Voice (technically speech) recognition is relatively close to being doable at acceptable quality with open source tools.

The hard parts of the Assistant, Alexa, etc., are basically everything else.