Steps:
- create a custom assistant with some kind of prompt that is meant to extract some kind of information from an image. For ex "tell me what animal is in this picture. That's the only information I want"
- use the custom assistant in Assistant
- do not type a prompt and just paste or add an image (for ex, of some animal)
What is supposed to happen:
It just says what animal is
What actually happens:
Gives a detailed explanation of the image you provided
What seems to be happening here that if there is no prompt, it is also not giving the custom assistant prompt to the llm api.
I have certain types of custom assistants that I shouldn't need to provide any text prompt with. purely an image should work but doesn't right now
This is very easily reproducible
