Hello recently i tried to make my own ai assistant using the wyoming protocol and home assistant assist so far whisper piper and the wyoming protocol work great but i wanted to make the assistant a bit smarter so i decided to run ollame 3.2 1b on my server and connect it to the assistant ollama 3.2 1b works fine when i tested in the command line yes its a bit slow but works but when i try to use it with assist it doesnt work at all it just stays stuck at 3 dots and after a while it gets timed out
#Ollama 3.2 1b stuck
viscid spoke
zealous folio
A 1b model probably won't work too well for this, even 7b and 14b models can struggle with some things. Additionally depending on the amount of entities exposed and scripts/automations, the context being passed to the llm can get large to the point that the model will crawl unless you have a strong gpu with a large amount of vram.