Advanced Features
Everything in this section is optional. The core of ChakataStat — import, define, analyse, chart, save — needs none of it. These pages are for when you want to go further: ask for analyses in plain language, make your work reproducible and scriptable, or reach the engine directly.
They share one foundation. Every analysis, chart and transform in ChakataStat is a single named command. The menus emit commands, and so the same commands can be typed, scripted, asked for by an AI, or sent by another program — all driving the one engine, all returning results (including errors) as data. The advanced features are different windows onto that one command stream.
The sections
-
The AI Assistant — connect a model backend (Claude or any OpenAI-compatible endpoint), ask for analyses in plain English, and read the resulting tables and charts inline. Choose and switch among several backends; add custom context about your data.
-
Reproducibility — the
.ckssyntax console that records and replays your work; the command palette; headless batch runs (--run) for CI and pipelines; exporting the output log to HTML/Markdown/LaTeX; and exporting a session as Python for the external Python client. -
Direct Engine Access — Advanced mode, the engine diagnostics panel, and driving ChakataStat from another program over MCP.
A word on safety
The powerful escape hatches are gated on purpose:
- The engine diagnostics panel is hidden until you enable Advanced mode in Settings.
- The AI assistant and MCP callers see only the tools you allow; a read-only profile lets them run analyses but blocks transforms that change data.
- Your API key is never stored — only your choice of backend is remembered.