Gerben007/Afrikaans-Transcriber

# Afrikaans Transcription Service A self-hosted Afrikaans audio transcription service built with faster-whisper, FastAPI, Celery, and Docker Compose. ## Architecture | Service | Role | |---------|------| | **FastAPI** | REST API + Web UI | | **Celery** | Task queue (transcription worker) | | **faster-whisper** | ASR inference engine (Systran/faster-whisper) | | **Redis** | Celery message broker | | **PostgreSQL** | Jobs database | | **MinIO** | S3-compatible file storage | | **Caddy** | Reverse proxy with automatic HTTPS (production) | ## Prerequisites - Docker Engine 20.10+ and Docker Compose v2 - ~6 GB disk space for the faster-whisper model - (Optional) NVIDIA GPU + NVIDIA Container Toolkit for GPU acceleration ## Quick Start ### 1. Clone and configure ```bash git clone github.com cd Afrikaans-Transcriber cp .env.example .env # Edit .env with your passwords and settings ``` ### 2. Download the whisper model ```bash pip install huggingface-hub huggingface-cli download Systran/faster-whisper-large-v3 \ --local-dir /opt/models/faster-whisper-large-v3 ``` Make sure `HOST_MODEL_PATH` in `.env` points to this directory. ### 3. Bring the stack up (development mode) ```bash docker compose up -d --build ``` This starts all services except Caddy (dev mode). The `minio-init` service automatically creates the required buckets (`audio-inbox` and `transcripts-out`). ### 4. Verify services are running ```bash # Check all containers docker compose ps # Health check curl localhost # Open the Web UI # Navigate to localhost in your browser ``` ### 5. Create MinIO buckets manually (if needed) The `minio-init` service handles this automatically. To do it manually: ```bash # Using docker docker compose run --rm minio-init # Or using mc locally mc alias set local localhost minioadmin CHANGE_ME_minio_secret mc mb --ignore-existing local/audio-inbox mc mb --ignore-existing local/t …