@livekit/agents 1.0.40 → 1.0.41
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/cli.cjs +20 -18
- package/dist/cli.cjs.map +1 -1
- package/dist/cli.d.ts.map +1 -1
- package/dist/cli.js +20 -18
- package/dist/cli.js.map +1 -1
- package/dist/index.cjs +5 -0
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.cts +1 -0
- package/dist/index.d.ts +1 -0
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +3 -0
- package/dist/index.js.map +1 -1
- package/dist/inference/stt.cjs +2 -1
- package/dist/inference/stt.cjs.map +1 -1
- package/dist/inference/stt.d.ts.map +1 -1
- package/dist/inference/stt.js +2 -1
- package/dist/inference/stt.js.map +1 -1
- package/dist/llm/realtime.cjs.map +1 -1
- package/dist/llm/realtime.d.cts +5 -1
- package/dist/llm/realtime.d.ts +5 -1
- package/dist/llm/realtime.d.ts.map +1 -1
- package/dist/llm/realtime.js.map +1 -1
- package/dist/tts/stream_adapter.cjs +15 -1
- package/dist/tts/stream_adapter.cjs.map +1 -1
- package/dist/tts/stream_adapter.d.ts.map +1 -1
- package/dist/tts/stream_adapter.js +15 -1
- package/dist/tts/stream_adapter.js.map +1 -1
- package/dist/tts/tts.cjs.map +1 -1
- package/dist/tts/tts.d.cts +9 -1
- package/dist/tts/tts.d.ts +9 -1
- package/dist/tts/tts.d.ts.map +1 -1
- package/dist/tts/tts.js.map +1 -1
- package/dist/types.cjs +3 -0
- package/dist/types.cjs.map +1 -1
- package/dist/types.d.cts +4 -0
- package/dist/types.d.ts +4 -0
- package/dist/types.d.ts.map +1 -1
- package/dist/types.js +2 -0
- package/dist/types.js.map +1 -1
- package/dist/voice/agent.cjs +11 -1
- package/dist/voice/agent.cjs.map +1 -1
- package/dist/voice/agent.d.cts +7 -3
- package/dist/voice/agent.d.ts +7 -3
- package/dist/voice/agent.d.ts.map +1 -1
- package/dist/voice/agent.js +11 -1
- package/dist/voice/agent.js.map +1 -1
- package/dist/voice/agent_activity.cjs +30 -14
- package/dist/voice/agent_activity.cjs.map +1 -1
- package/dist/voice/agent_activity.d.cts +1 -0
- package/dist/voice/agent_activity.d.ts +1 -0
- package/dist/voice/agent_activity.d.ts.map +1 -1
- package/dist/voice/agent_activity.js +30 -14
- package/dist/voice/agent_activity.js.map +1 -1
- package/dist/voice/agent_session.cjs +5 -1
- package/dist/voice/agent_session.cjs.map +1 -1
- package/dist/voice/agent_session.d.cts +2 -0
- package/dist/voice/agent_session.d.ts +2 -0
- package/dist/voice/agent_session.d.ts.map +1 -1
- package/dist/voice/agent_session.js +5 -1
- package/dist/voice/agent_session.js.map +1 -1
- package/dist/voice/background_audio.cjs +2 -1
- package/dist/voice/background_audio.cjs.map +1 -1
- package/dist/voice/background_audio.d.cts +4 -2
- package/dist/voice/background_audio.d.ts +4 -2
- package/dist/voice/background_audio.d.ts.map +1 -1
- package/dist/voice/background_audio.js +2 -1
- package/dist/voice/background_audio.js.map +1 -1
- package/dist/voice/generation.cjs +58 -5
- package/dist/voice/generation.cjs.map +1 -1
- package/dist/voice/generation.d.cts +17 -3
- package/dist/voice/generation.d.ts +17 -3
- package/dist/voice/generation.d.ts.map +1 -1
- package/dist/voice/generation.js +63 -6
- package/dist/voice/generation.js.map +1 -1
- package/dist/voice/index.cjs.map +1 -1
- package/dist/voice/index.d.cts +1 -1
- package/dist/voice/index.d.ts +1 -1
- package/dist/voice/index.d.ts.map +1 -1
- package/dist/voice/index.js.map +1 -1
- package/dist/voice/io.cjs +22 -2
- package/dist/voice/io.cjs.map +1 -1
- package/dist/voice/io.d.cts +21 -5
- package/dist/voice/io.d.ts +21 -5
- package/dist/voice/io.d.ts.map +1 -1
- package/dist/voice/io.js +18 -1
- package/dist/voice/io.js.map +1 -1
- package/dist/voice/room_io/_output.cjs +3 -2
- package/dist/voice/room_io/_output.cjs.map +1 -1
- package/dist/voice/room_io/_output.d.cts +3 -3
- package/dist/voice/room_io/_output.d.ts +3 -3
- package/dist/voice/room_io/_output.d.ts.map +1 -1
- package/dist/voice/room_io/_output.js +4 -3
- package/dist/voice/room_io/_output.js.map +1 -1
- package/dist/voice/transcription/synchronizer.cjs +137 -13
- package/dist/voice/transcription/synchronizer.cjs.map +1 -1
- package/dist/voice/transcription/synchronizer.d.cts +34 -4
- package/dist/voice/transcription/synchronizer.d.ts +34 -4
- package/dist/voice/transcription/synchronizer.d.ts.map +1 -1
- package/dist/voice/transcription/synchronizer.js +141 -14
- package/dist/voice/transcription/synchronizer.js.map +1 -1
- package/dist/voice/transcription/synchronizer.test.cjs +151 -0
- package/dist/voice/transcription/synchronizer.test.cjs.map +1 -0
- package/dist/voice/transcription/synchronizer.test.js +150 -0
- package/dist/voice/transcription/synchronizer.test.js.map +1 -0
- package/package.json +1 -1
- package/src/cli.ts +20 -18
- package/src/index.ts +1 -0
- package/src/inference/stt.ts +9 -8
- package/src/llm/realtime.ts +5 -1
- package/src/tts/stream_adapter.ts +23 -1
- package/src/tts/tts.ts +10 -1
- package/src/types.ts +5 -0
- package/src/voice/agent.ts +19 -4
- package/src/voice/agent_activity.ts +38 -13
- package/src/voice/agent_session.ts +6 -0
- package/src/voice/background_audio.ts +6 -3
- package/src/voice/generation.ts +115 -10
- package/src/voice/index.ts +1 -1
- package/src/voice/io.ts +40 -5
- package/src/voice/room_io/_output.ts +6 -5
- package/src/voice/transcription/synchronizer.test.ts +206 -0
- package/src/voice/transcription/synchronizer.ts +202 -17
package/dist/tts/tts.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../../src/tts/tts.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AACpD,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAGhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AAGtD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,oBAAoB,CAAC;AAGrD,OAAO,EAAE,KAAK,iBAAiB,EAAiD,MAAM,aAAa,CAAC;AACpG,OAAO,EAAE,kBAAkB,EAA0C,MAAM,aAAa,CAAC;
|
|
1
|
+
{"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../../src/tts/tts.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AACpD,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAGhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AAGtD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,oBAAoB,CAAC;AAGrD,OAAO,EAAE,KAAK,iBAAiB,EAAiD,MAAM,aAAa,CAAC;AACpG,OAAO,EAAE,kBAAkB,EAA0C,MAAM,aAAa,CAAC;AACzF,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,gBAAgB,CAAC;AAElD;;GAEG;AACH,MAAM,WAAW,gBAAgB;IAC/B,qEAAqE;IACrE,SAAS,EAAE,MAAM,CAAC;IAClB,uDAAuD;IACvD,SAAS,EAAE,MAAM,CAAC;IAClB,8BAA8B;IAC9B,KAAK,EAAE,UAAU,CAAC;IAClB,+CAA+C;IAC/C,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,qEAAqE;IACrE,KAAK,EAAE,OAAO,CAAC;IACf;;OAEG;IACH,gBAAgB,CAAC,EAAE,WAAW,EAAE,CAAC;CAClC;AAED;;;;;;GAMG;AACH,MAAM,WAAW,eAAe;IAC9B,SAAS,EAAE,OAAO,CAAC;IAEnB,iBAAiB,CAAC,EAAE,OAAO,CAAC;CAC7B;AAED,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,WAAW,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;IAClB,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,KAAK,CAAC;IACb,WAAW,EAAE,OAAO,CAAC;CACtB;AAED,MAAM,MAAM,YAAY,GAAG;IACzB,CAAC,mBAAmB,CAAC,EAAE,CAAC,OAAO,EAAE,UAAU,KAAK,IAAI,CAAC;IACrD,CAAC,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,CAAC;CACtC,CAAC;kCAS2D,aAAa,YAAY,CAAC;AAPvF;;;;;;GAMG;AACH,8BAAsB,GAAI,SAAQ,QAAsD;;IAItF,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;gBAEX,UAAU,EAAE,MAAM,EAAE,WAAW,EAAE,MAAM,EAAE,YAAY,EAAE,eAAe;IAOlF,sCAAsC;IACtC,IAAI,YAAY,IAAI,eAAe,CAElC;IAED,mEAAmE;IACnE,IAAI,UAAU,IAAI,MAAM,CAEvB;IAED,qEAAqE;IACrE,IAAI,WAAW,IAAI,MAAM,CAExB;IAED;;OAEG;IACH,QAAQ,CAAC,UAAU,CACjB,IAAI,EAAE,MAAM,EACZ,WAAW,CAAC,EAAE,iBAAiB,EAC/B,WAAW,CAAC,EAAE,WAAW,GACxB,aAAa;IAEhB;;;;OAIG;IACH,QAAQ,CAAC,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,gBAAgB;IAE1E,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;CAG7B;AAED;;;;;;;;;;;;;GAaG;AACH,8BAAsB,gBACpB,YAAW,qBAAqB,CAAC,gBAAgB,GAAG,OAAO,gBAAgB,CAAC,aAAa,CAAC;;IAE1F,SAAS,CAAC,MAAM,CAAC,QAAQ,CAAC,cAAc,gBAA4B;IACpE,MAAM,CAAC,QAAQ,CAAC,aAAa,gBAA2B;IACxD,SAAS,CAAC,KAAK,sEAA6E;IAC5F,SAAS,CAAC,KAAK,+EAEX;IACJ,SAAS,CAAC,MAAM,+EAEZ;IACJ,SAAS,CAAC,MAAM,UAAS;IACzB,SAAS,CAAC,WAAW,EAAE,iBAAiB,CAAC;IACzC,SAAS,CAAC,eAAe,kBAAyB;IAElD,OAAO,CAAC,mBAAmB,CAEzB;IACF,OAAO,CAAC,MAAM,CAAS;IAEvB,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;gBAQX,GAAG,EAAE,GAAG,EAAE,WAAW,GAAE,iBAA+C;IAqBlF,OAAO,CAAC,aAAa,CAoDnB;IAEF,OAAO,CAAC,QAAQ,CAIX;IAEL,OAAO,CAAC,SAAS;cAoBD,SAAS;cAuBT,cAAc;IAyD9B,SAAS,CAAC,QAAQ,CAAC,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IAEvC,iBAAiB,CAAC,IAAI,EAAE,cAAc,CAAC,MAAM,CAAC;IAI9C,uCAAuC;IACvC,kDAAkD;IAClD,QAAQ,CAAC,IAAI,EAAE,MAAM;IAgBrB,4DAA4D;IAC5D,KAAK;IAcL,2DAA2D;IAC3D,QAAQ;IAWR,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,gBAAgB,GAAG,OAAO,gBAAgB,CAAC,aAAa,CAAC,CAAC;IAIzF,IAAI,WAAW,IAAI,WAAW,CAE7B;IAED,wDAAwD;IACxD,KAAK;IAIL,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,gBAAgB;CAG3C;AAED;;;;;;;;;;;;;GAaG;AACH,8BAAsB,aAAc,YAAW,qBAAqB,CAAC,gBAAgB,CAAC;;IACpF,SAAS,CAAC,KAAK,uCAA8C;IAC7D,SAAS,CAAC,MAAM,uCAA8C;IAC9D,SAAS,CAAC,MAAM,UAAS;IACzB,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IAIvB,OAAO,CAAC,YAAY,CAAoB;IACxC,OAAO,CAAC,MAAM,CAAS;IAEvB,SAAS,CAAC,eAAe,kBAAyB;gBAGhD,IAAI,EAAE,MAAM,EACZ,GAAG,EAAE,GAAG,EACR,WAAW,GAAE,iBAA+C,EAC5D,WAAW,CAAC,EAAE,WAAW;IAmB3B,OAAO,CAAC,aAAa,CAoDnB;YAEY,QAAQ;IAOtB,OAAO,CAAC,SAAS;IAUjB,SAAS,CAAC,QAAQ,CAAC,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IAEvC,IAAI,SAAS,IAAI,MAAM,CAEtB;IAED,IAAI,WAAW,IAAI,WAAW,CAE7B;cAEe,cAAc;IAuC9B,oDAAoD;IAC9C,OAAO,IAAI,OAAO,CAAC,UAAU,CAAC;IAQpC,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC,gBAAgB,CAAC,CAAC;IAIjD,wDAAwD;IACxD,KAAK;IAOL,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,aAAa;CAGxC"}
|
package/dist/tts/tts.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/tts/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport type { Span } from '@opentelemetry/api';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { log } from '../log.js';\nimport type { TTSMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { recordException, traceTypes, tracer } from '../telemetry/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS, intervalForRetry } from '../types.js';\nimport { AsyncIterableQueue, delay, mergeFrames, startSoon, toError } from '../utils.js';\n\n/** SynthesizedAudio is a packet of speech synthesis as returned by the TTS. */\nexport interface SynthesizedAudio {\n /** Request ID (one segment could be made up of multiple requests) */\n requestId: string;\n /** Segment ID, each segment is separated by a flush */\n segmentId: string;\n /** Synthesized audio frame */\n frame: AudioFrame;\n /** Current segment of the synthesized audio */\n deltaText?: string;\n /** Whether this is the last frame of the segment (streaming only) */\n final: boolean;\n}\n\n/**\n * Describes the capabilities of the TTS provider.\n *\n * @remarks\n * At present, only `streaming` is supplied to this interface, and the framework only supports\n * providers that do have a streaming endpoint.\n */\nexport interface TTSCapabilities {\n streaming: boolean;\n}\n\nexport interface TTSError {\n type: 'tts_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type TTSCallbacks = {\n ['metrics_collected']: (metrics: TTSMetrics) => void;\n ['error']: (error: TTSError) => void;\n};\n\n/**\n * An instance of a text-to-speech adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child TTS class, which inherits this class's methods.\n */\nexport abstract class TTS extends (EventEmitter as new () => TypedEmitter<TTSCallbacks>) {\n #capabilities: TTSCapabilities;\n #sampleRate: number;\n #numChannels: number;\n abstract label: string;\n\n constructor(sampleRate: number, numChannels: number, capabilities: TTSCapabilities) {\n super();\n this.#capabilities = capabilities;\n this.#sampleRate = sampleRate;\n this.#numChannels = numChannels;\n }\n\n /** Returns this TTS's capabilities */\n get capabilities(): TTSCapabilities {\n return this.#capabilities;\n }\n\n /** Returns the sample rate of audio frames returned by this TTS */\n get sampleRate(): number {\n return this.#sampleRate;\n }\n\n /** Returns the channel count of audio frames returned by this TTS */\n get numChannels(): number {\n return this.#numChannels;\n }\n\n /**\n * Receives text and returns synthesis in the form of a {@link ChunkedStream}\n */\n abstract synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): ChunkedStream;\n\n /**\n * Returns a {@link SynthesizeStream} that can be used to push text and receive audio data\n *\n * @param options - Optional configuration including connection options\n */\n abstract stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a text-to-speech stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * await source.captureFrame(event.frame);\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SynthesizeStream class, which inherits this class's methods.\n */\nexport abstract class SynthesizeStream\n implements AsyncIterableIterator<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>\n{\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n static readonly END_OF_STREAM = Symbol('END_OF_STREAM');\n protected input = new AsyncIterableQueue<string | typeof SynthesizeStream.FLUSH_SENTINEL>();\n protected queue = new AsyncIterableQueue<\n SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n >();\n protected output = new AsyncIterableQueue<\n SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n >();\n protected closed = false;\n protected connOptions: APIConnectOptions;\n protected abortController = new AbortController();\n\n private deferredInputStream: DeferredReadableStream<\n string | typeof SynthesizeStream.FLUSH_SENTINEL\n >;\n private logger = log();\n\n abstract label: string;\n\n #tts: TTS;\n #metricsPendingTexts: string[] = [];\n #metricsText = '';\n #monitorMetricsTask?: Promise<void>;\n #ttsRequestSpan?: Span;\n\n constructor(tts: TTS, connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS) {\n this.#tts = tts;\n this.connOptions = connOptions;\n this.deferredInputStream = new DeferredReadableStream();\n this.pumpInput();\n\n this.abortController.signal.addEventListener('abort', () => {\n this.deferredInputStream.detachSource();\n // TODO (AJS-36) clean this up when we refactor with streams\n if (!this.input.closed) this.input.close();\n if (!this.output.closed) this.output.close();\n this.closed = true;\n });\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n startSoon(() => this.mainTask().finally(() => this.queue.close()));\n }\n\n private _mainTaskImpl = async (span: Span) => {\n this.