TLDR Tech / Practical
Controlling access to documents in RAG AI applications
This article outlines three methods for managing document permissions in retrieval-augmented generation (RAG) AI applications. The discussed approaches include sending user tokens with each request, using cached access control lists before source verification, and enabling parent entities to filter based on local parent IDs. The author stresses the importance of security by default and careful consideration of identity boundaries for various request paths.
Read the full story