport interface RimeOptions {}\n\nexport interface InworldOptions {}\n\ntype _TTSModels =\n | CartesiaModels\n | DeepgramTTSModels\n | ElevenlabsModels\n | RimeModels\n | InworldModels;\n\nexport type TTSModels =\n | CartesiaModels\n | DeepgramTTSModels\n | ElevenlabsModels\n | RimeModels\n | InworldModels\n | AnyString;\n\nexport type ModelWithVoice = `${_TTSModels}:${string}` | TTSModels;\n\nexport type TTSOptions<TModel extends TTSModels> = TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends DeepgramTTSModels\n ? DeepgramTTSOptions\n : TModel extends ElevenlabsModels\n ? ElevenlabsOptions\n : TModel extends RimeModels\n ? RimeOptions\n : TModel extends InworldModels\n ? InworldOptions\n : Record<string, unknown>;\n\n/** Parse a model string into [model, voice]. Voice is undefined if not specified. */\nexport function parseTTSModelString(model: string): [string, string | undefined] {\n const idx = model.lastIndexOf(':');\n if (idx !== -1) {\n return [model.slice(0, idx), model.slice(idx + 1)];\n }\n return [model, undefined];\n}\n\n/** A fallback model with optional extra configuration. Extra fields are passed through to the provider. */\nexport interface TTSFallbackModel {\n /** Model name (e.g. \"cartesia/sonic\", \"elevenlabs/eleven_flash_v2\", \"rime/arcana\"). */\n model: string;\n /** Voice to use for the model. */\n voice: string;\n /** Extra configuration for the model. */\n extraKwargs?: Record<string, unknown>;\n}\n\nexport type TTSFallbackModelType = TTSFallbackModel | string;\n\n/** Normalize a single or list of FallbackModelType into TTSFallbackModel[]. */\nexport function normalizeTTSFallback(\n fallback: TTSFallbackModelType | TTSFallbackModelType[],\n): TTSFallbackModel[] {\n const makeFallback = (model: TTSFallbackModelType): TTSFallbackModel => {\n if (typeof model === 'string') {\n const [name, voice] = parseTTSModelString(model);\n return { model: name, voice: voice ?? '' };\n }\n return model;\n };\n\n if (Array.isArray(fallback)) {\n return fallback.map(makeFallback);\n }\n return [makeFallback(fallback)];\n}\n\ntype TTSEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: TTSEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'https://agent-gateway.livekit.cloud/v1';\nconst NUM_CHANNELS = 1;\nconst DEFAULT_LANGUAGE = 'en';\n\nexport interface InferenceTTSOptions<TModel extends TTSModels> {\n model?: TModel;\n voice?: string;\n language?: string;\n encoding: TTSEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: TTSOptions<TModel>;\n fallback?: TTSFallbackModel[];\n connOptions?: APIConnectOptions;\n}\n\n/**\n * Livekit Cloud Inference TTS\n */\nexport class TTS<TModel extends TTSModels> extends BaseTTS {\n private opts: InferenceTTSOptions<TModel>;\n private streams: Set<SynthesizeStream<TModel>> = new Set();\n pool: ConnectionPool<WebSocket>;\n\n #logger = log();\n\n constructor(opts: {\n model: TModel;\n voice?: string;\n language?: string;\n baseURL?: string;\n encoding?: TTSEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: TTSOptions<TModel>;\n fallback?: TTSFallbackModelType | TTSFallbackModelType[];\n connOptions?: APIConnectOptions;\n }) {\n const sampleRate = opts?.sampleRate ?? DEFAULT_SAMPLE_RATE;\n super(sampleRate, 1, { streaming: true });\n\n const {\n model,\n voice,\n language = DEFAULT_LANGUAGE,\n baseURL,\n encoding = DEFAULT_ENCODING,\n apiKey,\n apiSecret,\n modelOptions = {} as TTSOptions<TModel>,\n fallback,\n connOptions,\n } = opts || {};\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n // read voice id from the model if provided: \"provider/model:voice_id\"\n let nextModel = model;\n let nextVoice = voice;\n if (typeof nextModel === 'string') {\n const idx = nextModel.lastIndexOf(':');\n if (idx !== -1) {\n const voiceFromModel = nextModel.slice(idx + 1);\n if (nextVoice && nextVoice !== voiceFromModel) {\n this.#logger.warn(\n '`voice` is provided via both argument and model, using the one from the argument',\n { voice: nextVoice, model: nextModel },\n );\n } else {\n nextVoice = voiceFromModel;\n }\n nextModel = nextModel.slice(0, idx) as TModel;\n }\n }\n\n const normalizedFallback = fallback ? normalizeTTSFallback(fallback) : undefined;\n\n this.opts = {\n model: nextModel,\n voice: nextVoice,\n language,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n fallback: normalizedFallback,\n connOptions: connOptions ?? DEFAULT_API_CONNECT_OPTIONS,\n };\n\n // Initialize connection pool\n this.pool = new ConnectionPool<WebSocket>({\n connectCb: (timeout) => this.connectWs(timeout),\n closeCb: (ws) => this.closeWs(ws),\n maxSessionDuration: 300_000,\n markRefreshedOnGet: true,\n connectTimeout: 10_000, // 10 seconds default\n });\n }\n\n get label() {\n return 'inference.TTS';\n }\n\n static fromModelString(modelString: string): TTS<AnyString> {\n const [model, voice] = parseTTSModelString(modelString);\n return new TTS({ model, voice: voice || undefined });\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n synthesize(_: string): ChunkedStream {\n throw new Error('ChunkedStream is not implemented');\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream<TModel> {\n const { connOptions = this.opts.connOptions ?? DEFAULT_API_CONNECT_OPTIONS } = options || {};\n const stream = new SynthesizeStream(this, { ...this.opts }, connOptions);\n this.streams.add(stream);\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/tts`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const params = {\n type: 'session.create',\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n } as Record<string, unknown>;\n\n if (this.opts.voice) (params as Record<string, unknown>).voice = this.opts.voice;\n if (this.opts.model) (params as Record<string, unknown>).model = this.opts.model;\n if (this.opts.language) (params as Record<string, unknown>).language = this.opts.language;\n\n if (this.opts.fallback?.length) {\n params.fallback = {\n models: this.opts.fallback.map((m) => ({\n model: m.model,\n voice: m.voice,\n extra: m.extraKwargs ?? {},\n })),\n };\n }\n\n if (this.opts.connOptions) {\n params.connection = {\n timeout: this.opts.connOptions.timeoutMs / 1000,\n retries: this.opts.connOptions.maxRetry,\n };\n }\n\n this.