Open Trading - AI Trading Arena

A C# port of the AI-Trader repository using the Ivy Framework. This application allows multiple AI agents to compete in a simulated stock trading environment using NASDAQ 100 stocks.

Features

• 🤖 Multiple AI Agents: Configure and run multiple AI models (Claude, GPT-4, etc.) simultaneously
• 📈 Real-time Trading: Watch agents make trading decisions in real-time
• 📊 Performance Analytics: Compare agent performance with charts and metrics
• 💾 Data Persistence: All positions and trading logs are saved in JSONL format
• 🎨 Modern UI: Built with Ivy Framework for a reactive, modern web interface
• ⚙️ Configurable: Easy configuration via JSON files

Architecture

Core Components

• Models: Data structures for positions, trading actions, stock prices, and logs
• Services: Business logic for trading, data fetching, position management, and AI agents
• Apps: Three main Ivy apps:
  • TradingArenaApp: Main trading dashboard
  • AgentPerformanceApp: Agent comparison and analytics
  • StockDataApp: Stock price data viewer

Key Services

• IStockDataService: Fetches and caches stock price data
• ITradingService: Executes buy/sell orders
• IPositionService: Tracks agent positions over time
• IAgentService: AI agent reasoning and decision-making
• IConfigService: Configuration management
• ILogService: Trading log persistence

Setup

Prerequisites

• .NET 9.0 SDK
• Anthropic API key (for Claude models) or OpenAI API key
• Ivy Framework (included via NuGet)

Configuration

1. Set up user secrets (recommended for API keys):

   # Set Alpha Vantage API key
   dotnet user-secrets set ALPHA_VANTAGE_API_KEY your-api-key-here
   
   # Set Anthropic API key (for AI agents)
   dotnet user-secrets set ANTHROPIC_API_KEY your-api-key-here

   Alternatively, you can set environment variables:

   export ANTHROPIC_API_KEY=your-api-key-here
   export ALPHA_VANTAGE_API_KEY=your-api-key-here

2. Edit Configs/default.json to configure:

   • Date range for backtesting
   • AI models to use
   • Agent parameters (max steps, initial cash, etc.)

Running

dotnet watch

The application will start and open in your browser at http://localhost:5000 (or the configured port).

Usage

Trading Arena

1. Select a date and agent from the dropdown
2. Click "Run Simulation" to start a trading session
3. Watch the agent make trading decisions
4. View current holdings and portfolio value

Agent Performance

1. Select a date range
2. Compare performance metrics across all enabled agents
3. View return percentages and trade counts

Stock Data

1. Select a stock symbol
2. Choose a date range
3. View price charts and historical data

API Keys

This application requires API keys for:

• Alpha Vantage (for stock price data): Get a free API key at Alpha Vantage
• Anthropic (for AI agents): Get an API key at Anthropic

Setting Up User Secrets

User secrets are stored securely and not committed to git:

# Set Alpha Vantage API key
dotnet user-secrets set ALPHA_VANTAGE_API_KEY your-key-here

# Set Anthropic API key  
dotnet user-secrets set ANTHROPIC_API_KEY your-key-here

# List all secrets
dotnet user-secrets list

# Remove a secret
dotnet user-secrets remove ALPHA_VANTAGE_API_KEY

The application will automatically fall back to mock data if API keys are not configured, but real data requires the Alpha Vantage API key.

Data Storage

All data is stored in the Data directory:

Data/
├── Agents/
│   └── {agent-id}/
│       ├── positions/
│       │   └── position.jsonl
│       └── logs/
│           └── {date}/
│               └── log.jsonl
└── Prices/
    └── {symbol}.jsonl

Configuration Format

{
  "agentType": "BaseAgent",
  "dateRange": {
    "initDate": "2025-01-01T00:00:00",
    "endDate": "2025-01-31T00:00:00"
  },
  "models": [
    {
      "name": "claude-sonnet-4",
      "baseModel": "claude-sonnet-4",
      "signature": "claude-sonnet-4",
      "enabled": true
    }
  ],
  "agentConfig": {
    "maxSteps": 30,
    "maxRetries": 3,
    "baseDelay": 1.0,
    "initialCash": 10000.0
  }
}

Development

Project Structure

OpenTrading/
├── Apps/              # Ivy UI applications
├── Services/          # Business logic services
├── Models/            # Data models
├── Data/              # Data storage
├── Configs/           # Configuration files
└── Program.cs         # Entry point

Building

dotnet build

Running Tests

dotnet test

Differences from Python Version

• No LangChain/MCP: Direct API integration with Anthropic/OpenAI
• Ivy UI: Modern reactive web UI instead of static HTML
• Service-based: Clean separation of concerns with dependency injection
• Type-safe: Full C# type safety throughout

License

MIT License

Developer Guide

See DEVELOPER_GUIDE.md for detailed development information, architecture, and contributing guidelines.

Features Implemented

✅ Core trading system with buy/sell logic ✅ Multiple AI agent support (Anthropic & OpenAI) ✅ Real stock data integration (Alpha Vantage API) ✅ User secrets for secure API key storage ✅ Multi-agent parallel execution ✅ Date range simulation support ✅ Portfolio tracking and visualization ✅ Trading logs and history ✅ Performance comparison across agents ✅ Retry logic and error handling ✅ Data persistence in JSONL format

Acknowledgments

• Original AI-Trader project
• Ivy Framework for the amazing web framework