@livekit/agents 0.7.8 → 0.7.9
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/constants.cjs +3 -0
- package/dist/constants.cjs.map +1 -1
- package/dist/constants.d.cts +1 -0
- package/dist/constants.d.ts +1 -0
- package/dist/constants.d.ts.map +1 -1
- package/dist/constants.js +2 -0
- package/dist/constants.js.map +1 -1
- package/dist/multimodal/multimodal_agent.cjs +41 -6
- package/dist/multimodal/multimodal_agent.cjs.map +1 -1
- package/dist/multimodal/multimodal_agent.d.ts.map +1 -1
- package/dist/multimodal/multimodal_agent.js +42 -6
- package/dist/multimodal/multimodal_agent.js.map +1 -1
- package/dist/pipeline/pipeline_agent.cjs +13 -3
- package/dist/pipeline/pipeline_agent.cjs.map +1 -1
- package/dist/pipeline/pipeline_agent.d.ts.map +1 -1
- package/dist/pipeline/pipeline_agent.js +14 -3
- package/dist/pipeline/pipeline_agent.js.map +1 -1
- package/package.json +1 -1
- package/src/constants.ts +1 -0
- package/src/multimodal/multimodal_agent.ts +39 -2
- package/src/pipeline/pipeline_agent.ts +12 -0
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/pipeline/pipeline_agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n LocalTrackPublication,\n NoiseCancellationOptions,\n RemoteParticipant,\n Room,\n} from '@livekit/rtc-node';\nimport {\n AudioSource,\n LocalAudioTrack,\n RoomEvent,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { randomUUID } from 'node:crypto';\nimport EventEmitter from 'node:events';\nimport {\n ATTRIBUTE_TRANSCRIPTION_FINAL,\n ATTRIBUTE_TRANSCRIPTION_TRACK_ID,\n TOPIC_TRANSCRIPTION,\n} from '../constants.js';\nimport type {\n CallableFunctionResult,\n FunctionCallInfo,\n FunctionContext,\n LLM,\n} from '../llm/index.js';\nimport { ChatContext, ChatMessage, ChatRole, LLMEvent, LLMStream } from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { AgentMetrics, PipelineEOUMetrics } from '../metrics/base.js';\nimport { type STT, StreamAdapter as STTStreamAdapter, SpeechEventType } from '../stt/index.js';\nimport {\n SentenceTokenizer as BasicSentenceTokenizer,\n WordTokenizer as BasicWordTokenizer,\n hyphenateWord,\n} from '../tokenize/basic/index.js';\nimport type { SentenceTokenizer, WordTokenizer } from '../tokenize/tokenizer.js';\nimport { TextAudioSynchronizer, defaultTextSyncOptions } from '../transcription.js';\nimport type { TTS } from '../tts/index.js';\nimport { TTSEvent, StreamAdapter as TTSStreamAdapter } from '../tts/index.js';\nimport { AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';\nimport { type VAD, type VADEvent, VADEventType } from '../vad.js';\nimport type { SpeechSource, SynthesisHandle } from './agent_output.js';\nimport { AgentOutput } from './agent_output.js';\nimport { AgentPlayout, AgentPlayoutEvent } from './agent_playout.js';\nimport { HumanInput, HumanInputEvent } from './human_input.js';\nimport { SpeechHandle } from './speech_handle.js';\n\nexport type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';\nexport const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';\nlet lastSpeechData: { sequenceId: string } | undefined;\nlet speechData: { sequenceId: string } | undefined;\n\nexport type BeforeLLMCallback = (\n agent: VoicePipelineAgent,\n chatCtx: ChatContext,\n) => LLMStream | false | void | Promise<LLMStream | false | void>;\n\nexport type BeforeTTSCallback = (\n agent: VoicePipelineAgent,\n source: string | AsyncIterable<string>,\n) => SpeechSource;\n\nexport enum VPAEvent {\n USER_STARTED_SPEAKING,\n USER_STOPPED_SPEAKING,\n AGENT_STARTED_SPEAKING,\n AGENT_STOPPED_SPEAKING,\n USER_SPEECH_COMMITTED,\n AGENT_SPEECH_COMMITTED,\n AGENT_SPEECH_INTERRUPTED,\n FUNCTION_CALLS_COLLECTED,\n FUNCTION_CALLS_FINISHED,\n METRICS_COLLECTED,\n}\n\nexport type VPACallbacks = {\n [VPAEvent.USER_STARTED_SPEAKING]: () => void;\n [VPAEvent.USER_STOPPED_SPEAKING]: () => void;\n [VPAEvent.AGENT_STARTED_SPEAKING]: () => void;\n [VPAEvent.AGENT_STOPPED_SPEAKING]: () => void;\n [VPAEvent.USER_SPEECH_COMMITTED]: (msg: ChatMessage) => void;\n [VPAEvent.AGENT_SPEECH_COMMITTED]: (msg: ChatMessage) => void;\n [VPAEvent.AGENT_SPEECH_INTERRUPTED]: (msg: ChatMessage) => void;\n [VPAEvent.FUNCTION_CALLS_COLLECTED]: (funcs: FunctionCallInfo[]) => void;\n [VPAEvent.FUNCTION_CALLS_FINISHED]: (funcs: CallableFunctionResult[]) => void;\n [VPAEvent.METRICS_COLLECTED]: (metrics: AgentMetrics) => void;\n};\n\ninterface TurnDetector {\n unlikelyThreshold: number;\n supportsLanguage: (language?: string) => boolean;\n predictEndOfTurn: (chatCtx: ChatContext) => Promise<number>;\n}\n\nexport class AgentCallContext {\n #agent: VoicePipelineAgent;\n #llmStream: LLMStream;\n #metadata = new Map<string, any>();\n #extraChatMessages: ChatMessage[] = [];\n static #current: AgentCallContext;\n\n constructor(agent: VoicePipelineAgent, llmStream: LLMStream) {\n this.#agent = agent;\n this.#llmStream = llmStream;\n AgentCallContext.#current = this;\n }\n\n static getCurrent(): AgentCallContext {\n return AgentCallContext.#current;\n }\n\n get agent(): VoicePipelineAgent {\n return this.#agent;\n }\n\n storeMetadata(key: string, value: any) {\n this.#metadata.set(key, value);\n }\n\n getMetadata(key: string, orDefault: any = undefined) {\n return this.#metadata.get(key) || orDefault;\n }\n\n get llmStream(): LLMStream {\n return this.#llmStream;\n }\n\n get extraChatMessages() {\n return this.#extraChatMessages;\n }\n\n addExtraChatMessage(message: ChatMessage) {\n this.#extraChatMessages.push(message);\n }\n}\n\nconst defaultBeforeLLMCallback: BeforeLLMCallback = (\n agent: VoicePipelineAgent,\n chatCtx: ChatContext,\n): LLMStream => {\n return agent.llm.chat({ chatCtx, fncCtx: agent.fncCtx });\n};\n\nconst defaultBeforeTTSCallback: BeforeTTSCallback = (\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n _: VoicePipelineAgent,\n text: string | AsyncIterable<string>,\n): string | AsyncIterable<string> => {\n return text;\n};\n\nexport interface AgentTranscriptionOptions {\n /** Whether to forward the user transcription to the client */\n userTranscription: boolean;\n /** Whether to forward the agent transcription to the client */\n agentTranscription: boolean;\n /**\n * The speed at which the agent's speech transcription is forwarded to the client.\n * We try to mimic the agent's speech speed by adjusting the transcription speed.\n */\n agentTranscriptionSpeech: number;\n /**\n * The tokenizer used to split the speech into sentences.\n * This is used to decide when to mark a transcript as final for the agent transcription.\n */\n sentenceTokenizer: SentenceTokenizer;\n /**\n * The tokenizer used to split the speech into words.\n * This is used to simulate the \"interim results\" of the agent transcription.\n */\n wordTokenizer: WordTokenizer;\n /**\n * A function that takes a string (word) as input and returns a list of strings,\n * representing the hyphenated parts of the word.\n */\n hyphenateWord: (word: string) => string[];\n}\n\nconst defaultAgentTranscriptionOptions: AgentTranscriptionOptions = {\n userTranscription: true,\n agentTranscription: true,\n agentTranscriptionSpeech: 1,\n sentenceTokenizer: new BasicSentenceTokenizer(),\n wordTokenizer: new BasicWordTokenizer(false),\n hyphenateWord: hyphenateWord,\n};\n\nexport interface VPAOptions {\n /** Chat context for the assistant. */\n chatCtx?: ChatContext;\n /** Function context for the assistant. */\n fncCtx?: FunctionContext;\n /** Whether to allow the user to interrupt the assistant. */\n allowInterruptions: boolean;\n /** Minimum duration of speech to consider for interruption. */\n interruptSpeechDuration: number;\n /** Minimum number of words to consider for interuption. This may increase latency. */\n interruptMinWords: number;\n /** Delay to wait before considering the user speech done. */\n minEndpointingDelay: number;\n maxNestedFncCalls: number;\n /* Whether to preemptively synthesize responses. */\n preemptiveSynthesis: boolean;\n /*\n * Callback called when the assistant is about to synthesize a reply.\n *\n * @remarks\n * Returning void will create a default LLM stream.\n * You can also return your own LLM stream by calling `llm.chat()`.\n * Returning `false` ill cancel the synthesis of the reply.\n */\n beforeLLMCallback: BeforeLLMCallback;\n /*\n * Callback called when the assistant is about to synthesize speech.\n *\n * @remarks\n * This can be used to customize text before synthesis\n * (e.g. editing the pronunciation of a word).\n */\n beforeTTSCallback: BeforeTTSCallback;\n /** Options for assistant transcription. */\n transcription: AgentTranscriptionOptions;\n /** Turn detection model to use. */\n turnDetector?: TurnDetector;\n /** Noise cancellation options. */\n noiseCancellation?: NoiseCancellationOptions;\n}\n\nconst defaultVPAOptions: VPAOptions = {\n chatCtx: new ChatContext(),\n allowInterruptions: true,\n interruptSpeechDuration: 50,\n interruptMinWords: 0,\n minEndpointingDelay: 500,\n maxNestedFncCalls: 1,\n preemptiveSynthesis: false,\n beforeLLMCallback: defaultBeforeLLMCallback,\n beforeTTSCallback: defaultBeforeTTSCallback,\n transcription: defaultAgentTranscriptionOptions,\n};\n\n/** A pipeline agent (VAD + STT + LLM + TTS) implementation. */\nexport class VoicePipelineAgent extends (EventEmitter as new () => TypedEmitter<VPACallbacks>) {\n /** Minimum time played for the user speech to be committed to the chat context. */\n readonly MIN_TIME_PLAYED_FOR_COMMIT = 1.5;\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n\n #vad: VAD;\n #stt: STT;\n #llm: LLM;\n #tts: TTS;\n #opts: VPAOptions;\n #humanInput?: HumanInput;\n #agentOutput?: AgentOutput;\n #trackPublishedFut = new Future();\n #pendingAgentReply?: SpeechHandle;\n #agentReplyTask?: CancellablePromise<void>;\n #playingSpeech?: SpeechHandle;\n transcribedText = '';\n #transcribedInterimText = '';\n #speechQueueOpen = new Future();\n #speechQueue = new AsyncIterableQueue<SpeechHandle | typeof VoicePipelineAgent.FLUSH_SENTINEL>();\n #updateStateTask?: CancellablePromise<void>;\n #started = false;\n #room?: Room;\n #participant: RemoteParticipant | string | null = null;\n #deferredValidation: DeferredReplyValidation;\n #logger = log();\n #agentPublication?: LocalTrackPublication;\n #lastFinalTranscriptTime?: number;\n #lastSpeechTime?: number;\n #transcriptionId?: string;\n #agentTranscribedText = '';\n\n constructor(\n /** Voice Activity Detection instance. */\n vad: VAD,\n /** Speech-to-Text instance. */\n stt: STT,\n /** Large Language Model instance. */\n llm: LLM,\n /** Text-to-Speech instance. */\n tts: TTS,\n /** Additional VoicePipelineAgent options. */\n opts: Partial<VPAOptions> = defaultVPAOptions,\n ) {\n super();\n\n this.#opts = { ...defaultVPAOptions, ...opts };\n\n if (!stt.capabilities.streaming) {\n stt = new STTStreamAdapter(stt, vad);\n }\n\n if (!tts.capabilities.streaming) {\n tts = new TTSStreamAdapter(tts, new BasicSentenceTokenizer());\n }\n\n this.#vad = vad;\n this.#stt = stt;\n this.#llm = llm;\n this.#tts = tts;\n\n this.#deferredValidation = new DeferredReplyValidation(\n this.#validateReplyIfPossible.bind(this),\n this.#opts.minEndpointingDelay,\n this,\n this.#opts.turnDetector,\n );\n }\n\n get fncCtx(): FunctionContext | undefined {\n return this.#opts.fncCtx;\n }\n\n set fncCtx(ctx: FunctionContext) {\n this.#opts.fncCtx = ctx;\n }\n\n get chatCtx(): ChatContext {\n return this.#opts.chatCtx!;\n }\n\n get llm(): LLM {\n return this.#llm;\n }\n\n get tts(): TTS {\n return this.#tts;\n }\n\n get stt(): STT {\n return this.#stt;\n }\n\n get vad(): VAD {\n return this.#vad;\n }\n\n /** Start the voice assistant. */\n start(\n /** The room to connect to. */\n room: Room,\n /**\n * The participant to listen to.\n *\n * @remarks\n * Can be a participant or an identity.\n * If omitted, the first participant in the room will be selected.\n */\n participant: RemoteParticipant | string | null = null,\n ) {\n if (this.#started) {\n throw new Error('voice assistant already started');\n }\n\n this.#stt.on(SpeechEventType.METRICS_COLLECTED, (metrics) => {\n this.emit(VPAEvent.METRICS_COLLECTED, metrics);\n });\n\n this.#tts.on(TTSEvent.METRICS_COLLECTED, (metrics) => {\n if (!speechData) return;\n this.emit(VPAEvent.METRICS_COLLECTED, { ...metrics, sequenceId: speechData.sequenceId });\n });\n\n this.#llm.on(LLMEvent.METRICS_COLLECTED, (metrics) => {\n const sequenceId = speechData ? speechData.sequenceId : lastSpeechData?.sequenceId;\n if (!sequenceId) return;\n\n this.emit(VPAEvent.METRICS_COLLECTED, { ...metrics, sequenceId });\n });\n\n this.#vad.on(VADEventType.METRICS_COLLECTED, (metrics) => {\n this.emit(VPAEvent.METRICS_COLLECTED, metrics);\n });\n\n room.on(RoomEvent.ParticipantConnected, (participant: RemoteParticipant) => {\n // automatically link to the first participant that connects, if not already linked\n if (this.#participant) {\n return;\n }\n this.#linkParticipant.call(this, participant.identity!);\n });\n\n this.#room = room;\n this.#participant = participant;\n\n if (participant) {\n if (typeof participant === 'string') {\n this.#linkParticipant(participant);\n } else {\n this.#linkParticipant(participant.identity!);\n }\n }\n\n this.#run();\n }\n\n /** Play a speech source through the voice assistant. */\n async say(\n source: string | LLMStream | AsyncIterable<string>,\n allowInterruptions = true,\n addToChatCtx = true,\n ): Promise<SpeechHandle> {\n await this.#trackPublishedFut.await;\n\n let callContext: AgentCallContext | undefined;\n let fncSource: string | AsyncIterable<string> | undefined;\n if (addToChatCtx) {\n callContext = AgentCallContext.getCurrent();\n if (source instanceof LLMStream) {\n this.#logger.warn('LLMStream will be ignored for function call chat context');\n } else if (typeof source === 'string') {\n fncSource = source;\n } else {\n fncSource = source;\n source = new AsyncIterableQueue<string>();\n }\n }\n\n const newHandle = SpeechHandle.createAssistantSpeech(allowInterruptions, addToChatCtx);\n const synthesisHandle = this.#synthesizeAgentSpeech(newHandle.id, source);\n newHandle.initialize(source, synthesisHandle);\n\n if (this.#playingSpeech && !this.#playingSpeech.nestedSpeechFinished) {\n this.#playingSpeech.addNestedSpeech(newHandle);\n } else {\n this.