Steve Dudenhoeffer
9e288954f2
Migrate speech-to-text transcription types and OpenAI transcriber implementation from v1. Types are defined in provider/ to avoid import cycles and re-exported via type aliases from the root package. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
101 lines
3.3 KiB
Go
101 lines
3.3 KiB
Go
package llm
|
|
|
|
import (
|
|
"context"
|
|
"fmt"
|
|
"os"
|
|
"os/exec"
|
|
"path/filepath"
|
|
"strings"
|
|
|
|
"gitea.stevedudenhoeffer.com/steve/go-llm/v2/provider"
|
|
)
|
|
|
|
// Transcriber abstracts a speech-to-text model implementation.
|
|
type Transcriber = provider.Transcriber
|
|
|
|
// TranscriptionResponseFormat controls the output format requested from a transcriber.
|
|
type TranscriptionResponseFormat = provider.TranscriptionResponseFormat
|
|
|
|
const (
|
|
TranscriptionResponseFormatJSON = provider.TranscriptionResponseFormatJSON
|
|
TranscriptionResponseFormatVerboseJSON = provider.TranscriptionResponseFormatVerboseJSON
|
|
TranscriptionResponseFormatText = provider.TranscriptionResponseFormatText
|
|
TranscriptionResponseFormatSRT = provider.TranscriptionResponseFormatSRT
|
|
TranscriptionResponseFormatVTT = provider.TranscriptionResponseFormatVTT
|
|
)
|
|
|
|
// TranscriptionTimestampGranularity defines the requested timestamp detail.
|
|
type TranscriptionTimestampGranularity = provider.TranscriptionTimestampGranularity
|
|
|
|
const (
|
|
TranscriptionTimestampGranularityWord = provider.TranscriptionTimestampGranularityWord
|
|
TranscriptionTimestampGranularitySegment = provider.TranscriptionTimestampGranularitySegment
|
|
)
|
|
|
|
// TranscriptionOptions configures transcription behavior.
|
|
type TranscriptionOptions = provider.TranscriptionOptions
|
|
|
|
// Transcription captures a normalized transcription result.
|
|
type Transcription = provider.Transcription
|
|
|
|
// TranscriptionSegment provides a coarse time-sliced transcription segment.
|
|
type TranscriptionSegment = provider.TranscriptionSegment
|
|
|
|
// TranscriptionWord provides a word-level timestamp.
|
|
type TranscriptionWord = provider.TranscriptionWord
|
|
|
|
// TranscriptionTokenLogprob captures token-level log probability details.
|
|
type TranscriptionTokenLogprob = provider.TranscriptionTokenLogprob
|
|
|
|
// TranscriptionUsage captures token or duration usage details.
|
|
type TranscriptionUsage = provider.TranscriptionUsage
|
|
|
|
// TranscribeFile converts an audio file to WAV (via ffmpeg) and transcribes it.
|
|
func TranscribeFile(ctx context.Context, filename string, transcriber Transcriber, opts TranscriptionOptions) (Transcription, error) {
|
|
if transcriber == nil {
|
|
return Transcription{}, fmt.Errorf("transcriber is nil")
|
|
}
|
|
|
|
wav, err := audioFileToWav(ctx, filename)
|
|
if err != nil {
|
|
return Transcription{}, err
|
|
}
|
|
|
|
return transcriber.Transcribe(ctx, wav, opts)
|
|
}
|
|
|
|
func audioFileToWav(ctx context.Context, filename string) ([]byte, error) {
|
|
if filename == "" {
|
|
return nil, fmt.Errorf("filename is empty")
|
|
}
|
|
|
|
if strings.EqualFold(filepath.Ext(filename), ".wav") {
|
|
data, err := os.ReadFile(filename)
|
|
if err != nil {
|
|
return nil, fmt.Errorf("read wav file: %w", err)
|
|
}
|
|
return data, nil
|
|
}
|
|
|
|
tempFile, err := os.CreateTemp("", "go-llm-audio-*.wav")
|
|
if err != nil {
|
|
return nil, fmt.Errorf("create temp wav file: %w", err)
|
|
}
|
|
tempPath := tempFile.Name()
|
|
_ = tempFile.Close()
|
|
defer os.Remove(tempPath)
|
|
|
|
cmd := exec.CommandContext(ctx, "ffmpeg", "-hide_banner", "-loglevel", "error", "-y", "-i", filename, "-vn", "-f", "wav", tempPath)
|
|
if output, err := cmd.CombinedOutput(); err != nil {
|
|
return nil, fmt.Errorf("ffmpeg convert failed: %w (output: %s)", err, strings.TrimSpace(string(output)))
|
|
}
|
|
|
|
data, err := os.ReadFile(tempPath)
|
|
if err != nil {
|
|
return nil, fmt.Errorf("read converted wav file: %w", err)
|
|
}
|
|
|
|
return data, nil
|
|
}
|