Apologies for the delayed response.
Theoretically, what you are saying should be possible with a MATRIX Creator hub and standalone MATRIX Voices as satellites placed all over the house, much like this MQTT audio streaming project we have but with multiple MATRIX Voices. The project linked used Snips which is no longer available for free use, but should be possible with any voice assistant that can take WAV files through MQTT.
The MATRIX Voice satellites in the project linked do indeed take in beamformed audio through this line of code.
On the Raspberry Pi, MATRIX devices can perform beamforming through our C++ programming layer called MATRIX HAL. However, this feature is not yet available in the MATRIX kernel modules which allow audio to be available through ALSA which is how most voice assistants running on the Pi receive and send audio.
In short, if your voice assistant interacts with ALSA on the Pi, your Pi+MATRIX Creator hub would not use beamforming for audio. If you use several MATRIX Voice satellites to communicate with said hub, those would send beamformed audio to the assistant.
Let me know if you have further questions!