#addSpeechForPlayout(newHandle);\n }\n\n if (callContext && fncSource) {\n let text: string;\n if (typeof source === 'string') {\n text = fncSource as string;\n } else {\n text = '';\n for await (const chunk of fncSource) {\n (source as AsyncIterableQueue<string>).put(chunk);\n text += chunk;\n }\n (source as AsyncIterableQueue<string>).close();\n }\n\n callContext.addExtraChatMessage(ChatMessage.create({ text, role: ChatRole.ASSISTANT }));\n this.#logger.child({ text }).debug('added speech to function call chat context');\n }\n\n return newHandle;\n }\n\n #updateState(state: AgentState, delay = 0) {\n const runTask = (delay: number): CancellablePromise<void> => {\n return new CancellablePromise(async (resolve, _, onCancel) => {\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n await new Promise((resolve) => setTimeout(resolve, delay));\n if (this.#room?.isConnected) {\n if (!cancelled) {\n await this.#room.localParticipant?.setAttributes({ [AGENT_STATE_ATTRIBUTE]: state });\n }\n }\n resolve();\n });\n };\n\n if (this.#updateStateTask) {\n this.#updateStateTask.cancel();\n }\n\n this.#updateStateTask = runTask(delay);\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.#participant = this.#room.remoteParticipants.get(participantIdentity) || null;\n if (!this.#participant) {\n this.#logger.error(`Participant with identity ${participantIdentity} not found`);\n return;\n }\n\n this.#humanInput = new HumanInput(\n this.#room,\n this.#vad,\n this.#stt,\n this.#participant,\n this.#opts.noiseCancellation,\n );\n this.#humanInput.on(HumanInputEvent.START_OF_SPEECH, (event) => {\n this.emit(VPAEvent.USER_STARTED_SPEAKING);\n this.#deferredValidation.onHumanStartOfSpeech(event);\n });\n this.#humanInput.on(HumanInputEvent.VAD_INFERENCE_DONE, (event) => {\n if (!this.#trackPublishedFut.done) {\n return;\n }\n if (!this.#agentOutput) {\n throw new Error('agent output is undefined');\n }\n\n let tv = 1;\n if (this.#opts.allowInterruptions) {\n tv = Math.max(0, 1 - event.probability);\n this.#agentOutput.playout.targetVolume = tv;\n }\n\n if (event.speechDuration >= this.#opts.interruptSpeechDuration) {\n this.#interruptIfPossible();\n }\n\n if (event.rawAccumulatedSpeech > 0) {\n this.#lastSpeechTime = Date.now() - event.rawAccumulatedSilence;\n }\n });\n this.#humanInput.on(HumanInputEvent.END_OF_SPEECH, (event) => {\n this.emit(VPAEvent.USER_STOPPED_SPEAKING);\n this.#deferredValidation.onHumanEndOfSpeech(event);\n });\n this.#humanInput.on(HumanInputEvent.INTERIM_TRANSCRIPT, async (event) => {\n if (!this.#transcriptionId) {\n this.#transcriptionId = randomUUID();\n }\n this.#transcribedInterimText = event.alternatives![0].text;\n\n await this.#publishTranscription(\n this.#humanInput!.participant.identity,\n this.#humanInput!.subscribedTrack!.sid!,\n this.#transcribedInterimText,\n false,\n this.#transcriptionId,\n );\n });\n this.#humanInput.on(HumanInputEvent.FINAL_TRANSCRIPT, async (event) => {\n const newTranscript = event.alternatives![0].text;\n if (!newTranscript) return;\n\n if (!this.#transcriptionId) {\n this.#transcriptionId = randomUUID();\n }\n\n this.#lastFinalTranscriptTime = Date.now();\n this.transcribedText += (this.transcribedText ? ' ' : '') + newTranscript;\n\n await this.#publishTranscription(\n this.#humanInput!.participant.identity,\n this.#humanInput!.subscribedTrack!.sid!,\n this.transcribedText,\n true,\n this.#transcriptionId,\n );\n\n this.#transcriptionId = undefined;\n\n if (\n this.#opts.preemptiveSynthesis &&\n (!this.#playingSpeech || this.#playingSpeech.allowInterruptions)\n ) {\n this.#synthesizeAgentReply();\n }\n\n this.#deferredValidation.onHumanFinalTranscript(newTranscript);\n\n const words = this.#opts.transcription.wordTokenizer.tokenize(newTranscript);\n if (words.length >= 3) {\n // VAD can sometimes not detect that the human is speaking.\n // to make the interruption more reliable, we also interrupt on the final transcript.\n this.#interruptIfPossible();\n }\n });\n }\n\n async #run() {\n this.#updateState('initializing');\n const audioSource = new AudioSource(this.#tts.sampleRate, this.#tts.numChannels);\n const track = LocalAudioTrack.createAudioTrack('assistant_voice', audioSource);\n this.#agentPublication = await this.#room?.localParticipant?.publishTrack(\n track,\n new TrackPublishOptions({ source: TrackSource.SOURCE_MICROPHONE }),\n );\n\n const agentPlayout = new AgentPlayout(audioSource);\n this.#agentOutput = new AgentOutput(agentPlayout, this.#tts);\n\n agentPlayout.on(AgentPlayoutEvent.PLAYOUT_STARTED, () => {\n this.emit(VPAEvent.AGENT_STARTED_SPEAKING);\n this.#updateState('speaking');\n });\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n agentPlayout.on(AgentPlayoutEvent.PLAYOUT_STOPPED, (_) => {\n this.emit(VPAEvent.AGENT_STOPPED_SPEAKING);\n this.#updateState('listening');\n });\n\n this.#trackPublishedFut.resolve();\n\n while (true) {\n await this.#speechQueueOpen.await;\n for await (const speech of this.#speechQueue) {\n if (speech === VoicePipelineAgent.FLUSH_SENTINEL) break;\n this.#playingSpeech = speech;\n await this.#playSpeech(speech);\n this.#playingSpeech = undefined;\n }\n this.#speechQueueOpen = new Future();\n }\n }\n\n #synthesizeAgentReply() {\n this.#pendingAgentReply?.cancel();\n if (this.#humanInput && this.#humanInput.speaking) {\n this.#updateState('thinking', 200);\n }\n\n this.#pendingAgentReply = SpeechHandle.createAssistantReply(\n this.#opts.allowInterruptions,\n true,\n this.transcribedText,\n );\n const newHandle = this.#pendingAgentReply;\n this.#agentReplyTask = this.#synthesizeAnswerTask(this.#agentReplyTask, newHandle);\n }\n\n #synthesizeAnswerTask(\n oldTask: CancellablePromise<void> | undefined,\n handle?: SpeechHandle,\n ): CancellablePromise<void> {\n return new CancellablePromise(async (resolve, _, onCancel) => {\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n\n if (oldTask) {\n await gracefullyCancel(oldTask);\n }\n\n const copiedCtx = this.chatCtx.copy();\n const playingSpeech = this.#playingSpeech;\n if (playingSpeech && playingSpeech.initialized) {\n if (\n (!playingSpeech.userQuestion || playingSpeech.userCommitted) &&\n !playingSpeech.speechCommitted\n ) {\n // the speech is playing but not committed yet,\n // add it to the chat context for this new reply synthesis\n copiedCtx.messages.push(\n ChatMessage.create({\n text: playingSpeech.synthesisHandle.text,\n role: ChatRole.ASSISTANT,\n }),\n );\n }\n }\n\n copiedCtx.messages.push(\n ChatMessage.create({\n text: handle?.userQuestion,\n role: ChatRole.USER,\n }),\n );\n\n speechData = { sequenceId: handle!.id };\n\n try {\n if (cancelled) resolve();\n let llmStream = await this.#opts.beforeLLMCallback(this, copiedCtx);\n if (llmStream === false) {\n handle?.cancel();\n return;\n }\n\n if (cancelled) resolve();\n // fallback to default impl if no custom/user stream is returned\n if (!(llmStream instanceof LLMStream)) {\n llmStream = (await defaultBeforeLLMCallback(this, copiedCtx)) as LLMStream;\n }\n\n if (handle!.interrupted) {\n return;\n }\n\n const synthesisHandle = this.#synthesizeAgentSpeech(handle!.id, llmStream);\n handle!.initialize(llmStream, synthesisHandle);\n } finally {\n lastSpeechData = speechData;\n speechData = undefined;\n }\n resolve();\n });\n }\n\n async #playSpeech(handle: SpeechHandle) {\n try {\n await handle.waitForInitialization();\n } catch {\n return;\n }\n await this.#agentPublication!.waitForSubscription();\n const synthesisHandle = handle.synthesisHandle;\n if (synthesisHandle.interrupted) return;\n\n const userQuestion = handle.userQuestion;\n const playHandle = synthesisHandle.play();\n const joinFut = playHandle.join();\n\n const commitUserQuestionIfNeeded = () => {\n if (!userQuestion || synthesisHandle.interrupted || handle.userCommitted) return;\n const isUsingTools =\n handle.source instanceof LLMStream && !!handle.source.functionCalls.length;\n\n // make sure at least some speech was played before committing the user message\n // since we try to validate as fast as possible it is possible the agent gets interrupted\n // really quickly (barely audible), we don't want to mark this question as \"answered\".\n if (\n handle.allowInterruptions &&\n !isUsingTools &&\n playHandle.timePlayed < this.MIN_TIME_PLAYED_FOR_COMMIT &&\n !joinFut.done\n ) {\n return;\n }\n\n this.#logger.child({ userTranscript: userQuestion }).debug('committed user transcript');\n const userMsg = ChatMessage.create({ text: userQuestion, role: ChatRole.USER });\n this.chatCtx.messages.push(userMsg);\n this.emit(VPAEvent.USER_SPEECH_COMMITTED, userMsg);\n\n this.transcribedText = this.transcribedText.slice(userQuestion.length);\n handle.markUserCommitted();\n };\n\n // wait for the playHandle to finish and check every 1s if user question should be committed\n commitUserQuestionIfNeeded();\n\n while (!joinFut.done) {\n await new Promise<void>(async (resolve) => {\n setTimeout(resolve, 500);\n await joinFut.await;\n resolve();\n });\n commitUserQuestionIfNeeded();\n if (handle.interrupted) break;\n }\n commitUserQuestionIfNeeded();\n\n let collectedText = this.#agentTranscribedText;\n const isUsingTools = handle.source instanceof LLMStream && !!handle.source.functionCalls.length;\n const interrupted = handle.interrupted;\n\n if (handle.addToChatCtx && (!userQuestion || handle.userCommitted)) {\n if (handle.extraToolsMessages) {\n this.chatCtx.messages.push(...handle.extraToolsMessages);\n }\n if (interrupted) {\n collectedText += '…';\n }\n\n const msg = ChatMessage.create({ text: collectedText, role: ChatRole.ASSISTANT });\n this.chatCtx.messages.push(msg);\n\n handle.markSpeechCommitted();\n if (interrupted) {\n this.emit(VPAEvent.AGENT_SPEECH_INTERRUPTED, msg);\n } else {\n this.emit(VPAEvent.AGENT_SPEECH_COMMITTED, msg);\n }\n\n this.#logger\n .child({\n agentTranscript: collectedText,\n interrupted,\n speechId: handle.id,\n })\n .debug('committed agent speech');\n\n handle.setDone();\n }\n\n const executeFunctionCalls = async () => {\n // if the answer is using tools, execute the functions and automatically generate\n // a response to the user question from the returned values\n if (!isUsingTools || interrupted) return;\n\n if (handle.fncNestedDepth >= this.#opts.maxNestedFncCalls) {\n this.#logger\n .child({ speechId: handle.id, fncNestedDepth: handle.fncNestedDepth })\n .warn('max function calls nested depth reached');\n return;\n }\n\n if (userQuestion && !handle.userCommitted) {\n throw new Error('user speech should have been committed before using tools');\n }\n const llmStream = handle.source;\n const newFunctionCalls = llmStream.functionCalls;\n\n new AgentCallContext(this, llmStream);\n\n this.emit(VPAEvent.FUNCTION_CALLS_COLLECTED, newFunctionCalls);\n const calledFuncs: FunctionCallInfo[] = [];\n for (const func of newFunctionCalls) {\n const task = func.func.execute(func.params).then(\n (result) => ({ name: func.name, toolCallId: func.toolCallId, result }),\n (error) => ({ name: func.name, toolCallId: func.toolCallId, error }),\n );\n calledFuncs.push({ ...func, task });\n this.#logger\n .child({ function: func.name, speechId: handle.id })\n .debug('executing AI function');\n try {\n await task;\n } catch {\n this.#logger\n .child({ function: func.name, speechId: handle.id })\n .error('error executing AI function');\n }\n }\n\n const toolCallsInfo = [];\n const toolCallsResults = [];\n for (const fnc of calledFuncs) {\n // ignore the function calls that return void\n const task = await fnc.task;\n if (!task || task.result === undefined) continue;\n toolCallsInfo.push(fnc);\n toolCallsResults.push(ChatMessage.createToolFromFunctionResult(task));\n }\n\n if (!toolCallsInfo.length) return;\n\n // generate an answer from the tool calls\n const extraToolsMessages = [ChatMessage.createToolCalls(toolCallsInfo, collectedText)];\n extraToolsMessages.push(...toolCallsResults);\n\n // create a nested speech handle\n const newSpeechHandle = SpeechHandle.createToolSpeech(\n handle.allowInterruptions,\n handle.addToChatCtx,\n handle.fncNestedDepth + 1,\n extraToolsMessages,\n );\n\n // synthesize the tool speech with the chat ctx from llmStream\n const chatCtx = handle.source.chatCtx.copy();\n chatCtx.messages.push(...extraToolsMessages);\n chatCtx.messages.push(...AgentCallContext.getCurrent().extraChatMessages);\n\n const answerLLMStream = this.llm.chat({\n chatCtx,\n fncCtx: this.fncCtx,\n });\n\n const answerSynthesis = this.#synthesizeAgentSpeech(newSpeechHandle.id, answerLLMStream);\n newSpeechHandle.initialize(answerLLMStream, answerSynthesis);\n handle.addNestedSpeech(newSpeechHandle);\n\n this.emit(VPAEvent.FUNCTION_CALLS_FINISHED, calledFuncs);\n };\n\n let finished = false;\n const task = executeFunctionCalls().then(() => {\n finished = true;\n });\n while (!handle.nestedSpeechFinished) {\n const changed = handle.nestedSpeechChanged();\n await Promise.race([changed, task]);\n while (handle.nestedSpeechHandles.length) {\n const speech = handle.nestedSpeechHandles[0]!;\n this.#playingSpeech = speech;\n await this.#playSpeech(speech);\n handle.nestedSpeechHandles.shift();\n this.#playingSpeech = handle;\n }\n\n handle.nestedSpeechHandles.forEach(() => handle.nestedSpeechHandles.pop());\n if (finished) {\n handle.markNestedSpeechFinished();\n }\n }\n handle.setDone();\n }\n\n async #publishTranscription(\n participantIdentity: string,\n trackSid: string,\n text: string,\n isFinal: boolean,\n id: string,\n ) {\n this.#room!.localParticipant!.publishTranscription({\n participantIdentity: participantIdentity,\n trackSid: trackSid,\n segments: [\n {\n text: text,\n final: isFinal,\n id: id,\n startTime: BigInt(0),\n endTime: BigInt(0),\n language: '',\n },\n ],\n });\n const stream = await this.#room!.localParticipant!.streamText({\n senderIdentity: participantIdentity,\n topic: TOPIC_TRANSCRIPTION,\n attributes: {\n [ATTRIBUTE_TRANSCRIPTION_TRACK_ID]: trackSid,\n [ATTRIBUTE_TRANSCRIPTION_FINAL]: isFinal.toString(),\n },\n });\n await stream.write(text);\n await stream.close();\n }\n\n #synthesizeAgentSpeech(\n speechId: string,\n source: string | LLMStream | AsyncIterable<string>,\n ): SynthesisHandle {\n const synchronizer = new TextAudioSynchronizer(defaultTextSyncOptions);\n // TODO: where possible we would want to use deltas instead of full text segments, esp for LLM streams over the streamText API\n synchronizer.on('textUpdated', async (text) => {\n this.