#logger.debug({ url }, 'inference.TTS creating new websocket connection (pool miss)');\n const socket = await connectWs(url, headers, timeout);\n socket.send(JSON.stringify(params));\n return socket;\n }\n\n async closeWs(ws: WebSocket) {\n await ws.close();\n }\n\n prewarm(): void {\n this.pool.prewarm();\n }\n\n async close() {\n for (const stream of this.streams) {\n await stream.close();\n }\n this.streams.clear();\n await this.pool.close();\n }\n}\n\nexport class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeStream {\n private opts: InferenceTTSOptions<TModel>;\n private tts: TTS<TModel>;\n\n #logger = log();\n\n constructor(tts: TTS<TModel>, opts: InferenceTTSOptions<TModel>, connOptions: APIConnectOptions) {\n super(tts, connOptions);\n this.opts = opts;\n this.tts = tts;\n }\n\n get label() {\n return 'inference.SynthesizeStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n }\n\n protected async run(): Promise<void> {\n let closing = false;\n let lastFrame: AudioFrame | undefined;\n\n const sendTokenizerStream = new tokenizeBasic.SentenceTokenizer().stream();\n const eventChannel = createStreamChannel<TtsServerEvent>();\n const requestId = shortuuid('tts_request_');\n const inputSentEvent = new Event();\n\n // Signal for protocol-driven completion (when 'done' message is received)\n const completionFuture = new Future<void>();\n\n const resourceCleanup = async () => {\n if (closing) return;\n closing = true;\n sendTokenizerStream.close();\n // close() returns a promise; don't leak it\n await eventChannel.close();\n };\n\n const sendClientEvent = async (event: TtsClientEvent, ws: WebSocket, signal: AbortSignal) => {\n // Don't send events to a closed WebSocket or aborted controller\n if (signal.aborted || closing) return;\n\n const validatedEvent = await ttsClientEventSchema.parseAsync(event);\n if (ws.readyState !== WebSocket.OPEN) {\n this.#logger.warn('Trying to send client TTS event to a closed WebSocket');\n return;\n }\n ws.send(JSON.stringify(validatedEvent));\n };\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const createInputTask = async (signal: AbortSignal) => {\n for await (const data of this.input) {\n if (signal.aborted || closing) break;\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n sendTokenizerStream.flush();\n continue;\n }\n sendTokenizerStream.pushText(data);\n }\n // Only call endInput if the stream hasn't been closed by cleanup\n if (!closing) {\n sendTokenizerStream.endInput();\n }\n };\n\n const createSentenceStreamTask = async (ws: WebSocket, signal: AbortSignal) => {\n for await (const ev of sendTokenizerStream) {\n if (signal.aborted || closing) break;\n\n await sendClientEvent(\n {\n type: 'input_transcript',\n transcript: ev.token + ' ',\n },\n ws,\n signal,\n );\n inputSentEvent.set();\n }\n\n await sendClientEvent({ type: 'session.flush' }, ws, signal);\n // needed in case empty input is sent\n inputSentEvent.set();\n };\n\n // Handles WebSocket message routing and error handling\n // Completes based on protocol messages, NOT on ws.close()\n const createWsListenerTask = async (ws: WebSocket, signal: AbortSignal) => {\n const onMessage = (data: Buffer) => {\n try {\n const eventJson = JSON.parse(data.toString()) as Record<string, unknown>;\n const validatedEvent = ttsServerEventSchema.parse(eventJson);\n // writer.write returns a promise; avoid unhandled rejections if stream is closed\n void eventChannel.write(validatedEvent).catch((error) => {\n this.#logger.debug(\n { error },\n 'Failed writing TTS event to stream channel (likely closed)',\n );\n });\n } catch (e) {\n this.#logger.error({ error: e }, 'Error parsing WebSocket message');\n }\n };\n\n const onError = (e: Error) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n void resourceCleanup();\n try {\n // If the ws is misbehaving, hard-stop it immediately to avoid buffering.\n ws.terminate?.();\n } catch {\n // ignore\n }\n // Ensure this ws is not reused\n this.tts.pool.remove(ws);\n completionFuture.reject(e);\n };\n\n const onClose = () => {\n // WebSocket closed unexpectedly (not by us)\n if (!closing) {\n this.#logger.error('WebSocket closed unexpectedly');\n void resourceCleanup();\n // Ensure this ws is not reused\n this.tts.pool.remove(ws);\n completionFuture.reject(\n new APIStatusError({\n message: 'Gateway connection closed unexpectedly',\n options: { requestId },\n }),\n );\n }\n };\n\n const onAbort = () => {\n void resourceCleanup();\n try {\n // On interruption/abort, close the websocket immediately so the server stops streaming\n // and the ws library doesn't buffer unread frames in memory.\n ws.terminate?.();\n } catch {\n // ignore\n }\n this.tts.pool.remove(ws);\n inputSentEvent.set();\n completionFuture.resolve();\n };\n\n // Attach listeners\n ws.on('message', onMessage);\n ws.on('error', onError);\n ws.on('close', onClose);\n signal.addEventListener('abort', onAbort);\n\n try {\n // Wait for protocol-driven completion or error\n await completionFuture.await;\n } finally {\n // IMPORTANT: Remove listeners so connection can be reused\n ws.off('message', onMessage);\n ws.off('error', onError);\n ws.off('close', onClose);\n signal.removeEventListener('abort', onAbort);\n }\n };\n\n const createRecvTask = async (signal: AbortSignal) => {\n let currentSessionId: string | null = null;\n\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n await inputSentEvent.wait();\n\n while (!this.closed && !signal.aborted) {\n const result = await reader.read();\n if (signal.aborted) return;\n if (result.done) return;\n\n const serverEvent = result.value;\n switch (serverEvent.type) {\n case 'session.created':\n currentSessionId = serverEvent.session_id;\n break;\n case 'output_audio':\n const base64Data = new Int8Array(Buffer.from(serverEvent.audio, 'base64'));\n for (const frame of bstream.write(base64Data.buffer)) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n break;\n case 'done':\n for (const frame of bstream.flush()) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n sendLastFrame(currentSessionId!, true);\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n await resourceCleanup();\n completionFuture.resolve();\n return;\n case 'session.closed':\n await resourceCleanup();\n completionFuture.resolve();\n return;\n case 'error':\n this.#logger.error(\n { serverEvent },\n 'Received error message from LiveKit TTS WebSocket',\n );\n await resourceCleanup();\n completionFuture.reject(\n new APIError(`LiveKit TTS returned error: ${serverEvent.message}`),\n );\n return;\n default:\n this.#logger.warn('Unexpected message %s', serverEvent);\n break;\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n await this.tts.pool.withConnection(\n async (ws: WebSocket) => {\n try {\n // IMPORTANT: don't cancel the stream's controller on normal completion,\n // otherwise the pool will remove+close the ws and every run becomes a pool miss.\n const runController = new AbortController();\n const onStreamAbort = () => runController.abort(this.abortController.signal.reason);\n this.abortController.signal.addEventListener('abort', onStreamAbort, { once: true });\n\n const tasks = [\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createInputTask(combined);\n },\n undefined,\n 'inference-tts-input',\n ),\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createSentenceStreamTask(ws, combined);\n },\n undefined,\n 'inference-tts-sentence',\n ),\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createWsListenerTask(ws, combined);\n },\n undefined,\n 'inference-tts-ws-listener',\n ),\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createRecvTask(combined);\n },\n undefined,\n 'inference-tts-recv',\n ),\n ];\n\n try {\n await Promise.all(tasks.map((t) => t.result));\n } finally {\n // Mirror python finally: unblock recv and cancel all tasks.