Skip to content

SYSTEM Cited by 1 source

Amazon Bedrock Knowledge Bases

Amazon Bedrock Knowledge Bases is an AWS managed service for building Retrieval-Augmented Generation (RAG) applications on top of Amazon Bedrock foundation models. A Knowledge Base takes raw documents from a customer-owned data source (commonly S3), chunks + embeds them using a Bedrock embedding model, stores the embeddings in a vector store of the customer's choice, and exposes a retrieval API that returns the most-relevant chunks for a query — typically wired into a Bedrock model call as context.

As of 2025-07-16, the selectable vector stores include Amazon S3 Vectors (new), alongside OpenSearch Serverless, Aurora, Pinecone, Redis Enterprise Cloud, and MongoDB Atlas.

(Source: sources/2025-07-16-aws-amazon-s3-vectors-preview-launch)

S3 Vectors integration (2025-07-16)

In the Bedrock console, when creating a Knowledge Base:

"In Step 3, you can choose the Vector store creation method either to create an S3 vector bucket and vector index or choose the existing S3 vector bucket and vector index that you've previously created."

The same integration is surfaced inside Amazon SageMaker Unified Studio's chat-agent authoring experience — S3 Vectors shows up as a Vector store option when a developer adds a Bedrock Knowledge Base component to a chat-agent app.

Position in the RAG stack

Raw docs (S3) → chunk → embed (Bedrock model, e.g. Titan V2)
                               │
                               ▼
                       Vector store (choose one)
                         ├─ S3 Vectors  ← new
                         ├─ OpenSearch Serverless
                         ├─ Aurora pgvector
                         └─ Pinecone / Redis / MongoDB

                              ▼
       Retrieve-by-query → context → Bedrock model → answer

Knowledge Bases owns the ingestion pipeline, chunking, embedding invocation, and retrieval API; the vector store is a pluggable backend.

Why S3 Vectors as a backend

AWS's pitch (Channy Yun, 2025-07-16):

"You can use S3 Vectors in Amazon Bedrock Knowledge Bases to simplify and reduce the cost of vector storage for RAG applications."

The combination is the cheapest paved RAG path on AWS at preview launch: storage-tier cost for embeddings, no provisioned vector cluster, one-click setup in the Bedrock console.

Caveats

  • The 2025-07-16 announcement is a feature addition to an existing service; Knowledge Bases itself was GA before. This page is stubby because the source focuses on the S3 Vectors integration.
  • No Bedrock-KB-specific numbers (query latency, chunk-size defaults, re-embedding triggers) are given in this source.

Seen in

Last updated · 200 distilled / 1,178 read