Storing Document Segments in RAG Systems: Embedded with Vectors or Using Separate Storage?

Data and Databases
1

Question

What is the optimal method for handling document fragments and metadata in AWS RAG setups? Should vectors and chunks be stored together or managed via distinct storage?

2

hey, im wonderin if combining them might save time, but a separate storage may boost flexibilty. what do u reckon about update speeds and cost impact? anyone tried a hybrid approach?

3

i lean towards seperat storage. it allows for better update control though it adds sync complexity. cost might be a bit higher but modularity really saves headaches when you need to update metadata without reindexing all vector data.