This project provides a proof-of-concept AI investment system for the A-share market, designed to assist in investment decision-making. It targets individuals interested in exploring AI-driven quantitative trading strategies and offers a modular, agent-based architecture for educational purposes.

How It Works

The system employs a multi-agent architecture, featuring specialized agents for market data analysis, technicals, fundamentals, sentiment, valuation, risk management, and portfolio management. A key innovation is the "Debate Room" module, which incorporates a "Bull Researcher" and "Bear Researcher" to analyze market data from opposing viewpoints. This is further enhanced by integrating a third-party LLM for objective evaluation and a hybrid confidence calculation that fuses traditional metrics with LLM scores, leading to more balanced and comprehensive investment decisions.

Quick Start & Requirements

• Install: Clone the repository, install Poetry, then run poetry install.
• Prerequisites: Python 3.x, Poetry. Requires API keys for Google Gemini or OpenAI-compatible services.
• Setup: Set environment variables (e.g., GEMINI_API_KEY) or configure the .env file.
• Usage:
  • Command line: poetry run python src/main.py --ticker <stock_code>
  • With reasoning: poetry run python src/main.py --ticker <stock_code> --show-reasoning
  • Backend API: poetry run python run_with_backend.py (access at http://localhost:8000/docs)
• Docs: Backend API Documentation (once backend is running)

Highlighted Details

• Supports multiple LLM providers (Google Gemini, OpenAI-compatible).
• Modular agent design for easy maintenance and extensibility.
• Includes a "Debate Room" with LLM-based third-party analysis for decision enhancement.
• Provides a backend API for asynchronous analysis and custom frontend development.
• Detailed logging for all API calls and agent reasoning processes.

Maintenance & Community

The project is a modification of ai-hedge-fund. The creator invites users to join a paid knowledge community for deeper insights and early access to code. Contact is via WeChat (PareidoliaX). A public WeChat account, "空指针指向量化 Agent," is also mentioned for content.

Licensing & Compatibility

MIT License. Permissive for commercial use and closed-source linking.

Limitations & Caveats

The project is explicitly for educational and research purposes only and is not suitable for actual trading. The backtesting system is still under testing. The README mentions a paid community for additional support and resources.