go-whisper-api/config/vad.go

package config

import (
    "fmt"
    "os"
    "path/filepath"
)

type VAD struct {
    Enabled        bool    `yaml:"enabled"`
    Model          string  `yaml:"model"`
    Threshold      float64 `yaml:"threshold"`
    MinSpeechMs    int     `yaml:"min_speech_duration_ms"`
    MinSilenceMs   int     `yaml:"min_silence_duration_ms"`
    MaxSpeechSec   float64 `yaml:"max_speech_duration_s"`
    SpeechPadMs    int     `yaml:"speech_pad_ms"`
    SamplesOverlap float64 `yaml:"samples_overlap"`
}

func DefaultVAD() VAD {
    return VAD{
        Threshold:      0.5,
        MinSpeechMs:    250,
        MinSilenceMs:   100,
        MaxSpeechSec:   0,
        SpeechPadMs:    30,
        SamplesOverlap: 0.1,
    }
}

func (v VAD) WithDefaults() VAD {
    d := DefaultVAD()
    if v.Threshold <= 0 {
        v.Threshold = d.Threshold
    }
    if v.MinSpeechMs <= 0 {
        v.MinSpeechMs = d.MinSpeechMs
    }
    if v.MinSilenceMs <= 0 {
        v.MinSilenceMs = d.MinSilenceMs
    }
    if v.SpeechPadMs <= 0 {
        v.SpeechPadMs = d.SpeechPadMs
    }
    if v.SamplesOverlap <= 0 {
        v.SamplesOverlap = d.SamplesOverlap
    }
    return v
}

func (v VAD) ResolveModelPath(modelsDir string) string {
    if v.Model == "" {
        return ""
    }
    if filepath.IsAbs(v.Model) {
        return v.Model
    }
    if modelsDir == "" {
        return v.Model
    }
    direct := filepath.Join(modelsDir, v.Model)
    if _, err := os.Stat(direct); err == nil {
        return direct
    }
    // VAD weights often live under models_dir/vad/ (not listed in /spr/models).
    base := filepath.Base(v.Model)
    inVAD := filepath.Join(modelsDir, "vad", base)
    if _, err := os.Stat(inVAD); err == nil {
        return inVAD
    }
    return direct
}

func (v VAD) Validate() error {
    if !v.Enabled {
        return nil
    }
    if v.Model == "" {
        return fmt.Errorf("vad.model is required when vad.enabled is true")
    }
    if _, err := os.Stat(v.Model); err != nil {
        return fmt.Errorf("vad model %q: %w", v.Model, err)
    }
    if v.Threshold < 0 || v.Threshold > 1 {
        return fmt.Errorf("vad.threshold must be between 0 and 1")
    }
    return nil
}