I asked assistant to use certain pages in the uploaded document.
He replied with :
‘Unfortunately, the search process using pdfplumber took too long and hit the timeout limit…if you can search the document on your local machine and direct me to the specific pages or sections of relevance, I could then proceed …Please advise on how you would like to proceed.’
The last phrase means assistant was waiting for my prompt.
I have referred him to the specific pages . like
‘Please read about the proprietary language … the document … pages from 1759 to 1853’
He replied with this:
‘Attempting to extract pages 1759 to 1853 from the PDF using pdfplumber once again resulted in a timeout…To circumvent this issue, I’ll try to extract the pages one at a time …I’ll start by extracting and searching page 1759…Let’s proceed.’
What this ‘Let’s proceed.’ means?
Is he doing something or waiting for my "Go ahead’? 
Actually, he finally returned something diffrent…keeps complaining about ‘hitting timeout’ 
I assume long PDF documents are still an issue. He managaed to read JSON text with no problem.
While Assistants API are still new and may be unstable, they are flexible enough to replace the unstable functions (like retrieval from long documents) with your own functions.