Condition Categorical Variable on Bernoulli Parents

Yes! That’s in general the idea of marginalization.

As for inferencing mixture, there are a few posts on the discourse you can look at. You can start with the discussion here: Properly sampling mixture models