I'm currently using OpenSearch as a vector store for Bedrock KB, but it's quite costly, especially since we have development, staging, and production environments with OpenSearch Serverless. I'm considering switching to S3 vector stores since we don't have an immediate need for a hybrid search option. Has anyone tried using S3 vector stores with Bedrock KB? If so, what has your experience been like?
2 Answers
I'm planning to test S3 against other options like Qdrant soon. My main goal is to see if I can build a solid LLM service on top of it. The simplicity of the Bedrock API is a huge advantage for me since I'm using Python with Boto3.
I'm curious to hear your results too! I'm looking into migrating next week myself, so I'll definitely share my findings once I make the switch.
Qdrant does seem intriguing, I've heard some great feedback about it. But for my setup, sticking with the Bedrock API really simplifies things.