Hacker News Reader

This is an interesting approach! Why not offload PDF extraction to other frameorks that apply OCR pdf -> .md

I may explore this when I implement the vectordb implementation I started.