Overview
OpenSearch k-NN Vector search provides scalable vector similarity search (k-nearest neighbors) with support for approximate and exact methods, sparse neural embeddings, and hybrid combinations with traditional keyword search.
Key Features:
- Approximate and exact k-NN search modes with automatic method/engine selection
- Neural sparse search using sparse embeddings and rank-features indexes
- Hybrid search pipeline combining vector semantic scores with traditional keyword matching and filters
Use Cases:
- Semantic document or semantic image retrieval at scale using ANN for low-latency search
- Filtered or custom-scored nearest-neighbor queries on smaller datasets with exact or Painless-scripted searches
- Enhanced relevance in search applications by combining vector semantics with keyword matching and aggregations
Benefits:
- Scales efficiently for large vector datasets with performance-optimized approximate methods
- Flexible: supports dense vectors, sparse neural embeddings, and complex scoring/filtering workflows
- Improves relevance by unifying semantic and lexical search signals in a configurable pipeline