Use Ollama to execute intent_script's action AND respond with the intent_script's preconfigured text | Home Assistant | Page 1

pliant perch Sep 28, 2025, 5:32 PM

#

Hi all - I finally have a setup that allows me to use local LLMs at home and have been enjoying dabbling in home assistant using it. Being able to naturally speak and have Ollama execute intents rather than me having to say exactly the preconfigured sentence is great. One thing I'd like to figure out is if it's possible to use a language model to interpret my intent, but respond with a pre-determined template. For example:

Before llm:

"Ok Nabu - I'm leaving home"
house shuts down
"The house is shut down. The temperature outside is 75 degrees. The window is still open, please make sure to close it."

With Ollama:
"Ok Nabu - I'm about to bounce"
house shuts down
"Got it, have a great day."

I lose the response telling me information important to me before I leave home. Is there a way to only have Ollama interpret an intent but still respond using the configured text I have setup?

wet marten Sep 29, 2025, 8:41 PM

#

Which model have you found to be the snappiest?

pliant perch Sep 30, 2025, 3:51 AM

#

Using gpt-oss:20b on an amd ai 395+ and it's decent enough. the assistant voice lag is rough.

kind gate Oct 6, 2025, 9:51 AM

#

indeed gpt 20b run local execute and run well with HA, i have the same

arctic oriole Oct 9, 2025, 2:20 PM

#

@pliant perch What hardware are you using exactly? I'd like to install a local AI at home too. So far, I've been playing around with my old gaming PC and an Nvidia 1080. I was planning to upgrade to a 3060, but then only 12-13b models would be feasible...

foggy swan Oct 10, 2025, 12:47 AM

#

I'm squeezing a qwen3:0.6b model onto a GTX 1050. I'm going to grab a Tesla K80 24gb for the rebuild

#Use Ollama to execute intent_script's action AND respond with the intent_script's preconfigured text