Single-Agent Quantitative Trading System

A professional, modular agent-based trading system built with Python. It uses ccxt for data fetching, custom technical analysis, and DeepSeek API for strategic scoring.

Architecture

• DataAgent: Fetches OHLCV data from Binance (or other exchanges supported by ccxt).
• FeatureAgent: Calculates technical indicators (RSI, SMA, Volatility, ROC).
• LLMAgent: Sends market summaries to DeepSeek LLM for scoring and selection (0-10 score).
• ExecutionAgent: Allocates capital (Equal Weight) to top 3 selected assets and logs trades.

Setup

1. Install Dependencies:

   pip install -r requirements.txt

   (Note: pandas-ta is optional, the system currently uses built-in pandas calculations for stability)

2. Configuration:

   • Edit config/config.yaml to set your target symbols and parameters.
   • Create a .env file (or edit the existing one) and add your API keys:

     DEEPSEEK_API_KEY=your_deepseek_key_here
     # Optional for public data
     BINANCE_API_KEY=your_binance_key
     BINANCE_SECRET=your_binance_secret
     

Usage

Run the main orchestrator:

python main.py

Logs & Data

• Market Data: Saved in data/*.csv for debugging.
• Trade Logs: Transactions are logged in data/trade_log.csv.
• System Logs: Detailed execution logs in agent.log.

Customization

• Strategy: Modify agents/llm_agent.py to change the prompt.
• Indicators: Add more indicators in agents/feature_agent.py.