1B Model not using tools when running from ollama

[CeyssensHelder]

When using the home-llm 1B model, it will correctly use the tools like turning off the lights. However it is very slow when the devices are added to the prompt, so I tried running the model on ollama so I could use a GPU.
When I ask the same question like:
"turn off bedroom bulb"
It should answer with:

"deactivating Bedroom Bulb as requested."

and turns off the lights, but with ollama it answers with:

"deactivating Bedroom Bulb as requested.
homeassistant {"service": "light.turn_off", "target_device": "light.bedroom_bulb"} "

(I removed one ` for githubs formatting)

It seems like it wants to use the tool, but somewhere it goes wrong.
I tried getting the settings the same as the non-ollama one, but the config menus are slightly different.
Did I miss something here?

[acon96]

Make sure in your settings that:

1. Selected LLM API is set to Home-LLM (v1-v3)
2. Service Call Regex is set to

```homeassistant\n([\S \t\n]*?)```

[CeyssensHelder]

Big thanks! I overlooked the Service Call Regex