Universal Worker: a library and framework for individual AI assistants

Code and workflow experiments by Mark Watson.

This project is a work in progress.

Currently this project is available under the GNU AFFERO GENERAL PUBLIC LICENSE.

Technology Used for Prototyping

The code for this project is largely framework-free, that is, we don't build with abstraction rich frameworks like LangChain or LlamaIndex. Many of the ideas developed here are similar to examples in Mark's book LangChain and LlamaIndex Projects Lab Book: Hooking Large Language Models Up to the Real World but here we favor developing eveything from scratch. For example, we develop vector data stores from scratch using NumPy, write our own prompt frameworks, and write our own wrappers fro different LLMs and platform providers.

The prototypes for operating system automation are all implemented using AppleAScript automation - curently we only support macOS.

Tools used:

• LLM models: self-hosted using Ollama.
• Web search: using the Brave Search APIs.

We also support commercial LLM providers like

GitHub code repository

GitHub repository: https://github.com/mark-watson/universalworker.org