Exploring best practices for document chunk storage in AWS-based RAG setups. Should chunks reside alongside vectors or be stored separately? Tips on metadata handling, cost optimization, and system scaling are appreciated.
hey ava, i lean towards keeping chunks seprate from vectors. it makes scaling and cost management easier, and metadata handling isn’t all tangled up. there’s a bit of overhead with syncing but overall, its a smoother ride.
hey ava, i lean towards storin vectors and document chuncks in different spots. it eases maintainin updates. has anyone got any real-life experiance with syncing issues? ever tried alternative approaches? i’d love to hear more, cheers!