#agentTranscribedText = text.text;\n await this.#publishTranscription(\n this.#room!.localParticipant!.identity!,\n this.#agentPublication?.sid ?? '',\n text.text,\n text.final,\n text.id,\n );\n });\n\n if (!this.#agentOutput) {\n throw new Error('agent output should be initialized when ready');\n }\n\n if (source instanceof LLMStream) {\n source = llmStreamToStringIterable(speechId, source);\n }\n\n const ogSource = source;\n if (!(typeof source === 'string')) {\n // TODO(nbsp): itertools.tee\n }\n\n const ttsSource = this.#opts.beforeTTSCallback(this, ogSource);\n if (!ttsSource) {\n throw new Error('beforeTTSCallback must return string or AsyncIterable<string>');\n }\n\n return this.#agentOutput.synthesize(speechId, ttsSource, synchronizer);\n }\n\n async #validateReplyIfPossible() {\n if (this.#playingSpeech && !this.#playingSpeech.allowInterruptions) {\n this.#logger\n .child({ speechId: this.#playingSpeech.id })\n .debug('skipping validation, agent is speaking and does not allow interruptions');\n return;\n }\n\n if (!this.#pendingAgentReply) {\n if (this.#opts.preemptiveSynthesis || !this.transcribedText) {\n return;\n }\n this.#synthesizeAgentReply();\n }\n\n if (!this.#pendingAgentReply) {\n throw new Error('pending agent reply is undefined');\n }\n\n // in some bad timimg, we could end up with two pushed agent replies inside the speech queue.\n // so make sure we directly interrupt every reply when validating a new one\n if (this.#speechQueueOpen.done) {\n for await (const speech of this.#speechQueue) {\n if (speech === VoicePipelineAgent.FLUSH_SENTINEL) break;\n if (!speech.isReply) continue;\n if (speech.allowInterruptions) speech.interrupt();\n }\n }\n\n this.#logger.child({ speechId: this.#pendingAgentReply.id }).debug('validated agent reply');\n\n if (this.#lastSpeechTime) {\n const timeSinceLastSpeech = Date.now() - this.#lastSpeechTime;\n const transcriptionDelay = Math.max(\n (this.#lastFinalTranscriptTime || 0) - this.#lastSpeechTime,\n 0,\n );\n const metrics: PipelineEOUMetrics = {\n timestamp: Date.now(),\n sequenceId: this.#pendingAgentReply.id,\n endOfUtteranceDelay: timeSinceLastSpeech,\n transcriptionDelay,\n };\n this.emit(VPAEvent.METRICS_COLLECTED, metrics);\n }\n\n this.#addSpeechForPlayout(this.#pendingAgentReply);\n this.#pendingAgentReply = undefined;\n this.#transcribedInterimText = '';\n }\n\n #interruptIfPossible() {\n if (\n !this.#playingSpeech ||\n !this.#playingSpeech.allowInterruptions ||\n this.#playingSpeech.interrupted\n ) {\n return;\n }\n\n if (this.#opts.interruptMinWords !== 0) {\n // check the final/interim transcribed text for the minimum word count\n // to interrupt the agent speech\n const interimWords = this.#opts.transcription.wordTokenizer.tokenize(\n this.#transcribedInterimText,\n );\n if (interimWords.length < this.#opts.interruptMinWords) {\n return;\n }\n }\n this.#playingSpeech.interrupt();\n }\n\n #addSpeechForPlayout(handle: SpeechHandle) {\n this.#speechQueue.put(handle);\n this.#speechQueue.put(VoicePipelineAgent.FLUSH_SENTINEL);\n this.#speechQueueOpen.resolve();\n }\n\n /** Close the voice assistant. */\n async close() {\n if (!this.#started) {\n return;\n }\n\n this.#room?.removeAllListeners(RoomEvent.ParticipantConnected);\n // TODO(nbsp): await this.#deferredValidation.close()\n }\n}\n\nasync function* llmStreamToStringIterable(\n speechId: string,\n stream: LLMStream,\n): AsyncIterable<string> {\n const startTime = Date.now();\n let firstFrame = true;\n for await (const chunk of stream) {\n const content = chunk.choices[0]?.delta.content;\n if (!content) continue;\n\n if (firstFrame) {\n firstFrame = false;\n log()\n .child({ speechId, elapsed: Math.round(Date.now() - startTime) })\n .debug('received first LLM token');\n }\n yield content;\n }\n}\n\n/** This class is used to try to find the best time to validate the agent reply. */\nclass DeferredReplyValidation {\n // if the STT gives us punctuation, we can try to validate the reply faster.\n readonly PUNCTUATION = '.!?';\n readonly PUNCTUATION_REDUCE_FACTOR = 0.75;\n readonly LATE_TRANSCRIPT_TOLERANCE = 1.5; // late compared to end of speech\n readonly UNLIKELY_ENDPOINT_DELAY = 6000;\n\n #validateFunc: () => Promise<void>;\n #validatingPromise?: Promise<void>;\n #validatingFuture = new Future();\n #lastFinalTranscript = '';\n #lastRecvEndOfSpeechTime = 0;\n #speaking = false;\n #endOfSpeechDelay: number;\n #finalTranscriptDelay: number;\n #turnDetector?: TurnDetector;\n #agent: VoicePipelineAgent;\n #abort?: AbortController;\n\n constructor(\n validateFunc: () => Promise<void>,\n minEndpointingDelay: number,\n agent: VoicePipelineAgent,\n turnDetector?: TurnDetector,\n ) {\n this.#validateFunc = validateFunc;\n this.#endOfSpeechDelay = minEndpointingDelay;\n this.#finalTranscriptDelay = minEndpointingDelay;\n this.#agent = agent;\n this.#turnDetector = turnDetector;\n }\n\n get validating(): boolean {\n return !this.#validatingFuture.done;\n }\n\n onHumanFinalTranscript(transcript: string) {\n this.#lastFinalTranscript = transcript.trim();\n if (this.#speaking) return;\n\n const hasRecentEndOfSpeech =\n Date.now() - this.#lastRecvEndOfSpeechTime < this.LATE_TRANSCRIPT_TOLERANCE;\n let delay = hasRecentEndOfSpeech ? this.#endOfSpeechDelay : this.#finalTranscriptDelay;\n delay = this.#endWithPunctuation() ? delay * this.PUNCTUATION_REDUCE_FACTOR : 1;\n\n this.#run(delay);\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n onHumanStartOfSpeech(_: VADEvent) {\n this.#speaking = true;\n if (this.validating) {\n this.#abort?.abort();\n }\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n onHumanEndOfSpeech(_: VADEvent) {\n this.#speaking = false;\n this.#lastRecvEndOfSpeechTime = Date.now();\n\n if (this.#lastFinalTranscript) {\n const delay = this.#endWithPunctuation()\n ? this.#endOfSpeechDelay * this.PUNCTUATION_REDUCE_FACTOR\n : 1_000;\n this.#run(delay);\n }\n }\n\n // TODO(nbsp): aclose\n\n #endWithPunctuation(): boolean {\n return (\n this.#lastFinalTranscript.length > 0 &&\n this.PUNCTUATION.includes(this.#lastFinalTranscript[this.#lastFinalTranscript.length - 1]!)\n );\n }\n\n #resetStates() {\n this.#lastFinalTranscript = '';\n this.#lastRecvEndOfSpeechTime = 0;\n }\n\n #run(delay: number) {\n const runTask = async (delay: number, chatCtx: ChatContext, signal: AbortSignal) => {\n if (this.#lastFinalTranscript && !this.#speaking && this.#turnDetector) {\n const startTime = Date.now();\n const eotProb = await this.#turnDetector.predictEndOfTurn(chatCtx);\n const unlikelyThreshold = this.#turnDetector.unlikelyThreshold;\n const elapsed = Date.now() - startTime;\n if (eotProb < unlikelyThreshold) {\n delay = this.UNLIKELY_ENDPOINT_DELAY;\n }\n delay = Math.max(0, delay - elapsed);\n }\n const timeout = setTimeout(() => {\n this.#resetStates();\n this.#validateFunc();\n }, delay);\n signal.addEventListener('abort', () => {\n clearTimeout(timeout);\n });\n };\n\n this.#abort?.abort();\n this.#abort = new AbortController();\n this.#validatingFuture = new Future();\n const detectCtx = this.#agent.chatCtx.copy();\n detectCtx.append({ text: this.#agent.transcribedText, role: ChatRole.USER });\n this.#validatingPromise = runTask(delay, detectCtx, this.#abort.signal);\n }\n}\n"],"mappings":"AASA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,kBAAkB;AAC3B,OAAO,kBAAkB;AACzB;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAOP,SAAS,aAAa,aAAa,UAAU,UAAU,iBAAiB;AACxE,SAAS,WAAW;AAEpB,SAAmB,iBAAiB,kBAAkB,uBAAuB;AAC7E;AAAA,EACE,qBAAqB;AAAA,EACrB,iBAAiB;AAAA,EACjB;AAAA,OACK;AAEP,SAAS,uBAAuB,8BAA8B;AAE9D,SAAS,UAAU,iBAAiB,wBAAwB;AAC5D,SAAS,oBAAoB,oBAAoB,QAAQ,wBAAwB;AACjF,SAAkC,oBAAoB;AAEtD,SAAS,mBAAmB;AAC5B,SAAS,cAAc,yBAAyB;AAChD,SAAS,YAAY,uBAAuB;AAC5C,SAAS,oBAAoB;AAGtB,MAAM,wBAAwB;AACrC,IAAI;AACJ,IAAI;AAYG,IAAK,WAAL,kBAAKA,cAAL;AACL,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AAVU,SAAAA;AAAA,GAAA;AAgCL,MAAM,iBAAiB;AAAA,EAC5B;AAAA,EACA;AAAA,EACA,YAAY,oBAAI,IAAiB;AAAA,EACjC,qBAAoC,CAAC;AAAA,EACrC,OAAO;AAAA,EAEP,YAAY,OAA2B,WAAsB;AAC3D,SAAK,SAAS;AACd,SAAK,aAAa;AAClB,qBAAiB,WAAW;AAAA,EAC9B;AAAA,EAEA,OAAO,aAA+B;AACpC,WAAO,iBAAiB;AAAA,EAC1B;AAAA,EAEA,IAAI,QAA4B;AAC9B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,KAAa,OAAY;AACrC,SAAK,UAAU,IAAI,KAAK,KAAK;AAAA,EAC/B;AAAA,EAEA,YAAY,KAAa,YAAiB,QAAW;AACnD,WAAO,KAAK,UAAU,IAAI,GAAG,KAAK;AAAA,EACpC;AAAA,EAEA,IAAI,YAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,oBAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,oBAAoB,SAAsB;AACxC,SAAK,mBAAmB,KAAK,OAAO;AAAA,EACtC;AACF;AAEA,MAAM,2BAA8C,CAClD,OACA,YACc;AACd,SAAO,MAAM,IAAI,KAAK,EAAE,SAAS,QAAQ,MAAM,OAAO,CAAC;AACzD;AAEA,MAAM,2BAA8C,CAElD,GACA,SACmC;AACnC,SAAO;AACT;AA6BA,MAAM,mCAA8D;AAAA,EAClE,mBAAmB;AAAA,EACnB,oBAAoB;AAAA,EACpB,0BAA0B;AAAA,EAC1B,mBAAmB,IAAI,uBAAuB;AAAA,EAC9C,eAAe,IAAI,mBAAmB,KAAK;AAAA,EAC3C;AACF;AA2CA,MAAM,oBAAgC;AAAA,EACpC,SAAS,IAAI,YAAY;AAAA,EACzB,oBAAoB;AAAA,EACpB,yBAAyB;AAAA,EACzB,mBAAmB;AAAA,EACnB,qBAAqB;AAAA,EACrB,mBAAmB;AAAA,EACnB,qBAAqB;AAAA,EACrB,mBAAmB;AAAA,EACnB,mBAAmB;AAAA,EACnB,eAAe;AACjB;AAGO,MAAM,2BAA4B,aAAsD;AAAA;AAAA,EAEpF,6BAA6B;AAAA,EACtC,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EAElE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,qBAAqB,IAAI,OAAO;AAAA,EAChC;AAAA,EACA;AAAA,EACA;AAAA,EACA,kBAAkB;AAAA,EAClB,0BAA0B;AAAA,EAC1B,mBAAmB,IAAI,OAAO;AAAA,EAC9B,eAAe,IAAI,mBAA4E;AAAA,EAC/F;AAAA,EACA,WAAW;AAAA,EACX;AAAA,EACA,eAAkD;AAAA,EAClD;AAAA,EACA,UAAU,IAAI;AAAA,EACd;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,wBAAwB;AAAA,EAExB,YAEE,KAEA,KAEA,KAEA,KAEA,OAA4B,mBAC5B;AACA,UAAM;AAEN,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,CAAC,IAAI,aAAa,WAAW;AAC/B,YAAM,IAAI,iBAAiB,KAAK,GAAG;AAAA,IACrC;AAEA,QAAI,CAAC,IAAI,aAAa,WAAW;AAC/B,YAAM,IAAI,iBAAiB,KAAK,IAAI,uBAAuB,CAAC;AAAA,IAC9D;AAEA,SAAK,OAAO;AACZ,SAAK,OAAO;AACZ,SAAK,OAAO;AACZ,SAAK,OAAO;AAEZ,SAAK,sBAAsB,IAAI;AAAA,MAC7B,KAAK,yBAAyB,KAAK,IAAI;AAAA,MACvC,KAAK,MAAM;AAAA,MACX;AAAA,MACA,KAAK,MAAM;AAAA,IACb;AAAA,EACF;AAAA,EAEA,IAAI,SAAsC;AACxC,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,OAAO,KAAsB;AAC/B,SAAK,MAAM,SAAS;AAAA,EACtB;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,MAAW;AACb,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAW;AACb,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAW;AACb,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAW;AACb,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,MAEE,MAQA,cAAiD,MACjD;AACA,QAAI,KAAK,UAAU;AACjB,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AAEA,SAAK,KAAK,GAAG,gBAAgB,mBAAmB,CAAC,YAAY;AAC3D,WAAK,KAAK,2BAA4B,OAAO;AAAA,IAC/C,CAAC;AAED,SAAK,KAAK,GAAG,SAAS,mBAAmB,CAAC,YAAY;AACpD,UAAI,CAAC,WAAY;AACjB,WAAK,KAAK,2BAA4B,EAAE,GAAG,SAAS,YAAY,WAAW,WAAW,CAAC;AAAA,IACzF,CAAC;AAED,SAAK,KAAK,GAAG,SAAS,mBAAmB,CAAC,YAAY;AACpD,YAAM,aAAa,aAAa,WAAW,aAAa,iDAAgB;AACxE,UAAI,CAAC,WAAY;AAEjB,WAAK,KAAK,2BAA4B,EAAE,GAAG,SAAS,WAAW,CAAC;AAAA,IAClE,CAAC;AAED,SAAK,KAAK,GAAG,aAAa,mBAAmB,CAAC,YAAY;AACxD,WAAK,KAAK,2BAA4B,OAAO;AAAA,IAC/C,CAAC;AAED,SAAK,GAAG,UAAU,sBAAsB,CAACC,iBAAmC;AAE1E,UAAI,KAAK,cAAc;AACrB;AAAA,MACF;AACA,WAAK,iBAAiB,KAAK,MAAMA,aAAY,QAAS;AAAA,IACxD,CAAC;AAED,SAAK,QAAQ;AACb,SAAK,eAAe;AAEpB,QAAI,aAAa;AACf,UAAI,OAAO,gBAAgB,UAAU;AACnC,aAAK,iBAAiB,WAAW;AAAA,MACnC,OAAO;AACL,aAAK,iBAAiB,YAAY,QAAS;AAAA,MAC7C;AAAA,IACF;AAEA,SAAK,KAAK;AAAA,EACZ;AAAA;AAAA,EAGA,MAAM,IACJ,QACA,qBAAqB,MACrB,eAAe,MACQ;AACvB,UAAM,KAAK,mBAAmB;AAE9B,QAAI;AACJ,QAAI;AACJ,QAAI,cAAc;AAChB,oBAAc,iBAAiB,WAAW;AAC1C,UAAI,kBAAkB,WAAW;AAC/B,aAAK,QAAQ,KAAK,0DAA0D;AAAA,MAC9E,WAAW,OAAO,WAAW,UAAU;AACrC,oBAAY;AAAA,MACd,OAAO;AACL,oBAAY;AACZ,iBAAS,IAAI,mBAA2B;AAAA,MAC1C;AAAA,IACF;AAEA,UAAM,YAAY,aAAa,sBAAsB,oBAAoB,YAAY;AACrF,UAAM,kBAAkB,KAAK,uBAAuB,UAAU,IAAI,MAAM;AACxE,cAAU,WAAW,QAAQ,eAAe;AAE5C,QAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,sBAAsB;AACpE,WAAK,eAAe,gBAAgB,SAAS;AAAA,IAC/C,OAAO;AACL,WAAK,qBAAqB,SAAS;AAAA,IACrC;AAEA,QAAI,eAAe,WAAW;AAC5B,UAAI;AACJ,UAAI,OAAO,WAAW,UAAU;AAC9B,eAAO;AAAA,MACT,OAAO;AACL,eAAO;AACP,yBAAiB,SAAS,WAAW;AACnC,UAAC,OAAsC,IAAI,KAAK;AAChD,kBAAQ;AAAA,QACV;AACA,QAAC,OAAsC,MAAM;AAAA,MAC/C;AAEA,kBAAY,oBAAoB,YAAY,OAAO,EAAE,MAAM,MAAM,SAAS,UAAU,CAAC,CAAC;AACtF,WAAK,QAAQ,MAAM,EAAE,KAAK,CAAC,EAAE,MAAM,4CAA4C;AAAA,IACjF;AAEA,WAAO;AAAA,EACT;AAAA,EAEA,aAAa,OAAmB,QAAQ,GAAG;AACzC,UAAM,UAAU,CAACC,WAA4C;AAC3D,aAAO,IAAI,mBAAmB,OAAO,SAAS,GAAG,aAAa;AAxcpE;AAycQ,YAAI,YAAY;AAChB,iBAAS,MAAM;AACb,sBAAY;AAAA,QACd,CAAC;AACD,cAAM,IAAI,QAAQ,CAACC,aAAY,WAAWA,UAASD,MAAK,CAAC;AACzD,aAAI,UAAK,UAAL,mBAAY,aAAa;AAC3B,cAAI,CAAC,WAAW;AACd,oBAAM,UAAK,MAAM,qBAAX,mBAA6B,cAAc,EAAE,CAAC,qBAAqB,GAAG,MAAM;AAAA,UACpF;AAAA,QACF;AACA,gBAAQ;AAAA,MACV,CAAC;AAAA,IACH;AAEA,QAAI,KAAK,kBAAkB;AACzB,WAAK,iBAAiB,OAAO;AAAA,IAC/B;AAEA,SAAK,mBAAmB,QAAQ,KAAK;AAAA,EACvC;AAAA,EAEA,iBAAiB,qBAAmC;AAClD,QAAI,CAAC,KAAK,OAAO;AACf,WAAK,QAAQ,MAAM,iBAAiB;AACpC;AAAA,IACF;AAEA,SAAK,eAAe,KAAK,MAAM,mBAAmB,IAAI,mBAAmB,KAAK;AAC9E,QAAI,CAAC,KAAK,cAAc;AACtB,WAAK,QAAQ,MAAM,6BAA6B,mBAAmB,YAAY;AAC/E;AAAA,IACF;AAEA,SAAK,cAAc,IAAI;AAAA,MACrB,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK,MAAM;AAAA,IACb;AACA,SAAK,YAAY,GAAG,gBAAgB,iBAAiB,CAAC,UAAU;AAC9D,WAAK,KAAK,6BAA8B;AACxC,WAAK,oBAAoB,qBAAqB,KAAK;AAAA,IACrD,CAAC;AACD,SAAK,YAAY,GAAG,gBAAgB,oBAAoB,CAAC,UAAU;AACjE,UAAI,CAAC,KAAK,mBAAmB,MAAM;AACjC;AAAA,MACF;AACA,UAAI,CAAC,KAAK,cAAc;AACtB,cAAM,IAAI,MAAM,2BAA2B;AAAA,MAC7C;AAEA,UAAI,KAAK;AACT,UAAI,KAAK,MAAM,oBAAoB;AACjC,aAAK,KAAK,IAAI,GAAG,IAAI,MAAM,WAAW;AACtC,aAAK,aAAa,QAAQ,eAAe;AAAA,MAC3C;AAEA,UAAI,MAAM,kBAAkB,KAAK,MAAM,yBAAyB;AAC9D,aAAK,qBAAqB;AAAA,MAC5B;AAEA,UAAI,MAAM,uBAAuB,GAAG;AAClC,aAAK,kBAAkB,KAAK,IAAI,IAAI,MAAM;AAAA,MAC5C;AAAA,IACF,CAAC;AACD,SAAK,YAAY,GAAG,gBAAgB,eAAe,CAAC,UAAU;AAC5D,WAAK,KAAK,6BAA8B;AACxC,WAAK,oBAAoB,mBAAmB,KAAK;AAAA,IACnD,CAAC;AACD,SAAK,YAAY,GAAG,gBAAgB,oBAAoB,OAAO,UAAU;AACvE,UAAI,CAAC,KAAK,kBAAkB;AAC1B,aAAK,mBAAmB,WAAW;AAAA,MACrC;AACA,WAAK,0BAA0B,MAAM,aAAc,CAAC,EAAE;AAEtD,YAAM,KAAK;AAAA,QACT,KAAK,YAAa,YAAY;AAAA,QAC9B,KAAK,YAAa,gBAAiB;AAAA,QACnC,KAAK;AAAA,QACL;AAAA,QACA,KAAK;AAAA,MACP;AAAA,IACF,CAAC;AACD,SAAK,YAAY,GAAG,gBAAgB,kBAAkB,OAAO,UAAU;AACrE,YAAM,gBAAgB,MAAM,aAAc,CAAC,EAAE;AAC7C,UAAI,CAAC,cAAe;AAEpB,UAAI,CAAC,KAAK,kBAAkB;AAC1B,aAAK,mBAAmB,WAAW;AAAA,MACrC;AAEA,WAAK,2BAA2B,KAAK,IAAI;AACzC,WAAK,oBAAoB,KAAK,kBAAkB,MAAM,MAAM;AAE5D,YAAM,KAAK;AAAA,QACT,KAAK,YAAa,YAAY;AAAA,QAC9B,KAAK,YAAa,gBAAiB;AAAA,QACnC,KAAK;AAAA,QACL;AAAA,QACA,KAAK;AAAA,MACP;AAEA,WAAK,mBAAmB;AAExB,UACE,KAAK,MAAM,wBACV,CAAC,KAAK,kBAAkB,KAAK,eAAe,qBAC7C;AACA,aAAK,sBAAsB;AAAA,MAC7B;AAEA,WAAK,oBAAoB,uBAAuB,aAAa;AAE7D,YAAM,QAAQ,KAAK,MAAM,cAAc,cAAc,SAAS,aAAa;AAC3E,UAAI,MAAM,UAAU,GAAG;AAGrB,aAAK,qBAAqB;AAAA,MAC5B;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,OAAO;AApkBf;AAqkBI,SAAK,aAAa,cAAc;AAChC,UAAM,cAAc,IAAI,YAAY,KAAK,KAAK,YAAY,KAAK,KAAK,WAAW;AAC/E,UAAM,QAAQ,gBAAgB,iBAAiB,mBAAmB,WAAW;AAC7E,SAAK,oBAAoB,QAAM,gBAAK,UAAL,mBAAY,qBAAZ,mBAA8B;AAAA,MAC3D;AAAA,MACA,IAAI,oBAAoB,EAAE,QAAQ,YAAY,kBAAkB,CAAC;AAAA;AAGnE,UAAM,eAAe,IAAI,aAAa,WAAW;AACjD,SAAK,eAAe,IAAI,YAAY,cAAc,KAAK,IAAI;AAE3D,iBAAa,GAAG,kBAAkB,iBAAiB,MAAM;AACvD,WAAK,KAAK,8BAA+B;AACzC,WAAK,aAAa,UAAU;AAAA,IAC9B,CAAC;AAED,iBAAa,GAAG,kBAAkB,iBAAiB,CAAC,MAAM;AACxD,WAAK,KAAK,8BAA+B;AACzC,WAAK,aAAa,WAAW;AAAA,IAC/B,CAAC;AAED,SAAK,mBAAmB,QAAQ;AAEhC,WAAO,MAAM;AACX,YAAM,KAAK,iBAAiB;AAC5B,uBAAiB,UAAU,KAAK,cAAc;AAC5C,YAAI,WAAW,mBAAmB,eAAgB;AAClD,aAAK,iBAAiB;AACtB,cAAM,KAAK,YAAY,MAAM;AAC7B,aAAK,iBAAiB;AAAA,MACxB;AACA,WAAK,mBAAmB,IAAI,OAAO;AAAA,IACrC;AAAA,EACF;AAAA,EAEA,wBAAwB;AAxmB1B;AAymBI,eAAK,uBAAL,mBAAyB;AACzB,QAAI,KAAK,eAAe,KAAK,YAAY,UAAU;AACjD,WAAK,aAAa,YAAY,GAAG;AAAA,IACnC;AAEA,SAAK,qBAAqB,aAAa;AAAA,MACrC,KAAK,MAAM;AAAA,MACX;AAAA,MACA,KAAK;AAAA,IACP;AACA,UAAM,YAAY,KAAK;AACvB,SAAK,kBAAkB,KAAK,sBAAsB,KAAK,iBAAiB,SAAS;AAAA,EACnF;AAAA,EAEA,sBACE,SACA,QAC0B;AAC1B,WAAO,IAAI,mBAAmB,OAAO,SAAS,GAAG,aAAa;AAC5D,UAAI,YAAY;AAChB,eAAS,MAAM;AACb,oBAAY;AAAA,MACd,CAAC;AAED,UAAI,SAAS;AACX,cAAM,iBAAiB,OAAO;AAAA,MAChC;AAEA,YAAM,YAAY,KAAK,QAAQ,KAAK;AACpC,YAAM,gBAAgB,KAAK;AAC3B,UAAI,iBAAiB,cAAc,aAAa;AAC9C,aACG,CAAC,cAAc,gBAAgB,cAAc,kBAC9C,CAAC,cAAc,iBACf;AAGA,oBAAU,SAAS;AAAA,YACjB,YAAY,OAAO;AAAA,cACjB,MAAM,cAAc,gBAAgB;AAAA,cACpC,MAAM,SAAS;AAAA,YACjB,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AAEA,gBAAU,SAAS;AAAA,QACjB,YAAY,OAAO;AAAA,UACjB,MAAM,iCAAQ;AAAA,UACd,MAAM,SAAS;AAAA,QACjB,CAAC;AAAA,MACH;AAEA,mBAAa,EAAE,YAAY,OAAQ,GAAG;AAEtC,UAAI;AACF,YAAI,UAAW,SAAQ;AACvB,YAAI,YAAY,MAAM,KAAK,MAAM,kBAAkB,MAAM,SAAS;AAClE,YAAI,cAAc,OAAO;AACvB,2CAAQ;AACR;AAAA,QACF;AAEA,YAAI,UAAW,SAAQ;AAEvB,YAAI,EAAE,qBAAqB,YAAY;AACrC,sBAAa,MAAM,yBAAyB,MAAM,SAAS;AAAA,QAC7D;AAEA,YAAI,OAAQ,aAAa;AACvB;AAAA,QACF;AAEA,cAAM,kBAAkB,KAAK,uBAAuB,OAAQ,IAAI,SAAS;AACzE,eAAQ,WAAW,WAAW,eAAe;AAAA,MAC/C,UAAE;AACA,yBAAiB;AACjB,qBAAa;AAAA,MACf;AACA,cAAQ;AAAA,IACV,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,YAAY,QAAsB;AACtC,QAAI;AACF,YAAM,OAAO,sBAAsB;AAAA,IACrC,QAAQ;AACN;AAAA,IACF;AACA,UAAM,KAAK,kBAAmB,oBAAoB;AAClD,UAAM,kBAAkB,OAAO;AAC/B,QAAI,gBAAgB,YAAa;AAEjC,UAAM,eAAe,OAAO;AAC5B,UAAM,aAAa,gBAAgB,KAAK;AACxC,UAAM,UAAU,WAAW,KAAK;AAEhC,UAAM,6BAA6B,MAAM;AACvC,UAAI,CAAC,gBAAgB,gBAAgB,eAAe,OAAO,cAAe;AAC1E,YAAME,gBACJ,OAAO,kBAAkB,aAAa,CAAC,CAAC,OAAO,OAAO,cAAc;AAKtE,UACE,OAAO,sBACP,CAACA,iBACD,WAAW,aAAa,KAAK,8BAC7B,CAAC,QAAQ,MACT;AACA;AAAA,MACF;AAEA,WAAK,QAAQ,MAAM,EAAE,gBAAgB,aAAa,CAAC,EAAE,MAAM,2BAA2B;AACtF,YAAM,UAAU,YAAY,OAAO,EAAE,MAAM,cAAc,MAAM,SAAS,KAAK,CAAC;AAC9E,WAAK,QAAQ,SAAS,KAAK,OAAO;AAClC,WAAK,KAAK,+BAAgC,OAAO;AAEjD,WAAK,kBAAkB,KAAK,gBAAgB,MAAM,aAAa,MAAM;AACrE,aAAO,kBAAkB;AAAA,IAC3B;AAGA,+BAA2B;AAE3B,WAAO,CAAC,QAAQ,MAAM;AACpB,YAAM,IAAI,QAAc,OAAO,YAAY;AACzC,mBAAW,SAAS,GAAG;AACvB,cAAM,QAAQ;AACd,gBAAQ;AAAA,MACV,CAAC;AACD,iCAA2B;AAC3B,UAAI,OAAO,YAAa;AAAA,IAC1B;AACA,+BAA2B;AAE3B,QAAI,gBAAgB,KAAK;AACzB,UAAM,eAAe,OAAO,kBAAkB,aAAa,CAAC,CAAC,OAAO,OAAO,cAAc;AACzF,UAAM,cAAc,OAAO;AAE3B,QAAI,OAAO,iBAAiB,CAAC,gBAAgB,OAAO,gBAAgB;AAClE,UAAI,OAAO,oBAAoB;AAC7B,aAAK,QAAQ,SAAS,KAAK,GAAG,OAAO,kBAAkB;AAAA,MACzD;AACA,UAAI,aAAa;AACf,yBAAiB;AAAA,MACnB;AAEA,YAAM,MAAM,YAAY,OAAO,EAAE,MAAM,eAAe,MAAM,SAAS,UAAU,CAAC;AAChF,WAAK,QAAQ,SAAS,KAAK,GAAG;AAE9B,aAAO,oBAAoB;AAC3B,UAAI,aAAa;AACf,aAAK,KAAK,kCAAmC,GAAG;AAAA,MAClD,OAAO;AACL,aAAK,KAAK,gCAAiC,GAAG;AAAA,MAChD;AAEA,WAAK,QACF,MAAM;AAAA,QACL,iBAAiB;AAAA,QACjB;AAAA,QACA,UAAU,OAAO;AAAA,MACnB,CAAC,EACA,MAAM,wBAAwB;AAEjC,aAAO,QAAQ;AAAA,IACjB;AAEA,UAAM,uBAAuB,YAAY;AAGvC,UAAI,CAAC,gBAAgB,YAAa;AAElC,UAAI,OAAO,kBAAkB,KAAK,MAAM,mBAAmB;AACzD,aAAK,QACF,MAAM,EAAE,UAAU,OAAO,IAAI,gBAAgB,OAAO,eAAe,CAAC,EACpE,KAAK,yCAAyC;AACjD;AAAA,MACF;AAEA,UAAI,gBAAgB,CAAC,OAAO,eAAe;AACzC,cAAM,IAAI,MAAM,2DAA2D;AAAA,MAC7E;AACA,YAAM,YAAY,OAAO;AACzB,YAAM,mBAAmB,UAAU;AAEnC,UAAI,iBAAiB,MAAM,SAAS;AAEpC,WAAK,KAAK,kCAAmC,gBAAgB;AAC7D,YAAM,cAAkC,CAAC;AACzC,iBAAW,QAAQ,kBAAkB;AACnC,cAAMC,QAAO,KAAK,KAAK,QAAQ,KAAK,MAAM,EAAE;AAAA,UAC1C,CAAC,YAAY,EAAE,MAAM,KAAK,MAAM,YAAY,KAAK,YAAY,OAAO;AAAA,UACpE,CAAC,WAAW,EAAE,MAAM,KAAK,MAAM,YAAY,KAAK,YAAY,MAAM;AAAA,QACpE;AACA,oBAAY,KAAK,EAAE,GAAG,MAAM,MAAAA,MAAK,CAAC;AAClC,aAAK,QACF,MAAM,EAAE,UAAU,KAAK,MAAM,UAAU,OAAO,GAAG,CAAC,EAClD,MAAM,uBAAuB;AAChC,YAAI;AACF,gBAAMA;AAAA,QACR,QAAQ;AACN,eAAK,QACF,MAAM,EAAE,UAAU,KAAK,MAAM,UAAU,OAAO,GAAG,CAAC,EAClD,MAAM,6BAA6B;AAAA,QACxC;AAAA,MACF;AAEA,YAAM,gBAAgB,CAAC;AACvB,YAAM,mBAAmB,CAAC;AAC1B,iBAAW,OAAO,aAAa;AAE7B,cAAMA,QAAO,MAAM,IAAI;AACvB,YAAI,CAACA,SAAQA,MAAK,WAAW,OAAW;AACxC,sBAAc,KAAK,GAAG;AACtB,yBAAiB,KAAK,YAAY,6BAA6BA,KAAI,CAAC;AAAA,MACtE;AAEA,UAAI,CAAC,cAAc,OAAQ;AAG3B,YAAM,qBAAqB,CAAC,YAAY,gBAAgB,eAAe,aAAa,CAAC;AACrF,yBAAmB,KAAK,GAAG,gBAAgB;AAG3C,YAAM,kBAAkB,aAAa;AAAA,QACnC,OAAO;AAAA,QACP,OAAO;AAAA,QACP,OAAO,iBAAiB;AAAA,QACxB;AAAA,MACF;AAGA,YAAM,UAAU,OAAO,OAAO,QAAQ,KAAK;AAC3C,cAAQ,SAAS,KAAK,GAAG,kBAAkB;AAC3C,cAAQ,SAAS,KAAK,GAAG,iBAAiB,WAAW,EAAE,iBAAiB;AAExE,YAAM,kBAAkB,KAAK,IAAI,KAAK;AAAA,QACpC;AAAA,QACA,QAAQ,KAAK;AAAA,MACf,CAAC;AAED,YAAM,kBAAkB,KAAK,uBAAuB,gBAAgB,IAAI,eAAe;AACvF,sBAAgB,WAAW,iBAAiB,eAAe;AAC3D,aAAO,gBAAgB,eAAe;AAEtC,WAAK,KAAK,iCAAkC,WAAW;AAAA,IACzD;AAEA,QAAI,WAAW;AACf,UAAM,OAAO,qBAAqB,EAAE,KAAK,MAAM;AAC7C,iBAAW;AAAA,IACb,CAAC;AACD,WAAO,CAAC,OAAO,sBAAsB;AACnC,YAAM,UAAU,OAAO,oBAAoB;AAC3C,YAAM,QAAQ,KAAK,CAAC,SAAS,IAAI,CAAC;AAClC,aAAO,OAAO,oBAAoB,QAAQ;AACxC,cAAM,SAAS,OAAO,oBAAoB,CAAC;AAC3C,aAAK,iBAAiB;AACtB,cAAM,KAAK,YAAY,MAAM;AAC7B,eAAO,oBAAoB,MAAM;AACjC,aAAK,iBAAiB;AAAA,MACxB;AAEA,aAAO,oBAAoB,QAAQ,MAAM,OAAO,oBAAoB,IAAI,CAAC;AACzE,UAAI,UAAU;AACZ,eAAO,yBAAyB;AAAA,MAClC;AAAA,IACF;AACA,WAAO,QAAQ;AAAA,EACjB;AAAA,EAEA,MAAM,sBACJ,qBACA,UACA,MACA,SACA,IACA;AACA,SAAK,MAAO,iBAAkB,qBAAqB;AAAA,MACjD;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;AACD,UAAM,SAAS,MAAM,KAAK,MAAO,iBAAkB,WAAW;AAAA,MAC5D,gBAAgB;AAAA,MAChB,OAAO;AAAA,MACP,YAAY;AAAA,QACV,CAAC,gCAAgC,GAAG;AAAA,QACpC,CAAC,6BAA6B,GAAG,QAAQ,SAAS;AAAA,MACpD;AAAA,IACF,CAAC;AACD,UAAM,OAAO,MAAM,IAAI;AACvB,UAAM,OAAO,MAAM;AAAA,EACrB;AAAA,EAEA,uBACE,UACA,QACiB;AACjB,UAAM,eAAe,IAAI,sBAAsB,sBAAsB;AAErE,iBAAa,GAAG,eAAe,OAAO,SAAS;AAl6BnD;AAm6BM,WAAK,wBAAwB,KAAK;AAClC,YAAM,KAAK;AAAA,QACT,KAAK,MAAO,iBAAkB;AAAA,UAC9B,UAAK,sBAAL,mBAAwB,QAAO;AAAA,QAC/B,KAAK;AAAA,QACL,KAAK;AAAA,QACL,KAAK;AAAA,MACP;AAAA,IACF,CAAC;AAED,QAAI,CAAC,KAAK,cAAc;AACtB,YAAM,IAAI,MAAM,+CAA+C;AAAA,IACjE;AAEA,QAAI,kBAAkB,WAAW;AAC/B,eAAS,0BAA0B,UAAU,MAAM;AAAA,IACrD;AAEA,UAAM,WAAW;AACjB,QAAI,EAAE,OAAO,WAAW,WAAW;AAAA,IAEnC;AAEA,UAAM,YAAY,KAAK,MAAM,kBAAkB,MAAM,QAAQ;AAC7D,QAAI,CAAC,WAAW;AACd,YAAM,IAAI,MAAM,+DAA+D;AAAA,IACjF;AAEA,WAAO,KAAK,aAAa,WAAW,UAAU,WAAW,YAAY;AAAA,EACvE;AAAA,EAEA,MAAM,2BAA2B;AAC/B,QAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,oBAAoB;AAClE,WAAK,QACF,MAAM,EAAE,UAAU,KAAK,eAAe,GAAG,CAAC,EAC1C,MAAM,yEAAyE;AAClF;AAAA,IACF;AAEA,QAAI,CAAC,KAAK,oBAAoB;AAC5B,UAAI,KAAK,MAAM,uBAAuB,CAAC,KAAK,iBAAiB;AAC3D;AAAA,MACF;AACA,WAAK,sBAAsB;AAAA,IAC7B;AAEA,QAAI,CAAC,KAAK,oBAAoB;AAC5B,YAAM,IAAI,MAAM,kCAAkC;AAAA,IACpD;AAIA,QAAI,KAAK,iBAAiB,MAAM;AAC9B,uBAAiB,UAAU,KAAK,cAAc;AAC5C,YAAI,WAAW,mBAAmB,eAAgB;AAClD,YAAI,CAAC,OAAO,QAAS;AACrB,YAAI,OAAO,mBAAoB,QAAO,UAAU;AAAA,MAClD;AAAA,IACF;AAEA,SAAK,QAAQ,MAAM,EAAE,UAAU,KAAK,mBAAmB,GAAG,CAAC,EAAE,MAAM,uBAAuB;AAE1F,QAAI,KAAK,iBAAiB;AACxB,YAAM,sBAAsB,KAAK,IAAI,IAAI,KAAK;AAC9C,YAAM,qBAAqB,KAAK;AAAA,SAC7B,KAAK,4BAA4B,KAAK,KAAK;AAAA,QAC5C;AAAA,MACF;AACA,YAAM,UAA8B;AAAA,QAClC,WAAW,KAAK,IAAI;AAAA,QACpB,YAAY,KAAK,mBAAmB;AAAA,QACpC,qBAAqB;AAAA,QACrB;AAAA,MACF;AACA,WAAK,KAAK,2BAA4B,OAAO;AAAA,IAC/C;AAEA,SAAK,qBAAqB,KAAK,kBAAkB;AACjD,SAAK,qBAAqB;AAC1B,SAAK,0BAA0B;AAAA,EACjC;AAAA,EAEA,uBAAuB;AACrB,QACE,CAAC,KAAK,kBACN,CAAC,KAAK,eAAe,sBACrB,KAAK,eAAe,aACpB;AACA;AAAA,IACF;AAEA,QAAI,KAAK,MAAM,sBAAsB,GAAG;AAGtC,YAAM,eAAe,KAAK,MAAM,cAAc,cAAc;AAAA,QAC1D,KAAK;AAAA,MACP;AACA,UAAI,aAAa,SAAS,KAAK,MAAM,mBAAmB;AACtD;AAAA,MACF;AAAA,IACF;AACA,SAAK,eAAe,UAAU;AAAA,EAChC;AAAA,EAEA,qBAAqB,QAAsB;AACzC,SAAK,aAAa,IAAI,MAAM;AAC5B,SAAK,aAAa,IAAI,mBAAmB,cAAc;AACvD,SAAK,iBAAiB,QAAQ;AAAA,EAChC;AAAA;AAAA,EAGA,MAAM,QAAQ;AAlhChB;AAmhCI,QAAI,CAAC,KAAK,UAAU;AAClB;AAAA,IACF;AAEA,eAAK,UAAL,mBAAY,mBAAmB,UAAU;AAAA,EAE3C;AACF;AAEA,gBAAgB,0BACd,UACA,QACuB;AA/hCzB;AAgiCE,QAAM,YAAY,KAAK,IAAI;AAC3B,MAAI,aAAa;AACjB,mBAAiB,SAAS,QAAQ;AAChC,UAAM,WAAU,WAAM,QAAQ,CAAC,MAAf,mBAAkB,MAAM;AACxC,QAAI,CAAC,QAAS;AAEd,QAAI,YAAY;AACd,mBAAa;AACb,UAAI,EACD,MAAM,EAAE,UAAU,SAAS,KAAK,MAAM,KAAK,IAAI,IAAI,SAAS,EAAE,CAAC,EAC/D,MAAM,0BAA0B;AAAA,IACrC;AACA,UAAM;AAAA,EACR;AACF;AAGA,MAAM,wBAAwB;AAAA;AAAA,EAEnB,cAAc;AAAA,EACd,4BAA4B;AAAA,EAC5B,4BAA4B;AAAA;AAAA,EAC5B,0BAA0B;AAAA,EAEnC;AAAA,EACA;AAAA,EACA,oBAAoB,IAAI,OAAO;AAAA,EAC/B,uBAAuB;AAAA,EACvB,2BAA2B;AAAA,EAC3B,YAAY;AAAA,EACZ;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YACE,cACA,qBACA,OACA,cACA;AACA,SAAK,gBAAgB;AACrB,SAAK,oBAAoB;AACzB,SAAK,wBAAwB;AAC7B,SAAK,SAAS;AACd,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,IAAI,aAAsB;AACxB,WAAO,CAAC,KAAK,kBAAkB;AAAA,EACjC;AAAA,EAEA,uBAAuB,YAAoB;AACzC,SAAK,uBAAuB,WAAW,KAAK;AAC5C,QAAI,KAAK,UAAW;AAEpB,UAAM,uBACJ,KAAK,IAAI,IAAI,KAAK,2BAA2B,KAAK;AACpD,QAAI,QAAQ,uBAAuB,KAAK,oBAAoB,KAAK;AACjE,YAAQ,KAAK,oBAAoB,IAAI,QAAQ,KAAK,4BAA4B;AAE9E,SAAK,KAAK,KAAK;AAAA,EACjB;AAAA;AAAA,EAGA,qBAAqB,GAAa;AAlmCpC;AAmmCI,SAAK,YAAY;AACjB,QAAI,KAAK,YAAY;AACnB,iBAAK,WAAL,mBAAa;AAAA,IACf;AAAA,EACF;AAAA;AAAA,EAGA,mBAAmB,GAAa;AAC9B,SAAK,YAAY;AACjB,SAAK,2BAA2B,KAAK,IAAI;AAEzC,QAAI,KAAK,sBAAsB;AAC7B,YAAM,QAAQ,KAAK,oBAAoB,IACnC,KAAK,oBAAoB,KAAK,4BAC9B;AACJ,WAAK,KAAK,KAAK;AAAA,IACjB;AAAA,EACF;AAAA;AAAA,EAIA,sBAA+B;AAC7B,WACE,KAAK,qBAAqB,SAAS,KACnC,KAAK,YAAY,SAAS,KAAK,qBAAqB,KAAK,qBAAqB,SAAS,CAAC,CAAE;AAAA,EAE9F;AAAA,EAEA,eAAe;AACb,SAAK,uBAAuB;AAC5B,SAAK,2BAA2B;AAAA,EAClC;AAAA,EAEA,KAAK,OAAe;AApoCtB;AAqoCI,UAAM,UAAU,OAAOH,QAAe,SAAsB,WAAwB;AAClF,UAAI,KAAK,wBAAwB,CAAC,KAAK,aAAa,KAAK,eAAe;AACtE,cAAM,YAAY,KAAK,IAAI;AAC3B,cAAM,UAAU,MAAM,KAAK,cAAc,iBAAiB,OAAO;AACjE,cAAM,oBAAoB,KAAK,cAAc;AAC7C,cAAM,UAAU,KAAK,IAAI,IAAI;AAC7B,YAAI,UAAU,mBAAmB;AAC/B,UAAAA,SAAQ,KAAK;AAAA,QACf;AACA,QAAAA,SAAQ,KAAK,IAAI,GAAGA,SAAQ,OAAO;AAAA,MACrC;AACA,YAAM,UAAU,WAAW,MAAM;AAC/B,aAAK,aAAa;AAClB,aAAK,cAAc;AAAA,MACrB,GAAGA,MAAK;AACR,aAAO,iBAAiB,SAAS,MAAM;AACrC,qBAAa,OAAO;AAAA,MACtB,CAAC;AAAA,IACH;AAEA,eAAK,WAAL,mBAAa;AACb,SAAK,SAAS,IAAI,gBAAgB;AAClC,SAAK,oBAAoB,IAAI,OAAO;AACpC,UAAM,YAAY,KAAK,OAAO,QAAQ,KAAK;AAC3C,cAAU,OAAO,EAAE,MAAM,KAAK,OAAO,iBAAiB,MAAM,SAAS,KAAK,CAAC;AAC3E,SAAK,qBAAqB,QAAQ,OAAO,WAAW,KAAK,OAAO,MAAM;AAAA,EACxE;AACF;","names":["VPAEvent","participant","delay","resolve","isUsingTools","task"]}
|
|
1
|
+
{"version":3,"sources":["../../src/pipeline/pipeline_agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n LocalTrackPublication,\n NoiseCancellationOptions,\n RemoteParticipant,\n Room,\n} from '@livekit/rtc-node';\nimport {\n AudioSource,\n LocalAudioTrack,\n RoomEvent,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { randomUUID } from 'node:crypto';\nimport EventEmitter from 'node:events';\nimport {\n ATTRIBUTE_SEGMENT_ID,\n ATTRIBUTE_TRANSCRIPTION_FINAL,\n ATTRIBUTE_TRANSCRIPTION_TRACK_ID,\n TOPIC_TRANSCRIPTION,\n} from '../constants.js';\nimport type {\n CallableFunctionResult,\n FunctionCallInfo,\n FunctionContext,\n LLM,\n} from '../llm/index.js';\nimport { ChatContext, ChatMessage, ChatRole, LLMEvent, LLMStream } from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { AgentMetrics, PipelineEOUMetrics } from '../metrics/base.js';\nimport { type STT, StreamAdapter as STTStreamAdapter, SpeechEventType } from '../stt/index.js';\nimport {\n SentenceTokenizer as BasicSentenceTokenizer,\n WordTokenizer as BasicWordTokenizer,\n hyphenateWord,\n} from '../tokenize/basic/index.js';\nimport type { SentenceTokenizer, WordTokenizer } from '../tokenize/tokenizer.js';\nimport { TextAudioSynchronizer, defaultTextSyncOptions } from '../transcription.js';\nimport type { TTS } from '../tts/index.js';\nimport { TTSEvent, StreamAdapter as TTSStreamAdapter } from '../tts/index.js';\nimport { AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';\nimport { type VAD, type VADEvent, VADEventType } from '../vad.js';\nimport type { SpeechSource, SynthesisHandle } from './agent_output.js';\nimport { AgentOutput } from './agent_output.js';\nimport { AgentPlayout, AgentPlayoutEvent } from './agent_playout.js';\nimport { HumanInput, HumanInputEvent } from './human_input.js';\nimport { SpeechHandle } from './speech_handle.js';\n\nexport type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';\nexport const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';\nlet lastSpeechData: { sequenceId: string } | undefined;\nlet speechData: { sequenceId: string } | undefined;\n\nexport type BeforeLLMCallback = (\n agent: VoicePipelineAgent,\n chatCtx: ChatContext,\n) => LLMStream | false | void | Promise<LLMStream | false | void>;\n\nexport type BeforeTTSCallback = (\n agent: VoicePipelineAgent,\n source: string | AsyncIterable<string>,\n) => SpeechSource;\n\nexport enum VPAEvent {\n USER_STARTED_SPEAKING,\n USER_STOPPED_SPEAKING,\n AGENT_STARTED_SPEAKING,\n AGENT_STOPPED_SPEAKING,\n USER_SPEECH_COMMITTED,\n AGENT_SPEECH_COMMITTED,\n AGENT_SPEECH_INTERRUPTED,\n FUNCTION_CALLS_COLLECTED,\n FUNCTION_CALLS_FINISHED,\n METRICS_COLLECTED,\n}\n\nexport type VPACallbacks = {\n [VPAEvent.USER_STARTED_SPEAKING]: () => void;\n [VPAEvent.USER_STOPPED_SPEAKING]: () => void;\n [VPAEvent.AGENT_STARTED_SPEAKING]: () => void;\n [VPAEvent.AGENT_STOPPED_SPEAKING]: () => void;\n [VPAEvent.USER_SPEECH_COMMITTED]: (msg: ChatMessage) => void;\n [VPAEvent.AGENT_SPEECH_COMMITTED]: (msg: ChatMessage) => void;\n [VPAEvent.AGENT_SPEECH_INTERRUPTED]: (msg: ChatMessage) => void;\n [VPAEvent.FUNCTION_CALLS_COLLECTED]: (funcs: FunctionCallInfo[]) => void;\n [VPAEvent.FUNCTION_CALLS_FINISHED]: (funcs: CallableFunctionResult[]) => void;\n [VPAEvent.METRICS_COLLECTED]: (metrics: AgentMetrics) => void;\n};\n\ninterface TurnDetector {\n unlikelyThreshold: number;\n supportsLanguage: (language?: string) => boolean;\n predictEndOfTurn: (chatCtx: ChatContext) => Promise<number>;\n}\n\nexport class AgentCallContext {\n #agent: VoicePipelineAgent;\n #llmStream: LLMStream;\n #metadata = new Map<string, any>();\n #extraChatMessages: ChatMessage[] = [];\n static #current: AgentCallContext;\n\n constructor(agent: VoicePipelineAgent, llmStream: LLMStream) {\n this.