I have no idea if voice dictation has already been handled, but if it hasn’t, we may be able to do something with an app called Kõnele.
Kõnele is a proper voice input app, which allows it to provide voice dictation for the AOSP keyboard and other apps. it’s also free and open-source.
While it’s default server doesn’t support English, private servers can be used as discussed here.
Firefox also has a voice dictation project in the works, but as far as I know, Kõnele is the only thing that is working right now.
Also, voice dictation is not the same as a voice assistant. I can’t find the edit button now.
You said it, /e/ devs should set up a Kõnele server(damn, it’s difficult to type that õ stuff), which called Kaldi gstreamer or PocketSphinx gstreamer.
I support this.
Meanwhile Mozilla is triyng to built his own voice recognition system. People can contribute by giving their voice or validating samples, here : https://voice.mozilla.org/