online-ml/river ? reverse-engineered prompt

Build me a Python library for online machine learning on streaming data, like River.

I want the core experience to be simple, with a learn one and predict one style API that lets a model update one example at a time without needing to retrain on the full dataset. It should include a few useful built in pieces like linear models, preprocessing, metrics, pipelines, drift detection, anomaly detection, and some example datasets so someone can try it quickly. Make it feel friendly and practical for real time use, and make sure it works well with Python 3.11 and above.

Please also add a small set of examples and docs so I can see how to train on a stream, track performance as data arrives, and use it for things like classification or forecasting. If you need to check current docs for any details, go ahead and look them up online.