As mentioned in the title, ollama models run kinda fast when asked on the server itself, but are REALLY slow when asked via HASS. The parameter size doesn't change anything. I have attached a video showing the issue.
#Ollama slow, but only on HASS
untold cedar
charred swift
Well in my instance is quite slow too as per it works in a separate machine and actually I discovered that if I speak to the satellite in the living room the bedroom one start to say something too.. something really strange... the response on my network takes about 30 second maybe longer depending on how long it is the replay.. of course few seccond from SSH the server where it runs..