Zaid-maker/urdu-text-utils ? reverse-engineered prompt

Build me a lightweight Urdu text processing toolkit for JavaScript and TypeScript that works in Node and the browser, with no runtime dependencies.

I want it to handle the main things people need when working with Urdu text, like normalizing different Unicode spellings into a consistent form, transliterating between Roman Urdu and Urdu, removing diacritics, detecting Urdu script, converting digits between English, Urdu, and Arabic style numerals, filtering stop words, splitting words and sentences, sorting Urdu text properly, and doing simple fuzzy search. It should feel easy to use from plain JavaScript or TypeScript, and the package should be fully typed and usable as both ESM and CommonJS.

Also set it up with good tests, docs, and a small playground or examples so I can see the output clearly. If it makes sense, include a PHP version too that matches the same behavior. Look up current docs online if you need to, and make sure the public API is clean and practical for real Urdu text.