Stock Wars is an advanced framework for autonomous stock trading, leveraging the power of Large Language Models (LLMs) and the OpenBB/Financial Modeling Prep (FMP) APIs. This innovative system facilitates comprehensive backtesting and reliable live trading, all while utilizing a ReAct (Reasoning + Acting) loop to effectively analyze market data and execute trades.

Key Features

• Autonomous Reasoning: Employs cutting-edge LLMs, including Chutes.ai (various models such as D-seek) and GPT-5, to analyze both technical and fundamental data.
• OpenBB/FMP Integration: Utilizes the FMP API and OpenBB to access high-quality financial data, including price, income, balance sheet, and news updates.
• Backtesting Engine: Simulate trading agent performance against historical data using backtesting/start_agent_backtest.py, ensuring strict adherence to lookahead constraints, thereby eliminating bias.
• Live Trading Daemon: Features a production-ready trading loop governed by PM2 for automatic restarts and enhanced reliability.
• MCP Support: Implements the Model Context Protocol (MCP) to standardize tool execution, promoting consistency in trading strategies.
• Portfolio Management: Monitors cash flow, open positions, and calculates both realized and unrealized profit and loss (P&L).
• Interactive Command Line Interface (CLI): Conveniently manage portfolios and monitor agent status via llm_stock_manager_cli.py.

Usage Guidelines

Backtesting

Assess your agent's past performance on historical datasets:

# General usage for backtesting
python3 backtesting/start_agent_backtest.py --symbol AAPL --start-date 2023-01-01 --end-date 2023-01-31
# Single day analysis
python3 backtesting/start_agent_backtest.py --symbol NVDA --date 2024-03-15

Live Trading (Production/Paper)

The live trading system is designed for continuous operation or one-time analysis.

• Multi-Stock Analysis: Manage up to five stocks concurrently, leveraging parallel processing for quicker analysis.
• Portfolio Allocation: Implement a waterfall allocation approach to tackle shared state issues regarding cash distribution. Trade decisions are ranked by confidence, with ties resolved based on sector categorization.

Monitoring Your System

To monitor and manage the daemon:

pm2 list                    # Check current status
pm2 logs trading-daemon     # Access live logging output
pm2 monit                   # Real-time performance dashboard

Project Structure

The repository features a well-organized structure:

stock_agent_eval_clean/
├── backtesting/                # Backtesting scripts and logic
│   └── start_agent_backtest.py # Main entry point for backtesting
├── custom_TradingBot/          # Core trading logic
│   ├── live_trade/             # Live trading components
│   │   ├── ReasoningAgent.py   # Main LLM agent class
│   │   ├── live_trading_loop.py# Main loop for live trading
│   │   ├── OpenBBMCPServer.py  # MCP Server for OpenBB tools
│   │   └── ...
│   └── requirements.txt        # Python dependencies
├── tests/                      # Unit and integration tests
├── PM2_SETUP.md                # PM2 setup instructions
├── QUICK_START.md              # Quick start guide
└── README.md                   # Documentation

Additional Resources

• QUICK_START.md: Guidance for rapid setup.
• PM2_SETUP.md: Comprehensive instructions for robust process management.
• PORTFOLIO_CONTEXT_AND_NEWS_STRATEGY.md: In-depth information on the agent's strategy and context.

Through its sophisticated module support and comprehensive trading functionality, Stock Wars is well-equipped to serve traders looking for an intelligent and automated approach to stock market engagement.