Can you run hybrid search in OpenSearch without a post-processing pipeline?

While you can technically run multiple queries, you need a post-processing pipeline to normalize and combine the disparate scoring scales (e.g., BM25 and neural sparse scores) into a single, accurately ranked result list.

Why use BM25 with neural sparse vector search?

Apache Iceberg is used to add a table layer on top of files in S3. It tracks schema, metadata, partitions, and snapshots so query engines can work with the data as a table instead of a loose set of raw files.

How to configure Kafka Connect Iceberg sink?

BM25 excels at exact keyword matching for specific terminology, while neural sparse vector search captures the semantic meaning of a query. Combining them ensures you capture both exact matches and broader conceptual intent.

What machine learning plugins are required for OpenSearch hybrid search?

To natively generate vector embeddings and handle neural queries, OpenSearch requires the ML Commons plugin. This plugin allows you to register and deploy the machine learning models needed for semantic search.

What plugins do you need to run a hybrid search on NetApp Instaclustr?

You’ll need the AI Search plugin enabled in order to run any semantic or hybrid searches. This plugin includes ML Commons and some useful ML tools for OpenSearch.

Instaclustr Blog

Featured blogs

How to create a hybrid search pipeline in OpenSearch®

Learn how to build a hybrid search in OpenSearch. This comprehensive guide covers ingest pipelines, index design, post-processing, and a Python implementation.

June 4, 2026

By Kassian Wren

Kafka® to Iceberg: Build a queryable data lake with ClickHouse®

Learn how to stream Kafka events to S3 with Iceberg and query them with ClickHouse in a practical step-by-step guide.

June 3, 2026

By Walt Ribeiro

Apache Iceberg explained: A better table format for modern data lakes

What is Apache Iceberg? Learn how Apache Iceberg improves data lakes with ACID transactions, schema evolution, and faster analytics.

June 2, 2026

By Walt Ribeiro

How to handle bad messages safely in Kafka Streams with KIP-1034 Dead Letter Queues

Kafka Streams dead letter queue explained: learn how KIP-1034 adds DLQ support for deserialization errors, raw byte preservation, metadata headers, and replay workflows.

June 1, 2026

By Paul Brebner

See all featured blogs

How to create a hybrid search pipeline in OpenSearch®

June 4, 2026 | By Kassian Wren

Learn how to build a hybrid search in OpenSearch. This comprehensive guide covers ingest pipelines, index design, post-processing, and a Python implementation.

June 4, 2026 | By Kassian Wren