bodhi-realtime-agent / AudioOutputObserver

Type Alias: AudioOutputObserver

AudioOutputObserver = (pcm, meta) => void

Defined in: src/types/session-seams.ts:11

Receives each chunk of native assistant audio (the model's own speech) as PCM, right before the session sends it to the client. Chunks dropped before delivery (by outputInterceptor.audio, in transcription mode, or for an already-finalized turn) are not observed, and neither is audio synthesized by an external TTS provider. Register with VoiceSession.observeAudioOutput.

pcm is the buffer the client receives; treat it as read-only. A throw is reported through hooks.onError and does not stop delivery.

Parameters

pcm

Buffer

meta

encoding

"pcm"

Always 'pcm' (16-bit little-endian mono): G.711 output is decoded first.

sampleRate

number

Sample rate of pcm: the transport's output rate.

turnId

string | null

Id of the turn the chunk belongs to.

Returns

void