\n inputSentEvent.set();\n await resourceCleanup();\n await cancelAndWait(tasks, 5000);\n this.abortController.signal.removeEventListener('abort', onStreamAbort);\n }\n } catch (e) {\n // If aborted, don't throw - let cleanup handle it\n if (e instanceof Error && e.name === 'AbortError') {\n return;\n }\n throw e;\n }\n },\n {\n timeout: this.connOptions.timeoutMs,\n },\n );\n } catch (e) {\n // Handle connection errors\n if (e instanceof Error && e.name === 'AbortError') {\n // Abort is expected during normal shutdown\n return;\n }\n throw e;\n } finally {\n // Ensure cleanup always runs (and don't leak the promise)\n await resourceCleanup();\n }\n }\n}\n"],"mappings":"AAIA,SAAS,iBAAiB;AAC1B,SAAS,UAAU,sBAAsB;AACzC,SAAS,uBAAuB;AAChC,SAAS,sBAAsB;AAC/B,SAAS,WAAW;AACpB,SAAS,2BAA2B;AACpC,SAAS,SAAS,qBAAqB;AAEvC,SAAS,oBAAoB,sBAAsB,OAAO,eAAe;AACzE,SAAiC,mCAAmC;AACpE,SAAS,OAAO,QAAQ,MAAM,eAAe,gBAAgB,iBAAiB;AAC9E;AAAA,EAGE;AAAA,EACA;AAAA,OACK;AACP,SAAyB,WAAW,yBAAyB;AA2EtD,SAAS,oBAAoB,OAA6C;AAC/E,QAAM,MAAM,MAAM,YAAY,GAAG;AACjC,MAAI,QAAQ,IAAI;AACd,WAAO,CAAC,MAAM,MAAM,GAAG,GAAG,GAAG,MAAM,MAAM,MAAM,CAAC,CAAC;AAAA,EACnD;AACA,SAAO,CAAC,OAAO,MAAS;AAC1B;AAeO,SAAS,qBACd,UACoB;AACpB,QAAM,eAAe,CAAC,UAAkD;AACtE,QAAI,OAAO,UAAU,UAAU;AAC7B,YAAM,CAAC,MAAM,KAAK,IAAI,oBAAoB,KAAK;AAC/C,aAAO,EAAE,OAAO,MAAM,OAAO,SAAS,GAAG;AAAA,IAC3C;AACA,WAAO;AAAA,EACT;AAEA,MAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,WAAO,SAAS,IAAI,YAAY;AAAA,EAClC;AACA,SAAO,CAAC,aAAa,QAAQ,CAAC;AAChC;AAIA,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,eAAe;AACrB,MAAM,mBAAmB;AAmBlB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAyC,oBAAI,IAAI;AAAA,EACzD;AAAA,EAEA,UAAU,IAAI;AAAA,EAEd,YAAY,MAYT;AACD,UAAM,cAAa,6BAAM,eAAc;AACvC,UAAM,YAAY,GAAG,EAAE,WAAW,KAAK,CAAC;AAExC,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA;AAAA,MACA,eAAe,CAAC;AAAA,MAChB;AAAA,MACA;AAAA,IACF,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,WAAW,QAAQ,IAAI,yBAAyB;AAClE,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAGA,QAAI,YAAY;AAChB,QAAI,YAAY;AAChB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,iBAAiB,UAAU,MAAM,MAAM,CAAC;AAC9C,YAAI,aAAa,cAAc,gBAAgB;AAC7C,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,OAAO,WAAW,OAAO,UAAU;AAAA,UACvC;AAAA,QACF,OAAO;AACL,sBAAY;AAAA,QACd;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AAEA,UAAM,qBAAqB,WAAW,qBAAqB,QAAQ,IAAI;AAEvE,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,OAAO;AAAA,MACP;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,MACA,UAAU;AAAA,MACV,aAAa,eAAe;AAAA,IAC9B;AAGA,SAAK,OAAO,IAAI,eAA0B;AAAA,MACxC,WAAW,CAAC,YAAY,KAAK,UAAU,OAAO;AAAA,MAC9C,SAAS,CAAC,OAAO,KAAK,QAAQ,EAAE;AAAA,MAChC,oBAAoB;AAAA,MACpB,oBAAoB;AAAA,MACpB,gBAAgB;AAAA;AAAA,IAClB,CAAC;AAAA,EACH;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,UAAM,CAAC,OAAO,KAAK,IAAI,oBAAoB,WAAW;AACtD,WAAO,IAAI,IAAI,EAAE,OAAO,OAAO,SAAS,OAAU,CAAC;AAAA,EACrD;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,WAAW,GAA0B;AACnC,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AAAA,EAEA,OAAO,SAAyE;AAC9E,UAAM,EAAE,cAAc,KAAK,KAAK,eAAe,4BAA4B,IAAI,WAAW,CAAC;AAC3F,UAAM,SAAS,IAAI,iBAAiB,MAAM,EAAE,GAAG,KAAK,KAAK,GAAG,WAAW;AACvE,SAAK,QAAQ,IAAI,MAAM;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AAvRvD;AAwRI,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,MAAM,kBAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS;AAAA,MACb,MAAM;AAAA,MACN,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,MACxC,UAAU,KAAK,KAAK;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,IACnB;AAEA,QAAI,KAAK,KAAK,MAAO,CAAC,OAAmC,QAAQ,KAAK,KAAK;AAC3E,QAAI,KAAK,KAAK,MAAO,CAAC,OAAmC,QAAQ,KAAK,KAAK;AAC3E,QAAI,KAAK,KAAK,SAAU,CAAC,OAAmC,WAAW,KAAK,KAAK;AAEjF,SAAI,UAAK,KAAK,aAAV,mBAAoB,QAAQ;AAC9B,aAAO,WAAW;AAAA,QAChB,QAAQ,KAAK,KAAK,SAAS,IAAI,CAAC,OAAO;AAAA,UACrC,OAAO,EAAE;AAAA,UACT,OAAO,EAAE;AAAA,UACT,OAAO,EAAE,eAAe,CAAC;AAAA,QAC3B,EAAE;AAAA,MACJ;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,aAAa;AACzB,aAAO,aAAa;AAAA,QAClB,SAAS,KAAK,KAAK,YAAY,YAAY;AAAA,QAC3C,SAAS,KAAK,KAAK,YAAY;AAAA,MACjC;AAAA,IACF;AAEA,SAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6DAA6D;AACzF,UAAM,SAAS,MAAM,UAAU,KAAK,SAAS,OAAO;AACpD,WAAO,KAAK,KAAK,UAAU,MAAM,CAAC;AAClC,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,QAAQ,IAAe;AAC3B,UAAM,GAAG,MAAM;AAAA,EACjB;AAAA,EAEA,UAAgB;AACd,SAAK,KAAK,QAAQ;AAAA,EACpB;AAAA,EAEA,MAAM,QAAQ;AACZ,eAAW,UAAU,KAAK,SAAS;AACjC,YAAM,OAAO,MAAM;AAAA,IACrB;AACA,SAAK,QAAQ,MAAM;AACnB,UAAM,KAAK,KAAK,MAAM;AAAA,EACxB;AACF;AAEO,MAAM,yBAAmD,qBAAqB;AAAA,EAC3E;AAAA,EACA;AAAA,EAER,UAAU,IAAI;AAAA,EAEd,YAAY,KAAkB,MAAmC,aAAgC;AAC/F,UAAM,KAAK,WAAW;AACtB,SAAK,OAAO;AACZ,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAAA,EACtC;AAAA,EAEA,MAAgB,MAAqB;AACnC,QAAI,UAAU;AACd,QAAI;AAEJ,UAAM,sBAAsB,IAAI,cAAc,kBAAkB,EAAE,OAAO;AACzE,UAAM,eAAe,oBAAoC;AACzD,UAAM,YAAY,UAAU,cAAc;AAC1C,UAAM,iBAAiB,IAAI,MAAM;AAGjC,UAAM,mBAAmB,IAAI,OAAa;AAE1C,UAAM,kBAAkB,YAAY;AAClC,UAAI,QAAS;AACb,gBAAU;AACV,0BAAoB,MAAM;AAE1B,YAAM,aAAa,MAAM;AAAA,IAC3B;AAEA,UAAM,kBAAkB,OAAO,OAAuB,IAAe,WAAwB;AAE3F,UAAI,OAAO,WAAW,QAAS;AAE/B,YAAM,iBAAiB,MAAM,qBAAqB,WAAW,KAAK;AAClE,UAAI,GAAG,eAAe,UAAU,MAAM;AACpC,aAAK,QAAQ,KAAK,uDAAuD;AACzE;AAAA,MACF;AACA,SAAG,KAAK,KAAK,UAAU,cAAc,CAAC;AAAA,IACxC;AAEA,UAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,UAAI,WAAW;AACb,aAAK,MAAM,IAAI,EAAE,WAAW,WAAW,OAAO,WAAW,MAAM,CAAC;AAChE,oBAAY;AAAA,MACd;AAAA,IACF;AAEA,UAAM,kBAAkB,OAAO,WAAwB;AACrD,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,OAAO,WAAW,QAAS;AAC/B,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,8BAAoB,MAAM;AAC1B;AAAA,QACF;AACA,4BAAoB,SAAS,IAAI;AAAA,MACnC;AAEA,UAAI,CAAC,SAAS;AACZ,4BAAoB,SAAS;AAAA,MAC/B;AAAA,IACF;AAEA,UAAM,2BAA2B,OAAO,IAAe,WAAwB;AAC7E,uBAAiB,MAAM,qBAAqB;AAC1C,YAAI,OAAO,WAAW,QAAS;AAE/B,cAAM;AAAA,UACJ;AAAA,YACE,MAAM;AAAA,YACN,YAAY,GAAG,QAAQ;AAAA,UACzB;AAAA,UACA;AAAA,UACA;AAAA,QACF;AACA,uBAAe,IAAI;AAAA,MACrB;AAEA,YAAM,gBAAgB,EAAE,MAAM,gBAAgB,GAAG,IAAI,MAAM;AAE3D,qBAAe,IAAI;AAAA,IACrB;AAIA,UAAM,uBAAuB,OAAO,IAAe,WAAwB;AACzE,YAAM,YAAY,CAAC,SAAiB;AAClC,YAAI;AACF,gBAAM,YAAY,KAAK,MAAM,KAAK,SAAS,CAAC;AAC5C,gBAAM,iBAAiB,qBAAqB,MAAM,SAAS;AAE3D,eAAK,aAAa,MAAM,cAAc,EAAE,MAAM,CAAC,UAAU;AACvD,iBAAK,QAAQ;AAAA,cACX,EAAE,MAAM;AAAA,cACR;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iCAAiC;AAAA,QACpE;AAAA,MACF;AAEA,YAAM,UAAU,CAAC,MAAa;AArcpC;AAscQ,aAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,aAAK,gBAAgB;AACrB,YAAI;AAEF,mBAAG,cAAH;AAAA,QACF,QAAQ;AAAA,QAER;AAEA,aAAK,IAAI,KAAK,OAAO,EAAE;AACvB,yBAAiB,OAAO,CAAC;AAAA,MAC3B;AAEA,YAAM,UAAU,MAAM;AAEpB,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,+BAA+B;AAClD,eAAK,gBAAgB;AAErB,eAAK,IAAI,KAAK,OAAO,EAAE;AACvB,2BAAiB;AAAA,YACf,IAAI,eAAe;AAAA,cACjB,SAAS;AAAA,cACT,SAAS,EAAE,UAAU;AAAA,YACvB,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AAEA,YAAM,UAAU,MAAM;AAne5B;AAoeQ,aAAK,gBAAgB;AACrB,YAAI;AAGF,mBAAG,cAAH;AAAA,QACF,QAAQ;AAAA,QAER;AACA,aAAK,IAAI,KAAK,OAAO,EAAE;AACvB,uBAAe,IAAI;AACnB,yBAAiB,QAAQ;AAAA,MAC3B;AAGA,SAAG,GAAG,WAAW,SAAS;AAC1B,SAAG,GAAG,SAAS,OAAO;AACtB,SAAG,GAAG,SAAS,OAAO;AACtB,aAAO,iBAAiB,SAAS,OAAO;AAExC,UAAI;AAEF,cAAM,iBAAiB;AAAA,MACzB,UAAE;AAEA,WAAG,IAAI,WAAW,SAAS;AAC3B,WAAG,IAAI,SAAS,OAAO;AACvB,WAAG,IAAI,SAAS,OAAO;AACvB,eAAO,oBAAoB,SAAS,OAAO;AAAA,MAC7C;AAAA,IACF;AAEA,UAAM,iBAAiB,OAAO,WAAwB;AACpD,UAAI,mBAAkC;AAEtC,YAAM,UAAU,IAAI,gBAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,YAAM,oBAAoB,aAAa,OAAO;AAC9C,YAAM,SAAS,kBAAkB,UAAU;AAE3C,UAAI;AACF,cAAM,eAAe,KAAK;AAE1B,eAAO,CAAC,KAAK,UAAU,CAAC,OAAO,SAAS;AACtC,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,OAAO,QAAS;AACpB,cAAI,OAAO,KAAM;AAEjB,gBAAM,cAAc,OAAO;AAC3B,kBAAQ,YAAY,MAAM;AAAA,YACxB,KAAK;AACH,iCAAmB,YAAY;AAC/B;AAAA,YACF,KAAK;AACH,oBAAM,aAAa,IAAI,UAAU,OAAO,KAAK,YAAY,OAAO,QAAQ,CAAC;AACzE,yBAAW,SAAS,QAAQ,MAAM,WAAW,MAAM,GAAG;AACpD,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA;AAAA,YACF,KAAK;AACH,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA,4BAAc,kBAAmB,IAAI;AACrC,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAC7C,oBAAM,gBAAgB;AACtB,+BAAiB,QAAQ;AACzB;AAAA,YACF,KAAK;AACH,oBAAM,gBAAgB;AACtB,+BAAiB,QAAQ;AACzB;AAAA,YACF,KAAK;AACH,mBAAK,QAAQ;AAAA,gBACX,EAAE,YAAY;AAAA,gBACd;AAAA,cACF;AACA,oBAAM,gBAAgB;AACtB,+BAAiB;AAAA,gBACf,IAAI,SAAS,+BAA+B,YAAY,OAAO,EAAE;AAAA,cACnE;AACA;AAAA,YACF;AACE,mBAAK,QAAQ,KAAK,yBAAyB,WAAW;AACtD;AAAA,UACJ;AAAA,QACF;AAAA,MACF,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,kBAAkB,OAAO;AAAA,QACjC,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,QACxF;AAAA,MACF;AAAA,IACF;AAEA,QAAI;AACF,YAAM,KAAK,IAAI,KAAK;AAAA,QAClB,OAAO,OAAkB;AACvB,cAAI;AAGF,kBAAM,gBAAgB,IAAI,gBAAgB;AAC1C,kBAAM,gBAAgB,MAAM,cAAc,MAAM,KAAK,gBAAgB,OAAO,MAAM;AAClF,iBAAK,gBAAgB,OAAO,iBAAiB,SAAS,eAAe,EAAE,MAAM,KAAK,CAAC;AAEnF,kBAAM,QAAQ;AAAA,cACZ,KAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,WAAW,eAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,gBAAgB,QAAQ;AAAA,gBAChC;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,cACA,KAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,WAAW,eAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,yBAAyB,IAAI,QAAQ;AAAA,gBAC7C;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,cACA,KAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,WAAW,eAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,qBAAqB,IAAI,QAAQ;AAAA,gBACzC;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,cACA,KAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,WAAW,eAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,eAAe,QAAQ;AAAA,gBAC/B;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,YACF;AAEA,gBAAI;AACF,oBAAM,QAAQ,IAAI,MAAM,IAAI,CAAC,MAAM,EAAE,MAAM,CAAC;AAAA,YAC9C,UAAE;AAEA,6BAAe,IAAI;AACnB,oBAAM,gBAAgB;AACtB,oBAAM,cAAc,OAAO,GAAI;AAC/B,mBAAK,gBAAgB,OAAO,oBAAoB,SAAS,aAAa;AAAA,YACxE;AAAA,UACF,SAAS,GAAG;AAEV,gBAAI,aAAa,SAAS,EAAE,SAAS,cAAc;AACjD;AAAA,YACF;AACA,kBAAM;AAAA,UACR;AAAA,QACF;AAAA,QACA;AAAA,UACE,SAAS,KAAK,YAAY;AAAA,QAC5B;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AAEV,UAAI,aAAa,SAAS,EAAE,SAAS,cAAc;AAEjD;AAAA,MACF;AACA,YAAM;AAAA,IACR,UAAE;AAEA,YAAM,gBAAgB;AAAA,IACxB;AAAA,EACF;AACF;","names":[]}
1
+ {"version":3,"sources":["../../src/inference/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { APIError, APIStatusError } from '../_exceptions.js';\nimport { AudioByteStream } from '../audio.js';\nimport { ConnectionPool } from '../connection_pool.js';\nimport { log } from '../log.js';\nimport { createStreamChannel } from '../stream/stream_channel.js';\nimport { basic as tokenizeBasic } from '../tokenize/index.js';\nimport type { ChunkedStream } from '../tts/index.js';\nimport { SynthesizeStream as BaseSynthesizeStream, TTS as BaseTTS } from '../tts/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { Event, Future, Task, cancelAndWait, combineSignals, shortuuid } from '../utils.js';\nimport {\n type TtsClientEvent,\n type TtsServerEvent,\n ttsClientEventSchema,\n ttsServerEventSchema,\n} from './api_protos.js';\nimport { type AnyString, connectWs, createAccessToken } from './utils.js';\n\nexport type CartesiaModels =\n | 'cartesia/sonic-3'\n | 'cartesia/sonic-2'\n | 'cartesia/sonic-turbo'\n | 'cartesia/sonic';\n\nexport type DeepgramTTSModels = 'deepgram/aura' | 'deepgram/aura-2';\n\nexport type ElevenlabsModels =\n | 'elevenlabs/eleven_flash_v2'\n | 'elevenlabs/eleven_flash_v2_5'\n | 'elevenlabs/eleven_turbo_v2'\n | 'elevenlabs/eleven_turbo_v2_5'\n | 'elevenlabs/eleven_multilingual_v2';\n\nexport type InworldModels =\n | 'inworld/inworld-tts-1.5-max'\n | 'inworld/inworld-tts-1.5-mini'\n | 'inworld/inworld-tts-1-max'\n | 'inworld/inworld-tts-1';\n\nexport type RimeModels = 'rime/arcana' | 'rime/mistv2';\n\nexport interface CartesiaOptions {\n /** Maximum duration of audio in seconds. */\n duration?: number;\n /** Speech speed. Default: not specified. */\n speed?: 'slow' | 'normal' | 'fast';\n}\n\nexport interface ElevenlabsOptions {\n /** Inactivity timeout in seconds. Default: 60. */\n inactivity_timeout?: number;\n /** Text normalization mode. Default: \"auto\". */\n apply_text_normalization?: 'auto' | 'off' | 'on';\n}\n\nexport interface DeepgramTTSOptions {}\n\nexport interface RimeOptions {}\n\nexport interface InworldOptions {\n /** Controls how fast the voice speaks. 1.0 is normal speed, 0.5 is half, 1.5 is 1.5x. Default: 1.0. */\n speaking_rate?: number;\n /** Controls randomness in the output. Recommended between 0.6 and 1.1. Default: 1.1. */\n temperature?: number;\n /** Controls text normalization. \"ON\" expands numbers, dates, abbreviations. \"OFF\" reads text as written. Default: \"ON\". */\n text_normalization?: 'ON' | 'OFF';\n}\n\ntype _TTSModels =\n | CartesiaModels\n | DeepgramTTSModels\n | ElevenlabsModels\n | RimeModels\n | InworldModels;\n\nexport type TTSModels =\n | CartesiaModels\n | DeepgramTTSModels\n | ElevenlabsModels\n | RimeModels\n | InworldModels\n | AnyString;\n\nexport type ModelWithVoice = `${_TTSModels}:${string}` | TTSModels;\n\nexport type TTSOptions<TModel extends TTSModels> = TModel extends CartesiaModels\n ? CartesiaOptions\n : TModel extends DeepgramTTSModels\n ? DeepgramTTSOptions\n : TModel extends ElevenlabsModels\n ? ElevenlabsOptions\n : TModel extends RimeModels\n ? RimeOptions\n : TModel extends InworldModels\n ? InworldOptions\n : Record<string, unknown>;\n\n/** Parse a model string into [model, voice]. Voice is undefined if not specified. */\nexport function parseTTSModelString(model: string): [string, string | undefined] {\n const idx = model.lastIndexOf(':');\n if (idx !