♟️ Shadow-Code

Shadow-Code is a feature-rich chess game that supports both local AI battles and online multiplayer. Play against a Deep Reinforcement Learning (DRL) agent trained from scratch or connect with friends over a real-time socket-based network—all with an elegant Pygame-based interface and animated visuals.

---

🚀 Features

• ✅ Local Play vs AI
  Play against a trained Deep Q-Network (DQN)-based DRL agent.

• 🌐 Online Multiplayer
  Host or join real-time matches using TCP sockets.

• 🎞️ Smooth Animations
  Enjoy sliding piece animations for a polished experience.

• 👑 Promotion Interface
  Promote pawns with a clean graphical selection.

• 🔊 Sound Effects
  Custom sounds for moves, captures, and promotions.

• 📋 Move History Sidebar
  Track your game progress with a live move list.

---

🛠 Installation

1. Clone the repository

git clone https://github.com/MrTBK/Shadow-Code.git
cd Shadow-Code

2. Install dependencies

Install the required packages via pip:

pip install -r requirements.txt

Note: Requires Python 3.8+ and PyTorch.

---

🧠 Running the Game

To start the game:

python main.py

---

🎮 Modes of Play

🧑‍💻 Local Play (vs DRL)

• Select Local from the menu.
• Play against a Deep Q-Learning agent trained on legal chess states.

🌐 Online Multiplayer

Host a Game

• Select Online → Host.
• Wait for an opponent to connect.
• Game starts automatically after connection.

Join a Game

• Select Online → Join.
• Enter the IP address of the host machine.
• Connect and start playing.

---

📁 Project Structure

Shadow-Code/
├── assets/               # Piece images & sound files
├── dqn.py                # DQN agent definition and utilities
├── game.py               # Game logic and rendering
├── interface.py          # UI handling (board, promotion UI, sounds)
├── server.py             # Socket-based networking
├── main.py               # Entry point
├── README.md
└── requirements.txt

---

🧠 DRL Agent Training

Our AI uses a Deep Q-Network with a flat board tensor input. Training code and logs are included in the project for reproducibility. If no dqn_checkpoint.pt exists, training can be resumed automatically.

---