scribe-worker

Worker for the Sunet transcription service (Sunet Scribe).

Author

This project is developed by Sunet. Contributor: Kristofer Hallin.

License

This project is licensed under the Apache License, Version 2.0. See LICENSE for details.

Copyright (c) 2025-2026 Sunet. Contributor: Kristofer Hallin.

Contributing

Contributions are welcome! Please feel free to open issues or submit pull requests.

Features

• Transcription Processing: Processes audio/video transcription jobs from the backend queue
• Whisper.cpp Integration: Uses whisper.cpp for efficient local transcription
• Multiple Output Formats: Generates JSON and SRT transcription outputs
• Multi-worker Support: Run multiple workers in parallel for increased throughput

Requirements

• Python 3.13+
• uv (recommended package manager)
• whisper.cpp (must be built separately)
• FFmpeg (for audio/video processing)

Development Environment Setup

1. Clone and Install Dependencies

git clone <repository-url>
cd scribe-worker
uv sync

2. Build whisper.cpp

Build and install whisper.cpp from source. See https://github.com/ggml-org/whisper.cpp for detailed instructions.

3. Download Whisper Models

./download_models.sh

4. Configure Environment Variables

Create a .env file in the project root with the following settings:

# Debug mode
DEBUG=True

# Backend API configuration
API_BACKEND_URL="http://localhost:8000"
API_VERSION="v1"

# Worker configuration
WORKERS=2
WHISPER_CPP_PATH=<Path to whisper.cpp>
FILE_STORAGE_DIR=<Your file storage directory>

5. Run the Worker

uv run main.py --foreground --debug

Docker

Build and run with Docker:

docker build -t scribe-worker .
docker run --env-file .env scribe-worker

Project Structure

scribe-worker/
├── main.py              # Worker entry point
├── utils/               # Utility modules
├── models/              # Whisper model files
└── downloaded/          # Downloaded files for processing