christopherkenny/gptzeror ? reverse-engineered prompt

Build me an R package that lets me check whether a piece of writing looks like it was generated by AI using the GPTZero API.

I want to be able to pass in plain text directly, or upload a local file like a txt, docx, or pdf, and get back a tidy table with the prediction results for the whole document, each paragraph, and each sentence. It should be easy to use from R, with simple functions for text and file inputs, and the output should be something I can inspect or pipe into other tidyverse code.

Please include clear documentation, examples, and tests, and make it feel like a normal small R package that someone could install from GitHub and use right away. If anything about the current API needs checking, look up the latest docs online and match the behavior to that.