#ttsRequestSpan = span;\n span.setAttributes({\n [traceTypes.ATTR_TTS_STREAMING]: true,\n [traceTypes.ATTR_TTS_LABEL]: this.#tts.label,\n });\n\n for (let i = 0; i < this.connOptions.maxRetry + 1; i++) {\n try {\n return await tracer.startActiveSpan(\n async (attemptSpan) => {\n attemptSpan.setAttribute(traceTypes.ATTR_RETRY_COUNT, i);\n try {\n return await this.run();\n } catch (error) {\n recordException(attemptSpan, toError(error));\n throw error;\n }\n },\n { name: 'tts_request_run' },\n );\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = intervalForRetry(this.connOptions, i);\n\n if (this.connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this.connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to generate TTS completion after ${this.connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#tts.label, attempt: i + 1, error },\n `failed to synthesize speech, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n };\n\n private mainTask = async () =>\n tracer.startActiveSpan(async (span) => this._mainTaskImpl(span), {\n name: 'tts_request',\n endOnExit: false,\n });\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#tts.emit('error', {\n type: 'tts_error',\n timestamp: Date.now(),\n label: this.#tts.label,\n error,\n recoverable,\n });\n }\n\n // NOTE(AJS-37): The implementation below uses an AsyncIterableQueue (`this.input`)\n // bridged from a DeferredReadableStream (`this.deferredInputStream`) rather than\n // consuming the stream directly.\n //\n // A full refactor to native Web Streams was considered but is currently deferred.\n // The primary reason is to maintain architectural parity with the Python SDK,\n // which is a key design goal for the project. This ensures a consistent developer\n // experience across both platforms.\n //\n // For more context, see the discussion in GitHub issue # 844.\n protected async pumpInput() {\n const reader = this.deferredInputStream.stream.getReader();\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done || value === SynthesizeStream.FLUSH_SENTINEL) {\n break;\n }\n this.pushText(value);\n }\n this.endInput();\n } catch (error) {\n this.logger.error(error, 'Error reading deferred input stream');\n } finally {\n reader.releaseLock();\n // Ensure output is closed when the stream ends\n if (!this.#monitorMetricsTask) {\n // No text was received, close the output directly\n this.output.close();\n }\n }\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let audioDurationMs = 0;\n let ttfb: bigint = BigInt(-1);\n let requestId = '';\n\n const emit = () => {\n if (this.#metricsPendingTexts.length) {\n const text = this.#metricsPendingTexts.shift()!;\n const duration = process.hrtime.bigint() - startTime;\n const roundedAudioDurationMs = Math.round(audioDurationMs);\n const metrics: TTSMetrics = {\n type: 'tts_metrics',\n timestamp: Date.now(),\n requestId,\n ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n charactersCount: text.length,\n audioDurationMs: roundedAudioDurationMs,\n cancelled: this.abortController.signal.aborted,\n label: this.#tts.label,\n streamed: false,\n };\n if (this.#ttsRequestSpan) {\n this.#ttsRequestSpan.setAttribute(traceTypes.ATTR_TTS_METRICS, JSON.stringify(metrics));\n }\n this.#tts.emit('metrics_collected', metrics);\n }\n };\n\n for await (const audio of this.queue) {\n if (this.abortController.signal.aborted) {\n break;\n }\n this.output.put(audio);\n if (audio === SynthesizeStream.END_OF_STREAM) continue;\n requestId = audio.requestId;\n if (ttfb === BigInt(-1)) {\n ttfb = process.hrtime.bigint() - startTime;\n }\n // TODO(AJS-102): use frame.durationMs once available in rtc-node\n audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n if (audio.final) {\n emit();\n }\n }\n\n if (requestId) {\n emit();\n }\n\n if (this.#ttsRequestSpan) {\n this.#ttsRequestSpan.end();\n this.#ttsRequestSpan = undefined;\n }\n }\n\n protected abstract run(): Promise<void>;\n\n updateInputStream(text: ReadableStream<string>) {\n this.deferredInputStream.setSource(text);\n }\n\n /** Push a string of text to the TTS */\n /** @deprecated Use `updateInputStream` instead */\n pushText(text: string) {\n if (!this.#monitorMetricsTask) {\n this.#monitorMetricsTask = this.monitorMetrics();\n // Close output when metrics task completes\n this.#monitorMetricsTask.finally(() => this.output.close());\n }\n this.#metricsText += text;\n\n if (this.input.closed || this.closed) {\n // Stream was aborted/closed, silently skip\n return;\n }\n\n this.input.put(text);\n }\n\n /** Flush the TTS, causing it to process all pending text */\n flush() {\n if (this.#metricsText) {\n this.#metricsPendingTexts.push(this.#metricsText);\n this.#metricsText = '';\n }\n\n if (this.input.closed || this.closed) {\n // Stream was aborted/closed, silently skip\n return;\n }\n\n this.input.put(SynthesizeStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n this.flush();\n\n if (this.input.closed || this.closed) {\n // Stream was aborted/closed, silently skip\n return;\n }\n\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>> {\n return this.output.next();\n }\n\n get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n /** Close both the input and output of the TTS stream */\n close() {\n this.abortController.abort();\n }\n\n [Symbol.asyncIterator](): SynthesizeStream {\n return this;\n }\n}\n\n/**\n * An instance of a text-to-speech response, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * await source.captureFrame(event.frame);\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child ChunkedStream class, which inherits this class's methods.\n */\nexport abstract class ChunkedStream implements AsyncIterableIterator<SynthesizedAudio> {\n protected queue = new AsyncIterableQueue<SynthesizedAudio>();\n protected output = new AsyncIterableQueue<SynthesizedAudio>();\n protected closed = false;\n abstract label: string;\n #text: string;\n #tts: TTS;\n #ttsRequestSpan?: Span;\n private _connOptions: APIConnectOptions;\n private logger = log();\n\n protected abortController = new AbortController();\n\n constructor(\n text: string,\n tts: TTS,\n connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n abortSignal?: AbortSignal,\n ) {\n this.#text = text;\n this.#tts = tts;\n this._connOptions = connOptions;\n\n if (abortSignal) {\n abortSignal.addEventListener('abort', () => this.abortController.abort(), { once: true });\n }\n\n this.monitorMetrics();\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n Promise.resolve().then(() => this.mainTask().finally(() => this.queue.close()));\n }\n\n private _mainTaskImpl = async (span: Span) => {\n this.#ttsRequestSpan = span;\n span.setAttributes({\n [traceTypes.ATTR_TTS_STREAMING]: false,\n [traceTypes.ATTR_TTS_LABEL]: this.#tts.label,\n });\n\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await tracer.startActiveSpan(\n async (attemptSpan) => {\n attemptSpan.setAttribute(traceTypes.ATTR_RETRY_COUNT, i);\n try {\n return await this.run();\n } catch (error) {\n recordException(attemptSpan, toError(error));\n throw error;\n }\n },\n { name: 'tts_request_run' },\n );\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = intervalForRetry(this._connOptions, i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to generate TTS completion after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#tts.label, attempt: i + 1, error },\n `failed to generate TTS completion, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n };\n\n private async mainTask() {\n return tracer.startActiveSpan(async (span) => this._mainTaskImpl(span), {\n name: 'tts_request',\n endOnExit: false,\n });\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#tts.emit('error', {\n type: 'tts_error',\n timestamp: Date.now(),\n label: this.#tts.label,\n error,\n recoverable,\n });\n }\n\n protected abstract run(): Promise<void>;\n\n get inputText(): string {\n return this.#text;\n }\n\n get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let audioDurationMs = 0;\n let ttfb: bigint = BigInt(-1);\n let requestId = '';\n\n for await (const audio of this.queue) {\n this.output.put(audio);\n requestId = audio.requestId;\n if (ttfb === BigInt(-1)) {\n ttfb = process.hrtime.bigint() - startTime;\n }\n audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n }\n this.output.close();\n\n const duration = process.hrtime.bigint() - startTime;\n const metrics: TTSMetrics = {\n type: 'tts_metrics',\n timestamp: Date.now(),\n requestId,\n ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n charactersCount: this.#text.length,\n audioDurationMs: Math.round(audioDurationMs),\n cancelled: false, // TODO(AJS-186): support ChunkedStream with 1.0 - add this.abortController.signal.aborted here\n label: this.#tts.label,\n streamed: false,\n };\n\n if (this.#ttsRequestSpan) {\n this.#ttsRequestSpan.setAttribute(traceTypes.ATTR_TTS_METRICS, JSON.stringify(metrics));\n this.#ttsRequestSpan.end();\n this.#ttsRequestSpan = undefined;\n }\n\n this.#tts.emit('metrics_collected', metrics);\n }\n\n /** Collect every frame into one in a single call */\n async collect(): Promise<AudioFrame> {\n const frames = [];\n for await (const event of this) {\n frames.push(event.frame);\n }\n return mergeFrames(frames);\n }\n\n next(): Promise<IteratorResult<SynthesizedAudio>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the TTS stream */\n close() {\n if (!this.queue.closed) this.queue.close();\n if (!this.output.closed) this.output.close();\n if (!this.abortController.signal.aborted) this.abortController.abort();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): ChunkedStream {\n return this;\n }\n}\n"],"mappings":"AAMA,SAAS,oBAAoB;AAE7B,SAAS,oBAAoB,gBAAgB;AAC7C,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAS,iBAAiB,YAAY,cAAc;AACpD,SAAiC,6BAA6B,wBAAwB;AACtF,SAAS,oBAAoB,OAAO,aAAa,WAAW,eAAe;AA+CpE,MAAe,YAAa,aAAsD;AAAA,EACvF;AAAA,EACA;AAAA,EACA;AAAA,EAGA,YAAY,YAAoB,aAAqB,cAA+B;AAClF,UAAM;AACN,SAAK,gBAAgB;AACrB,SAAK,cAAc;AACnB,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,aAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,cAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAkBA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAgBO,MAAe,iBAEtB;AAAA,EACE,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EAClE,OAAgB,gBAAgB,OAAO,eAAe;AAAA,EAC5C,QAAQ,IAAI,mBAAoE;AAAA,EAChF,QAAQ,IAAI,mBAEpB;AAAA,EACQ,SAAS,IAAI,mBAErB;AAAA,EACQ,SAAS;AAAA,EACT;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAExC;AAAA,EAGA,SAAS,IAAI;AAAA,EAIrB;AAAA,EACA,uBAAiC,CAAC;AAAA,EAClC,eAAe;AAAA,EACf;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,cAAiC,6BAA6B;AAClF,SAAK,OAAO;AACZ,SAAK,cAAc;AACnB,SAAK,sBAAsB,IAAI,uBAAuB;AACtD,SAAK,UAAU;AAEf,SAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,WAAK,oBAAoB,aAAa;AAEtC,UAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,UAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,WAAK,SAAS;AAAA,IAChB,CAAC;AAMD,cAAU,MAAM,KAAK,SAAS,EAAE,QAAQ,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EACnE;AAAA,EAEQ,gBAAgB,OAAO,SAAe;AAC5C,SAAK,kBAAkB;AACvB,SAAK,cAAc;AAAA,MACjB,CAAC,WAAW,kBAAkB,GAAG;AAAA,MACjC,CAAC,WAAW,cAAc,GAAG,KAAK,KAAK;AAAA,IACzC,CAAC;AAED,aAAS,IAAI,GAAG,IAAI,KAAK,YAAY,WAAW,GAAG,KAAK;AACtD,UAAI;AACF,eAAO,MAAM,OAAO;AAAA,UAClB,OAAO,gBAAgB;AACrB,wBAAY,aAAa,WAAW,kBAAkB,CAAC;AACvD,gBAAI;AACF,qBAAO,MAAM,KAAK,IAAI;AAAA,YACxB,SAAS,OAAO;AACd,8BAAgB,aAAa,QAAQ,KAAK,CAAC;AAC3C,oBAAM;AAAA,YACR;AAAA,UACF;AAAA,UACA,EAAE,MAAM,kBAAkB;AAAA,QAC5B;AAAA,MACF,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,iBAAiB,KAAK,aAAa,CAAC;AAE1D,cAAI,KAAK,YAAY,aAAa,KAAK,CAAC,MAAM,WAAW;AACvD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,YAAY,UAAU;AAC1C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,mBAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,YAAY,WAAW,CAAC;AAAA,cACjF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,6CAA6C,aAAa;AAAA,YAC5D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,WAAW,YACjB,OAAO,gBAAgB,OAAO,SAAS,KAAK,cAAc,IAAI,GAAG;AAAA,IAC/D,MAAM;AAAA,IACN,WAAW;AAAA,EACb,CAAC;AAAA,EAEK,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,MAAgB,YAAY;AAC1B,UAAM,SAAS,KAAK,oBAAoB,OAAO,UAAU;AACzD,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,QAAQ,UAAU,iBAAiB,gBAAgB;AACrD;AAAA,QACF;AACA,aAAK,SAAS,KAAK;AAAA,MACrB;AACA,WAAK,SAAS;AAAA,IAChB,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,OAAO,qCAAqC;AAAA,IAChE,UAAE;AACA,aAAO,YAAY;AAEnB,UAAI,CAAC,KAAK,qBAAqB;AAE7B,aAAK,OAAO,MAAM;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,UAAM,OAAO,MAAM;AACjB,UAAI,KAAK,qBAAqB,QAAQ;AACpC,cAAM,OAAO,KAAK,qBAAqB,MAAM;AAC7C,cAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,cAAM,yBAAyB,KAAK,MAAM,eAAe;AACzD,cAAM,UAAsB;AAAA,UAC1B,MAAM;AAAA,UACN,WAAW,KAAK,IAAI;AAAA,UACpB;AAAA,UACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,UAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,UACzD,iBAAiB,KAAK;AAAA,UACtB,iBAAiB;AAAA,UACjB,WAAW,KAAK,gBAAgB,OAAO;AAAA,UACvC,OAAO,KAAK,KAAK;AAAA,UACjB,UAAU;AAAA,QACZ;AACA,YAAI,KAAK,iBAAiB;AACxB,eAAK,gBAAgB,aAAa,WAAW,kBAAkB,KAAK,UAAU,OAAO,CAAC;AAAA,QACxF;AACA,aAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,MAC7C;AAAA,IACF;AAEA,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,KAAK,gBAAgB,OAAO,SAAS;AACvC;AAAA,MACF;AACA,WAAK,OAAO,IAAI,KAAK;AACrB,UAAI,UAAU,iBAAiB,cAAe;AAC9C,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AAEA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAC9E,UAAI,MAAM,OAAO;AACf,aAAK;AAAA,MACP;AAAA,IACF;AAEA,QAAI,WAAW;AACb,WAAK;AAAA,IACP;AAEA,QAAI,KAAK,iBAAiB;AACxB,WAAK,gBAAgB,IAAI;AACzB,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;AAAA,EAIA,kBAAkB,MAA8B;AAC9C,SAAK,oBAAoB,UAAU,IAAI;AAAA,EACzC;AAAA;AAAA;AAAA,EAIA,SAAS,MAAc;AACrB,QAAI,CAAC,KAAK,qBAAqB;AAC7B,WAAK,sBAAsB,KAAK,eAAe;AAE/C,WAAK,oBAAoB,QAAQ,MAAM,KAAK,OAAO,MAAM,CAAC;AAAA,IAC5D;AACA,SAAK,gBAAgB;AAErB,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,IAAI,IAAI;AAAA,EACrB;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,cAAc;AACrB,WAAK,qBAAqB,KAAK,KAAK,YAAY;AAChD,WAAK,eAAe;AAAA,IACtB;AAEA,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,IAAI,iBAAiB,cAAc;AAAA,EAChD;AAAA;AAAA,EAGA,WAAW;AACT,SAAK,MAAM;AAEX,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA0F;AACxF,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA,EAEA,IAAI,cAA2B;AAC7B,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA,EAEA,CAAC,OAAO,aAAa,IAAsB;AACzC,WAAO;AAAA,EACT;AACF;AAgBO,MAAe,cAAiE;AAAA,EAC3E,QAAQ,IAAI,mBAAqC;AAAA,EACjD,SAAS,IAAI,mBAAqC;AAAA,EAClD,SAAS;AAAA,EAEnB;AAAA,EACA;AAAA,EACA;AAAA,EACQ;AAAA,EACA,SAAS,IAAI;AAAA,EAEX,kBAAkB,IAAI,gBAAgB;AAAA,EAEhD,YACE,MACA,KACA,cAAiC,6BACjC,aACA;AACA,SAAK,QAAQ;AACb,SAAK,OAAO;AACZ,SAAK,eAAe;AAEpB,QAAI,aAAa;AACf,kBAAY,iBAAiB,SAAS,MAAM,KAAK,gBAAgB,MAAM,GAAG,EAAE,MAAM,KAAK,CAAC;AAAA,IAC1F;AAEA,SAAK,eAAe;AAMpB,YAAQ,QAAQ,EAAE,KAAK,MAAM,KAAK,SAAS,EAAE,QAAQ,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChF;AAAA,EAEQ,gBAAgB,OAAO,SAAe;AAC5C,SAAK,kBAAkB;AACvB,SAAK,cAAc;AAAA,MACjB,CAAC,WAAW,kBAAkB,GAAG;AAAA,MACjC,CAAC,WAAW,cAAc,GAAG,KAAK,KAAK;AAAA,IACzC,CAAC;AAED,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,OAAO;AAAA,UAClB,OAAO,gBAAgB;AACrB,wBAAY,aAAa,WAAW,kBAAkB,CAAC;AACvD,gBAAI;AACF,qBAAO,MAAM,KAAK,IAAI;AAAA,YACxB,SAAS,OAAO;AACd,8BAAgB,aAAa,QAAQ,KAAK,CAAC;AAC3C,oBAAM;AAAA,YACR;AAAA,UACF;AAAA,UACA,EAAE,MAAM,kBAAkB;AAAA,QAC5B;AAAA,MACF,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,iBAAiB,KAAK,cAAc,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,mBAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,kDAAkD,aAAa;AAAA,YACjE;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,WAAW;AACvB,WAAO,OAAO,gBAAgB,OAAO,SAAS,KAAK,cAAc,IAAI,GAAG;AAAA,MACtE,MAAM;AAAA,MACN,WAAW;AAAA,IACb,CAAC;AAAA,EACH;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAIA,IAAI,YAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAA2B;AAC7B,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,qBAAiB,SAAS,KAAK,OAAO;AACpC,WAAK,OAAO,IAAI,KAAK;AACrB,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AACA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAAA,IAChF;AACA,SAAK,OAAO,MAAM;AAElB,UAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,UAAM,UAAsB;AAAA,MAC1B,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,MAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,MACzD,iBAAiB,KAAK,MAAM;AAAA,MAC5B,iBAAiB,KAAK,MAAM,eAAe;AAAA,MAC3C,WAAW;AAAA;AAAA,MACX,OAAO,KAAK,KAAK;AAAA,MACjB,UAAU;AAAA,IACZ;AAEA,QAAI,KAAK,iBAAiB;AACxB,WAAK,gBAAgB,aAAa,WAAW,kBAAkB,KAAK,UAAU,OAAO,CAAC;AACtF,WAAK,gBAAgB,IAAI;AACzB,WAAK,kBAAkB;AAAA,IACzB;AAEA,SAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,EAC7C;AAAA;AAAA,EAGA,MAAM,UAA+B;AACnC,UAAM,SAAS,CAAC;AAChB,qBAAiB,SAAS,MAAM;AAC9B,aAAO,KAAK,MAAM,KAAK;AAAA,IACzB;AACA,WAAO,YAAY,MAAM;AAAA,EAC3B;AAAA,EAEA,OAAkD;AAChD,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,QAAI,CAAC,KAAK,gBAAgB,OAAO,QAAS,MAAK,gBAAgB,MAAM;AACrE,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAmB;AACtC,WAAO;AAAA,EACT;AACF;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../../src/tts/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport type { Span } from '@opentelemetry/api';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { log } from '../log.js';\nimport type { TTSMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { recordException, traceTypes, tracer } from '../telemetry/index.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS, intervalForRetry } from '../types.js';\nimport { AsyncIterableQueue, delay, mergeFrames, startSoon, toError } from '../utils.js';\nimport type { TimedString } from '../voice/io.js';\n\n/**\n * SynthesizedAudio is a packet of speech synthesis as returned by the TTS.\n */\nexport interface SynthesizedAudio {\n /** Request ID (one segment could be made up of multiple requests) */\n requestId: string;\n /** Segment ID, each segment is separated by a flush */\n segmentId: string;\n /** Synthesized audio frame */\n frame: AudioFrame;\n /** Current segment of the synthesized audio */\n deltaText?: string;\n /** Whether this is the last frame of the segment (streaming only) */\n final: boolean;\n /**\n * Timed transcripts associated with this audio packet (word-level timestamps).\n */\n timedTranscripts?: TimedString[];\n}\n\n/**\n * Describes the capabilities of the TTS provider.\n *\n * @remarks\n * At present, only `streaming` is supplied to this interface, and the framework only supports\n * providers that do have a streaming endpoint.\n */\nexport interface TTSCapabilities {\n streaming: boolean;\n // Whether this TTS supports aligned transcripts (word-level timestamps).\n alignedTranscript?: boolean;\n}\n\nexport interface TTSError {\n type: 'tts_error';\n timestamp: number;\n label: string;\n error: Error;\n recoverable: boolean;\n}\n\nexport type TTSCallbacks = {\n ['metrics_collected']: (metrics: TTSMetrics) => void;\n ['error']: (error: TTSError) => void;\n};\n\n/**\n * An instance of a text-to-speech adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child TTS class, which inherits this class's methods.\n */\nexport abstract class TTS extends (EventEmitter as new () => TypedEmitter<TTSCallbacks>) {\n #capabilities: TTSCapabilities;\n #sampleRate: number;\n #numChannels: number;\n abstract label: string;\n\n constructor(sampleRate: number, numChannels: number, capabilities: TTSCapabilities) {\n super();\n this.#capabilities = capabilities;\n this.#sampleRate = sampleRate;\n this.#numChannels = numChannels;\n }\n\n /** Returns this TTS's capabilities */\n get capabilities(): TTSCapabilities {\n return this.#capabilities;\n }\n\n /** Returns the sample rate of audio frames returned by this TTS */\n get sampleRate(): number {\n return this.#sampleRate;\n }\n\n /** Returns the channel count of audio frames returned by this TTS */\n get numChannels(): number {\n return this.#numChannels;\n }\n\n /**\n * Receives text and returns synthesis in the form of a {@link ChunkedStream}\n */\n abstract synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): ChunkedStream;\n\n /**\n * Returns a {@link SynthesizeStream} that can be used to push text and receive audio data\n *\n * @param options - Optional configuration including connection options\n */\n abstract stream(options?: { connOptions?: APIConnectOptions }): SynthesizeStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\n/**\n * An instance of a text-to-speech stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * await source.captureFrame(event.frame);\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SynthesizeStream class, which inherits this class's methods.\n */\nexport abstract class SynthesizeStream\n implements AsyncIterableIterator<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>\n{\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n static readonly END_OF_STREAM = Symbol('END_OF_STREAM');\n protected input = new AsyncIterableQueue<string | typeof SynthesizeStream.FLUSH_SENTINEL>();\n protected queue = new AsyncIterableQueue<\n SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n >();\n protected output = new AsyncIterableQueue<\n SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n >();\n protected closed = false;\n protected connOptions: APIConnectOptions;\n protected abortController = new AbortController();\n\n private deferredInputStream: DeferredReadableStream<\n string | typeof SynthesizeStream.FLUSH_SENTINEL\n >;\n private logger = log();\n\n abstract label: string;\n\n #tts: TTS;\n #metricsPendingTexts: string[] = [];\n #metricsText = '';\n #monitorMetricsTask?: Promise<void>;\n #ttsRequestSpan?: Span;\n\n constructor(tts: TTS, connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS) {\n this.#tts = tts;\n this.connOptions = connOptions;\n this.deferredInputStream = new DeferredReadableStream();\n this.pumpInput();\n\n this.abortController.signal.addEventListener('abort', () => {\n this.deferredInputStream.detachSource();\n // TODO (AJS-36) clean this up when we refactor with streams\n if (!this.input.closed) this.input.close();\n if (!this.output.closed) this.output.close();\n this.closed = true;\n });\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n startSoon(() => this.mainTask().finally(() => this.queue.close()));\n }\n\n private _mainTaskImpl = async (span: Span) => {\n this.#ttsRequestSpan = span;\n span.setAttributes({\n [traceTypes.ATTR_TTS_STREAMING]: true,\n [traceTypes.ATTR_TTS_LABEL]: this.#tts.label,\n });\n\n for (let i = 0; i < this.connOptions.maxRetry + 1; i++) {\n try {\n return await tracer.startActiveSpan(\n async (attemptSpan) => {\n attemptSpan.setAttribute(traceTypes.ATTR_RETRY_COUNT, i);\n try {\n return await this.run();\n } catch (error) {\n recordException(attemptSpan, toError(error));\n throw error;\n }\n },\n { name: 'tts_request_run' },\n );\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = intervalForRetry(this.connOptions, i);\n\n if (this.connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this.connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to generate TTS completion after ${this.connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#tts.label, attempt: i + 1, error },\n `failed to synthesize speech, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n };\n\n private mainTask = async () =>\n tracer.startActiveSpan(async (span) => this._mainTaskImpl(span), {\n name: 'tts_request',\n endOnExit: false,\n });\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#tts.emit('error', {\n type: 'tts_error',\n timestamp: Date.now(),\n label: this.#tts.label,\n error,\n recoverable,\n });\n }\n\n // NOTE(AJS-37): The implementation below uses an AsyncIterableQueue (`this.input`)\n // bridged from a DeferredReadableStream (`this.deferredInputStream`) rather than\n // consuming the stream directly.\n //\n // A full refactor to native Web Streams was considered but is currently deferred.\n // The primary reason is to maintain architectural parity with the Python SDK,\n // which is a key design goal for the project. This ensures a consistent developer\n // experience across both platforms.\n //\n // For more context, see the discussion in GitHub issue # 844.\n protected async pumpInput() {\n const reader = this.deferredInputStream.stream.getReader();\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done || value === SynthesizeStream.FLUSH_SENTINEL) {\n break;\n }\n this.pushText(value);\n }\n this.endInput();\n } catch (error) {\n this.logger.error(error, 'Error reading deferred input stream');\n } finally {\n reader.releaseLock();\n // Ensure output is closed when the stream ends\n if (!this.#monitorMetricsTask) {\n // No text was received, close the output directly\n this.output.close();\n }\n }\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let audioDurationMs = 0;\n let ttfb: bigint = BigInt(-1);\n let requestId = '';\n\n const emit = () => {\n if (this.#metricsPendingTexts.length) {\n const text = this.#metricsPendingTexts.shift()!;\n const duration = process.hrtime.bigint() - startTime;\n const roundedAudioDurationMs = Math.round(audioDurationMs);\n const metrics: TTSMetrics = {\n type: 'tts_metrics',\n timestamp: Date.now(),\n requestId,\n ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n charactersCount: text.length,\n audioDurationMs: roundedAudioDurationMs,\n cancelled: this.abortController.signal.aborted,\n label: this.#tts.label,\n streamed: false,\n };\n if (this.#ttsRequestSpan) {\n this.#ttsRequestSpan.setAttribute(traceTypes.ATTR_TTS_METRICS, JSON.stringify(metrics));\n }\n this.#tts.emit('metrics_collected', metrics);\n }\n };\n\n for await (const audio of this.queue) {\n if (this.abortController.signal.aborted) {\n break;\n }\n this.output.put(audio);\n if (audio === SynthesizeStream.END_OF_STREAM) continue;\n requestId = audio.requestId;\n if (ttfb === BigInt(-1)) {\n ttfb = process.hrtime.bigint() - startTime;\n }\n // TODO(AJS-102): use frame.durationMs once available in rtc-node\n audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n if (audio.final) {\n emit();\n }\n }\n\n if (requestId) {\n emit();\n }\n\n if (this.#ttsRequestSpan) {\n this.#ttsRequestSpan.end();\n this.#ttsRequestSpan = undefined;\n }\n }\n\n protected abstract run(): Promise<void>;\n\n updateInputStream(text: ReadableStream<string>) {\n this.deferredInputStream.setSource(text);\n }\n\n /** Push a string of text to the TTS */\n /** @deprecated Use `updateInputStream` instead */\n pushText(text: string) {\n if (!this.#monitorMetricsTask) {\n this.#monitorMetricsTask = this.monitorMetrics();\n // Close output when metrics task completes\n this.#monitorMetricsTask.finally(() => this.output.close());\n }\n this.#metricsText += text;\n\n if (this.input.closed || this.closed) {\n // Stream was aborted/closed, silently skip\n return;\n }\n\n this.input.put(text);\n }\n\n /** Flush the TTS, causing it to process all pending text */\n flush() {\n if (this.#metricsText) {\n this.#metricsPendingTexts.push(this.#metricsText);\n this.#metricsText = '';\n }\n\n if (this.input.closed || this.closed) {\n // Stream was aborted/closed, silently skip\n return;\n }\n\n this.input.put(SynthesizeStream.FLUSH_SENTINEL);\n }\n\n /** Mark the input as ended and forbid additional pushes */\n endInput() {\n this.flush();\n\n if (this.input.closed || this.closed) {\n // Stream was aborted/closed, silently skip\n return;\n }\n\n this.input.close();\n }\n\n next(): Promise<IteratorResult<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>> {\n return this.output.next();\n }\n\n get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n /** Close both the input and output of the TTS stream */\n close() {\n this.abortController.abort();\n }\n\n [Symbol.asyncIterator](): SynthesizeStream {\n return this;\n }\n}\n\n/**\n * An instance of a text-to-speech response, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n * await source.captureFrame(event.frame);\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child ChunkedStream class, which inherits this class's methods.