Kenzy-Zero/kenze ? reverse-engineered prompt

Build me a Python tool called kenze that helps me clean and reshape huge CSV, Parquet, and JSON files without running out of memory.

I want both a friendly interactive shell and a simple command line version, so I can load a big file, filter, keep or drop columns, rename, cast, deduplicate, sort, count, join, and export the result to a new file. It should work smoothly on files bigger than RAM by using DuckDB behind the scenes, with automatic memory handling and streaming so it doesn’t crash.

The shell should feel smart and easy to use, with live previews as I build steps, autocomplete for file names, column names, and even values from the data, plus a quick command menu. I also want basic checks like assert unique, assert not null, validate schema, and a way to convert between CSV, Parquet, Excel, and GeoJSON. If possible, add simple terminal plots and a history of recent runs too.