Description
Trading terminals lock the screening logic, the conditions and the alerts inside their own frame — try to filter a batch of stocks your own way, watch a few signals and backtest to check, and you are either boxed in by their fixed conditions or paywalled behind a membership. TSP (tick-stock-panel) is a solo open-source project that moves the whole screen-monitor-backtest loop onto your own machine, with the strategies and data sources entirely yours — self-hosted, zero-ops.
It builds the things a quant workflow leans on into panels: a market dashboard for prices and capital flow at a glance, a screener that filters the whole market by conditions you write, a monitoring centre that watches your signals and alerts in real time, and backtesting that validates a strategy against history and lets you review it. The data layer takes third-party feeds and your own custom data, rather than being tied to one vendor. It also wires in an LLM, so natural language customises strategies and drives per-stock analysis and review without translating every rule into code.
FastAPI, Polars and DuckDB on the backend, React on the front, with TongDaXin data support; MIT-licensed.
Screener: filter the whole market by conditions you write, with the logic yours to define rather than a terminal's fixed options.
Monitoring centre: watch the stocks and signals you care about with real-time alerts, handing the time sink of watching the tape to the machine.
Backtest and review: validate a strategy against historical data, then replay every trade to judge whether it actually holds up.
LLM-driven: an integrated model lets natural language customise strategies and run per-stock analysis and review, without coding each rule.
Open data sources: third-party feeds like TongDaXin plus your own custom data, so the data layer is not tied to a single vendor.
It builds the things a quant workflow leans on into panels: a market dashboard for prices and capital flow at a glance, a screener that filters the whole market by conditions you write, a monitoring centre that watches your signals and alerts in real time, and backtesting that validates a strategy against history and lets you review it. The data layer takes third-party feeds and your own custom data, rather than being tied to one vendor. It also wires in an LLM, so natural language customises strategies and drives per-stock analysis and review without translating every rule into code.
FastAPI, Polars and DuckDB on the backend, React on the front, with TongDaXin data support; MIT-licensed.
Features
Screener: filter the whole market by conditions you write, with the logic yours to define rather than a terminal's fixed options.
Monitoring centre: watch the stocks and signals you care about with real-time alerts, handing the time sink of watching the tape to the machine.
Backtest and review: validate a strategy against historical data, then replay every trade to judge whether it actually holds up.
LLM-driven: an integrated model lets natural language customise strategies and run per-stock analysis and review, without coding each rule.
Open data sources: third-party feeds like TongDaXin plus your own custom data, so the data layer is not tied to a single vendor.
