How to estimate the conditional density p(y|x) if y and x contain both continuous and discrete variables

I guess that KDE can solve such a problem, but I still want to find some alternative solution. Is there any method in this library that I can use to solve this problem? I'm a beginner in this field, and I appreciate any help, Thank You!