shy3130/tick-stock-panel
Tick Stock Panel is a self-hosted A-share research stack with screening, backtesting, monitoring, and an AI assistant whose 18 read-only tools expose auditable data calls.
Tick Stock Panel packages A-share screening, monitoring, and backtesting in a self-hosted Docker app. It includes **25 strategies**, computes **68 indicator and signal columns**, and gives its assistant **18 read-only tools** with inspectable call parameters, timing, and result summaries.
Agent builders can study its fail-closed tool access, visible tool traces, local Parquet storage, and extension model: backend and frontend modules are discovered through isolated directories and can be removed without changing core files. Data-source capabilities are routed independently so providers can change without altering metric or backtest conventions.
Tick Stock Panel packages A-share screening, monitoring, and backtesting in a self-hosted Docker app. It includes **25 strategies**, computes **68 indicator and signal columns**, and gives its assistant **18 read-only tools** with inspectable call parameters, timing, and result summaries. Agent builders can study its fail-closed tool access, visible tool traces, local Parquet storage, and extension model: backend and frontend modules are discovered through isolated directories and can be removed without changing core files. Data-source capabilities are routed independently so providers can change without altering metric or backtest conventions. It is a personal open-source research project, not trading or recommendation software. Some data sources require separate API keys, the packaged image omits certain plugins, and Codex CLI mode mounts local credentials read-only and should be used only on a trusted machine.
This is a concrete domain-agent implementation of bounded, inspectable tooling: read-only assistant actions, visible traces, stable local data conventions, and removable modules constrain behavior without hiding extensibility. It supports structured domain interfaces over generic tool access, while its personal-project status and credential assumptions limit how far the pattern can be generalized to production or multi-user systems.