# Devin - Your Personal AI Assistant Welcome to Devin, your very own Jarvis-inspired AI assistant! This project aims to create a powerful, extensible AI system that can understand and respond to your questions while gradually building up capabilities through tools and worker agents. ## Project Overview Devin is designed to be a personal AI assistant that: - Connects to local LLMs for natural language understanding and generation - Responds to user questions in a conversational manner - Scales from simple question answering to complex multi-agent workflows - Provides a foundation for building sophisticated AI tools and systems ## Getting Started ### Prerequisites Before running Devin, you'll need to set up your local LLM environment. Refer to the [Configuration](#configuration) section below. ### Installation 1. Clone/Fork this repository 1. Use UV to set up requirements and environment: ```bash uv sync ``` ### Running Devin To start Devin, simply run: ```bash uv run main.py ``` This will initialize the AI assistant and begin listening for your questions. ## Configuration Devin uses a configuration file to connect to your local LLM. The configuration should be set up in `config.yaml` ```yaml llm: model_path: "path/to/your/local/model" api_base: "http://localhost:1234/v1" ``` ### Local LLM Setup Devin currently supports connecting to local LLMs such as: - [Ollama](https://ollama.com/) - [LM Studio](https://lmstudio.ai/) - [LocalAI](https://localai.io/) Make sure your chosen LLM is running and accessible before starting Devin. ## Roadmap ### Phase 1: Core Functionality - [ ] Basic LLM connection and response - [ ] Conversation history management - [ ] Enhanced response formatting ### Phase 2: Tool Integration - [ ] Web search capabilities - [ ] File system operations - [ ] Code execution environment ### Phase 3: Multi-Agent System - [ ] Worker LLM coordination - [ ] Task delegation and management - [ ] Complex workflow automation ## License TBC - non commercial / not for profit --- *Devin - Your Personal AI Assistant* *Powered by local LLM technology*