Amazon Bedrock Knowledge Bases now supports hybrid search for Aurora PostgreSQL and MongoDB Atlas vector stores

Amazon Bedrock Knowledge Bases now extends support for hybrid search to knowledge bases created using Amazon Aurora PostgreSQL and MongoDB Atlas vector stores. This capability, which can improve relevance of the results, previously only worked with Opensearch Serverless and Opensearch Managed Clusters in Bedrock Knowledge Bases.

Retrieval augmented generation (RAG) applications use semantic search, based on vectors, to search unstructured text. These vectors are created using foundation models to capture contextual and linguistic meaning within data to answer human-like questions. Hybrid search merges semantic and full-text search methods, executing dual queries and combining results. This approach improves results relevance by retrieving documents that match conceptually from semantic search or that contain specific keywords found in full-text search. The wider search scope enhances result quality, particularly for keyword-based queries.

You can enable hybrid search through the Knowledge Base APIs or through the Bedrock console. In the console, you can select hybrid search as your preferred search option within Knowledge Bases, or choose the default search option to use semantic search only. Hybrid search with Aurora PostgreSQL is available in all AWS Regions where Bedrock Knowledge Bases is available, excluding Europe (Zurich) and GovCloud (US) Regions. Hybrid search with Mongo DB Atlas is available in the US West (Oregon) and US East (N. Virginia) AWS Regions. To learn more, refer to Bedrock Knowledge Bases documentation. To get started, visit the Amazon Bedrock console.