Ok, i managed to compile Sendspin-enabled firmware for my Koala satellite. It works. So we’re prepared for release.
I don’t want to make alpha/beta things. But i can make some branch with modified config, if someone is curious.
That’s very nice! I would like to see it! Would it be a lot of extra work to make it work for the ReSpeaker Lite integration? Maybe if I have a look at the Koala I might be able to do it for the ReSpeaker Lite integration
I have an automation system at home that deodorizes rooms with a deodorizer and a fan. The entire automation cycle takes about 5 minutes. When I call the automation, the Raspberry Lite remains active until the automation ends (then for 5 minutes with the LED flashing). This doesn’t happen with the other two voice assistants (which aren’t Raspberry Lite).
That’s very odd. I don’t see this behaviour at all, except for the cases when automation is actually doing some big job like calling LLM, and prepares the answer.
If you have automation that actually runs for 5 minutes blocking, I’d suggest rethinking it.
However, all this is true for other assist devices that are PE based.
It is simple, but waits are in general not great, especially if it’s automation and not script.
Basically you’re adding software blocking. So everything that is stuck on that block - will be stick for 5 minutes.
As I said, I can’t imagine why other satellites behavior is different for you.
If I have to do blocking job somewhere, my last resort (apart from timers, which are great for this!) is creating a script, and then using it from automation in non-blocking way (this is crucial, there are two ways to launch script - via direct action call or via script.turn_on action, latter is launching the script in background).
at this point it occurs to me…
that perhaps when the automation is managed by HA locally everything is fine… whereas when the automation is managed by an LLM then it remains blocked. Maybe that’s why…
what can I put in an automation instead of “wait 5 minutes”?
Use timer entity. Create 5-min timer, then on triggering start your devices, start timer - and that’s it. Then add another trigger on timer stop - and on that trigger stop your device. That will basically create two automations - one for start, one for stop.
Move automation actions to the script. Then in automation call that script as a target of script.turn_on action. This way script will be holding, but automation itself will finish right away.
FYI, read there is some bug in the new Music Assistant 2.7.0 release that can cause Sendspin to not work as it should so be sure to update to Music Assistant 2.7.1 patch release with some fixes:
Thank you. I’m aware of that, but i have issues with backbone provider, which results in Github and ghcr.io unavailability… Cannot even download the image for 2.7.1…
I am trying to put together a bit more powerful smart speaker using a TPA3118 amplifier board. I supply with 15V and step down to 5V to the ReSpeaker… However it is quite noisy and my basic filtrering doesn’t do much.
Of course when i power the ReSpeaker and amplifier using separate power supplies there is not noise.
Has anyone succeded putting a bit more powerful amplifier on the same ground as the ReSpeaker? What did you use, preferably a solution my basic soldering skills can do
Uhh, will listen. I have other board that has so much noise. Don’t use Respeaker with big speakers. Tried 3.5mm with some battery powered speaker, and of course it was ok… But not more that that.
Hm, it’s not a big speaker, but more than what the 5W from the ReSpeaker can handle… It sound great when powered from separate supplies but maybe the ReSpeaker is too noisy after all… Gonna do some more testing with some filtering.
Can’t wait to test it out. It’s the missing piece to ditch Google for me.