@livekit/agents 0.5.1 → 0.6.0
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.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/llm/index.cjs +2 -0
- package/dist/llm/index.cjs.map +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 +47 -3
- package/dist/llm/llm.cjs.map +1 -1
- package/dist/llm/llm.d.ts +15 -2
- package/dist/llm/llm.d.ts.map +1 -1
- package/dist/llm/llm.js +46 -3
- package/dist/llm/llm.js.map +1 -1
- package/dist/metrics/base.cjs +44 -0
- package/dist/metrics/base.cjs.map +1 -0
- package/dist/metrics/base.d.ts +96 -0
- package/dist/metrics/base.d.ts.map +1 -0
- package/dist/metrics/base.js +20 -0
- package/dist/metrics/base.js.map +1 -0
- package/dist/metrics/index.cjs +35 -0
- package/dist/metrics/index.cjs.map +1 -0
- package/dist/metrics/index.d.ts +5 -0
- package/dist/metrics/index.d.ts.map +1 -0
- package/dist/metrics/index.js +9 -0
- package/dist/metrics/index.js.map +1 -0
- package/dist/metrics/usage_collector.cjs +53 -0
- package/dist/metrics/usage_collector.cjs.map +1 -0
- package/dist/metrics/usage_collector.d.ts +14 -0
- package/dist/metrics/usage_collector.d.ts.map +1 -0
- package/dist/metrics/usage_collector.js +29 -0
- package/dist/metrics/usage_collector.js.map +1 -0
- package/dist/metrics/utils.cjs +104 -0
- package/dist/metrics/utils.cjs.map +1 -0
- package/dist/metrics/utils.d.ts +10 -0
- package/dist/metrics/utils.d.ts.map +1 -0
- package/dist/metrics/utils.js +73 -0
- package/dist/metrics/utils.js.map +1 -0
- package/dist/multimodal/multimodal_agent.cjs +7 -13
- package/dist/multimodal/multimodal_agent.cjs.map +1 -1
- package/dist/multimodal/multimodal_agent.d.ts +1 -4
- package/dist/multimodal/multimodal_agent.d.ts.map +1 -1
- package/dist/multimodal/multimodal_agent.js +7 -13
- package/dist/multimodal/multimodal_agent.js.map +1 -1
- package/dist/pipeline/agent_output.cjs +9 -2
- package/dist/pipeline/agent_output.cjs.map +1 -1
- package/dist/pipeline/agent_output.d.ts +1 -0
- package/dist/pipeline/agent_output.d.ts.map +1 -1
- package/dist/pipeline/agent_output.js +9 -2
- package/dist/pipeline/agent_output.js.map +1 -1
- package/dist/pipeline/index.cjs +2 -0
- package/dist/pipeline/index.cjs.map +1 -1
- package/dist/pipeline/index.d.ts +1 -1
- package/dist/pipeline/index.d.ts.map +1 -1
- package/dist/pipeline/index.js +3 -1
- package/dist/pipeline/index.js.map +1 -1
- package/dist/pipeline/pipeline_agent.cjs +168 -70
- package/dist/pipeline/pipeline_agent.cjs.map +1 -1
- package/dist/pipeline/pipeline_agent.d.ts +10 -4
- package/dist/pipeline/pipeline_agent.d.ts.map +1 -1
- package/dist/pipeline/pipeline_agent.js +171 -73
- package/dist/pipeline/pipeline_agent.js.map +1 -1
- package/dist/pipeline/speech_handle.cjs +49 -1
- package/dist/pipeline/speech_handle.cjs.map +1 -1
- package/dist/pipeline/speech_handle.d.ts +12 -2
- package/dist/pipeline/speech_handle.d.ts.map +1 -1
- package/dist/pipeline/speech_handle.js +50 -2
- package/dist/pipeline/speech_handle.js.map +1 -1
- package/dist/stt/index.cjs.map +1 -1
- package/dist/stt/index.d.ts +1 -1
- package/dist/stt/index.d.ts.map +1 -1
- package/dist/stt/index.js.map +1 -1
- package/dist/stt/stream_adapter.cjs +15 -5
- package/dist/stt/stream_adapter.cjs.map +1 -1
- package/dist/stt/stream_adapter.d.ts +4 -1
- package/dist/stt/stream_adapter.d.ts.map +1 -1
- package/dist/stt/stream_adapter.js +15 -5
- package/dist/stt/stream_adapter.js.map +1 -1
- package/dist/stt/stt.cjs +46 -2
- package/dist/stt/stt.cjs.map +1 -1
- package/dist/stt/stt.d.ts +25 -3
- package/dist/stt/stt.d.ts.map +1 -1
- package/dist/stt/stt.js +46 -2
- package/dist/stt/stt.js.map +1 -1
- package/dist/tts/index.cjs +4 -2
- package/dist/tts/index.cjs.map +1 -1
- package/dist/tts/index.d.ts +1 -1
- package/dist/tts/index.d.ts.map +1 -1
- package/dist/tts/index.js +3 -1
- package/dist/tts/index.js.map +1 -1
- package/dist/tts/stream_adapter.cjs +14 -3
- package/dist/tts/stream_adapter.cjs.map +1 -1
- package/dist/tts/stream_adapter.d.ts +3 -0
- package/dist/tts/stream_adapter.d.ts.map +1 -1
- package/dist/tts/stream_adapter.js +15 -4
- package/dist/tts/stream_adapter.js.map +1 -1
- package/dist/tts/tts.cjs +109 -6
- package/dist/tts/tts.cjs.map +1 -1
- package/dist/tts/tts.d.ts +24 -1
- package/dist/tts/tts.d.ts.map +1 -1
- package/dist/tts/tts.js +107 -5
- package/dist/tts/tts.js.map +1 -1
- package/dist/vad.cjs +43 -2
- package/dist/vad.cjs.map +1 -1
- package/dist/vad.d.ts +21 -4
- package/dist/vad.d.ts.map +1 -1
- package/dist/vad.js +43 -2
- package/dist/vad.js.map +1 -1
- package/package.json +1 -1
- package/src/index.ts +2 -1
- package/src/llm/index.ts +2 -0
- package/src/llm/llm.ts +55 -3
- package/src/metrics/base.ts +127 -0
- package/src/metrics/index.ts +20 -0
- package/src/metrics/usage_collector.ts +40 -0
- package/src/metrics/utils.ts +100 -0
- package/src/multimodal/multimodal_agent.ts +12 -17
- package/src/pipeline/agent_output.ts +14 -7
- package/src/pipeline/index.ts +1 -1
- package/src/pipeline/pipeline_agent.ts +210 -95
- package/src/pipeline/speech_handle.ts +67 -2
- package/src/stt/index.ts +2 -0
- package/src/stt/stream_adapter.ts +17 -5
- package/src/stt/stt.ts +67 -3
- package/src/tts/index.ts +2 -0
- package/src/tts/stream_adapter.ts +17 -4
- package/src/tts/tts.ts +127 -4
- package/src/vad.ts +61 -4
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/multimodal/multimodal_agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n LocalTrackPublication,\n RemoteAudioTrack,\n RemoteParticipant,\n RemoteTrack,\n RemoteTrackPublication,\n Room,\n} from '@livekit/rtc-node';\nimport {\n AudioSource,\n AudioStream,\n LocalAudioTrack,\n RoomEvent,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport * as llm from '../llm/index.js';\nimport { log } from '../log.js';\nimport { BasicTranscriptionForwarder } from '../transcription.js';\nimport { findMicroTrackId } from '../utils.js';\nimport { AgentPlayout, type PlayoutHandle } from './agent_playout.js';\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeSession extends EventEmitter {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract conversation: any; // openai.realtime.Conversation\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract inputAudioBuffer: any; // openai.realtime.InputAudioBuffer\n abstract fncCtx: llm.FunctionContext | undefined;\n}\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeModel {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract session(options: any): RealtimeSession; // openai.realtime.ModelOptions\n abstract close(): Promise<void>;\n abstract sampleRate: number;\n abstract numChannels: number;\n abstract inFrameSize: number;\n abstract outFrameSize: number;\n}\n\nexport type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';\nexport const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';\n\n/** @beta */\nexport class MultimodalAgent extends EventEmitter {\n model: RealtimeModel;\n room: Room | null = null;\n linkedParticipant: RemoteParticipant | null = null;\n subscribedTrack: RemoteAudioTrack | null = null;\n readMicroTask: { promise: Promise<void>; cancel: () => void } | null = null;\n\n constructor({\n model,\n chatCtx,\n fncCtx,\n }: {\n model: RealtimeModel;\n chatCtx?: llm.ChatContext;\n fncCtx?: llm.FunctionContext;\n }) {\n super();\n this.model = model;\n this.#chatCtx = chatCtx;\n this.#fncCtx = fncCtx;\n }\n\n #participant: RemoteParticipant | string | null = null;\n #agentPublication: LocalTrackPublication | null = null;\n #localTrackSid: string | null = null;\n #localSource: AudioSource | null = null;\n #agentPlayout: AgentPlayout | null = null;\n #playingHandle: PlayoutHandle | undefined = undefined;\n #logger = log();\n #session: RealtimeSession | null = null;\n #fncCtx: llm.FunctionContext | undefined = undefined;\n #chatCtx: llm.ChatContext | undefined = undefined;\n\n #_started: boolean = false;\n #_pendingFunctionCalls: Set<string> = new Set();\n #_speaking: boolean = false;\n\n get fncCtx(): llm.FunctionContext | undefined {\n return this.#fncCtx;\n }\n\n set fncCtx(ctx: llm.FunctionContext | undefined) {\n this.#fncCtx = ctx;\n if (this.#session) {\n this.#session.fncCtx = ctx;\n }\n }\n\n get #pendingFunctionCalls(): Set<string> {\n return this.#_pendingFunctionCalls;\n }\n\n set #pendingFunctionCalls(calls: Set<string>) {\n this.#_pendingFunctionCalls = calls;\n this.#updateState();\n }\n\n get #speaking(): boolean {\n return this.#_speaking;\n }\n\n set #speaking(isSpeaking: boolean) {\n this.#_speaking = isSpeaking;\n this.#updateState();\n }\n\n get #started(): boolean {\n return this.#_started;\n }\n\n set #started(started: boolean) {\n this.#_started = started;\n this.#updateState();\n }\n\n start(\n room: Room,\n participant: RemoteParticipant | string | null = null,\n ): Promise<RealtimeSession> {\n return new Promise(async (resolve, reject) => {\n if (this.#started) {\n reject(new Error('MultimodalAgent already started'));\n }\n this.#updateState();\n\n room.on(RoomEvent.ParticipantConnected, (participant: RemoteParticipant) => {\n // automatically link to the first participant that connects, if not already linked\n if (this.linkedParticipant) {\n return;\n }\n this.