adennng/stock_strategy_lab

Stock Strategy Lab is a Python-based research workbench for quantitative trading, designed for studying A-share stocks (Chinese domestic equity market), ETFs (exchange-traded funds, baskets of assets you can buy and sell like a stock), and market indices. Rather than offering fixed pre-built strategies, the tool lets AI agents explore, write, backtest (test against historical data), and iteratively refine trading strategies through a closed loop of trial and improvement. The system is organized into three layers of AI agents. The Signal Layer agent focuses on a single asset and tries to answer questions like: when should this asset be bought, in what quantity, and when should it be sold? The Budget Layer agent works at the portfolio level and determines how capital should be divided across a pool of assets. The Portfolio Layer agent fuses the two by combining signal-layer timing decisions with budget-layer allocation logic into a final tradeable strategy. Each layer can run independently or be chained together to form a complete research pipeline. All three agents keep session memory, they track which strategies have been tried and what to try next, and allow the user to intervene at any point to change preferences, adjust risk limits, or steer the direction of exploration. The command-line interface supports saving and resuming research sessions. The tool is written in Python and requires Python 3.11 or later. Data capabilities rely on MiniQMT and xtquant, which are mainly available in a Windows environment. It supports plugging in OpenAI-compatible language model APIs, with DeepSeek and Moonshot/Kimi listed as options in the configuration.