My personal toolkit for exploring LLMs' chess playing abilities.
- Modular Player Framework: Implement various chess-playing agents, including LLM-based players and chess engines (Stockfish).
- Multi-LLM Support: Compatible with OpenAI (and OpenAI API compatible models, such as DeepSeek), Google Gemini, and SGLang models.
- Chess Game Management: Utilities for managing chess games, positions, and move validation. See the game examples notebook.
- Prompt Engineering: Customizable prompt templates for guiding LLMs in chess gameplay. See the prompt examples notebook.
- Game Logging: Comprehensive logging of games in PGN format.
- Calibration Options: Tools for calibrating LLM performance with different starting positions. See the gpt-3.5-turbo-instruct calibration notebook.
pip install -e '.[test]'
pre-commit install