audiobuffer

Documentation

Overview

Package audiobuffer records a conversation's audio. A Processor taps the audio flowing through the pipeline — the user's InputAudioRawFrames and the bot's Output/TTS audio — keeps the two tracks time-aligned by padding silence for wall-clock gaps, and delivers the recording through callbacks: a merged mix (mono) or interleaved stereo (user left, bot right), and the separate tracks.

Recording is off until started, either automatically at pipeline start (AutoStart) or by StartRecording / an AudioBufferStartRecordingFrame. It is stopped by StopRecording, an AudioBufferStopRecordingFrame, or pipeline shutdown, which flushes the final audio and fires OnRecordingStopped.

Index

• type Config
• type Processor
  • func New(cfg Config) *Processor
  • func (p *Processor) Cleanup(ctx context.Context) error
  • func (p *Processor) ProcessFrame(ctx context.Context, f frames.Frame, dir processor.Direction) error
  • func (p *Processor) Recording() bool
  • func (p *Processor) StartRecording()
  • func (p *Processor) StopRecording()

Types

type Config

type Config struct {
	// SampleRate overrides the recording sample rate in Hz; 0 uses the pipeline
	// output rate from the StartFrame.
	SampleRate int
	// NumChannels is 1 for a mono mix of user and bot, or 2 for stereo with the
	// user on the left and the bot on the right; 0 uses 1.
	NumChannels int
	// BufferSize is the byte threshold at which OnAudioData and OnTrackAudioData
	// fire with the audio accumulated so far; 0 emits only when recording stops.
	BufferSize int
	// AutoStart begins recording as soon as the pipeline starts.
	AutoStart bool

	// OnRecordingStarted is called when recording transitions to active.
	OnRecordingStarted func()
	// OnRecordingStopped is called after recording stops and the final audio has
	// been delivered.
	OnRecordingStopped func()
	// OnAudioData receives the merged recording (mono mix or interleaved stereo)
	// at each BufferSize boundary and once more on stop.
	OnAudioData func(audio []byte, sampleRate, numChannels int)
	// OnTrackAudioData receives the separate, time-aligned user and bot tracks at
	// each BufferSize boundary and once more on stop.
	OnTrackAudioData func(user, bot []byte, sampleRate, numChannels int)
}

Config configures a recording Processor.

type Processor

type Processor struct {
	*processor.Base
	// contains filtered or unexported fields
}

Processor records the pipeline's audio. It is placed downstream, typically after the output transport, so it sees both the user's input audio and the bot's output audio.

func New

func New(cfg Config) *Processor

New builds a recording Processor.

func (*Processor) Cleanup

func (p *Processor) Cleanup(ctx context.Context) error

Cleanup releases the resamplers after the processor stops.

func (*Processor) ProcessFrame

func (p *Processor) ProcessFrame(ctx context.Context, f frames.Frame, dir processor.Direction) error

ProcessFrame taps audio for recording and forwards every frame untouched.

func (*Processor) Recording

func (p *Processor) Recording() bool

Recording reports whether recording is active.

func (*Processor) StartRecording

func (p *Processor) StartRecording()

StartRecording begins recording. It does nothing if already recording.

func (*Processor) StopRecording

func (p *Processor) StopRecording()

StopRecording flushes the buffered audio and stops recording. It does nothing if not recording.