cdk-s3-vectors 0.3.0

A CDK construct library for Amazon S3 Vectors. This construct simplifies the creation of vector buckets, vector indexes with full configuration options, and Amazon Bedrock knowledge bases using S3 Vectors as the underlying vector store.

cdk-s3-vectors

⚠️ Maintenance Notice: This library is intended as a temporary solution and will only be maintained until AWS CloudFormation and CDK introduce native support for Amazon S3 Vectors. Once native support is available, users are encouraged to migrate to the official AWS constructs.

Reference Documentation

Amazon S3 Vectors User Guide

Supported Languages

Language	Package
Python	cdk-s3-vectors
TypeScript	cdk-s3-vectors
Java	io.github.bimnett:cdk-s3-vectors
.NET	bimnett.CdkS3Vectors

Overview

Amazon S3 Vectors is in preview release and provides native vector storage and similarity search capabilities within Amazon S3.

This AWS CDK construct library provides high-level constructs for Amazon S3 Vectors, enabling you to create vector buckets, indexes, and knowledge bases for AI/ML applications.

The library includes three main constructs:

• Bucket: Creates S3 vector buckets with optional encryption
• Index: Creates vector indexes within buckets for similarity search
• KnowledgeBase: Creates Amazon Bedrock knowledge bases using S3 Vectors as the vector store

IMPORTANT: The constructs library version must be >= 10.0.5.

Examples

For complete, deployable examples in all supported languages, see the examples directory.

Bucket Construct Props

Name	Type	Description
vectorBucketName	string	The name of the vector bucket to create.
encryptionConfiguration?	EncryptionConfiguration	Optional encryption configuration for the vector bucket. Defaults to AES256.

Index Construct Props

Name	Type	Description
vectorBucketName	string	The name of the vector bucket to create the index in.
indexName	string	The name of the vector index to create.
dataType	'float32'	The data type of the vectors in the index.
dimension	number	The dimensions of the vectors (1-4096).
distanceMetric	'euclidean' | 'cosine'	The distance metric for similarity search.
metadataConfiguration?	MetadataConfiguration	Optional metadata configuration for the index.

KnowledgeBase Construct Props

Name	Type	Description
knowledgeBaseName	string	The name of the knowledge base to create.
knowledgeBaseConfiguration	KnowledgeBaseConfiguration	Vector embeddings configuration details.
vectorBucketArn	string	The ARN of the S3 vector bucket.
indexArn	string	The ARN of the vector index.
description?	string	Optional description of the knowledge base.
clientToken?	string	Optional idempotency token (≥33 characters).

Pattern Properties

Name	Type	Description
vectorBucketName	string	Returns the name of the created vector bucket
vectorBucketArn	string	Returns the ARN of the created vector bucket
indexArn	string	Returns the ARN of the created vector index
indexName	string	Returns the name of the created vector index
knowledgeBaseId	string	Returns the ID of the created knowledge base
knowledgeBaseArn	string	Returns the ARN of the created knowledge base

Default Settings

Out of the box implementation of the constructs without any override will set the following defaults:

Amazon S3 Vector Bucket

• Server-side encryption with Amazon S3 managed keys (SSE-S3) using AES256
• Least privilege IAM permissions for vector operations
• Custom resource handlers for bucket lifecycle management

Vector Index

• Support for float32 data type vectors
• Configurable dimensions (1-4096)
• Choice of euclidean or cosine distance metrics
• Optional metadata configuration for enhanced search capabilities

Amazon Bedrock Knowledge Base

• Integration with S3 Vectors as the vector store
• Configurable embedding models
• IAM role with least privilege permissions
• Support for various embedding data types and dimensions

Architecture

License

This project is licensed under the Apache License 2.0. See LICENSE file for details.