\n */\nexport abstract class ChunkedStream implements AsyncIterableIterator<SynthesizedAudio> {\n protected queue = new AsyncIterableQueue<SynthesizedAudio>();\n protected output = new AsyncIterableQueue<SynthesizedAudio>();\n protected closed = false;\n abstract label: string;\n #text: string;\n #tts: TTS;\n #ttsRequestSpan?: Span;\n private _connOptions: APIConnectOptions;\n private logger = log();\n\n protected abortController = new AbortController();\n\n constructor(\n text: string,\n tts: TTS,\n connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n abortSignal?: AbortSignal,\n ) {\n this.#text = text;\n this.#tts = tts;\n this._connOptions = connOptions;\n\n if (abortSignal) {\n abortSignal.addEventListener('abort', () => this.abortController.abort(), { once: true });\n }\n\n this.monitorMetrics();\n\n // this is a hack to immitate asyncio.create_task so that mainTask\n // is run **after** the constructor has finished. Otherwise we get\n // runtime error when trying to access class variables in the\n // `run` method.\n Promise.resolve().then(() => this.mainTask().finally(() => this.queue.close()));\n }\n\n private _mainTaskImpl = async (span: Span) => {\n this.#ttsRequestSpan = span;\n span.setAttributes({\n [traceTypes.ATTR_TTS_STREAMING]: false,\n [traceTypes.ATTR_TTS_LABEL]: this.#tts.label,\n });\n\n for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n try {\n return await tracer.startActiveSpan(\n async (attemptSpan) => {\n attemptSpan.setAttribute(traceTypes.ATTR_RETRY_COUNT, i);\n try {\n return await this.run();\n } catch (error) {\n recordException(attemptSpan, toError(error));\n throw error;\n }\n },\n { name: 'tts_request_run' },\n );\n } catch (error) {\n if (error instanceof APIError) {\n const retryInterval = intervalForRetry(this._connOptions, i);\n\n if (this._connOptions.maxRetry === 0 || !error.retryable) {\n this.emitError({ error, recoverable: false });\n throw error;\n } else if (i === this._connOptions.maxRetry) {\n this.emitError({ error, recoverable: false });\n throw new APIConnectionError({\n message: `failed to generate TTS completion after ${this._connOptions.maxRetry + 1} attempts`,\n options: { retryable: false },\n });\n } else {\n // Don't emit error event for recoverable errors during retry loop\n // to avoid ERR_UNHANDLED_ERROR or premature session termination\n this.logger.warn(\n { tts: this.#tts.label, attempt: i + 1, error },\n `failed to generate TTS completion, retrying in ${retryInterval}s`,\n );\n }\n\n if (retryInterval > 0) {\n await delay(retryInterval);\n }\n } else {\n this.emitError({ error: toError(error), recoverable: false });\n throw error;\n }\n }\n }\n };\n\n private async mainTask() {\n return tracer.startActiveSpan(async (span) => this._mainTaskImpl(span), {\n name: 'tts_request',\n endOnExit: false,\n });\n }\n\n private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n this.#tts.emit('error', {\n type: 'tts_error',\n timestamp: Date.now(),\n label: this.#tts.label,\n error,\n recoverable,\n });\n }\n\n protected abstract run(): Promise<void>;\n\n get inputText(): string {\n return this.#text;\n }\n\n get abortSignal(): AbortSignal {\n return this.abortController.signal;\n }\n\n protected async monitorMetrics() {\n const startTime = process.hrtime.bigint();\n let audioDurationMs = 0;\n let ttfb: bigint = BigInt(-1);\n let requestId = '';\n\n for await (const audio of this.queue) {\n this.output.put(audio);\n requestId = audio.requestId;\n if (ttfb === BigInt(-1)) {\n ttfb = process.hrtime.bigint() - startTime;\n }\n audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n }\n this.output.close();\n\n const duration = process.hrtime.bigint() - startTime;\n const metrics: TTSMetrics = {\n type: 'tts_metrics',\n timestamp: Date.now(),\n requestId,\n ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n charactersCount: this.#text.length,\n audioDurationMs: Math.round(audioDurationMs),\n cancelled: false, // TODO(AJS-186): support ChunkedStream with 1.0 - add this.abortController.signal.aborted here\n label: this.#tts.label,\n streamed: false,\n };\n\n if (this.#ttsRequestSpan) {\n this.#ttsRequestSpan.setAttribute(traceTypes.ATTR_TTS_METRICS, JSON.stringify(metrics));\n this.#ttsRequestSpan.end();\n this.#ttsRequestSpan = undefined;\n }\n\n this.#tts.emit('metrics_collected', metrics);\n }\n\n /** Collect every frame into one in a single call */\n async collect(): Promise<AudioFrame> {\n const frames = [];\n for await (const event of this) {\n frames.push(event.frame);\n }\n return mergeFrames(frames);\n }\n\n next(): Promise<IteratorResult<SynthesizedAudio>> {\n return this.output.next();\n }\n\n /** Close both the input and output of the TTS stream */\n close() {\n if (!this.queue.closed) this.queue.close();\n if (!this.output.closed) this.output.close();\n if (!this.abortController.signal.aborted) this.abortController.abort();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): ChunkedStream {\n return this;\n }\n}\n"],"mappings":"AAMA,SAAS,oBAAoB;AAE7B,SAAS,oBAAoB,gBAAgB;AAC7C,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAS,iBAAiB,YAAY,cAAc;AACpD,SAAiC,6BAA6B,wBAAwB;AACtF,SAAS,oBAAoB,OAAO,aAAa,WAAW,eAAe;AAwDpE,MAAe,YAAa,aAAsD;AAAA,EACvF;AAAA,EACA;AAAA,EACA;AAAA,EAGA,YAAY,YAAoB,aAAqB,cAA+B;AAClF,UAAM;AACN,SAAK,gBAAgB;AACrB,SAAK,cAAc;AACnB,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,aAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,cAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAkBA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAgBO,MAAe,iBAEtB;AAAA,EACE,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EAClE,OAAgB,gBAAgB,OAAO,eAAe;AAAA,EAC5C,QAAQ,IAAI,mBAAoE;AAAA,EAChF,QAAQ,IAAI,mBAEpB;AAAA,EACQ,SAAS,IAAI,mBAErB;AAAA,EACQ,SAAS;AAAA,EACT;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAExC;AAAA,EAGA,SAAS,IAAI;AAAA,EAIrB;AAAA,EACA,uBAAiC,CAAC;AAAA,EAClC,eAAe;AAAA,EACf;AAAA,EACA;AAAA,EAEA,YAAY,KAAU,cAAiC,6BAA6B;AAClF,SAAK,OAAO;AACZ,SAAK,cAAc;AACnB,SAAK,sBAAsB,IAAI,uBAAuB;AACtD,SAAK,UAAU;AAEf,SAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,WAAK,oBAAoB,aAAa;AAEtC,UAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,UAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,WAAK,SAAS;AAAA,IAChB,CAAC;AAMD,cAAU,MAAM,KAAK,SAAS,EAAE,QAAQ,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EACnE;AAAA,EAEQ,gBAAgB,OAAO,SAAe;AAC5C,SAAK,kBAAkB;AACvB,SAAK,cAAc;AAAA,MACjB,CAAC,WAAW,kBAAkB,GAAG;AAAA,MACjC,CAAC,WAAW,cAAc,GAAG,KAAK,KAAK;AAAA,IACzC,CAAC;AAED,aAAS,IAAI,GAAG,IAAI,KAAK,YAAY,WAAW,GAAG,KAAK;AACtD,UAAI;AACF,eAAO,MAAM,OAAO;AAAA,UAClB,OAAO,gBAAgB;AACrB,wBAAY,aAAa,WAAW,kBAAkB,CAAC;AACvD,gBAAI;AACF,qBAAO,MAAM,KAAK,IAAI;AAAA,YACxB,SAAS,OAAO;AACd,8BAAgB,aAAa,QAAQ,KAAK,CAAC;AAC3C,oBAAM;AAAA,YACR;AAAA,UACF;AAAA,UACA,EAAE,MAAM,kBAAkB;AAAA,QAC5B;AAAA,MACF,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,iBAAiB,KAAK,aAAa,CAAC;AAE1D,cAAI,KAAK,YAAY,aAAa,KAAK,CAAC,MAAM,WAAW;AACvD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,YAAY,UAAU;AAC1C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,mBAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,YAAY,WAAW,CAAC;AAAA,cACjF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,6CAA6C,aAAa;AAAA,YAC5D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,WAAW,YACjB,OAAO,gBAAgB,OAAO,SAAS,KAAK,cAAc,IAAI,GAAG;AAAA,IAC/D,MAAM;AAAA,IACN,WAAW;AAAA,EACb,CAAC;AAAA,EAEK,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,MAAgB,YAAY;AAC1B,UAAM,SAAS,KAAK,oBAAoB,OAAO,UAAU;AACzD,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,QAAQ,UAAU,iBAAiB,gBAAgB;AACrD;AAAA,QACF;AACA,aAAK,SAAS,KAAK;AAAA,MACrB;AACA,WAAK,SAAS;AAAA,IAChB,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,OAAO,qCAAqC;AAAA,IAChE,UAAE;AACA,aAAO,YAAY;AAEnB,UAAI,CAAC,KAAK,qBAAqB;AAE7B,aAAK,OAAO,MAAM;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,UAAM,OAAO,MAAM;AACjB,UAAI,KAAK,qBAAqB,QAAQ;AACpC,cAAM,OAAO,KAAK,qBAAqB,MAAM;AAC7C,cAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,cAAM,yBAAyB,KAAK,MAAM,eAAe;AACzD,cAAM,UAAsB;AAAA,UAC1B,MAAM;AAAA,UACN,WAAW,KAAK,IAAI;AAAA,UACpB;AAAA,UACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,UAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,UACzD,iBAAiB,KAAK;AAAA,UACtB,iBAAiB;AAAA,UACjB,WAAW,KAAK,gBAAgB,OAAO;AAAA,UACvC,OAAO,KAAK,KAAK;AAAA,UACjB,UAAU;AAAA,QACZ;AACA,YAAI,KAAK,iBAAiB;AACxB,eAAK,gBAAgB,aAAa,WAAW,kBAAkB,KAAK,UAAU,OAAO,CAAC;AAAA,QACxF;AACA,aAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,MAC7C;AAAA,IACF;AAEA,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,KAAK,gBAAgB,OAAO,SAAS;AACvC;AAAA,MACF;AACA,WAAK,OAAO,IAAI,KAAK;AACrB,UAAI,UAAU,iBAAiB,cAAe;AAC9C,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AAEA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAC9E,UAAI,MAAM,OAAO;AACf,aAAK;AAAA,MACP;AAAA,IACF;AAEA,QAAI,WAAW;AACb,WAAK;AAAA,IACP;AAEA,QAAI,KAAK,iBAAiB;AACxB,WAAK,gBAAgB,IAAI;AACzB,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;AAAA,EAIA,kBAAkB,MAA8B;AAC9C,SAAK,oBAAoB,UAAU,IAAI;AAAA,EACzC;AAAA;AAAA;AAAA,EAIA,SAAS,MAAc;AACrB,QAAI,CAAC,KAAK,qBAAqB;AAC7B,WAAK,sBAAsB,KAAK,eAAe;AAE/C,WAAK,oBAAoB,QAAQ,MAAM,KAAK,OAAO,MAAM,CAAC;AAAA,IAC5D;AACA,SAAK,gBAAgB;AAErB,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,IAAI,IAAI;AAAA,EACrB;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,cAAc;AACrB,WAAK,qBAAqB,KAAK,KAAK,YAAY;AAChD,WAAK,eAAe;AAAA,IACtB;AAEA,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,IAAI,iBAAiB,cAAc;AAAA,EAChD;AAAA;AAAA,EAGA,WAAW;AACT,SAAK,MAAM;AAEX,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA0F;AACxF,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA,EAEA,IAAI,cAA2B;AAC7B,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA,EAEA,CAAC,OAAO,aAAa,IAAsB;AACzC,WAAO;AAAA,EACT;AACF;AAgBO,MAAe,cAAiE;AAAA,EAC3E,QAAQ,IAAI,mBAAqC;AAAA,EACjD,SAAS,IAAI,mBAAqC;AAAA,EAClD,SAAS;AAAA,EAEnB;AAAA,EACA;AAAA,EACA;AAAA,EACQ;AAAA,EACA,SAAS,IAAI;AAAA,EAEX,kBAAkB,IAAI,gBAAgB;AAAA,EAEhD,YACE,MACA,KACA,cAAiC,6BACjC,aACA;AACA,SAAK,QAAQ;AACb,SAAK,OAAO;AACZ,SAAK,eAAe;AAEpB,QAAI,aAAa;AACf,kBAAY,iBAAiB,SAAS,MAAM,KAAK,gBAAgB,MAAM,GAAG,EAAE,MAAM,KAAK,CAAC;AAAA,IAC1F;AAEA,SAAK,eAAe;AAMpB,YAAQ,QAAQ,EAAE,KAAK,MAAM,KAAK,SAAS,EAAE,QAAQ,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChF;AAAA,EAEQ,gBAAgB,OAAO,SAAe;AAC5C,SAAK,kBAAkB;AACvB,SAAK,cAAc;AAAA,MACjB,CAAC,WAAW,kBAAkB,GAAG;AAAA,MACjC,CAAC,WAAW,cAAc,GAAG,KAAK,KAAK;AAAA,IACzC,CAAC;AAED,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AACF,eAAO,MAAM,OAAO;AAAA,UAClB,OAAO,gBAAgB;AACrB,wBAAY,aAAa,WAAW,kBAAkB,CAAC;AACvD,gBAAI;AACF,qBAAO,MAAM,KAAK,IAAI;AAAA,YACxB,SAAS,OAAO;AACd,8BAAgB,aAAa,QAAQ,KAAK,CAAC;AAC3C,oBAAM;AAAA,YACR;AAAA,UACF;AAAA,UACA,EAAE,MAAM,kBAAkB;AAAA,QAC5B;AAAA,MACF,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,iBAAiB,KAAK,cAAc,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,mBAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,kDAAkD,aAAa;AAAA,YACjE;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,WAAW;AACvB,WAAO,OAAO,gBAAgB,OAAO,SAAS,KAAK,cAAc,IAAI,GAAG;AAAA,MACtE,MAAM;AAAA,MACN,WAAW;AAAA,IACb,CAAC;AAAA,EACH;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAIA,IAAI,YAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAA2B;AAC7B,WAAO,KAAK,gBAAgB;AAAA,EAC9B;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,qBAAiB,SAAS,KAAK,OAAO;AACpC,WAAK,OAAO,IAAI,KAAK;AACrB,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AACA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAAA,IAChF;AACA,SAAK,OAAO,MAAM;AAElB,UAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,UAAM,UAAsB;AAAA,MAC1B,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,MAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,MACzD,iBAAiB,KAAK,MAAM;AAAA,MAC5B,iBAAiB,KAAK,MAAM,eAAe;AAAA,MAC3C,WAAW;AAAA;AAAA,MACX,OAAO,KAAK,KAAK;AAAA,MACjB,UAAU;AAAA,IACZ;AAEA,QAAI,KAAK,iBAAiB;AACxB,WAAK,gBAAgB,aAAa,WAAW,kBAAkB,KAAK,UAAU,OAAO,CAAC;AACtF,WAAK,gBAAgB,IAAI;AACzB,WAAK,kBAAkB;AAAA,IACzB;AAEA,SAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,EAC7C;AAAA;AAAA,EAGA,MAAM,UAA+B;AACnC,UAAM,SAAS,CAAC;AAChB,qBAAiB,SAAS,MAAM;AAC9B,aAAO,KAAK,MAAM,KAAK;AAAA,IACzB;AACA,WAAO,YAAY,MAAM;AAAA,EAC3B;AAAA,EAEA,OAAkD;AAChD,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,CAAC,KAAK,MAAM,OAAQ,MAAK,MAAM,MAAM;AACzC,QAAI,CAAC,KAAK,OAAO,OAAQ,MAAK,OAAO,MAAM;AAC3C,QAAI,CAAC,KAAK,gBAAgB,OAAO,QAAS,MAAK,gBAAgB,MAAM;AACrE,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAmB;AACtC,WAAO;AAAA,EACT;AACF;","names":[]}
|
package/dist/types.cjs
CHANGED
|
@@ -20,9 +20,11 @@ var types_exports = {};
|
|
|
20
20
|
__export(types_exports, {
|
|
21
21
|
DEFAULT_API_CONNECT_OPTIONS: () => DEFAULT_API_CONNECT_OPTIONS,
|
|
22
22
|
DEFAULT_SESSION_CONNECT_OPTIONS: () => DEFAULT_SESSION_CONNECT_OPTIONS,
|
|
23
|
+
USERDATA_TIMED_TRANSCRIPT: () => USERDATA_TIMED_TRANSCRIPT,
|
|
23
24
|
intervalForRetry: () => intervalForRetry
|
|
24
25
|
});
|
|
25
26
|
module.exports = __toCommonJS(types_exports);
|
|
27
|
+
const USERDATA_TIMED_TRANSCRIPT = "lk.timed_transcripts";
|
|
26
28
|
const DEFAULT_API_CONNECT_OPTIONS = {
|
|
27
29
|
maxRetry: 3,
|
|
28
30
|
retryIntervalMs: 2e3,
|
|
@@ -44,6 +46,7 @@ const DEFAULT_SESSION_CONNECT_OPTIONS = {
|
|
|
44
46
|
0 && (module.exports = {
|
|
45
47
|
DEFAULT_API_CONNECT_OPTIONS,
|
|
46
48
|
DEFAULT_SESSION_CONNECT_OPTIONS,
|
|
49
|
+
USERDATA_TIMED_TRANSCRIPT,
|
|
47
50
|
intervalForRetry
|
|
48
51
|
});
|
|
49
52
|
//# sourceMappingURL=types.cjs.map
|
