This is an educational 'game' to teach people about machine learning in the context of a Tetris AI.

The Rust back end is both available as a CLI and as WASM library.

First, run cargo build --release to compile the code. The executable will then be available at ./target/release/tetris-ai.

The binary can be run like:

tetris-ai run|train [preset|criterion]

The available weight presets are: score and levels. The available training criteria are: score,levels and tetrisses.

A devenv is provided which creates a development environment with all the dependencies.

For developing the backend:

• Rust 1.68 or higher

For developing front end:

• wasm32-unknown-unknown target for Rust
• Node.js with the pnpm package manager

For running the tests:

• Python 3.12 with the numpy package

After cloning the repository, ensure the dependencies are installed:

devenv shell # or `direnv allow`, only if you are using devenv
pnpm wasm # bootstraps the backend
pnpm install # installs JavaScript dependencies

Now you can run pnpm dev which opens a live preview of the site which updates as you make changes.

To run the back end go inside the tetris-ai subdirectory and run cargo run.

Edit the JSON files in src/lib/translations/[lang].json.

Edit the configurations in src/lib/levels.ts.

1. Create a function that implements it insie tetris-ai/src/feature.rs
2. Add it to the FEATURE_LOOKUP array at the top of the file
3. Now the feature should be available for use. Optionally you can add the translations to the JSON translation files under feature.[name] where the name corresponds to the string you defined in the lookup table.

First install the dependencies as described in the Install section. Clone the repository and run the following commands:

pnpm wasm 
pnpm install --frozen-lockfile
export BASE_PATH=/tetris-ai # set to the path where the site will be deployed e.g. example.com/tetris-ai
pnpm build

The files will be build to the build directory which you can upload to your web host.

The fallback page is set to 404.html, this works well with GitHub pages. If it is different on your web host make sure to change it in svelte.config.ts. See also the instructions from SvelteKit.

An automatically build version is available from GitHub Releases, however note that this version has the /tetris-ai path built into it and only works if you deploy to a domain like example.com/tetris-ai.

This project is licensed under the MIT License.

The Rust code was originally based on a Python implementation. This repository still includes a simplified copy of the Python code used for testing. All credit goes to the original authors.