[BYOM]: Display error messages in the UI (for OpenAI and more in general)

[LorenzoMinto]

Description

If a request fails we currently show a standardized message and it's not easy for the user to understand the issue, i.e. missing credits in API.

Steps to reproduce

1. Connect OpenAI API to Leo
2. Put faulty key
3. Make request

Actual result

General error message

Expected result

Error message with explanation

Channel information

• [ ] release (stable)
• [ ] beta
• [X] nightly

[petemill]

This is the error format:

{
    "error": {
        "message": "The model `gpt-4-turbo-preview` does not exist or you do not have access to it.",
        "type": "invalid_request_error",
        "param": null,
        "code": "model_not_found"
    }
}

[jonathansampson]

I was looking into this today for Ollama as well. Ollama uses the same shape, but only fills the response object partially in some cases: