5.9K ·GitHub
What is this?
TSP is a self-hosted, zero-ops quantitative trading workbench for China's A-share stock market that combines stock screening, real-time anomaly monitoring, and strategy backtesting, with optional LLM-assisted analysis.
TSP is a self-hosted, zero-ops quantitative trading workbench for China's A-share stock market that combines stock screening, real-time anomaly monitoring, and strategy backtesting, with optional LLM-assisted analysis.Is this a trading bot or a research and screening tool?It is not an automated trading bot that places orders or predicts limit-ups; it is a self-hosted dashboard (FastAPI backend + React frontend) for screening, backtesting, and monitoring A-share stocks.
It is not an automated trading bot that places orders or predicts limit-ups; it is a self-hosted dashboard (FastAPI backend + React frontend) for screening, backtesting, and monitoring A-share stocks.How do I get it running?The easiest way is pulling the prebuilt multi-arch GHCR image and running it with Docker (no build needed); Docker Compose is available for a full local build, and a Dev mode exists for active development.
The easiest way is pulling the prebuilt multi-arch GHCR image and running it with Docker (no build needed); Docker Compose is available for a full local build, and a Dev mode exists for active development.Do I need Python or Node installed?Only for Dev mode (method D), which needs Python >=3.11, Node >=20, uv and pnpm; the Docker-based methods (A/B/C) need nothing but Docker itself.
Only for Dev mode (method D), which needs Python >=3.11, Node >=20, uv and pnpm; the Docker-based methods (A/B/C) need nothing but Docker itself.Can I use this for a commercial trading service?No — despite being MIT-licensed, the README explicitly limits the project to personal learning and research and forbids commercial use; it's also not meant as investment or stock-picking software.
No — despite being MIT-licensed, the README explicitly limits the project to personal learning and research and forbids commercial use; it's also not meant as investment or stock-picking software.What LLM providers does the AI assistance support?Any OpenAI-compatible endpoint, including DeepSeek, Alibaba's Tongyi (Qwen), and local Ollama models.
Any OpenAI-compatible endpoint, including DeepSeek, Alibaba's Tongyi (Qwen), and local Ollama models.What kind of built-in strategies does it ship with?25 built-in screening strategies plus minute-level strategies, custom DSL signals, and AI-generated strategies, scanned across the whole A-share market in milliseconds using Polars.
25 built-in screening strategies plus minute-level strategies, custom DSL signals, and AI-generated strategies, scanned across the whole A-share market in milliseconds using Polars.Can I plug in my own data sources?Yes — data sources are plugin-based (TickFlow SDK, fuyao/THS REST, or a custom YAML source), and extra fields can be wired straight into a first-class analysis page.
Yes — data sources are plugin-based (TickFlow SDK, fuyao/THS REST, or a custom YAML source), and extra fields can be wired straight into a first-class analysis page.Does the AI feature pick stocks for you?No — the project explicitly avoids built-in 'AI stock picks' or 'limit-up predictions'; AI is used for strategy drafting, four-dimensional stock analysis, and end-of-day review summaries instead.
No — the project explicitly avoids built-in 'AI stock picks' or 'limit-up predictions'; AI is used for strategy drafting, four-dimensional stock analysis, and end-of-day review summaries instead.What are similar projects?daily_stock_analysis, PanWatch, QuantDinger, abu, akshare, easy-stock, astra-quant-agent, qlib
daily_stock_analysis, PanWatch, QuantDinger, abu, akshare, easy-stock, astra-quant-agent, qlibWhich topics is it in?trade, polars, stock, stock-analysis, quantitative-trading, aigc, duckdb, backtesting, quant, algorithmic-trading, stock-market, training
trade, polars, stock, stock-analysis, quantitative-trading, aigc, duckdb, backtesting, quant, algorithmic-trading, stock-market, training