== -1) {\n return [model.slice(0, idx), model.slice(idx + 1)];\n }\n return [model, undefined];\n}\n\n/** A fallback model with optional extra configuration. Extra fields are passed through to the provider. */\nexport interface TTSFallbackModel {\n /** Model name (e.g. \"cartesia/sonic\", \"elevenlabs/eleven_flash_v2\", \"rime/arcana\"). */\n model: string;\n /** Voice to use for the model. */\n voice: string;\n /** Extra configuration for the model. */\n extraKwargs?: Record<string, unknown>;\n}\n\nexport type TTSFallbackModelType = TTSFallbackModel | string;\n\n/** Normalize a single or list of FallbackModelType into TTSFallbackModel[]. */\nexport function normalizeTTSFallback(\n fallback: TTSFallbackModelType | TTSFallbackModelType[],\n): TTSFallbackModel[] {\n const makeFallback = (model: TTSFallbackModelType): TTSFallbackModel => {\n if (typeof model === 'string') {\n const [name, voice] = parseTTSModelString(model);\n return { model: name, voice: voice ?? '' };\n }\n return model;\n };\n\n if (Array.isArray(fallback)) {\n return fallback.map(makeFallback);\n }\n return [makeFallback(fallback)];\n}\n\ntype TTSEncoding = 'pcm_s16le';\n\nconst DEFAULT_ENCODING: TTSEncoding = 'pcm_s16le';\nconst DEFAULT_SAMPLE_RATE = 16000;\nconst DEFAULT_BASE_URL = 'https://agent-gateway.livekit.cloud/v1';\nconst NUM_CHANNELS = 1;\nconst DEFAULT_LANGUAGE = 'en';\n\nexport interface InferenceTTSOptions<TModel extends TTSModels> {\n model?: TModel;\n voice?: string;\n language?: string;\n encoding: TTSEncoding;\n sampleRate: number;\n baseURL: string;\n apiKey: string;\n apiSecret: string;\n modelOptions: TTSOptions<TModel>;\n fallback?: TTSFallbackModel[];\n connOptions?: APIConnectOptions;\n}\n\n/**\n * Livekit Cloud Inference TTS\n */\nexport class TTS<TModel extends TTSModels> extends BaseTTS {\n private opts: InferenceTTSOptions<TModel>;\n private streams: Set<SynthesizeStream<TModel>> = new Set();\n pool: ConnectionPool<WebSocket>;\n\n #logger = log();\n\n constructor(opts: {\n model: TModel;\n voice?: string;\n language?: string;\n baseURL?: string;\n encoding?: TTSEncoding;\n sampleRate?: number;\n apiKey?: string;\n apiSecret?: string;\n modelOptions?: TTSOptions<TModel>;\n fallback?: TTSFallbackModelType | TTSFallbackModelType[];\n connOptions?: APIConnectOptions;\n }) {\n const sampleRate = opts?.sampleRate ?? DEFAULT_SAMPLE_RATE;\n super(sampleRate, 1, { streaming: true });\n\n const {\n model,\n voice,\n language = DEFAULT_LANGUAGE,\n baseURL,\n encoding = DEFAULT_ENCODING,\n apiKey,\n apiSecret,\n modelOptions = {} as TTSOptions<TModel>,\n fallback,\n connOptions,\n } = opts || {};\n\n const lkBaseURL = baseURL || process.env.LIVEKIT_INFERENCE_URL || DEFAULT_BASE_URL;\n const lkApiKey = apiKey || process.env.LIVEKIT_INFERENCE_API_KEY || process.env.LIVEKIT_API_KEY;\n if (!lkApiKey) {\n throw new Error('apiKey is required: pass apiKey or set LIVEKIT_API_KEY');\n }\n\n const lkApiSecret =\n apiSecret || process.env.LIVEKIT_INFERENCE_API_SECRET || process.env.LIVEKIT_API_SECRET;\n if (!lkApiSecret) {\n throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');\n }\n\n // read voice id from the model if provided: \"provider/model:voice_id\"\n let nextModel = model;\n let nextVoice = voice;\n if (typeof nextModel === 'string') {\n const idx = nextModel.lastIndexOf(':');\n if (idx !== -1) {\n const voiceFromModel = nextModel.slice(idx + 1);\n if (nextVoice && nextVoice !== voiceFromModel) {\n this.#logger.warn(\n '`voice` is provided via both argument and model, using the one from the argument',\n { voice: nextVoice, model: nextModel },\n );\n } else {\n nextVoice = voiceFromModel;\n }\n nextModel = nextModel.slice(0, idx) as TModel;\n }\n }\n\n const normalizedFallback = fallback ? normalizeTTSFallback(fallback) : undefined;\n\n this.opts = {\n model: nextModel,\n voice: nextVoice,\n language,\n encoding,\n sampleRate,\n baseURL: lkBaseURL,\n apiKey: lkApiKey,\n apiSecret: lkApiSecret,\n modelOptions,\n fallback: normalizedFallback,\n connOptions: connOptions ?? DEFAULT_API_CONNECT_OPTIONS,\n };\n\n // Initialize connection pool\n this.pool = new ConnectionPool<WebSocket>({\n connectCb: (timeout) => this.connectWs(timeout),\n closeCb: (ws) => this.closeWs(ws),\n maxSessionDuration: 300_000,\n markRefreshedOnGet: true,\n connectTimeout: 10_000, // 10 seconds default\n });\n }\n\n get label() {\n return 'inference.TTS';\n }\n\n static fromModelString(modelString: string): TTS<AnyString> {\n const [model, voice] = parseTTSModelString(modelString);\n return new TTS({ model, voice: voice || undefined });\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n for (const stream of this.streams) {\n stream.updateOptions(opts);\n }\n }\n\n synthesize(_: string): ChunkedStream {\n throw new Error('ChunkedStream is not implemented');\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream<TModel> {\n const { connOptions = this.opts.connOptions ?? DEFAULT_API_CONNECT_OPTIONS } = options || {};\n const stream = new SynthesizeStream(this, { ...this.opts }, connOptions);\n this.streams.add(stream);\n return stream;\n }\n\n async connectWs(timeout: number): Promise<WebSocket> {\n let baseURL = this.opts.baseURL;\n if (baseURL.startsWith('http://') || baseURL.startsWith('https://')) {\n baseURL = baseURL.replace('http', 'ws');\n }\n\n const token = await createAccessToken(this.opts.apiKey, this.opts.apiSecret);\n const url = `${baseURL}/tts`;\n const headers = { Authorization: `Bearer ${token}` } as Record<string, string>;\n\n const params = {\n type: 'session.create',\n sample_rate: String(this.opts.sampleRate),\n encoding: this.opts.encoding,\n extra: this.opts.modelOptions,\n } as Record<string, unknown>;\n\n if (this.opts.voice) (params as Record<string, unknown>).voice = this.opts.voice;\n if (this.opts.model) (params as Record<string, unknown>).model = this.opts.model;\n if (this.opts.language) (params as Record<string, unknown>).language = this.opts.language;\n\n if (this.opts.fallback?.length) {\n params.fallback = {\n models: this.opts.fallback.map((m) => ({\n model: m.model,\n voice: m.voice,\n extra: m.extraKwargs ?? {},\n })),\n };\n }\n\n if (this.opts.connOptions) {\n params.connection = {\n timeout: this.opts.connOptions.timeoutMs / 1000,\n retries: this.opts.connOptions.maxRetry,\n };\n }\n\n this.#logger.debug({ url }, 'inference.TTS creating new websocket connection (pool miss)');\n const socket = await connectWs(url, headers, timeout);\n socket.send(JSON.stringify(params));\n return socket;\n }\n\n async closeWs(ws: WebSocket) {\n await ws.close();\n }\n\n prewarm(): void {\n this.pool.prewarm();\n }\n\n async close() {\n for (const stream of this.streams) {\n await stream.close();\n }\n this.streams.clear();\n await this.pool.close();\n }\n}\n\nexport class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeStream {\n private opts: InferenceTTSOptions<TModel>;\n private tts: TTS<TModel>;\n\n #logger = log();\n\n constructor(tts: TTS<TModel>, opts: InferenceTTSOptions<TModel>, connOptions: APIConnectOptions) {\n super(tts, connOptions);\n this.opts = opts;\n this.tts = tts;\n }\n\n get label() {\n return 'inference.SynthesizeStream';\n }\n\n updateOptions(opts: Partial<Pick<InferenceTTSOptions<TModel>, 'model' | 'voice' | 'language'>>) {\n this.opts = { ...this.opts, ...opts };\n }\n\n protected async run(): Promise<void> {\n let closing = false;\n let lastFrame: AudioFrame | undefined;\n\n const sendTokenizerStream = new tokenizeBasic.SentenceTokenizer().stream();\n const eventChannel = createStreamChannel<TtsServerEvent>();\n const requestId = shortuuid('tts_request_');\n const inputSentEvent = new Event();\n\n // Signal for protocol-driven completion (when 'done' message is received)\n const completionFuture = new Future<void>();\n\n const resourceCleanup = async () => {\n if (closing) return;\n closing = true;\n sendTokenizerStream.close();\n // close() returns a promise; don't leak it\n await eventChannel.close();\n };\n\n const sendClientEvent = async (event: TtsClientEvent, ws: WebSocket, signal: AbortSignal) => {\n // Don't send events to a closed WebSocket or aborted controller\n if (signal.aborted || closing) return;\n\n const validatedEvent = await ttsClientEventSchema.parseAsync(event);\n if (ws.readyState !