A FastAPI server for document processing using the marker-pdf library for production environments, inspired by the official Marker server script.
- Docker with NVIDIA GPU support
- NVIDIA drivers and CUDA installed on the host
- Docker Compose v2+
The project includes a compose.yml file for easy deployment:
docker compose up --detachThe following environment variables can be configured:
HOST: Host address to bind to (default: 0.0.0.0)PORT: Port to listen on (default: 8000)WORKERS: Number of worker processes (default: 1)WORKER_CONNECTIONS: Maximum number of connections for each worker (default: 1000)TIMEOUT: Worker timeout in seconds (default: 30)KEEP_ALIVE: Keep-alive timeout in seconds (default: 75)GRACEFUL_TIMEOUT: Graceful timeout in seconds (default: 75)LOG_LEVEL: Logging level (default: info)TORCH_DEVICE: Device to run marker on (default: cuda)GUNICORN_CMD_ARGS: Additional gunicorn command arguments (default: "")API_KEY: Secret key for API authentication. If not set, authentication is disabled.
Example in compose.yml:
services:
marker:
# other configuration...
environment:
- TORCH_DEVICE=cuda
- WORKERS=2
- LOG_LEVEL=debug
- TIMEOUT=60
- API_KEY=your_secret_api_key.
├── server/ # Main FastAPI application code
│ ├── main.py # Application entry point
│ ├── schemas.py # Pydantic data models
│ └── ...
├── scripts/ # Utility scripts
│ └── startup.sh # Container startup script
└── ...
Health check endpoint.
Process a document.
Parameters:
file: The PDF file to process (required)page_range: Page range to convert (e.g. "0,5-10,20")languages: Comma separated list of languages for OCRforce_ocr: Force OCR on all pages (default: false)paginate_output: Whether to paginate the output (default: false)output_format: Output format - "markdown", "json", or "html" (default: "markdown")use_llm: Use LLM to improve conversion accuracy (default: false)
Example:
curl -X POST "http://localhost:8000/parse" \
-H "accept: application/json" \
-H "Content-Type: multipart/form-data" \
-F "file=@document.pdf" \
-F "output_format=markdown" \
-F "force_ocr=false"Authentication:
If API_KEY is set, all requests must include the Bearer token:
curl -X POST "http://localhost:8000/parse" \
-H "accept: application/json" \
-H "Authorization: Bearer your_secret_api_key" \
-H "Content-Type: multipart/form-data" \
-F "file=@document.pdf"