package/dist/types.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/types.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\n/**\n * Connection options for API calls, controlling retry and timeout behavior.\n */\nexport interface APIConnectOptions {\n /** Maximum number of retries to connect to the API. Default: 3 */\n maxRetry: number;\n /** Interval between retries to connect to the API in milliseconds. Default: 2000 */\n retryIntervalMs: number;\n /** Timeout for connecting to the API in milliseconds. Default: 10000 */\n timeoutMs: number;\n}\n\nexport const DEFAULT_API_CONNECT_OPTIONS: APIConnectOptions = {\n maxRetry: 3,\n retryIntervalMs: 2000,\n timeoutMs: 10000,\n};\n\n/**\n * Return the interval for the given number of retries.\n * The first retry is immediate, and then uses specified retryIntervalMs.\n * @internal\n */\nexport function intervalForRetry(connOptions: APIConnectOptions, numRetries: number): number {\n if (numRetries === 0) {\n return 0.1;\n }\n return connOptions.retryIntervalMs;\n}\n\n/**\n * Connection options for the agent session, controlling retry and timeout behavior\n * for STT, LLM, and TTS connections.\n */\nexport interface SessionConnectOptions {\n /** Connection options for speech-to-text. */\n sttConnOptions?: Partial<APIConnectOptions>;\n /** Connection options for the language model. */\n llmConnOptions?: Partial<APIConnectOptions>;\n /** Connection options for text-to-speech. */\n ttsConnOptions?: Partial<APIConnectOptions>;\n /** Maximum number of consecutive unrecoverable errors from LLM or TTS before closing the session. Default: 3 */\n maxUnrecoverableErrors?: number;\n}\n\n/**\n * Resolved session connect options with all values populated.\n * @internal\n */\nexport interface ResolvedSessionConnectOptions {\n sttConnOptions: APIConnectOptions;\n llmConnOptions: APIConnectOptions;\n ttsConnOptions: APIConnectOptions;\n maxUnrecoverableErrors: number;\n}\n\nexport const DEFAULT_SESSION_CONNECT_OPTIONS: ResolvedSessionConnectOptions = {\n sttConnOptions: DEFAULT_API_CONNECT_OPTIONS,\n llmConnOptions: DEFAULT_API_CONNECT_OPTIONS,\n ttsConnOptions: DEFAULT_API_CONNECT_OPTIONS,\n maxUnrecoverableErrors: 3,\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;
|
|
1
|
+
{"version":3,"sources":["../src/types.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\n/**\n * Key used to store timed transcripts in AudioFrame.userdata.\n */\nexport const USERDATA_TIMED_TRANSCRIPT = 'lk.timed_transcripts';\n\n/**\n * Connection options for API calls, controlling retry and timeout behavior.\n */\nexport interface APIConnectOptions {\n /** Maximum number of retries to connect to the API. Default: 3 */\n maxRetry: number;\n /** Interval between retries to connect to the API in milliseconds. Default: 2000 */\n retryIntervalMs: number;\n /** Timeout for connecting to the API in milliseconds. Default: 10000 */\n timeoutMs: number;\n}\n\nexport const DEFAULT_API_CONNECT_OPTIONS: APIConnectOptions = {\n maxRetry: 3,\n retryIntervalMs: 2000,\n timeoutMs: 10000,\n};\n\n/**\n * Return the interval for the given number of retries.\n * The first retry is immediate, and then uses specified retryIntervalMs.\n * @internal\n */\nexport function intervalForRetry(connOptions: APIConnectOptions, numRetries: number): number {\n if (numRetries === 0) {\n return 0.1;\n }\n return connOptions.retryIntervalMs;\n}\n\n/**\n * Connection options for the agent session, controlling retry and timeout behavior\n * for STT, LLM, and TTS connections.\n */\nexport interface SessionConnectOptions {\n /** Connection options for speech-to-text. */\n sttConnOptions?: Partial<APIConnectOptions>;\n /** Connection options for the language model. */\n llmConnOptions?: Partial<APIConnectOptions>;\n /** Connection options for text-to-speech. */\n ttsConnOptions?: Partial<APIConnectOptions>;\n /** Maximum number of consecutive unrecoverable errors from LLM or TTS before closing the session. Default: 3 */\n maxUnrecoverableErrors?: number;\n}\n\n/**\n * Resolved session connect options with all values populated.\n * @internal\n */\nexport interface ResolvedSessionConnectOptions {\n sttConnOptions: APIConnectOptions;\n llmConnOptions: APIConnectOptions;\n ttsConnOptions: APIConnectOptions;\n maxUnrecoverableErrors: number;\n}\n\nexport const DEFAULT_SESSION_CONNECT_OPTIONS: ResolvedSessionConnectOptions = {\n sttConnOptions: DEFAULT_API_CONNECT_OPTIONS,\n llmConnOptions: DEFAULT_API_CONNECT_OPTIONS,\n ttsConnOptions: DEFAULT_API_CONNECT_OPTIONS,\n maxUnrecoverableErrors: 3,\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAOO,MAAM,4BAA4B;AAclC,MAAM,8BAAiD;AAAA,EAC5D,UAAU;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AACb;AAOO,SAAS,iBAAiB,aAAgC,YAA4B;AAC3F,MAAI,eAAe,GAAG;AACpB,WAAO;AAAA,EACT;AACA,SAAO,YAAY;AACrB;AA4BO,MAAM,kCAAiE;AAAA,EAC5E,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,wBAAwB;AAC1B;","names":[]}
|
package/dist/types.d.cts
CHANGED
package/dist/types.d.ts
CHANGED
package/dist/types.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAIA;;GAEG;AACH,MAAM,WAAW,iBAAiB;IAChC,kEAAkE;IAClE,QAAQ,EAAE,MAAM,CAAC;IACjB,oFAAoF;IACpF,eAAe,EAAE,MAAM,CAAC;IACxB,wEAAwE;IACxE,SAAS,EAAE,MAAM,CAAC;CACnB;AAED,eAAO,MAAM,2BAA2B,EAAE,iBAIzC,CAAC;AAEF;;;;GAIG;AACH,wBAAgB,gBAAgB,CAAC,WAAW,EAAE,iBAAiB,EAAE,UAAU,EAAE,MAAM,GAAG,MAAM,CAK3F;AAED;;;GAGG;AACH,MAAM,WAAW,qBAAqB;IACpC,6CAA6C;IAC7C,cAAc,CAAC,EAAE,OAAO,CAAC,iBAAiB,CAAC,CAAC;IAC5C,iDAAiD;IACjD,cAAc,CAAC,EAAE,OAAO,CAAC,iBAAiB,CAAC,CAAC;IAC5C,6CAA6C;IAC7C,cAAc,CAAC,EAAE,OAAO,CAAC,iBAAiB,CAAC,CAAC;IAC5C,gHAAgH;IAChH,sBAAsB,CAAC,EAAE,MAAM,CAAC;CACjC;AAED;;;GAGG;AACH,MAAM,WAAW,6BAA6B;IAC5C,cAAc,EAAE,iBAAiB,CAAC;IAClC,cAAc,EAAE,iBAAiB,CAAC;IAClC,cAAc,EAAE,iBAAiB,CAAC;IAClC,sBAAsB,EAAE,MAAM,CAAC;CAChC;AAED,eAAO,MAAM,+BAA+B,EAAE,6BAK7C,CAAC"}
|
|
1
|
+
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAIA;;GAEG;AACH,eAAO,MAAM,yBAAyB,yBAAyB,CAAC;AAEhE;;GAEG;AACH,MAAM,WAAW,iBAAiB;IAChC,kEAAkE;IAClE,QAAQ,EAAE,MAAM,CAAC;IACjB,oFAAoF;IACpF,eAAe,EAAE,MAAM,CAAC;IACxB,wEAAwE;IACxE,SAAS,EAAE,MAAM,CAAC;CACnB;AAED,eAAO,MAAM,2BAA2B,EAAE,iBAIzC,CAAC;AAEF;;;;GAIG;AACH,wBAAgB,gBAAgB,CAAC,WAAW,EAAE,iBAAiB,EAAE,UAAU,EAAE,MAAM,GAAG,MAAM,CAK3F;AAED;;;GAGG;AACH,MAAM,WAAW,qBAAqB;IACpC,6CAA6C;IAC7C,cAAc,CAAC,EAAE,OAAO,CAAC,iBAAiB,CAAC,CAAC;IAC5C,iDAAiD;IACjD,cAAc,CAAC,EAAE,OAAO,CAAC,iBAAiB,CAAC,CAAC;IAC5C,6CAA6C;IAC7C,cAAc,CAAC,EAAE,OAAO,CAAC,iBAAiB,CAAC,CAAC;IAC5C,gHAAgH;IAChH,sBAAsB,CAAC,EAAE,MAAM,CAAC;CACjC;AAED;;;GAGG;AACH,MAAM,WAAW,6BAA6B;IAC5C,cAAc,EAAE,iBAAiB,CAAC;IAClC,cAAc,EAAE,iBAAiB,CAAC;IAClC,cAAc,EAAE,iBAAiB,CAAC;IAClC,sBAAsB,EAAE,MAAM,CAAC;CAChC;AAED,eAAO,MAAM,+BAA+B,EAAE,6BAK7C,CAAC"}
|
package/dist/types.js
CHANGED
|
@@ -1,3 +1,4 @@
|
|
|
1
|
+
const USERDATA_TIMED_TRANSCRIPT = "lk.timed_transcripts";
|
|
1
2
|
const DEFAULT_API_CONNECT_OPTIONS = {
|
|
2
3
|
maxRetry: 3,
|
|
3
4
|
retryIntervalMs: 2e3,
|
|
@@ -18,6 +19,7 @@ const DEFAULT_SESSION_CONNECT_OPTIONS = {
|
|
|
18
19
|
export {
|
|
19
20
|
DEFAULT_API_CONNECT_OPTIONS,
|
|
20
21
|
DEFAULT_SESSION_CONNECT_OPTIONS,
|
|
22
|
+
USERDATA_TIMED_TRANSCRIPT,
|
|
21
23
|
intervalForRetry
|
|
22
24
|
};
|
|
23
25
|
//# sourceMappingURL=types.js.map
|
package/dist/types.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/types.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\n/**\n * Connection options for API calls, controlling retry and timeout behavior.\n */\nexport interface APIConnectOptions {\n /** Maximum number of retries to connect to the API. Default: 3 */\n maxRetry: number;\n /** Interval between retries to connect to the API in milliseconds. Default: 2000 */\n retryIntervalMs: number;\n /** Timeout for connecting to the API in milliseconds. Default: 10000 */\n timeoutMs: number;\n}\n\nexport const DEFAULT_API_CONNECT_OPTIONS: APIConnectOptions = {\n maxRetry: 3,\n retryIntervalMs: 2000,\n timeoutMs: 10000,\n};\n\n/**\n * Return the interval for the given number of retries.\n * The first retry is immediate, and then uses specified retryIntervalMs.\n * @internal\n */\nexport function intervalForRetry(connOptions: APIConnectOptions, numRetries: number): number {\n if (numRetries === 0) {\n return 0.1;\n }\n return connOptions.retryIntervalMs;\n}\n\n/**\n * Connection options for the agent session, controlling retry and timeout behavior\n * for STT, LLM, and TTS connections.\n */\nexport interface SessionConnectOptions {\n /** Connection options for speech-to-text. */\n sttConnOptions?: Partial<APIConnectOptions>;\n /** Connection options for the language model. */\n llmConnOptions?: Partial<APIConnectOptions>;\n /** Connection options for text-to-speech. */\n ttsConnOptions?: Partial<APIConnectOptions>;\n /** Maximum number of consecutive unrecoverable errors from LLM or TTS before closing the session. Default: 3 */\n maxUnrecoverableErrors?: number;\n}\n\n/**\n * Resolved session connect options with all values populated.\n * @internal\n */\nexport interface ResolvedSessionConnectOptions {\n sttConnOptions: APIConnectOptions;\n llmConnOptions: APIConnectOptions;\n ttsConnOptions: APIConnectOptions;\n maxUnrecoverableErrors: number;\n}\n\nexport const DEFAULT_SESSION_CONNECT_OPTIONS: ResolvedSessionConnectOptions = {\n sttConnOptions: DEFAULT_API_CONNECT_OPTIONS,\n llmConnOptions: DEFAULT_API_CONNECT_OPTIONS,\n ttsConnOptions: DEFAULT_API_CONNECT_OPTIONS,\n maxUnrecoverableErrors: 3,\n};\n"],"mappings":"
|
|
1
|
+
{"version":3,"sources":["../src/types.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\n/**\n * Key used to store timed transcripts in AudioFrame.userdata.\n */\nexport const USERDATA_TIMED_TRANSCRIPT = 'lk.timed_transcripts';\n\n/**\n * Connection options for API calls, controlling retry and timeout behavior.\n */\nexport interface APIConnectOptions {\n /** Maximum number of retries to connect to the API. Default: 3 */\n maxRetry: number;\n /** Interval between retries to connect to the API in milliseconds. Default: 2000 */\n retryIntervalMs: number;\n /** Timeout for connecting to the API in milliseconds. Default: 10000 */\n timeoutMs: number;\n}\n\nexport const DEFAULT_API_CONNECT_OPTIONS: APIConnectOptions = {\n maxRetry: 3,\n retryIntervalMs: 2000,\n timeoutMs: 10000,\n};\n\n/**\n * Return the interval for the given number of retries.\n * The first retry is immediate, and then uses specified retryIntervalMs.\n * @internal\n */\nexport function intervalForRetry(connOptions: APIConnectOptions, numRetries: number): number {\n if (numRetries === 0) {\n return 0.1;\n }\n return connOptions.retryIntervalMs;\n}\n\n/**\n * Connection options for the agent session, controlling retry and timeout behavior\n * for STT, LLM, and TTS connections.\n */\nexport interface SessionConnectOptions {\n /** Connection options for speech-to-text. */\n sttConnOptions?: Partial<APIConnectOptions>;\n /** Connection options for the language model. */\n llmConnOptions?: Partial<APIConnectOptions>;\n /** Connection options for text-to-speech. */\n ttsConnOptions?: Partial<APIConnectOptions>;\n /** Maximum number of consecutive unrecoverable errors from LLM or TTS before closing the session. Default: 3 */\n maxUnrecoverableErrors?: number;\n}\n\n/**\n * Resolved session connect options with all values populated.\n * @internal\n */\nexport interface ResolvedSessionConnectOptions {\n sttConnOptions: APIConnectOptions;\n llmConnOptions: APIConnectOptions;\n ttsConnOptions: APIConnectOptions;\n maxUnrecoverableErrors: number;\n}\n\nexport const DEFAULT_SESSION_CONNECT_OPTIONS: ResolvedSessionConnectOptions = {\n sttConnOptions: DEFAULT_API_CONNECT_OPTIONS,\n llmConnOptions: DEFAULT_API_CONNECT_OPTIONS,\n ttsConnOptions: DEFAULT_API_CONNECT_OPTIONS,\n maxUnrecoverableErrors: 3,\n};\n"],"mappings":"AAOO,MAAM,4BAA4B;AAclC,MAAM,8BAAiD;AAAA,EAC5D,UAAU;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AACb;AAOO,SAAS,iBAAiB,aAAgC,YAA4B;AAC3F,MAAI,eAAe,GAAG;AACpB,WAAO;AAAA,EACT;AACA,SAAO,YAAY;AACrB;AA4BO,MAAM,kCAAiE;AAAA,EAC5E,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,wBAAwB;AAC1B;","names":[]}
|
package/dist/voice/agent.cjs
CHANGED
|
@@ -33,6 +33,7 @@ var import_llm = require("../llm/index.cjs");
|
|
|
33
33
|
var import_stt = require("../stt/index.cjs");
|
|
34
34
|
var import_basic = require("../tokenize/basic/index.cjs");
|
|
35
35
|
var import_tts = require("../tts/index.cjs");
|
|
36
|
+
var import_types = require("../types.cjs");
|
|
36
37
|
const asyncLocalStorage = new import_node_async_hooks.AsyncLocalStorage();
|
|
37
38
|
const STOP_RESPONSE_SYMBOL = Symbol("StopResponse");
|
|
38
39
|
class StopResponse extends Error {
|
|
@@ -54,6 +55,7 @@ class Agent {
|
|
|
54
55
|
_vad;
|
|
55
56
|
_llm;
|
|
56
57
|
_tts;
|
|
58
|
+
_useTtsAlignedTranscript;
|
|
57
59
|
/** @internal */
|
|
58
60
|
_agentActivity;
|
|
59
61
|
/** @internal */
|
|
@@ -71,7 +73,8 @@ class Agent {
|
|
|
71
73
|
stt,
|
|
72
74
|
vad,
|
|
73
75
|
llm,
|
|
74
|
-
tts
|
|
76
|
+
tts,
|
|
77
|
+
useTtsAlignedTranscript
|
|
75
78
|
}) {
|
|
76
79
|
if (id) {
|
|
77
80
|
this._id = id;
|
|
@@ -105,6 +108,7 @@ class Agent {
|
|
|
105
108
|
} else {
|
|
106
109
|
this._tts = tts;
|
|
107
110
|
}
|
|
111
|
+
this._useTtsAlignedTranscript = useTtsAlignedTranscript;
|
|
108
112
|
this._agentActivity = void 0;
|
|
109
113
|
}
|
|
110
114
|
get vad() {
|
|
@@ -119,6 +123,9 @@ class Agent {
|
|
|
119
123
|
get tts() {
|
|
120
124
|
return this._tts;
|
|
121
125
|
}
|
|
126
|
+
get useTtsAlignedTranscript() {
|
|
127
|
+
return this._useTtsAlignedTranscript;
|
|
128
|
+
}
|
|
122
129
|
get chatCtx() {
|
|
123
130
|
return new import_chat_context.ReadonlyChatContext(this._chatCtx.items);
|
|
124
131
|
}
|
|
@@ -281,6 +288,9 @@ class Agent {
|
|
|
281
288
|
if (chunk === import_tts.SynthesizeStream.END_OF_STREAM) {
|
|
282
289
|
break;
|
|
283
290
|
}
|
|
291
|
+
if (chunk.timedTranscripts && chunk.timedTranscripts.length > 0) {
|
|
292
|
+
chunk.frame.userdata[import_types.USERDATA_TIMED_TRANSCRIPT] = chunk.timedTranscripts;
|
|
293
|
+
}
|
|
284
294
|
controller.enqueue(chunk.frame);
|
|
285
295
|
}
|
|
286
296
|
controller.close();
|