#agent = agent;\n this.#llmStream = llmStream;\n AgentCallContext.#current = this;\n }\n\n static getCurrent(): AgentCallContext {\n return AgentCallContext.#current;\n }\n\n get agent(): VoicePipelineAgent {\n return this.#agent;\n }\n\n storeMetadata(key: string, value: any) {\n this.#metadata.set(key, value);\n }\n\n getMetadata(key: string, orDefault: any = undefined) {\n return this.#metadata.get(key) || orDefault;\n }\n\n get llmStream(): LLMStream {\n return this.#llmStream;\n }\n\n get extraChatMessages() {\n return this.#extraChatMessages;\n }\n\n addExtraChatMessage(message: ChatMessage) {\n this.#extraChatMessages.push(message);\n }\n}\n\nconst defaultBeforeLLMCallback: BeforeLLMCallback = (\n agent: VoicePipelineAgent,\n chatCtx: ChatContext,\n): LLMStream => {\n return agent.llm.chat({ chatCtx, fncCtx: agent.fncCtx });\n};\n\nconst defaultBeforeTTSCallback: BeforeTTSCallback = (\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n _: VoicePipelineAgent,\n text: string | AsyncIterable<string>,\n): string | AsyncIterable<string> => {\n return text;\n};\n\nexport interface AgentTranscriptionOptions {\n /** Whether to forward the user transcription to the client */\n userTranscription: boolean;\n /** Whether to forward the agent transcription to the client */\n agentTranscription: boolean;\n /**\n * The speed at which the agent's speech transcription is forwarded to the client.\n * We try to mimic the agent's speech speed by adjusting the transcription speed.\n */\n agentTranscriptionSpeech: number;\n /**\n * The tokenizer used to split the speech into sentences.\n * This is used to decide when to mark a transcript as final for the agent transcription.\n */\n sentenceTokenizer: SentenceTokenizer;\n /**\n * The tokenizer used to split the speech into words.\n * This is used to simulate the \"interim results\" of the agent transcription.\n */\n wordTokenizer: WordTokenizer;\n /**\n * A function that takes a string (word) as input and returns a list of strings,\n * representing the hyphenated parts of the word.\n */\n hyphenateWord: (word: string) => string[];\n}\n\nconst defaultAgentTranscriptionOptions: AgentTranscriptionOptions = {\n userTranscription: true,\n agentTranscription: true,\n agentTranscriptionSpeech: 1,\n sentenceTokenizer: new BasicSentenceTokenizer(),\n wordTokenizer: new BasicWordTokenizer(false),\n hyphenateWord: hyphenateWord,\n};\n\nexport interface VPAOptions {\n /** Chat context for the assistant. */\n chatCtx?: ChatContext;\n /** Function context for the assistant. */\n fncCtx?: FunctionContext;\n /** Whether to allow the user to interrupt the assistant. */\n allowInterruptions: boolean;\n /** Minimum duration of speech to consider for interruption. */\n interruptSpeechDuration: number;\n /** Minimum number of words to consider for interuption. This may increase latency. */\n interruptMinWords: number;\n /** Delay to wait before considering the user speech done. */\n minEndpointingDelay: number;\n maxNestedFncCalls: number;\n /* Whether to preemptively synthesize responses. */\n preemptiveSynthesis: boolean;\n /*\n * Callback called when the assistant is about to synthesize a reply.\n *\n * @remarks\n * Returning void will create a default LLM stream.\n * You can also return your own LLM stream by calling `llm.chat()`.\n * Returning `false` ill cancel the synthesis of the reply.\n */\n beforeLLMCallback: BeforeLLMCallback;\n /*\n * Callback called when the assistant is about to synthesize speech.\n *\n * @remarks\n * This can be used to customize text before synthesis\n * (e.g. editing the pronunciation of a word).\n */\n beforeTTSCallback: BeforeTTSCallback;\n /** Options for assistant transcription. */\n transcription: AgentTranscriptionOptions;\n /** Turn detection model to use. */\n turnDetector?: TurnDetector;\n /** Noise cancellation options. */\n noiseCancellation?: NoiseCancellationOptions;\n}\n\nconst defaultVPAOptions: VPAOptions = {\n chatCtx: new ChatContext(),\n allowInterruptions: true,\n interruptSpeechDuration: 50,\n interruptMinWords: 0,\n minEndpointingDelay: 500,\n maxNestedFncCalls: 1,\n preemptiveSynthesis: false,\n beforeLLMCallback: defaultBeforeLLMCallback,\n beforeTTSCallback: defaultBeforeTTSCallback,\n transcription: defaultAgentTranscriptionOptions,\n};\n\n/** A pipeline agent (VAD + STT + LLM + TTS) implementation. */\nexport class VoicePipelineAgent extends (EventEmitter as new () => TypedEmitter<VPACallbacks>) {\n /** Minimum time played for the user speech to be committed to the chat context. */\n readonly MIN_TIME_PLAYED_FOR_COMMIT = 1.5;\n protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n\n #vad: VAD;\n #stt: STT;\n #llm: LLM;\n #tts: TTS;\n #opts: VPAOptions;\n #humanInput?: HumanInput;\n #agentOutput?: AgentOutput;\n #trackPublishedFut = new Future();\n #pendingAgentReply?: SpeechHandle;\n #agentReplyTask?: CancellablePromise<void>;\n #playingSpeech?: SpeechHandle;\n transcribedText = '';\n #transcribedInterimText = '';\n #speechQueueOpen = new Future();\n #speechQueue = new AsyncIterableQueue<SpeechHandle | typeof VoicePipelineAgent.FLUSH_SENTINEL>();\n #updateStateTask?: CancellablePromise<void>;\n #started = false;\n #room?: Room;\n #participant: RemoteParticipant | string | null = null;\n #deferredValidation: DeferredReplyValidation;\n #logger = log();\n #agentPublication?: LocalTrackPublication;\n #lastFinalTranscriptTime?: number;\n #lastSpeechTime?: number;\n #transcriptionId?: string;\n #agentTranscribedText = '';\n\n constructor(\n /** Voice Activity Detection instance. */\n vad: VAD,\n /** Speech-to-Text instance. */\n stt: STT,\n /** Large Language Model instance. */\n llm: LLM,\n /** Text-to-Speech instance. */\n tts: TTS,\n /** Additional VoicePipelineAgent options. */\n opts: Partial<VPAOptions> = defaultVPAOptions,\n ) {\n super();\n\n this.#opts = { ...defaultVPAOptions, ...opts };\n\n if (!stt.capabilities.streaming) {\n stt = new STTStreamAdapter(stt, vad);\n }\n\n if (!tts.capabilities.streaming) {\n tts = new TTSStreamAdapter(tts, new BasicSentenceTokenizer());\n }\n\n this.#vad = vad;\n this.#stt = stt;\n this.#llm = llm;\n this.#tts = tts;\n\n this.#deferredValidation = new DeferredReplyValidation(\n this.#validateReplyIfPossible.bind(this),\n this.#opts.minEndpointingDelay,\n this,\n this.#opts.turnDetector,\n );\n }\n\n get fncCtx(): FunctionContext | undefined {\n return this.#opts.fncCtx;\n }\n\n set fncCtx(ctx: FunctionContext) {\n this.#opts.fncCtx = ctx;\n }\n\n get chatCtx(): ChatContext {\n return this.#opts.chatCtx!;\n }\n\n get llm(): LLM {\n return this.#llm;\n }\n\n get tts(): TTS {\n return this.#tts;\n }\n\n get stt(): STT {\n return this.#stt;\n }\n\n get vad(): VAD {\n return this.#vad;\n }\n\n /** Start the voice assistant. */\n start(\n /** The room to connect to. */\n room: Room,\n /**\n * The participant to listen to.\n *\n * @remarks\n * Can be a participant or an identity.\n * If omitted, the first participant in the room will be selected.\n */\n participant: RemoteParticipant | string | null = null,\n ) {\n if (this.#started) {\n throw new Error('voice assistant already started');\n }\n\n this.#stt.on(SpeechEventType.METRICS_COLLECTED, (metrics) => {\n this.emit(VPAEvent.METRICS_COLLECTED, metrics);\n });\n\n this.#tts.on(TTSEvent.METRICS_COLLECTED, (metrics) => {\n if (!speechData) return;\n this.emit(VPAEvent.METRICS_COLLECTED, { ...metrics, sequenceId: speechData.sequenceId });\n });\n\n this.#llm.on(LLMEvent.METRICS_COLLECTED, (metrics) => {\n const sequenceId = speechData ? speechData.sequenceId : lastSpeechData?.sequenceId;\n if (!sequenceId) return;\n\n this.emit(VPAEvent.METRICS_COLLECTED, { ...metrics, sequenceId });\n });\n\n this.#vad.on(VADEventType.METRICS_COLLECTED, (metrics) => {\n this.emit(VPAEvent.METRICS_COLLECTED, metrics);\n });\n\n room.on(RoomEvent.ParticipantConnected, (participant: RemoteParticipant) => {\n // automatically link to the first participant that connects, if not already linked\n if (this.#participant) {\n return;\n }\n this.#linkParticipant.call(this, participant.identity!);\n });\n\n this.#room = room;\n this.#participant = participant;\n\n if (participant) {\n if (typeof participant === 'string') {\n this.#linkParticipant(participant);\n } else {\n this.#linkParticipant(participant.identity!);\n }\n }\n\n this.#run();\n }\n\n /** Play a speech source through the voice assistant. */\n async say(\n source: string | LLMStream | AsyncIterable<string>,\n allowInterruptions = true,\n addToChatCtx = true,\n ): Promise<SpeechHandle> {\n await this.#trackPublishedFut.await;\n\n let callContext: AgentCallContext | undefined;\n let fncSource: string | AsyncIterable<string> | undefined;\n if (addToChatCtx) {\n callContext = AgentCallContext.getCurrent();\n if (source instanceof LLMStream) {\n this.#logger.warn('LLMStream will be ignored for function call chat context');\n } else if (typeof source === 'string') {\n fncSource = source;\n } else {\n fncSource = source;\n source = new AsyncIterableQueue<string>();\n }\n }\n\n const newHandle = SpeechHandle.createAssistantSpeech(allowInterruptions, addToChatCtx);\n const synthesisHandle = this.#synthesizeAgentSpeech(newHandle.id, source);\n newHandle.initialize(source, synthesisHandle);\n\n if (this.#playingSpeech && !this.#playingSpeech.nestedSpeechFinished) {\n this.#playingSpeech.addNestedSpeech(newHandle);\n } else {\n this.#addSpeechForPlayout(newHandle);\n }\n\n if (callContext && fncSource) {\n let text: string;\n if (typeof source === 'string') {\n text = fncSource as string;\n } else {\n text = '';\n for await (const chunk of fncSource) {\n (source as AsyncIterableQueue<string>).put(chunk);\n text += chunk;\n }\n (source as AsyncIterableQueue<string>).close();\n }\n\n callContext.addExtraChatMessage(ChatMessage.create({ text, role: ChatRole.ASSISTANT }));\n this.#logger.child({ text }).debug('added speech to function call chat context');\n }\n\n return newHandle;\n }\n\n #updateState(state: AgentState, delay = 0) {\n const runTask = (delay: number): CancellablePromise<void> => {\n return new CancellablePromise(async (resolve, _, onCancel) => {\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n await new Promise((resolve) => setTimeout(resolve, delay));\n if (this.#room?.isConnected) {\n if (!cancelled) {\n await this.#room.localParticipant?.setAttributes({ [AGENT_STATE_ATTRIBUTE]: state });\n }\n }\n resolve();\n });\n };\n\n if (this.#updateStateTask) {\n this.#updateStateTask.cancel();\n }\n\n this.#updateStateTask = runTask(delay);\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.#participant = this.#room.remoteParticipants.get(participantIdentity) || null;\n if (!this.#participant) {\n this.#logger.error(`Participant with identity ${participantIdentity} not found`);\n return;\n }\n\n this.#humanInput = new HumanInput(\n this.#room,\n this.#vad,\n this.#stt,\n this.#participant,\n this.#opts.noiseCancellation,\n );\n this.#humanInput.on(HumanInputEvent.START_OF_SPEECH, (event) => {\n this.emit(VPAEvent.USER_STARTED_SPEAKING);\n this.#deferredValidation.onHumanStartOfSpeech(event);\n });\n this.#humanInput.on(HumanInputEvent.VAD_INFERENCE_DONE, (event) => {\n if (!this.#trackPublishedFut.done) {\n return;\n }\n if (!this.#agentOutput) {\n throw new Error('agent output is undefined');\n }\n\n let tv = 1;\n if (this.#opts.allowInterruptions) {\n tv = Math.max(0, 1 - event.probability);\n this.#agentOutput.playout.targetVolume = tv;\n }\n\n if (event.speechDuration >= this.#opts.interruptSpeechDuration) {\n this.#interruptIfPossible();\n }\n\n if (event.rawAccumulatedSpeech > 0) {\n this.#lastSpeechTime = Date.now() - event.rawAccumulatedSilence;\n }\n });\n this.#humanInput.on(HumanInputEvent.END_OF_SPEECH, (event) => {\n this.emit(VPAEvent.USER_STOPPED_SPEAKING);\n this.#deferredValidation.onHumanEndOfSpeech(event);\n });\n this.#humanInput.on(HumanInputEvent.INTERIM_TRANSCRIPT, async (event) => {\n if (!this.#transcriptionId) {\n this.#transcriptionId = randomUUID();\n }\n this.#transcribedInterimText = event.alternatives![0].text;\n\n await this.#publishTranscription(\n this.#humanInput!.participant.identity,\n this.#humanInput!.subscribedTrack!.sid!,\n this.#transcribedInterimText,\n false,\n this.#transcriptionId,\n this.#transcriptionId,\n );\n });\n this.#humanInput.on(HumanInputEvent.FINAL_TRANSCRIPT, async (event) => {\n const newTranscript = event.alternatives![0].text;\n if (!newTranscript) return;\n\n if (!this.#transcriptionId) {\n this.#transcriptionId = randomUUID();\n }\n\n this.#lastFinalTranscriptTime = Date.now();\n this.transcribedText += (this.transcribedText ? ' ' : '') + newTranscript;\n\n await this.#publishTranscription(\n this.#humanInput!.participant.identity,\n this.#humanInput!.subscribedTrack!.sid!,\n this.transcribedText,\n true,\n this.#transcriptionId,\n this.#transcriptionId,\n );\n\n this.#transcriptionId = undefined;\n\n if (\n this.#opts.preemptiveSynthesis &&\n (!this.#playingSpeech || this.#playingSpeech.allowInterruptions)\n ) {\n this.#synthesizeAgentReply();\n }\n\n this.#deferredValidation.onHumanFinalTranscript(newTranscript);\n\n const words = this.#opts.transcription.wordTokenizer.tokenize(newTranscript);\n if (words.length >= 3) {\n // VAD can sometimes not detect that the human is speaking.\n // to make the interruption more reliable, we also interrupt on the final transcript.\n this.#interruptIfPossible();\n }\n });\n }\n\n async #run() {\n this.#updateState('initializing');\n const audioSource = new AudioSource(this.#tts.sampleRate, this.#tts.numChannels);\n const track = LocalAudioTrack.createAudioTrack('assistant_voice', audioSource);\n this.#agentPublication = await this.#room?.localParticipant?.publishTrack(\n track,\n new TrackPublishOptions({ source: TrackSource.SOURCE_MICROPHONE }),\n );\n\n const agentPlayout = new AgentPlayout(audioSource);\n this.#agentOutput = new AgentOutput(agentPlayout, this.