Struct PipecatSmartTurn 

pub struct PipecatSmartTurn {}

Pipecat Smart Turn v3 detector.

Buffers up to 8 seconds of audio internally. When predict is called, it takes the last 8s (zero-padded at front if shorter), extracts Whisper log-mel features, and runs ONNX inference.

Implementations

impl PipecatSmartTurn

pub fn new() -> Result<Self, TurnError>

Create a new Smart Turn detector, loading the ONNX model.

Trait Implementations

impl AudioTurnDetector for PipecatSmartTurn

fn push_audio(&mut self, _frame: &AudioFrame<'_>)

Feed audio into the internal buffer.

fn predict(&mut self) -> Result<TurnPrediction, TurnError>

Run prediction on buffered audio.

fn reset(&mut self)

Clear the internal buffer. Call when a new speech turn begins.