In a traditional database, you search by exact values: "SELECT * WHERE category = 'shoes'." But what if a user searches for "comfortable walking footwear for rainy days"? SQL can't handle that. Vector databases store embeddings alongside your data and let you query by similarity, "find the 10 items whose meaning is closest to this query." Pinecone, Weaviate, Chroma, Qdrant, and pgvector have emerged as purpose-built systems for this task, each optimized for storing millions or billions of vectors and searching them in milliseconds. They're the infrastructure layer that makes RAG systems, semantic search, and recommendation engines possible at production scale.