Embeddings and vector retrieval for RAG: chunking documents, embedding models and similarity search

Keyword search finds documents that share words with the query. That fails when users describe the same idea with different words: a search for 'car won't start' may miss a passage about 'engine fails to turn over'. Vector search solves this by comparing meaning. An embedding model turns a piece of text into a vector, a list of hundreds or thousands of numbers, placed so that texts with similar meaning have vectors that are close together.