doesn’t look like work on Elivia, a client/server model, ever progressed beyond proof-of-concept.
My impression is on-device stt-engines progressed to the point of being usable. I played with Vosk inside Dicio not long ago and thought it to be fun for simple assistant tasks - [FEATURE PROPOSAL] Open Source Voice Assistant on /e/ - #49 by tcecyk - it’s on fdroid.
If you look around on github, Apps depending on Vosk or DeepSpeech and local language models pop up, I think it’s a space to watch, rather then off-devices services.