#linkParticipant(participant.identity);\n });\n room.on(\n RoomEvent.TrackPublished,\n (trackPublication: RemoteTrackPublication, participant: RemoteParticipant) => {\n if (\n this.linkedParticipant &&\n participant.identity === this.linkedParticipant.identity &&\n trackPublication.source === TrackSource.SOURCE_MICROPHONE &&\n !trackPublication.subscribed\n ) {\n trackPublication.setSubscribed(true);\n }\n },\n );\n room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));\n\n this.room = room;\n this.#participant = participant;\n\n this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);\n this.#agentPlayout = new AgentPlayout(\n this.#localSource,\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n this.model.outFrameSize,\n );\n const onPlayoutStarted = () => {\n this.emit('agent_started_speaking');\n this.#speaking = true;\n };\n\n const onPlayoutStopped = (interrupted: boolean) => {\n this.emit('agent_stopped_speaking');\n this.#speaking = false;\n if (this.#playingHandle) {\n let text = this.#playingHandle.transcriptionFwd.text;\n if (interrupted) {\n text += '…';\n }\n const msg = llm.ChatMessage.create({\n role: llm.ChatRole.ASSISTANT,\n text,\n });\n\n if (interrupted) {\n this.emit('agent_speech_interrupted', msg);\n } else {\n this.emit('agent_speech_committed', msg);\n }\n this.#logger.child({ transcription: text, interrupted }).debug('committed agent speech');\n }\n };\n\n this.#agentPlayout.on('playout_started', onPlayoutStarted);\n this.#agentPlayout.on('playout_stopped', onPlayoutStopped);\n\n const track = LocalAudioTrack.createAudioTrack('assistant_voice', this.#localSource);\n const options = new TrackPublishOptions();\n options.source = TrackSource.SOURCE_MICROPHONE;\n this.#agentPublication = (await room.localParticipant?.publishTrack(track, options)) || null;\n if (!this.#agentPublication) {\n this.#logger.error('Failed to publish track');\n reject(new Error('Failed to publish track'));\n return;\n }\n\n await this.#agentPublication.waitForSubscription();\n\n if (participant) {\n if (typeof participant === 'string') {\n this.#linkParticipant(participant);\n } else {\n this.#linkParticipant(participant.identity);\n }\n } else {\n // No participant specified, try to find the first participant in the room\n for (const participant of room.remoteParticipants.values()) {\n this.#linkParticipant(participant.identity);\n break;\n }\n }\n\n this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });\n this.#started = true;\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_added', (message: any) => {\n // openai.realtime.RealtimeContent\n const trFwd = new BasicTranscriptionForwarder(\n this.room!,\n this.room!.localParticipant!.identity,\n this.#getLocalTrackSid()!,\n message.responseId,\n );\n\n const handle = this.#agentPlayout?.play(\n message.itemId,\n message.contentIndex,\n trFwd,\n message.textStream,\n message.audioStream,\n );\n this.#playingHandle = handle;\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_committed', (ev: any) => {\n // openai.realtime.InputSpeechCommittedEvent\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_transcription_completed', (ev: any) => {\n // openai.realtime.InputSpeechTranscriptionCompletedEvent\n const transcription = ev.transcript;\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, transcription, true, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n const userMsg = llm.ChatMessage.create({\n role: llm.ChatRole.USER,\n text: transcription,\n });\n this.emit('user_speech_committed', userMsg);\n this.#logger.child({ transcription }).debug('committed user speech');\n });\n\n this.#session.on('input_speech_started', (ev: any) => {\n if (this.#playingHandle && !this.#playingHandle.done) {\n this.#playingHandle.interrupt();\n\n this.#session!.conversation.item.truncate(\n this.#playingHandle.itemId,\n this.#playingHandle.contentIndex,\n Math.floor((this.#playingHandle.audioSamples / 24000) * 1000),\n );\n\n this.#playingHandle = undefined;\n }\n\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n this.#session.on('input_speech_stopped', (ev: any) => {\n this.emit('user_stopped_speaking');\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_started', (ev: any) => {\n this.#pendingFunctionCalls.add(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_completed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_failed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n resolve(this.#session);\n });\n }\n\n #linkParticipant(participantIdentity: string): void {\n if (!this.room) {\n this.#logger.error('Room is not set');\n return;\n }\n\n this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;\n if (!this.linkedParticipant) {\n this.#logger.error(`Participant with identity ${participantIdentity} not found`);\n return;\n }\n\n if (this.linkedParticipant.trackPublications.size > 0) {\n this.#subscribeToMicrophone();\n }\n\n // also check if already subscribed\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {\n this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);\n break;\n }\n }\n }\n\n #subscribeToMicrophone(): void {\n if (!this.linkedParticipant) {\n this.#logger.error('Participant is not set');\n return;\n }\n\n let microphonePublication: RemoteTrackPublication | undefined = undefined;\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n microphonePublication = publication;\n break;\n }\n }\n if (!microphonePublication) {\n return;\n }\n\n if (!microphonePublication.subscribed) {\n microphonePublication.setSubscribed(true);\n }\n }\n\n #handleTrackSubscription(\n track: RemoteTrack,\n publication: RemoteTrackPublication,\n participant: RemoteParticipant,\n ) {\n if (\n publication.source !== TrackSource.SOURCE_MICROPHONE ||\n participant.identity !== this.linkedParticipant?.identity\n ) {\n return;\n }\n const readAudioStreamTask = async (audioStream: AudioStream) => {\n const bstream = new AudioByteStream(\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n );\n\n for await (const frame of audioStream) {\n const audioData = frame.data;\n for (const frame of bstream.write(audioData.buffer)) {\n this.#session!.inputAudioBuffer.append(frame);\n }\n }\n };\n this.subscribedTrack = track;\n\n if (this.readMicroTask) {\n this.readMicroTask.cancel();\n }\n\n let cancel: () => void;\n this.readMicroTask = {\n promise: new Promise<void>((resolve, reject) => {\n cancel = () => {\n reject(new Error('Task cancelled'));\n };\n readAudioStreamTask(new AudioStream(track, this.model.sampleRate, this.model.numChannels))\n .then(resolve)\n .catch(reject);\n }),\n cancel: () => cancel(),\n };\n }\n\n #getLocalTrackSid(): string | null {\n if (!this.#localTrackSid && this.room && this.room.localParticipant) {\n this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant?.identity);\n }\n return this.#localTrackSid;\n }\n\n #publishTranscription(\n participantIdentity: string,\n trackSid: string,\n text: string,\n isFinal: boolean,\n id: string,\n ): void {\n this.#logger.debug(\n `Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`,\n );\n if (!this.room?.localParticipant) {\n this.#logger.error('Room or local participant not set');\n return;\n }\n\n this.room.localParticipant.publishTranscription({\n participantIdentity,\n trackSid,\n segments: [\n {\n text,\n final: isFinal,\n id,\n startTime: BigInt(0),\n endTime: BigInt(0),\n language: '',\n },\n ],\n });\n }\n\n #updateState() {\n let newState: AgentState = 'initializing';\n if (this.#pendingFunctionCalls.size > 0) {\n newState = 'thinking';\n } else if (this.#speaking) {\n newState = 'speaking';\n } else if (this.#started) {\n newState = 'listening';\n }\n\n this.#setState(newState);\n }\n\n #setState(state: AgentState) {\n if (this.room?.isConnected && this.room.localParticipant) {\n const currentState = this.room.localParticipant.attributes[AGENT_STATE_ATTRIBUTE];\n if (currentState !