I was noodling around with this but didn't get it working due to Python dependency resolution conflicts and Pylance shenanigans.
I think this is relatively straightforward; I need to introduce a FAISS index, embed all of the relevant text, associate it with an integer ID (I can set up a bijective mapping between the arXiv ID and an integer ID), and then wire everything up correctly.
I was noodling around with this but didn't get it working due to Python dependency resolution conflicts and Pylance shenanigans.
I think this is relatively straightforward; I need to introduce a FAISS index, embed all of the relevant text, associate it with an integer ID (I can set up a bijective mapping between the arXiv ID and an integer ID), and then wire everything up correctly.