@livekit/agents 1.0.18 → 1.0.20
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/index.cjs +3 -0
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.cts +2 -1
- package/dist/index.d.ts +2 -1
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +2 -0
- package/dist/index.js.map +1 -1
- package/dist/inference/api_protos.d.cts +12 -12
- package/dist/inference/api_protos.d.ts +12 -12
- package/dist/inference/tts.cjs +1 -1
- package/dist/inference/tts.cjs.map +1 -1
- package/dist/inference/tts.js +1 -1
- package/dist/inference/tts.js.map +1 -1
- package/dist/ipc/job_proc_lazy_main.cjs +6 -2
- package/dist/ipc/job_proc_lazy_main.cjs.map +1 -1
- package/dist/ipc/job_proc_lazy_main.js +6 -2
- package/dist/ipc/job_proc_lazy_main.js.map +1 -1
- package/dist/job.cjs +31 -0
- package/dist/job.cjs.map +1 -1
- package/dist/job.d.cts +6 -0
- package/dist/job.d.ts +6 -0
- package/dist/job.d.ts.map +1 -1
- package/dist/job.js +31 -0
- package/dist/job.js.map +1 -1
- package/dist/llm/chat_context.cjs +33 -0
- package/dist/llm/chat_context.cjs.map +1 -1
- package/dist/llm/chat_context.d.cts +22 -2
- package/dist/llm/chat_context.d.ts +22 -2
- package/dist/llm/chat_context.d.ts.map +1 -1
- package/dist/llm/chat_context.js +32 -0
- package/dist/llm/chat_context.js.map +1 -1
- package/dist/llm/index.cjs +2 -0
- package/dist/llm/index.cjs.map +1 -1
- package/dist/llm/index.d.cts +1 -1
- package/dist/llm/index.d.ts +1 -1
- package/dist/llm/index.d.ts.map +1 -1
- package/dist/llm/index.js +2 -0
- package/dist/llm/index.js.map +1 -1
- package/dist/llm/llm.cjs.map +1 -1
- package/dist/llm/llm.d.ts.map +1 -1
- package/dist/llm/llm.js.map +1 -1
- package/dist/llm/provider_format/google.test.cjs +48 -0
- package/dist/llm/provider_format/google.test.cjs.map +1 -1
- package/dist/llm/provider_format/google.test.js +54 -1
- package/dist/llm/provider_format/google.test.js.map +1 -1
- package/dist/llm/provider_format/index.d.cts +1 -1
- package/dist/llm/provider_format/index.d.ts +1 -1
- package/dist/llm/provider_format/openai.cjs +1 -2
- package/dist/llm/provider_format/openai.cjs.map +1 -1
- package/dist/llm/provider_format/openai.js +1 -2
- package/dist/llm/provider_format/openai.js.map +1 -1
- package/dist/llm/provider_format/openai.test.cjs +32 -0
- package/dist/llm/provider_format/openai.test.cjs.map +1 -1
- package/dist/llm/provider_format/openai.test.js +38 -1
- package/dist/llm/provider_format/openai.test.js.map +1 -1
- package/dist/log.cjs.map +1 -1
- package/dist/log.d.ts.map +1 -1
- package/dist/log.js.map +1 -1
- package/dist/stt/stt.cjs +3 -0
- package/dist/stt/stt.cjs.map +1 -1
- package/dist/stt/stt.d.cts +1 -0
- package/dist/stt/stt.d.ts +1 -0
- package/dist/stt/stt.d.ts.map +1 -1
- package/dist/stt/stt.js +3 -0
- package/dist/stt/stt.js.map +1 -1
- package/dist/telemetry/index.cjs +51 -0
- package/dist/telemetry/index.cjs.map +1 -0
- package/dist/telemetry/index.d.cts +4 -0
- package/dist/telemetry/index.d.ts +4 -0
- package/dist/telemetry/index.d.ts.map +1 -0
- package/dist/telemetry/index.js +12 -0
- package/dist/telemetry/index.js.map +1 -0
- package/dist/telemetry/trace_types.cjs +191 -0
- package/dist/telemetry/trace_types.cjs.map +1 -0
- package/dist/telemetry/trace_types.d.cts +56 -0
- package/dist/telemetry/trace_types.d.ts +56 -0
- package/dist/telemetry/trace_types.d.ts.map +1 -0
- package/dist/telemetry/trace_types.js +113 -0
- package/dist/telemetry/trace_types.js.map +1 -0
- package/dist/telemetry/traces.cjs +196 -0
- package/dist/telemetry/traces.cjs.map +1 -0
- package/dist/telemetry/traces.d.cts +97 -0
- package/dist/telemetry/traces.d.ts +97 -0
- package/dist/telemetry/traces.d.ts.map +1 -0
- package/dist/telemetry/traces.js +173 -0
- package/dist/telemetry/traces.js.map +1 -0
- package/dist/telemetry/utils.cjs +86 -0
- package/dist/telemetry/utils.cjs.map +1 -0
- package/dist/telemetry/utils.d.cts +5 -0
- package/dist/telemetry/utils.d.ts +5 -0
- package/dist/telemetry/utils.d.ts.map +1 -0
- package/dist/telemetry/utils.js +51 -0
- package/dist/telemetry/utils.js.map +1 -0
- package/dist/tts/tts.cjs +3 -0
- package/dist/tts/tts.cjs.map +1 -1
- package/dist/tts/tts.d.cts +1 -0
- package/dist/tts/tts.d.ts +1 -0
- package/dist/tts/tts.d.ts.map +1 -1
- package/dist/tts/tts.js +3 -0
- package/dist/tts/tts.js.map +1 -1
- package/dist/vad.cjs +3 -0
- package/dist/vad.cjs.map +1 -1
- package/dist/vad.d.cts +1 -0
- package/dist/vad.d.ts +1 -0
- package/dist/vad.d.ts.map +1 -1
- package/dist/vad.js +3 -0
- package/dist/vad.js.map +1 -1
- package/dist/voice/agent.cjs +15 -0
- package/dist/voice/agent.cjs.map +1 -1
- package/dist/voice/agent.d.cts +4 -1
- package/dist/voice/agent.d.ts +4 -1
- package/dist/voice/agent.d.ts.map +1 -1
- package/dist/voice/agent.js +15 -0
- package/dist/voice/agent.js.map +1 -1
- package/dist/voice/agent_activity.cjs +5 -0
- package/dist/voice/agent_activity.cjs.map +1 -1
- package/dist/voice/agent_activity.d.ts.map +1 -1
- package/dist/voice/agent_activity.js +5 -0
- package/dist/voice/agent_activity.js.map +1 -1
- package/dist/voice/agent_session.cjs +29 -1
- package/dist/voice/agent_session.cjs.map +1 -1
- package/dist/voice/agent_session.d.cts +6 -2
- package/dist/voice/agent_session.d.ts +6 -2
- package/dist/voice/agent_session.d.ts.map +1 -1
- package/dist/voice/agent_session.js +30 -2
- package/dist/voice/agent_session.js.map +1 -1
- package/dist/voice/audio_recognition.cjs +1 -1
- package/dist/voice/audio_recognition.cjs.map +1 -1
- package/dist/voice/audio_recognition.d.ts.map +1 -1
- package/dist/voice/audio_recognition.js +1 -1
- package/dist/voice/audio_recognition.js.map +1 -1
- package/dist/voice/generation.cjs.map +1 -1
- package/dist/voice/generation.d.ts.map +1 -1
- package/dist/voice/generation.js.map +1 -1
- package/dist/voice/index.cjs +2 -0
- package/dist/voice/index.cjs.map +1 -1
- package/dist/voice/index.d.cts +1 -0
- package/dist/voice/index.d.ts +1 -0
- package/dist/voice/index.d.ts.map +1 -1
- package/dist/voice/index.js +1 -0
