Why this matters
Individual retail quants and hobbyist researchers often juggle multiple tools: slow full‑market scans, fragmented backtests, and ad‑hoc monitoring. This project bundles a local, zero‑ops A‑share workflow that runs Polars‑native full‑market scans, stores enriched Parquet snapshots, performs constrained backtests, and adds rule‑based realtime monitoring — with optional LLM features for generating strategy code and producing stock-level explanations.
What Sets It Apart
- Millisecond full‑market screening with Polars: vectorized expressions scan enriched Parquet snapshots across A‑share universe, so you can iterate strategies without heavy DB tuning.
- Reproducible, constraint‑aware backtesting via vectorbt: supports T+1, fees, slippage, stoploss and streaming SSE progress, so backtests reflect realistic trading rules.
- Integrated monitoring and persistence: four rule types (strategy / per‑stock signal / price / market abnormality), alert cooling, voice/Feishu push and durable alert logs for operational use.
- Data plugin architecture and local storage: TickFlow SDK + plugin examples let you merge third‑party sources into DuckDB/Parquet for same‑platform analysis, preserving data control and privacy.
- Guarded LLM integration: optional AI generator creates single Python Polars strategy files; backend performs AST safety checks and generated code is sandboxed into an AI strategies directory.
Who It's For
Great fit if you are an advanced retail quant or developer who wants a single self‑hosted environment to iterate screening → backtest → monitor loops on A‑share data, value low‑latency vectorized scans, and may optionally prototype LLM‑assisted signals. Look elsewhere if you need a commercial, regulated trading terminal, out‑of‑the‑box broker connectivity, or turnkey buy/sell recommendations — the project is explicitly research/demo oriented and disclaims investment advice.
Tradeoffs and Practical Notes
- Requires local resources and some ops familiarity (Python ≥3.11, Node, Docker recommended).
- AI features are optional and gated by API keys; generated strategies follow strict import and file‑scope rules to limit risk.
- Not an official TickFlow product; data access and quota depend on TickFlow credentials and paid tiers.
