Distributed architecture

Rhasspy Voice Assistant - distributed architecture

Hi all, I see Rhasspy as a growing community and would like to confirm I have my thinking correct in installing a system wide Voice Assistant before I join in. Documentation is excellent but is vague over how to run on a distributed architecture with multiple satellites.

Current setup:
10 satellites and one central location - 200 devices communicating via mqtt
Home Assistant as a Gui and NodeRed for automation
Google is my current Voice Assistant which I am looking to replace

Proposed setup:

  1. 10 x satellite devices with Pi3+/ReSpeakHat (MIC) and Rhaspy + Wake Word (WW)
  2. mqtt uses Hermes protocol to transport .wav files
  3. Server 1 - VirtualBox + Ubuntu + Rhasspy + Speech to Text (STT) + Intent Recognition (IR)
  4. mqtt carries JSON with Intent
  5. Server 2 / Server 3 (existing) - Home Assistant (HA) / NodeRed (NR) to action intent
  6. mqtt returns JSON with speech response which is forwarded to:
  7. Server 4 - Text to Speech (TTS) - No Rhasspy ?
  8. mqtt uses Hermes protocol to transport .wav files back to satellite device
  9. Rhasspy plays on satellite speaker (SPKR)

am i missing anything ?

1 Like