Semantic Search

Semantic search goes beyond traditional keyword matching by understanding the meaning behind search queries. It uses vector embeddings to represent both queries and documents as points in high-dimensional space, finding results based on conceptual similarity.

For example, a semantic search for "how to deploy containers" would also find documents about "Kubernetes deployment," "Docker orchestration," and "container management" — even if they don't contain the exact search terms.

The typical semantic search pipeline involves: embedding documents during indexing, embedding the query at search time, performing approximate nearest neighbor search in a vector database, and optionally re-ranking results for relevance.

Hybrid search combines semantic search with traditional keyword search (BM25) for the best of both worlds. This is particularly effective when users mix conceptual queries with specific terms like error codes or product names.