WarAgent is an LLM-powered multi-agent system designed to simulate historical international conflicts, including World War I, World War II, and the Warring States Period. It aims to provide data-driven, AI-augmented insights into the triggers, conditions, and consequences of war, offering a novel perspective for conflict resolution and peacekeeping strategies. The system is targeted at researchers, policymakers, and anyone interested in understanding complex human behaviors and historical conflicts through AI.

How It Works

WarAgent utilizes a multi-agent architecture where each country is represented by a "Country Agent." These agents interact with each other and with a "Secretary Agent" that verifies the appropriateness and logical consistency of their actions. A "Board" component manages international relationships, while a "Stick" component serves as an internal record-keeping system for each country's domestic statutes. This approach leverages the advanced reasoning capabilities of LLMs to simulate nuanced decision-making and emergent interactions in complex geopolitical scenarios.

Quick Start & Requirements

• Installation: Requires Python 3.9, conda, and git. Install via pip install -r requirements.txt after cloning the repository.
• Dependencies: Supports OpenAI's GPT-4 and Anthropic's Claude-2. Requires setting OPENAI_API_KEY or CLAUDE_API_KEY environment variables.
• Running: Execute simulations using python main.py --model <model_choice> --scenario <scenario_name>.
• Demo: A video demo is available via a provided link.

Highlighted Details

• Simulates historical conflicts like WWI, WWII, and the Warring States Period.
• Employs a multi-agent system with dedicated Secretary Agents for action verification.
• Supports advanced LLMs like GPT-4 and Claude-2 for nuanced simulation.
• Offers insights into war triggers and conditions through emergent agent interactions.

Maintenance & Community

The initial version was released on November 28, 2023, including source code, data, and evaluation metrics. Further community or roadmap information is not detailed in the README.

Licensing & Compatibility

The source code is licensed under Apache 2.0, intended solely for research use.

Limitations & Caveats

The system's primary intended purpose is research. Compatibility for commercial use or integration into closed-source projects is not specified.