Skip to content

About

Autonomous 3D AI agent inspired by Caine from The Amazing Digital Circus, featuring perception, memory, and LLM-driven behavior.

Topics

Resources

Stars

3 stars

Watchers

1 watching

Forks

Repository files navigation

Caine World

An experimental AI recreation of Caine from The Amazing Digital Circus.

Project Status

Caine World (Three.js version) is now considered an experimental prototype.

During development, I reached several technical limitations related to procedural locomotion and inverse kinematics.

The current Three.js architecture reached several limitations regarding:

  • inverse kinematics (IK),
  • physical foot-ground interaction,
  • procedural locomotion systems,
  • stable rig resting pose handling.

Because of this, the current procedural walking system became increasingly unstable and difficult to evolve realistically.

Instead of forcing complex workarounds, the project is now moving toward a new architecture based on Babylon.js and Node.js, with a stronger focus on:

  • procedural animation,
  • physical interaction,
  • autonomous behavior systems,
  • scalable AI experimentation.

This repository remains available as an archive of the original experimental prototype.

A new repository for the Babylon.js version is currently in development.

Features

  • Autonomous movement
  • Procedural body animation
  • Object creation and stacking
  • Episodic memory
  • Vision system
  • Local LLM reasoning with Ollama
  • Dynamic exploration

Requirements

  • Node.js
  • Ollama
  • VS Code
  • Live Server extension
  • Prettier extension

Setup

Install Ollama:

https://ollama.com

Install the model:

ollama pull llama3.2

Install dependencies:

npm install

Launch

Run:

lancer-caine.bat

Then launch Live Server on index.html.

Technologies

  • Three.js
  • JavaScript
  • Ollama
  • Local LLMs

Disclaimer

Fan project inspired by The Amazing Digital Circus.

About

Autonomous 3D AI agent inspired by Caine from The Amazing Digital Circus, featuring perception, memory, and LLM-driven behavior.

Topics

Resources

Stars

3 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages