Built for the researcher who has to defend the number.

Research runs on statistics that are easy to get subtly wrong, whatever the field: the wrong test for the data's shape, an assumption that quietly fails, a result that can't be reproduced six months later when a reviewer asks for the script. None of that is a knowledge problem so much as a tooling problem - the correct path should be the easiest one to take.

Ingest a dataset, and Celeus profiles it, flags identifiers before anything else touches the data, suggests the statistically appropriate method, runs it deterministically, and interprets the output in plain language - flagging assumption violations and other problems along the way. Every run ships as a reproducible package: the plan, the exact script, the seed, session info, and an audit log. The changelog lists what has changed, newest first.

Celeus is a research statistics tool, not clinical decision support - it never produces diagnosis or treatment recommendations for individual patients. It's built to make published research more defensible, not to make clinical calls.