== state) {\n this.room.localParticipant.setAttributes({\n [AGENT_STATE_ATTRIBUTE]: state,\n });\n this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);\n }\n }\n }\n}\n"],"mappings":"AAWA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,oBAAoB;AAC7B,SAAS,uBAAuB;AAChC,YAAY,SAAS;AACrB,SAAS,WAAW;AACpB,SAAS,mCAAmC;AAC5C,SAAS,wBAAwB;AACjC,SAAS,oBAAwC;AAM1C,MAAe,wBAAwB,aAAa;AAM3D;AAMO,MAAe,cAAc;AAQpC;AAGO,MAAM,wBAAwB;AAG9B,MAAM,wBAAwB,aAAa;AAAA,EAChD;AAAA,EACA,OAAoB;AAAA,EACpB,oBAA8C;AAAA,EAC9C,kBAA2C;AAAA,EAC3C,gBAAuE;AAAA,EAEvE,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAIG;AACD,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,WAAW;AAChB,SAAK,UAAU;AAAA,EACjB;AAAA,EAEA,eAAkD;AAAA,EAClD,oBAAkD;AAAA,EAClD,iBAAgC;AAAA,EAChC,eAAmC;AAAA,EACnC,gBAAqC;AAAA,EACrC,iBAA4C;AAAA,EAC5C,UAAU,IAAI;AAAA,EACd,WAAmC;AAAA,EACnC,UAA2C;AAAA,EAC3C,WAAwC;AAAA,EAExC,YAAqB;AAAA,EACrB,yBAAsC,oBAAI,IAAI;AAAA,EAC9C,aAAsB;AAAA,EAEtB,IAAI,SAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAO,KAAsC;AAC/C,SAAK,UAAU;AACf,QAAI,KAAK,UAAU;AACjB,WAAK,SAAS,SAAS;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,IAAI,wBAAqC;AACvC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAAsB,OAAoB;AAC5C,SAAK,yBAAyB;AAC9B,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAU,YAAqB;AACjC,SAAK,aAAa;AAClB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,SAAkB;AAC7B,SAAK,YAAY;AACjB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,MACE,MACA,cAAiD,MACvB;AAC1B,WAAO,IAAI,QAAQ,OAAO,SAAS,WAAW;AAxIlD;AAyIM,UAAI,KAAK,UAAU;AACjB,eAAO,IAAI,MAAM,iCAAiC,CAAC;AAAA,MACrD;AACA,WAAK,aAAa;AAElB,WAAK,GAAG,UAAU,sBAAsB,CAACA,iBAAmC;AAE1E,YAAI,KAAK,mBAAmB;AAC1B;AAAA,QACF;AACA,aAAK,iBAAiBA,aAAY,QAAQ;AAAA,MAC5C,CAAC;AACD,WAAK;AAAA,QACH,UAAU;AAAA,QACV,CAAC,kBAA0CA,iBAAmC;AAC5E,cACE,KAAK,qBACLA,aAAY,aAAa,KAAK,kBAAkB,YAChD,iBAAiB,WAAW,YAAY,qBACxC,CAAC,iBAAiB,YAClB;AACA,6BAAiB,cAAc,IAAI;AAAA,UACrC;AAAA,QACF;AAAA,MACF;AACA,WAAK,GAAG,UAAU,iBAAiB,KAAK,yBAAyB,KAAK,IAAI,CAAC;AAE3E,WAAK,OAAO;AACZ,WAAK,eAAe;AAEpB,WAAK,eAAe,IAAI,YAAY,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW;AACjF,WAAK,gBAAgB,IAAI;AAAA,QACvB,KAAK;AAAA,QACL,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AACA,YAAM,mBAAmB,MAAM;AAC7B,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AAAA,MACnB;AAEA,YAAM,mBAAmB,CAAC,gBAAyB;AACjD,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AACjB,YAAI,KAAK,gBAAgB;AACvB,cAAI,OAAO,KAAK,eAAe,iBAAiB;AAChD,cAAI,aAAa;AACf,oBAAQ;AAAA,UACV;AACA,gBAAM,MAAM,IAAI,YAAY,OAAO;AAAA,YACjC,MAAM,IAAI,SAAS;AAAA,YACnB;AAAA,UACF,CAAC;AAED,cAAI,aAAa;AACf,iBAAK,KAAK,4BAA4B,GAAG;AAAA,UAC3C,OAAO;AACL,iBAAK,KAAK,0BAA0B,GAAG;AAAA,UACzC;AACA,eAAK,QAAQ,MAAM,EAAE,eAAe,MAAM,YAAY,CAAC,EAAE,MAAM,wBAAwB;AAAA,QACzF;AAAA,MACF;AAEA,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AACzD,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AAEzD,YAAM,QAAQ,gBAAgB,iBAAiB,mBAAmB,KAAK,YAAY;AACnF,YAAM,UAAU,IAAI,oBAAoB;AACxC,cAAQ,SAAS,YAAY;AAC7B,WAAK,oBAAqB,QAAM,UAAK,qBAAL,mBAAuB,aAAa,OAAO,aAAa;AACxF,UAAI,CAAC,KAAK,mBAAmB;AAC3B,aAAK,QAAQ,MAAM,yBAAyB;AAC5C,eAAO,IAAI,MAAM,yBAAyB,CAAC;AAC3C;AAAA,MACF;AAEA,YAAM,KAAK,kBAAkB,oBAAoB;AAEjD,UAAI,aAAa;AACf,YAAI,OAAO,gBAAgB,UAAU;AACnC,eAAK,iBAAiB,WAAW;AAAA,QACnC,OAAO;AACL,eAAK,iBAAiB,YAAY,QAAQ;AAAA,QAC5C;AAAA,MACF,OAAO;AAEL,mBAAWA,gBAAe,KAAK,mBAAmB,OAAO,GAAG;AAC1D,eAAK,iBAAiBA,aAAY,QAAQ;AAC1C;AAAA,QACF;AAAA,MACF;AAEA,WAAK,WAAW,KAAK,MAAM,QAAQ,EAAE,QAAQ,KAAK,SAAS,SAAS,KAAK,SAAS,CAAC;AACnF,WAAK,WAAW;AAGhB,WAAK,SAAS,GAAG,0BAA0B,CAAC,YAAiB;AA3OnE,YAAAC;AA6OQ,cAAM,QAAQ,IAAI;AAAA,UAChB,KAAK;AAAA,UACL,KAAK,KAAM,iBAAkB;AAAA,UAC7B,KAAK,kBAAkB;AAAA,UACvB,QAAQ;AAAA,QACV;AAEA,cAAM,UAASA,MAAA,KAAK,kBAAL,gBAAAA,IAAoB;AAAA,UACjC,QAAQ;AAAA,UACR,QAAQ;AAAA,UACR;AAAA,UACA,QAAQ;AAAA,UACR,QAAQ;AAAA;AAEV,aAAK,iBAAiB;AAAA,MACxB,CAAC;AAGD,WAAK,SAAS,GAAG,0BAA0B,CAAC,OAAY;AA/P9D,YAAAA,KAAA;AAiQQ,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wCAAwC,CAAC,OAAY;AA3Q5E,YAAAA,KAAA;AA6QQ,cAAM,gBAAgB,GAAG;AACzB,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,eAAe,MAAM,GAAG,MAAM;AAAA,QAC1F,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AACA,cAAM,UAAU,IAAI,YAAY,OAAO;AAAA,UACrC,MAAM,IAAI,SAAS;AAAA,UACnB,MAAM;AAAA,QACR,CAAC;AACD,aAAK,KAAK,yBAAyB,OAAO;AAC1C,aAAK,QAAQ,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,uBAAuB;AAAA,MACrE,CAAC;AAED,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AA7R5D,YAAAA,KAAA;AA8RQ,YAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,MAAM;AACpD,eAAK,eAAe,UAAU;AAE9B,eAAK,SAAU,aAAa,KAAK;AAAA,YAC/B,KAAK,eAAe;AAAA,YACpB,KAAK,eAAe;AAAA,YACpB,KAAK,MAAO,KAAK,eAAe,eAAe,OAAS,GAAI;AAAA,UAC9D;AAEA,eAAK,iBAAiB;AAAA,QACxB;AAEA,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,KAAK,uBAAuB;AAAA,MACnC,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,OAAY;AACrD,aAAK,sBAAsB,IAAI,GAAG,MAAM;AACxC,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,2BAA2B,CAAC,OAAY;AACvD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAED,cAAQ,KAAK,QAAQ;AAAA,IACvB,CAAC;AAAA,EACH;AAAA,EAEA,iBAAiB,qBAAmC;AAClD,QAAI,CAAC,KAAK,MAAM;AACd,WAAK,QAAQ,MAAM,iBAAiB;AACpC;AAAA,IACF;AAEA,SAAK,oBAAoB,KAAK,KAAK,mBAAmB,IAAI,mBAAmB,KAAK;AAClF,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,6BAA6B,mBAAmB,YAAY;AAC/E;AAAA,IACF;AAEA,QAAI,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AACrD,WAAK,uBAAuB;AAAA,IAC9B;AAGA,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,YAAY,qBAAqB,YAAY,OAAO;AAC7E,aAAK,yBAAyB,YAAY,OAAO,aAAa,KAAK,iBAAiB;AACpF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,YAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAAA,EACF;AAAA,EAEA,yBACE,OACA,aACA,aACA;AA/XJ;AAgYI,QACE,YAAY,WAAW,YAAY,qBACnC,YAAY,eAAa,UAAK,sBAAL,mBAAwB,WACjD;AACA;AAAA,IACF;AACA,UAAM,sBAAsB,OAAO,gBAA6B;AAC9D,YAAM,UAAU,IAAI;AAAA,QAClB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AAEA,uBAAiB,SAAS,aAAa;AACrC,cAAM,YAAY,MAAM;AACxB,mBAAWC,UAAS,QAAQ,MAAM,UAAU,MAAM,GAAG;AACnD,eAAK,SAAU,iBAAiB,OAAOA,MAAK;AAAA,QAC9C;AAAA,MACF;AAAA,IACF;AACA,SAAK,kBAAkB;AAEvB,QAAI,KAAK,eAAe;AACtB,WAAK,cAAc,OAAO;AAAA,IAC5B;AAEA,QAAI;AACJ,SAAK,gBAAgB;AAAA,MACnB,SAAS,IAAI,QAAc,CAAC,SAAS,WAAW;AAC9C,iBAAS,MAAM;AACb,iBAAO,IAAI,MAAM,gBAAgB,CAAC;AAAA,QACpC;AACA,4BAAoB,IAAI,YAAY,OAAO,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW,CAAC,EACtF,KAAK,OAAO,EACZ,MAAM,MAAM;AAAA,MACjB,CAAC;AAAA,MACD,QAAQ,MAAM,OAAO;AAAA,IACvB;AAAA,EACF;AAAA,EAEA,oBAAmC;AAxarC;AAyaI,QAAI,CAAC,KAAK,kBAAkB,KAAK,QAAQ,KAAK,KAAK,kBAAkB;AACnE,WAAK,iBAAiB,iBAAiB,KAAK,OAAM,UAAK,KAAK,qBAAV,mBAA4B,QAAQ;AAAA,IACxF;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,sBACE,qBACA,UACA,MACA,SACA,IACM;AArbV;AAsbI,SAAK,QAAQ;AAAA,MACX,4BAA4B,mBAAmB,IAAI,QAAQ,IAAI,IAAI,IAAI,OAAO,IAAI,EAAE;AAAA,IACtF;AACA,QAAI,GAAC,UAAK,SAAL,mBAAW,mBAAkB;AAChC,WAAK,QAAQ,MAAM,mCAAmC;AACtD;AAAA,IACF;AAEA,SAAK,KAAK,iBAAiB,qBAAqB;AAAA,MAC9C;AAAA,MACA;AAAA,MACA,UAAU;AAAA,QACR;AAAA,UACE;AAAA,UACA,OAAO;AAAA,UACP;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,UAAU;AAAA,QACZ;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,eAAe;AACb,QAAI,WAAuB;AAC3B,QAAI,KAAK,sBAAsB,OAAO,GAAG;AACvC,iBAAW;AAAA,IACb,WAAW,KAAK,WAAW;AACzB,iBAAW;AAAA,IACb,WAAW,KAAK,UAAU;AACxB,iBAAW;AAAA,IACb;AAEA,SAAK,UAAU,QAAQ;AAAA,EACzB;AAAA,EAEA,UAAU,OAAmB;AA3d/B;AA4dI,UAAI,UAAK,SAAL,mBAAW,gBAAe,KAAK,KAAK,kBAAkB;AACxD,YAAM,eAAe,KAAK,KAAK,iBAAiB,WAAW,qBAAqB;AAChF,UAAI,iBAAiB,OAAO;AAC1B,aAAK,KAAK,iBAAiB,cAAc;AAAA,UACvC,CAAC,qBAAqB,GAAG;AAAA,QAC3B,CAAC;AACD,aAAK,QAAQ,MAAM,GAAG,qBAAqB,KAAK,YAAY,MAAM,KAAK,EAAE;AAAA,MAC3E;AAAA,IACF;AAAA,EACF;AACF;","names":["participant","_a","frame"]}
|
|
1
|
+
{"version":3,"sources":["../../src/multimodal/multimodal_agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n LocalTrackPublication,\n RemoteAudioTrack,\n RemoteParticipant,\n RemoteTrack,\n RemoteTrackPublication,\n Room,\n} from '@livekit/rtc-node';\nimport {\n AudioSource,\n AudioStream,\n LocalAudioTrack,\n RoomEvent,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport * as llm from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { MultimodalLLMMetrics } from '../metrics/base.js';\nimport { BasicTranscriptionForwarder } from '../transcription.js';\nimport { findMicroTrackId } from '../utils.js';\nimport { AgentPlayout, type PlayoutHandle } from './agent_playout.js';\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeSession extends EventEmitter {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract conversation: any; // openai.realtime.Conversation\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract inputAudioBuffer: any; // openai.realtime.InputAudioBuffer\n abstract fncCtx: llm.FunctionContext | undefined;\n}\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeModel {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract session(options: any): RealtimeSession; // openai.realtime.ModelOptions\n abstract close(): Promise<void>;\n abstract sampleRate: number;\n abstract numChannels: number;\n abstract inFrameSize: number;\n abstract outFrameSize: number;\n}\n\nexport type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';\nexport const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';\n\n/** @beta */\nexport class MultimodalAgent extends EventEmitter {\n model: RealtimeModel;\n room: Room | null = null;\n linkedParticipant: RemoteParticipant | null = null;\n subscribedTrack: RemoteAudioTrack | null = null;\n readMicroTask: Promise<void> | null = null;\n\n constructor({\n model,\n chatCtx,\n fncCtx,\n }: {\n model: RealtimeModel;\n chatCtx?: llm.ChatContext;\n fncCtx?: llm.FunctionContext;\n }) {\n super();\n this.model = model;\n this.#chatCtx = chatCtx;\n this.#fncCtx = fncCtx;\n }\n\n #participant: RemoteParticipant | string | null = null;\n #agentPublication: LocalTrackPublication | null = null;\n #localTrackSid: string | null = null;\n #localSource: AudioSource | null = null;\n #agentPlayout: AgentPlayout | null = null;\n #playingHandle: PlayoutHandle | undefined = undefined;\n #logger = log();\n #session: RealtimeSession | null = null;\n #fncCtx: llm.FunctionContext | undefined = undefined;\n #chatCtx: llm.ChatContext | undefined = undefined;\n\n #_started: boolean = false;\n #_pendingFunctionCalls: Set<string> = new Set();\n #_speaking: boolean = false;\n\n get fncCtx(): llm.FunctionContext | undefined {\n return this.#fncCtx;\n }\n\n set fncCtx(ctx: llm.FunctionContext | undefined) {\n this.#fncCtx = ctx;\n if (this.#session) {\n this.#session.fncCtx = ctx;\n }\n }\n\n get #pendingFunctionCalls(): Set<string> {\n return this.#_pendingFunctionCalls;\n }\n\n set #pendingFunctionCalls(calls: Set<string>) {\n this.