stock-market-analysis

Run the stock_9layer_fetch.py script with the stock code inside the quantmind container, then follow the nine-layer runbook covering market, valuation, financials, technicals, capital flow, microstructure, industry, models, and news. The output is written as Markdown and converted to PDF in the trading_agents report directory.

Call the /api/v1/selection/daily endpoint with an auth token, then parse the candidates array into a CSV with utf-8-sig encoding. Similar export snippets exist for full feature sets, batch stock comparisons, and risk scorecards.

Financial statements and margin trading data live in parquet files, not the REST API, and the required libraries (pandas, duckdb) only exist inside the quantmind container. Scripts must be copied in with docker cp and executed via docker exec.

The l2_factors partitions stopped updating at 20260227 on the vendor side, so recent flow, chip, and microstructure fields are legitimately empty. Reports must mark this as missing data rather than fabricating values.

Stock volume is in shares while index volume is in lots, amounts are in ten-thousand yuan, valuation close is unadjusted while technical indicators use back-adjusted prices, and dividend_rate switched to percentage format on 20260814. Always verify against the quantdb-fields reference before computing.

Yes, the platform supports A-share, Hong Kong, US, futures, and crypto markets, with multi-market K-line endpoints available. However, the deepest nine-layer pipeline with parquet financials and L2 microstructure is built primarily around A-share QuantDB data.