- package/dist/voice/index.js.map +1 -1
- package/dist/voice/report.cjs +69 -0
- package/dist/voice/report.cjs.map +1 -0
- package/dist/voice/report.d.cts +26 -0
- package/dist/voice/report.d.ts +26 -0
- package/dist/voice/report.d.ts.map +1 -0
- package/dist/voice/report.js +44 -0
- package/dist/voice/report.js.map +1 -0
- package/package.json +10 -3
- package/src/index.ts +2 -1
- package/src/inference/tts.ts +1 -1
- package/src/ipc/job_proc_lazy_main.ts +10 -2
- package/src/job.ts +48 -0
- package/src/llm/chat_context.ts +53 -1
- package/src/llm/index.ts +1 -0
- package/src/llm/llm.ts +2 -0
- package/src/llm/provider_format/google.test.ts +72 -1
- package/src/llm/provider_format/openai.test.ts +55 -1
- package/src/llm/provider_format/openai.ts +3 -2
- package/src/log.ts +1 -0
- package/src/stt/stt.ts +4 -0
- package/src/telemetry/index.ts +10 -0
- package/src/telemetry/trace_types.ts +88 -0
- package/src/telemetry/traces.ts +266 -0
- package/src/telemetry/utils.ts +61 -0
- package/src/tts/tts.ts +8 -0
- package/src/vad.ts +4 -0
- package/src/voice/agent.ts +22 -0
- package/src/voice/agent_activity.ts +9 -0
- package/src/voice/agent_session.ts +44 -1
- package/src/voice/audio_recognition.ts +3 -1
- package/src/voice/generation.ts +3 -0
- package/src/voice/index.ts +1 -0
- package/src/voice/report.ts +77 -0
package/dist/vad.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/vad.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type {\n ReadableStream,\n ReadableStreamDefaultReader,\n WritableStreamDefaultWriter,\n} from 'node:stream/web';\nimport { log } from './log.js';\nimport type { VADMetrics } from './metrics/base.js';\nimport { DeferredReadableStream } from './stream/deferred_stream.js';\nimport { IdentityTransform } from './stream/identity_transform.js';\n\nexport enum VADEventType {\n START_OF_SPEECH,\n INFERENCE_DONE,\n END_OF_SPEECH,\n METRICS_COLLECTED,\n}\n\nexport interface VADEvent {\n /** Type of the VAD event (e.g., start of speech, end of speech, inference done). */\n type: VADEventType;\n /**\n * Index of the audio sample where the event occurred, relative to the inference sample rate.\n */\n samplesIndex: number;\n /** Timestamp when the event was fired. */\n timestamp: number;\n /** Duration of the speech segment. */\n speechDuration: number;\n /** Duration of the silence segment. */\n silenceDuration: number;\n /**\n * List of audio frames associated with the speech.\n *\n * @remarks\n * - For `start_of_speech` events, this contains the audio chunks that triggered the detection.\n * - For `inference_done` events, this contains the audio chunks that were processed.\n * - For `end_of_speech` events, this contains the complete user speech.\n */\n frames: AudioFrame[];\n /** Probability that speech is present (only for `INFERENCE_DONE` events). */\n probability: number;\n /** Time taken to perform the inference, in seconds (only for `INFERENCE_DONE` events). */\n inferenceDuration: number;\n /** Indicates whether speech was detected in the frames. */\n speaking: boolean;\n /** Threshold used to detect silence. */\n rawAccumulatedSilence: number;\n /** Threshold used to detect speech. */\n rawAccumulatedSpeech: number;\n}\n\nexport interface VADCapabilities {\n updateInterval: number;\n}\n\nexport type VADCallbacks = {\n ['metrics_collected']: (metrics: VADMetrics) => void;\n};\n\nexport abstract class VAD extends (EventEmitter as new () => TypedEmitter<VADCallbacks>) {\n #capabilities: VADCapabilities;\n abstract label: string;\n\n constructor(capabilities: VADCapabilities) {\n super();\n this.#capabilities = capabilities;\n }\n\n get capabilities(): VADCapabilities {\n return this.#capabilities;\n }\n\n /**\n * Returns a {@link VADStream} that can be used to push audio frames and receive VAD events.\n */\n abstract stream(): VADStream;\n}\n\nexport abstract class VADStream implements AsyncIterableIterator<VADEvent> {\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n protected input = new IdentityTransform<AudioFrame | typeof VADStream.FLUSH_SENTINEL>();\n protected output = new IdentityTransform<VADEvent>();\n protected inputWriter: WritableStreamDefaultWriter<AudioFrame | typeof VADStream.FLUSH_SENTINEL>;\n protected inputReader: ReadableStreamDefaultReader<AudioFrame | typeof VADStream.FLUSH_SENTINEL>;\n protected outputWriter: WritableStreamDefaultWriter<VADEvent>;\n protected outputReader: ReadableStreamDefaultReader<VADEvent>;\n protected closed = false;\n protected inputClosed = false;\n\n #vad: VAD;\n #lastActivityTime = BigInt(0);\n private logger = log();\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n\n private metricsStream: ReadableStream<VADEvent>;\n constructor(vad: VAD) {\n this.#vad = vad;\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n\n this.inputWriter = this.input.writable.getWriter();\n this.inputReader = this.input.readable.getReader();\n this.outputWriter = this.output.writable.getWriter();\n\n const [outputStream, metricsStream] = this.output.readable.tee();\n this.metricsStream = metricsStream;\n this.outputReader = outputStream.getReader();\n\n this.pumpDeferredStream();\n this.monitorMetrics();\n }\n\n /**\n * Reads from the deferred input stream and forwards chunks to the input writer.\n *\n * Note: we can't just do this.deferredInputStream.stream.pipeTo(this.input.writable)\n * because the inputWriter locks the this.input.writable stream. All writes must go through\n * the inputWriter.\n */\n private async pumpDeferredStream() {\n const reader = this.deferredInputStream.stream.getReader();\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done) break;\n await this.inputWriter.write(value);\n }\n } catch (e) {\n this.logger.error(`Error pumping deferred stream: ${e}`);\n throw e;\n } finally {\n reader.releaseLock();\n }\n }\n\n protected async monitorMetrics() {\n let inferenceDurationTotalMs = 0;\n let inferenceCount = 0;\n const metricsReader = this.metricsStream.getReader();\n while (true) {\n const { done, value } = await metricsReader.read();\n if (done) {\n break;\n }\n switch (value.type) {\n case VADEventType.START_OF_SPEECH:\n inferenceCount++;\n if (inferenceCount >= 1 / this.