#_pendingFunctionCalls = calls;\n this.#updateState();\n }\n\n get #speaking(): boolean {\n return this.#_speaking;\n }\n\n set #speaking(isSpeaking: boolean) {\n this.#_speaking = isSpeaking;\n this.#updateState();\n }\n\n get #started(): boolean {\n return this.#_started;\n }\n\n set #started(started: boolean) {\n this.#_started = started;\n this.#updateState();\n }\n\n start(\n room: Room,\n participant: RemoteParticipant | string | null = null,\n ): Promise<RealtimeSession> {\n return new Promise(async (resolve, reject) => {\n if (this.#started) {\n reject(new Error('MultimodalAgent already started'));\n }\n this.#updateState();\n\n room.on(RoomEvent.ParticipantConnected, (participant: RemoteParticipant) => {\n // automatically link to the first participant that connects, if not already linked\n if (this.linkedParticipant) {\n return;\n }\n this.#linkParticipant(participant.identity);\n });\n room.on(\n RoomEvent.TrackPublished,\n (trackPublication: RemoteTrackPublication, participant: RemoteParticipant) => {\n if (\n this.linkedParticipant &&\n participant.identity === this.linkedParticipant.identity &&\n trackPublication.source === TrackSource.SOURCE_MICROPHONE &&\n !trackPublication.subscribed\n ) {\n trackPublication.setSubscribed(true);\n }\n },\n );\n room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));\n\n this.room = room;\n this.#participant = participant;\n\n this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);\n this.#agentPlayout = new AgentPlayout(\n this.#localSource,\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n this.model.outFrameSize,\n );\n const onPlayoutStarted = () => {\n this.emit('agent_started_speaking');\n this.#speaking = true;\n };\n\n const onPlayoutStopped = (interrupted: boolean) => {\n this.emit('agent_stopped_speaking');\n this.#speaking = false;\n if (this.#playingHandle) {\n let text = this.#playingHandle.transcriptionFwd.text;\n if (interrupted) {\n text += '…';\n }\n const msg = llm.ChatMessage.create({\n role: llm.ChatRole.ASSISTANT,\n text,\n });\n\n if (interrupted) {\n this.emit('agent_speech_interrupted', msg);\n } else {\n this.emit('agent_speech_committed', msg);\n }\n this.#logger.child({ transcription: text, interrupted }).debug('committed agent speech');\n }\n };\n\n this.#agentPlayout.on('playout_started', onPlayoutStarted);\n this.#agentPlayout.on('playout_stopped', onPlayoutStopped);\n\n const track = LocalAudioTrack.createAudioTrack('assistant_voice', this.#localSource);\n const options = new TrackPublishOptions();\n options.source = TrackSource.SOURCE_MICROPHONE;\n this.#agentPublication = (await room.localParticipant?.publishTrack(track, options)) || null;\n if (!this.#agentPublication) {\n this.#logger.error('Failed to publish track');\n reject(new Error('Failed to publish track'));\n return;\n }\n\n await this.#agentPublication.waitForSubscription();\n\n if (participant) {\n if (typeof participant === 'string') {\n this.#linkParticipant(participant);\n } else {\n this.#linkParticipant(participant.identity);\n }\n } else {\n // No participant specified, try to find the first participant in the room\n for (const participant of room.remoteParticipants.values()) {\n this.#linkParticipant(participant.identity);\n break;\n }\n }\n\n this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });\n this.#started = true;\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_added', (message: any) => {\n // openai.realtime.RealtimeContent\n const trFwd = new BasicTranscriptionForwarder(\n this.room!,\n this.room!.localParticipant!.identity,\n this.#getLocalTrackSid()!,\n message.responseId,\n );\n\n const handle = this.#agentPlayout?.play(\n message.itemId,\n message.contentIndex,\n trFwd,\n message.textStream,\n message.audioStream,\n );\n this.#playingHandle = handle;\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_committed', (ev: any) => {\n // openai.realtime.InputSpeechCommittedEvent\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_transcription_completed', (ev: any) => {\n // openai.realtime.InputSpeechTranscriptionCompletedEvent\n const transcription = ev.transcript;\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, transcription, true, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n const userMsg = llm.ChatMessage.create({\n role: llm.ChatRole.USER,\n text: transcription,\n });\n this.emit('user_speech_committed', userMsg);\n this.#logger.child({ transcription }).debug('committed user speech');\n });\n\n this.#session.on('input_speech_started', (ev: any) => {\n this.emit('user_started_speaking');\n if (this.#playingHandle && !this.#playingHandle.done) {\n this.#playingHandle.interrupt();\n\n this.#session!.conversation.item.truncate(\n this.#playingHandle.itemId,\n this.#playingHandle.contentIndex,\n Math.floor((this.#playingHandle.audioSamples / 24000) * 1000),\n );\n\n this.#playingHandle = undefined;\n }\n\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n this.#session.on('input_speech_stopped', (ev: any) => {\n this.emit('user_stopped_speaking');\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_started', (ev: any) => {\n this.#pendingFunctionCalls.add(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_completed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_failed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n this.#session.on('metrics_collected', (metrics: MultimodalLLMMetrics) => {\n this.emit('metrics_collected', metrics);\n });\n\n resolve(this.#session);\n });\n }\n\n #linkParticipant(participantIdentity: string): void {\n if (!this.room) {\n this.#logger.error('Room is not set');\n return;\n }\n\n this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;\n if (!this.linkedParticipant) {\n this.#logger.error(`Participant with identity ${participantIdentity} not found`);\n return;\n }\n\n if (this.linkedParticipant.trackPublications.size > 0) {\n this.#subscribeToMicrophone();\n }\n\n // also check if already subscribed\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {\n this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);\n break;\n }\n }\n }\n\n #subscribeToMicrophone(): void {\n if (!this.linkedParticipant) {\n this.#logger.error('Participant is not set');\n return;\n }\n\n let microphonePublication: RemoteTrackPublication | undefined = undefined;\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n microphonePublication = publication;\n break;\n }\n }\n if (!microphonePublication) {\n return;\n }\n\n if (!microphonePublication.subscribed) {\n microphonePublication.setSubscribed(true);\n }\n }\n\n #handleTrackSubscription(\n track: RemoteTrack,\n publication: RemoteTrackPublication,\n participant: RemoteParticipant,\n ) {\n if (\n publication.source !== TrackSource.SOURCE_MICROPHONE ||\n participant.identity !== this.linkedParticipant?.identity\n ) {\n return;\n }\n const readAudioStreamTask = async (audioStream: AudioStream) => {\n const bstream = new AudioByteStream(\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n );\n\n for await (const frame of audioStream) {\n const audioData = frame.data;\n for (const frame of bstream.write(audioData.buffer)) {\n this.#session!.inputAudioBuffer.append(frame);\n }\n }\n };\n this.subscribedTrack = track;\n\n this.readMicroTask = new Promise<void>((resolve, reject) => {\n readAudioStreamTask(new AudioStream(track, this.model.sampleRate, this.model.numChannels))\n .then(resolve)\n .catch(reject);\n });\n }\n\n #getLocalTrackSid(): string | null {\n if (!this.#localTrackSid && this.room && this.room.localParticipant) {\n this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant?.identity);\n }\n return this.#localTrackSid;\n }\n\n #publishTranscription(\n participantIdentity: string,\n trackSid: string,\n text: string,\n isFinal: boolean,\n id: string,\n ): void {\n this.#logger.debug(\n `Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`,\n );\n if (!this.room?.localParticipant) {\n this.#logger.error('Room or local participant not set');\n return;\n }\n\n this.room.localParticipant.publishTranscription({\n participantIdentity,\n trackSid,\n segments: [\n {\n text,\n final: isFinal,\n id,\n startTime: BigInt(0),\n endTime: BigInt(0),\n language: '',\n },\n ],\n });\n }\n\n #updateState() {\n let newState: AgentState = 'initializing';\n if (this.#pendingFunctionCalls.size > 0) {\n newState = 'thinking';\n } else if (this.#speaking) {\n newState = 'speaking';\n } else if (this.#started) {\n newState = 'listening';\n }\n\n this.#setState(newState);\n }\n\n #setState(state: AgentState) {\n if (this.room?.isConnected && this.room.localParticipant) {\n const currentState = this.room.localParticipant.attributes[AGENT_STATE_ATTRIBUTE];\n if (currentState !