Before passing the input to GPT, we should pass the URL contents (BATCHED if too large) to a separate GPT request, just to write a summary of the contents in markdown, this summary can then be used for the quiz generation (less tokens, less garbage like OCR image output). This might even allow smaller models like GPT-3 to generate acceptable quizzes.
This summary can also be displayed, very useful to just look at what you are about to learn.
Before passing the input to GPT, we should pass the URL contents (BATCHED if too large) to a separate GPT request, just to write a summary of the contents in markdown, this summary can then be used for the quiz generation (less tokens, less garbage like OCR image output). This might even allow smaller models like GPT-3 to generate acceptable quizzes.
This summary can also be displayed, very useful to just look at what you are about to learn.