Vector Databases 101
Traditional databases find exact matches. Vector databases find similar meanings. This changes everything about how AI retrieves information.
The problem: You search a traditional database for "joyful" and it returns nothing — because the data contains "happy," not "joyful." A vector database finds "happy" because it understands they mean the same thing. Type a query below to see the difference.
How vector databases work under the hood: They use special index structures (like HNSW — Hierarchical Navigable Small World graphs) to search billions of vectors in milliseconds. Instead of scanning every row, they navigate a graph of connected vectors to find the nearest neighbors quickly.
Popular Vector Databases
Pinecone
Fully managed, serverless vector DB. Great for getting started fast. Auto-scales.
Weaviate
Open-source with built-in vectorization. Supports hybrid search natively.
Chroma
Lightweight, open-source, runs locally. Perfect for prototyping RAG apps.
pgvector
PostgreSQL extension. Use vectors in your existing Postgres DB. Supabase supports it.
Qdrant
Rust-based, high-performance. Great filtering + payload support.
Milvus
Enterprise-grade, handles billions of vectors. Used by major tech companies.
Match the Vector Database
Tap one on the left, then its match on the right
How a Vector Search Works — Put the Steps in Order
Arrange these steps in the correct sequence for a vector similarity search
1Compare query vector against all stored document vectors using cosine similarity
2Rank results by similarity score
3Convert user query to an embedding vector
4Return the top-K most similar document chunks