NannyML/nannyml ? reverse-engineered prompt

Build me an open source Python library for monitoring machine learning models after they’re deployed.

I want it to help data scientists check whether a model is quietly getting worse even when the true labels arrive late or not at all. It should estimate performance over time, detect data drift, and show how drift might be connected to changes in model quality. Support tabular use cases, both classification and regression, and keep it model agnostic so it works with different kinds of trained models.

Please include an easy to use API, clear documentation, and some interactive charts or visual reports so people can understand what is happening without digging through raw numbers. It should feel practical for real world monitoring, with sensible defaults, but still let users inspect the details when they need to. If you need to look up current library best practices or docs online, go ahead.