The vector engine showdown nobody was telling the truth about
At OpenSearchCon Europe 2026, Fernando Rejon Barrera presented a study comparing Lucene’s HNSW, Faiss, and jVector under realistic production workloads rather than isolated benchma…
Back to all articles
Category
Blog
66 articles · Page 2 of 6
No more zero results with vector query relaxing in OpenSearch
This article discusses how vector search in OpenSearch acts as a semantic safety net to prevent zero-result pages when traditional lexical search strategies fail to match user inte…
From user to contributor to maintainer: My journey with OpenSearch search relevance
This article chronicles an individual's journey from being a regular OpenSearch user at Cloudera to becoming a maintainer of the project's search relevance plugin. By starting with…
Vectors vs. hallucinations: How OpenSearch keeps GenAI grounded
At OpenSearchCon Europe 2026, Laysa Uchoa argued that generative AI hallucinations are primarily a retrieval problem rather than an inherent limitation of Large Language Models. Sh…
Introducing LLM as a Judge: Scaling search relevance evaluation with AI
OpenSearch 3.5 introduces "LLM as a Judge," a new feature that leverages large language models to automatically and at scale evaluate search result relevance. This approach provide…
Bringing intelligence to OpenSearch: Introducing the OpenSearch agent server
The OpenSearch agent server is an experimental multi-agent orchestration platform that enables users to build and deploy specialized AI agents for specific tasks within OpenSearch.…
Explore OpenSearch 3.7
OpenSearch 3.7 introduces significant enhancements for observability and AI-powered search, focusing on unifying telemetry and accelerating performance. Key updates include native …
OpenSearch in the Agentic Era: What Carl Meadows announced at OpenSearchCon Europe 2026
At OpenSearchCon Europe 2026, Carl Meadows announced three key innovations—OpenSearch Launchpad, the Relevance Agent, and a unified Observability Stack—to position OpenSearch as a …
How a single space broke OpenSearch backups—and how Aiven fixed it
Aiven engineers identified a bug where a single space in a KNN field name caused OpenSearch snapshot backups to fail because the snapshot repository rejected the resulting filename…
Asymmetric model support: Optimizing semantic search for queries and documents
OpenSearch 3.5 introduces support for asymmetric embedding models, which enhance semantic search relevance by applying distinct encoding strategies to short queries and long, infor…
Unpacked at Open Source Summit NA 2026: Inside OpenSearch’s Massive Leap into the Agentic Era
OpenSearch 3.6 introduces a significant shift from traditional search to supporting autonomous, agentic AI architectures through new features like pre-packaged Agent Skills and nat…
Agentic SDLC: How OpenSearch accelerates engineering with its own engine
The OpenSearch team has developed four AI agents—Atlas, Ralph, Nitro, and Sentinel—to automate repetitive tasks throughout the software development lifecycle, ranging from knowledg…