Decentral Voice Assistant with central processing

Hello,
I’m pretty sure my question is answered somehow already, most likely in some documentation, but honestly I’m currently a bit lost in information and perhaps any potential answer may help others in future.

I’m currently running some decentral rhasspy instances (running on Raspberry Pis with GPIO microphones and a speaker). These decentral rhasspy instances are more or less just waiting for the wakewords, take the sound and forward to a central rhasspy instance to process. The processed result is forwarded to home assistant to issue what someone said. Thats actually working quite good for my needs. I haven’t touched this setup for quite a while.

Now I am planning a new voice infrastrucutre and I recognized that there is a lot of going on in home assistant (year of voice I have read :slight_smile: ). I have seen that there is not a lot of effort in rhasspy anymore so I worked briefly through the home assistant documentation and found a lot of stuff, but what I haven’t found is a way to build satellites as I currently have. I would again prefer a solution with some decentral pis (i need them anyway) and a central component to process.

Can someone perhaps point me in the right direction?

Thanks
Martin