Vector Databases

Vector databases are specialized databases designed to store and search high-dimensional vectors (embeddings). They're essential for RAG, semantic search, and recommendation systems.

What are Embeddings?

Embeddings are numerical representations of text, images, or other data. Similar items have similar vectors, enabling semantic search.

"cat" → [0.2, 0.8, 0.1, ...]  (768 dimensions)
"dog" → [0.3, 0.7, 0.2, ...]  (similar to cat!)
"car" → [0.9, 0.1, 0.8, ...]  (different)

Popular Vector Databases

Pinecone

Fully managed, serverless. Easy to use, scales automatically.

Best for: Production apps, no ops

Weaviate

Open-source, GraphQL API. Built-in vectorization.

Best for: Flexibility, self-hosting

Chroma

Lightweight, Python-first. Great for prototyping.

Best for: Local dev, quick start

Qdrant

Rust-based, high performance. Advanced filtering.

Best for: Speed, complex queries

Vector Search Example

Simplified example of similarity search using cosine similarity.

python

Output:

Click "Run Code" to see output

Key Concepts

ANN (Approximate Nearest Neighbors): Fast search algorithm (HNSW, IVF)
Indexing: Organize vectors for efficient retrieval
Metadata Filtering: Combine vector search with traditional filters
Hybrid Search: Combine semantic (vector) + keyword (BM25) search