ML//RAG//vector database

Stores vectors paired with resources (chunked to ~500 token paragraphs)

Stores vectors paired with resources (chunked to ~500 token paragraphs)

Core operation: "given a vector, find the N nearest vectors fast".

The math: cosine similarity between the question (embedded on-the-fly) and the stored vectors.

HNSW (Hierarchical Navigable Small World graphs): builds a graph where similar vectors are neighbors — hops through ~40 nodes instead of checking all 10M vectors, O(n) → O(log n)

IVF (Inverted File Index): clusters vectors into buckets first, searches only the closest ones — one drawer instead of the whole filing cabinet.

Implementations: Pinecone, Weaviate, Chroma, Qdrant, Supabase, FAISS.