Class VoicePipelineAgent

constructor

• new VoicePipelineAgent(vad, stt, llm, tts, opts?): VoicePipelineAgent

• Parameters

  • vad: VAD

    Voice Activity Detection instance.
  • stt: stt.STT

    Speech-to-Text instance.
  • llm: llm.LLM

    Large Language Model instance.
  • tts: tts.TTS

    Text-to-Speech instance.
  • opts: Partial<VPAOptions> = defaultVPAOptions

    Additional VoicePipelineAgent options.

  Returns VoicePipelineAgent

Readonly MIN_TIME_PLAYED_FOR_COMMIT

chatCtx

• get chatCtx(): ChatContext

• Returns ChatContext

fncCtx

• get fncCtx(): undefined | FunctionContext

• Returns undefined | FunctionContext

• set fncCtx(ctx): void

• Parameters

  • ctx: FunctionContext

  Returns void

llm

• get llm(): llm.LLM

• Returns llm.LLM

stt

• get stt(): stt.STT

• Returns stt.STT

tts

• get tts(): tts.TTS

• Returns tts.TTS

vad

• get vad(): VAD

• Returns VAD

close

• close(): Promise<void>

• Close the voice assistant.

  Returns Promise<void>

say

• say(source, allowInterruptions?, addToChatCtx?): Promise<void>

• Play a speech source through the voice assistant.

  Parameters

  • source: string | llm.LLMStream | AsyncIterable<string>
  • allowInterruptions: boolean = true
  • addToChatCtx: boolean = true

  Returns Promise<void>

start

• start(room, participant?): void

• Start the voice assistant.

  Parameters

  • room: Room

    The room to connect to.
  • participant: null | string | RemoteParticipant = null

    The participant to listen to.

    Remarks

    Can be a participant or an identity. If omitted, the first participant in the room will be selected.

  Returns void