package/dist/voice/agent.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/voice/agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { AsyncLocalStorage } from 'node:async_hooks';\nimport { ReadableStream } from 'node:stream/web';\nimport {\n LLM as InferenceLLM,\n STT as InferenceSTT,\n TTS as InferenceTTS,\n type LLMModels,\n type STTModelString,\n type TTSModelString,\n} from '../inference/index.js';\nimport { ReadonlyChatContext } from '../llm/chat_context.js';\nimport type { ChatMessage, FunctionCall, RealtimeModel } from '../llm/index.js';\nimport {\n type ChatChunk,\n ChatContext,\n LLM,\n type ToolChoice,\n type ToolContext,\n} from '../llm/index.js';\nimport type { STT, SpeechEvent } from '../stt/index.js';\nimport { StreamAdapter as STTStreamAdapter } from '../stt/index.js';\nimport { SentenceTokenizer as BasicSentenceTokenizer } from '../tokenize/basic/index.js';\nimport type { TTS } from '../tts/index.js';\nimport { SynthesizeStream, StreamAdapter as TTSStreamAdapter } from '../tts/index.js';\nimport type { VAD } from '../vad.js';\nimport type { AgentActivity } from './agent_activity.js';\nimport type { AgentSession, TurnDetectionMode } from './agent_session.js';\n\nexport const asyncLocalStorage = new AsyncLocalStorage<{ functionCall?: FunctionCall }>();\nexport const STOP_RESPONSE_SYMBOL = Symbol('StopResponse');\n\nexport class StopResponse extends Error {\n constructor() {\n super();\n this.name = 'StopResponse';\n\n Object.defineProperty(this, STOP_RESPONSE_SYMBOL, {\n value: true,\n });\n }\n}\n\nexport function isStopResponse(value: unknown): value is StopResponse {\n return (\n value !== undefined &&\n value !== null &&\n typeof value === 'object' &&\n STOP_RESPONSE_SYMBOL in value\n );\n}\n\nexport interface ModelSettings {\n /** The tool choice to use when calling the LLM. */\n toolChoice?: ToolChoice;\n}\n\nexport interface AgentOptions<UserData> {\n id?: string;\n instructions: string;\n chatCtx?: ChatContext;\n tools?: ToolContext<UserData>;\n turnDetection?: TurnDetectionMode;\n stt?: STT | STTModelString;\n vad?: VAD;\n llm?: LLM | RealtimeModel | LLMModels;\n tts?: TTS | TTSModelString;\n allowInterruptions?: boolean;\n minConsecutiveSpeechDelay?: number;\n}\n\nexport class Agent<UserData = any> {\n private _id: string;\n private turnDetection?: TurnDetectionMode;\n private _stt?: STT;\n private _vad?: VAD;\n private _llm?: LLM | RealtimeModel;\n private _tts?: TTS;\n\n /** @internal */\n _agentActivity?: AgentActivity;\n\n /** @internal */\n _chatCtx: ChatContext;\n\n /** @internal */\n _instructions: string;\n\n /** @internal */\n _tools?: ToolContext<UserData>;\n\n constructor({\n id,\n instructions,\n chatCtx,\n tools,\n turnDetection,\n stt,\n vad,\n llm,\n tts,\n }: AgentOptions<UserData>) {\n if (id) {\n this._id = id;\n } else {\n // Convert class name to snake_case\n const className = this.constructor.name;\n if (className === 'Agent') {\n this._id = 'default_agent';\n } else {\n this._id = className\n .replace(/([A-Z])/g, '_$1')\n .toLowerCase()\n .replace(/^_/, '');\n }\n }\n\n this._instructions = instructions;\n this._tools = { ...tools };\n this._chatCtx = chatCtx\n ? chatCtx.copy({\n toolCtx: this._tools,\n })\n : ChatContext.empty();\n\n this.turnDetection = turnDetection;\n this._vad = vad;\n\n if (typeof stt === 'string') {\n this._stt = InferenceSTT.fromModelString(stt);\n } else {\n this._stt = stt;\n }\n\n if (typeof llm === 'string') {\n this._llm = InferenceLLM.fromModelString(llm);\n } else {\n this._llm = llm;\n }\n\n if (typeof tts === 'string') {\n this._tts = InferenceTTS.fromModelString(tts);\n } else {\n this._tts = tts;\n }\n\n this._agentActivity = undefined;\n }\n\n get vad(): VAD | undefined {\n return this._vad;\n }\n\n get stt(): STT | undefined {\n return this._stt;\n }\n\n get llm(): LLM | RealtimeModel | undefined {\n return this._llm;\n }\n\n get tts(): TTS | undefined {\n return this._tts;\n }\n\n get chatCtx(): ReadonlyChatContext {\n return new ReadonlyChatContext(this._chatCtx.items);\n }\n\n get id(): string {\n return this._id;\n }\n\n get instructions(): string {\n return this._instructions;\n }\n\n get toolCtx(): ToolContext<UserData> {\n return { ...this._tools };\n }\n\n get session(): AgentSession<UserData> {\n return this.getActivityOrThrow().agentSession as AgentSession<UserData>;\n }\n\n async onEnter(): Promise<void> {}\n\n async onExit(): Promise<void> {}\n\n async transcriptionNode(\n text: ReadableStream<string>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<string> | null> {\n return Agent.default.transcriptionNode(this, text, modelSettings);\n }\n\n async onUserTurnCompleted(_chatCtx: ChatContext, _newMessage: ChatMessage): Promise<void> {}\n\n async sttNode(\n audio: ReadableStream<AudioFrame>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<SpeechEvent | string> | null> {\n return Agent.default.sttNode(this, audio, modelSettings);\n }\n\n async llmNode(\n chatCtx: ChatContext,\n toolCtx: ToolContext,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<ChatChunk | string> | null> {\n return Agent.default.llmNode(this, chatCtx, toolCtx, modelSettings);\n }\n\n async ttsNode(\n text: ReadableStream<string>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return Agent.default.ttsNode(this, text, modelSettings);\n }\n\n async realtimeAudioOutputNode(\n audio: ReadableStream<AudioFrame>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return Agent.default.realtimeAudioOutputNode(this, audio, modelSettings);\n }\n\n // realtime_audio_output_node\n\n getActivityOrThrow(): AgentActivity {\n if (!this._agentActivity) {\n throw new Error('Agent activity not found');\n }\n return this._agentActivity;\n }\n\n async updateChatCtx(chatCtx: ChatContext): Promise<void> {\n if (!this._agentActivity) {\n this._chatCtx = chatCtx.copy({ toolCtx: this.toolCtx });\n return;\n }\n\n this._agentActivity.updateChatCtx(chatCtx);\n }\n\n static default = {\n async sttNode(\n agent: Agent,\n audio: ReadableStream<AudioFrame>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<SpeechEvent | string> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.stt) {\n throw new Error('sttNode called but no STT node is available');\n }\n\n let wrapped_stt = activity.stt;\n\n if (!wrapped_stt.capabilities.streaming) {\n const vad = agent.vad || activity.vad;\n if (!vad) {\n throw new Error(\n 'STT does not support streaming, add a VAD to the AgentTask/VoiceAgent to enable streaming',\n );\n }\n wrapped_stt = new STTStreamAdapter(wrapped_stt, vad);\n }\n\n const connOptions = activity.agentSession.connOptions.sttConnOptions;\n const stream = wrapped_stt.stream({ connOptions });\n\n // Set startTimeOffset to provide linear timestamps across reconnections\n const audioInputStartedAt =\n activity.agentSession._recorderIO?.recordingStartedAt ?? // Use recording start time if available\n activity.agentSession._startedAt ?? // Fallback to session start time\n Date.now(); // Fallback to current time\n\n stream.startTimeOffset = (Date.now() - audioInputStartedAt) / 1000;\n\n stream.updateInputStream(audio);\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.detachInputStream();\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const event of stream) {\n controller.enqueue(event);\n }\n controller.close();\n } finally {\n // Always clean up the STT stream, whether it ends naturally or is cancelled\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async llmNode(\n agent: Agent,\n chatCtx: ChatContext,\n toolCtx: ToolContext,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<ChatChunk | string> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.llm) {\n throw new Error('llmNode called but no LLM node is available');\n }\n\n if (!(activity.llm instanceof LLM)) {\n throw new Error(\n 'llmNode should only be used with LLM (non-multimodal/realtime APIs) nodes',\n );\n }\n\n const { toolChoice } = modelSettings;\n const connOptions = activity.agentSession.connOptions.llmConnOptions;\n\n // parallelToolCalls is not passed here - it will use the value from LLM's modelOptions\n // This allows users to configure it via: new inference.LLM({ modelOptions: { parallel_tool_calls: false } })\n const stream = activity.llm.chat({\n chatCtx,\n toolCtx,\n toolChoice,\n connOptions,\n });\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const chunk of stream) {\n controller.enqueue(chunk);\n }\n controller.close();\n } finally {\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async ttsNode(\n agent: Agent,\n text: ReadableStream<string>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.tts) {\n throw new Error('ttsNode called but no TTS node is available');\n }\n\n let wrapped_tts = activity.tts;\n\n if (!activity.tts.capabilities.streaming) {\n wrapped_tts = new TTSStreamAdapter(wrapped_tts, new BasicSentenceTokenizer());\n }\n\n const connOptions = activity.agentSession.connOptions.ttsConnOptions;\n const stream = wrapped_tts.stream({ connOptions });\n stream.updateInputStream(text);\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const chunk of stream) {\n if (chunk === SynthesizeStream.END_OF_STREAM) {\n break;\n }\n controller.enqueue(chunk.frame);\n }\n controller.close();\n } finally {\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async transcriptionNode(\n agent: Agent,\n text: ReadableStream<string>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<string> | null> {\n return text;\n },\n\n async realtimeAudioOutputNode(\n _agent: Agent,\n audio: ReadableStream<AudioFrame>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return audio;\n },\n };\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,8BAAkC;AAClC,iBAA+B;AAC/B,uBAOO;AACP,0BAAoC;AAEpC,iBAMO;AAEP,iBAAkD;AAClD,mBAA4D;AAE5D,iBAAoE;AAK7D,MAAM,oBAAoB,IAAI,0CAAmD;AACjF,MAAM,uBAAuB,OAAO,cAAc;AAElD,MAAM,qBAAqB,MAAM;AAAA,EACtC,cAAc;AACZ,UAAM;AACN,SAAK,OAAO;AAEZ,WAAO,eAAe,MAAM,sBAAsB;AAAA,MAChD,OAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AAEO,SAAS,eAAe,OAAuC;AACpE,SACE,UAAU,UACV,UAAU,QACV,OAAO,UAAU,YACjB,wBAAwB;AAE5B;AAqBO,MAAM,MAAsB;AAAA,EACzB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGR;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAA2B;AACzB,QAAI,IAAI;AACN,WAAK,MAAM;AAAA,IACb,OAAO;AAEL,YAAM,YAAY,KAAK,YAAY;AACnC,UAAI,cAAc,SAAS;AACzB,aAAK,MAAM;AAAA,MACb,OAAO;AACL,aAAK,MAAM,UACR,QAAQ,YAAY,KAAK,EACzB,YAAY,EACZ,QAAQ,MAAM,EAAE;AAAA,MACrB;AAAA,IACF;AAEA,SAAK,gBAAgB;AACrB,SAAK,SAAS,EAAE,GAAG,MAAM;AACzB,SAAK,WAAW,UACZ,QAAQ,KAAK;AAAA,MACX,SAAS,KAAK;AAAA,IAChB,CAAC,IACD,uBAAY,MAAM;AAEtB,SAAK,gBAAgB;AACrB,SAAK,OAAO;AAEZ,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,iBAAAA,IAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,SAAK,iBAAiB;AAAA,EACxB;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuC;AACzC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAA+B;AACjC,WAAO,IAAI,wCAAoB,KAAK,SAAS,KAAK;AAAA,EACpD;AAAA,EAEA,IAAI,KAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAiC;AACnC,WAAO,EAAE,GAAG,KAAK,OAAO;AAAA,EAC1B;AAAA,EAEA,IAAI,UAAkC;AACpC,WAAO,KAAK,mBAAmB,EAAE;AAAA,EACnC;AAAA,EAEA,MAAM,UAAyB;AAAA,EAAC;AAAA,EAEhC,MAAM,SAAwB;AAAA,EAAC;AAAA,EAE/B,MAAM,kBACJ,MACA,eACwC;AACxC,WAAO,MAAM,QAAQ,kBAAkB,MAAM,MAAM,aAAa;AAAA,EAClE;AAAA,EAEA,MAAM,oBAAoB,UAAuB,aAAyC;AAAA,EAAC;AAAA,EAE3F,MAAM,QACJ,OACA,eACsD;AACtD,WAAO,MAAM,QAAQ,QAAQ,MAAM,OAAO,aAAa;AAAA,EACzD;AAAA,EAEA,MAAM,QACJ,SACA,SACA,eACoD;AACpD,WAAO,MAAM,QAAQ,QAAQ,MAAM,SAAS,SAAS,aAAa;AAAA,EACpE;AAAA,EAEA,MAAM,QACJ,MACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,QAAQ,MAAM,MAAM,aAAa;AAAA,EACxD;AAAA,EAEA,MAAM,wBACJ,OACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,wBAAwB,MAAM,OAAO,aAAa;AAAA,EACzE;AAAA;AAAA,EAIA,qBAAoC;AAClC,QAAI,CAAC,KAAK,gBAAgB;AACxB,YAAM,IAAI,MAAM,0BAA0B;AAAA,IAC5C;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,cAAc,SAAqC;AACvD,QAAI,CAAC,KAAK,gBAAgB;AACxB,WAAK,WAAW,QAAQ,KAAK,EAAE,SAAS,KAAK,QAAQ,CAAC;AACtD;AAAA,IACF;AAEA,SAAK,eAAe,cAAc,OAAO;AAAA,EAC3C;AAAA,EAEA,OAAO,UAAU;AAAA,IACf,MAAM,QACJ,OACA,OACA,gBACsD;AA7P5D;AA8PM,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,cAAc,SAAS;AAE3B,UAAI,CAAC,YAAY,aAAa,WAAW;AACvC,cAAM,MAAM,MAAM,OAAO,SAAS;AAClC,YAAI,CAAC,KAAK;AACR,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,sBAAc,IAAI,WAAAC,cAAiB,aAAa,GAAG;AAAA,MACrD;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,YAAY,OAAO,EAAE,YAAY,CAAC;AAGjD,YAAM,wBACJ,cAAS,aAAa,gBAAtB,mBAAmC;AAAA,MACnC,SAAS,aAAa;AAAA,MACtB,KAAK,IAAI;AAEX,aAAO,mBAAmB,KAAK,IAAI,IAAI,uBAAuB;AAE9D,aAAO,kBAAkB,KAAK;AAE9B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,kBAAkB;AACzB,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AAEA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,SACA,SACA,eACoD;AACpD,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,EAAE,SAAS,eAAe,iBAAM;AAClC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAEA,YAAM,EAAE,WAAW,IAAI;AACvB,YAAM,cAAc,SAAS,aAAa,YAAY;AAItD,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,CAAC;AAED,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,MACA,gBAC4C;AAC5C,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,cAAc,SAAS;AAE3B,UAAI,CAAC,SAAS,IAAI,aAAa,WAAW;AACxC,sBAAc,IAAI,WAAAC,cAAiB,aAAa,IAAI,aAAAC,kBAAuB,CAAC;AAAA,MAC9E;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,YAAY,OAAO,EAAE,YAAY,CAAC;AACjD,aAAO,kBAAkB,IAAI;AAE7B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,kBAAI,UAAU,4BAAiB,eAAe;AAC5C;AAAA,cACF;AACA,yBAAW,QAAQ,MAAM,KAAK;AAAA,YAChC;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,kBACJ,OACA,MACA,gBACwC;AACxC,aAAO;AAAA,IACT;AAAA,IAEA,MAAM,wBACJ,QACA,OACA,gBAC4C;AAC5C,aAAO;AAAA,IACT;AAAA,EACF;AACF;","names":["InferenceSTT","InferenceLLM","InferenceTTS","STTStreamAdapter","TTSStreamAdapter","BasicSentenceTokenizer"]}
|
|
1
|
+
