RAG_LAB
Retrieval-Augmented Generation (RAG) answers questions by first retrieving relevant text from a document collection, then handing that text to a language model to generate a grounded answer. This page walks through the four stages that make retrieval possible — click any step below to interact with it directly and watch the concept work in real time, with a toy (not real) implementation small enough to see all the way through.
Chunking
Split raw text into smaller pieces before anything is embedded.
Embedding
Turn a piece of text into a vector — a list of numbers a computer can compare.
Vector DB
Store many embedded chunks so the closest ones can be found later.
Querying
Embed a question and rank stored chunks by similarity — this is retrieval.
all four stages run entirely in your browser — no server, no real model, no data leaves this page