A retail trader’s AI analyst team for surviving China’s A-share market

What it does This Python/Streamlit application chains together several OpenAI-compatible LLMs to act as a virtual securities analyst team covering China’s A-share market. It ingests real-time and macro data—from local TDX feeds, National Bureau of Statistics releases, and 20 social/news platforms—and runs them through agent-driven workflows for stock picking, sector rotation alerts, macro cycle analysis, and automated trading decisions. The system also integrates with miniQMT for order execution and fires alerts via DingTalk, email, or webhook.

The interesting bit The author openly calls the project a cobbled-together experiment born from years of expensive trading courses and losses. It leans into that energy: one agent runs Kondratiev wave analysis on official macro data, while another hunts for sub-¥10 Shenzhen stocks with a README-typo-level net-profit-growth threshold of 65,100%.

Key highlights

• Multi-agent macro analysis using official National Bureau of Statistics data mapped to A-share sectors and specific stock picks.
• Several built-in retail-trader strategies: low-valuation value investing, small-cap growth, net-profit growth, and a “low-price bull capture” screener.
• AI盯盘 module with real-time monitoring, interactive K-line charts, and automated miniQMT trade execution respecting T+1 rules.
• News sentiment pipeline monitoring Baidu, Weibo, East Money, TikTok, Bilibili, and others for hot-topic impact on stocks.
• Model-agnostic backend: swap DeepSeek, Qwen, GPT-4o, or any OpenAI-compatible endpoint via a single .env variable.

Caveats

• The author explicitly calls it a pieced-together “small program,” and the codebase shows it: some strategies rely on rudimentary sell signals like MA5 crossing MA20 while promising KDJ upgrades later.
• The hosted test site warns that heavy traffic gets data sources blocked, so local deployment with a private TDX Docker instance is effectively mandatory.
• You’ll need to bring your own LLM API keys, and the README warns that leaving a DeepSeek test key in the public demo risks leakage.

Verdict Worth a look if you’re a mainland A-share retail investor who wants a hackable, all-in-one AI research dashboard and can tolerate self-hosting. Skip it if you need institutional-grade quant infrastructure or trade markets outside China.