#vad.capabilities.updateInterval) {\n this.#vad.emit('metrics_collected', {\n type: 'vad_metrics',\n timestamp: Date.now(),\n idleTimeMs: Math.trunc(\n Number((process.hrtime.bigint() - this.#lastActivityTime) / BigInt(1000000)),\n ),\n inferenceDurationTotalMs,\n inferenceCount,\n label: this.#vad.label,\n });\n\n inferenceCount = 0;\n inferenceDurationTotalMs = 0;\n }\n break;\n case VADEventType.INFERENCE_DONE:\n inferenceDurationTotalMs += Math.round(value.inferenceDuration);\n this.#lastActivityTime = process.hrtime.bigint();\n break;\n case VADEventType.END_OF_SPEECH:\n this.#lastActivityTime = process.hrtime.bigint();\n break;\n }\n }\n }\n\n /**\n * Safely send a VAD event to the output stream, handling writer release errors during shutdown.\n * @returns true if the event was sent, false if the stream is closing\n * @throws Error if an unexpected error occurs\n */\n protected sendVADEvent(event: VADEvent): boolean {\n if (this.closed) {\n return false;\n }\n\n try {\n this.outputWriter.write(event);\n return true;\n } catch (e) {\n throw e;\n }\n }\n\n updateInputStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputStream() {\n this.deferredInputStream.detachSource();\n }\n\n /** @deprecated Use `updateInputStream` instead */\n pushFrame(frame: AudioFrame) {\n // TODO(AJS-395): remove this method\n if (this.inputClosed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.inputWriter.write(frame);\n }\n\n flush() {\n if (this.inputClosed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.inputWriter.write(VADStream.FLUSH_SENTINEL);\n }\n\n endInput() {\n if (this.inputClosed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.inputClosed = true;\n this.input.writable.close();\n }\n\n async next(): Promise<IteratorResult<VADEvent>> {\n return this.outputReader.read().then(({ done, value }) => {\n if (done) {\n return { done: true, value: undefined };\n }\n return { done: false, value };\n });\n }\n\n close() {\n this.outputWriter.releaseLock();\n this.outputReader.cancel();\n this.output.writable.close();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): VADStream {\n return this;\n }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAM7B,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAS,yBAAyB;AAE3B,IAAK,eAAL,kBAAKA,kBAAL;AACL,EAAAA,4BAAA;AACA,EAAAA,4BAAA;AACA,EAAAA,4BAAA;AACA,EAAAA,4BAAA;AAJU,SAAAA;AAAA,GAAA;AAiDL,MAAe,YAAa,aAAsD;AAAA,EACvF;AAAA,EAGA,YAAY,cAA+B;AACzC,UAAM;AACN,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAMF;AAEO,MAAe,UAAqD;AAAA,EACzE,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,kBAAgE;AAAA,EAC5E,SAAS,IAAI,kBAA4B;AAAA,EACzC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,SAAS;AAAA,EACT,cAAc;AAAA,EAExB;AAAA,EACA,oBAAoB,OAAO,CAAC;AAAA,EACpB,SAAS,IAAI;AAAA,EACb;AAAA,EAEA;AAAA,EACR,YAAY,KAAU;AACpB,SAAK,OAAO;AACZ,SAAK,sBAAsB,IAAI,uBAAmC;AAElE,SAAK,cAAc,KAAK,MAAM,SAAS,UAAU;AACjD,SAAK,cAAc,KAAK,MAAM,SAAS,UAAU;AACjD,SAAK,eAAe,KAAK,OAAO,SAAS,UAAU;AAEnD,UAAM,CAAC,cAAc,aAAa,IAAI,KAAK,OAAO,SAAS,IAAI;AAC/D,SAAK,gBAAgB;AACrB,SAAK,eAAe,aAAa,UAAU;AAE3C,SAAK,mBAAmB;AACxB,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,MAAc,qBAAqB;AACjC,UAAM,SAAS,KAAK,oBAAoB,OAAO,UAAU;AACzD,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AACV,cAAM,KAAK,YAAY,MAAM,KAAK;AAAA,MACpC;AAAA,IACF,SAAS,GAAG;AACV,WAAK,OAAO,MAAM,kCAAkC,CAAC,EAAE;AACvD,YAAM;AAAA,IACR,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,QAAI,2BAA2B;AAC/B,QAAI,iBAAiB;AACrB,UAAM,gBAAgB,KAAK,cAAc,UAAU;AACnD,WAAO,MAAM;AACX,YAAM,EAAE,MAAM,MAAM,IAAI,MAAM,cAAc,KAAK;AACjD,UAAI,MAAM;AACR;AAAA,MACF;AACA,cAAQ,MAAM,MAAM;AAAA,QAClB,KAAK;AACH;AACA,cAAI,kBAAkB,IAAI,KAAK,KAAK,aAAa,gBAAgB;AAC/D,iBAAK,KAAK,KAAK,qBAAqB;AAAA,cAClC,MAAM;AAAA,cACN,WAAW,KAAK,IAAI;AAAA,cACpB,YAAY,KAAK;AAAA,gBACf,QAAQ,QAAQ,OAAO,OAAO,IAAI,KAAK,qBAAqB,OAAO,GAAO,CAAC;AAAA,cAC7E;AAAA,cACA;AAAA,cACA;AAAA,cACA,OAAO,KAAK,KAAK;AAAA,YACnB,CAAC;AAED,6BAAiB;AACjB,uCAA2B;AAAA,UAC7B;AACA;AAAA,QACF,KAAK;AACH,sCAA4B,KAAK,MAAM,MAAM,iBAAiB;AAC9D,eAAK,oBAAoB,QAAQ,OAAO,OAAO;AAC/C;AAAA,QACF,KAAK;AACH,eAAK,oBAAoB,QAAQ,OAAO,OAAO;AAC/C;AAAA,MACJ;AAAA,IACF;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOU,aAAa,OAA0B;AAC/C,QAAI,KAAK,QAAQ;AACf,aAAO;AAAA,IACT;AAEA,QAAI;AACF,WAAK,aAAa,MAAM,KAAK;AAC7B,aAAO;AAAA,IACT,SAAS,GAAG;AACV,YAAM;AAAA,IACR;AAAA,EACF;AAAA,EAEA,kBAAkB,aAAyC;AACzD,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,oBAAoB;AAClB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA;AAAA,EAGA,UAAU,OAAmB;AAE3B,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,YAAY,MAAM,KAAK;AAAA,EAC9B;AAAA,EAEA,QAAQ;AACN,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,YAAY,MAAM,UAAU,cAAc;AAAA,EACjD;AAAA,EAEA,WAAW;AACT,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,cAAc;AACnB,SAAK,MAAM,SAAS,MAAM;AAAA,EAC5B;AAAA,EAEA,MAAM,OAA0C;AAC9C,WAAO,KAAK,aAAa,KAAK,EAAE,KAAK,CAAC,EAAE,MAAM,MAAM,MAAM;AACxD,UAAI,MAAM;AACR,eAAO,EAAE,MAAM,MAAM,OAAO,OAAU;AAAA,MACxC;AACA,aAAO,EAAE,MAAM,OAAO,MAAM;AAAA,IAC9B,CAAC;AAAA,EACH;AAAA,EAEA,QAAQ;AACN,SAAK,aAAa,YAAY;AAC9B,SAAK,aAAa,OAAO;AACzB,SAAK,OAAO,SAAS,MAAM;AAC3B,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAe;AAClC,WAAO;AAAA,EACT;AACF;","names":["VADEventType"]}
|
|
1
|
+