#tts);\n\n agentPlayout.on(AgentPlayoutEvent.PLAYOUT_STARTED, () => {\n this.emit(VPAEvent.AGENT_STARTED_SPEAKING);\n this.#updateState('speaking');\n });\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n agentPlayout.on(AgentPlayoutEvent.PLAYOUT_STOPPED, (_) => {\n this.emit(VPAEvent.AGENT_STOPPED_SPEAKING);\n this.#updateState('listening');\n });\n\n this.#trackPublishedFut.resolve();\n\n while (true) {\n await this.#speechQueueOpen.await;\n for await (const speech of this.#speechQueue) {\n if (speech === VoicePipelineAgent.FLUSH_SENTINEL) break;\n this.#playingSpeech = speech;\n await this.#playSpeech(speech);\n this.#playingSpeech = undefined;\n }\n this.#speechQueueOpen = new Future();\n }\n }\n\n #synthesizeAgentReply() {\n this.#pendingAgentReply?.cancel();\n if (this.#humanInput && this.#humanInput.speaking) {\n this.#updateState('thinking', 200);\n }\n\n this.#pendingAgentReply = SpeechHandle.createAssistantReply(\n this.#opts.allowInterruptions,\n true,\n this.transcribedText,\n );\n const newHandle = this.#pendingAgentReply;\n this.#agentReplyTask = this.#synthesizeAnswerTask(this.#agentReplyTask, newHandle);\n }\n\n #synthesizeAnswerTask(\n oldTask: CancellablePromise<void> | undefined,\n handle?: SpeechHandle,\n ): CancellablePromise<void> {\n return new CancellablePromise(async (resolve, _, onCancel) => {\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n\n if (oldTask) {\n await gracefullyCancel(oldTask);\n }\n\n const copiedCtx = this.chatCtx.copy();\n const playingSpeech = this.#playingSpeech;\n if (playingSpeech && playingSpeech.initialized) {\n if (\n (!playingSpeech.userQuestion || playingSpeech.userCommitted) &&\n !playingSpeech.speechCommitted\n ) {\n // the speech is playing but not committed yet,\n // add it to the chat context for this new reply synthesis\n copiedCtx.messages.push(\n ChatMessage.create({\n text: playingSpeech.synthesisHandle.text,\n role: ChatRole.ASSISTANT,\n }),\n );\n }\n }\n\n copiedCtx.messages.push(\n ChatMessage.create({\n text: handle?.userQuestion,\n role: ChatRole.USER,\n }),\n );\n\n speechData = { sequenceId: handle!.id };\n\n try {\n if (cancelled) resolve();\n let llmStream = await this.#opts.beforeLLMCallback(this, copiedCtx);\n if (llmStream === false) {\n handle?.cancel();\n return;\n }\n\n if (cancelled) resolve();\n // fallback to default impl if no custom/user stream is returned\n if (!(llmStream instanceof LLMStream)) {\n llmStream = (await defaultBeforeLLMCallback(this, copiedCtx)) as LLMStream;\n }\n\n if (handle!.interrupted) {\n return;\n }\n\n const synthesisHandle = this.#synthesizeAgentSpeech(handle!.id, llmStream);\n handle!.initialize(llmStream, synthesisHandle);\n } finally {\n lastSpeechData = speechData;\n speechData = undefined;\n }\n resolve();\n });\n }\n\n async #playSpeech(handle: SpeechHandle) {\n try {\n await handle.waitForInitialization();\n } catch {\n return;\n }\n await this.#agentPublication!.waitForSubscription();\n const synthesisHandle = handle.synthesisHandle;\n if (synthesisHandle.interrupted) return;\n\n const userQuestion = handle.userQuestion;\n const playHandle = synthesisHandle.play();\n const joinFut = playHandle.join();\n\n const commitUserQuestionIfNeeded = () => {\n if (!userQuestion || synthesisHandle.interrupted || handle.userCommitted) return;\n const isUsingTools =\n handle.source instanceof LLMStream && !!handle.source.functionCalls.length;\n\n // make sure at least some speech was played before committing the user message\n // since we try to validate as fast as possible it is possible the agent gets interrupted\n // really quickly (barely audible), we don't want to mark this question as \"answered\".\n if (\n handle.allowInterruptions &&\n !isUsingTools &&\n playHandle.timePlayed < this.MIN_TIME_PLAYED_FOR_COMMIT &&\n !joinFut.done\n ) {\n return;\n }\n\n this.#logger.child({ userTranscript: userQuestion }).debug('committed user transcript');\n const userMsg = ChatMessage.create({ text: userQuestion, role: ChatRole.USER });\n this.chatCtx.messages.push(userMsg);\n this.emit(VPAEvent.USER_SPEECH_COMMITTED, userMsg);\n\n this.transcribedText = this.transcribedText.slice(userQuestion.length);\n handle.markUserCommitted();\n };\n\n // wait for the playHandle to finish and check every 1s if user question should be committed\n commitUserQuestionIfNeeded();\n\n while (!joinFut.done) {\n await new Promise<void>(async (resolve) => {\n setTimeout(resolve, 500);\n await joinFut.await;\n resolve();\n });\n commitUserQuestionIfNeeded();\n if (handle.interrupted) break;\n }\n commitUserQuestionIfNeeded();\n\n let collectedText = this.#agentTranscribedText;\n const isUsingTools = handle.source instanceof LLMStream && !!handle.source.functionCalls.length;\n const interrupted = handle.interrupted;\n\n if (handle.addToChatCtx && (!userQuestion || handle.userCommitted)) {\n if (handle.extraToolsMessages) {\n this.chatCtx.messages.push(...handle.extraToolsMessages);\n }\n if (interrupted) {\n collectedText += '…';\n }\n\n const msg = ChatMessage.create({ text: collectedText, role: ChatRole.ASSISTANT });\n this.chatCtx.messages.push(msg);\n\n handle.markSpeechCommitted();\n if (interrupted) {\n this.emit(VPAEvent.AGENT_SPEECH_INTERRUPTED, msg);\n } else {\n this.emit(VPAEvent.AGENT_SPEECH_COMMITTED, msg);\n }\n\n this.#logger\n .child({\n agentTranscript: collectedText,\n interrupted,\n speechId: handle.id,\n })\n .debug('committed agent speech');\n\n handle.setDone();\n }\n\n const executeFunctionCalls = async () => {\n // if the answer is using tools, execute the functions and automatically generate\n // a response to the user question from the returned values\n if (!isUsingTools || interrupted) return;\n\n if (handle.fncNestedDepth >= this.#opts.maxNestedFncCalls) {\n this.#logger\n .child({ speechId: handle.id, fncNestedDepth: handle.fncNestedDepth })\n .warn('max function calls nested depth reached');\n return;\n }\n\n if (userQuestion && !handle.userCommitted) {\n throw new Error('user speech should have been committed before using tools');\n }\n const llmStream = handle.source;\n const newFunctionCalls = llmStream.functionCalls;\n\n new AgentCallContext(this, llmStream);\n\n this.emit(VPAEvent.FUNCTION_CALLS_COLLECTED, newFunctionCalls);\n const calledFuncs: FunctionCallInfo[] = [];\n for (const func of newFunctionCalls) {\n const task = func.func.execute(func.params).then(\n (result) => ({ name: func.name, toolCallId: func.toolCallId, result }),\n (error) => ({ name: func.name, toolCallId: func.toolCallId, error }),\n );\n calledFuncs.push({ ...func, task });\n this.#logger\n .child({ function: func.name, speechId: handle.id })\n .debug('executing AI function');\n try {\n await task;\n } catch {\n this.#logger\n .child({ function: func.name, speechId: handle.id })\n .error('error executing AI function');\n }\n }\n\n const toolCallsInfo = [];\n const toolCallsResults = [];\n for (const fnc of calledFuncs) {\n // ignore the function calls that return void\n const task = await fnc.task;\n if (!task || task.result === undefined) continue;\n toolCallsInfo.push(fnc);\n toolCallsResults.push(ChatMessage.createToolFromFunctionResult(task));\n }\n\n if (!toolCallsInfo.length) return;\n\n // generate an answer from the tool calls\n const extraToolsMessages = [ChatMessage.createToolCalls(toolCallsInfo, collectedText)];\n extraToolsMessages.push(...toolCallsResults);\n\n // create a nested speech handle\n const newSpeechHandle = SpeechHandle.createToolSpeech(\n handle.allowInterruptions,\n handle.addToChatCtx,\n handle.fncNestedDepth + 1,\n extraToolsMessages,\n );\n\n // synthesize the tool speech with the chat ctx from llmStream\n const chatCtx = handle.source.chatCtx.copy();\n chatCtx.messages.push(...extraToolsMessages);\n chatCtx.messages.push(...AgentCallContext.getCurrent().extraChatMessages);\n\n const answerLLMStream = this.llm.chat({\n chatCtx,\n fncCtx: this.fncCtx,\n });\n\n const answerSynthesis = this.#synthesizeAgentSpeech(newSpeechHandle.id, answerLLMStream);\n newSpeechHandle.initialize(answerLLMStream, answerSynthesis);\n handle.addNestedSpeech(newSpeechHandle);\n\n this.emit(VPAEvent.FUNCTION_CALLS_FINISHED, calledFuncs);\n };\n\n let finished = false;\n const task = executeFunctionCalls().then(() => {\n finished = true;\n });\n while (!handle.nestedSpeechFinished) {\n const changed = handle.nestedSpeechChanged();\n await Promise.race([changed, task]);\n while (handle.nestedSpeechHandles.length) {\n const speech = handle.nestedSpeechHandles[0]!;\n this.#playingSpeech = speech;\n await this.#playSpeech(speech);\n handle.nestedSpeechHandles.shift();\n this.#playingSpeech = handle;\n }\n\n handle.nestedSpeechHandles.forEach(() => handle.nestedSpeechHandles.pop());\n if (finished) {\n handle.markNestedSpeechFinished();\n }\n }\n handle.setDone();\n }\n\n async #publishTranscription(\n participantIdentity: string,\n trackSid: string,\n text: string,\n isFinal: boolean,\n id: string,\n segmentId: string,\n ) {\n this.#room!.localParticipant!.publishTranscription({\n participantIdentity: participantIdentity,\n trackSid: trackSid,\n segments: [\n {\n text: text,\n final: isFinal,\n id: id,\n startTime: BigInt(0),\n endTime: BigInt(0),\n language: '',\n },\n ],\n });\n const stream = await this.#room!.localParticipant!.streamText({\n senderIdentity: participantIdentity,\n topic: TOPIC_TRANSCRIPTION,\n attributes: {\n [ATTRIBUTE_TRANSCRIPTION_TRACK_ID]: trackSid,\n [ATTRIBUTE_TRANSCRIPTION_FINAL]: isFinal.toString(),\n [ATTRIBUTE_SEGMENT_ID]: segmentId,\n },\n });\n await stream.write(text);\n await stream.close();\n }\n\n #synthesizeAgentSpeech(\n speechId: string,\n source: string | LLMStream | AsyncIterable<string>,\n ): SynthesisHandle {\n const synchronizer = new TextAudioSynchronizer(defaultTextSyncOptions);\n // TODO: where possible we would want to use deltas instead of full text segments, esp for LLM streams over the streamText API\n synchronizer.on('textUpdated', async (text) => {\n this.#agentTranscribedText = text.text;\n if (!this.#transcriptionId) {\n this.#transcriptionId = randomUUID();\n }\n await this.#publishTranscription(\n this.#room!.localParticipant!.identity!,\n this.#agentPublication?.sid ?? '',\n text.text,\n text.final,\n text.id,\n this.#transcriptionId,\n );\n if (text.final) {\n this.#transcriptionId = undefined;\n }\n });\n\n if (!this.#agentOutput) {\n throw new Error('agent output should be initialized when ready');\n }\n\n if (source instanceof LLMStream) {\n source = llmStreamToStringIterable(speechId, source);\n }\n\n const ogSource = source;\n if (!(typeof source === 'string')) {\n // TODO(nbsp): itertools.tee\n }\n\n const ttsSource = this.#opts.beforeTTSCallback(this, ogSource);\n if (!ttsSource) {\n throw new Error('beforeTTSCallback must return string or AsyncIterable<string>');\n }\n\n return this.#agentOutput.synthesize(speechId, ttsSource, synchronizer);\n }\n\n async #validateReplyIfPossible() {\n if (this.#playingSpeech && !this.#playingSpeech.allowInterruptions) {\n this.#logger\n .child({ speechId: this.#playingSpeech.id })\n .debug('skipping validation, agent is speaking and does not allow interruptions');\n return;\n }\n\n if (!this.#pendingAgentReply) {\n if (this.#opts.preemptiveSynthesis || !this.transcribedText) {\n return;\n }\n this.#synthesizeAgentReply();\n }\n\n if (!this.#pendingAgentReply) {\n throw new Error('pending agent reply is undefined');\n }\n\n // in some bad timimg, we could end up with two pushed agent replies inside the speech queue.\n // so make sure we directly interrupt every reply when validating a new one\n if (this.#speechQueueOpen.done) {\n for await (const speech of this.#speechQueue) {\n if (speech === VoicePipelineAgent.FLUSH_SENTINEL) break;\n if (!speech.isReply) continue;\n if (speech.allowInterruptions) speech.interrupt();\n }\n }\n\n this.#logger.child({ speechId: this.#pendingAgentReply.id }).debug('validated agent reply');\n\n if (this.#lastSpeechTime) {\n const timeSinceLastSpeech = Date.now() - this.#lastSpeechTime;\n const transcriptionDelay = Math.max(\n (this.#lastFinalTranscriptTime || 0) - this.#lastSpeechTime,\n 0,\n );\n const metrics: PipelineEOUMetrics = {\n timestamp: Date.now(),\n sequenceId: this.#pendingAgentReply.id,\n endOfUtteranceDelay: timeSinceLastSpeech,\n transcriptionDelay,\n };\n this.emit(VPAEvent.METRICS_COLLECTED, metrics);\n }\n\n this.#addSpeechForPlayout(this.#pendingAgentReply);\n this.#pendingAgentReply = undefined;\n this.#transcribedInterimText = '';\n }\n\n #interruptIfPossible() {\n if (\n !this.#playingSpeech ||\n !this.#playingSpeech.allowInterruptions ||\n this.#playingSpeech.interrupted\n ) {\n return;\n }\n\n if (this.#opts.interruptMinWords !== 0) {\n // check the final/interim transcribed text for the minimum word count\n // to interrupt the agent speech\n const interimWords = this.#opts.transcription.wordTokenizer.tokenize(\n this.#transcribedInterimText,\n );\n if (interimWords.length < this.#opts.interruptMinWords) {\n return;\n }\n }\n this.#playingSpeech.interrupt();\n }\n\n #addSpeechForPlayout(handle: SpeechHandle) {\n this.#speechQueue.put(handle);\n this.#speechQueue.put(VoicePipelineAgent.FLUSH_SENTINEL);\n this.#speechQueueOpen.resolve();\n }\n\n /** Close the voice assistant. */\n async close() {\n if (!this.#started) {\n return;\n }\n\n this.#room?.removeAllListeners(RoomEvent.ParticipantConnected);\n // TODO(nbsp): await this.#deferredValidation.close()\n }\n}\n\nasync function* llmStreamToStringIterable(\n speechId: string,\n stream: LLMStream,\n): AsyncIterable<string> {\n const startTime = Date.now();\n let firstFrame = true;\n for await (const chunk of stream) {\n const content = chunk.choices[0]?.delta.content;\n if (!content) continue;\n\n if (firstFrame) {\n firstFrame = false;\n log()\n .child({ speechId, elapsed: Math.round(Date.now() - startTime) })\n .debug('received first LLM token');\n }\n yield content;\n }\n}\n\n/** This class is used to try to find the best time to validate the agent reply. */\nclass DeferredReplyValidation {\n // if the STT gives us punctuation, we can try to validate the reply faster.