README.md

metadata

title: Agent Arena
emoji: 🌍
colorFrom: green
colorTo: gray
sdk: docker
short_description: LLM-powered agents compete in an automated DeFi market

Agent Arena

Multi-agent LLM simulation in DeFi markets with emergent strategic arms races.

Overview

Agent Arena is a simulation where AI agents powered by MiniMax-M2.1 compete in an automated DeFi market. Agents make strategic trading decisions, form alliances, and their behaviors evolve over time.

Features

• AI agents powered by MiniMax-M2.1 with reasoning transparency
• Constant product AMM pool mechanics (like Uniswap)
• Real-time metrics including Gini coefficient, cooperation rates, and pool stability
• Strategic decision making with thinking traces
• Persistent storage with Supabase

Architecture

defi-agents/
├── api/              # API clients (MiniMax, Supabase)
├── core/             # Core simulation (Agent, Pool, Simulation, Analyzer)
├── web/              # FastAPI backend
├── frontend/         # React dashboard (Vite + Tailwind)
├── scripts/          # Database schema
└── config.py         # Configuration

Getting Started

Prerequisites

• Python 3.11+
• uv (Python package manager)
• MiniMax API key
• Supabase project

Installation

# Clone the repository
git clone https://github.com/nice-bills/agent-arena.git
cd agent-arena

# Install dependencies
uv sync

# Set up environment variables
cp .env.example .env
# Edit .env with your API keys

Running Locally

# Start the backend
uv run python web/app.py

# In another terminal, start the frontend
cd frontend
npm install
npm run dev

API Endpoints

Method	Endpoint	Description
GET	/health	Health check
POST	/api/runs	Start a new simulation run
GET	/api/runs	List all runs
GET	/api/runs/{id}	Get run details
GET	/api/analysis/trends	Get trend analysis
GET	/api/thinking/{action_id}	Get thinking trace

Environment Variables

Variable	Description
MINIMAX_API_KEY	MiniMax API key
SUPABASE_URL	Supabase project URL
SUPABASE_KEY	Supabase anon key
NUM_AGENTS	Number of agents per run (default: 5)
TURNS_PER_RUN	Turns per simulation (default: 10)

License

MIT