{"version":3,"sources":["../src/vad.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type {\n ReadableStream,\n ReadableStreamDefaultReader,\n WritableStreamDefaultWriter,\n} from 'node:stream/web';\nimport { log } from './log.js';\nimport type { VADMetrics } from './metrics/base.js';\nimport { DeferredReadableStream } from './stream/deferred_stream.js';\nimport { IdentityTransform } from './stream/identity_transform.js';\n\nexport enum VADEventType {\n START_OF_SPEECH,\n INFERENCE_DONE,\n END_OF_SPEECH,\n METRICS_COLLECTED,\n}\n\nexport interface VADEvent {\n /** Type of the VAD event (e.g., start of speech, end of speech, inference done). */\n type: VADEventType;\n /**\n * Index of the audio sample where the event occurred, relative to the inference sample rate.\n */\n samplesIndex: number;\n /** Timestamp when the event was fired. */\n timestamp: number;\n /** Duration of the speech segment. */\n speechDuration: number;\n /** Duration of the silence segment. */\n silenceDuration: number;\n /**\n * List of audio frames associated with the speech.\n *\n * @remarks\n * - For `start_of_speech` events, this contains the audio chunks that triggered the detection.\n * - For `inference_done` events, this contains the audio chunks that were processed.\n * - For `end_of_speech` events, this contains the complete user speech.\n */\n frames: AudioFrame[];\n /** Probability that speech is present (only for `INFERENCE_DONE` events). */\n probability: number;\n /** Time taken to perform the inference, in seconds (only for `INFERENCE_DONE` events). */\n inferenceDuration: number;\n /** Indicates whether speech was detected in the frames. */\n speaking: boolean;\n /** Threshold used to detect silence. */\n rawAccumulatedSilence: number;\n /** Threshold used to detect speech. */\n rawAccumulatedSpeech: number;\n}\n\nexport interface VADCapabilities {\n updateInterval: number;\n}\n\nexport type VADCallbacks = {\n ['metrics_collected']: (metrics: VADMetrics) => void;\n};\n\nexport abstract class VAD extends (EventEmitter as new () => TypedEmitter<VADCallbacks>) {\n #capabilities: VADCapabilities;\n abstract label: string;\n\n constructor(capabilities: VADCapabilities) {\n super();\n this.#capabilities = capabilities;\n }\n\n get capabilities(): VADCapabilities {\n return this.#capabilities;\n }\n\n /**\n * Returns a {@link VADStream} that can be used to push audio frames and receive VAD events.\n */\n abstract stream(): VADStream;\n\n async close(): Promise<void> {\n return;\n }\n}\n\nexport abstract class VADStream implements AsyncIterableIterator<VADEvent> {\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n protected input = new IdentityTransform<AudioFrame | typeof VADStream.FLUSH_SENTINEL>();\n protected output = new IdentityTransform<VADEvent>();\n protected inputWriter: WritableStreamDefaultWriter<AudioFrame | typeof VADStream.FLUSH_SENTINEL>;\n protected inputReader: ReadableStreamDefaultReader<AudioFrame | typeof VADStream.FLUSH_SENTINEL>;\n protected outputWriter: WritableStreamDefaultWriter<VADEvent>;\n protected outputReader: ReadableStreamDefaultReader<VADEvent>;\n protected closed = false;\n protected inputClosed = false;\n\n #vad: VAD;\n #lastActivityTime = BigInt(0);\n private logger = log();\n private deferredInputStream: DeferredReadableStream<AudioFrame>;\n\n private metricsStream: ReadableStream<VADEvent>;\n constructor(vad: VAD) {\n this.#vad = vad;\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n\n this.inputWriter = this.input.writable.getWriter();\n this.inputReader = this.input.readable.getReader();\n this.outputWriter = this.output.writable.getWriter();\n\n const [outputStream, metricsStream] = this.output.readable.tee();\n this.metricsStream = metricsStream;\n this.outputReader = outputStream.getReader();\n\n this.pumpDeferredStream();\n this.monitorMetrics();\n }\n\n /**\n * Reads from the deferred input stream and forwards chunks to the input writer.\n *\n * Note: we can't just do this.deferredInputStream.stream.pipeTo(this.input.writable)\n * because the inputWriter locks the this.input.writable stream. All writes must go through\n * the inputWriter.\n */\n private async pumpDeferredStream() {\n const reader = this.deferredInputStream.stream.getReader();\n try {\n while (true) {\n const { done, value } = await reader.read();\n if (done) break;\n await this.inputWriter.write(value);\n }\n } catch (e) {\n this.logger.error(`Error pumping deferred stream: ${e}`);\n throw e;\n } finally {\n reader.releaseLock();\n }\n }\n\n protected async monitorMetrics() {\n let inferenceDurationTotalMs = 0;\n let inferenceCount = 0;\n const metricsReader = this.metricsStream.getReader();\n while (true) {\n const { done, value } = await metricsReader.read();\n if (done) {\n break;\n }\n switch (value.type) {\n case VADEventType.START_OF_SPEECH:\n inferenceCount++;\n if (inferenceCount >= 1 / this.#vad.capabilities.updateInterval) {\n this.#vad.emit('metrics_collected', {\n type: 'vad_metrics',\n timestamp: Date.now(),\n idleTimeMs: Math.trunc(\n Number((process.hrtime.bigint() - this.#lastActivityTime) / BigInt(1000000)),\n ),\n inferenceDurationTotalMs,\n inferenceCount,\n label: this.#vad.label,\n });\n\n inferenceCount = 0;\n inferenceDurationTotalMs = 0;\n }\n break;\n case VADEventType.INFERENCE_DONE:\n inferenceDurationTotalMs += Math.round(value.inferenceDuration);\n this.#lastActivityTime = process.hrtime.bigint();\n break;\n case VADEventType.END_OF_SPEECH:\n this.#lastActivityTime = process.hrtime.bigint();\n break;\n }\n }\n }\n\n /**\n * Safely send a VAD event to the output stream, handling writer release errors during shutdown.\n * @returns true if the event was sent, false if the stream is closing\n * @throws Error if an unexpected error occurs\n */\n protected sendVADEvent(event: VADEvent): boolean {\n if (this.closed) {\n return false;\n }\n\n try {\n this.outputWriter.write(event);\n return true;\n } catch (e) {\n throw e;\n }\n }\n\n updateInputStream(audioStream: ReadableStream<AudioFrame>) {\n this.deferredInputStream.setSource(audioStream);\n }\n\n detachInputStream() {\n this.deferredInputStream.detachSource();\n }\n\n /** @deprecated Use `updateInputStream` instead */\n pushFrame(frame: AudioFrame) {\n // TODO(AJS-395): remove this method\n if (this.inputClosed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.inputWriter.write(frame);\n }\n\n flush() {\n if (this.inputClosed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.inputWriter.write(VADStream.FLUSH_SENTINEL);\n }\n\n endInput() {\n if (this.inputClosed) {\n throw new Error('Input is closed');\n }\n if (this.closed) {\n throw new Error('Stream is closed');\n }\n this.inputClosed = true;\n this.input.writable.close();\n }\n\n async next(): Promise<IteratorResult<VADEvent>> {\n return this.outputReader.read().then(({ done, value }) => {\n if (done) {\n return { done: true, value: undefined };\n }\n return { done: false, value };\n });\n }\n\n close() {\n this.outputWriter.releaseLock();\n this.outputReader.cancel();\n this.output.writable.close();\n this.closed = true;\n }\n\n [Symbol.asyncIterator](): VADStream {\n return this;\n }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAM7B,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAS,yBAAyB;AAE3B,IAAK,eAAL,kBAAKA,kBAAL;AACL,EAAAA,4BAAA;AACA,EAAAA,4BAAA;AACA,EAAAA,4BAAA;AACA,EAAAA,4BAAA;AAJU,SAAAA;AAAA,GAAA;AAiDL,MAAe,YAAa,aAAsD;AAAA,EACvF;AAAA,EAGA,YAAY,cAA+B;AACzC,UAAM;AACN,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA,EAOA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAEO,MAAe,UAAqD;AAAA,EACzE,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,kBAAgE;AAAA,EAC5E,SAAS,IAAI,kBAA4B;AAAA,EACzC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,SAAS;AAAA,EACT,cAAc;AAAA,EAExB;AAAA,EACA,oBAAoB,OAAO,CAAC;AAAA,EACpB,SAAS,IAAI;AAAA,EACb;AAAA,EAEA;AAAA,EACR,YAAY,KAAU;AACpB,SAAK,OAAO;AACZ,SAAK,sBAAsB,IAAI,uBAAmC;AAElE,SAAK,cAAc,KAAK,MAAM,SAAS,UAAU;AACjD,SAAK,cAAc,KAAK,MAAM,SAAS,UAAU;AACjD,SAAK,eAAe,KAAK,OAAO,SAAS,UAAU;AAEnD,UAAM,CAAC,cAAc,aAAa,IAAI,KAAK,OAAO,SAAS,IAAI;AAC/D,SAAK,gBAAgB;AACrB,SAAK,eAAe,aAAa,UAAU;AAE3C,SAAK,mBAAmB;AACxB,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,MAAc,qBAAqB;AACjC,UAAM,SAAS,KAAK,oBAAoB,OAAO,UAAU;AACzD,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AACV,cAAM,KAAK,YAAY,MAAM,KAAK;AAAA,MACpC;AAAA,IACF,SAAS,GAAG;AACV,WAAK,OAAO,MAAM,kCAAkC,CAAC,EAAE;AACvD,YAAM;AAAA,IACR,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,QAAI,2BAA2B;AAC/B,QAAI,iBAAiB;AACrB,UAAM,gBAAgB,KAAK,cAAc,UAAU;AACnD,WAAO,MAAM;AACX,YAAM,EAAE,MAAM,MAAM,IAAI,MAAM,cAAc,KAAK;AACjD,UAAI,MAAM;AACR;AAAA,MACF;AACA,cAAQ,MAAM,MAAM;AAAA,QAClB,KAAK;AACH;AACA,cAAI,kBAAkB,IAAI,KAAK,KAAK,aAAa,gBAAgB;AAC/D,iBAAK,KAAK,KAAK,qBAAqB;AAAA,cAClC,MAAM;AAAA,cACN,WAAW,KAAK,IAAI;AAAA,cACpB,YAAY,KAAK;AAAA,gBACf,QAAQ,QAAQ,OAAO,OAAO,IAAI,KAAK,qBAAqB,OAAO,GAAO,CAAC;AAAA,cAC7E;AAAA,cACA;AAAA,cACA;AAAA,cACA,OAAO,KAAK,KAAK;AAAA,YACnB,CAAC;AAED,6BAAiB;AACjB,uCAA2B;AAAA,UAC7B;AACA;AAAA,QACF,KAAK;AACH,sCAA4B,KAAK,MAAM,MAAM,iBAAiB;AAC9D,eAAK,oBAAoB,QAAQ,OAAO,OAAO;AAC/C;AAAA,QACF,KAAK;AACH,eAAK,oBAAoB,QAAQ,OAAO,OAAO;AAC/C;AAAA,MACJ;AAAA,IACF;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOU,aAAa,OAA0B;AAC/C,QAAI,KAAK,QAAQ;AACf,aAAO;AAAA,IACT;AAEA,QAAI;AACF,WAAK,aAAa,MAAM,KAAK;AAC7B,aAAO;AAAA,IACT,SAAS,GAAG;AACV,YAAM;AAAA,IACR;AAAA,EACF;AAAA,EAEA,kBAAkB,aAAyC;AACzD,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,oBAAoB;AAClB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA;AAAA,EAGA,UAAU,OAAmB;AAE3B,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,YAAY,MAAM,KAAK;AAAA,EAC9B;AAAA,EAEA,QAAQ;AACN,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,YAAY,MAAM,UAAU,cAAc;AAAA,EACjD;AAAA,EAEA,WAAW;AACT,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,cAAc;AACnB,SAAK,MAAM,SAAS,MAAM;AAAA,EAC5B;AAAA,EAEA,MAAM,OAA0C;AAC9C,WAAO,KAAK,aAAa,KAAK,EAAE,KAAK,CAAC,EAAE,MAAM,MAAM,MAAM;AACxD,UAAI,MAAM;AACR,eAAO,EAAE,MAAM,MAAM,OAAO,OAAU;AAAA,MACxC;AACA,aAAO,EAAE,MAAM,OAAO,MAAM;AAAA,IAC9B,CAAC;AAAA,EACH;AAAA,EAEA,QAAQ;AACN,SAAK,aAAa,YAAY;AAC9B,SAAK,aAAa,OAAO;AACzB,SAAK,OAAO,SAAS,MAAM;AAC3B,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAe;AAClC,WAAO;AAAA,EACT;AACF;","names":["VADEventType"]}
|
package/dist/voice/agent.cjs
CHANGED
|
@@ -48,6 +48,7 @@ function isStopResponse(value) {
|
|
|
48
48
|
return value !== void 0 && value !== null && typeof value === "object" && STOP_RESPONSE_SYMBOL in value;
|
|
49
49
|
}
|
|
50
50
|
class Agent {
|
|
51
|
+
_id;
|
|
51
52
|
turnDetection;
|
|
52
53
|
_stt;
|
|
53
54
|
_vad;
|
|
@@ -62,6 +63,7 @@ class Agent {
|
|
|
62
63
|
/** @internal */
|
|
63
64
|
_tools;
|
|
64
65
|
constructor({
|
|
66
|
+
id,
|
|
65
67
|
instructions,
|
|
66
68
|
chatCtx,
|
|
67
69
|
tools,
|
|
@@ -71,6 +73,16 @@ class Agent {
|
|
|
71
73
|
llm,
|
|
72
74
|
tts
|
|
73
75
|
}) {
|
|
76
|
+
if (id) {
|
|
77
|
+
this._id = id;
|
|
78
|
+
} else {
|
|
79
|
+
const className = this.constructor.name;
|
|
80
|
+
if (className === "Agent") {
|
|
81
|
+
this._id = "default_agent";
|
|
82
|
+
} else {
|
|
83
|
+
this._id = className.replace(/([A-Z])/g, "_$1").toLowerCase().replace(/^_/, "");
|
|
84
|
+
}
|
|
85
|
+
}
|
|
74
86
|
this._instructions = instructions;
|
|
75
87
|
this._tools = { ...tools };
|
|
76
88
|
this._chatCtx = chatCtx ? chatCtx.copy({
|
|
@@ -110,6 +122,9 @@ class Agent {
|
|
|
110
122
|
get chatCtx() {
|
|
111
123
|
return new import_chat_context.ReadonlyChatContext(this._chatCtx.items);
|
|
112
124
|
}
|
|
125
|
+
get id() {
|
|
126
|
+
return this._id;
|
|
127
|
+
}
|
|
113
128
|
get instructions() {
|
|
114
129
|
return this._instructions;
|
|
115
130
|
}
|
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 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 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 instructions,\n chatCtx,\n tools,\n turnDetection,\n stt,\n vad,\n llm,\n tts,\n }: AgentOptions<UserData>) {\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 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 if (!agent.