Overview
DynamoDB vector search uses native vector-search capabilities to find semantically relevant records stored in Amazon DynamoDB.1 The source describes it as a way to keep operational document data and embedding vectors in one table, instead of maintaining a separate vector database alongside the document store.1 The stated benefits include less infrastructure sprawl, reduced synchronization overhead, and a lower risk of retrieval against stale embeddings.1
Architecture and Retrieval Flow
In the described agent design, a Bedrock agent routes a user request to a Lambda action group, and a retrieval Lambda generates a query embedding with Amazon Titan Text Embeddings V2.1 The Lambda then calls DynamoDB’s SearchVectors API to perform semantic retrieval.1 A separate Lambda action handles ordinary CRUD operations against the same table, so semantic search and direct document access use the shared data store.1 The table uses a composite primary key and represents stored embeddings as lists of numbers.1
Indexing and Maintenance
The vector index can use a moderate-cardinality partition key, such as category or tenant_id, according to the implementation guidance summarized in the source.1 The index must become fully active before it is ready for use, and initial ValidationException errors during index creation may be retryable.1 DynamoDB Streams can trigger an embedding pipeline when content changes, keeping the vectors synchronized with the stored content without requiring manual updates.1
Applications
The unified approach is especially relevant to knowledge-management systems that combine natural-language search with direct document lookup and modification.1 It offers an architectural alternative to splitting documents and embeddings across separate databases, with the tradeoff that table and vector-index design remain important implementation concerns.1
Limits and example parameters
“Maximum five vector indexes per table, with up to 4,096 dimensions each.”
“SearchVectors responses are limited to 16 MB and don’t support pagination.”
“DynamoDB table with vector index stores documents, metadata, and 1,024-dimension embeddings in one place.”
Related Pages
- Amazon DynamoDB — DynamoDB Vector Search is a capability of Amazon DynamoDB, which stores operational records and supports vector search.