feat: initial commit

internal/infrastructure/speech/audio_converter.go

@@ -0,0 +1,63 @@
+package speech
+
+import (
+	"bytes"
+	"context"
+	"fmt"
+	"os"
+	"os/exec"
+	"path/filepath"
+
+	"github.com/google/uuid"
+)
+
+type FFmpegConverter struct {
+	ffmpegPath string
+	tempDir    string
+}
+
+func NewFFmpegConverter(ffmpegPath, tempDir string) *FFmpegConverter {
+	if ffmpegPath == "" {
+		ffmpegPath = "ffmpeg"
+	}
+	if tempDir == "" {
+		tempDir = os.TempDir()
+	}
+	return &FFmpegConverter{ffmpegPath: ffmpegPath, tempDir: tempDir}
+}
+
+func (c *FFmpegConverter) Convert(ctx context.Context, input []byte, fromMime, toMime string) ([]byte, error) {
+	id := uuid.New().String()
+	inFile := filepath.Join(c.tempDir, id+".input")
+	outFile := filepath.Join(c.tempDir, id+".wav")
+
+	defer os.Remove(inFile)
+	defer os.Remove(outFile)
+
+	if err := os.WriteFile(inFile, input, 0600); err != nil {
+		return nil, fmt.Errorf("write temp input: %w", err)
+	}
+
+	cmd := exec.CommandContext(ctx, c.ffmpegPath,
+		"-i", inFile,
+		"-ar", "16000",
+		"-ac", "1",
+		"-f", "wav",
+		"-y",
+		outFile,
+	)
+
+	var stderr bytes.Buffer
+	cmd.Stderr = &stderr
+
+	if err := cmd.Run(); err != nil {
+		return nil, fmt.Errorf("ffmpeg conversion: %w: %s", err, stderr.String())
+	}
+
+	out, err := os.ReadFile(outFile)
+	if err != nil {
+		return nil, fmt.Errorf("read converted file: %w", err)
+	}
+
+	return out, nil
+}