Chore: stt reworks [WIP]

bot.go

@@ -606,6 +606,6 @@ func init() {
 		orator = extra.NewOrator(logger, cfg)
 	}
 	if cfg.STT_ENABLED {
-		asr = extra.NewWhisperSTT(logger, cfg.STT_URL, 16000)
+		asr = extra.NewSTT(logger, cfg)
 	}
 }

config/config.go

@@ -56,8 +56,13 @@ type Config struct {
 	TTS_ENABLED bool    `toml:"TTS_ENABLED"`
 	TTS_SPEED   float32 `toml:"TTS_SPEED"`
 	// STT
+	STT_TYPE          string `toml:"STT_TYPE"` // WHISPER_SERVER, WHISPER_BINARY
 	STT_URL           string `toml:"STT_URL"`
+	STT_SR            int    `toml:"STT_SR"`
 	STT_ENABLED       bool   `toml:"STT_ENABLED"`
+	WhisperBinaryPath string `toml:"WhisperBinaryPath"`
+	WhisperModelPath  string `toml:"WhisperModelPath"`
+	STT_LANG          string `toml:"STT_LANG"`
 	DBPATH            string `toml:"DBPATH"`
 }
 
@@ -93,6 +98,7 @@ func LoadConfigOrDefault(fn string) *Config {
 		config.TTS_ENABLED = false
 		config.TTS_URL = "http://localhost:8880/v1/audio/speech"
 		config.FetchModelNameAPI = "http://localhost:8080/v1/models"
+		config.STT_SR = 16000
 	}
 	config.CurrentAPI = config.ChatAPI
 	config.APIMap = map[string]string{

extra/stt.go

@@ -5,6 +5,7 @@ import (
 	"encoding/binary"
 	"errors"
 	"fmt"
+	"gf-lt/config"
 	"io"
 	"log/slog"
 	"mime/multipart"
@@ -27,7 +28,16 @@ type StreamCloser interface {
 	Close() error
 }
 
-type WhisperSTT struct {
+func NewSTT(logger *slog.Logger, cfg *config.Config) STT {
+	switch cfg.STT_TYPE {
+	case "WHISPER_BINARY":
+	case "WHISPER_SERVER":
+		return NewWhisperServer(logger, cfg)
+	}
+	return NewWhisperServer(logger, cfg)
+}
+
+type WhisperServer struct {
 	logger      *slog.Logger
 	ServerURL   string
 	SampleRate  int
@@ -35,16 +45,16 @@ type WhisperSTT struct {
 	recording   bool
 }
 
-func NewWhisperSTT(logger *slog.Logger, serverURL string, sampleRate int) *WhisperSTT {
-	return &WhisperSTT{
+func NewWhisperServer(logger *slog.Logger, cfg *config.Config) *WhisperServer {
+	return &WhisperServer{
 		logger:      logger,
-		ServerURL:   serverURL,
-		SampleRate:  sampleRate,
+		ServerURL:   cfg.STT_URL,
+		SampleRate:  cfg.STT_SR,
 		AudioBuffer: new(bytes.Buffer),
 	}
 }
 
-func (stt *WhisperSTT) StartRecording() error {
+func (stt *WhisperServer) StartRecording() error {
 	if err := stt.microphoneStream(stt.SampleRate); err != nil {
 		return fmt.Errorf("failed to init microphone: %w", err)
 	}
@@ -52,7 +62,7 @@ func (stt *WhisperSTT) StartRecording() error {
 	return nil
 }
 
-func (stt *WhisperSTT) StopRecording() (string, error) {
+func (stt *WhisperServer) StopRecording() (string, error) {
 	stt.recording = false
 	// wait loop to finish?
 	if stt.AudioBuffer == nil {
@@ -107,7 +117,7 @@ func (stt *WhisperSTT) StopRecording() (string, error) {
 	return strings.TrimSpace(strings.ReplaceAll(resptext, "\n ", "\n")), nil
 }
 
-func (stt *WhisperSTT) writeWavHeader(w io.Writer, dataSize int) {
+func (stt *WhisperServer) writeWavHeader(w io.Writer, dataSize int) {
 	header := make([]byte, 44)
 	copy(header[0:4], "RIFF")
 	binary.LittleEndian.PutUint32(header[4:8], uint32(36+dataSize))
@@ -127,11 +137,11 @@ func (stt *WhisperSTT) writeWavHeader(w io.Writer, dataSize int) {
 	}
 }
 
-func (stt *WhisperSTT) IsRecording() bool {
+func (stt *WhisperServer) IsRecording() bool {
 	return stt.recording
 }
 
-func (stt *WhisperSTT) microphoneStream(sampleRate int) error {
+func (stt *WhisperServer) microphoneStream(sampleRate int) error {
 	if err := portaudio.Initialize(); err != nil {
 		return fmt.Errorf("portaudio init failed: %w", err)
 	}

extra/whisper_binary.go

@@ -0,0 +1,31 @@
+package extra
+
+import (
+	"context"
+	"gf-lt/config"
+	"log/slog"
+	"os/exec"
+	"sync"
+)
+
+type WhisperBinary struct {
+	whisperPath string
+	modelPath   string
+	lang        string
+	ctx         context.Context
+	cancel      context.CancelFunc
+	mu          sync.Mutex
+	running     bool
+	cmd         *exec.Cmd
+}
+
+func NewWhisperBinary(logger *slog.Logger, cfg *config.Config) *WhisperBinary {
+	ctx, cancel := context.WithCancel(context.Background())
+	return &WhisperBinary{
+		whisperPath: cfg.WhisperBinaryPath,
+		modelPath:   cfg.WhisperModelPath,
+		lang:        cfg.STT_LANG,
+		ctx:         ctx,
+		cancel:      cancel,
+	}
+}