{"version":3,"sources":["../../src/voice/agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { AsyncLocalStorage } from 'node:async_hooks';\nimport { ReadableStream } from 'node:stream/web';\nimport {\n LLM as InferenceLLM,\n STT as InferenceSTT,\n TTS as InferenceTTS,\n type LLMModels,\n type STTModelString,\n type TTSModelString,\n} from '../inference/index.js';\nimport { ReadonlyChatContext } from '../llm/chat_context.js';\nimport type { ChatMessage, FunctionCall, RealtimeModel } from '../llm/index.js';\nimport {\n type ChatChunk,\n ChatContext,\n LLM,\n type ToolChoice,\n type ToolContext,\n} from '../llm/index.js';\nimport type { STT, SpeechEvent } from '../stt/index.js';\nimport { StreamAdapter as STTStreamAdapter } from '../stt/index.js';\nimport { SentenceTokenizer as BasicSentenceTokenizer } from '../tokenize/basic/index.js';\nimport type { TTS } from '../tts/index.js';\nimport { SynthesizeStream, StreamAdapter as TTSStreamAdapter } from '../tts/index.js';\nimport { USERDATA_TIMED_TRANSCRIPT } from '../types.js';\nimport type { VAD } from '../vad.js';\nimport type { AgentActivity } from './agent_activity.js';\nimport type { AgentSession, TurnDetectionMode } from './agent_session.js';\nimport type { TimedString } from './io.js';\n\nexport const asyncLocalStorage = new AsyncLocalStorage<{ functionCall?: FunctionCall }>();\nexport const STOP_RESPONSE_SYMBOL = Symbol('StopResponse');\n\nexport class StopResponse extends Error {\n constructor() {\n super();\n this.name = 'StopResponse';\n\n Object.defineProperty(this, STOP_RESPONSE_SYMBOL, {\n value: true,\n });\n }\n}\n\nexport function isStopResponse(value: unknown): value is StopResponse {\n return (\n value !== undefined &&\n value !== null &&\n typeof value === 'object' &&\n STOP_RESPONSE_SYMBOL in value\n );\n}\n\nexport interface ModelSettings {\n /** The tool choice to use when calling the LLM. */\n toolChoice?: ToolChoice;\n}\n\nexport interface AgentOptions<UserData> {\n id?: string;\n instructions: string;\n chatCtx?: ChatContext;\n tools?: ToolContext<UserData>;\n turnDetection?: TurnDetectionMode;\n stt?: STT | STTModelString;\n vad?: VAD;\n llm?: LLM | RealtimeModel | LLMModels;\n tts?: TTS | TTSModelString;\n allowInterruptions?: boolean;\n minConsecutiveSpeechDelay?: number;\n useTtsAlignedTranscript?: boolean;\n}\n\nexport class Agent<UserData = any> {\n private _id: string;\n private turnDetection?: TurnDetectionMode;\n private _stt?: STT;\n private _vad?: VAD;\n private _llm?: LLM | RealtimeModel;\n private _tts?: TTS;\n private _useTtsAlignedTranscript?: boolean;\n\n /** @internal */\n _agentActivity?: AgentActivity;\n\n /** @internal */\n _chatCtx: ChatContext;\n\n /** @internal */\n _instructions: string;\n\n /** @internal */\n _tools?: ToolContext<UserData>;\n\n constructor({\n id,\n instructions,\n chatCtx,\n tools,\n turnDetection,\n stt,\n vad,\n llm,\n tts,\n useTtsAlignedTranscript,\n }: AgentOptions<UserData>) {\n if (id) {\n this._id = id;\n } else {\n // Convert class name to snake_case\n const className = this.constructor.name;\n if (className === 'Agent') {\n this._id = 'default_agent';\n } else {\n this._id = className\n .replace(/([A-Z])/g, '_$1')\n .toLowerCase()\n .replace(/^_/, '');\n }\n }\n\n this._instructions = instructions;\n this._tools = { ...tools };\n this._chatCtx = chatCtx\n ? chatCtx.copy({\n toolCtx: this._tools,\n })\n : ChatContext.empty();\n\n this.turnDetection = turnDetection;\n this._vad = vad;\n\n if (typeof stt === 'string') {\n this._stt = InferenceSTT.fromModelString(stt);\n } else {\n this._stt = stt;\n }\n\n if (typeof llm === 'string') {\n this._llm = InferenceLLM.fromModelString(llm);\n } else {\n this._llm = llm;\n }\n\n if (typeof tts === 'string') {\n this._tts = InferenceTTS.fromModelString(tts);\n } else {\n this._tts = tts;\n }\n\n this._useTtsAlignedTranscript = useTtsAlignedTranscript;\n\n this._agentActivity = undefined;\n }\n\n get vad(): VAD | undefined {\n return this._vad;\n }\n\n get stt(): STT | undefined {\n return this._stt;\n }\n\n get llm(): LLM | RealtimeModel | undefined {\n return this._llm;\n }\n\n get tts(): TTS | undefined {\n return this._tts;\n }\n\n get useTtsAlignedTranscript(): boolean | undefined {\n return this._useTtsAlignedTranscript;\n }\n\n get chatCtx(): ReadonlyChatContext {\n return new ReadonlyChatContext(this._chatCtx.items);\n }\n\n get id(): string {\n return this._id;\n }\n\n get instructions(): string {\n return this._instructions;\n }\n\n get toolCtx(): ToolContext<UserData> {\n return { ...this._tools };\n }\n\n get session(): AgentSession<UserData> {\n return this.getActivityOrThrow().agentSession as AgentSession<UserData>;\n }\n\n async onEnter(): Promise<void> {}\n\n async onExit(): Promise<void> {}\n\n async transcriptionNode(\n text: ReadableStream<string | TimedString>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<string | TimedString> | null> {\n return Agent.default.transcriptionNode(this, text, modelSettings);\n }\n\n async onUserTurnCompleted(_chatCtx: ChatContext, _newMessage: ChatMessage): Promise<void> {}\n\n async sttNode(\n audio: ReadableStream<AudioFrame>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<SpeechEvent | string> | null> {\n return Agent.default.sttNode(this, audio, modelSettings);\n }\n\n async llmNode(\n chatCtx: ChatContext,\n toolCtx: ToolContext,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<ChatChunk | string> | null> {\n return Agent.default.llmNode(this, chatCtx, toolCtx, modelSettings);\n }\n\n async ttsNode(\n text: ReadableStream<string>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return Agent.default.ttsNode(this, text, modelSettings);\n }\n\n async realtimeAudioOutputNode(\n audio: ReadableStream<AudioFrame>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return Agent.default.realtimeAudioOutputNode(this, audio, modelSettings);\n }\n\n // realtime_audio_output_node\n\n getActivityOrThrow(): AgentActivity {\n if (!this._agentActivity) {\n throw new Error('Agent activity not found');\n }\n return this._agentActivity;\n }\n\n async updateChatCtx(chatCtx: ChatContext): Promise<void> {\n if (!this._agentActivity) {\n this._chatCtx = chatCtx.copy({ toolCtx: this.toolCtx });\n return;\n }\n\n this._agentActivity.updateChatCtx(chatCtx);\n }\n\n static default = {\n async sttNode(\n agent: Agent,\n audio: ReadableStream<AudioFrame>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<SpeechEvent | string> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.stt) {\n throw new Error('sttNode called but no STT node is available');\n }\n\n let wrapped_stt = activity.stt;\n\n if (!wrapped_stt.capabilities.streaming) {\n const vad = agent.vad || activity.vad;\n if (!vad) {\n throw new Error(\n 'STT does not support streaming, add a VAD to the AgentTask/VoiceAgent to enable streaming',\n );\n }\n wrapped_stt = new STTStreamAdapter(wrapped_stt, vad);\n }\n\n const connOptions = activity.agentSession.connOptions.sttConnOptions;\n const stream = wrapped_stt.stream({ connOptions });\n\n // Set startTimeOffset to provide linear timestamps across reconnections\n const audioInputStartedAt =\n activity.agentSession._recorderIO?.recordingStartedAt ?? // Use recording start time if available\n activity.agentSession._startedAt ?? // Fallback to session start time\n Date.now(); // Fallback to current time\n\n stream.startTimeOffset = (Date.now() - audioInputStartedAt) / 1000;\n\n stream.updateInputStream(audio);\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.detachInputStream();\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const event of stream) {\n controller.enqueue(event);\n }\n controller.close();\n } finally {\n // Always clean up the STT stream, whether it ends naturally or is cancelled\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async llmNode(\n agent: Agent,\n chatCtx: ChatContext,\n toolCtx: ToolContext,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<ChatChunk | string> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.llm) {\n throw new Error('llmNode called but no LLM node is available');\n }\n\n if (!(activity.llm instanceof LLM)) {\n throw new Error(\n 'llmNode should only be used with LLM (non-multimodal/realtime APIs) nodes',\n );\n }\n\n const { toolChoice } = modelSettings;\n const connOptions = activity.agentSession.connOptions.llmConnOptions;\n\n // parallelToolCalls is not passed here - it will use the value from LLM's modelOptions\n // This allows users to configure it via: new inference.LLM({ modelOptions: { parallel_tool_calls: false } })\n const stream = activity.llm.chat({\n chatCtx,\n toolCtx,\n toolChoice,\n connOptions,\n });\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const chunk of stream) {\n controller.enqueue(chunk);\n }\n controller.close();\n } finally {\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async ttsNode(\n agent: Agent,\n text: ReadableStream<string>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.tts) {\n throw new Error('ttsNode called but no TTS node is available');\n }\n\n let wrapped_tts = activity.tts;\n\n if (!activity.tts.capabilities.streaming) {\n wrapped_tts = new TTSStreamAdapter(wrapped_tts, new BasicSentenceTokenizer());\n }\n\n const connOptions = activity.agentSession.connOptions.ttsConnOptions;\n const stream = wrapped_tts.stream({ connOptions });\n stream.updateInputStream(text);\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const chunk of stream) {\n if (chunk === SynthesizeStream.END_OF_STREAM) {\n break;\n }\n // Attach timed transcripts to frame.userdata\n if (chunk.timedTranscripts && chunk.timedTranscripts.length > 0) {\n chunk.frame.userdata[USERDATA_TIMED_TRANSCRIPT] = chunk.timedTranscripts;\n }\n controller.enqueue(chunk.frame);\n }\n controller.close();\n } finally {\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async transcriptionNode(\n agent: Agent,\n text: ReadableStream<string | TimedString>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<string | TimedString> | null> {\n return text;\n },\n\n async realtimeAudioOutputNode(\n _agent: Agent,\n audio: ReadableStream<AudioFrame>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return audio;\n },\n };\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,8BAAkC;AAClC,iBAA+B;AAC/B,uBAOO;AACP,0BAAoC;AAEpC,iBAMO;AAEP,iBAAkD;AAClD,mBAA4D;AAE5D,iBAAoE;AACpE,mBAA0C;AAMnC,MAAM,oBAAoB,IAAI,0CAAmD;AACjF,MAAM,uBAAuB,OAAO,cAAc;AAElD,MAAM,qBAAqB,MAAM;AAAA,EACtC,cAAc;AACZ,UAAM;AACN,SAAK,OAAO;AAEZ,WAAO,eAAe,MAAM,sBAAsB;AAAA,MAChD,OAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AAEO,SAAS,eAAe,OAAuC;AACpE,SACE,UAAU,UACV,UAAU,QACV,OAAO,UAAU,YACjB,wBAAwB;AAE5B;AAsBO,MAAM,MAAsB;AAAA,EACzB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGR;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAA2B;AACzB,QAAI,IAAI;AACN,WAAK,MAAM;AAAA,IACb,OAAO;AAEL,YAAM,YAAY,KAAK,YAAY;AACnC,UAAI,cAAc,SAAS;AACzB,aAAK,MAAM;AAAA,MACb,OAAO;AACL,aAAK,MAAM,UACR,QAAQ,YAAY,KAAK,EACzB,YAAY,EACZ,QAAQ,MAAM,EAAE;AAAA,MACrB;AAAA,IACF;AAEA,SAAK,gBAAgB;AACrB,SAAK,SAAS,EAAE,GAAG,MAAM;AACzB,SAAK,WAAW,UACZ,QAAQ,KAAK;AAAA,MACX,SAAS,KAAK;AAAA,IAChB,CAAC,IACD,uBAAY,MAAM;AAEtB,SAAK,gBAAgB;AACrB,SAAK,OAAO;AAEZ,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,iBAAAA,IAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,SAAK,2BAA2B;AAEhC,SAAK,iBAAiB;AAAA,EACxB;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuC;AACzC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,0BAA+C;AACjD,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAA+B;AACjC,WAAO,IAAI,wCAAoB,KAAK,SAAS,KAAK;AAAA,EACpD;AAAA,EAEA,IAAI,KAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAiC;AACnC,WAAO,EAAE,GAAG,KAAK,OAAO;AAAA,EAC1B;AAAA,EAEA,IAAI,UAAkC;AACpC,WAAO,KAAK,mBAAmB,EAAE;AAAA,EACnC;AAAA,EAEA,MAAM,UAAyB;AAAA,EAAC;AAAA,EAEhC,MAAM,SAAwB;AAAA,EAAC;AAAA,EAE/B,MAAM,kBACJ,MACA,eACsD;AACtD,WAAO,MAAM,QAAQ,kBAAkB,MAAM,MAAM,aAAa;AAAA,EAClE;AAAA,EAEA,MAAM,oBAAoB,UAAuB,aAAyC;AAAA,EAAC;AAAA,EAE3F,MAAM,QACJ,OACA,eACsD;AACtD,WAAO,MAAM,QAAQ,QAAQ,MAAM,OAAO,aAAa;AAAA,EACzD;AAAA,EAEA,MAAM,QACJ,SACA,SACA,eACoD;AACpD,WAAO,MAAM,QAAQ,QAAQ,MAAM,SAAS,SAAS,aAAa;AAAA,EACpE;AAAA,EAEA,MAAM,QACJ,MACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,QAAQ,MAAM,MAAM,aAAa;AAAA,EACxD;AAAA,EAEA,MAAM,wBACJ,OACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,wBAAwB,MAAM,OAAO,aAAa;AAAA,EACzE;AAAA;AAAA,EAIA,qBAAoC;AAClC,QAAI,CAAC,KAAK,gBAAgB;AACxB,YAAM,IAAI,MAAM,0BAA0B;AAAA,IAC5C;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,cAAc,SAAqC;AACvD,QAAI,CAAC,KAAK,gBAAgB;AACxB,WAAK,WAAW,QAAQ,KAAK,EAAE,SAAS,KAAK,QAAQ,CAAC;AACtD;AAAA,IACF;AAEA,SAAK,eAAe,cAAc,OAAO;AAAA,EAC3C;AAAA,EAEA,OAAO,UAAU;AAAA,IACf,MAAM,QACJ,OACA,OACA,gBACsD;AAxQ5D;AAyQM,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,cAAc,SAAS;AAE3B,UAAI,CAAC,YAAY,aAAa,WAAW;AACvC,cAAM,MAAM,MAAM,OAAO,SAAS;AAClC,YAAI,CAAC,KAAK;AACR,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,sBAAc,IAAI,WAAAC,cAAiB,aAAa,GAAG;AAAA,MACrD;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,YAAY,OAAO,EAAE,YAAY,CAAC;AAGjD,YAAM,wBACJ,cAAS,aAAa,gBAAtB,mBAAmC;AAAA,MACnC,SAAS,aAAa;AAAA,MACtB,KAAK,IAAI;AAEX,aAAO,mBAAmB,KAAK,IAAI,IAAI,uBAAuB;AAE9D,aAAO,kBAAkB,KAAK;AAE9B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,kBAAkB;AACzB,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AAEA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,SACA,SACA,eACoD;AACpD,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,EAAE,SAAS,eAAe,iBAAM;AAClC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAEA,YAAM,EAAE,WAAW,IAAI;AACvB,YAAM,cAAc,SAAS,aAAa,YAAY;AAItD,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,CAAC;AAED,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,MACA,gBAC4C;AAC5C,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,cAAc,SAAS;AAE3B,UAAI,CAAC,SAAS,IAAI,aAAa,WAAW;AACxC,sBAAc,IAAI,WAAAC,cAAiB,aAAa,IAAI,aAAAC,kBAAuB,CAAC;AAAA,MAC9E;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,YAAY,OAAO,EAAE,YAAY,CAAC;AACjD,aAAO,kBAAkB,IAAI;AAE7B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,kBAAI,UAAU,4BAAiB,eAAe;AAC5C;AAAA,cACF;AAEA,kBAAI,MAAM,oBAAoB,MAAM,iBAAiB,SAAS,GAAG;AAC/D,sBAAM,MAAM,SAAS,sCAAyB,IAAI,MAAM;AAAA,cAC1D;AACA,yBAAW,QAAQ,MAAM,KAAK;AAAA,YAChC;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,kBACJ,OACA,MACA,gBACsD;AACtD,aAAO;AAAA,IACT;AAAA,IAEA,MAAM,wBACJ,QACA,OACA,gBAC4C;AAC5C,aAAO;AAAA,IACT;AAAA,EACF;AACF;","names":["InferenceSTT","InferenceLLM","InferenceTTS","STTStreamAdapter","TTSStreamAdapter","BasicSentenceTokenizer"]}
|
package/dist/voice/agent.d.cts
CHANGED
|
@@ -12,6 +12,7 @@ import type { TTS } from '../tts/index.js';
|
|
|
12
12
|
import type { VAD } from '../vad.js';
|
|
13
13
|
import type { AgentActivity } from './agent_activity.js';
|
|
14
14
|
import type { AgentSession, TurnDetectionMode } from './agent_session.js';
|
|
15
|
+
import type { TimedString } from './io.js';
|
|
15
16
|
export declare const asyncLocalStorage: AsyncLocalStorage<{
|
|
16
17
|
functionCall?: FunctionCall | undefined;
|
|
17
18
|
}>;
|
|
@@ -36,6 +37,7 @@ export interface AgentOptions<UserData> {
|
|
|
36
37
|
tts?: TTS | TTSModelString;
|
|
37
38
|
