Minimal configuration for satellite

[tlt_atro]

Hi,

My first device, with the complete installation of Rhasspy, is used to process the STT and the Intent Recognition.

My second device will just be used to send audio data directly to the server, and wait for the intent json after the intent recognition.

Does the second device needs also a complete installation of Rhasspy ? Or is it possible to stream the audio input directly to the server and to listen for an mqtt response after the intent recognition (with a minimal Rhasspy installation or without Rhasspy at all) ?

[romkabouter]

@tlt_atro

Does the second device needs also a complete installation of Rhasspy ?

Nope.

@tlt_atro

Or is it possible to stream the audio input directly to the server and to listen for an mqtt response after the intent recognition (with a minimal Rhasspy installation or without Rhasspy at all) ?

Yep, have a look at my satellite code for the ESP32 here:

GitHub - Romkabouter/ESP32-Rhasspy-Satellite: The repo has implementing an...

This code used the mic to stream audio data and is subscribed to various topics for handling interaction with Rhasspy.

Maybe it will help you with your satellite

[tlt_atro]

Thanks ! I have check your code, sure it will helps me a lot !

[donburch]

@tlt_atro

Does the second device needs also a complete installation of Rhasspy ? Or is it possible to stream the audio input directly to the server and to listen for an mqtt response after the intent recognition (with a minimal Rhasspy installation or without Rhasspy at all) ?

I haven’t looked at romkabouter’s ESP32 code, so can’t comment on that.

The official documentation has instructions for setting up Rhasspy satellites which basicly comes down to: install the same Rhasspy, but only configure audio input and output, and wakeword on the satellite; and forward all other modules on the satellite to the base machine.
I personally use MQTT and node-red on the base machine for communication and actioning the intents.

[tlt_atro]

Thanks for your reply !
Finally, I use a “custom” code for that.
Starting from the Python Skill code in Tutorials - Rhasspy, I convert the wav data to bytes and publish them to “hermes/audioServer/default/audioFrame”.