[BUG] Desktop model selector remains empty after setup until frontend reload #1197

Description

On first launch of the OpenJarvis desktop app (I downloaded the "OpenJarvis_1.0.1_x64-setup.exe" in the latest release "Desktop desktop-v1.0.2"), the model selector can remain empty even after setup completes and the backend is running. Reloading the frontend with Ctrl+R makes the models appear, after which chat works normally.

The model-fetch effect in frontend/src/App.tsx runs on mount, while the desktop app is still displaying SetupScreen. If the request fails during startup, the error handler sets the models to an empty list. The effect has an empty dependency array, so it does not retry when setup finishes. basically, it appears to be a startup race between the frontend and backend.

I also verified that http://127.0.0.1:8000/v1/models returns the installed models once the backend is running. Reloading the frontend resolves the issue.

Steps to Reproduce

1. Install and launch the OpenJarvis desktop app on Windows.
2. Allow the initial setup and backend startup process to complete.
3. Open the model selector in the frontend and observe that no models are available.
4. Attempt to send a message. The request may fail with "Error during generation: Model '' not found in any engine".

5. Reload the frontend with Ctrl+R.
6. Observe that the models appear and chat works as attached.

Expected Behavior

The desktop app should load the available models once setup and backend initialization have completed. Users should not need to manually reload the frontend.

Actual Behavior

The initial model fetch appears to happen before the backend is ready. The fetch fails, and the frontend retains an empty model list after setup completes.

Operating System

Windows

Python Version

3.13

Hardware

NVIDIA GPU

Engine

Ollama

Logs / Traceback