== state) {\n this.room.localParticipant.setAttributes({\n [AGENT_STATE_ATTRIBUTE]: state,\n });\n this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);\n }\n }\n }\n}\n"],"mappings":"AAWA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,oBAAoB;AAC7B,SAAS,uBAAuB;AAChC,YAAY,SAAS;AACrB,SAAS,WAAW;AAEpB,SAAS,mCAAmC;AAC5C,SAAS,wBAAwB;AACjC,SAAS,oBAAwC;AAM1C,MAAe,wBAAwB,aAAa;AAM3D;AAMO,MAAe,cAAc;AAQpC;AAGO,MAAM,wBAAwB;AAG9B,MAAM,wBAAwB,aAAa;AAAA,EAChD;AAAA,EACA,OAAoB;AAAA,EACpB,oBAA8C;AAAA,EAC9C,kBAA2C;AAAA,EAC3C,gBAAsC;AAAA,EAEtC,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAIG;AACD,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,WAAW;AAChB,SAAK,UAAU;AAAA,EACjB;AAAA,EAEA,eAAkD;AAAA,EAClD,oBAAkD;AAAA,EAClD,iBAAgC;AAAA,EAChC,eAAmC;AAAA,EACnC,gBAAqC;AAAA,EACrC,iBAA4C;AAAA,EAC5C,UAAU,IAAI;AAAA,EACd,WAAmC;AAAA,EACnC,UAA2C;AAAA,EAC3C,WAAwC;AAAA,EAExC,YAAqB;AAAA,EACrB,yBAAsC,oBAAI,IAAI;AAAA,EAC9C,aAAsB;AAAA,EAEtB,IAAI,SAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAO,KAAsC;AAC/C,SAAK,UAAU;AACf,QAAI,KAAK,UAAU;AACjB,WAAK,SAAS,SAAS;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,IAAI,wBAAqC;AACvC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAAsB,OAAoB;AAC5C,SAAK,yBAAyB;AAC9B,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAU,YAAqB;AACjC,SAAK,aAAa;AAClB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,SAAkB;AAC7B,SAAK,YAAY;AACjB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,MACE,MACA,cAAiD,MACvB;AAC1B,WAAO,IAAI,QAAQ,OAAO,SAAS,WAAW;AAzIlD;AA0IM,UAAI,KAAK,UAAU;AACjB,eAAO,IAAI,MAAM,iCAAiC,CAAC;AAAA,MACrD;AACA,WAAK,aAAa;AAElB,WAAK,GAAG,UAAU,sBAAsB,CAACA,iBAAmC;AAE1E,YAAI,KAAK,mBAAmB;AAC1B;AAAA,QACF;AACA,aAAK,iBAAiBA,aAAY,QAAQ;AAAA,MAC5C,CAAC;AACD,WAAK;AAAA,QACH,UAAU;AAAA,QACV,CAAC,kBAA0CA,iBAAmC;AAC5E,cACE,KAAK,qBACLA,aAAY,aAAa,KAAK,kBAAkB,YAChD,iBAAiB,WAAW,YAAY,qBACxC,CAAC,iBAAiB,YAClB;AACA,6BAAiB,cAAc,IAAI;AAAA,UACrC;AAAA,QACF;AAAA,MACF;AACA,WAAK,GAAG,UAAU,iBAAiB,KAAK,yBAAyB,KAAK,IAAI,CAAC;AAE3E,WAAK,OAAO;AACZ,WAAK,eAAe;AAEpB,WAAK,eAAe,IAAI,YAAY,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW;AACjF,WAAK,gBAAgB,IAAI;AAAA,QACvB,KAAK;AAAA,QACL,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AACA,YAAM,mBAAmB,MAAM;AAC7B,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AAAA,MACnB;AAEA,YAAM,mBAAmB,CAAC,gBAAyB;AACjD,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AACjB,YAAI,KAAK,gBAAgB;AACvB,cAAI,OAAO,KAAK,eAAe,iBAAiB;AAChD,cAAI,aAAa;AACf,oBAAQ;AAAA,UACV;AACA,gBAAM,MAAM,IAAI,YAAY,OAAO;AAAA,YACjC,MAAM,IAAI,SAAS;AAAA,YACnB;AAAA,UACF,CAAC;AAED,cAAI,aAAa;AACf,iBAAK,KAAK,4BAA4B,GAAG;AAAA,UAC3C,OAAO;AACL,iBAAK,KAAK,0BAA0B,GAAG;AAAA,UACzC;AACA,eAAK,QAAQ,MAAM,EAAE,eAAe,MAAM,YAAY,CAAC,EAAE,MAAM,wBAAwB;AAAA,QACzF;AAAA,MACF;AAEA,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AACzD,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AAEzD,YAAM,QAAQ,gBAAgB,iBAAiB,mBAAmB,KAAK,YAAY;AACnF,YAAM,UAAU,IAAI,oBAAoB;AACxC,cAAQ,SAAS,YAAY;AAC7B,WAAK,oBAAqB,QAAM,UAAK,qBAAL,mBAAuB,aAAa,OAAO,aAAa;AACxF,UAAI,CAAC,KAAK,mBAAmB;AAC3B,aAAK,QAAQ,MAAM,yBAAyB;AAC5C,eAAO,IAAI,MAAM,yBAAyB,CAAC;AAC3C;AAAA,MACF;AAEA,YAAM,KAAK,kBAAkB,oBAAoB;AAEjD,UAAI,aAAa;AACf,YAAI,OAAO,gBAAgB,UAAU;AACnC,eAAK,iBAAiB,WAAW;AAAA,QACnC,OAAO;AACL,eAAK,iBAAiB,YAAY,QAAQ;AAAA,QAC5C;AAAA,MACF,OAAO;AAEL,mBAAWA,gBAAe,KAAK,mBAAmB,OAAO,GAAG;AAC1D,eAAK,iBAAiBA,aAAY,QAAQ;AAC1C;AAAA,QACF;AAAA,MACF;AAEA,WAAK,WAAW,KAAK,MAAM,QAAQ,EAAE,QAAQ,KAAK,SAAS,SAAS,KAAK,SAAS,CAAC;AACnF,WAAK,WAAW;AAGhB,WAAK,SAAS,GAAG,0BAA0B,CAAC,YAAiB;AA5OnE,YAAAC;AA8OQ,cAAM,QAAQ,IAAI;AAAA,UAChB,KAAK;AAAA,UACL,KAAK,KAAM,iBAAkB;AAAA,UAC7B,KAAK,kBAAkB;AAAA,UACvB,QAAQ;AAAA,QACV;AAEA,cAAM,UAASA,MAAA,KAAK,kBAAL,gBAAAA,IAAoB;AAAA,UACjC,QAAQ;AAAA,UACR,QAAQ;AAAA,UACR;AAAA,UACA,QAAQ;AAAA,UACR,QAAQ;AAAA;AAEV,aAAK,iBAAiB;AAAA,MACxB,CAAC;AAGD,WAAK,SAAS,GAAG,0BAA0B,CAAC,OAAY;AAhQ9D,YAAAA,KAAA;AAkQQ,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wCAAwC,CAAC,OAAY;AA5Q5E,YAAAA,KAAA;AA8QQ,cAAM,gBAAgB,GAAG;AACzB,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,eAAe,MAAM,GAAG,MAAM;AAAA,QAC1F,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AACA,cAAM,UAAU,IAAI,YAAY,OAAO;AAAA,UACrC,MAAM,IAAI,SAAS;AAAA,UACnB,MAAM;AAAA,QACR,CAAC;AACD,aAAK,KAAK,yBAAyB,OAAO;AAC1C,aAAK,QAAQ,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,uBAAuB;AAAA,MACrE,CAAC;AAED,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AA9R5D,YAAAA,KAAA;AA+RQ,aAAK,KAAK,uBAAuB;AACjC,YAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,MAAM;AACpD,eAAK,eAAe,UAAU;AAE9B,eAAK,SAAU,aAAa,KAAK;AAAA,YAC/B,KAAK,eAAe;AAAA,YACpB,KAAK,eAAe;AAAA,YACpB,KAAK,MAAO,KAAK,eAAe,eAAe,OAAS,GAAI;AAAA,UAC9D;AAEA,eAAK,iBAAiB;AAAA,QACxB;AAEA,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,KAAK,uBAAuB;AAAA,MACnC,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,OAAY;AACrD,aAAK,sBAAsB,IAAI,GAAG,MAAM;AACxC,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,2BAA2B,CAAC,OAAY;AACvD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAED,WAAK,SAAS,GAAG,qBAAqB,CAAC,YAAkC;AACvE,aAAK,KAAK,qBAAqB,OAAO;AAAA,MACxC,CAAC;AAED,cAAQ,KAAK,QAAQ;AAAA,IACvB,CAAC;AAAA,EACH;AAAA,EAEA,iBAAiB,qBAAmC;AAClD,QAAI,CAAC,KAAK,MAAM;AACd,WAAK,QAAQ,MAAM,iBAAiB;AACpC;AAAA,IACF;AAEA,SAAK,oBAAoB,KAAK,KAAK,mBAAmB,IAAI,mBAAmB,KAAK;AAClF,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,6BAA6B,mBAAmB,YAAY;AAC/E;AAAA,IACF;AAEA,QAAI,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AACrD,WAAK,uBAAuB;AAAA,IAC9B;AAGA,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,YAAY,qBAAqB,YAAY,OAAO;AAC7E,aAAK,yBAAyB,YAAY,OAAO,aAAa,KAAK,iBAAiB;AACpF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,YAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAAA,EACF;AAAA,EAEA,yBACE,OACA,aACA,aACA;AArYJ;AAsYI,QACE,YAAY,WAAW,YAAY,qBACnC,YAAY,eAAa,UAAK,sBAAL,mBAAwB,WACjD;AACA;AAAA,IACF;AACA,UAAM,sBAAsB,OAAO,gBAA6B;AAC9D,YAAM,UAAU,IAAI;AAAA,QAClB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AAEA,uBAAiB,SAAS,aAAa;AACrC,cAAM,YAAY,MAAM;AACxB,mBAAWC,UAAS,QAAQ,MAAM,UAAU,MAAM,GAAG;AACnD,eAAK,SAAU,iBAAiB,OAAOA,MAAK;AAAA,QAC9C;AAAA,MACF;AAAA,IACF;AACA,SAAK,kBAAkB;AAEvB,SAAK,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC1D,0BAAoB,IAAI,YAAY,OAAO,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW,CAAC,EACtF,KAAK,OAAO,EACZ,MAAM,MAAM;AAAA,IACjB,CAAC;AAAA,EACH;AAAA,EAEA,oBAAmC;AAnarC;AAoaI,QAAI,CAAC,KAAK,kBAAkB,KAAK,QAAQ,KAAK,KAAK,kBAAkB;AACnE,WAAK,iBAAiB,iBAAiB,KAAK,OAAM,UAAK,KAAK,qBAAV,mBAA4B,QAAQ;AAAA,IACxF;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,sBACE,qBACA,UACA,MACA,SACA,IACM;AAhbV;AAibI,SAAK,QAAQ;AAAA,MACX,4BAA4B,mBAAmB,IAAI,QAAQ,IAAI,IAAI,IAAI,OAAO,IAAI,EAAE;AAAA,IACtF;AACA,QAAI,GAAC,UAAK,SAAL,mBAAW,mBAAkB;AAChC,WAAK,QAAQ,MAAM,mCAAmC;AACtD;AAAA,IACF;AAEA,SAAK,KAAK,iBAAiB,qBAAqB;AAAA,MAC9C;AAAA,MACA;AAAA,MACA,UAAU;AAAA,QACR;AAAA,UACE;AAAA,UACA,OAAO;AAAA,UACP;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,UAAU;AAAA,QACZ;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,eAAe;AACb,QAAI,WAAuB;AAC3B,QAAI,KAAK,sBAAsB,OAAO,GAAG;AACvC,iBAAW;AAAA,IACb,WAAW,KAAK,WAAW;AACzB,iBAAW;AAAA,IACb,WAAW,KAAK,UAAU;AACxB,iBAAW;AAAA,IACb;AAEA,SAAK,UAAU,QAAQ;AAAA,EACzB;AAAA,EAEA,UAAU,OAAmB;AAtd/B;AAudI,UAAI,UAAK,SAAL,mBAAW,gBAAe,KAAK,KAAK,kBAAkB;AACxD,YAAM,eAAe,KAAK,KAAK,iBAAiB,WAAW,qBAAqB;AAChF,UAAI,iBAAiB,OAAO;AAC1B,aAAK,KAAK,iBAAiB,cAAc;AAAA,UACvC,CAAC,qBAAqB,GAAG;AAAA,QAC3B,CAAC;AACD,aAAK,QAAQ,MAAM,GAAG,qBAAqB,KAAK,YAAY,MAAM,KAAK,EAAE;AAAA,MAC3E;AAAA,IACF;AAAA,EACF;AACF;","names":["participant","_a","frame"]}
|
|
@@ -28,6 +28,7 @@ var import_utils = require("../utils.cjs");
|
|
|
28
28
|
class SynthesisHandle {
|
|
29
29
|
static FLUSH_SENTINEL = Symbol("FLUSH_SENTINEL");
|
|
30
30
|
#speechId;
|
|
31
|
+
text;
|
|
31
32
|
ttsSource;
|
|
32
33
|
#agentPlayout;
|
|
33
34
|
tts;
|
|
@@ -111,6 +112,10 @@ class AgentOutput {
|
|
|
111
112
|
} finally {
|
|
112
113
|
if (handle.intFut.done) {
|
|
113
114
|
(0, import_utils.gracefullyCancel)(task);
|
|
115
|
+
} else {
|
|
116
|
+
task.then((text) => {
|
|
117
|
+
handle.text = text;
|
|
118
|
+
});
|
|
114
119
|
}
|
|
115
120
|
}
|
|
116
121
|
resolve();
|
|
@@ -134,11 +139,12 @@ const stringSynthesisTask = (text, handle) => {
|
|
|
134
139
|
handle.queue.put(audio.frame);
|
|
135
140
|
}
|
|
136
141
|
handle.queue.put(SynthesisHandle.FLUSH_SENTINEL);
|
|
137
|
-
resolve();
|
|
142
|
+
resolve(text);
|
|
138
143
|
});
|
|
139
144
|
};
|
|
140
145
|
const streamSynthesisTask = (stream, handle) => {
|
|
141
146
|
return new import_utils.CancellablePromise(async (resolve, _, onCancel) => {
|
|
147
|
+
let fullText = "";
|
|
142
148
|
let cancelled = false;
|
|
143
149
|
onCancel(() => {
|
|
144
150
|
cancelled = true;
|
|
@@ -156,12 +162,13 @@ const streamSynthesisTask = (stream, handle) => {
|
|
|
156
162
|
};
|
|
157
163
|
readGeneratedAudio();
|
|
158
164
|
for await (const text of stream) {
|
|
165
|
+
fullText += text;
|
|
159
166
|
if (cancelled) break;
|
|
160
167
|
ttsStream.pushText(text);
|
|
161
168
|
}
|
|
162
169
|
ttsStream.flush();
|
|
163
170
|
ttsStream.endInput();
|
|
164
|
-
resolve();
|
|
171
|
+
resolve(fullText);
|
|
165
172
|
});
|
|
166
173
|
};
|
|
167
174
|
// Annotate the CommonJS export names for ESM import in node:
