Send `LLM_MODEL` to frontend

[yimothysu]

The LLM_MODEL setting defines the default model to use when no other model is selected in the frontend.

When first initialized, the frontend needs a model selected in the settings dialog. Currently this is hardcoded to gpt-4-0125-preview. Since the frontend now sends this model as an initialization argument to the backend, the LLM_MODEL setting is effectively not respected.

To fix, the backend should send LLM_MODEL to the frontend. If there is no model key in localStorage, it should be initialized with the value of LLM_MODEL.

[foragerr]

How does the backend send the value of LLM_MODEL to the frontend? Are you thinking that's a new endpoint?

[yimothysu]

Yes.