FastAPI-based server for uploading and validating neural network files was used by Stockfish chess engine. The server accepts files, validates their content, and stores them g-zip compressed on the server.

• Upload neural network files
• Validate file content using SHA-256 hash
• Save g-zipped network files
• Handle various error scenarios with appropriate HTTP status codes

• Python 3.13 or higher
• FastAPI
• Gunicorn + Uvicorn (for running the server)
• uv (for project management)

1. Clone the repository:

   git clone https://github.com/ppigazzini/net-server.git
   cd net-server

2. Sync the project with production dependencies:

   uv sync --group prod

Use this systemd unit file to run the server with a nginx proxy server:

[Unit]
Description=fastapi server for chess engine networks
After=network.target

[Service]
Type=simple
ExecStart=/home/<username>/net-server/.venv/bin/gunicorn main:app --timeout 120 --workers 4 --worker-class uvicorn.workers.UvicornWorker
Restart=on-failure
RestartSec=3
User=<username>
WorkingDirectory=/home/<username>/net-server/app

[Install]
WantedBy=multi-user.target