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/pipeline/agent_output.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { log } from '../log.js';\nimport { SynthesizeStream, type TTS } from '../tts/index.js';\nimport { AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';\nimport type { AgentPlayout, PlayoutHandle } from './agent_playout.js';\n\nexport type SpeechSource = AsyncIterable<string> | string | Promise<string>;\n\nexport class SynthesisHandle {\n static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n\n #speechId: string;\n ttsSource: SpeechSource;\n #agentPlayout: AgentPlayout;\n tts: TTS;\n queue = new AsyncIterableQueue<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>();\n #playHandle?: PlayoutHandle;\n intFut = new Future();\n #logger = log();\n\n constructor(speechId: string, ttsSource: SpeechSource, agentPlayout: AgentPlayout, tts: TTS) {\n this.#speechId = speechId;\n this.ttsSource = ttsSource;\n this.#agentPlayout = agentPlayout;\n this.tts = tts;\n }\n\n get speechId(): string {\n return this.#speechId;\n }\n\n get validated(): boolean {\n return !!this.#playHandle;\n }\n\n get interrupted(): boolean {\n return this.intFut.done;\n }\n\n get playHandle(): PlayoutHandle | undefined {\n return this.#playHandle;\n }\n\n /** Validate the speech for playout. */\n play(): PlayoutHandle {\n if (this.interrupted) {\n throw new Error('synthesis was interrupted');\n }\n\n this.#playHandle = this.#agentPlayout.play(this.#speechId, this.queue);\n return this.#playHandle;\n }\n\n /** Interrupt the speech. */\n interrupt() {\n if (this.interrupted) {\n return;\n }\n\n this.#logger.child({ speechId: this.#speechId }).debug('interrupting synthesis/playout');\n this.#playHandle?.interrupt();\n this.intFut.resolve();\n }\n}\n\nexport class AgentOutput {\n #agentPlayout: AgentPlayout;\n #tts: TTS;\n #tasks: CancellablePromise<void>[] = [];\n\n constructor(agentPlayout: AgentPlayout, tts: TTS) {\n this.#agentPlayout = agentPlayout;\n this.#tts = tts;\n }\n\n get playout(): AgentPlayout {\n return this.#agentPlayout;\n }\n\n async close() {\n this.#tasks.forEach((task) => task.cancel());\n await Promise.all(this.#tasks);\n }\n\n synthesize(speechId: string, ttsSource: SpeechSource): SynthesisHandle {\n const handle = new SynthesisHandle(speechId, ttsSource, this.#agentPlayout, this.#tts);\n const task = this.#synthesize(handle);\n this.#tasks.push(task);\n task.finally(() => this.#tasks.splice(this.#tasks.indexOf(task)));\n return handle;\n }\n\n #synthesize(handle: SynthesisHandle): CancellablePromise<void> {\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n return new CancellablePromise(async (resolve, _, onCancel) => {\n const ttsSource = await handle.ttsSource;\n let task: CancellablePromise<
|
|
1
|
+
{"version":3,"sources":["../../src/pipeline/agent_output.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { log } from '../log.js';\nimport { SynthesizeStream, type TTS } from '../tts/index.js';\nimport { AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';\nimport type { AgentPlayout, PlayoutHandle } from './agent_playout.js';\n\nexport type SpeechSource = AsyncIterable<string> | string | Promise<string>;\n\nexport class SynthesisHandle {\n static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n\n #speechId: string;\n text?: string;\n ttsSource: SpeechSource;\n #agentPlayout: AgentPlayout;\n tts: TTS;\n queue = new AsyncIterableQueue<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>();\n #playHandle?: PlayoutHandle;\n intFut = new Future();\n #logger = log();\n\n constructor(speechId: string, ttsSource: SpeechSource, agentPlayout: AgentPlayout, tts: TTS) {\n this.#speechId = speechId;\n this.ttsSource = ttsSource;\n this.#agentPlayout = agentPlayout;\n this.tts = tts;\n }\n\n get speechId(): string {\n return this.#speechId;\n }\n\n get validated(): boolean {\n return !!this.#playHandle;\n }\n\n get interrupted(): boolean {\n return this.intFut.done;\n }\n\n get playHandle(): PlayoutHandle | undefined {\n return this.#playHandle;\n }\n\n /** Validate the speech for playout. */\n play(): PlayoutHandle {\n if (this.interrupted) {\n throw new Error('synthesis was interrupted');\n }\n\n this.#playHandle = this.#agentPlayout.play(this.#speechId, this.queue);\n return this.#playHandle;\n }\n\n /** Interrupt the speech. */\n interrupt() {\n if (this.interrupted) {\n return;\n }\n\n this.#logger.child({ speechId: this.#speechId }).debug('interrupting synthesis/playout');\n this.#playHandle?.interrupt();\n this.intFut.resolve();\n }\n}\n\nexport class AgentOutput {\n #agentPlayout: AgentPlayout;\n #tts: TTS;\n #tasks: CancellablePromise<void>[] = [];\n\n constructor(agentPlayout: AgentPlayout, tts: TTS) {\n this.#agentPlayout = agentPlayout;\n this.#tts = tts;\n }\n\n get playout(): AgentPlayout {\n return this.#agentPlayout;\n }\n\n async close() {\n this.#tasks.forEach((task) => task.cancel());\n await Promise.all(this.#tasks);\n }\n\n synthesize(speechId: string, ttsSource: SpeechSource): SynthesisHandle {\n const handle = new SynthesisHandle(speechId, ttsSource, this.#agentPlayout, this.#tts);\n const task = this.#synthesize(handle);\n this.#tasks.push(task);\n task.finally(() => this.#tasks.splice(this.#tasks.indexOf(task)));\n return handle;\n }\n\n #synthesize(handle: SynthesisHandle): CancellablePromise<void> {\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n return new CancellablePromise(async (resolve, _, onCancel) => {\n const ttsSource = await handle.ttsSource;\n let task: CancellablePromise<string>;\n if (typeof ttsSource === 'string') {\n task = stringSynthesisTask(ttsSource, handle);\n } else {\n task = streamSynthesisTask(ttsSource, handle);\n }\n\n onCancel(() => {\n gracefullyCancel(task);\n });\n\n try {\n await Promise.any([task, handle.intFut.await]);\n } finally {\n if (handle.intFut.done) {\n gracefullyCancel(task);\n } else {\n task.then((text) => {\n handle.text = text;\n });\n }\n }\n\n resolve();\n });\n }\n}\n\nconst stringSynthesisTask = (text: string, handle: SynthesisHandle): CancellablePromise<string> => {\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n return new CancellablePromise(async (resolve, _, onCancel) => {\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n\n const ttsStream = handle.tts.stream();\n ttsStream.pushText(text);\n ttsStream.flush();\n ttsStream.endInput();\n for await (const audio of ttsStream) {\n if (cancelled || audio === SynthesizeStream.END_OF_STREAM) {\n break;\n }\n handle.queue.put(audio.frame);\n }\n handle.queue.put(SynthesisHandle.FLUSH_SENTINEL);\n\n resolve(text);\n });\n};\n\nconst streamSynthesisTask = (\n stream: AsyncIterable<string>,\n handle: SynthesisHandle,\n): CancellablePromise<string> => {\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n return new CancellablePromise(async (resolve, _, onCancel) => {\n let fullText = '';\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n\n const ttsStream = handle.tts.stream();\n const readGeneratedAudio = async () => {\n for await (const audio of ttsStream) {\n if (cancelled) break;\n if (audio === SynthesizeStream.END_OF_STREAM) {\n break;\n }\n handle.queue.put(audio.frame);\n }\n handle.queue.put(SynthesisHandle.FLUSH_SENTINEL);\n };\n readGeneratedAudio();\n\n for await (const text of stream) {\n fullText += text;\n if (cancelled) break;\n ttsStream.pushText(text);\n }\n ttsStream.flush();\n ttsStream.endInput();\n\n resolve(fullText);\n });\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,iBAAoB;AACpB,iBAA2C;AAC3C,mBAAiF;AAK1E,MAAM,gBAAgB;AAAA,EAC3B,OAAgB,iBAAiB,OAAO,gBAAgB;AAAA,EAExD;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,QAAQ,IAAI,gCAAuE;AAAA,EACnF;AAAA,EACA,SAAS,IAAI,oBAAO;AAAA,EACpB,cAAU,gBAAI;AAAA,EAEd,YAAY,UAAkB,WAAyB,cAA4B,KAAU;AAC3F,SAAK,YAAY;AACjB,SAAK,YAAY;AACjB,SAAK,gBAAgB;AACrB,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,CAAC,CAAC,KAAK;AAAA,EAChB;AAAA,EAEA,IAAI,cAAuB;AACzB,WAAO,KAAK,OAAO;AAAA,EACrB;AAAA,EAEA,IAAI,aAAwC;AAC1C,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,OAAsB;AACpB,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,2BAA2B;AAAA,IAC7C;AAEA,SAAK,cAAc,KAAK,cAAc,KAAK,KAAK,WAAW,KAAK,KAAK;AACrE,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,YAAY;AA1Dd;AA2DI,QAAI,KAAK,aAAa;AACpB;AAAA,IACF;AAEA,SAAK,QAAQ,MAAM,EAAE,UAAU,KAAK,UAAU,CAAC,EAAE,MAAM,gCAAgC;AACvF,eAAK,gBAAL,mBAAkB;AAClB,SAAK,OAAO,QAAQ;AAAA,EACtB;AACF;AAEO,MAAM,YAAY;AAAA,EACvB;AAAA,EACA;AAAA,EACA,SAAqC,CAAC;AAAA,EAEtC,YAAY,cAA4B,KAAU;AAChD,SAAK,gBAAgB;AACrB,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,IAAI,UAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,OAAO,QAAQ,CAAC,SAAS,KAAK,OAAO,CAAC;AAC3C,UAAM,QAAQ,IAAI,KAAK,MAAM;AAAA,EAC/B;AAAA,EAEA,WAAW,UAAkB,WAA0C;AACrE,UAAM,SAAS,IAAI,gBAAgB,UAAU,WAAW,KAAK,eAAe,KAAK,IAAI;AACrF,UAAM,OAAO,KAAK,YAAY,MAAM;AACpC,SAAK,OAAO,KAAK,IAAI;AACrB,SAAK,QAAQ,MAAM,KAAK,OAAO,OAAO,KAAK,OAAO,QAAQ,IAAI,CAAC,CAAC;AAChE,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,QAAmD;AAE7D,WAAO,IAAI,gCAAmB,OAAO,SAAS,GAAG,aAAa;AAC5D,YAAM,YAAY,MAAM,OAAO;AAC/B,UAAI;AACJ,UAAI,OAAO,cAAc,UAAU;AACjC,eAAO,oBAAoB,WAAW,MAAM;AAAA,MAC9C,OAAO;AACL,eAAO,oBAAoB,WAAW,MAAM;AAAA,MAC9C;AAEA,eAAS,MAAM;AACb,2CAAiB,IAAI;AAAA,MACvB,CAAC;AAED,UAAI;AACF,cAAM,QAAQ,IAAI,CAAC,MAAM,OAAO,OAAO,KAAK,CAAC;AAAA,MAC/C,UAAE;AACA,YAAI,OAAO,OAAO,MAAM;AACtB,6CAAiB,IAAI;AAAA,QACvB,OAAO;AACL,eAAK,KAAK,CAAC,SAAS;AAClB,mBAAO,OAAO;AAAA,UAChB,CAAC;AAAA,QACH;AAAA,MACF;AAEA,cAAQ;AAAA,IACV,CAAC;AAAA,EACH;AACF;AAEA,MAAM,sBAAsB,CAAC,MAAc,WAAwD;AAEjG,SAAO,IAAI,gCAAmB,OAAO,SAAS,GAAG,aAAa;AAC5D,QAAI,YAAY;AAChB,aAAS,MAAM;AACb,kBAAY;AAAA,IACd,CAAC;AAED,UAAM,YAAY,OAAO,IAAI,OAAO;AACpC,cAAU,SAAS,IAAI;AACvB,cAAU,MAAM;AAChB,cAAU,SAAS;AACnB,qBAAiB,SAAS,WAAW;AACnC,UAAI,aAAa,UAAU,4BAAiB,eAAe;AACzD;AAAA,MACF;AACA,aAAO,MAAM,IAAI,MAAM,KAAK;AAAA,IAC9B;AACA,WAAO,MAAM,IAAI,gBAAgB,cAAc;AAE/C,YAAQ,IAAI;AAAA,EACd,CAAC;AACH;AAEA,MAAM,sBAAsB,CAC1B,QACA,WAC+B;AAE/B,SAAO,IAAI,gCAAmB,OAAO,SAAS,GAAG,aAAa;AAC5D,QAAI,WAAW;AACf,QAAI,YAAY;AAChB,aAAS,MAAM;AACb,kBAAY;AAAA,IACd,CAAC;AAED,UAAM,YAAY,OAAO,IAAI,OAAO;AACpC,UAAM,qBAAqB,YAAY;AACrC,uBAAiB,SAAS,WAAW;AACnC,YAAI,UAAW;AACf,YAAI,UAAU,4BAAiB,eAAe;AAC5C;AAAA,QACF;AACA,eAAO,MAAM,IAAI,MAAM,KAAK;AAAA,MAC9B;AACA,aAAO,MAAM,IAAI,gBAAgB,cAAc;AAAA,IACjD;AACA,uBAAmB;AAEnB,qBAAiB,QAAQ,QAAQ;AAC/B,kBAAY;AACZ,UAAI,UAAW;AACf,gBAAU,SAAS,IAAI;AAAA,IACzB;AACA,cAAU,MAAM;AAChB,cAAU,SAAS;AAEnB,YAAQ,QAAQ;AAAA,EAClB,CAAC;AACH;","names":[]}
