What is fleur?

Fleur is a research and analysis platform for the Chinese A share stock market, built almost entirely through AI assisted coding according to its own description. The project is explicit that it is for research and learning only: it does not handle real money, and none of its data, indicators, or AI generated output should be treated as investment advice. The platform covers several stages of stock research in one place. It collects public market and financial data from Chinese sources, calculates technical indicators like moving averages, RSI, MACD, and Bollinger Bands, lets a user build a stock screening strategy visually without writing code, backtests that strategy against historical data, and then monitors a simulated portfolio built from the results. The stated goals are keeping the data trustworthy and traceable back to its original source, making every indicator calculation explainable rather than a black box, letting people build strategies through a visual interface, and running the entire thing for free. Under the hood, data comes from public Chinese financial data sources including BaoStock, EastMoney, Sina, and others, none of which the project charges for or requires special access to. A Python based pipeline handles fetching, scheduling, and reprocessing that data, storing it first as Parquet files in S3 compatible storage and then loading it into ClickHouse, a database built for fast analytical queries. A separate Rust based engine calculates the technical indicators at high speed. A backtesting engine and portfolio evaluation system produce statistics like Sharpe ratio, maximum drawdown, and win rate. The user facing workbench is a React web application that walks a person through picking stocks by indicator, previewing the resulting stock pool, simulating a position, and backtesting the results. The technology stack spans Python for the data pipeline, Rust for the calculation engine, and React with TypeScript for the frontend, tied together with tools like Dagster for scheduling and dbt for data modeling. The project is licensed under MIT.