LLM Learning Hub

workspace/llm-course/home

Similarity Search

Similarity search is the operation of finding the k nearest vectors to a query embedding in a vector database, returning the most semantically similar stored passages.

What is Similarity Search?

Similarity search is the operation of finding the k nearest vectors to a query embedding in a vector database, returning the most semantically similar stored passages.

Similarity search is the operation of finding the k nearest vectors to a query embedding in a vector database, returning the most semantically similar stored passages.

Where is it used?

Every RAG pipeline performs similarity search before generation; FAISS `IndexFlatIP.search`, Qdrant `.search()`, and pgvector `<=>` operator implement this under the hood.

How to build it

Build a FAISS `IndexFlatL2(d)`, call `.add(vectors)` and `.search(query_vec, k=5)` to get indices and distances, then map indices back to original chunk text for the prompt.