\n readonly PUNCTUATION = '.!?';\n readonly PUNCTUATION_REDUCE_FACTOR = 0.75;\n readonly LATE_TRANSCRIPT_TOLERANCE = 1.5; // late compared to end of speech\n readonly UNLIKELY_ENDPOINT_DELAY = 6000;\n\n #validateFunc: () => Promise<void>;\n #validatingPromise?: Promise<void>;\n #validatingFuture = new Future();\n #lastFinalTranscript = '';\n #lastRecvEndOfSpeechTime = 0;\n #speaking = false;\n #endOfSpeechDelay: number;\n #finalTranscriptDelay: number;\n #turnDetector?: TurnDetector;\n #agent: VoicePipelineAgent;\n #abort?: AbortController;\n\n constructor(\n validateFunc: () => Promise<void>,\n minEndpointingDelay: number,\n agent: VoicePipelineAgent,\n turnDetector?: TurnDetector,\n ) {\n this.#validateFunc = validateFunc;\n this.#endOfSpeechDelay = minEndpointingDelay;\n this.#finalTranscriptDelay = minEndpointingDelay;\n this.#agent = agent;\n this.#turnDetector = turnDetector;\n }\n\n get validating(): boolean {\n return !this.#validatingFuture.done;\n }\n\n onHumanFinalTranscript(transcript: string) {\n this.#lastFinalTranscript = transcript.trim();\n if (this.#speaking) return;\n\n const hasRecentEndOfSpeech =\n Date.now() - this.#lastRecvEndOfSpeechTime < this.LATE_TRANSCRIPT_TOLERANCE;\n let delay = hasRecentEndOfSpeech ? this.#endOfSpeechDelay : this.#finalTranscriptDelay;\n delay = this.#endWithPunctuation() ? delay * this.PUNCTUATION_REDUCE_FACTOR : 1;\n\n this.#run(delay);\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n onHumanStartOfSpeech(_: VADEvent) {\n this.#speaking = true;\n if (this.validating) {\n this.#abort?.abort();\n }\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n onHumanEndOfSpeech(_: VADEvent) {\n this.#speaking = false;\n this.#lastRecvEndOfSpeechTime = Date.now();\n\n if (this.#lastFinalTranscript) {\n const delay = this.#endWithPunctuation()\n ? this.#endOfSpeechDelay * this.PUNCTUATION_REDUCE_FACTOR\n : 1_000;\n this.#run(delay);\n }\n }\n\n // TODO(nbsp): aclose\n\n #endWithPunctuation(): boolean {\n return (\n this.#lastFinalTranscript.length > 0 &&\n this.PUNCTUATION.includes(this.#lastFinalTranscript[this.#lastFinalTranscript.length - 1]!)\n );\n }\n\n #resetStates() {\n this.#lastFinalTranscript = '';\n this.#lastRecvEndOfSpeechTime = 0;\n }\n\n #run(delay: number) {\n const runTask = async (delay: number, chatCtx: ChatContext, signal: AbortSignal) => {\n if (this.#lastFinalTranscript && !this.#speaking && this.#turnDetector) {\n const startTime = Date.now();\n const eotProb = await this.#turnDetector.predictEndOfTurn(chatCtx);\n const unlikelyThreshold = this.#turnDetector.unlikelyThreshold;\n const elapsed = Date.now() - startTime;\n if (eotProb < unlikelyThreshold) {\n delay = this.UNLIKELY_ENDPOINT_DELAY;\n }\n delay = Math.max(0, delay - elapsed);\n }\n const timeout = setTimeout(() => {\n this.#resetStates();\n this.#validateFunc();\n }, delay);\n signal.addEventListener('abort', () => {\n clearTimeout(timeout);\n });\n };\n\n this.#abort?.abort();\n this.#abort = new AbortController();\n this.#validatingFuture = new Future();\n const detectCtx = this.#agent.chatCtx.copy();\n detectCtx.append({ text: this.#agent.transcribedText, role: ChatRole.USER });\n this.#validatingPromise = runTask(delay, detectCtx, this.#abort.signal);\n }\n}\n"],"mappings":"AASA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,kBAAkB;AAC3B,OAAO,kBAAkB;AACzB;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAOP,SAAS,aAAa,aAAa,UAAU,UAAU,iBAAiB;AACxE,SAAS,WAAW;AAEpB,SAAmB,iBAAiB,kBAAkB,uBAAuB;AAC7E;AAAA,EACE,qBAAqB;AAAA,EACrB,iBAAiB;AAAA,EACjB;AAAA,OACK;AAEP,SAAS,uBAAuB,8BAA8B;AAE9D,SAAS,UAAU,iBAAiB,wBAAwB;AAC5D,SAAS,oBAAoB,oBAAoB,QAAQ,wBAAwB;AACjF,SAAkC,oBAAoB;AAEtD,SAAS,mBAAmB;AAC5B,SAAS,cAAc,yBAAyB;AAChD,SAAS,YAAY,uBAAuB;AAC5C,SAAS,oBAAoB;AAGtB,MAAM,wBAAwB;AACrC,IAAI;AACJ,IAAI;AAYG,IAAK,WAAL,kBAAKA,cAAL;AACL,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AACA,EAAAA,oBAAA;AAVU,SAAAA;AAAA,GAAA;AAgCL,MAAM,iBAAiB;AAAA,EAC5B;AAAA,EACA;AAAA,EACA,YAAY,oBAAI,IAAiB;AAAA,EACjC,qBAAoC,CAAC;AAAA,EACrC,OAAO;AAAA,EAEP,YAAY,OAA2B,WAAsB;AAC3D,SAAK,SAAS;AACd,SAAK,aAAa;AAClB,qBAAiB,WAAW;AAAA,EAC9B;AAAA,EAEA,OAAO,aAA+B;AACpC,WAAO,iBAAiB;AAAA,EAC1B;AAAA,EAEA,IAAI,QAA4B;AAC9B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,KAAa,OAAY;AACrC,SAAK,UAAU,IAAI,KAAK,KAAK;AAAA,EAC/B;AAAA,EAEA,YAAY,KAAa,YAAiB,QAAW;AACnD,WAAO,KAAK,UAAU,IAAI,GAAG,KAAK;AAAA,EACpC;AAAA,EAEA,IAAI,YAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,oBAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,oBAAoB,SAAsB;AACxC,SAAK,mBAAmB,KAAK,OAAO;AAAA,EACtC;AACF;AAEA,MAAM,2BAA8C,CAClD,OACA,YACc;AACd,SAAO,MAAM,IAAI,KAAK,EAAE,SAAS,QAAQ,MAAM,OAAO,CAAC;AACzD;AAEA,MAAM,2BAA8C,CAElD,GACA,SACmC;AACnC,SAAO;AACT;AA6BA,MAAM,mCAA8D;AAAA,EAClE,mBAAmB;AAAA,EACnB,oBAAoB;AAAA,EACpB,0BAA0B;AAAA,EAC1B,mBAAmB,IAAI,uBAAuB;AAAA,EAC9C,eAAe,IAAI,mBAAmB,KAAK;AAAA,EAC3C;AACF;AA2CA,MAAM,oBAAgC;AAAA,EACpC,SAAS,IAAI,YAAY;AAAA,EACzB,oBAAoB;AAAA,EACpB,yBAAyB;AAAA,EACzB,mBAAmB;AAAA,EACnB,qBAAqB;AAAA,EACrB,mBAAmB;AAAA,EACnB,qBAAqB;AAAA,EACrB,mBAAmB;AAAA,EACnB,mBAAmB;AAAA,EACnB,eAAe;AACjB;AAGO,MAAM,2BAA4B,aAAsD;AAAA;AAAA,EAEpF,6BAA6B;AAAA,EACtC,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EAElE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,qBAAqB,IAAI,OAAO;AAAA,EAChC;AAAA,EACA;AAAA,EACA;AAAA,EACA,kBAAkB;AAAA,EAClB,0BAA0B;AAAA,EAC1B,mBAAmB,IAAI,OAAO;AAAA,EAC9B,eAAe,IAAI,mBAA4E;AAAA,EAC/F;AAAA,EACA,WAAW;AAAA,EACX;AAAA,EACA,eAAkD;AAAA,EAClD;AAAA,EACA,UAAU,IAAI;AAAA,EACd;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,wBAAwB;AAAA,EAExB,YAEE,KAEA,KAEA,KAEA,KAEA,OAA4B,mBAC5B;AACA,UAAM;AAEN,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,CAAC,IAAI,aAAa,WAAW;AAC/B,YAAM,IAAI,iBAAiB,KAAK,GAAG;AAAA,IACrC;AAEA,QAAI,CAAC,IAAI,aAAa,WAAW;AAC/B,YAAM,IAAI,iBAAiB,KAAK,IAAI,uBAAuB,CAAC;AAAA,IAC9D;AAEA,SAAK,OAAO;AACZ,SAAK,OAAO;AACZ,SAAK,OAAO;AACZ,SAAK,OAAO;AAEZ,SAAK,sBAAsB,IAAI;AAAA,MAC7B,KAAK,yBAAyB,KAAK,IAAI;AAAA,MACvC,KAAK,MAAM;AAAA,MACX;AAAA,MACA,KAAK,MAAM;AAAA,IACb;AAAA,EACF;AAAA,EAEA,IAAI,SAAsC;AACxC,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,OAAO,KAAsB;AAC/B,SAAK,MAAM,SAAS;AAAA,EACtB;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,MAAW;AACb,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAW;AACb,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAW;AACb,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAW;AACb,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,MAEE,MAQA,cAAiD,MACjD;AACA,QAAI,KAAK,UAAU;AACjB,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AAEA,SAAK,KAAK,GAAG,gBAAgB,mBAAmB,CAAC,YAAY;AAC3D,WAAK,KAAK,2BAA4B,OAAO;AAAA,IAC/C,CAAC;AAED,SAAK,KAAK,GAAG,SAAS,mBAAmB,CAAC,YAAY;AACpD,UAAI,CAAC,WAAY;AACjB,WAAK,KAAK,2BAA4B,EAAE,GAAG,SAAS,YAAY,WAAW,WAAW,CAAC;AAAA,IACzF,CAAC;AAED,SAAK,KAAK,GAAG,SAAS,mBAAmB,CAAC,YAAY;AACpD,YAAM,aAAa,aAAa,WAAW,aAAa,iDAAgB;AACxE,UAAI,CAAC,WAAY;AAEjB,WAAK,KAAK,2BAA4B,EAAE,GAAG,SAAS,WAAW,CAAC;AAAA,IAClE,CAAC;AAED,SAAK,KAAK,GAAG,aAAa,mBAAmB,CAAC,YAAY;AACxD,WAAK,KAAK,2BAA4B,OAAO;AAAA,IAC/C,CAAC;AAED,SAAK,GAAG,UAAU,sBAAsB,CAACC,iBAAmC;AAE1E,UAAI,KAAK,cAAc;AACrB;AAAA,MACF;AACA,WAAK,iBAAiB,KAAK,MAAMA,aAAY,QAAS;AAAA,IACxD,CAAC;AAED,SAAK,QAAQ;AACb,SAAK,eAAe;AAEpB,QAAI,aAAa;AACf,UAAI,OAAO,gBAAgB,UAAU;AACnC,aAAK,iBAAiB,WAAW;AAAA,MACnC,OAAO;AACL,aAAK,iBAAiB,YAAY,QAAS;AAAA,MAC7C;AAAA,IACF;AAEA,SAAK,KAAK;AAAA,EACZ;AAAA;AAAA,EAGA,MAAM,IACJ,QACA,qBAAqB,MACrB,eAAe,MACQ;AACvB,UAAM,KAAK,mBAAmB;AAE9B,QAAI;AACJ,QAAI;AACJ,QAAI,cAAc;AAChB,oBAAc,iBAAiB,WAAW;AAC1C,UAAI,kBAAkB,WAAW;AAC/B,aAAK,QAAQ,KAAK,0DAA0D;AAAA,MAC9E,WAAW,OAAO,WAAW,UAAU;AACrC,oBAAY;AAAA,MACd,OAAO;AACL,oBAAY;AACZ,iBAAS,IAAI,mBAA2B;AAAA,MAC1C;AAAA,IACF;AAEA,UAAM,YAAY,aAAa,sBAAsB,oBAAoB,YAAY;AACrF,UAAM,kBAAkB,KAAK,uBAAuB,UAAU,IAAI,MAAM;AACxE,cAAU,WAAW,QAAQ,eAAe;AAE5C,QAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,sBAAsB;AACpE,WAAK,eAAe,gBAAgB,SAAS;AAAA,IAC/C,OAAO;AACL,WAAK,qBAAqB,SAAS;AAAA,IACrC;AAEA,QAAI,eAAe,WAAW;AAC5B,UAAI;AACJ,UAAI,OAAO,WAAW,UAAU;AAC9B,eAAO;AAAA,MACT,OAAO;AACL,eAAO;AACP,yBAAiB,SAAS,WAAW;AACnC,UAAC,OAAsC,IAAI,KAAK;AAChD,kBAAQ;AAAA,QACV;AACA,QAAC,OAAsC,MAAM;AAAA,MAC/C;AAEA,kBAAY,oBAAoB,YAAY,OAAO,EAAE,MAAM,MAAM,SAAS,UAAU,CAAC,CAAC;AACtF,WAAK,QAAQ,MAAM,EAAE,KAAK,CAAC,EAAE,MAAM,4CAA4C;AAAA,IACjF;AAEA,WAAO;AAAA,EACT;AAAA,EAEA,aAAa,OAAmB,QAAQ,GAAG;AACzC,UAAM,UAAU,CAACC,WAA4C;AAC3D,aAAO,IAAI,mBAAmB,OAAO,SAAS,GAAG,aAAa;AAzcpE;AA0cQ,YAAI,YAAY;AAChB,iBAAS,MAAM;AACb,sBAAY;AAAA,QACd,CAAC;AACD,cAAM,IAAI,QAAQ,CAACC,aAAY,WAAWA,UAASD,MAAK,CAAC;AACzD,aAAI,UAAK,UAAL,mBAAY,aAAa;AAC3B,cAAI,CAAC,WAAW;AACd,oBAAM,UAAK,MAAM,qBAAX,mBAA6B,cAAc,EAAE,CAAC,qBAAqB,GAAG,MAAM;AAAA,UACpF;AAAA,QACF;AACA,gBAAQ;AAAA,MACV,CAAC;AAAA,IACH;AAEA,QAAI,KAAK,kBAAkB;AACzB,WAAK,iBAAiB,OAAO;AAAA,IAC/B;AAEA,SAAK,mBAAmB,QAAQ,KAAK;AAAA,EACvC;AAAA,EAEA,iBAAiB,qBAAmC;AAClD,QAAI,CAAC,KAAK,OAAO;AACf,WAAK,QAAQ,MAAM,iBAAiB;AACpC;AAAA,IACF;AAEA,SAAK,eAAe,KAAK,MAAM,mBAAmB,IAAI,mBAAmB,KAAK;AAC9E,QAAI,CAAC,KAAK,cAAc;AACtB,WAAK,QAAQ,MAAM,6BAA6B,mBAAmB,YAAY;AAC/E;AAAA,IACF;AAEA,SAAK,cAAc,IAAI;AAAA,MACrB,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK,MAAM;AAAA,IACb;AACA,SAAK,YAAY,GAAG,gBAAgB,iBAAiB,CAAC,UAAU;AAC9D,WAAK,KAAK,6BAA8B;AACxC,WAAK,oBAAoB,qBAAqB,KAAK;AAAA,IACrD,CAAC;AACD,SAAK,YAAY,GAAG,gBAAgB,oBAAoB,CAAC,UAAU;AACjE,UAAI,CAAC,KAAK,mBAAmB,MAAM;AACjC;AAAA,MACF;AACA,UAAI,CAAC,KAAK,cAAc;AACtB,cAAM,IAAI,MAAM,2BAA2B;AAAA,MAC7C;AAEA,UAAI,KAAK;AACT,UAAI,KAAK,MAAM,oBAAoB;AACjC,aAAK,KAAK,IAAI,GAAG,IAAI,MAAM,WAAW;AACtC,aAAK,aAAa,QAAQ,eAAe;AAAA,MAC3C;AAEA,UAAI,MAAM,kBAAkB,KAAK,MAAM,yBAAyB;AAC9D,aAAK,qBAAqB;AAAA,MAC5B;AAEA,UAAI,MAAM,uBAAuB,GAAG;AAClC,aAAK,kBAAkB,KAAK,IAAI,IAAI,MAAM;AAAA,MAC5C;AAAA,IACF,CAAC;AACD,SAAK,YAAY,GAAG,gBAAgB,eAAe,CAAC,UAAU;AAC5D,WAAK,KAAK,6BAA8B;AACxC,WAAK,oBAAoB,mBAAmB,KAAK;AAAA,IACnD,CAAC;AACD,SAAK,YAAY,GAAG,gBAAgB,oBAAoB,OAAO,UAAU;AACvE,UAAI,CAAC,KAAK,kBAAkB;AAC1B,aAAK,mBAAmB,WAAW;AAAA,MACrC;AACA,WAAK,0BAA0B,MAAM,aAAc,CAAC,EAAE;AAEtD,YAAM,KAAK;AAAA,QACT,KAAK,YAAa,YAAY;AAAA,QAC9B,KAAK,YAAa,gBAAiB;AAAA,QACnC,KAAK;AAAA,QACL;AAAA,QACA,KAAK;AAAA,QACL,KAAK;AAAA,MACP;AAAA,IACF,CAAC;AACD,SAAK,YAAY,GAAG,gBAAgB,kBAAkB,OAAO,UAAU;AACrE,YAAM,gBAAgB,MAAM,aAAc,CAAC,EAAE;AAC7C,UAAI,CAAC,cAAe;AAEpB,UAAI,CAAC,KAAK,kBAAkB;AAC1B,aAAK,mBAAmB,WAAW;AAAA,MACrC;AAEA,WAAK,2BAA2B,KAAK,IAAI;AACzC,WAAK,oBAAoB,KAAK,kBAAkB,MAAM,MAAM;AAE5D,YAAM,KAAK;AAAA,QACT,KAAK,YAAa,YAAY;AAAA,QAC9B,KAAK,YAAa,gBAAiB;AAAA,QACnC,KAAK;AAAA,QACL;AAAA,QACA,KAAK;AAAA,QACL,KAAK;AAAA,MACP;AAEA,WAAK,mBAAmB;AAExB,UACE,KAAK,MAAM,wBACV,CAAC,KAAK,kBAAkB,KAAK,eAAe,qBAC7C;AACA,aAAK,sBAAsB;AAAA,MAC7B;AAEA,WAAK,oBAAoB,uBAAuB,aAAa;AAE7D,YAAM,QAAQ,KAAK,MAAM,cAAc,cAAc,SAAS,aAAa;AAC3E,UAAI,MAAM,UAAU,GAAG;AAGrB,aAAK,qBAAqB;AAAA,MAC5B;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,OAAO;AAvkBf;AAwkBI,SAAK,aAAa,cAAc;AAChC,UAAM,cAAc,IAAI,YAAY,KAAK,KAAK,YAAY,KAAK,KAAK,WAAW;AAC/E,UAAM,QAAQ,gBAAgB,iBAAiB,mBAAmB,WAAW;AAC7E,SAAK,oBAAoB,QAAM,gBAAK,UAAL,mBAAY,qBAAZ,mBAA8B;AAAA,MAC3D;AAAA,MACA,IAAI,oBAAoB,EAAE,QAAQ,YAAY,kBAAkB,CAAC;AAAA;AAGnE,UAAM,eAAe,IAAI,aAAa,WAAW;AACjD,SAAK,eAAe,IAAI,YAAY,cAAc,KAAK,IAAI;AAE3D,iBAAa,GAAG,kBAAkB,iBAAiB,MAAM;AACvD,WAAK,KAAK,8BAA+B;AACzC,WAAK,aAAa,UAAU;AAAA,IAC9B,CAAC;AAED,iBAAa,GAAG,kBAAkB,iBAAiB,CAAC,MAAM;AACxD,WAAK,KAAK,8BAA+B;AACzC,WAAK,aAAa,WAAW;AAAA,IAC/B,CAAC;AAED,SAAK,mBAAmB,QAAQ;AAEhC,WAAO,MAAM;AACX,YAAM,KAAK,iBAAiB;AAC5B,uBAAiB,UAAU,KAAK,cAAc;AAC5C,YAAI,WAAW,mBAAmB,eAAgB;AAClD,aAAK,iBAAiB;AACtB,cAAM,KAAK,YAAY,MAAM;AAC7B,aAAK,iBAAiB;AAAA,MACxB;AACA,WAAK,mBAAmB,IAAI,OAAO;AAAA,IACrC;AAAA,EACF;AAAA,EAEA,wBAAwB;AA3mB1B;AA4mBI,eAAK,uBAAL,mBAAyB;AACzB,QAAI,KAAK,eAAe,KAAK,YAAY,UAAU;AACjD,WAAK,aAAa,YAAY,GAAG;AAAA,IACnC;AAEA,SAAK,qBAAqB,aAAa;AAAA,MACrC,KAAK,MAAM;AAAA,MACX;AAAA,MACA,KAAK;AAAA,IACP;AACA,UAAM,YAAY,KAAK;AACvB,SAAK,kBAAkB,KAAK,sBAAsB,KAAK,iBAAiB,SAAS;AAAA,EACnF;AAAA,EAEA,sBACE,SACA,QAC0B;AAC1B,WAAO,IAAI,mBAAmB,OAAO,SAAS,GAAG,aAAa;AAC5D,UAAI,YAAY;AAChB,eAAS,MAAM;AACb,oBAAY;AAAA,MACd,CAAC;AAED,UAAI,SAAS;AACX,cAAM,iBAAiB,OAAO;AAAA,MAChC;AAEA,YAAM,YAAY,KAAK,QAAQ,KAAK;AACpC,YAAM,gBAAgB,KAAK;AAC3B,UAAI,iBAAiB,cAAc,aAAa;AAC9C,aACG,CAAC,cAAc,gBAAgB,cAAc,kBAC9C,CAAC,cAAc,iBACf;AAGA,oBAAU,SAAS;AAAA,YACjB,YAAY,OAAO;AAAA,cACjB,MAAM,cAAc,gBAAgB;AAAA,cACpC,MAAM,SAAS;AAAA,YACjB,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AAEA,gBAAU,SAAS;AAAA,QACjB,YAAY,OAAO;AAAA,UACjB,MAAM,iCAAQ;AAAA,UACd,MAAM,SAAS;AAAA,QACjB,CAAC;AAAA,MACH;AAEA,mBAAa,EAAE,YAAY,OAAQ,GAAG;AAEtC,UAAI;AACF,YAAI,UAAW,SAAQ;AACvB,YAAI,YAAY,MAAM,KAAK,MAAM,kBAAkB,MAAM,SAAS;AAClE,YAAI,cAAc,OAAO;AACvB,2CAAQ;AACR;AAAA,QACF;AAEA,YAAI,UAAW,SAAQ;AAEvB,YAAI,EAAE,qBAAqB,YAAY;AACrC,sBAAa,MAAM,yBAAyB,MAAM,SAAS;AAAA,QAC7D;AAEA,YAAI,OAAQ,aAAa;AACvB;AAAA,QACF;AAEA,cAAM,kBAAkB,KAAK,uBAAuB,OAAQ,IAAI,SAAS;AACzE,eAAQ,WAAW,WAAW,eAAe;AAAA,MAC/C,UAAE;AACA,yBAAiB;AACjB,qBAAa;AAAA,MACf;AACA,cAAQ;AAAA,IACV,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,YAAY,QAAsB;AACtC,QAAI;AACF,YAAM,OAAO,sBAAsB;AAAA,IACrC,QAAQ;AACN;AAAA,IACF;AACA,UAAM,KAAK,kBAAmB,oBAAoB;AAClD,UAAM,kBAAkB,OAAO;AAC/B,QAAI,gBAAgB,YAAa;AAEjC,UAAM,eAAe,OAAO;AAC5B,UAAM,aAAa,gBAAgB,KAAK;AACxC,UAAM,UAAU,WAAW,KAAK;AAEhC,UAAM,6BAA6B,MAAM;AACvC,UAAI,CAAC,gBAAgB,gBAAgB,eAAe,OAAO,cAAe;AAC1E,YAAME,gBACJ,OAAO,kBAAkB,aAAa,CAAC,CAAC,OAAO,OAAO,cAAc;AAKtE,UACE,OAAO,sBACP,CAACA,iBACD,WAAW,aAAa,KAAK,8BAC7B,CAAC,QAAQ,MACT;AACA;AAAA,MACF;AAEA,WAAK,QAAQ,MAAM,EAAE,gBAAgB,aAAa,CAAC,EAAE,MAAM,2BAA2B;AACtF,YAAM,UAAU,YAAY,OAAO,EAAE,MAAM,cAAc,MAAM,SAAS,KAAK,CAAC;AAC9E,WAAK,QAAQ,SAAS,KAAK,OAAO;AAClC,WAAK,KAAK,+BAAgC,OAAO;AAEjD,WAAK,kBAAkB,KAAK,gBAAgB,MAAM,aAAa,MAAM;AACrE,aAAO,kBAAkB;AAAA,IAC3B;AAGA,+BAA2B;AAE3B,WAAO,CAAC,QAAQ,MAAM;AACpB,YAAM,IAAI,QAAc,OAAO,YAAY;AACzC,mBAAW,SAAS,GAAG;AACvB,cAAM,QAAQ;AACd,gBAAQ;AAAA,MACV,CAAC;AACD,iCAA2B;AAC3B,UAAI,OAAO,YAAa;AAAA,IAC1B;AACA,+BAA2B;AAE3B,QAAI,gBAAgB,KAAK;AACzB,UAAM,eAAe,OAAO,kBAAkB,aAAa,CAAC,CAAC,OAAO,OAAO,cAAc;AACzF,UAAM,cAAc,OAAO;AAE3B,QAAI,OAAO,iBAAiB,CAAC,gBAAgB,OAAO,gBAAgB;AAClE,UAAI,OAAO,oBAAoB;AAC7B,aAAK,QAAQ,SAAS,KAAK,GAAG,OAAO,kBAAkB;AAAA,MACzD;AACA,UAAI,aAAa;AACf,yBAAiB;AAAA,MACnB;AAEA,YAAM,MAAM,YAAY,OAAO,EAAE,MAAM,eAAe,MAAM,SAAS,UAAU,CAAC;AAChF,WAAK,QAAQ,SAAS,KAAK,GAAG;AAE9B,aAAO,oBAAoB;AAC3B,UAAI,aAAa;AACf,aAAK,KAAK,kCAAmC,GAAG;AAAA,MAClD,OAAO;AACL,aAAK,KAAK,gCAAiC,GAAG;AAAA,MAChD;AAEA,WAAK,QACF,MAAM;AAAA,QACL,iBAAiB;AAAA,QACjB;AAAA,QACA,UAAU,OAAO;AAAA,MACnB,CAAC,EACA,MAAM,wBAAwB;AAEjC,aAAO,QAAQ;AAAA,IACjB;AAEA,UAAM,uBAAuB,YAAY;AAGvC,UAAI,CAAC,gBAAgB,YAAa;AAElC,UAAI,OAAO,kBAAkB,KAAK,MAAM,mBAAmB;AACzD,aAAK,QACF,MAAM,EAAE,UAAU,OAAO,IAAI,gBAAgB,OAAO,eAAe,CAAC,EACpE,KAAK,yCAAyC;AACjD;AAAA,MACF;AAEA,UAAI,gBAAgB,CAAC,OAAO,eAAe;AACzC,cAAM,IAAI,MAAM,2DAA2D;AAAA,MAC7E;AACA,YAAM,YAAY,OAAO;AACzB,YAAM,mBAAmB,UAAU;AAEnC,UAAI,iBAAiB,MAAM,SAAS;AAEpC,WAAK,KAAK,kCAAmC,gBAAgB;AAC7D,YAAM,cAAkC,CAAC;AACzC,iBAAW,QAAQ,kBAAkB;AACnC,cAAMC,QAAO,KAAK,KAAK,QAAQ,KAAK,MAAM,EAAE;AAAA,UAC1C,CAAC,YAAY,EAAE,MAAM,KAAK,MAAM,YAAY,KAAK,YAAY,OAAO;AAAA,UACpE,CAAC,WAAW,EAAE,MAAM,KAAK,MAAM,YAAY,KAAK,YAAY,MAAM;AAAA,QACpE;AACA,oBAAY,KAAK,EAAE,GAAG,MAAM,MAAAA,MAAK,CAAC;AAClC,aAAK,QACF,MAAM,EAAE,UAAU,KAAK,MAAM,UAAU,OAAO,GAAG,CAAC,EAClD,MAAM,uBAAuB;AAChC,YAAI;AACF,gBAAMA;AAAA,QACR,QAAQ;AACN,eAAK,QACF,MAAM,EAAE,UAAU,KAAK,MAAM,UAAU,OAAO,GAAG,CAAC,EAClD,MAAM,6BAA6B;AAAA,QACxC;AAAA,MACF;AAEA,YAAM,gBAAgB,CAAC;AACvB,YAAM,mBAAmB,CAAC;AAC1B,iBAAW,OAAO,aAAa;AAE7B,cAAMA,QAAO,MAAM,IAAI;AACvB,YAAI,CAACA,SAAQA,MAAK,WAAW,OAAW;AACxC,sBAAc,KAAK,GAAG;AACtB,yBAAiB,KAAK,YAAY,6BAA6BA,KAAI,CAAC;AAAA,MACtE;AAEA,UAAI,CAAC,cAAc,OAAQ;AAG3B,YAAM,qBAAqB,CAAC,YAAY,gBAAgB,eAAe,aAAa,CAAC;AACrF,yBAAmB,KAAK,GAAG,gBAAgB;AAG3C,YAAM,kBAAkB,aAAa;AAAA,QACnC,OAAO;AAAA,QACP,OAAO;AAAA,QACP,OAAO,iBAAiB;AAAA,QACxB;AAAA,MACF;AAGA,YAAM,UAAU,OAAO,OAAO,QAAQ,KAAK;AAC3C,cAAQ,SAAS,KAAK,GAAG,kBAAkB;AAC3C,cAAQ,SAAS,KAAK,GAAG,iBAAiB,WAAW,EAAE,iBAAiB;AAExE,YAAM,kBAAkB,KAAK,IAAI,KAAK;AAAA,QACpC;AAAA,QACA,QAAQ,KAAK;AAAA,MACf,CAAC;AAED,YAAM,kBAAkB,KAAK,uBAAuB,gBAAgB,IAAI,eAAe;AACvF,sBAAgB,WAAW,iBAAiB,eAAe;AAC3D,aAAO,gBAAgB,eAAe;AAEtC,WAAK,KAAK,iCAAkC,WAAW;AAAA,IACzD;AAEA,QAAI,WAAW;AACf,UAAM,OAAO,qBAAqB,EAAE,KAAK,MAAM;AAC7C,iBAAW;AAAA,IACb,CAAC;AACD,WAAO,CAAC,OAAO,sBAAsB;AACnC,YAAM,UAAU,OAAO,oBAAoB;AAC3C,YAAM,QAAQ,KAAK,CAAC,SAAS,IAAI,CAAC;AAClC,aAAO,OAAO,oBAAoB,QAAQ;AACxC,cAAM,SAAS,OAAO,oBAAoB,CAAC;AAC3C,aAAK,iBAAiB;AACtB,cAAM,KAAK,YAAY,MAAM;AAC7B,eAAO,oBAAoB,MAAM;AACjC,aAAK,iBAAiB;AAAA,MACxB;AAEA,aAAO,oBAAoB,QAAQ,MAAM,OAAO,oBAAoB,IAAI,CAAC;AACzE,UAAI,UAAU;AACZ,eAAO,yBAAyB;AAAA,MAClC;AAAA,IACF;AACA,WAAO,QAAQ;AAAA,EACjB;AAAA,EAEA,MAAM,sBACJ,qBACA,UACA,MACA,SACA,IACA,WACA;AACA,SAAK,MAAO,iBAAkB,qBAAqB;AAAA,MACjD;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;AACD,UAAM,SAAS,MAAM,KAAK,MAAO,iBAAkB,WAAW;AAAA,MAC5D,gBAAgB;AAAA,MAChB,OAAO;AAAA,MACP,YAAY;AAAA,QACV,CAAC,gCAAgC,GAAG;AAAA,QACpC,CAAC,6BAA6B,GAAG,QAAQ,SAAS;AAAA,QAClD,CAAC,oBAAoB,GAAG;AAAA,MAC1B;AAAA,IACF,CAAC;AACD,UAAM,OAAO,MAAM,IAAI;AACvB,UAAM,OAAO,MAAM;AAAA,EACrB;AAAA,EAEA,uBACE,UACA,QACiB;AACjB,UAAM,eAAe,IAAI,sBAAsB,sBAAsB;AAErE,iBAAa,GAAG,eAAe,OAAO,SAAS;AAv6BnD;AAw6BM,WAAK,wBAAwB,KAAK;AAClC,UAAI,CAAC,KAAK,kBAAkB;AAC1B,aAAK,mBAAmB,WAAW;AAAA,MACrC;AACA,YAAM,KAAK;AAAA,QACT,KAAK,MAAO,iBAAkB;AAAA,UAC9B,UAAK,sBAAL,mBAAwB,QAAO;AAAA,QAC/B,KAAK;AAAA,QACL,KAAK;AAAA,QACL,KAAK;AAAA,QACL,KAAK;AAAA,MACP;AACA,UAAI,KAAK,OAAO;AACd,aAAK,mBAAmB;AAAA,MAC1B;AAAA,IACF,CAAC;AAED,QAAI,CAAC,KAAK,cAAc;AACtB,YAAM,IAAI,MAAM,+CAA+C;AAAA,IACjE;AAEA,QAAI,kBAAkB,WAAW;AAC/B,eAAS,0BAA0B,UAAU,MAAM;AAAA,IACrD;AAEA,UAAM,WAAW;AACjB,QAAI,EAAE,OAAO,WAAW,WAAW;AAAA,IAEnC;AAEA,UAAM,YAAY,KAAK,MAAM,kBAAkB,MAAM,QAAQ;AAC7D,QAAI,CAAC,WAAW;AACd,YAAM,IAAI,MAAM,+DAA+D;AAAA,IACjF;AAEA,WAAO,KAAK,aAAa,WAAW,UAAU,WAAW,YAAY;AAAA,EACvE;AAAA,EAEA,MAAM,2BAA2B;AAC/B,QAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,oBAAoB;AAClE,WAAK,QACF,MAAM,EAAE,UAAU,KAAK,eAAe,GAAG,CAAC,EAC1C,MAAM,yEAAyE;AAClF;AAAA,IACF;AAEA,QAAI,CAAC,KAAK,oBAAoB;AAC5B,UAAI,KAAK,MAAM,uBAAuB,CAAC,KAAK,iBAAiB;AAC3D;AAAA,MACF;AACA,WAAK,sBAAsB;AAAA,IAC7B;AAEA,QAAI,CAAC,KAAK,oBAAoB;AAC5B,YAAM,IAAI,MAAM,kCAAkC;AAAA,IACpD;AAIA,QAAI,KAAK,iBAAiB,MAAM;AAC9B,uBAAiB,UAAU,KAAK,cAAc;AAC5C,YAAI,WAAW,mBAAmB,eAAgB;AAClD,YAAI,CAAC,OAAO,QAAS;AACrB,YAAI,OAAO,mBAAoB,QAAO,UAAU;AAAA,MAClD;AAAA,IACF;AAEA,SAAK,QAAQ,MAAM,EAAE,UAAU,KAAK,mBAAmB,GAAG,CAAC,EAAE,MAAM,uBAAuB;AAE1F,QAAI,KAAK,iBAAiB;AACxB,YAAM,sBAAsB,KAAK,IAAI,IAAI,KAAK;AAC9C,YAAM,qBAAqB,KAAK;AAAA,SAC7B,KAAK,4BAA4B,KAAK,KAAK;AAAA,QAC5C;AAAA,MACF;AACA,YAAM,UAA8B;AAAA,QAClC,WAAW,KAAK,IAAI;AAAA,QACpB,YAAY,KAAK,mBAAmB;AAAA,QACpC,qBAAqB;AAAA,QACrB;AAAA,MACF;AACA,WAAK,KAAK,2BAA4B,OAAO;AAAA,IAC/C;AAEA,SAAK,qBAAqB,KAAK,kBAAkB;AACjD,SAAK,qBAAqB;AAC1B,SAAK,0BAA0B;AAAA,EACjC;AAAA,EAEA,uBAAuB;AACrB,QACE,CAAC,KAAK,kBACN,CAAC,KAAK,eAAe,sBACrB,KAAK,eAAe,aACpB;AACA;AAAA,IACF;AAEA,QAAI,KAAK,MAAM,sBAAsB,GAAG;AAGtC,YAAM,eAAe,KAAK,MAAM,cAAc,cAAc;AAAA,QAC1D,KAAK;AAAA,MACP;AACA,UAAI,aAAa,SAAS,KAAK,MAAM,mBAAmB;AACtD;AAAA,MACF;AAAA,IACF;AACA,SAAK,eAAe,UAAU;AAAA,EAChC;AAAA,EAEA,qBAAqB,QAAsB;AACzC,SAAK,aAAa,IAAI,MAAM;AAC5B,SAAK,aAAa,IAAI,mBAAmB,cAAc;AACvD,SAAK,iBAAiB,QAAQ;AAAA,EAChC;AAAA;AAAA,EAGA,MAAM,QAAQ;AA9hChB;AA+hCI,QAAI,CAAC,KAAK,UAAU;AAClB;AAAA,IACF;AAEA,eAAK,UAAL,mBAAY,mBAAmB,UAAU;AAAA,EAE3C;AACF;AAEA,gBAAgB,0BACd,UACA,QACuB;AA3iCzB;AA4iCE,QAAM,YAAY,KAAK,IAAI;AAC3B,MAAI,aAAa;AACjB,mBAAiB,SAAS,QAAQ;AAChC,UAAM,WAAU,WAAM,QAAQ,CAAC,MAAf,mBAAkB,MAAM;AACxC,QAAI,CAAC,QAAS;AAEd,QAAI,YAAY;AACd,mBAAa;AACb,UAAI,EACD,MAAM,EAAE,UAAU,SAAS,KAAK,MAAM,KAAK,IAAI,IAAI,SAAS,EAAE,CAAC,EAC/D,MAAM,0BAA0B;AAAA,IACrC;AACA,UAAM;AAAA,EACR;AACF;AAGA,MAAM,wBAAwB;AAAA;AAAA,EAEnB,cAAc;AAAA,EACd,4BAA4B;AAAA,EAC5B,4BAA4B;AAAA;AAAA,EAC5B,0BAA0B;AAAA,EAEnC;AAAA,EACA;AAAA,EACA,oBAAoB,IAAI,OAAO;AAAA,EAC/B,uBAAuB;AAAA,EACvB,2BAA2B;AAAA,EAC3B,YAAY;AAAA,EACZ;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YACE,cACA,qBACA,OACA,cACA;AACA,SAAK,gBAAgB;AACrB,SAAK,oBAAoB;AACzB,SAAK,wBAAwB;AAC7B,SAAK,SAAS;AACd,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,IAAI,aAAsB;AACxB,WAAO,CAAC,KAAK,kBAAkB;AAAA,EACjC;AAAA,EAEA,uBAAuB,YAAoB;AACzC,SAAK,uBAAuB,WAAW,KAAK;AAC5C,QAAI,KAAK,UAAW;AAEpB,UAAM,uBACJ,KAAK,IAAI,IAAI,KAAK,2BAA2B,KAAK;AACpD,QAAI,QAAQ,uBAAuB,KAAK,oBAAoB,KAAK;AACjE,YAAQ,KAAK,oBAAoB,IAAI,QAAQ,KAAK,4BAA4B;AAE9E,SAAK,KAAK,KAAK;AAAA,EACjB;AAAA;AAAA,EAGA,qBAAqB,GAAa;AA9mCpC;AA+mCI,SAAK,YAAY;AACjB,QAAI,KAAK,YAAY;AACnB,iBAAK,WAAL,mBAAa;AAAA,IACf;AAAA,EACF;AAAA;AAAA,EAGA,mBAAmB,GAAa;AAC9B,SAAK,YAAY;AACjB,SAAK,2BAA2B,KAAK,IAAI;AAEzC,QAAI,KAAK,sBAAsB;AAC7B,YAAM,QAAQ,KAAK,oBAAoB,IACnC,KAAK,oBAAoB,KAAK,4BAC9B;AACJ,WAAK,KAAK,KAAK;AAAA,IACjB;AAAA,EACF;AAAA;AAAA,EAIA,sBAA+B;AAC7B,WACE,KAAK,qBAAqB,SAAS,KACnC,KAAK,YAAY,SAAS,KAAK,qBAAqB,KAAK,qBAAqB,SAAS,CAAC,CAAE;AAAA,EAE9F;AAAA,EAEA,eAAe;AACb,SAAK,uBAAuB;AAC5B,SAAK,2BAA2B;AAAA,EAClC;AAAA,EAEA,KAAK,OAAe;AAhpCtB;AAipCI,UAAM,UAAU,OAAOH,QAAe,SAAsB,WAAwB;AAClF,UAAI,KAAK,wBAAwB,CAAC,KAAK,aAAa,KAAK,eAAe;AACtE,cAAM,YAAY,KAAK,IAAI;AAC3B,cAAM,UAAU,MAAM,KAAK,cAAc,iBAAiB,OAAO;AACjE,cAAM,oBAAoB,KAAK,cAAc;AAC7C,cAAM,UAAU,KAAK,IAAI,IAAI;AAC7B,YAAI,UAAU,mBAAmB;AAC/B,UAAAA,SAAQ,KAAK;AAAA,QACf;AACA,QAAAA,SAAQ,KAAK,IAAI,GAAGA,SAAQ,OAAO;AAAA,MACrC;AACA,YAAM,UAAU,WAAW,MAAM;AAC/B,aAAK,aAAa;AAClB,aAAK,cAAc;AAAA,MACrB,GAAGA,MAAK;AACR,aAAO,iBAAiB,SAAS,MAAM;AACrC,qBAAa,OAAO;AAAA,MACtB,CAAC;AAAA,IACH;AAEA,eAAK,WAAL,mBAAa;AACb,SAAK,SAAS,IAAI,gBAAgB;AAClC,SAAK,oBAAoB,IAAI,OAAO;AACpC,UAAM,YAAY,KAAK,OAAO,QAAQ,KAAK;AAC3C,cAAU,OAAO,EAAE,MAAM,KAAK,OAAO,iBAAiB,MAAM,SAAS,KAAK,CAAC;AAC3E,SAAK,qBAAqB,QAAQ,OAAO,WAAW,KAAK,OAAO,MAAM;AAAA,EACxE;AACF;","names":["VPAEvent","participant","delay","resolve","isUsingTools","task"]}
|
package/package.json
CHANGED
package/src/constants.ts
CHANGED
|
@@ -4,4 +4,5 @@
|
|
|
4
4
|
export const ATTRIBUTE_TRANSCRIPTION_TRACK_ID = 'lk.transcribed_track_id';
|
|
5
5
|
export const ATTRIBUTE_TRANSCRIPTION_FINAL = 'lk.transcription_final';
|
|
6
6
|
export const TOPIC_TRANSCRIPTION = 'lk.transcription';
|
|
7
|
+
export const ATTRIBUTE_SEGMENT_ID = 'lk.segment_id';
|
|
7
8
|
export const TOPIC_CHAT = 'lk.chat';
|
|
@@ -18,9 +18,11 @@ import {
|
|
|
18
18
|
TrackPublishOptions,
|
|
19
19
|
TrackSource,
|
|
20
20
|
} from '@livekit/rtc-node';
|
|
21
|
+
import { randomUUID } from 'node:crypto';
|
|
21
22
|
import { EventEmitter } from 'node:events';
|
|
22
23
|
import { AudioByteStream } from '../audio.js';
|
|
23
24
|
import {
|
|
25
|
+
ATTRIBUTE_SEGMENT_ID,
|
|
24
26
|
ATTRIBUTE_TRANSCRIPTION_FINAL,
|
|
25
27
|
ATTRIBUTE_TRANSCRIPTION_TRACK_ID,
|
|
26
28
|
TOPIC_TRANSCRIPTION,
|
|
@@ -72,6 +74,7 @@ export class MultimodalAgent extends EventEmitter {
|
|
|
72
74
|
|
|
73
75
|
#textResponseRetries = 0;
|
|
74
76
|
#maxTextResponseRetries: number;
|
|
77
|
+
#transcriptionId?: string;
|
|
75
78
|
|
|
76
79
|
constructor({
|
|
77
80
|
model,
|
|
@@ -257,13 +260,20 @@ export class MultimodalAgent extends EventEmitter {
|
|
|
257
260
|
|
|
258
261
|
const synchronizer = new TextAudioSynchronizer(defaultTextSyncOptions);
|
|
259
262
|
synchronizer.on('textUpdated', async (text) => {
|
|
263
|
+
if (!this.#transcriptionId) {
|
|
264
|
+
this.#transcriptionId = randomUUID();
|
|
265
|
+
}
|
|
260
266
|
await this.#publishTranscription(
|
|
261
267
|
this.room!.localParticipant!.identity!,
|
|
262
268
|
this.#getLocalTrackSid()!,
|
|
263
269
|
text.text,
|
|
264
270
|
text.final,
|
|
265
271
|
text.id,
|
|
272
|
+
this.#transcriptionId,
|
|
266
273
|
);
|
|
274
|
+
if (text.final) {
|
|
275
|
+
this.#transcriptionId = undefined;
|
|
276
|
+
}
|
|
267
277
|
});
|
|
268
278
|
|
|
269
279
|
const handle = this.#agentPlayout?.play(
|
|
@@ -312,7 +322,17 @@ export class MultimodalAgent extends EventEmitter {
|
|
|
312
322
|
const participantIdentity = this.linkedParticipant?.identity;
|
|
313
323
|
const trackSid = this.subscribedTrack?.sid;
|
|
314
324
|
if (participantIdentity && trackSid) {
|
|
315
|
-
|
|
325
|
+
if (!this.#transcriptionId) {
|
|
326
|
+
this.#transcriptionId = randomUUID();
|
|
327
|
+
}
|
|
328
|
+
await this.#publishTranscription(
|
|
329
|
+
participantIdentity,
|
|
330
|
+
trackSid,
|
|
331
|
+
'…',
|
|
332
|
+
false,
|
|
333
|
+
ev.itemId,
|
|
334
|
+
this.#transcriptionId,
|
|
335
|
+
);
|
|
316
336
|
} else {
|
|
317
337
|
this.#logger.error('Participant or track not set');
|
|
318
338
|
}
|
|
@@ -325,13 +345,18 @@ export class MultimodalAgent extends EventEmitter {
|
|
|
325
345
|
const participantIdentity = this.linkedParticipant?.identity;
|
|
326
346
|
const trackSid = this.subscribedTrack?.sid;
|
|
327
347
|
if (participantIdentity && trackSid) {
|
|
348
|
+
if (!this.#transcriptionId) {
|
|
349
|
+
this.#transcriptionId = randomUUID();
|
|
350
|
+
}
|
|
328
351
|
await this.#publishTranscription(
|
|
329
352
|
participantIdentity,
|
|
330
353
|
trackSid,
|
|
331
354
|
transcription,
|
|
332
355
|
true,
|
|
333
356
|
ev.itemId,
|
|
357
|
+
this.#transcriptionId,
|
|
334
358
|
);
|
|
359
|
+
this.#transcriptionId = undefined;
|
|
335
360
|
} else {
|
|
336
361
|
this.#logger.error('Participant or track not set');
|
|
337
362
|
}
|
|
@@ -360,7 +385,17 @@ export class MultimodalAgent extends EventEmitter {
|
|
|
360
385
|
const participantIdentity = this.linkedParticipant?.identity;
|
|
361
386
|
const trackSid = this.subscribedTrack?.sid;
|
|
362
387
|
if (participantIdentity && trackSid) {
|
|
363
|
-
|
|
388
|
+
if (!this.#transcriptionId) {
|
|
389
|
+
this.#transcriptionId = randomUUID();
|
|
390
|
+
}
|
|
391
|
+
await this.#publishTranscription(
|
|
392
|
+
participantIdentity,
|
|
393
|
+
trackSid,
|
|
394
|
+
'…',
|
|
395
|
+
false,
|
|
396
|
+
ev.itemId,
|
|
397
|
+
this.#transcriptionId,
|
|
398
|
+
);
|
|
364
399
|
}
|
|
365
400
|
});
|
|
366
401
|
|
|
@@ -492,6 +527,7 @@ export class MultimodalAgent extends EventEmitter {
|
|
|
492
527
|
text: string,
|
|
493
528
|
isFinal: boolean,
|
|
494
529
|
id: string,
|
|
530
|
+
segmentId: string,
|
|
495
531
|
): Promise<void> {
|
|
496
532
|
this.#logger.debug(
|
|
497
533
|
`Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`,
|
|
@@ -522,6 +558,7 @@ export class MultimodalAgent extends EventEmitter {
|
|
|
522
558
|
attributes: {
|
|
523
559
|
[ATTRIBUTE_TRANSCRIPTION_TRACK_ID]: trackSid,
|
|
524
560
|
[ATTRIBUTE_TRANSCRIPTION_FINAL]: isFinal.toString(),
|
|
561
|
+
[ATTRIBUTE_SEGMENT_ID]: segmentId,
|
|
525
562
|
},
|
|
526
563
|
});
|
|
527
564
|
await stream.write(text);
|
|
@@ -18,6 +18,7 @@ import type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';
|
|
|
18
18
|
import { randomUUID } from 'node:crypto';
|
|
19
19
|
import EventEmitter from 'node:events';
|
|
20
20
|
import {
|
|
21
|
+
ATTRIBUTE_SEGMENT_ID,
|
|
21
22
|
ATTRIBUTE_TRANSCRIPTION_FINAL,
|
|
22
23
|
ATTRIBUTE_TRANSCRIPTION_TRACK_ID,
|
|
23
24
|
TOPIC_TRANSCRIPTION,
|
|
@@ -537,6 +538,7 @@ export class VoicePipelineAgent extends (EventEmitter as new () => TypedEmitter<
|
|
|
537
538
|
this.#transcribedInterimText,
|
|
538
539
|
false,
|
|
539
540
|
this.#transcriptionId,
|
|
541
|
+
this.#transcriptionId,
|
|
540
542
|
);
|
|
541
543
|
});
|
|
542
544
|
this.#humanInput.on(HumanInputEvent.FINAL_TRANSCRIPT, async (event) => {
|
|
@@ -556,6 +558,7 @@ export class VoicePipelineAgent extends (EventEmitter as new () => TypedEmitter<
|
|
|
556
558
|
this.transcribedText,
|
|
557
559
|
true,
|
|
558
560
|
this.#transcriptionId,
|
|
561
|
+
this.#transcriptionId,
|
|
559
562
|
);
|
|
560
563
|
|
|
561
564
|
this.#transcriptionId = undefined;
|
|
@@ -895,6 +898,7 @@ export class VoicePipelineAgent extends (EventEmitter as new () => TypedEmitter<
|
|
|
895
898
|
text: string,
|
|
896
899
|
isFinal: boolean,
|
|
897
900
|
id: string,
|
|
901
|
+
segmentId: string,
|
|
898
902
|
) {
|
|
899
903
|
this.#room!.localParticipant!.publishTranscription({
|
|
900
904
|
participantIdentity: participantIdentity,
|
|
@@ -916,6 +920,7 @@ export class VoicePipelineAgent extends (EventEmitter as new () => TypedEmitter<
|
|
|
916
920
|
attributes: {
|
|
917
921
|
[ATTRIBUTE_TRANSCRIPTION_TRACK_ID]: trackSid,
|
|
918
922
|
[ATTRIBUTE_TRANSCRIPTION_FINAL]: isFinal.toString(),
|
|
923
|
+
[ATTRIBUTE_SEGMENT_ID]: segmentId,
|
|
919
924
|
},
|
|
920
925
|
});
|
|
921
926
|
await stream.write(text);
|
|
@@ -930,13 +935,20 @@ export class VoicePipelineAgent extends (EventEmitter as new () => TypedEmitter<
|
|
|
930
935
|
// TODO: where possible we would want to use deltas instead of full text segments, esp for LLM streams over the streamText API
|
|
931
936
|
synchronizer.on('textUpdated', async (text) => {
|
|
932
937
|
this.#agentTranscribedText = text.text;
|
|
938
|
+
if (!this.#transcriptionId) {
|
|
939
|
+
this.#transcriptionId = randomUUID();
|
|
940
|
+
}
|
|
933
941
|
await this.#publishTranscription(
|
|
934
942
|
this.#room!.localParticipant!.identity!,
|
|
935
943
|
this.#agentPublication?.sid ?? '',
|
|
936
944
|
text.text,
|
|
937
945
|
text.final,
|
|
938
946
|
text.id,
|
|
947
|
+
this.#transcriptionId,
|
|
939
948
|
);
|
|
949
|
+
if (text.final) {
|
|
950
|
+
this.#transcriptionId = undefined;
|
|
951
|
+
}
|
|
940
952
|
});
|
|
941
953
|
|
|
942
954
|
if (!this.#agentOutput) {
|