What is the Optimal Storage Approach for Document Chunks in RAG Systems?

Data and Databases
1

Query

Building an AWS RAG system: should document chunks be stored alongside vectors or in separate storage (e.g., S3)? How do you manage metadata links?

2

i lean towards keeping chunks with vectors using a db for metadata linking. s3 can be used for backup storage, though it sometimes add more overhead. i personally find the integrated approach reusable and simple to debug.