sngyai/Sequoia-X

Sequoia-X V2 is a Python system for screening China's A-share market with technical strategies. After each trading day's close it updates daily price data, runs six pattern scanners, and pushes the selected stocks to a Feishu group chat. It is aimed at users who already understand A-share trading and want a repeatable, self-run screening routine rather than a hosted signal service.

How it works

The data layer pulls historical and incremental daily K-lines in back-adjusted (hfq) form from baostock — described as free, registration-free and rate-limit-free — and stores them in a local SQLite file, avoiding anti-scraping issues from other sources README. Back-adjusted prices keep past prices fixed, which the author says suits incremental storage and avoids distortion around dividend/split adjustments.

The six built-in strategies are TurtleTrade (20-day high, turnover above 100 million, bullish candle filter), MaVolume, HighTightFlag, LimitUpShakeout, UptrendLimitDown, and RpsBreakout README. Each is a separate module behind a common strategy base class, so adding a scanner means adding a class rather than editing the pipeline.

Running it

Requirements are Python 3.10 or newer; dependencies install via uv sync or pip install. After copying .env.example to .env and entering a Feishu webhook URL, the first step is --backfill, which loads roughly 5,200 A-share histories in about 12 minutes. Daily runs use eight parallel processes to update data and execute strategies in about 2–3 minutes README. The README suggests a weekday 19:15 cron job to automate this after the close.

Backfill is a one-time cost, but the daily path depends on baostock being reachable and on the local SQLite file remaining intact. The database can be copied to another machine, which makes migration simple.

Constraints to weigh

• Python-only, self-hosted workflow. There is no packaged app or hosted interface; you run the script and manage scheduling yourself repository.
• Feishu is the delivery channel. Pushing results requires a Feishu webhook; users not on Feishu would need to change the notification layer.
• No declared repository license even though the README closes with MIT — check the actual license file before redistributing or building on it.
• Screening, not trading. The system selects stocks by technical pattern; the README does not document order execution, backtesting results, or performance validation.

HotRepo's take

HotRepo recorded the repository at 5,978 stars on 2026-09-02 and 8,391 on 2026-10-31, appearing on the daily trending board on three sampled days repository. Those are dated observations only; the available board history is too sparse to say whether interest is rising or fading.