Team project · Python / Software Engineering
Chess broadcasting
A live chess backend driven by recognised board positions.
- FastAPI
- Redis
- python-chess
- Python
- PostgreSQL
- SQLAlchemy
- Alembic
- WebSocket
- Docker
Project and contribution
Developed recognition-result ingestion, move reconstruction, game state and WebSocket updates. Worked on transactions, Redis queues and locks, PGN export and Lichess Broadcast integration.
Purpose
The team solution turns a game on a physical chessboard into a digital broadcast. Computer vision recognises the board position; the backend converts results into game state, moves and spectator updates.
My work
- Developed the API, models, migrations and game lifecycle.
- Processed recognised positions and checked possible moves with python-chess.
- Worked on stabilising recognition results and recovering move sequences.
- Used Redis for queues, locking and duplicate-processing protection.
- Connected game changes to database transactions and WebSocket updates.
- Added PGN generation, Lichess Broadcast synchronisation and a simulated recognition stream for integration checks.
Technical approach
FastAPI receives data, PostgreSQL stores games, and python-chess validates positions and moves. Queues and locks coordinate incoming events; clients receive current state through WebSocket.
My work covered the backend and computer-vision result integration. Board recognition was a separate part of the team solution.