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, agent.vad);\n }\n\n const stream = wrapped_stt.stream();\n stream.updateInputStream(audio);\n\n return new ReadableStream({\n async start(controller) {\n for await (const event of stream) {\n controller.enqueue(event);\n }\n controller.close();\n },\n cancel() {\n stream.detachInputStream();\n stream.close();\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 // TODO(brian): make parallelToolCalls configurable\n const { toolChoice } = modelSettings;\n\n const stream = activity.llm.chat({\n chatCtx,\n toolCtx,\n toolChoice,\n parallelToolCalls: true,\n });\n return new ReadableStream({\n async start(controller) {\n for await (const chunk of stream) {\n controller.enqueue(chunk);\n }\n controller.close();\n },\n cancel() {\n stream.close();\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 stream = wrapped_tts.stream();\n stream.updateInputStream(text);\n\n return new ReadableStream({\n async start(controller) {\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 },\n cancel() {\n stream.close();\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;AAoBO,MAAM,MAAsB;AAAA,EACzB;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,EACF,GAA2B;AACzB,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,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;AACtD,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,YAAI,CAAC,MAAM,KAAK;AACd,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,sBAAc,IAAI,WAAAC,cAAiB,aAAa,MAAM,GAAG;AAAA,MAC3D;AAEA,YAAM,SAAS,YAAY,OAAO;AAClC,aAAO,kBAAkB,KAAK;AAE9B,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,uBAAW,QAAQ,KAAK;AAAA,UAC1B;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,kBAAkB;AACzB,iBAAO,MAAM;AAAA,QACf;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;AAGA,YAAM,EAAE,WAAW,IAAI;AAEvB,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA,mBAAmB;AAAA,MACrB,CAAC;AACD,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,uBAAW,QAAQ,KAAK;AAAA,UAC1B;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,MAAM;AAAA,QACf;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,SAAS,YAAY,OAAO;AAClC,aAAO,kBAAkB,IAAI;AAE7B,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,gBAAI,UAAU,4BAAiB,eAAe;AAC5C;AAAA,YACF;AACA,uBAAW,QAAQ,MAAM,KAAK;AAAA,UAChC;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,MAAM;AAAA,QACf;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 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 if (!agent.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, agent.vad);\n }\n\n const stream = wrapped_stt.stream();\n stream.updateInputStream(audio);\n\n return new ReadableStream({\n async start(controller) {\n for await (const event of stream) {\n controller.enqueue(event);\n }\n controller.close();\n },\n cancel() {\n stream.detachInputStream();\n stream.close();\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 // TODO(brian): make parallelToolCalls configurable\n const { toolChoice } = modelSettings;\n\n const stream = activity.llm.chat({\n chatCtx,\n toolCtx,\n toolChoice,\n parallelToolCalls: true,\n });\n return new ReadableStream({\n async start(controller) {\n for await (const chunk of stream) {\n controller.enqueue(chunk);\n }\n controller.close();\n },\n cancel() {\n stream.close();\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 stream = wrapped_tts.stream();\n stream.updateInputStream(text);\n\n return new ReadableStream({\n async start(controller) {\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 },\n cancel() {\n stream.close();\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;AACtD,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,YAAI,CAAC,MAAM,KAAK;AACd,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,sBAAc,IAAI,WAAAC,cAAiB,aAAa,MAAM,GAAG;AAAA,MAC3D;AAEA,YAAM,SAAS,YAAY,OAAO;AAClC,aAAO,kBAAkB,KAAK;AAE9B,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,uBAAW,QAAQ,KAAK;AAAA,UAC1B;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,kBAAkB;AACzB,iBAAO,MAAM;AAAA,QACf;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;AAGA,YAAM,EAAE,WAAW,IAAI;AAEvB,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA,mBAAmB;AAAA,MACrB,CAAC;AACD,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,uBAAW,QAAQ,KAAK;AAAA,UAC1B;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,MAAM;AAAA,QACf;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,SAAS,YAAY,OAAO;AAClC,aAAO,kBAAkB,IAAI;AAE7B,aAAO,IAAI,0BAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,gBAAI,UAAU,4BAAiB,eAAe;AAC5C;AAAA,YACF;AACA,uBAAW,QAAQ,MAAM,KAAK;AAAA,UAChC;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,MAAM;AAAA,QACf;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"]}
|
package/dist/voice/agent.d.cts
CHANGED
|
@@ -25,6 +25,7 @@ export interface ModelSettings {
|
|
|
25
25
|
toolChoice?: ToolChoice;
|
|
26
26
|
}
|
|
27
27
|
export interface AgentOptions<UserData> {
|
|
28
|
+
id?: string;
|
|
28
29
|
instructions: string;
|
|
29
30
|
chatCtx?: ChatContext;
|
|
30
31
|
tools?: ToolContext<UserData>;
|
|
@@ -37,6 +38,7 @@ export interface AgentOptions<UserData> {
|
|
|
37
38
|
minConsecutiveSpeechDelay?: number;
|
|
38
39
|
}
|
|
39
40
|
export declare class Agent<UserData = any> {
|
|
41
|
+
private _id;
|
|
40
42
|
private turnDetection?;
|
|
41
43
|
private _stt?;
|
|
42
44
|
private _vad?;
|
|
@@ -50,12 +52,13 @@ export declare class Agent<UserData = any> {
|
|
|
50
52
|
_instructions: string;
|
|
51
53
|
/** @internal */
|
|
52
54
|
_tools?: ToolContext<UserData>;
|
|
53
|
-
constructor({ instructions, chatCtx, tools, turnDetection, stt, vad, llm, tts, }: AgentOptions<UserData>);
|
|
55
|
+
constructor({ id, instructions, chatCtx, tools, turnDetection, stt, vad, llm, tts, }: AgentOptions<UserData>);
|
|
54
56
|
get vad(): VAD | undefined;
|
|
55
57
|
get stt(): STT | undefined;
|
|
56
58
|
get llm(): LLM | RealtimeModel | undefined;
|
|
57
59
|
get tts(): TTS | undefined;
|
|
58
60
|
get chatCtx(): ReadonlyChatContext;
|
|
61
|
+
get id(): string;
|
|
59
62
|
get instructions(): string;
|
|
60
63
|
get toolCtx(): ToolContext<UserData>;
|
|
61
64
|
get session(): AgentSession<UserData>;
|
package/dist/voice/agent.d.ts
CHANGED
|
@@ -25,6 +25,7 @@ export interface ModelSettings {
|
|
|
25
25
|
toolChoice?: ToolChoice;
|
|
26
26
|
}
|
|
27
27
|
export interface AgentOptions<UserData> {
|
|
28
|
+
id?: string;
|
|
28
29
|
instructions: string;
|
|
29
30
|
chatCtx?: ChatContext;
|
|
30
31
|
tools?: ToolContext<UserData>;
|
|
@@ -37,6 +38,7 @@ export interface AgentOptions<UserData> {
|
|
|
37
38
|
minConsecutiveSpeechDelay?: number;
|
|
38
39
|
}
|
|
39
40
|
