audio

type AnalyzeConfig struct {
	// TargetDuration is the desired segment length in seconds
	TargetDuration float64

	// StepSize is the sliding window step in seconds
	StepSize float64

	// FrameSize is the analysis frame size in seconds
	FrameSize float64

	// SilenceThreshold is the RMS threshold for silence detection (0-1)
	SilenceThreshold float64

	// TopN returns the top N segments
	TopN int
}

type SegmentMetrics struct {
	// MeanEnergy is the average RMS energy (higher = louder speech)
	MeanEnergy float64

	// EnergyVariance measures consistency (lower = more consistent)
	EnergyVariance float64

	// SpeechRatio is the fraction of frames with speech activity
	SpeechRatio float64

	// SilenceGaps is the number of silence gaps > 0.5s
	SilenceGaps int

	// MaxSilenceGap is the longest silence gap in seconds
	MaxSilenceGap float64

	// PeakClipping is the fraction of samples near clipping
	PeakClipping float64
}

type SegmentScore struct {
	StartSample int
	EndSample   int
	StartTime   float64
	EndTime     float64
	Score       float64
	Metrics     SegmentMetrics
}

type WAVHeader struct {
	SampleRate    uint32
	NumChannels   uint16
	BitsPerSample uint16
	DataSize      uint32
}