This project provides a parallel AI agent system for solving International Mathematical Olympiad (IMO) problems using Google's Gemini API. It's designed for researchers and enthusiasts exploring AI's capabilities in complex problem-solving domains, offering a structured approach to tackle challenging mathematical questions.

How It Works

The system comprises individual AI agents (agent.py) that leverage Google's Gemini 2.5 Pro model to reason through and solve IMO problems. A parallel execution manager (run_parallel.py) orchestrates multiple agents concurrently, increasing the probability of finding a solution. Agents follow a structured reasoning process, with the system identifying successful solutions by detecting specific output phrases.

Quick Start & Requirements

• Install: pip install requests
• Prerequisites: Python 3.7+, Google Gemini API key.
• Setup: Clone the repository, create a .env file with GOOGLE_API_KEY=your_api_key_here or set the environment variable.
• Usage: python agent.py problem.txt [options] for single agent, python run_parallel.py problem.txt [options] for parallel execution.
• Documentation: problems folder for problem file format.

Highlighted Details

• Utilizes Google's Gemini 2.5 Pro model for problem-solving.
• Supports parallel execution of multiple agents to enhance solution discovery.
• Allows customization of agent count, timeouts, and additional prompts.
• Logs detailed reasoning for each agent and provides a summary of parallel runs.

Maintenance & Community

The project is maintained by Lin Yang and Yichen Huang. Contributions via issues, feature requests, or pull requests are welcomed.

Licensing & Compatibility

MIT License. Free to copy, modify, and distribute with attribution. Compatible with commercial and closed-source use.

Limitations & Caveats

Success is dependent on problem difficulty and AI model capabilities; not all problems may be solvable. Users should be mindful of Google API rate limits and associated costs.