export declare class Agent<UserData = any> {
|
|
41
|
+
private _id;
|
|
40
42
|
private turnDetection?;
|
|
41
43
|
private _stt?;
|
|
42
44
|
private _vad?;
|
|
@@ -50,12 +52,13 @@ export declare class Agent<UserData = any> {
|
|
|
50
52
|
_instructions: string;
|
|
51
53
|
/** @internal */
|
|
52
54
|
_tools?: ToolContext<UserData>;
|
|
53
|
-
constructor({ instructions, chatCtx, tools, turnDetection, stt, vad, llm, tts, }: AgentOptions<UserData>);
|
|
55
|
+
constructor({ id, instructions, chatCtx, tools, turnDetection, stt, vad, llm, tts, }: AgentOptions<UserData>);
|
|
54
56
|
get vad(): VAD | undefined;
|
|
55
57
|
get stt(): STT | undefined;
|
|
56
58
|
get llm(): LLM | RealtimeModel | undefined;
|
|
57
59
|
get tts(): TTS | undefined;
|
|
58
60
|
get chatCtx(): ReadonlyChatContext;
|
|
61
|
+
get id(): string;
|
|
59
62
|
get instructions(): string;
|
|
60
63
|
get toolCtx(): ToolContext<UserData>;
|
|
61
64
|
get session(): AgentSession<UserData>;
|
|
@@ -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;AAE3C,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;AAE1E,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,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;CACpC;AAED,qBAAa,KAAK,CAAC,QAAQ,GAAG,GAAG;IAC/B,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;IAEnB,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,YAAY,EACZ,OAAO,EACP,KAAK,EACL,aAAa,EACb,GAAG,EACH,GAAG,EACH,GAAG,EACH,GAAG,GACJ,EAAE,YAAY,CAAC,QAAQ,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;AAE3C,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;AAE1E,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;CACpC;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;IAEnB,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,GACJ,EAAE,YAAY,CAAC,QAAQ,CAAC;IAgDzB,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,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,CAAC,EAC5B,aAAa,EAAE,aAAa,GAC3B,OAAO,CAAC,cAAc,CAAC,MAAM,CAAC,GAAG,IAAI,CAAC;IAInC,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;uBAmC9C,KAAK,WACH,WAAW,WACX,WAAW,iBACL,aAAa,GAC3B,QAAQ,eAAe,SAAS,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC;uBAmC5C,KAAK,QACN,eAAe,MAAM,CAAC,kBACZ,aAAa,GAC5B,QAAQ,eAAe,UAAU,CAAC,GAAG,IAAI,CAAC;iCAgCpC,KAAK,QACN,eAAe,MAAM,CAAC,kBACZ,aAAa,GAC5B,QAAQ,eAAe,MAAM,CAAC,GAAG,IAAI,CAAC;wCAK/B,KAAK,SACN,eAAe,UAAU,CAAC,kBACjB,aAAa,GAC5B,QAAQ,eAAe,UAAU,CAAC,GAAG,IAAI,CAAC;MAG7C;CACH"}
|
package/dist/voice/agent.js
CHANGED
|
@@ -28,6 +28,7 @@ function isStopResponse(value) {
|
|
|
28
28
|
return value !== void 0 && value !== null && typeof value === "object" && STOP_RESPONSE_SYMBOL in value;
|
|
29
29
|
}
|
|
30
30
|
class Agent {
|
|
31
|
+
_id;
|
|
31
32
|
turnDetection;
|
|
32
33
|
_stt;
|
|
33
34
|
_vad;
|
|
@@ -42,6 +43,7 @@ class Agent {
|
|
|
42
43
|
/** @internal */
|
|
43
44
|
_tools;
|
|
44
45
|
constructor({
|
|
46
|
+
id,
|
|
45
47
|
instructions,
|
|
46
48
|
chatCtx,
|
|
47
49
|
tools,
|
|
@@ -51,6 +53,16 @@ class Agent {
|
|
|
51
53
|
llm,
|
|
52
54
|
tts
|
|
53
55
|
}) {
|
|
56
|
+
if (id) {
|
|
57
|
+
this._id = id;
|
|
58
|
+
} else {
|
|
59
|
+
const className = this.constructor.name;
|
|
60
|
+
if (className === "Agent") {
|
|
61
|
+
this._id = "default_agent";
|
|
62
|
+
} else {
|
|
63
|
+
this._id = className.replace(/([A-Z])/g, "_$1").toLowerCase().replace(/^_/, "");
|
|
64
|
+
}
|
|
65
|
+
}
|
|
54
66
|
this._instructions = instructions;
|
|
55
67
|
this._tools = { ...tools };
|
|
56
68
|
this._chatCtx = chatCtx ? chatCtx.copy({
|
|
@@ -90,6 +102,9 @@ class Agent {
|
|
|
90
102
|
get chatCtx() {
|
|
91
103
|
return new ReadonlyChatContext(this._chatCtx.items);
|
|
92
104
|
}
|
|
105
|
+
get id() {
|
|
106
|
+
return this._id;
|
|
107
|
+
}
|
|
93
108
|
get instructions() {
|
|
94
109
|
return this._instructions;
|
|
95
110
|
}
|
package/dist/voice/agent.js.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 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 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 instructions,\n chatCtx,\n tools,\n turnDetection,\n stt,\n vad,\n llm,\n tts,\n }: AgentOptions<UserData>) {\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 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 if (!agent.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, agent.vad);\n }\n\n const stream = wrapped_stt.stream();\n stream.updateInputStream(audio);\n\n return new ReadableStream({\n async start(controller) {\n for await (const event of stream) {\n controller.enqueue(event);\n }\n controller.close();\n },\n cancel() {\n stream.detachInputStream();\n stream.close();\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 // TODO(brian): make parallelToolCalls configurable\n const { toolChoice } = modelSettings;\n\n const stream = activity.llm.chat({\n chatCtx,\n toolCtx,\n toolChoice,\n parallelToolCalls: true,\n });\n return new ReadableStream({\n async start(controller) {\n for await (const chunk of stream) {\n controller.enqueue(chunk);\n }\n controller.close();\n },\n cancel() {\n stream.close();\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 stream = wrapped_tts.stream();\n stream.updateInputStream(text);\n\n return new ReadableStream({\n async start(controller) {\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 },\n cancel() {\n