== WebSocket.OPEN) {\n this.#logger.warn('Trying to send client TTS event to a closed WebSocket');\n return;\n }\n ws.send(JSON.stringify(validatedEvent));\n };\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const createInputTask = async (signal: AbortSignal) => {\n for await (const data of this.input) {\n if (signal.aborted || closing) break;\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n sendTokenizerStream.flush();\n continue;\n }\n sendTokenizerStream.pushText(data);\n }\n // Only call endInput if the stream hasn't been closed by cleanup\n if (!closing) {\n sendTokenizerStream.endInput();\n }\n };\n\n const createSentenceStreamTask = async (ws: WebSocket, signal: AbortSignal) => {\n for await (const ev of sendTokenizerStream) {\n if (signal.aborted || closing) break;\n\n await sendClientEvent(\n {\n type: 'input_transcript',\n transcript: ev.token + ' ',\n },\n ws,\n signal,\n );\n inputSentEvent.set();\n }\n\n await sendClientEvent({ type: 'session.flush' }, ws, signal);\n // needed in case empty input is sent\n inputSentEvent.set();\n };\n\n // Handles WebSocket message routing and error handling\n // Completes based on protocol messages, NOT on ws.close()\n const createWsListenerTask = async (ws: WebSocket, signal: AbortSignal) => {\n const onMessage = (data: Buffer) => {\n try {\n const eventJson = JSON.parse(data.toString()) as Record<string, unknown>;\n const validatedEvent = ttsServerEventSchema.parse(eventJson);\n // writer.write returns a promise; avoid unhandled rejections if stream is closed\n void eventChannel.write(validatedEvent).catch((error) => {\n this.#logger.debug(\n { error },\n 'Failed writing TTS event to stream channel (likely closed)',\n );\n });\n } catch (e) {\n this.#logger.error({ error: e }, 'Error parsing WebSocket message');\n }\n };\n\n const onError = (e: Error) => {\n this.#logger.error({ error: e }, 'WebSocket error');\n void resourceCleanup();\n try {\n // If the ws is misbehaving, hard-stop it immediately to avoid buffering.\n ws.terminate?.();\n } catch {\n // ignore\n }\n // Ensure this ws is not reused\n this.tts.pool.remove(ws);\n completionFuture.reject(e);\n };\n\n const onClose = () => {\n // WebSocket closed unexpectedly (not by us)\n if (!closing) {\n this.#logger.error('WebSocket closed unexpectedly');\n void resourceCleanup();\n // Ensure this ws is not reused\n this.tts.pool.remove(ws);\n completionFuture.reject(\n new APIStatusError({\n message: 'Gateway connection closed unexpectedly',\n options: { requestId },\n }),\n );\n }\n };\n\n const onAbort = () => {\n void resourceCleanup();\n try {\n // On interruption/abort, close the websocket immediately so the server stops streaming\n // and the ws library doesn't buffer unread frames in memory.\n ws.terminate?.();\n } catch {\n // ignore\n }\n this.tts.pool.remove(ws);\n inputSentEvent.set();\n completionFuture.resolve();\n };\n\n // Attach listeners\n ws.on('message', onMessage);\n ws.on('error', onError);\n ws.on('close', onClose);\n signal.addEventListener('abort', onAbort);\n\n try {\n // Wait for protocol-driven completion or error\n await completionFuture.await;\n } finally {\n // IMPORTANT: Remove listeners so connection can be reused\n ws.off('message', onMessage);\n ws.off('error', onError);\n ws.off('close', onClose);\n signal.removeEventListener('abort', onAbort);\n }\n };\n\n const createRecvTask = async (signal: AbortSignal) => {\n let currentSessionId: string | null = null;\n\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const serverEventStream = eventChannel.stream();\n const reader = serverEventStream.getReader();\n\n try {\n await inputSentEvent.wait();\n\n while (!this.closed && !signal.aborted) {\n const result = await reader.read();\n if (signal.aborted) return;\n if (result.done) return;\n\n const serverEvent = result.value;\n switch (serverEvent.type) {\n case 'session.created':\n currentSessionId = serverEvent.session_id;\n break;\n case 'output_audio':\n const base64Data = new Int8Array(Buffer.from(serverEvent.audio, 'base64'));\n for (const frame of bstream.write(base64Data.buffer)) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n break;\n case 'done':\n for (const frame of bstream.flush()) {\n sendLastFrame(currentSessionId!, false);\n lastFrame = frame;\n }\n sendLastFrame(currentSessionId!, true);\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n await resourceCleanup();\n completionFuture.resolve();\n return;\n case 'session.closed':\n await resourceCleanup();\n completionFuture.resolve();\n return;\n case 'error':\n this.#logger.error(\n { serverEvent },\n 'Received error message from LiveKit TTS WebSocket',\n );\n await resourceCleanup();\n completionFuture.reject(\n new APIError(`LiveKit TTS returned error: ${serverEvent.message}`),\n );\n return;\n default:\n this.#logger.warn('Unexpected message %s', serverEvent);\n break;\n }\n }\n } finally {\n reader.releaseLock();\n try {\n await serverEventStream.cancel();\n } catch (e) {\n this.#logger.debug('Error cancelling serverEventStream (may already be cancelled):', e);\n }\n }\n };\n\n try {\n await this.tts.pool.withConnection(\n async (ws: WebSocket) => {\n try {\n // IMPORTANT: don't cancel the stream's controller on normal completion,\n // otherwise the pool will remove+close the ws and every run becomes a pool miss.\n const runController = new AbortController();\n const onStreamAbort = () => runController.abort(this.abortController.signal.reason);\n this.abortController.signal.addEventListener('abort', onStreamAbort, { once: true });\n\n const tasks = [\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createInputTask(combined);\n },\n undefined,\n 'inference-tts-input',\n ),\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createSentenceStreamTask(ws, combined);\n },\n undefined,\n 'inference-tts-sentence',\n ),\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createWsListenerTask(ws, combined);\n },\n undefined,\n 'inference-tts-ws-listener',\n ),\n Task.from(\n async (controller) => {\n const combined = combineSignals(runController.signal, controller.signal);\n await createRecvTask(combined);\n },\n undefined,\n 'inference-tts-recv',\n ),\n ];\n\n try {\n await Promise.all(tasks.map((t) => t.result));\n } finally {\n // Mirror python finally: unblock recv and cancel all tasks.