AI plays Emulators

(For now only Nintendo 64)

This is a ready-to-use AI environment that runs on your browser. Simply connect with

Running it from Web

https://ai-plays-emulators.vercel.app and set your Vitrus API key and your world ID

Creating a player

bun add vitrus 

Running it locally

• emulator-environment/: Contains the React frontend application using EmulatorJS.
• ai-player/: Contains an example of an AI agent.

Getting Started in 3 steps

1. Install dependencies for the entire monorepo:

   cd ./emulator-environment
   bun install

2. Setup Environment Variables To help developers focus only in the AI player development, we abstracted the ai-to-emulator communication using Vitrus Agent-World-Actor (AwA) infrastructure.

   🔑 You can get your API Keys here.

   cd ./emulator-environment
   touch .env

   And inside of the .env file define your credentials and API

   VITE_VITRUS_API_KEY = vitrus-<api-key>
   VITE_VITRUS_WORLD = <world-id>
   ROM_URL = https://path/your_file_here.n64 #Your game of preference

   Alternativelly you may fork this repo and use a custom AwA layer like Redis, PeerJs, etc. But we've made Vitrus SDK to simplify this.

3. Running the Frontend (Vite Development Server): Navigate to the emulator-environment directory and run:

   cd ./emulator-environment
   bun run dev

   This will typically start the frontend on http://localhost:5173. You can defined the port by adding --port 3000.

4. (optional) Running the example Player (typescript)

   cd ./ai-player
   bun install
   touch .env

   Add the same API Keys on .env but without the VITE_ prefix.

   VITRUS_API_KEY = vitrus-<api-key>
   VITRUS_WORLD = <world-id> #Make sure it's the same world as the emulator

   With the browser running, run the player

   bun dev

   It should play now!! 🎉

Gamepad Buttons map