allowInterruptions?: boolean;
|
|
38
39
|
minConsecutiveSpeechDelay?: number;
|
|
40
|
+
useTtsAlignedTranscript?: boolean;
|
|
39
41
|
}
|
|
40
42
|
export declare class Agent<UserData = any> {
|
|
41
43
|
private _id;
|
|
@@ -44,6 +46,7 @@ export declare class Agent<UserData = any> {
|
|
|
44
46
|
private _vad?;
|
|
45
47
|
private _llm?;
|
|
46
48
|
private _tts?;
|
|
49
|
+
private _useTtsAlignedTranscript?;
|
|
47
50
|
/** @internal */
|
|
48
51
|
_agentActivity?: AgentActivity;
|
|
49
52
|
/** @internal */
|
|
@@ -52,11 +55,12 @@ export declare class Agent<UserData = any> {
|
|
|
52
55
|
_instructions: string;
|
|
53
56
|
/** @internal */
|
|
54
57
|
_tools?: ToolContext<UserData>;
|
|
55
|
-
constructor({ id, instructions, chatCtx, tools, turnDetection, stt, vad, llm, tts, }: AgentOptions<UserData>);
|
|
58
|
+
constructor({ id, instructions, chatCtx, tools, turnDetection, stt, vad, llm, tts, useTtsAlignedTranscript, }: AgentOptions<UserData>);
|
|
56
59
|
get vad(): VAD | undefined;
|
|
57
60
|
get stt(): STT | undefined;
|
|
58
61
|
get llm(): LLM | RealtimeModel | undefined;
|
|
59
62
|
get tts(): TTS | undefined;
|
|
63
|
+
get useTtsAlignedTranscript(): boolean | undefined;
|
|
60
64
|
get chatCtx(): ReadonlyChatContext;
|
|
61
65
|
get id(): string;
|
|
62
66
|
get instructions(): string;
|
|
@@ -64,7 +68,7 @@ export declare class Agent<UserData = any> {
|
|
|
64
68
|
get session(): AgentSession<UserData>;
|
|
65
69
|
onEnter(): Promise<void>;
|
|
66
70
|
onExit(): Promise<void>;
|
|
67
|
-
transcriptionNode(text: ReadableStream<string>, modelSettings: ModelSettings): Promise<ReadableStream<string> | null>;
|
|
71
|
+
transcriptionNode(text: ReadableStream<string | TimedString>, modelSettings: ModelSettings): Promise<ReadableStream<string | TimedString> | null>;
|
|
68
72
|
onUserTurnCompleted(_chatCtx: ChatContext, _newMessage: ChatMessage): Promise<void>;
|
|
69
73
|
sttNode(audio: ReadableStream<AudioFrame>, modelSettings: ModelSettings): Promise<ReadableStream<SpeechEvent | string> | null>;
|
|
70
74
|
llmNode(chatCtx: ChatContext, toolCtx: ToolContext, modelSettings: ModelSettings): Promise<ReadableStream<ChatChunk | string> | null>;
|
|
@@ -76,7 +80,7 @@ export declare class Agent<UserData = any> {
|
|
|
76
80
|
sttNode(agent: Agent, audio: ReadableStream<AudioFrame>, _modelSettings: ModelSettings): Promise<ReadableStream<SpeechEvent | string> | null>;
|
|
77
81
|
llmNode(agent: Agent, chatCtx: ChatContext, toolCtx: ToolContext, modelSettings: ModelSettings): Promise<ReadableStream<ChatChunk | string> | null>;
|
|
78
82
|
ttsNode(agent: Agent, text: ReadableStream<string>, _modelSettings: ModelSettings): Promise<ReadableStream<AudioFrame> | null>;
|
|
79
|
-
transcriptionNode(agent: Agent, text: ReadableStream<string>, _modelSettings: ModelSettings): Promise<ReadableStream<string> | null>;
|
|
83
|
+
transcriptionNode(agent: Agent, text: ReadableStream<string | TimedString>, _modelSettings: ModelSettings): Promise<ReadableStream<string | TimedString> | null>;
|
|
80
84
|
realtimeAudioOutputNode(_agent: Agent, audio: ReadableStream<AudioFrame>, _modelSettings: ModelSettings): Promise<ReadableStream<AudioFrame> | null>;
|
|
81
85
|
};
|
|
82
86
|
}
|
package/dist/voice/agent.d.ts
CHANGED
|
@@ -12,6 +12,7 @@ import type { TTS } from '../tts/index.js';
|
|
|
12
12
|
import type { VAD } from '../vad.js';
|
|
13
13
|
import type { AgentActivity } from './agent_activity.js';
|
|
14
14
|
import type { AgentSession, TurnDetectionMode } from './agent_session.js';
|
|
15
|
+
import type { TimedString } from './io.js';
|
|
15
16
|
export declare const asyncLocalStorage: AsyncLocalStorage<{
|
|
16
17
|
functionCall?: FunctionCall | undefined;
|
|
17
18
|
}>;
|
|
@@ -36,6 +37,7 @@ export interface AgentOptions<UserData> {
|
|
|
36
37
|
tts?: TTS | TTSModelString;
|
|
37
38
|
allowInterruptions?: boolean;
|
|
38
39
|
minConsecutiveSpeechDelay?: number;
|
|
40
|
+
useTtsAlignedTranscript?: boolean;
|
|
39
41
|
}
|
|
40
42
|
export declare class Agent<UserData = any> {
|
|
41
43
|
private _id;
|
|
@@ -44,6 +46,7 @@ export declare class Agent<UserData = any> {
|
|
|
44
46
|
private _vad?;
|
|
45
47
|
private _llm?;
|
|
46
48
|
private _tts?;
|
|
49
|
+
private _useTtsAlignedTranscript?;
|
|
47
50
|
/** @internal */
|
|
48
51
|
_agentActivity?: AgentActivity;
|
|
49
52
|
/** @internal */
|
|
@@ -52,11 +55,12 @@ export declare class Agent<UserData = any> {
|
|
|
52
55
|
_instructions: string;
|
|
53
56
|
/** @internal */
|
|
54
57
|
_tools?: ToolContext<UserData>;
|
|
55
|
-
constructor({ id, instructions, chatCtx, tools, turnDetection, stt, vad, llm, tts, }: AgentOptions<UserData>);
|
|
58
|
+
constructor({ id, instructions, chatCtx, tools, turnDetection, stt, vad, llm, tts, useTtsAlignedTranscript, }: AgentOptions<UserData>);
|
|
56
59
|
get vad(): VAD | undefined;
|
|
57
60
|
get stt(): STT | undefined;
|
|
58
61
|
get llm(): LLM | RealtimeModel | undefined;
|
|
59
62
|
get tts(): TTS | undefined;
|
|
63
|
+
get useTtsAlignedTranscript(): boolean | undefined;
|
|
60
64
|
get chatCtx(): ReadonlyChatContext;
|
|
61
65
|
get id(): string;
|
|
62
66
|
get instructions(): string;
|
|
@@ -64,7 +68,7 @@ export declare class Agent<UserData = any> {
|
|
|
64
68
|
get session(): AgentSession<UserData>;
|
|
65
69
|
onEnter(): Promise<void>;
|
|
66
70
|
onExit(): Promise<void>;
|
|
67
|
-
transcriptionNode(text: ReadableStream<string>, modelSettings: ModelSettings): Promise<ReadableStream<string> | null>;
|
|
71
|
+
transcriptionNode(text: ReadableStream<string | TimedString>, modelSettings: ModelSettings): Promise<ReadableStream<string | TimedString> | null>;
|
|
68
72
|
onUserTurnCompleted(_chatCtx: ChatContext, _newMessage: ChatMessage): Promise<void>;
|
|
69
73
|
sttNode(audio: ReadableStream<AudioFrame>, modelSettings: ModelSettings): Promise<ReadableStream<SpeechEvent | string> | null>;
|
|
70
74
|
llmNode(chatCtx: ChatContext, toolCtx: ToolContext, modelSettings: ModelSettings): Promise<ReadableStream<ChatChunk | string> | null>;
|
|
@@ -76,7 +80,7 @@ export declare class Agent<UserData = any> {
|
|
|
76
80
|
sttNode(agent: Agent, audio: ReadableStream<AudioFrame>, _modelSettings: ModelSettings): Promise<ReadableStream<SpeechEvent | string> | null>;
|
|
77
81
|
llmNode(agent: Agent, chatCtx: ChatContext, toolCtx: ToolContext, modelSettings: ModelSettings): Promise<ReadableStream<ChatChunk | string> | null>;
|
|
78
82
|
ttsNode(agent: Agent, text: ReadableStream<string>, _modelSettings: ModelSettings): Promise<ReadableStream<AudioFrame> | null>;
|
|
79
|
-
transcriptionNode(agent: Agent, text: ReadableStream<string>, _modelSettings: ModelSettings): Promise<ReadableStream<string> | null>;
|
|
83
|
+
transcriptionNode(agent: Agent, text: ReadableStream<string | TimedString>, _modelSettings: ModelSettings): Promise<ReadableStream<string | TimedString> | null>;
|
|
80
84
|
realtimeAudioOutputNode(_agent: Agent, audio: ReadableStream<AudioFrame>, _modelSettings: ModelSettings): Promise<ReadableStream<AudioFrame> | null>;
|
|
81
85
|
};
|
|
82
86
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"agent.d.ts","sourceRoot":"","sources":["../../src/voice/agent.ts"],"names":[],"mappings":";;AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AACpD,OAAO,EAAE,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AACrD,OAAO,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,EAIL,KAAK,SAAS,EACd,KAAK,cAAc,EACnB,KAAK,cAAc,EACpB,MAAM,uBAAuB,CAAC;AAC/B,OAAO,EAAE,mBAAmB,EAAE,MAAM,wBAAwB,CAAC;AAC7D,OAAO,KAAK,EAAE,WAAW,EAAE,YAAY,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AAChF,OAAO,EACL,KAAK,SAAS,EACd,WAAW,EACX,GAAG,EACH,KAAK,UAAU,EACf,KAAK,WAAW,EACjB,MAAM,iBAAiB,CAAC;AACzB,OAAO,KAAK,EAAE,GAAG,EAAE,WAAW,EAAE,MAAM,iBAAiB,CAAC;AAGxD,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;
|
|
1
|
+
{"version":3,"file":"agent.d.ts","sourceRoot":"","sources":["../../src/voice/agent.ts"],"names":[],"mappings":";;AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AACpD,OAAO,EAAE,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AACrD,OAAO,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,EAIL,KAAK,SAAS,EACd,KAAK,cAAc,EACnB,KAAK,cAAc,EACpB,MAAM,uBAAuB,CAAC;AAC/B,OAAO,EAAE,mBAAmB,EAAE,MAAM,wBAAwB,CAAC;AAC7D,OAAO,KAAK,EAAE,WAAW,EAAE,YAAY,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AAChF,OAAO,EACL,KAAK,SAAS,EACd,WAAW,EACX,GAAG,EACH,KAAK,UAAU,EACf,KAAK,WAAW,EACjB,MAAM,iBAAiB,CAAC;AACzB,OAAO,KAAK,EAAE,GAAG,EAAE,WAAW,EAAE,MAAM,iBAAiB,CAAC;AAGxD,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAG3C,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AACrC,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,qBAAqB,CAAC;AACzD,OAAO,KAAK,EAAE,YAAY,EAAE,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAC1E,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,SAAS,CAAC;AAE3C,eAAO,MAAM,iBAAiB;;EAA2D,CAAC;AAC1F,eAAO,MAAM,oBAAoB,eAAyB,CAAC;AAE3D,qBAAa,YAAa,SAAQ,KAAK;;CAStC;AAED,wBAAgB,cAAc,CAAC,KAAK,EAAE,OAAO,GAAG,KAAK,IAAI,YAAY,CAOpE;AAED,MAAM,WAAW,aAAa;IAC5B,mDAAmD;IACnD,UAAU,CAAC,EAAE,UAAU,CAAC;CACzB;AAED,MAAM,WAAW,YAAY,CAAC,QAAQ;IACpC,EAAE,CAAC,EAAE,MAAM,CAAC;IACZ,YAAY,EAAE,MAAM,CAAC;IACrB,OAAO,CAAC,EAAE,WAAW,CAAC;IACtB,KAAK,CAAC,EAAE,WAAW,CAAC,QAAQ,CAAC,CAAC;IAC9B,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAClC,GAAG,CAAC,EAAE,GAAG,GAAG,cAAc,CAAC;IAC3B,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,GAAG,aAAa,GAAG,SAAS,CAAC;IACtC,GAAG,CAAC,EAAE,GAAG,GAAG,cAAc,CAAC;IAC3B,kBAAkB,CAAC,EAAE,OAAO,CAAC;IAC7B,yBAAyB,CAAC,EAAE,MAAM,CAAC;IACnC,uBAAuB,CAAC,EAAE,OAAO,CAAC;CACnC;AAED,qBAAa,KAAK,CAAC,QAAQ,GAAG,GAAG;IAC/B,OAAO,CAAC,GAAG,CAAS;IACpB,OAAO,CAAC,aAAa,CAAC,CAAoB;IAC1C,OAAO,CAAC,IAAI,CAAC,CAAM;IACnB,OAAO,CAAC,IAAI,CAAC,CAAM;IACnB,OAAO,CAAC,IAAI,CAAC,CAAsB;IACnC,OAAO,CAAC,IAAI,CAAC,CAAM;IACnB,OAAO,CAAC,wBAAwB,CAAC,CAAU;IAE3C,gBAAgB;IAChB,cAAc,CAAC,EAAE,aAAa,CAAC;IAE/B,gBAAgB;IAChB,QAAQ,EAAE,WAAW,CAAC;IAEtB,gBAAgB;IAChB,aAAa,EAAE,MAAM,CAAC;IAEtB,gBAAgB;IAChB,MAAM,CAAC,EAAE,WAAW,CAAC,QAAQ,CAAC,CAAC;gBAEnB,EACV,EAAE,EACF,YAAY,EACZ,OAAO,EACP,KAAK,EACL,aAAa,EACb,GAAG,EACH,GAAG,EACH,GAAG,EACH,GAAG,EACH,uBAAuB,GACxB,EAAE,YAAY,CAAC,QAAQ,CAAC;IAkDzB,IAAI,GAAG,IAAI,GAAG,GAAG,SAAS,CAEzB;IAED,IAAI,GAAG,IAAI,GAAG,GAAG,SAAS,CAEzB;IAED,IAAI,GAAG,IAAI,GAAG,GAAG,aAAa,GAAG,SAAS,CAEzC;IAED,IAAI,GAAG,IAAI,GAAG,GAAG,SAAS,CAEzB;IAED,IAAI,uBAAuB,IAAI,OAAO,GAAG,SAAS,CAEjD;IAED,IAAI,OAAO,IAAI,mBAAmB,CAEjC;IAED,IAAI,EAAE,IAAI,MAAM,CAEf;IAED,IAAI,YAAY,IAAI,MAAM,CAEzB;IAED,IAAI,OAAO,IAAI,WAAW,CAAC,QAAQ,CAAC,CAEnC;IAED,IAAI,OAAO,IAAI,YAAY,CAAC,QAAQ,CAAC,CAEpC;IAEK,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC;IAExB,MAAM,IAAI,OAAO,CAAC,IAAI,CAAC;IAEvB,iBAAiB,CACrB,IAAI,EAAE,cAAc,CAAC,MAAM,GAAG,WAAW,CAAC,EAC1C,aAAa,EAAE,aAAa,GAC3B,OAAO,CAAC,cAAc,CAAC,MAAM,GAAG,WAAW,CAAC,GAAG,IAAI,CAAC;IAIjD,mBAAmB,CAAC,QAAQ,EAAE,WAAW,EAAE,WAAW,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC;IAEnF,OAAO,CACX,KAAK,EAAE,cAAc,CAAC,UAAU,CAAC,EACjC,aAAa,EAAE,aAAa,GAC3B,OAAO,CAAC,cAAc,CAAC,WAAW,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC;IAIjD,OAAO,CACX,OAAO,EAAE,WAAW,EACpB,OAAO,EAAE,WAAW,EACpB,aAAa,EAAE,aAAa,GAC3B,OAAO,CAAC,cAAc,CAAC,SAAS,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC;IAI/C,OAAO,CACX,IAAI,EAAE,cAAc,CAAC,MAAM,CAAC,EAC5B,aAAa,EAAE,aAAa,GAC3B,OAAO,CAAC,cAAc,CAAC,UAAU,CAAC,GAAG,IAAI,CAAC;IAIvC,uBAAuB,CAC3B,KAAK,EAAE,cAAc,CAAC,UAAU,CAAC,EACjC,aAAa,EAAE,aAAa,GAC3B,OAAO,CAAC,cAAc,CAAC,UAAU,CAAC,GAAG,IAAI,CAAC;IAM7C,kBAAkB,IAAI,aAAa;IAO7B,aAAa,CAAC,OAAO,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC;IASxD,MAAM,CAAC,OAAO;uBAEH,KAAK,SACL,eAAe,UAAU,CAAC,kBACjB,aAAa,GAC5B,QAAQ,eAAe,WAAW,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC;uBA0D9C,KAAK,WACH,WAAW,WACX,WAAW,iBACL,aAAa,GAC3B,QAAQ,eAAe,SAAS,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC;uBAiD5C,KAAK,QACN,eAAe,MAAM,CAAC,kBACZ,aAAa,GAC5B,QAAQ,eAAe,UAAU,CAAC,GAAG,IAAI,CAAC;iCAgDpC,KAAK,QACN,eAAe,MAAM,GAAG,WAAW,CAAC,kBAC1B,aAAa,GAC5B,QAAQ,eAAe,MAAM,GAAG,WAAW,CAAC,GAAG,IAAI,CAAC;wCAK7C,KAAK,SACN,eAAe,UAAU,CAAC,kBACjB,aAAa,GAC5B,QAAQ,eAAe,UAAU,CAAC,GAAG,IAAI,CAAC;MAG7C;CACH"}
|
package/dist/voice/agent.js
CHANGED
|
@@ -13,6 +13,7 @@ import {
|
|
|
13
13
|
import { StreamAdapter as STTStreamAdapter } from "../stt/index.js";
|
|
14
14
|
import { SentenceTokenizer as BasicSentenceTokenizer } from "../tokenize/basic/index.js";
|
|
15
15
|
import { SynthesizeStream, StreamAdapter as TTSStreamAdapter } from "../tts/index.js";
|
|
16
|
+
import { USERDATA_TIMED_TRANSCRIPT } from "../types.js";
|
|
16
17
|
const asyncLocalStorage = new AsyncLocalStorage();
|
|
17
18
|
const STOP_RESPONSE_SYMBOL = Symbol("StopResponse");
|
|
18
19
|
class StopResponse extends Error {
|
|
@@ -34,6 +35,7 @@ class Agent {
|
|
|
34
35
|
_vad;
|
|
35
36
|
_llm;
|
|
36
37
|
_tts;
|
|
38
|
+
_useTtsAlignedTranscript;
|
|
37
39
|
/** @internal */
|
|
38
40
|
_agentActivity;
|
|
39
41
|
/** @internal */
|
|
@@ -51,7 +53,8 @@ class Agent {
|
|
|
51
53
|
stt,
|
|
52
54
|
vad,
|
|
53
55
|
llm,
|
|
54
|
-
tts
|
|
56
|
+
tts,
|
|
57
|
+
useTtsAlignedTranscript
|
|
55
58
|
}) {
|
|
56
59
|
if (id) {
|
|
57
60
|
this._id = id;
|
|
@@ -85,6 +88,7 @@ class Agent {
|
|
|
85
88
|
} else {
|
|
86
89
|
this._tts = tts;
|
|
87
90
|
}
|
|
91
|
+
this._useTtsAlignedTranscript = useTtsAlignedTranscript;
|
|
88
92
|
this._agentActivity = void 0;
|
|
89
93
|
}
|
|
90
94
|
get vad() {
|
|
@@ -99,6 +103,9 @@ class Agent {
|
|
|
99
103
|
get tts() {
|
|
100
104
|
return this._tts;
|
|
101
105
|
}
|
|
106
|
+
get useTtsAlignedTranscript() {
|
|
107
|
+
return this._useTtsAlignedTranscript;
|
|
108
|
+
}
|
|
102
109
|
get chatCtx() {
|
|
103
110
|
return new ReadonlyChatContext(this._chatCtx.items);
|
|
104
111
|
}
|
|
@@ -261,6 +268,9 @@ class Agent {
|
|
|
261
268
|
if (chunk === SynthesizeStream.END_OF_STREAM) {
|
|
262
269
|
break;
|
|
263
270
|
}
|
|
271
|
+
if (chunk.timedTranscripts && chunk.timedTranscripts.length > 0) {
|
|
272
|
+
chunk.frame.userdata[USERDATA_TIMED_TRANSCRIPT] = chunk.timedTranscripts;
|
|
273
|
+
}
|
|
264
274
|
controller.enqueue(chunk.frame);
|
|
265
275
|
}
|
|
266
276
|
controller.close();
|