Catan

front

• run ng serve
• to deploy:
  • from front dir
  • ng build --configuration production (builds with production settings)
  • from root dir
  • firebase login
  • firebase deploy

back

• from back dir
• run shuttle run
• deploy shuttle deploy

simulation

The project includes a CLI tool for simulating Catan games between AI players:

• Build: cargo build --bin simulate
• Run: cargo run --bin simulate -- [OPTIONS]
• Run (optimized): cargo run --release --bin simulate -- [OPTIONS]

Options

• -p, --players <CONFIG>: Player types (e.g., "MR" for MCTS vs Random)
  • R: Random player
  • M: MCTS player
• -n, --num_games <N>: Number of games to simulate (default: 1)
• -v, --verbose: Show detailed game logs

Examples

• MCTS vs Random (single game): cargo run --bin simulate -- -p MR
• Random vs MCTS (10 games): cargo run --bin simulate -- -p RM -n 10
• Random vs Random with logs: cargo run --bin simulate -- -p RR -v

Attribution

Inspired by bcollazo's Catanatron. Licensed under GPL-3.0.

TODO

• Build the greatest catan bot player of all time
• Implement DB functionality to track user count and game history
  • MCP integration
• Robber getting moved after first turn to wood 10 tile every time?

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-keyring_1.1-1_all.deb sudo dpkg -i cuda-keyring_1.1-1_all.deb sudo apt-get update sudo apt-get -y install cuda-toolkit-12-5

WSL2 GPU (CUDA) setup for Candle

Follow these steps inside your WSL2 Ubuntu terminal to enable CUDA for Candle.

1. Install CUDA toolkit in WSL2 (Ubuntu 24.04)

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt-get update
sudo apt-get -y install cuda-toolkit-12-5

2. Point canonical symlink to the installed version (12.5)

sudo ln -sfn /usr/local/cuda-12.5 /usr/local/cuda

3. Add CUDA to PATH and runtime libraries (make persistent)

echo 'export PATH=/usr/local/cuda/bin:$PATH' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/lib/wsl/lib:$LD_LIBRARY_PATH' >> ~/.bashrc
source ~/.bashrc

4. Verify GPU visibility in WSL

nvidia-smi
which nvcc && nvcc --version
ls -l /usr/lib/wsl/lib/libcuda.so.1

5. Enable Candle CUDA features (already done in this repo)

# back/Cargo.toml
[dependencies]
candle-core = { version = "0.9.1", features = ["cuda"] }
candle-nn   = { version = "0.9.1", features = ["cuda"] }

6. Rebuild and run

cargo clean
cargo run --bin train

Expected logs:

• [device] using CUDA:0
• [infer] device=Device(Cuda(0))

Troubleshooting

• Do not set CUDARC_CUDA_VERSION.

  • If you previously exported it, unset and rebuild so cudarc can auto-detect from nvcc:

    unset CUDARC_CUDA_VERSION
    cargo clean && cargo run --bin train

• Still seeing CPU device in logs

  • Confirm WSL2 usage and nvidia-smi works inside WSL.
  • Ensure /usr/lib/wsl/lib/libcuda.so.1 exists and LD_LIBRARY_PATH includes /usr/lib/wsl/lib.
  • Re-run with a fresh shell or source ~/.bashrc.

• Build is slow while testing

  • Use a short training run: SELF_PLAY_GAMES=1 (default here).
  • Try optimized build: cargo run --release --bin train.