Special thanks to xpilot and Enzyme of the Slippi Discord for the help they gave me in setting this up.

In this guide, we’ll get example.py working of xpilot's fork of LibMelee. xpilot actively maintains this version of LibMelee and it works with the latest versions of Slippi. This is the ideal build to use for machine learning.

To begin with, make a folder called MeleeBot.

Making a virtual environment

1. Run python -m venv venv
2. Run source venv/bin/activate

There should now be a folder called venv inside MeleeBot. Your bash should also have (venv) in front of the computer and folder name, e.g. (venv) premelon@premelon_pc %

The file structure right now:

MeleeBot
├─ venv

Getting the right AppImages

You will need a Melee ISO. This guide will not cover how to procure one. You can put this in the root of the MeleeBot directory.

You will want two different AppImages.

• Headless: https://github.com/vladfi1/slippi-Ishiiruka/releases/download/exi-ai-0.1.0/Slippi_Online-x86_64-ExiAI.AppImage
• NoGUI: https://github.com/vladfi1/dolphin/releases/tag/slippi-nogui-v0.1.0

You can put each in a folder called “AppImages”. Your file structure should look like this (AppImages renamed for convenience):

MeleeBot
├─ melee.iso
├─ venv
├─ AppImages
   ├─ headless.AppImage
   ├─ nogui.AppImage