Generate FEN strings given a chess board image.

FENGEN handles the image uploads, FEN-to-board state translation, and communication with the process running predict.py.

The predict.py script is ran by a GenServer worker/process, which is "replicated", and managed by poolboy. So task distribution is handled by it. The results are flushed to stdout, which notifies the port that there's a prediction that has been made.

Container image is used to deploy to DO. Dependencies are handled by nix. There are better ways to shrink the image size (erlang & tensorflow are rather large), but this will have to do for now.

Ensure you have the nix package manager. If you're using NixOS then you can skip this.

curl -L https://nixos.org/nix/install | sh

Activate the nix environment to fetch necessary dependencies.

# If you don't need the libraries/packages from the external environment
nix-shell --pure

# Otherwise
nix-shell

If you want to open your IDE/text editor within the environment, you can do so within it.

# If you're using VSCode. Unfortunately, you'll need more 
# configuration if you've installed it through Flatpak.
code .

Setup the remaining things.

mix setup

Run the server

mix phx.server

Now you can visit localhost:4000 from your browser.

FENGEN's code is licensed under MIT.

• Pavel Koryakin for the dataset https://www.kaggle.com/koryakinp/chess-positions
• Colin M.L. Burnett for the chess set assets. (LICENSED UNDER GPLv2+)