Does not detect LmStudio server

[JaziGuj]

It does not detect and connect to the LmStudio server. I did not try Ollama yet.

[haiphucnguyen]

Thank you @JaziGuj for reporting issue, recent refactors causes the connection issue with lmstudio. Because we are using HTTP 2 as default while LmStudio is using HTTP 1.1. After modifying to HTTP 1.1, then I am able to connect to LmStudio. I will release a fix in the next release at this weekend.

[star26bsd]

Hi @haiphucnguyen - which version will be the fixed one?

[haiphucnguyen]

@star26bsd, it hasn’t been released yet. I’ll fix it in the next version, which I plan to deliver this weekend.

[haiphucnguyen]

Hi @JaziGuj , @star26bsd now the askimo can work with lmstudio with the latest release https://github.com/haiphucnguyen/askimo/releases/tag/v1.2.25. Let me know if the problem exists.

[star26bsd]

Thanks @haiphucnguyen, testing 1.2.25 as we speak.

1. Starting LM Studio on Windows x64, starting Askimo.
2. Opening settings => dialog for configuring AI provider, chosing Lmstudio as provider
3. accepting default of http://localhost:1234/v1
4. Clicking "Test Connection"
5. Error: No models available for lmstudio

At this point, I suspected I need to load a model in LM Studio first, which is counter intuitive. When the dialog offers to "test connection", I'd assume a bare connection test of being able to talk to the API of lm studio, no matter whether or not any pre-conditions are fulfilled.

My expectation was to get presented with a list of available models within LM studio, especially /api/v1/models is available.

6. Loading a model in LM Studio and repeating 4) => Works.

On a seperate note, using LM Studios native API layout http://localhost:1234/api/v1 does not work for testing a connection, even though LM Studio says

2026-04-05 19:43:04 [DEBUG] Received request: GET to /api/v1/models
2026-04-05 19:43:04  [INFO] Returning 9 models from v1 API

Thanks for this quick fix!

[haiphucnguyen]

Hi @star26bsd ,

I'd assume a bare connection test of being able to talk to the API of lm studio, no matter whether or not any pre-conditions are fulfilled.

That's the very good point for user convenient. I am using LmStudio 0.49+1 and when I downloaded the new model, it is automatically loaded the model (I tested on MacOS though). I created a ticket https://github.com/haiphucnguyen/askimo/issues/306. Askimo uses the lmstudio endpoint /v1/models and lmstudio spec clearly say it only returns the loaded models only. So it is a minor issue we should fix to improve the user experience for lmstudio.

On a seperate note, using LM Studios native API layout http://localhost:1234/api/v1 does not work for testing a connection, even though LM Studio says ...

Yes, it is expected. The reason because we use the OpenAI compatible endpoint. It is another user experience issue is logged in another LmStudio issue https://github.com/haiphucnguyen/askimo/issues/307.