♟ Hikaru Chess Bot

A simple chess-playing AI bot built with TensorFlow. The project includes a Jupyter Notebook for model training and a Flask API to play against the trained bot.

The original training was done on Hikaru Nakamura games. Due to a lack of resources, the model was trained with only 2000 games.

📁 Project Structure

.
├── hikaru.ipynb # Jupyter Notebook for training the model
├── api.py # Flask API to interact with the bot
├── model/ # Folder to store trained model files
├── requirements.txt # Dependencies
└── README.md # You're here!

🚀 Features

• Train a chess-playing model using historical game data (e.g., from Hikaru Nakamura).
• Expose the bot via a simple REST API using Flask.
• Play moves and respond with the AI's move.

🧠 Training the Model

Open the Jupyter Notebook:

jupyter notebook hikaru.ipynb

Make sure to have your dataset ready and follow the notebook steps to:

• Load and preprocess the data.

• Build a model with TensorFlow.

• Train and evaluate the model.

• Save the trained model to the model/ directory.

🌐 Running the Flask API

After training and saving the model:

python api.py

The server will start at http://127.0.0.1:8000.

API Endpoint

POST /bestmove

Request JSON:

{
  "pgn": "{Game PGN as string}"
}

Response JSON:

{
  "move": "e2e4"
}

📦 Installation

Install dependencies:

pip install -r requirements.txt

Make sure you have Python 3.7 or above.

📁 Requirements

Python 3.7+

TensorFlow

python-chess

Flask

flask-cors