\n inputSentEvent.set();\n await resourceCleanup();\n await cancelAndWait(tasks, 5000);\n this.abortController.signal.removeEventListener('abort', onStreamAbort);\n }\n } catch (e) {\n // If aborted, don't throw - let cleanup handle it\n if (e instanceof Error && e.name === 'AbortError') {\n return;\n }\n throw e;\n }\n },\n {\n timeout: this.connOptions.timeoutMs,\n },\n );\n } catch (e) {\n // Handle connection errors\n if (e instanceof Error && e.name === 'AbortError') {\n // Abort is expected during normal shutdown\n return;\n }\n throw e;\n } finally {\n // Ensure cleanup always runs (and don't leak the promise)\n await resourceCleanup();\n }\n }\n}\n"],"mappings":"AAIA,SAAS,iBAAiB;AAC1B,SAAS,UAAU,sBAAsB;AACzC,SAAS,uBAAuB;AAChC,SAAS,sBAAsB;AAC/B,SAAS,WAAW;AACpB,SAAS,2BAA2B;AACpC,SAAS,SAAS,qBAAqB;AAEvC,SAAS,oBAAoB,sBAAsB,OAAO,eAAe;AACzE,SAAiC,mCAAmC;AACpE,SAAS,OAAO,QAAQ,MAAM,eAAe,gBAAgB,iBAAiB;AAC9E;AAAA,EAGE;AAAA,EACA;AAAA,OACK;AACP,SAAyB,WAAW,yBAAyB;AAkFtD,SAAS,oBAAoB,OAA6C;AAC/E,QAAM,MAAM,MAAM,YAAY,GAAG;AACjC,MAAI,QAAQ,IAAI;AACd,WAAO,CAAC,MAAM,MAAM,GAAG,GAAG,GAAG,MAAM,MAAM,MAAM,CAAC,CAAC;AAAA,EACnD;AACA,SAAO,CAAC,OAAO,MAAS;AAC1B;AAeO,SAAS,qBACd,UACoB;AACpB,QAAM,eAAe,CAAC,UAAkD;AACtE,QAAI,OAAO,UAAU,UAAU;AAC7B,YAAM,CAAC,MAAM,KAAK,IAAI,oBAAoB,KAAK;AAC/C,aAAO,EAAE,OAAO,MAAM,OAAO,SAAS,GAAG;AAAA,IAC3C;AACA,WAAO;AAAA,EACT;AAEA,MAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,WAAO,SAAS,IAAI,YAAY;AAAA,EAClC;AACA,SAAO,CAAC,aAAa,QAAQ,CAAC;AAChC;AAIA,MAAM,mBAAgC;AACtC,MAAM,sBAAsB;AAC5B,MAAM,mBAAmB;AACzB,MAAM,eAAe;AACrB,MAAM,mBAAmB;AAmBlB,MAAM,YAAsC,QAAQ;AAAA,EACjD;AAAA,EACA,UAAyC,oBAAI,IAAI;AAAA,EACzD;AAAA,EAEA,UAAU,IAAI;AAAA,EAEd,YAAY,MAYT;AACD,UAAM,cAAa,6BAAM,eAAc;AACvC,UAAM,YAAY,GAAG,EAAE,WAAW,KAAK,CAAC;AAExC,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA;AAAA,MACA,eAAe,CAAC;AAAA,MAChB;AAAA,MACA;AAAA,IACF,IAAI,QAAQ,CAAC;AAEb,UAAM,YAAY,WAAW,QAAQ,IAAI,yBAAyB;AAClE,UAAM,WAAW,UAAU,QAAQ,IAAI,6BAA6B,QAAQ,IAAI;AAChF,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,UAAM,cACJ,aAAa,QAAQ,IAAI,gCAAgC,QAAQ,IAAI;AACvE,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AAGA,QAAI,YAAY;AAChB,QAAI,YAAY;AAChB,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,MAAM,UAAU,YAAY,GAAG;AACrC,UAAI,QAAQ,IAAI;AACd,cAAM,iBAAiB,UAAU,MAAM,MAAM,CAAC;AAC9C,YAAI,aAAa,cAAc,gBAAgB;AAC7C,eAAK,QAAQ;AAAA,YACX;AAAA,YACA,EAAE,OAAO,WAAW,OAAO,UAAU;AAAA,UACvC;AAAA,QACF,OAAO;AACL,sBAAY;AAAA,QACd;AACA,oBAAY,UAAU,MAAM,GAAG,GAAG;AAAA,MACpC;AAAA,IACF;AAEA,UAAM,qBAAqB,WAAW,qBAAqB,QAAQ,IAAI;AAEvE,SAAK,OAAO;AAAA,MACV,OAAO;AAAA,MACP,OAAO;AAAA,MACP;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,WAAW;AAAA,MACX;AAAA,MACA,UAAU;AAAA,MACV,aAAa,eAAe;AAAA,IAC9B;AAGA,SAAK,OAAO,IAAI,eAA0B;AAAA,MACxC,WAAW,CAAC,YAAY,KAAK,UAAU,OAAO;AAAA,MAC9C,SAAS,CAAC,OAAO,KAAK,QAAQ,EAAE;AAAA,MAChC,oBAAoB;AAAA,MACpB,oBAAoB;AAAA,MACpB,gBAAgB;AAAA;AAAA,IAClB,CAAC;AAAA,EACH;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,gBAAgB,aAAqC;AAC1D,UAAM,CAAC,OAAO,KAAK,IAAI,oBAAoB,WAAW;AACtD,WAAO,IAAI,IAAI,EAAE,OAAO,OAAO,SAAS,OAAU,CAAC;AAAA,EACrD;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AACpC,eAAW,UAAU,KAAK,SAAS;AACjC,aAAO,cAAc,IAAI;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,WAAW,GAA0B;AACnC,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AAAA,EAEA,OAAO,SAAyE;AAC9E,UAAM,EAAE,cAAc,KAAK,KAAK,eAAe,4BAA4B,IAAI,WAAW,CAAC;AAC3F,UAAM,SAAS,IAAI,iBAAiB,MAAM,EAAE,GAAG,KAAK,KAAK,GAAG,WAAW;AACvE,SAAK,QAAQ,IAAI,MAAM;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,UAAU,SAAqC;AA9RvD;AA+RI,QAAI,UAAU,KAAK,KAAK;AACxB,QAAI,QAAQ,WAAW,SAAS,KAAK,QAAQ,WAAW,UAAU,GAAG;AACnE,gBAAU,QAAQ,QAAQ,QAAQ,IAAI;AAAA,IACxC;AAEA,UAAM,QAAQ,MAAM,kBAAkB,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS;AAC3E,UAAM,MAAM,GAAG,OAAO;AACtB,UAAM,UAAU,EAAE,eAAe,UAAU,KAAK,GAAG;AAEnD,UAAM,SAAS;AAAA,MACb,MAAM;AAAA,MACN,aAAa,OAAO,KAAK,KAAK,UAAU;AAAA,MACxC,UAAU,KAAK,KAAK;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,IACnB;AAEA,QAAI,KAAK,KAAK,MAAO,CAAC,OAAmC,QAAQ,KAAK,KAAK;AAC3E,QAAI,KAAK,KAAK,MAAO,CAAC,OAAmC,QAAQ,KAAK,KAAK;AAC3E,QAAI,KAAK,KAAK,SAAU,CAAC,OAAmC,WAAW,KAAK,KAAK;AAEjF,SAAI,UAAK,KAAK,aAAV,mBAAoB,QAAQ;AAC9B,aAAO,WAAW;AAAA,QAChB,QAAQ,KAAK,KAAK,SAAS,IAAI,CAAC,OAAO;AAAA,UACrC,OAAO,EAAE;AAAA,UACT,OAAO,EAAE;AAAA,UACT,OAAO,EAAE,eAAe,CAAC;AAAA,QAC3B,EAAE;AAAA,MACJ;AAAA,IACF;AAEA,QAAI,KAAK,KAAK,aAAa;AACzB,aAAO,aAAa;AAAA,QAClB,SAAS,KAAK,KAAK,YAAY,YAAY;AAAA,QAC3C,SAAS,KAAK,KAAK,YAAY;AAAA,MACjC;AAAA,IACF;AAEA,SAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6DAA6D;AACzF,UAAM,SAAS,MAAM,UAAU,KAAK,SAAS,OAAO;AACpD,WAAO,KAAK,KAAK,UAAU,MAAM,CAAC;AAClC,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,QAAQ,IAAe;AAC3B,UAAM,GAAG,MAAM;AAAA,EACjB;AAAA,EAEA,UAAgB;AACd,SAAK,KAAK,QAAQ;AAAA,EACpB;AAAA,EAEA,MAAM,QAAQ;AACZ,eAAW,UAAU,KAAK,SAAS;AACjC,YAAM,OAAO,MAAM;AAAA,IACrB;AACA,SAAK,QAAQ,MAAM;AACnB,UAAM,KAAK,KAAK,MAAM;AAAA,EACxB;AACF;AAEO,MAAM,yBAAmD,qBAAqB;AAAA,EAC3E;AAAA,EACA;AAAA,EAER,UAAU,IAAI;AAAA,EAEd,YAAY,KAAkB,MAAmC,aAAgC;AAC/F,UAAM,KAAK,WAAW;AACtB,SAAK,OAAO;AACZ,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,IAAI,QAAQ;AACV,WAAO;AAAA,EACT;AAAA,EAEA,cAAc,MAAkF;AAC9F,SAAK,OAAO,EAAE,GAAG,KAAK,MAAM,GAAG,KAAK;AAAA,EACtC;AAAA,EAEA,MAAgB,MAAqB;AACnC,QAAI,UAAU;AACd,QAAI;AAEJ,UAAM,sBAAsB,IAAI,cAAc,kBAAkB,EAAE,OAAO;AACzE,UAAM,eAAe,oBAAoC;AACzD,UAAM,YAAY,UAAU,cAAc;AAC1C,UAAM,iBAAiB,IAAI,MAAM;AAGjC,UAAM,mBAAmB,IAAI,OAAa;AAE1C,UAAM,kBAAkB,YAAY;AAClC,UAAI,QAAS;AACb,gBAAU;AACV,0BAAoB,MAAM;AAE1B,YAAM,aAAa,MAAM;AAAA,IAC3B;AAEA,UAAM,kBAAkB,OAAO,OAAuB,IAAe,WAAwB;AAE3F,UAAI,OAAO,WAAW,QAAS;AAE/B,YAAM,iBAAiB,MAAM,qBAAqB,WAAW,KAAK;AAClE,UAAI,GAAG,eAAe,UAAU,MAAM;AACpC,aAAK,QAAQ,KAAK,uDAAuD;AACzE;AAAA,MACF;AACA,SAAG,KAAK,KAAK,UAAU,cAAc,CAAC;AAAA,IACxC;AAEA,UAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,UAAI,WAAW;AACb,aAAK,MAAM,IAAI,EAAE,WAAW,WAAW,OAAO,WAAW,MAAM,CAAC;AAChE,oBAAY;AAAA,MACd;AAAA,IACF;AAEA,UAAM,kBAAkB,OAAO,WAAwB;AACrD,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,OAAO,WAAW,QAAS;AAC/B,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,8BAAoB,MAAM;AAC1B;AAAA,QACF;AACA,4BAAoB,SAAS,IAAI;AAAA,MACnC;AAEA,UAAI,CAAC,SAAS;AACZ,4BAAoB,SAAS;AAAA,MAC/B;AAAA,IACF;AAEA,UAAM,2BAA2B,OAAO,IAAe,WAAwB;AAC7E,uBAAiB,MAAM,qBAAqB;AAC1C,YAAI,OAAO,WAAW,QAAS;AAE/B,cAAM;AAAA,UACJ;AAAA,YACE,MAAM;AAAA,YACN,YAAY,GAAG,QAAQ;AAAA,UACzB;AAAA,UACA;AAAA,UACA;AAAA,QACF;AACA,uBAAe,IAAI;AAAA,MACrB;AAEA,YAAM,gBAAgB,EAAE,MAAM,gBAAgB,GAAG,IAAI,MAAM;AAE3D,qBAAe,IAAI;AAAA,IACrB;AAIA,UAAM,uBAAuB,OAAO,IAAe,WAAwB;AACzE,YAAM,YAAY,CAAC,SAAiB;AAClC,YAAI;AACF,gBAAM,YAAY,KAAK,MAAM,KAAK,SAAS,CAAC;AAC5C,gBAAM,iBAAiB,qBAAqB,MAAM,SAAS;AAE3D,eAAK,aAAa,MAAM,cAAc,EAAE,MAAM,CAAC,UAAU;AACvD,iBAAK,QAAQ;AAAA,cACX,EAAE,MAAM;AAAA,cACR;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iCAAiC;AAAA,QACpE;AAAA,MACF;AAEA,YAAM,UAAU,CAAC,MAAa;AA5cpC;AA6cQ,aAAK,QAAQ,MAAM,EAAE,OAAO,EAAE,GAAG,iBAAiB;AAClD,aAAK,gBAAgB;AACrB,YAAI;AAEF,mBAAG,cAAH;AAAA,QACF,QAAQ;AAAA,QAER;AAEA,aAAK,IAAI,KAAK,OAAO,EAAE;AACvB,yBAAiB,OAAO,CAAC;AAAA,MAC3B;AAEA,YAAM,UAAU,MAAM;AAEpB,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,+BAA+B;AAClD,eAAK,gBAAgB;AAErB,eAAK,IAAI,KAAK,OAAO,EAAE;AACvB,2BAAiB;AAAA,YACf,IAAI,eAAe;AAAA,cACjB,SAAS;AAAA,cACT,SAAS,EAAE,UAAU;AAAA,YACvB,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AAEA,YAAM,UAAU,MAAM;AA1e5B;AA2eQ,aAAK,gBAAgB;AACrB,YAAI;AAGF,mBAAG,cAAH;AAAA,QACF,QAAQ;AAAA,QAER;AACA,aAAK,IAAI,KAAK,OAAO,EAAE;AACvB,uBAAe,IAAI;AACnB,