if you upload a large pdf/txt of a book, it's able to be aware of specific content on any page far beyond the context window limit as per the API. so they must be handling the files differently than loading the contents into the prompt. what techniques are there to handle a large amount of info like all the text from a book? this isn't available through the API
1 comments
Retrieval Augmented Generation, probably.