|
|
@@ -6,6 +6,7 @@ export type SpeechSource = AsyncIterable<string> | string | Promise<string>;
|
|
|
6
6
|
export declare class SynthesisHandle {
|
|
7
7
|
#private;
|
|
8
8
|
static readonly FLUSH_SENTINEL: unique symbol;
|
|
9
|
+
text?: string;
|
|
9
10
|
ttsSource: SpeechSource;
|
|
10
11
|
tts: TTS;
|
|
11
12
|
queue: AsyncIterableQueue<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"agent_output.d.ts","sourceRoot":"","sources":["../../src/pipeline/agent_output.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAEpD,OAAO,EAAoB,KAAK,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAC7D,OAAO,EAAE,kBAAkB,EAAsB,MAAM,EAAoB,MAAM,aAAa,CAAC;AAC/F,OAAO,KAAK,EAAE,YAAY,EAAE,aAAa,EAAE,MAAM,oBAAoB,CAAC;AAEtE,MAAM,MAAM,YAAY,GAAG,aAAa,CAAC,MAAM,CAAC,GAAG,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;AAE5E,qBAAa,eAAe;;IAC1B,MAAM,CAAC,QAAQ,CAAC,cAAc,gBAA4B;IAG1D,SAAS,EAAE,YAAY,CAAC;IAExB,GAAG,EAAE,GAAG,CAAC;IACT,KAAK,yEAAgF;IAErF,MAAM,SAAgB;gBAGV,QAAQ,EAAE,MAAM,EAAE,SAAS,EAAE,YAAY,EAAE,YAAY,EAAE,YAAY,EAAE,GAAG,EAAE,GAAG;IAO3F,IAAI,QAAQ,IAAI,MAAM,CAErB;IAED,IAAI,SAAS,IAAI,OAAO,CAEvB;IAED,IAAI,WAAW,IAAI,OAAO,CAEzB;IAED,IAAI,UAAU,IAAI,aAAa,GAAG,SAAS,CAE1C;IAED,uCAAuC;IACvC,IAAI,IAAI,aAAa;IASrB,4BAA4B;IAC5B,SAAS;CASV;AAED,qBAAa,WAAW;;gBAKV,YAAY,EAAE,YAAY,EAAE,GAAG,EAAE,GAAG;IAKhD,IAAI,OAAO,IAAI,YAAY,CAE1B;IAEK,KAAK;IAKX,UAAU,CAAC,QAAQ,EAAE,MAAM,EAAE,SAAS,EAAE,YAAY,GAAG,eAAe;
|
|
1
|
+
{"version":3,"file":"agent_output.d.ts","sourceRoot":"","sources":["../../src/pipeline/agent_output.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAEpD,OAAO,EAAoB,KAAK,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAC7D,OAAO,EAAE,kBAAkB,EAAsB,MAAM,EAAoB,MAAM,aAAa,CAAC;AAC/F,OAAO,KAAK,EAAE,YAAY,EAAE,aAAa,EAAE,MAAM,oBAAoB,CAAC;AAEtE,MAAM,MAAM,YAAY,GAAG,aAAa,CAAC,MAAM,CAAC,GAAG,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;AAE5E,qBAAa,eAAe;;IAC1B,MAAM,CAAC,QAAQ,CAAC,cAAc,gBAA4B;IAG1D,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,SAAS,EAAE,YAAY,CAAC;IAExB,GAAG,EAAE,GAAG,CAAC;IACT,KAAK,yEAAgF;IAErF,MAAM,SAAgB;gBAGV,QAAQ,EAAE,MAAM,EAAE,SAAS,EAAE,YAAY,EAAE,YAAY,EAAE,YAAY,EAAE,GAAG,EAAE,GAAG;IAO3F,IAAI,QAAQ,IAAI,MAAM,CAErB;IAED,IAAI,SAAS,IAAI,OAAO,CAEvB;IAED,IAAI,WAAW,IAAI,OAAO,CAEzB;IAED,IAAI,UAAU,IAAI,aAAa,GAAG,SAAS,CAE1C;IAED,uCAAuC;IACvC,IAAI,IAAI,aAAa;IASrB,4BAA4B;IAC5B,SAAS;CASV;AAED,qBAAa,WAAW;;gBAKV,YAAY,EAAE,YAAY,EAAE,GAAG,EAAE,GAAG;IAKhD,IAAI,OAAO,IAAI,YAAY,CAE1B;IAEK,KAAK;IAKX,UAAU,CAAC,QAAQ,EAAE,MAAM,EAAE,SAAS,EAAE,YAAY,GAAG,eAAe;CAsCvE"}
|
|
@@ -4,6 +4,7 @@ import { AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from
|
|
|
4
4
|
class SynthesisHandle {
|
|
5
5
|
static FLUSH_SENTINEL = Symbol("FLUSH_SENTINEL");
|
|
6
6
|
#speechId;
|
|
7
|
+
text;
|
|
7
8
|
ttsSource;
|
|
8
9
|
#agentPlayout;
|
|
9
10
|
tts;
|
|
@@ -87,6 +88,10 @@ class AgentOutput {
|
|
|
87
88
|
} finally {
|
|
88
89
|
if (handle.intFut.done) {
|
|
89
90
|
gracefullyCancel(task);
|
|
91
|
+
} else {
|
|
92
|
+
task.then((text) => {
|
|
93
|
+
handle.text = text;
|
|
94
|
+
});
|
|
90
95
|
}
|
|
91
96
|
}
|
|
92
97
|
resolve();
|
|
@@ -110,11 +115,12 @@ const stringSynthesisTask = (text, handle) => {
|
|
|
110
115
|
handle.queue.put(audio.frame);
|
|
111
116
|
}
|
|
112
117
|
handle.queue.put(SynthesisHandle.FLUSH_SENTINEL);
|
|
113
|
-
resolve();
|
|
118
|
+
resolve(text);
|
|
114
119
|
});
|
|
115
120
|
};
|
|
116
121
|
const streamSynthesisTask = (stream, handle) => {
|
|
117
122
|
return new CancellablePromise(async (resolve, _, onCancel) => {
|
|
123
|
+
let fullText = "";
|
|
118
124
|
let cancelled = false;
|
|
119
125
|
onCancel(() => {
|
|
120
126
|
cancelled = true;
|
|
@@ -132,12 +138,13 @@ const streamSynthesisTask = (stream, handle) => {
|
|
|
132
138
|
};
|
|
133
139
|
readGeneratedAudio();
|
|
134
140
|
for await (const text of stream) {
|
|
141
|
+
fullText += text;
|
|
135
142
|
if (cancelled) break;
|
|
136
143
|
ttsStream.pushText(text);
|
|
137
144
|
}
|
|
138
145
|
ttsStream.flush();
|
|
139
146
|
ttsStream.endInput();
|
|
140
|
-
resolve();
|
|
147
|
+
resolve(fullText);
|
|
141
148
|
});
|
|
142
149
|
};
|
|
143
150
|
export {
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/pipeline/agent_output.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { log } from '../log.js';\nimport { SynthesizeStream, type TTS } from '../tts/index.js';\nimport { AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';\nimport type { AgentPlayout, PlayoutHandle } from './agent_playout.js';\n\nexport type SpeechSource = AsyncIterable<string> | string | Promise<string>;\n\nexport class SynthesisHandle {\n static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n\n #speechId: string;\n ttsSource: SpeechSource;\n #agentPlayout: AgentPlayout;\n tts: TTS;\n queue = new AsyncIterableQueue<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>();\n #playHandle?: PlayoutHandle;\n intFut = new Future();\n #logger = log();\n\n constructor(speechId: string, ttsSource: SpeechSource, agentPlayout: AgentPlayout, tts: TTS) {\n this.#speechId = speechId;\n this.ttsSource = ttsSource;\n this.#agentPlayout = agentPlayout;\n this.tts = tts;\n }\n\n get speechId(): string {\n return this.#speechId;\n }\n\n get validated(): boolean {\n return !!this.#playHandle;\n }\n\n get interrupted(): boolean {\n return this.intFut.done;\n }\n\n get playHandle(): PlayoutHandle | undefined {\n return this.#playHandle;\n }\n\n /** Validate the speech for playout. */\n play(): PlayoutHandle {\n if (this.interrupted) {\n throw new Error('synthesis was interrupted');\n }\n\n this.#playHandle = this.#agentPlayout.play(this.#speechId, this.queue);\n return this.#playHandle;\n }\n\n /** Interrupt the speech. */\n interrupt() {\n if (this.interrupted) {\n return;\n }\n\n this.#logger.child({ speechId: this.#speechId }).debug('interrupting synthesis/playout');\n this.#playHandle?.interrupt();\n this.intFut.resolve();\n }\n}\n\nexport class AgentOutput {\n #agentPlayout: AgentPlayout;\n #tts: TTS;\n #tasks: CancellablePromise<void>[] = [];\n\n constructor(agentPlayout: AgentPlayout, tts: TTS) {\n this.#agentPlayout = agentPlayout;\n this.#tts = tts;\n }\n\n get playout(): AgentPlayout {\n return this.#agentPlayout;\n }\n\n async close() {\n this.#tasks.forEach((task) => task.cancel());\n await Promise.all(this.#tasks);\n }\n\n synthesize(speechId: string, ttsSource: SpeechSource): SynthesisHandle {\n const handle = new SynthesisHandle(speechId, ttsSource, this.#agentPlayout, this.#tts);\n const task = this.#synthesize(handle);\n this.#tasks.push(task);\n task.finally(() => this.#tasks.splice(this.#tasks.indexOf(task)));\n return handle;\n }\n\n #synthesize(handle: SynthesisHandle): CancellablePromise<void> {\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n return new CancellablePromise(async (resolve, _, onCancel) => {\n const ttsSource = await handle.ttsSource;\n let task: CancellablePromise<
|
|
1
|
+
{"version":3,"sources":["../../src/pipeline/agent_output.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { log } from '../log.js';\nimport { SynthesizeStream, type TTS } from '../tts/index.js';\nimport { AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';\nimport type { AgentPlayout, PlayoutHandle } from './agent_playout.js';\n\nexport type SpeechSource = AsyncIterable<string> | string | Promise<string>;\n\nexport class SynthesisHandle {\n static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n\n #speechId: string;\n text?: string;\n ttsSource: SpeechSource;\n #agentPlayout: AgentPlayout;\n tts: TTS;\n queue = new AsyncIterableQueue<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>();\n #playHandle?: PlayoutHandle;\n intFut = new Future();\n #logger = log();\n\n constructor(speechId: string, ttsSource: SpeechSource, agentPlayout: AgentPlayout, tts: TTS) {\n this.#speechId = speechId;\n this.ttsSource = ttsSource;\n this.#agentPlayout = agentPlayout;\n this.tts = tts;\n }\n\n get speechId(): string {\n return this.#speechId;\n }\n\n get validated(): boolean {\n return !!this.#playHandle;\n }\n\n get interrupted(): boolean {\n return this.intFut.done;\n }\n\n get playHandle(): PlayoutHandle | undefined {\n return this.#playHandle;\n }\n\n /** Validate the speech for playout. */\n play(): PlayoutHandle {\n if (this.interrupted) {\n throw new Error('synthesis was interrupted');\n }\n\n this.#playHandle = this.#agentPlayout.play(this.#speechId, this.queue);\n return this.