[Feature]: Show indication of model defaulting
At times when using GPT-4 the model will default to 3.5 in case of, for example, a breach of Openai's community guidelines. It would be nice if there is a indication of that.
I'm sorry, I was not aware of that. If I understand correctly, you're meaning that even when using GPT-4, the OpenAI API might respond using GPT-3 in certain cases? Is there any documentation to clarify this?
I can't seem to quickly find documentation on that but looking at the openai community forum, it it a thing that happens. In Chatwizard it seems to switch the model for a chat. What I suspect is happening is that openai sends a response as gpt-3 and that makes chat wizard pick it up as the model for the chat. So consequent prompts are send to gpt-3 without the user knowing it. Making the Ai suddenly seem a bit drunk.
- Just as a idea, would it be relatively easy to label the response with the model as it is send by the api? Makes it easier to spot when this happens.