Cannot select model on custom endpoint

[erickrf]

Describe the bug

If I want to run a self-hosted LLM (e.g., running on ollama), I cannot specify the model name.

To Reproduce
Steps to reproduce the behavior:

1. Go to Settings -> Language Models
2. Select Dictation Cleanup, then Self-hosted
3. The model name is fixed, there is no way to change it. Other options (e.g., for disabling thinking) are also not available.

Expected behavior
I expected to be able to select at least the model name; maybe even more arguments to the server.

Screenshots

Desktop (please complete the following information):

• OS: macOS 26.4
• OpenWhispr Version 1.6.10

[erickrf]

This is related to #148

[gabrielste1n]

@erickrf i think i need to improve the UX here, can you go to cloud providers and select the custom option - that should be what you are looking for

[erickrf]

Thanks, that seemed to work. Is there an option to customize the API call? I see that the models are still going into thinking mode, which takes a lot of time. It could be avoided by an API parameter.

[Today20092]

yes, i came looking for where to use ollama as well.