tahahssn/sanitizepy ? reverse-engineered prompt

Build me a Python library that helps me clean messy pandas data in a way that is easy to understand and safe to use.

I want one main Cleaner entry point that can inspect a dataframe, tell me what is wrong in plain English, suggest a cleaning plan, and then either preview or apply the changes. It should support missing values, duplicates, bad datatypes, simple quality checks, and a health score so I can quickly see how dirty the dataset is. When it cleans data, I want a dry run option, a real apply option, and a clear audit trail showing exactly what changed and why.

Please also include simple feature engineering helpers like interaction, ratio, polynomial, log, and datetime features, plus a report output that can be shown in the terminal or exported as text or JSON. If it helps, look up current pandas and pydantic docs online. Make it feel polished, reliable, and easy for someone with basic Python knowledge to use on a CSV file.