Agent Evolve

A comprehensive toolkit for evolving and tracking AI agents through evolutionary programming and automated optimization.

Features

• 🧬 Evolutionary Optimization: Automatically improve your AI agents using OpenEvolve
• 📊 Interactive Dashboard: Real-time tracking and visualization of evolution progress
• 🎯 Smart Extraction: Automatically find and extract functions marked for evolution
• 🔧 Auto-Generation: Generate evaluators, training data, and configurations automatically
• 📈 Metrics Tracking: Comprehensive metrics collection and analysis
• 🚀 CLI Interface: Easy-to-use command-line tools for all operations

Quick Start

Installation

pip install agent-evolve

Basic Usage

1. Mark functions for evolution by adding the @evolve decorator:

from agent_evolve import evolve

@evolve()
def my_function(x: int) -> str:
    return f"Hello {x}"

2. Extract evolution targets:

agent-evolve extract /path/to/your/code

3. Run the evolution pipeline:

agent-evolve pipeline my_function

4. View results in the dashboard:

agent-evolve dashboard

CLI Commands

• extract - Extract functions marked with @evolve decorator
• generate-training-data - Generate training data for extracted functions
• generate-evaluators - Generate evaluation functions
• generate-configs - Generate OpenEvolve configuration files
• evolve - Run evolution optimization on a specific tool
• pipeline - Run the complete pipeline (training data → evaluator → config → evolution)
• dashboard - Launch the interactive Streamlit dashboard
• analyze - Analyze tracking data

Dashboard Features

• Evolution Pipeline: Step-by-step progress tracking
• Code Comparison: Side-by-side diffs of original vs evolved code
• Metrics & Timeline: Performance improvements over time
• Interactive Tools: Generate evaluators, run evolution, and more

Requirements

• Python 3.8+
• OpenAI API key (for evaluator generation)
• OpenEvolve package

License

MIT License - see LICENSE file for details.

Contributing

Contributions welcome! Please read our contributing guidelines and submit pull requests.