yBAAiB,QAAQ;AAAA,MAC3B;AAGA,SAAG,GAAG,WAAW,SAAS;AAC1B,SAAG,GAAG,SAAS,OAAO;AACtB,SAAG,GAAG,SAAS,OAAO;AACtB,aAAO,iBAAiB,SAAS,OAAO;AAExC,UAAI;AAEF,cAAM,iBAAiB;AAAA,MACzB,UAAE;AAEA,WAAG,IAAI,WAAW,SAAS;AAC3B,WAAG,IAAI,SAAS,OAAO;AACvB,WAAG,IAAI,SAAS,OAAO;AACvB,eAAO,oBAAoB,SAAS,OAAO;AAAA,MAC7C;AAAA,IACF;AAEA,UAAM,iBAAiB,OAAO,WAAwB;AACpD,UAAI,mBAAkC;AAEtC,YAAM,UAAU,IAAI,gBAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,YAAM,oBAAoB,aAAa,OAAO;AAC9C,YAAM,SAAS,kBAAkB,UAAU;AAE3C,UAAI;AACF,cAAM,eAAe,KAAK;AAE1B,eAAO,CAAC,KAAK,UAAU,CAAC,OAAO,SAAS;AACtC,gBAAM,SAAS,MAAM,OAAO,KAAK;AACjC,cAAI,OAAO,QAAS;AACpB,cAAI,OAAO,KAAM;AAEjB,gBAAM,cAAc,OAAO;AAC3B,kBAAQ,YAAY,MAAM;AAAA,YACxB,KAAK;AACH,iCAAmB,YAAY;AAC/B;AAAA,YACF,KAAK;AACH,oBAAM,aAAa,IAAI,UAAU,OAAO,KAAK,YAAY,OAAO,QAAQ,CAAC;AACzE,yBAAW,SAAS,QAAQ,MAAM,WAAW,MAAM,GAAG;AACpD,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA;AAAA,YACF,KAAK;AACH,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,kBAAmB,KAAK;AACtC,4BAAY;AAAA,cACd;AACA,4BAAc,kBAAmB,IAAI;AACrC,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAC7C,oBAAM,gBAAgB;AACtB,+BAAiB,QAAQ;AACzB;AAAA,YACF,KAAK;AACH,oBAAM,gBAAgB;AACtB,+BAAiB,QAAQ;AACzB;AAAA,YACF,KAAK;AACH,mBAAK,QAAQ;AAAA,gBACX,EAAE,YAAY;AAAA,gBACd;AAAA,cACF;AACA,oBAAM,gBAAgB;AACtB,+BAAiB;AAAA,gBACf,IAAI,SAAS,+BAA+B,YAAY,OAAO,EAAE;AAAA,cACnE;AACA;AAAA,YACF;AACE,mBAAK,QAAQ,KAAK,yBAAyB,WAAW;AACtD;AAAA,UACJ;AAAA,QACF;AAAA,MACF,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,kBAAkB,OAAO;AAAA,QACjC,SAAS,GAAG;AACV,eAAK,QAAQ,MAAM,kEAAkE,CAAC;AAAA,QACxF;AAAA,MACF;AAAA,IACF;AAEA,QAAI;AACF,YAAM,KAAK,IAAI,KAAK;AAAA,QAClB,OAAO,OAAkB;AACvB,cAAI;AAGF,kBAAM,gBAAgB,IAAI,gBAAgB;AAC1C,kBAAM,gBAAgB,MAAM,cAAc,MAAM,KAAK,gBAAgB,OAAO,MAAM;AAClF,iBAAK,gBAAgB,OAAO,iBAAiB,SAAS,eAAe,EAAE,MAAM,KAAK,CAAC;AAEnF,kBAAM,QAAQ;AAAA,cACZ,KAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,WAAW,eAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,gBAAgB,QAAQ;AAAA,gBAChC;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,cACA,KAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,WAAW,eAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,yBAAyB,IAAI,QAAQ;AAAA,gBAC7C;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,cACA,KAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,WAAW,eAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,qBAAqB,IAAI,QAAQ;AAAA,gBACzC;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,cACA,KAAK;AAAA,gBACH,OAAO,eAAe;AACpB,wBAAM,WAAW,eAAe,cAAc,QAAQ,WAAW,MAAM;AACvE,wBAAM,eAAe,QAAQ;AAAA,gBAC/B;AAAA,gBACA;AAAA,gBACA;AAAA,cACF;AAAA,YACF;AAEA,gBAAI;AACF,oBAAM,QAAQ,IAAI,MAAM,IAAI,CAAC,MAAM,EAAE,MAAM,CAAC;AAAA,YAC9C,UAAE;AAEA,6BAAe,IAAI;AACnB,oBAAM,gBAAgB;AACtB,oBAAM,cAAc,OAAO,GAAI;AAC/B,mBAAK,gBAAgB,OAAO,oBAAoB,SAAS,aAAa;AAAA,YACxE;AAAA,UACF,SAAS,GAAG;AAEV,gBAAI,aAAa,SAAS,EAAE,SAAS,cAAc;AACjD;AAAA,YACF;AACA,kBAAM;AAAA,UACR;AAAA,QACF;AAAA,QACA;AAAA,UACE,SAAS,KAAK,YAAY;AAAA,QAC5B;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AAEV,UAAI,aAAa,SAAS,EAAE,SAAS,cAAc;AAEjD;AAAA,MACF;AACA,YAAM;AAAA,IACR,UAAE;AAEA,YAAM,gBAAgB;AAAA,IACxB;AAAA,EACF;AACF;","names":[]}
package/dist/version.cjs CHANGED
@@ -21,7 +21,7 @@ __export(version_exports, {
21
21
  version: () => version
22
22
  });
23
23
  module.exports = __toCommonJS(version_exports);
24
- const version = "1.0.48";
24
+ const version = "1.0.49";
25
25
  // Annotate the CommonJS export names for ESM import in node:
26
26
  0 && (module.exports = {
27
27
  version
package/dist/version.js CHANGED
@@ -1,4 +1,4 @@
1
- const version = "1.0.48";
1
+ const version = "1.0.49";
2
2
  export {
3
3
  version
4
4
  };
@@ -341,13 +341,27 @@ class AgentActivity {
341
341
  attachAudioInput(audioStream) {
342
342
  void this.audioStream.close();
343
343
  this.audioStream = new import_multi_input_stream.MultiInputStream();
344
+ const aecWarmupAudioFilter = new import_web.TransformStream({
345
+ transform: (frame, controller) => {
346
+ const shouldDiscardForAecWarmup = this.agentSession.agentState === "speaking" && this.agentSession._aecWarmupRemaining > 0;
347
+ if (!shouldDiscardForAecWarmup) {
348
+ controller.enqueue(frame);
349
+ }
350
+ }
351
+ });
344
352
  this.audioStreamId = this.audioStream.addInputStream(audioStream);
345
- const [realtimeAudioStream, recognitionAudioStream] = this.audioStream.stream.tee();
346
- if (this.realtimeSession) {
353
+ if (this.realtimeSession && this.audioRecognition) {
354
+ const [realtimeAudioStream, recognitionAudioStream] = this.audioStream.stream.pipeThrough(aecWarmupAudioFilter).tee();
347
355
  this.realtimeSession.setInputAudioStream(realtimeAudioStream);
348
- }
349
- if (this.audioRecognition) {
350
356
  this.audioRecognition.setInputAudioStream(recognitionAudioStream);
357
+ } else if (this.realtimeSession) {
358
+ this.realtimeSession.setInputAudioStream(
359
+ this.audioStream.stream.pipeThrough(aecWarmupAudioFilter)
360
+ );
361
+ } else if (this.audioRecognition) {
362
+ this.audioRecognition.setInputAudioStream(
363
+ this.audioStream.stream.pipeThrough(aecWarmupAudioFilter)
364
+ );
351
365
  }
352
366
  }
353
367
  detachAudioInput() {
@@ -543,6 +557,9 @@ class AgentActivity {
543
557
  }
544
558
  interruptByAudioActivity() {
545
559
  var _a, _b;
560
+ if (this.agentSession._aecWarmupRemaining > 0) {
561
+ return;
562
+ }
546
563
  if (this.llm instanceof import_llm.RealtimeModel && this.llm.capabilities.turnDetection) {
547
564
  return;
548
565
  }
@@ -861,7 +878,16 @@ ${instructions}`;
861
878
  speech.interrupt(force);
862
879
  }
863
880
  (_a = this.realtimeSession) == null ? void 0 : _a.interrupt();
864
- if (currentSpeech === void 0) {
881
+ if (force) {
882
+ for (const task of this.speechTasks) {
883
+ task.cancel();
884
+ }
885
+ if (currentSpeech && !currentSpeech.done()) {
886
+ currentSpeech._markDone();
887
+ }
888
+ this.speechQueue.clear();
889
+ future.resolve();
890
+ } else if (currentSpeech === void 0) {
865
891
  future.resolve();
866
892
  } else {
867
893
  currentSpeech.addDoneCallback(() => {
@@ -1229,9 +1255,7 @@ ${instructions}`;
1229
1255
  audioOutput.clearBuffer();
1230
1256
  }
1231
1257
  replyAbortController.abort();
1232
- await Promise.allSettled(
1233
- tasks.map((task) => task.cancelAndWait(AgentActivity.REPLY_TASK_CANCEL_TIMEOUT))
1234
- );
1258
+ await (0, import_utils.cancelAndWait)(tasks, AgentActivity.REPLY_TASK_CANCEL_TIMEOUT);
1235
1259
  let forwardedText = (textOut == null ? void 0 : textOut.text) || "";
1236
1260
  if (audioOutput) {
1237
1261
  const playbackEv = await audioOutput.waitForPlayout();
@@ -1861,6 +1885,10 @@ ${instructions}`;
1861
1885
  const unlock = await this.lock.lock();
1862
1886
  try {
1863
1887
  this.cancelPreemptiveGeneration();
1888
+ await (0, import_utils.cancelAndWait)(Array.from(this.speechTasks), AgentActivity.REPLY_TASK_CANCEL_TIMEOUT);
1889
+ if (this._currentSpeech && !this._currentSpeech.done()) {
1890
+ this._currentSpeech._markDone();
1891
+ }
1864
1892
  await this._closeSessionResources();
1865
1893
  if (this._mainTask) {
1866
1894
  await this._mainTask.cancelAndWait();