Vector Search

Definition

Vector search is a retrieval method that finds items by comparing the numerical similarity of their vector embeddings rather than matching exact keywords.

Overview

Vector search works by converting text, images, or other data into high-dimensional numeric vectors — embeddings — that capture meaning rather than surface form. A query is embedded the same way, and the system finds the stored vectors closest to it using a distance metric such as cosine similarity or dot product. This is fundamentally different from keyword indexing, because two pieces of content can be considered similar even if they share no words in common. At scale, exact nearest-neighbor comparison across millions of vectors is too slow, so production systems use approximate nearest neighbor (ANN) indexes such as HNSW or IVF to trade a small amount of accuracy for large speed gains. Purpose-built vector databases like Pinecone and general-purpose engines like Elasticsearch both offer vector search capabilities alongside traditional indexing. Vector search is the retrieval backbone of modern semantic search systems and of retrieval-augmented generation (RAG) pipelines, where relevant documents are fetched and fed into a large language model's context window before it generates an answer. The Retrieval-Augmented Generation course covers building these pipelines end to end, and the blog post Vector Databases Explained walks through the underlying storage layer in more depth.

Key Concepts

• Compares meaning-based numeric embeddings instead of exact text matches
• Uses distance metrics like cosine similarity, dot product, or Euclidean distance
• Relies on approximate nearest neighbor (ANN) indexes for speed at scale
• Works across text, images, audio, and other embeddable data types
• Often combined with metadata filters for hybrid search
• Core retrieval layer for RAG pipelines and AI assistants
• Supported by dedicated vector databases and by extensions to traditional databases

Use Cases

Frequently Asked Questions

Related in AI Fundamentals