micro-agi

It is TypeScript framework designed to facilitate the integration of autonomous AI agents with open-source Large Language Models (LLMs). Drawing inspiration from AutoGPT and BabyAGI, this framework is developed to support both browser and server applications.

Overview

micro-agi aims to provide a straightforward and effective solution for developers looking to incorporate AI functionalities into their applications. It is built with a focus on compatibility and ease of use, catering to a wide range of development needs.

Key Features

Integration with Open Source LLMs: Enables connection with various Large Language Models via LM Studio.
Browser and Server Compatibility: Designed to work in both web browsers and server-side via Node.js.
Local File Management via Web APIs: This feature allows the manipulation and retrieval of files directly from your computer, utilizing the latest Web File System Access APIs for enhanced interaction with the local file system.
Deployment and Operation of Additional GPT Agents: Enables the creation and execution of multiple GPT-based AI agents, offering flexibility in managing diverse AI functionalities within the same environment.
Automated Code Generation: Facilitates the automatic creation of code snippets or entire code blocks, streamlining the development process and reducing manual coding efforts.
Limited Duration Memory Capabilities: Possesses a short-term memory feature, enabling the AI to temporarily store and recall recent interactions or data, useful for context-aware processing.
Integrated Search via Duck Duck Go: Incorporates a search functionality using Duck Duck Go, which involves indirect retrieval of search results through a server proxy, ensuring privacy-focused and unbiased search results.
Non-persistent Web Browsing: Offers a stateless method for accessing web content, where visits to URLs are conducted without retaining any session information, done through server-proxied requests to ensure privacy and security.

Future Plans

Wait for LM Studio support Function Calls
Integrate Agent Protocol
Integrate e2b sandbox
Integrate e2b assistants API
Dockerize
Websocket server that streams events
State manager

Ideas

Ink CLI (?)
Integration with e2b for agent deployment (?)
BentoLM integration for LLM cloud deployment (?)
Web-browsing with Puppeteer (?)
Support local vector embeddings (memgpt?) (?)
Support interaction with OpenAI / other plugins (?)

How to

To install dependencies:

bun install

To run:

bun run index.ts

This project was created using bun init in bun v1.0.6. Bun is a fast all-in-one JavaScript runtime.

Credits

This project was inspired by: