Only with certificate: Long-Term Support for OpenSearch

Contents

At its European home conference, the OpenSearch Foundation announced a Long-Term Support (LTS) version of OpenSearch. Initially, the current version 3.6 and OpenSearch 2.19 as the final release of the previous major version will receive LTS status. In the future, there will be an LTS variant per major version, which will receive at least 18 months of support.

Certified third parties take over LTS program

To prevent vendor lock-in, users can decide for themselves which provider will provide the LTS service. While there are service providers that offer OpenSearch, certification by the OpenSearch Foundation is required for the provision of long-term support. Companies must demonstrate reliable deadlines for providing patches, upgrade options, and basic work in the community.

Specifically, the OpenSearch Foundation promises that security patches and bug fixes will be available within a maximum of 60 days and that upgrades can be applied during operation without having to shut down OpenSearch services. Currently, there are three certified providers: Bigdata Boutique, Eliatra, and Resolve.

OpenSearch 3.6: AI agent builds search applications autonomously

The main innovations in OpenSearch 3.6 are in the area of agent-based AI. With the new OpenSearch Launchpad, users will be able to create their own search applications within minutes without needing technical OpenSearch knowledge. Instead, users can submit their requirements in documents or via chat to the platform, which then independently sets up a local application with a graphical interface.

Launchpad is the first capability of the new OpenSearch Agent Skills. They contain and link the technical knowledge for building search applications. All tools contained within them are to be integrated with various development environments via the Model Context Protocol (MCP), including Claude Code, Cursor, and Kiro.

Connection to external language models

Another innovation is the OpenSearch Agent Server. It is part of OpenSearch's AI agent infrastructure. The Agent Server forwards requests, communicates with OpenSearch via MCP, and can communicate with LLM providers such as AWS Bedrock or Ollama. The technical prerequisites are Python 3.12 or newer, OpenSearch 2.x or newer, and a connection to the preferred language model.

The Agent Server includes the also new, still experimental OpenSearch Relevance Agent. With it, users can evaluate search usage behavior using natural language instructions, generate data-driven hypotheses, and evaluate improvement suggestions. To achieve this, the Relevance Agent orchestrates communication with specialized AI agents and is also integrated into the OpenSearch cockpit via the chat interface. In addition to forwarding messages, it evaluates the received responses and outputs a summary in the cockpit.

The new features of OpenSearch 3.6 can be found in the developer blog of the project. After a license dispute, OpenSearch split off from Elasticsearch in 2021 as an open-source fork. AWS operated the project until its transfer to the Linux Foundation in 2024.

(vbr)