flairNLP/flair ? reverse-engineered prompt

Build me a simple Python NLP library like Flair that makes it easy to tag text and train your own models. I want to be able to load a prebuilt model and run things like named entity recognition, sentiment analysis, part of speech tagging, text classification, and word embeddings on a sentence with just a few lines of code.

It should feel friendly for people who are new to NLP, but still powerful enough for training custom models in PyTorch. Please include a clean API for sentences, annotations, model loading, prediction, and embedding use, plus a few working examples and tutorials so I can try it right away. If there are modern docs or model hub examples that would help, look those up online and follow current best practices.

Also make sure it supports a few languages, includes biomedical text support if that fits naturally, and has tests so I know the basic pieces work.