How does ai apps handle memory

I see a lot of app make use of GPT for there applications and got me thinking on how all these apps handle Multi document search. how do you pass the right data to the applications. I understand that the most common approach is to use a Vector db. But it does not seem to be the most scaleable approach.