How to create an OpenSearch (Vector Database) On Demand instance for Bedrock Knowledge Base

0

I am trying to create a Knowledge base to be used with AWS Bedrock for a very Proof of Concept solution. The knowledge base needs some form of Vector Database to run off of, which one of the options is "Vector engine for Amazon OpenSearch Serverless"

I already setup using the default serverless option and ended up incurring high billing costs, but the representative told me that there is an option when creating your OpenSearch Vector DB to run it in full On Demand mode rather than serverless. I have tried searching all across AWS for how to specifically do this in console and cannot find anything.

Could someone give instructions for how to setup an OpenSearch Vector engine in On Demand mode? Since I am doing testing I do not want to have a server that will be always running from the moment I start it up, as I will only need to run it for 1-2 hour periods per week as I am testing and trying to build out my solution. Thank you! And if

2 Answers
0

I wanted to give a partial solution I found / more like an explanation. It seems like AWS does not let you setup a knowledgebase using an On Demand Open Search Vector DB... When creating the knowledgebase, the option for using OpenSearch even says Serverless Vector DB (implying you have to use the Serverless mode only).

I would still like to know how to setup an OpenSearch Database in On Demand mode, as I was told over the phone that that is possible... But for my Knowledge base solution I ended up going with another Vector Database service instead of OpenSearch.

answered 2 months ago
0

Hi,

Currently Amazon Bedrock Knowledgebase does not support OpenSearch in On Demand mode. Currently it supports only the following vector stores.

  • Amazon OpenSearch Serverless
  • Amazon Aurora
  • Pinecone
  • Redis Enterprise Cloud

https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base-setup.html

profile pictureAWS
answered 2 months ago

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions