DISTRIBUTED MINDS
Interactive walkthrough

Anatomy of a RAG request — from question to grounded answer

Retrieval-Augmented Generation (RAG) gives an LLM access to facts it wasn't trained on by fetching relevant documents at request time and stuffing them into the prompt before the model ever generates a word. Press Send Request and watch a single query travel through the pipeline — or click any step below to jump straight to it.

Client sends the question Embedding model query → vector Vector store similarity search Context builder assembles the prompt LLM generates the answer
trace log
$ waiting for request…
payload inspector
Nothing on the wire yet. Press "Send request" to begin.