L♾️pGPT is a modular, Python-native framework for building AI agents, designed as a more extensible and efficient alternative to Auto-GPT. It targets developers and researchers seeking to create custom AI agents with minimal prompt overhead and enhanced control, offering features like human-in-the-loop correction and flexible tool integration.

How It Works

L♾️pGPT is built around a core Agent class that orchestrates task execution using large language models (LLMs). Its modular design allows developers to easily add custom tools by inheriting from BaseTool, enabling agents to interact with external services or perform specific actions. The framework prioritizes efficient token usage and supports state serialization for resuming agent progress, reducing reliance on external databases.

Quick Start & Requirements

• Install via pip: pip install loopgpt
• Requires Python 3.8+ and an OpenAI API Key.
• Optional: Google API Key and Custom Search Engine ID for Google Search integration; otherwise, DuckDuckGo is used.
• Setup involves setting the OPENAI_API_KEY environment variable or via a .env file.
• Example usage: Create an Agent instance, define its name, description, and goals, then run agent.cli().
• Docker images are available for development and execution.

Highlighted Details

• "Plug N Play" API for easy integration of new features and custom agent capabilities via Python code.
• Optimized for GPT-3.5, providing better results than Auto-GPT for users without GPT-4 access.
• Human-in-the-loop "course correction" allows users to guide agents when they deviate.
• Full state serialization supports saving and loading agent progress, memory, and tool states without external databases.

Maintenance & Community

• The project actively seeks contributions via issues and pull requests.
• A Discord server is available for community support and discussion.

Licensing & Compatibility

• The project appears to be licensed under the MIT License, allowing for broad commercial and closed-source use.

Limitations & Caveats

• Continuous mode (continuous=True) carries a risk of infinite loops.
• Official Google Search functionality requires specific API key setup; fallback to DuckDuckGo may limit search capabilities.