OpenSearch

What’s new in OpenSearch 3.8

OpenSearch 3.8 introduces significant performance and functionality upgrades across search, AI, and observability workflows. Key enhancements include up to 4x faster vector ingesti…

Agentic relevance tuning: Letting LLM agents do the search engineering work

Introduced at OpenSearchCon Europe 2026, Agentic Relevance Tuning (ART) is a new framework that uses specialized LLM agents to automate the entire search relevance lifecycle in Ope…

Connect OpenSearch to private ML endpoints

OpenSearch ML Commons enables secure, low-latency connections to machine learning endpoints hosted within private VPCs or behind firewalls. While self-managed OpenSearch can be con…

The forbidden index: Building privacy-preserving search with OpenSearch

At OpenSearchCon Europe 2026, Unnati Mishra and Akshat Khanna presented a method for handling search privacy during the data ingest phase rather than through retroactive access con…

OpenSearchCon North America 2026 to Showcase Five Years of Innovation Powering Enterprise Search, Observability and Analytics

The OpenSearch Software Foundation has announced the schedule for OpenSearchCon North America 2026, taking place from September 22-24 in San Jose, California. The event will featur…

Scaling Kubernetes Workloads with the OpenSearch KEDA Scaler

The new OpenSearch KEDA scaler, introduced in KEDA 2.20, allows Kubernetes workloads to scale directly based on any arbitrary query result from OpenSearch, such as error rates or l…

How a 60% speedup on Lucene vector search came from the CPU, not the algorithm

At OpenSearchCon Europe 2026, Amazon’s Shubham Chaudhary demonstrated how bypassing JVM limitations to implement native SIMD architecture achieved a 60% reduction in Lucene vector …

Code that ships: The OpenSearch Agent Skills Hackathon

The OpenSearch Project has launched the Agent Skills Hackathon, inviting developers to create production-ready AI skills for search and observability applications. Running from Jul…

OpenSearch at five: Building sustainability in open source infrastructure

Having surpassed 2 billion downloads, the OpenSearch Project demonstrates the long-term viability and structural durability of a vendor-neutral, community-driven governance model u…

Retrieve vectors 5x faster with docvalue_fields in OpenSearch

OpenSearch 3.7 introduces the ability to retrieve vectors via `docvaluefields` rather than the traditional `source` path, allowing for direct reading from the columnar store. This …

OpenSearch Ambassador applications are now open year-round

The OpenSearch Ambassador Program is transitioning to a rolling application model, allowing community members to apply year-round rather than during specific windows. Applications …

Intent is the thing: Dom Couldwell’s case for a different kind of enterprise search

At OpenSearchCon Europe 2026, Dom Couldwell argued that effective enterprise search must move beyond simple keywords to address user intent across three levels: query, journey, and…