explosion/spaCy ? reverse-engineered prompt

Build me a Python library for industrial strength natural language processing, something like spaCy.

I want it to take plain text and turn it into useful structured data, like tokens, parts of speech, named entities, sentence boundaries, and text classifications. It should support training custom models, loading pretrained language pipelines, and working well on many languages. Make it fast, practical, and ready for real projects, not just a demo.

Please include clear examples showing how to process text, train a model on custom data, and save and load a pipeline. If it makes sense, add a small docs site or simple website with usage guides and examples. Use current best practices for Python packaging and look up current docs online if you need to.

Keep the API clean and friendly, so someone can install it and start using it quickly.