Lex: An Autonomous AI Assistant 🤖

Lex is a multi-agent, autonomous desktop assistant built with a sophisticated agentic architecture. It's designed to go beyond simple conversation, enabling complex task execution, self-correction, and learning from past interactions through a persistent memory system.

---

🚀 Live Demo

How it Works :

Step - 1 : Load app.py

Step - 2 : Load gui.py

Step 3 : Activate Hot-Keys ['l' , 'e' , 'x'] to Activate the Multi Agent.

---

✨ Key Features

Lex isn't just a chatbot; it's a powerful agent with a unique set of capabilities:

• 🧠 Multi-Agent Architecture: A central "router" agent analyzes user prompts and delegates tasks to specialized agents—a conversational agent for Q&A and a coding agent for autonomous task execution on the host system.
• 🛠️ Self-Correction Loop: Lex can debug itself! A "Child" agent analyzes the code execution output for errors. If an error is found, it provides iterative feedback to the coding agent, allowing it to fix its own code and re-attempt the task.
• 📚 RAG-Based Memory: Equipped with a FAISS vector store, Lex has long-term memory. It can recall past interactions, learn user preferences, and provide contextually aware responses, becoming more personalized over time.
• 🖥️ Full-Stack Desktop Application: Lex is a complete application with a Flask API backend serving the agent's logic and a responsive, non-blocking PyQt6 GUI for a polished user experience.
• ⚡ Hotkey Activated: The GUI can be summoned and dismissed instantly with a global hotkey (l e x), making it a seamless part of any workflow.

---

🛠️ Tech Stack

This project leverages a modern, powerful stack to achieve its agentic capabilities:

• AI Core: LangGraph, LangChain
• LLM: Groq (Llama3-70b)
• Memory: FAISS (Vector Store), SerpAPI (Web Search)
• Backend: Python, Flask
• GUI: PyQt6
• Tools: Git, GitHub, VS Code

---

🏛️ Architecture Overview

Lex operates on a clear, modular architecture:

1. GUI (gui.py): The user-facing interface, built with PyQt6. It captures user input and displays results in a non-blocking manner by running API calls in a separate thread.
2. API Server (app.py): A lightweight Flask server that exposes the agent's functionality via a simple /invoke endpoint. It acts as the bridge between the GUI and the agent's brain.
3. Agent Brain (Lex_Lang_Graph.py): The core of the application. It contains the LangGraph state machine, all the agent definitions (router, coder, critic), the memory system, and the tool integrations.

graph TD
    A[User via PyQt6 GUI] -->|HTTP Request| B(Flask API Server);
    B -->|Invokes Agent| C{LangGraph State Machine};
    C --> D{Router Agent};
    D -->|Casual Task| E[Conversational Agent];
    D -->|Coding Task| F[Coding Agent];
    F -->|Executes Code| G[Python REPL Tool];
    G -->|Output w/ Error| H(Critic Agent);
    H -->|Feedback| F;
    G -->|Success| I[Memory Node];
    E --> I;
    I -->|Saves to FAISS| J[(Vector Store)];
    I -->|Final State| B;
    B -->|HTTP Response| A;

---

🚀 Getting Started

To get a local copy up and running, follow these simple steps.

Prerequisites

• Python 3.10+
• An API key from Groq
• An API key from SerpAPI

Installation

1. Clone the repo sh git clone https://github.com/akshayabalan/lex.git cd lex-agent
2. Install Python packages sh pip install -r requirements.txt
3. Set up your environment variables
   • Create a file named .env in the root directory.
   • Add your API keys to this file: env GROQ_API_KEY="your_groq_api_key" SERPAPI_API_KEY="your_serpapi_api_key"
4. Run the application
   • First, start the agent server in one terminal: sh python app.py
   • Then, start the GUI in a separate terminal: sh python gui.py
5. Activate Lex
   • Type the hotkey word lex anywhere on your computer to summon the GUI.

---

📞 Contact

Akshaya Balan - [[email protected]]

Project Link: [https://github.com/akshayabalan/lex]

Built With

• python