An open-source implementation of nof1.ai's Alpha Arena - A benchmark platform for evaluating AI models' cryptocurrency trading capabilities with real money in real markets.
Alpha Arena is a revolutionary benchmark that tests AI models where it matters most: real financial markets with real money. Unlike traditional AI benchmarks that rely on static datasets, Alpha Arena evaluates AI models by giving each one an initial capital of $10,000 and letting them trade cryptocurrency perpetual contracts on live markets.
Why markets are the ultimate test of intelligence:
- Markets are dynamic, adversarial, and open-ended
- They challenge AI in ways that static benchmarks cannot
- Real-time decision-making under uncertainty reveals true capabilities
- Risk management and strategic thinking are essential
This open-source implementation currently focuses on running the DeepSeek trading model with the following capabilities:
- 🔄 Real-time Trading: Automated cryptocurrency trading on Binance via CCXT
- 📊 Live Dashboard: Beautiful real-time charts showing account performance
- 🧠 AI Decision Making: Complete chain-of-thought reasoning for every trade
- 💹 Multi-Asset Support: Trade BTC, ETH, SOL, BNB, and DOGE
- 📈 Performance Tracking: Detailed metrics, trade history, and profit/loss tracking
- 🔍 Full Transparency: Every decision, prompt, and reasoning is logged and visible
- ⚡ Cron Jobs: Automated 20-second metric collection and 3-minute trading intervals
- Framework: Next.js 15 with App Router & Turbopack
- AI SDK: Vercel AI SDK with DeepSeek integration
- Database: PostgreSQL with Prisma ORM
- Trading: CCXT for exchange connectivity
- Charts: Recharts with shadcn/ui
- Styling: Tailwind CSS v4
- Runtime: Bun for fast package management
- Bun installed
- PostgreSQL database
- Binance API credentials (for live trading)
- DeepSeek API key
-
Clone the repository
git clone https://github.com/snowingfox/open-nof1.ai.git cd open-nof1.ai -
Install dependencies
bun install
-
Set up environment variables
cp .env.example .env
Fill in your
.envfile:# Application NEXT_PUBLIC_URL="http://localhost:3000" # Database DATABASE_URL="postgresql://postgres:password@localhost:5432/nof1" # AI Models DEEPSEEK_API_KEY="your_deepseek_api_key" OPENROUTER_API_KEY="your_openrouter_api_key" # Optional: for additional models # Market Research (Optional) EXA_API_KEY="your_exa_api_key" # For enhanced market analysis # Trading (Binance) BINANCE_API_KEY="your_binance_api_key" BINANCE_API_SECRET="your_binance_secret" BINANCE_USE_SANDBOX="true" # Set to "false" for live trading # Trading Configuration START_MONEY=10000 # Initial capital in USDT (e.g., 10000 = $10,000 USDT) # Cron Job Authentication CRON_SECRET_KEY="your_secret_token"
-
Set up the database
bunx prisma generate bunx prisma db push
-
Run the development server
bun dev
-
Set up cron jobs (for automated trading)
You'll need to set up external cron jobs or use a service like Vercel Cron to call these endpoints:
POST /api/cron/20-seconds-metrics-interval- Collect metrics every 20 secondsPOST /api/cron/3-minutes-run-interval- Execute trading logic every 3 minutes
Example crontab:
# Metrics collection (every 20 seconds) * * * * * curl -X POST http://localhost:3000/api/cron/20-seconds-metrics-interval -H "Authorization: Bearer YOUR_CRON_SECRET_KEY" * * * * * sleep 20 && curl -X POST http://localhost:3000/api/cron/20-seconds-metrics-interval -H "Authorization: Bearer YOUR_CRON_SECRET_KEY" * * * * * sleep 40 && curl -X POST http://localhost:3000/api/cron/20-seconds-metrics-interval -H "Authorization: Bearer YOUR_CRON_SECRET_KEY" # Trading execution (every 3 minutes) */3 * * * * curl -X POST http://localhost:3000/api/cron/3-minutes-run-interval -H "Authorization: Bearer YOUR_CRON_SECRET_KEY"
Open http://localhost:3000 to view the dashboard.
open-nof1.ai/
├── app/
│ ├── api/
│ │ ├── cron/ # Automated job endpoints
│ │ ├── metrics/ # Metrics data API
│ │ ├── pricing/ # Crypto pricing API
│ │ └── model/chat/ # Chat history API
│ ├── page.tsx # Main dashboard
│ └── globals.css
├── components/
│ ├── ui/ # shadcn/ui components
│ ├── metrics-chart.tsx # Account value chart
│ ├── models-view.tsx # Trade & chat history
│ └── crypto-card.tsx # Price display cards
├── lib/
│ ├── ai/
│ │ ├── agent/ # AI agent logic
│ │ ├── model.ts # Model configurations
│ │ ├── prompt.ts # Trading prompts
│ │ └── tool.ts # AI tools/functions
│ ├── trading/
│ │ ├── buy.ts # Buy execution
│ │ ├── sell.ts # Sell execution
│ │ ├── current-market-state.ts
│ │ └── account-information-and-performance.ts
│ └── types/ # TypeScript types
└── prisma/
└── schema.prisma # Database schema
- Data Collection: Every 20 seconds, the system collects account metrics (balance, positions, PnL)
- AI Decision Making: Every 3 minutes, the AI analyzes market data and makes trading decisions
- Execution: Approved trades are executed via Binance API
- Transparency: All reasoning, prompts, and decisions are stored in the database
- Visualization: The dashboard displays real-time performance and trade history
Starting Capital (START_MONEY)
- Set your initial trading capital in USDT (Tether stablecoin)
- Example:
START_MONEY=10000means you start with $10,000 USDT - Recommended for testing: Start with small amounts (e.g.,
START_MONEY=30) - The AI will make trading decisions based on this available capital
- All profits and losses are calculated relative to this starting amount
Currently supported:
- DeepSeek V3 Chat - Primary trading model
- DeepSeek R1 - Advanced reasoning model (optional)
Want to add more models? Check out the AI SDK providers and add them to lib/ai/model.ts!
- Live Crypto Prices: Real-time prices for BTC, ETH, SOL, BNB, DOGE
- Account Performance Chart: Interactive chart showing total account value over time
- Completed Trades: Detailed history of all buy/sell operations
- Model Chat: Full transparency into AI's chain-of-thought and decision-making
- Positions: Current open positions (coming soon)
This is educational/research software. Trading cryptocurrencies involves substantial risk of loss.
- Start with small amounts or paper trading
- The AI model may make poor decisions
- Past performance doesn't guarantee future results
- You are responsible for any financial losses
- Review and test thoroughly before using real money
The original nof1.ai Alpha Arena is a closed competition. This open-source version aims to:
- Democratize AI trading research - Anyone can experiment with AI trading agents
- Educational purposes - Learn how AI agents make financial decisions
- Transparency - Full visibility into prompts, reasoning, and execution
- Community innovation - Improve and iterate on trading strategies together
MIT License - See LICENSE file for details
- Inspired by nof1.ai's Alpha Arena
- Built with shadcn/ui components
- Powered by DeepSeek AI models
- Trading via CCXT
⚡ Built with Bun + Next.js 15 + DeepSeek + CCXT
Markets are the ultimate test of intelligence. Let's find out if LLMs are good enough.