vector-search-setup

What problem does it solve?

Provides a clear, end-to-end workflow to add vector and semantic search capabilities to Elasticsearch so applications can retrieve semantically relevant documents rather than relying solely on keyword matches.

Core Features & Use Cases

• Index Provisioning: Create indices with dense_vector mappings sized for your embedding model (e.g., 384 or 768 dimensions).
• Embedding Integration: Support for inference endpoints or ingest pipelines to generate embeddings at ingest time or from the application.
• Search & Validation: Index documents with vectors and run kNN or hybrid keyword+vector queries, with guidance on tuning size and num_candidates.
• Use Case: Add semantic search to a customer support knowledge base to surface relevant articles by meaning rather than exact keyword overlap.

Quick Start

Create a dense_vector-enabled index, ensure documents are indexed with embeddings, and run a sample kNN search to verify results.