poke-env

Why we included this project

Poke-env turns Pokemon Showdown into a Python training ground for reinforcement learning, and it goes well beyond the usual gridworlds: the game is turn-based, has a large action space, and gives you a clear win or loss at the end. You work with a Python API built around players, battles, Pokemon, and moves, with a Farama Gymnasium interface that lets you reuse existing RL code instead of rewriting it. That suits scripted bots, self-play experiments, and coursework alike. A built-in RandomPlayer gets a first local battle running in a few lines, so you can confirm your setup works before training agents of your own.