byrizki/rusto-rs ? reverse-engineered prompt

Build me a fast, pure Rust OCR library that can read text from images and work on Linux, macOS, Windows, iOS, and Android.

I want a simple API that lets me load an OCR model, run it on an image file or image bytes, and get back the recognized text, confidence score, and bounding box for each result. Please also include a small command line tool that can output OCR results as JSON, plain text, or TSV.

If it makes sense, add support for document layout analysis too, so it can process things like titles, headers, tables, figures, and generate a markdown style summary of a document. Keep it memory safe and as dependency light as possible, with a pure Rust image processing path, but it’s okay to make OpenCV optional if needed.

Use the PaddleOCR model format with MNN inference, and make the project easy to use as a library plus from other languages through C bindings. If you need to check current docs for anything, look them up online.