stream.close();\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":"AAIA,SAAS,yBAAyB;AAClC,SAAS,sBAAsB;AAC/B;AAAA,EACE,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,OAIF;AACP,SAAS,2BAA2B;AAEpC;AAAA,EAEE;AAAA,EACA;AAAA,OAGK;AAEP,SAAS,iBAAiB,wBAAwB;AAClD,SAAS,qBAAqB,8BAA8B;AAE5D,SAAS,kBAAkB,iBAAiB,wBAAwB;AAK7D,MAAM,oBAAoB,IAAI,kBAAmD;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;AAoBO,MAAM,MAAsB;AAAA,EACzB;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,EACF,GAA2B;AACzB,SAAK,gBAAgB;AACrB,SAAK,SAAS,EAAE,GAAG,MAAM;AACzB,SAAK,WAAW,UACZ,QAAQ,KAAK;AAAA,MACX,SAAS,KAAK;AAAA,IAChB,CAAC,IACD,YAAY,MAAM;AAEtB,SAAK,gBAAgB;AACrB,SAAK,OAAO;AAEZ,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,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,oBAAoB,KAAK,SAAS,KAAK;AAAA,EACpD;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;AACtD,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,YAAI,CAAC,MAAM,KAAK;AACd,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,sBAAc,IAAI,iBAAiB,aAAa,MAAM,GAAG;AAAA,MAC3D;AAEA,YAAM,SAAS,YAAY,OAAO;AAClC,aAAO,kBAAkB,KAAK;AAE9B,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,uBAAW,QAAQ,KAAK;AAAA,UAC1B;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,kBAAkB;AACzB,iBAAO,MAAM;AAAA,QACf;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,MAAM;AAClC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAGA,YAAM,EAAE,WAAW,IAAI;AAEvB,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA,mBAAmB;AAAA,MACrB,CAAC;AACD,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,uBAAW,QAAQ,KAAK;AAAA,UAC1B;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,MAAM;AAAA,QACf;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,iBAAiB,aAAa,IAAI,uBAAuB,CAAC;AAAA,MAC9E;AAEA,YAAM,SAAS,YAAY,OAAO;AAClC,aAAO,kBAAkB,IAAI;AAE7B,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,gBAAI,UAAU,iBAAiB,eAAe;AAC5C;AAAA,YACF;AACA,uBAAW,QAAQ,MAAM,KAAK;AAAA,UAChC;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,MAAM;AAAA,QACf;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":[]}
|
|
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 if (!agent.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, agent.vad);\n }\n\n const stream = wrapped_stt.stream();\n stream.updateInputStream(audio);\n\n return new ReadableStream({\n async start(controller) {\n for await (const event of stream) {\n controller.enqueue(event);\n }\n controller.close();\n },\n cancel() {\n stream.detachInputStream();\n stream.close();\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 // TODO(brian): make parallelToolCalls configurable\n const { toolChoice } = modelSettings;\n\n const stream = activity.llm.chat({\n chatCtx,\n toolCtx,\n toolChoice,\n parallelToolCalls: true,\n });\n return new ReadableStream({\n async start(controller) {\n for await (const chunk of stream) {\n controller.enqueue(chunk);\n }\n controller.close();\n },\n cancel() {\n stream.close();\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 stream = wrapped_tts.stream();\n stream.updateInputStream(text);\n\n return new ReadableStream({\n async start(controller) {\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 },\n cancel() {\n stream.close();\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":"AAIA,SAAS,yBAAyB;AAClC,SAAS,sBAAsB;AAC/B;AAAA,EACE,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,OAIF;AACP,SAAS,2BAA2B;AAEpC;AAAA,EAEE;AAAA,EACA;AAAA,OAGK;AAEP,SAAS,iBAAiB,wBAAwB;AAClD,SAAS,qBAAqB,8BAA8B;AAE5D,SAAS,kBAAkB,iBAAiB,wBAAwB;AAK7D,MAAM,oBAAoB,IAAI,kBAAmD;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,YAAY,MAAM;AAEtB,SAAK,gBAAgB;AACrB,SAAK,OAAO;AAEZ,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,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,oBAAoB,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;AACtD,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,YAAI,CAAC,MAAM,KAAK;AACd,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,sBAAc,IAAI,iBAAiB,aAAa,MAAM,GAAG;AAAA,MAC3D;AAEA,YAAM,SAAS,YAAY,OAAO;AAClC,aAAO,kBAAkB,KAAK;AAE9B,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,uBAAW,QAAQ,KAAK;AAAA,UAC1B;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,kBAAkB;AACzB,iBAAO,MAAM;AAAA,QACf;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,MAAM;AAClC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAGA,YAAM,EAAE,WAAW,IAAI;AAEvB,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA,mBAAmB;AAAA,MACrB,CAAC;AACD,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,uBAAW,QAAQ,KAAK;AAAA,UAC1B;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,MAAM;AAAA,QACf;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,iBAAiB,aAAa,IAAI,uBAAuB,CAAC;AAAA,MAC9E;AAEA,YAAM,SAAS,YAAY,OAAO;AAClC,aAAO,kBAAkB,IAAI;AAE7B,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,gBAAI,UAAU,iBAAiB,eAAe;AAC5C;AAAA,YACF;AACA,uBAAW,QAAQ,MAAM,KAAK;AAAA,UAChC;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,MAAM;AAAA,QACf;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":[]}
|
|
@@ -903,6 +903,7 @@ ${instructions}` : instructions,
|
|
|
903
903
|
this.agentSession._updateAgentState("listening");
|
|
904
904
|
}
|
|
905
905
|
}
|
|
906
|
+
// TODO(brian): PR3 - Wrap entire pipelineReplyTask() method with tracer.startActiveSpan('agent_turn')
|
|
906
907
|
async pipelineReplyTask(speechHandle, chatCtx, toolCtx, modelSettings, replyAbortController, instructions, newMessage, toolsMessages) {
|
|
907
908
|
var _a, _b, _c;
|
|
908
909
|
speechHandleStorage.enterWith(speechHandle);
|
|
@@ -1559,6 +1560,7 @@ ${instructions}` : instructions,
|
|
|
1559
1560
|
speechHandle._markScheduled();
|
|
1560
1561
|
this.wakeupMainTask();
|
|
1561
1562
|
}
|
|
1563
|
+
// TODO(brian): PR3 - Wrap entire drain() method with tracer.startActiveSpan('drain_agent_activity', { attributes: { 'lk.agent_label': this.agent.label } })
|
|
1562
1564
|
async drain() {
|
|
1563
1565
|
var _a;
|
|
1564
1566
|
const unlock = await this.lock.lock();
|
|
@@ -1599,12 +1601,15 @@ ${instructions}` : instructions,
|
|
|
1599
1601
|
}
|
|
1600
1602
|
if (this.stt instanceof import_stt.STT) {
|
|
1601
1603
|
this.stt.off("metrics_collected", this.onMetricsCollected);
|
|
1604
|
+
await this.stt.close();
|
|
1602
1605
|
}
|
|
1603
1606
|
if (this.tts instanceof import_tts.TTS) {
|
|
1604
1607
|
this.tts.off("metrics_collected", this.onMetricsCollected);
|
|
1608
|
+
await this.tts.close();
|
|
1605
1609
|
}
|
|
1606
1610
|
if (this.vad instanceof import_vad.VAD) {
|
|
1607
1611
|
this.vad.off("metrics_collected", this.onMetricsCollected);
|
|
1612
|
+
await this.vad.close();
|
|
1608
1613
|
}
|
|
1609
1614
|
this.detachAudioInput();
|
|
1610
1615
|
await ((_a = this.realtimeSession) == null ? void 0 : _a.close());
|