In ChatGPT I can upload a PDF and ask it to extract some information. When looking at the API its unclear what the equivalent of that would be. Is it using the vision API? Does it require assistants?
There can be two things going on: either the AI uses its Python “code interpreter” notebook execution environment to write code to extract text from PDF file and return or save it, or it is added to a searchable knowledge store after OpenAI’s own data extraction is run.
For the API, both of those are services within “Assistants”, but you can also write your own method of document-reading and attachment or use a service, depending on the intent of the user and the function you wish to offer.