OpenSearch 2.4.0 has been officially released. This release brings support for Windows, security analytics, new geospatial capabilities, and several upgrades for search, analytics, and observability use cases, the announcement reads.

OpenSearch 2.4.0, the last minor update for 2022, extends the OpenSearch project toolkit with more than 25 features and enhancements across a range of use cases for geospatial, search, security analytics, observability, and more. Support for Windows environments and other first-of-its-kind features are now ready for production use.

download link

• Deploy directly on Windows

The OpenSearch 2.4.0 release fulfills a long-standing community request for a Windows x64 version of OpenSearch and OpenSearch Dashboards. This is the first time that developers can deploy the OpenSearch suite directly in their Windows environment.

The distribution is provided in ZIP format, includes support for all OpenSearch plugins (except the performance analyzer), and uses a common UI with other distributions. See the reference documentation for details.download link

• Use experimental semantic search tools

For the first time, OpenSearch users can run automated semantic searches on their documents, enhancing search results with additional meaning and context provided by machine learning models. The experimental set of new features in 2.4.0 allow users to combine traditional BM25 lexical search with deep learning-driven semantic search, providing new ways to tune search queries to improve relevance, recall, and accuracy.

Version 2.4.0 introduces the Model Serving Framework, an experimental feature that allows users to unlock the ability to upload their own ML models to OpenSearch. In this release, the framework supports integrating text embedding models to enhance semantic search. This release is the first step toward making ML on OpenSearch easier to integrate and operate to support various use cases such as visual and audio search, natural language processing (NLP)-driven analytics, and others.

• Support querying data from a specific point in time

Another new feature of 2.4.0 is support for time search (Point in Time, PIT). Previously, it was not possible to run queries on the same dataset and produce the same results; as the underlying data changed, so did the search results, making comparing and analyzing query results challenging. It is now possible to specify a PIT and run different queries against the same fixed data set against that PIT to obtain consistent results. This feature is accessible through the API in 2.4.0 and is planned to be enabled in a future version of the UI.

See the release announcement and Release Notes for details.