Amazon Bedrock Knowledge Bases

aws Search indexaws-bedrock-knowledge-base

Managed retrieval-augmented generation index over your data sources.

Provider: aws
Kind: Search index
Category: Data
Official docs: https://docs.aws.amazon.com/bedrock/latest/userguide/kb-chunking-parsing.html

Properties · 3 set by this component

property	type	values	required	kind default	this component
chunkSizeTokens · component property	integer	—	no	300	inherited
chunkingStrategy · component property	enum	default \| fixed-size \| hierarchical \| semantic \| none	no	default	inherited
documentLevelPermissions · component property	boolean	—	no	false	inherited
geoScheme · advanced	enum	none \| s2 \| h3 \| geohash \| quadtree	no	none	inherited
indexing	enum	sync \| async \| batch	no	async	▸ batch
knowledgeBaseType · component property	enum	managed \| customer-managed	no	—	inherited
managed	boolean	—	no	false	▸ true
rebuildCadenceHours · advanced	integer	—	no	—	inherited
replicas	integer	—	no	—	inherited
shardingStrategy · advanced	enum	none \| hash \| geo \| time \| prefix-range	no	none	inherited
structure	enum	inverted \| trie \| vector \| geo-cell	no	inverted	▸ vector
vectorStore · component property	enum	opensearch-serverless \| aurora \| neptune-analytics \| s3-vectors \| kendra-genai-index	no	—	inherited

Rules that apply · 1

Every rule whose selector covers this component's kind or its category. Whether a rule fires depends on the design: its properties, its edges and its groups.

Search index without an asynchronous feederror

derived-store-without-cdc-or-rebuild

A search index is derived data and needs an asynchronous feed from a stream, change data capture or an indexing worker. Without one it is written inline on the request path, and it can never be rebuilt after it drifts.

— this does not hold:
— the node has at least one asynchronous incoming edge from a stream, change-data-capture or worker

Use it in a design

nodes:
  - { id: "bedrock-knowledge-base", component: "aws-bedrock-knowledge-base" }

Other Search index products · 31

Alibaba Cloud Elasticsearchalibaba-elasticsearchalibabacloudsearch-index OpenSearchalibaba-open-searchalibabacloudsearch-index Amazon CloudSearchaws-cloudsearchawssearch-index Amazon Kendraaws-kendraawssearch-index Amazon OpenSearch Serviceaws-opensearchawssearch-index Azure AI Searchazure-ai-searchazuresearch-index Foundry IQazure-foundry-iqazuresearch-index AI Searchcloudflare-ai-searchcloudflaresearch-index Algoliaalgoliasearch-index Coveocoveosearch-index Elastic Cloudelastic-cloudsearch-index Elasticsearchelasticsearchsearch-index Geo cell indexgeo-cell-indexsearch-index Manticore Searchmanticoresearch-index Meilisearchmeilisearchsearch-index OpenSearchopensearchsearch-index Quickwitquickwitsearch-index Generic search indexsearch-indexsearch-index Apache Solrsolrsearch-index Sonicsonicsearch-index Sphinxsphinxsearch-index Trie indextrie-indexsearch-index Typesensetypesensesearch-index Vespavespasearch-index Cloud Talent Solutiongoogle-cloud-talent-solutiongcpsearch-index Vertex AI Searchgoogle-vertex-ai-searchgcpsearch-index AI Commerce Searchgoogle-vertex-ai-search-commercegcpsearch-index Vision API Product Searchgoogle-vision-product-searchgcpsearch-index IBM Cloud Databases for Elasticsearchibm-databases-for-elasticsearchibmsearch-index IBM Watson Discoveryibm-watson-discoveryibmsearch-index OCI Search with OpenSearchoci-opensearchocisearch-index

Source

1# yaml-language-server: $schema=../../schema/component.schema.json
2# Amazon Bedrock Knowledge Bases: managed RAG index that chunks and embeds source documents into a
3# vector store and serves retrieval queries. Per "How content chunking works for knowledge bases" the
4# default chunking strategy splits content into chunks of approximately 300 tokens on sentence
5# boundaries; data sources are loaded by ingestion jobs.
6schemaVersion: "1"
7type: "component"
8id: "aws-bedrock-knowledge-base"
9kind: "search-index"
10provider: "aws"
11docs: "https://docs.aws.amazon.com/bedrock/latest/userguide/kb-chunking-parsing.html"
12defaults:
13  managed: true
14  structure: "vector"
15  indexing: "batch"
16properties:
17  knowledgeBaseType:
18    type: "enum"
19    values: ["managed", "customer-managed"]
20  vectorStore:
21    type: "enum"
22    values: ["opensearch-serverless", "aurora", "neptune-analytics", "s3-vectors", "kendra-genai-index"]
23  chunkingStrategy:
24    type: "enum"
25    values: ["default", "fixed-size", "hierarchical", "semantic", "none"]
26    default: "default"
27  chunkSizeTokens:
28    type: "integer"
29    default: 300
30  documentLevelPermissions:
31    type: "boolean"
32    default: false

The same page as markdown, for an agent: /catalogue/components/aws-bedrock-knowledge-base.md