#playHandle;\n }\n\n /** Interrupt the speech. */\n interrupt() {\n if (this.interrupted) {\n return;\n }\n\n this.#logger.child({ speechId: this.#speechId }).debug('interrupting synthesis/playout');\n this.#playHandle?.interrupt();\n this.intFut.resolve();\n }\n}\n\nexport class AgentOutput {\n #agentPlayout: AgentPlayout;\n #tts: TTS;\n #tasks: CancellablePromise<void>[] = [];\n\n constructor(agentPlayout: AgentPlayout, tts: TTS) {\n this.#agentPlayout = agentPlayout;\n this.#tts = tts;\n }\n\n get playout(): AgentPlayout {\n return this.#agentPlayout;\n }\n\n async close() {\n this.#tasks.forEach((task) => task.cancel());\n await Promise.all(this.#tasks);\n }\n\n synthesize(speechId: string, ttsSource: SpeechSource): SynthesisHandle {\n const handle = new SynthesisHandle(speechId, ttsSource, this.#agentPlayout, this.#tts);\n const task = this.#synthesize(handle);\n this.#tasks.push(task);\n task.finally(() => this.#tasks.splice(this.#tasks.indexOf(task)));\n return handle;\n }\n\n #synthesize(handle: SynthesisHandle): CancellablePromise<void> {\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n return new CancellablePromise(async (resolve, _, onCancel) => {\n const ttsSource = await handle.ttsSource;\n let task: CancellablePromise<string>;\n if (typeof ttsSource === 'string') {\n task = stringSynthesisTask(ttsSource, handle);\n } else {\n task = streamSynthesisTask(ttsSource, handle);\n }\n\n onCancel(() => {\n gracefullyCancel(task);\n });\n\n try {\n await Promise.any([task, handle.intFut.await]);\n } finally {\n if (handle.intFut.done) {\n gracefullyCancel(task);\n } else {\n task.then((text) => {\n handle.text = text;\n });\n }\n }\n\n resolve();\n });\n }\n}\n\nconst stringSynthesisTask = (text: string, handle: SynthesisHandle): CancellablePromise<string> => {\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n return new CancellablePromise(async (resolve, _, onCancel) => {\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n\n const ttsStream = handle.tts.stream();\n ttsStream.pushText(text);\n ttsStream.flush();\n ttsStream.endInput();\n for await (const audio of ttsStream) {\n if (cancelled || audio === SynthesizeStream.END_OF_STREAM) {\n break;\n }\n handle.queue.put(audio.frame);\n }\n handle.queue.put(SynthesisHandle.FLUSH_SENTINEL);\n\n resolve(text);\n });\n};\n\nconst streamSynthesisTask = (\n stream: AsyncIterable<string>,\n handle: SynthesisHandle,\n): CancellablePromise<string> => {\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n return new CancellablePromise(async (resolve, _, onCancel) => {\n let fullText = '';\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n\n const ttsStream = handle.tts.stream();\n const readGeneratedAudio = async () => {\n for await (const audio of ttsStream) {\n if (cancelled) break;\n if (audio === SynthesizeStream.END_OF_STREAM) {\n break;\n }\n handle.queue.put(audio.frame);\n }\n handle.queue.put(SynthesisHandle.FLUSH_SENTINEL);\n };\n readGeneratedAudio();\n\n for await (const text of stream) {\n fullText += text;\n if (cancelled) break;\n ttsStream.pushText(text);\n }\n ttsStream.flush();\n ttsStream.endInput();\n\n resolve(fullText);\n });\n};\n"],"mappings":"AAIA,SAAS,WAAW;AACpB,SAAS,wBAAkC;AAC3C,SAAS,oBAAoB,oBAAoB,QAAQ,wBAAwB;AAK1E,MAAM,gBAAgB;AAAA,EAC3B,OAAgB,iBAAiB,OAAO,gBAAgB;AAAA,EAExD;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,QAAQ,IAAI,mBAAuE;AAAA,EACnF;AAAA,EACA,SAAS,IAAI,OAAO;AAAA,EACpB,UAAU,IAAI;AAAA,EAEd,YAAY,UAAkB,WAAyB,cAA4B,KAAU;AAC3F,SAAK,YAAY;AACjB,SAAK,YAAY;AACjB,SAAK,gBAAgB;AACrB,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,CAAC,CAAC,KAAK;AAAA,EAChB;AAAA,EAEA,IAAI,cAAuB;AACzB,WAAO,KAAK,OAAO;AAAA,EACrB;AAAA,EAEA,IAAI,aAAwC;AAC1C,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,OAAsB;AACpB,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,2BAA2B;AAAA,IAC7C;AAEA,SAAK,cAAc,KAAK,cAAc,KAAK,KAAK,WAAW,KAAK,KAAK;AACrE,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,YAAY;AA1Dd;AA2DI,QAAI,KAAK,aAAa;AACpB;AAAA,IACF;AAEA,SAAK,QAAQ,MAAM,EAAE,UAAU,KAAK,UAAU,CAAC,EAAE,MAAM,gCAAgC;AACvF,eAAK,gBAAL,mBAAkB;AAClB,SAAK,OAAO,QAAQ;AAAA,EACtB;AACF;AAEO,MAAM,YAAY;AAAA,EACvB;AAAA,EACA;AAAA,EACA,SAAqC,CAAC;AAAA,EAEtC,YAAY,cAA4B,KAAU;AAChD,SAAK,gBAAgB;AACrB,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,IAAI,UAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,OAAO,QAAQ,CAAC,SAAS,KAAK,OAAO,CAAC;AAC3C,UAAM,QAAQ,IAAI,KAAK,MAAM;AAAA,EAC/B;AAAA,EAEA,WAAW,UAAkB,WAA0C;AACrE,UAAM,SAAS,IAAI,gBAAgB,UAAU,WAAW,KAAK,eAAe,KAAK,IAAI;AACrF,UAAM,OAAO,KAAK,YAAY,MAAM;AACpC,SAAK,OAAO,KAAK,IAAI;AACrB,SAAK,QAAQ,MAAM,KAAK,OAAO,OAAO,KAAK,OAAO,QAAQ,IAAI,CAAC,CAAC;AAChE,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,QAAmD;AAE7D,WAAO,IAAI,mBAAmB,OAAO,SAAS,GAAG,aAAa;AAC5D,YAAM,YAAY,MAAM,OAAO;AAC/B,UAAI;AACJ,UAAI,OAAO,cAAc,UAAU;AACjC,eAAO,oBAAoB,WAAW,MAAM;AAAA,MAC9C,OAAO;AACL,eAAO,oBAAoB,WAAW,MAAM;AAAA,MAC9C;AAEA,eAAS,MAAM;AACb,yBAAiB,IAAI;AAAA,MACvB,CAAC;AAED,UAAI;AACF,cAAM,QAAQ,IAAI,CAAC,MAAM,OAAO,OAAO,KAAK,CAAC;AAAA,MAC/C,UAAE;AACA,YAAI,OAAO,OAAO,MAAM;AACtB,2BAAiB,IAAI;AAAA,QACvB,OAAO;AACL,eAAK,KAAK,CAAC,SAAS;AAClB,mBAAO,OAAO;AAAA,UAChB,CAAC;AAAA,QACH;AAAA,MACF;AAEA,cAAQ;AAAA,IACV,CAAC;AAAA,EACH;AACF;AAEA,MAAM,sBAAsB,CAAC,MAAc,WAAwD;AAEjG,SAAO,IAAI,mBAAmB,OAAO,SAAS,GAAG,aAAa;AAC5D,QAAI,YAAY;AAChB,aAAS,MAAM;AACb,kBAAY;AAAA,IACd,CAAC;AAED,UAAM,YAAY,OAAO,IAAI,OAAO;AACpC,cAAU,SAAS,IAAI;AACvB,cAAU,MAAM;AAChB,cAAU,SAAS;AACnB,qBAAiB,SAAS,WAAW;AACnC,UAAI,aAAa,UAAU,iBAAiB,eAAe;AACzD;AAAA,MACF;AACA,aAAO,MAAM,IAAI,MAAM,KAAK;AAAA,IAC9B;AACA,WAAO,MAAM,IAAI,gBAAgB,cAAc;AAE/C,YAAQ,IAAI;AAAA,EACd,CAAC;AACH;AAEA,MAAM,sBAAsB,CAC1B,QACA,WAC+B;AAE/B,SAAO,IAAI,mBAAmB,OAAO,SAAS,GAAG,aAAa;AAC5D,QAAI,WAAW;AACf,QAAI,YAAY;AAChB,aAAS,MAAM;AACb,kBAAY;AAAA,IACd,CAAC;AAED,UAAM,YAAY,OAAO,IAAI,OAAO;AACpC,UAAM,qBAAqB,YAAY;AACrC,uBAAiB,SAAS,WAAW;AACnC,YAAI,UAAW;AACf,YAAI,UAAU,iBAAiB,eAAe;AAC5C;AAAA,QACF;AACA,eAAO,MAAM,IAAI,MAAM,KAAK;AAAA,MAC9B;AACA,aAAO,MAAM,IAAI,gBAAgB,cAAc;AAAA,IACjD;AACA,uBAAmB;AAEnB,qBAAiB,QAAQ,QAAQ;AAC/B,kBAAY;AACZ,UAAI,UAAW;AACf,gBAAU,SAAS,IAAI;AAAA,IACzB;AACA,cAAU,MAAM;AAChB,cAAU,SAAS;AAEnB,YAAQ,QAAQ;AAAA,EAClB,CAAC;AACH;","names":[]}
|
package/dist/pipeline/index.cjs
CHANGED
|
@@ -18,6 +18,7 @@ var __copyProps = (to, from, except, desc) => {
|
|
|
18
18
|
var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
|
|
19
19
|
var pipeline_exports = {};
|
|
20
20
|
__export(pipeline_exports, {
|
|
21
|
+
AgentCallContext: () => import_pipeline_agent.AgentCallContext,
|
|
21
22
|
VPAEvent: () => import_pipeline_agent.VPAEvent,
|
|
22
23
|
VoicePipelineAgent: () => import_pipeline_agent.VoicePipelineAgent
|
|
23
24
|
});
|
|
@@ -25,6 +26,7 @@ module.exports = __toCommonJS(pipeline_exports);
|
|
|
25
26
|
var import_pipeline_agent = require("./pipeline_agent.cjs");
|
|
26
27
|
// Annotate the CommonJS export names for ESM import in node:
|
|
27
28
|
0 && (module.exports = {
|
|
29
|
+
AgentCallContext,
|
|
28
30
|
VPAEvent,
|
|
29
31
|
VoicePipelineAgent
|
|
30
32
|
});
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/pipeline/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport {\n type AgentState,\n type BeforeTTSCallback,\n type BeforeLLMCallback,\n type VPACallbacks,\n type
|
|
1
|
+
{"version":3,"sources":["../../src/pipeline/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport {\n type AgentState,\n type BeforeTTSCallback,\n type BeforeLLMCallback,\n type VPACallbacks,\n type AgentTranscriptionOptions,\n type VPAOptions,\n VPAEvent,\n VoicePipelineAgent,\n AgentCallContext,\n} from './pipeline_agent.js';\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,4BAUO;","names":[]}
|
package/dist/pipeline/index.d.ts
CHANGED
|
@@ -1,2 +1,2 @@
|
|
|
1
|
-
export { type AgentState, type BeforeTTSCallback, type BeforeLLMCallback, type VPACallbacks, type
|
|
1
|
+
export { type AgentState, type BeforeTTSCallback, type BeforeLLMCallback, type VPACallbacks, type AgentTranscriptionOptions, type VPAOptions, VPAEvent, VoicePipelineAgent, AgentCallContext, } from './pipeline_agent.js';
|
|
2
2
|
//# sourceMappingURL=index.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/pipeline/index.ts"],"names":[],"mappings":"AAIA,OAAO,EACL,KAAK,UAAU,EACf,KAAK,iBAAiB,EACtB,KAAK,iBAAiB,EACtB,KAAK,YAAY,EACjB,KAAK,
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/pipeline/index.ts"],"names":[],"mappings":"AAIA,OAAO,EACL,KAAK,UAAU,EACf,KAAK,iBAAiB,EACtB,KAAK,iBAAiB,EACtB,KAAK,YAAY,EACjB,KAAK,yBAAyB,EAC9B,KAAK,UAAU,EACf,QAAQ,EACR,kBAAkB,EAClB,gBAAgB,GACjB,MAAM,qBAAqB,CAAC"}
|
package/dist/pipeline/index.js
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/pipeline/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport {\n type AgentState,\n type BeforeTTSCallback,\n type BeforeLLMCallback,\n type VPACallbacks,\n type
|
|
1
|
+
{"version":3,"sources":["../../src/pipeline/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport {\n type AgentState,\n type BeforeTTSCallback,\n type BeforeLLMCallback,\n type VPACallbacks,\n type AgentTranscriptionOptions,\n type VPAOptions,\n VPAEvent,\n VoicePipelineAgent,\n AgentCallContext,\n} from './pipeline_agent.js';\n"],"mappings":"AAIA;AAAA,EAOE;AAAA,EACA;AAAA,EACA;AAAA,OACK;","names":[]}
|