Everytime, I say, it shoudl turn something on, it responds with "xxxx does not exist"
#Voice-Pe doesn't know my devices
olive glade
olive glade
Are you using a LLM?
No. Not yet.
olive glade
and is the device exposed to assist?
wdym by exposed? (I'm new, sorry)
wheat yarrow
So any device you add to HA is not by default exposed to the assist, meaning it can't see it, for privacy reasons
you'd want to go to your voice assist area, and look at the Expose Devices
So only devices exposed can be seen or controlled by Assist 🙂
olive glade
Yes, those are exposed
wheat yarrow
Have you tried this using the assist chat window? And have you checked your voice asst debug? (three dots on my screenshot above for the pipeline you are using)
wheat yarrow
What are you using for STT?
olive glade
stt.faster_whisper
wheat yarrow
What model size?
olive glade
idk, the default installed one. Where can I check?
wheat yarrow
Basically if you want anything on-par with cloud stt, you'd need to use the large-v3 model
hmm not sure, I don't use the addon 😅
I imagine it is in the addon config
olive glade
Can't find it.... I've an ollama instance running on another VM. Can I use this?
wheat yarrow
You'd need to use wyoming faster whisper
olive glade
I read sth about ollama, I think
wheat yarrow
ollama can be used for LLM stuff, but not for STT/TTS
olive glade
ah okay...
wheat yarrow
if you are using the HA addon, I am betting you are running the tiny or small model
As performance on anything larger would likely be pretty slow, even with a beefy cpu
olive glade
Yes, I guess. How can I host it on a differen mashine?
wheat yarrow
There's tons of topics on that in the discord, but it basically involves running the wyoming-faster-whisper docker container on a separate server with GPU toolkit added to docker
I use https://docs.linuxserver.io/images/docker-faster-whisper/ for instance, but I run it on my k8s cluster 😅