Dragonfly/dragonfly — reverse-engineered prompt

Build me a Python library for scalable Bayesian optimization, basically something I can use when evaluating a function is expensive and I want the tool to intelligently search for the best inputs.

I want it to work both from Python code and from a simple command line script. The core experience should be easy, where I can call functions to minimize or maximize a black box function over a domain and get back the best value, best point, and a history of the search. It should also support more advanced real world cases like high dimensional search, parallel evaluations, multi fidelity optimization, multi objective optimization, and an ask tell workflow where I can feed in previous results and get the next recommendation.

Please include runnable examples and basic docs so someone can install it, test the install with a tiny toy function, and try a few sample optimization problems. Cross platform support for Linux, Mac, and Windows would be great. If you need details, look up the current docs online and mirror the overall behavior closely.