feat: add Groq Whisper API support for transcription and translation

Add support for Groq Whisper API as an alternative to OpenAI:
- groq-transcription: Fast transcription using whisper-large-v3 or whisper-large-v3-turbo
- groq-translation: Translation to English (whisper-large-v3 only)

Features:
- New transcription adapters for both Groq services
- Config validation with provider-specific model restrictions
- API key support via config file or GROQ_API_KEY environment variable
- Updated interactive configuration wizard with provider selection
- Comprehensive test coverage for all providers
- Shared WAV conversion utility for audio preprocessing

Breaking changes: None
Backward compatibility: Existing OpenAI configurations continue to work unchanged
This commit is contained in:
Ben Nasedkin
2025-10-18 19:19:03 +03:00
parent 937c15e4de
commit 025ff7fc55
10 changed files with 676 additions and 58 deletions
+1 -38
View File
@@ -3,7 +3,6 @@ package transcriber
import (
"bytes"
"context"
"encoding/binary"
"fmt"
"log"
"time"
@@ -31,7 +30,7 @@ func (a *OpenAIAdapter) Transcribe(ctx context.Context, audioData []byte) (strin
}
// Convert raw PCM to WAV format
wavData, err := a.convertToWAV(audioData)
wavData, err := convertToWAV(audioData)
if err != nil {
return "", fmt.Errorf("convert to WAV: %w", err)
}
@@ -56,39 +55,3 @@ func (a *OpenAIAdapter) Transcribe(ctx context.Context, audioData []byte) (strin
log.Printf("openai-adapter: transcribed %d bytes in %v: %q", len(audioData), duration, resp.Text)
return resp.Text, nil
}
// convertToWAV converts raw 16-bit PCM audio to WAV format
func (a *OpenAIAdapter) convertToWAV(rawAudio []byte) ([]byte, error) {
var buf bytes.Buffer
const sampleRate = 16000
const channels = 1
const bitsPerSample = 16
const byteRate = sampleRate * channels * bitsPerSample / 8
const blockAlign = channels * bitsPerSample / 8
dataSize := len(rawAudio)
fileSize := 36 + dataSize
// WAV header
buf.WriteString("RIFF")
binary.Write(&buf, binary.LittleEndian, uint32(fileSize))
buf.WriteString("WAVE")
// fmt chunk
buf.WriteString("fmt ")
binary.Write(&buf, binary.LittleEndian, uint32(16)) // fmt chunk size
binary.Write(&buf, binary.LittleEndian, uint16(1)) // PCM format
binary.Write(&buf, binary.LittleEndian, uint16(channels)) // number of channels
binary.Write(&buf, binary.LittleEndian, uint32(sampleRate)) // sample rate
binary.Write(&buf, binary.LittleEndian, uint32(byteRate)) // byte rate
binary.Write(&buf, binary.LittleEndian, uint16(blockAlign)) // block align
binary.Write(&buf, binary.LittleEndian, uint16(bitsPerSample)) // bits per sample
// data chunk
buf.WriteString("data")
binary.Write(&buf, binary.LittleEndian, uint32(dataSize))
buf.Write(rawAudio)
return buf.Bytes(), nil
}