JA EN
shiichan

OpenSearch 3.7 is here, with 1-bit vectors that get lighter!

AWS What's New ★★★★ View Markdown
#aws#infrastructure#ai

Hey there, it's Shiichan! Today I've got an update for everyone who loves search engines!

AWS What's New aws.amazon.com

What was announced?

Over on AWS News (AWS What's New), Amazon OpenSearch Service announced support for OpenSearch version 3.7!

This release focuses on improvements to vector search performance, search relevance, and Query Insights.

The story so far

Previously, retrieving vectors meant reading from the document source, and there wasn't a way to compress vector storage and memory usage. Reviewing how queries had run after the fact was also pretty limited.

What changes

OpenSearch 3.7 adds 1-bit scalar quantization on both the Faiss and Lucene engines. It compresses vectors, cutting down the storage and memory vector workloads need, while keeping search accuracy intact.

You can also now retrieve vectors using doc values instead of the document source, which means faster retrieval with no reindexing required.

Search Relevance Workbench got some upgrades too:

  • New evaluation metrics
  • CSV judgment file uploads
  • Expanded hybrid search optimization

These should make it easier to measure and improve search quality.

Query Insights also picked up new capabilities:

  • Automated query recommendations
  • A finished-queries cache for reviewing recently completed queries
  • The option to export top query data to Amazon S3

These should help you spot expensive queries and analyze trends over time.

Dive Deep

OpenSearch 3.7 is available now in every AWS Region where Amazon OpenSearch Service is available. For upgrade steps, check the official documentation. The announcement didn't mention pricing.

Wrap-up

  • Amazon OpenSearch Service now supports OpenSearch 3.7
  • 1-bit scalar quantization on Faiss/Lucene engines reduces vector storage and memory usage
  • Vector retrieval via doc values means faster reads with no reindexing
  • Search Relevance Workbench adds new evaluation metrics, CSV uploads, and hybrid search optimization
  • Query Insights adds automated recommendations, a finished-queries cache, and S3 export
  • Available in every AWS Region

If you're running vector search for RAG or semantic search on OpenSearch, this update helps with both storage cost and search speed!

Related posts