Butler.AI

Butler.AI is a local fully uncensored AI assistant, features include :

• Customizable Avatars
• Poses
• Talking capabilities
• Fully Local (You are always in control)
• Uncensored (Free to use abliterated models or jb-prompts)

Demo

Prerequisites

• Node.js 20+ (recommended) for running locally
• Docker Desktop (optional) for containerized runs
• Dedicated CUDA enabled GPU for faster inference (optional, you need to uncomment gpu capabilities in docker compose file)

Models and Actions

• Models are generated via https://readyplayer.me/ or glb files also supported as long as its rigged. Just place them in the static/avatars and update the SettingsPanel.svelte
• Actions are plain Mixamo Actions, same drill to add more.

Run locally (recommended for development)

cd Ui
npm install
npm run dev

Open http://localhost:4173

Run with Docker

UI only

docker compose --profile ui up

Useful if you want to run ollama externally, openai compatible APIs are supported, change the OLLAMA_BASE_URL env variable accordingly.

UI + Ollama (optional):

docker compose --profile full up

Ollama will be reachable at http://localhost:11434.

(You must have models downloaded manually in ollama)

docker exec butlerai-ollama-1 ollama pull llama3.2:3b

Development build (local)

SvelteKit can be built and previewed locally:

cd Ui
npm run build
npm run dev

Notes

• The UI talks to Ollama using docker. If you run Ollama elsewhere, adjust your environment accordingly in docker-compose.yml the env variable OLLAMA_BASE_URL.

• If you face wasm errors, just reload the page.

Acknowledgements

This project would not have been possible without Talking Head & HeadTTS