CDDLeiden/DrugEx — reverse-engineered prompt

Build me a usable Python package for AI driven small molecule generation, something a drug discovery researcher could actually run without digging through the code. I want it to handle molecule cleaning and encoding, let me train or load generative models, including the older recurrent models and the newer transformer based ones, and support multi objective reinforcement learning so generated molecules can be pushed toward desired scoring goals. It should work both from Python and from a simple command line interface.

Please make the main user flow feel practical, install the package cleanly, include sensible defaults, and add a couple of working tutorial notebooks or examples that show how to preprocess data, train or fine tune a model, generate molecules, and score or optimize them. If there are pretrained models already referenced, make it easy to plug those in too. Optional integrations for external scoring tools are great if available, but the core library should still work without them. Check the current docs online if you need to fill in gaps.