The key feature I haven't seen any of these opensource projects implement is microphone response coordination: If you have multiple microphones and speakers, which one responds?

My google home's are terrible at this: often one in another room responds, but at least it's only one. When I tried to run Genie (https://genie.stanford.edu/) I had multiple devices responding simultaneously. It was a disaster.

For me, this is the core feature that will enable me to swap out my corporate listening devices for an opensource, cloud-free alternative.