Clicked Gallery

What is Retrieval-Augmented Generation (RAG)?

Highlighted from a real engineering doc. Explained by Clicked.

Used in a sentence

Instead of fine-tuning the base LLM, we implemented a RAG pipeline to pull direct facts from internal company files.

The reader highlighted one word in the docs. Clicked made the technical term “RAG pipeline” easy to understand:

Explained in three depths

●○○

Overview

▾

RAG is a way to make an AI model answer from specific documents. Instead of relying only on what it learned in training, the system first retrieves the relevant files and hands them to the model as context for its answer.

●●○

Detail

▾

The pipeline has two stages. Retrieval: the question is used to search a document store — company files, manuals, a knowledge base — and the most relevant passages are pulled out. Generation: those passages are inserted into the model's prompt, and it answers using them. The appeal over fine-tuning is that training data is frozen and generic, while RAG adds fresh or private knowledge without retraining, and you can see which passages fed the answer. The hard limit: answer quality is capped by retrieval quality, so if the search step pulls the wrong passages, the model confidently summarizes the wrong thing.

●●●

Analogy

▾

Two ways to staff a helpdesk for your company's policy manual. Option one: train an employee for months until the manual lives in their head, which works until the manual changes. Option two: hire a quick reader who keeps the manual on the desk and flips to the right page for each question.

AI explanations may contain errors · Not professional advice

Same facts, different vibe — Slang mode 😎

●○○

Overview

▾

RAG is giving the AI a cheat sheet instead of making it memorize the whole library. Question comes in, the system grabs the right documents, slides them across the table, and the model answers off those. Swapping in new docs takes seconds, and nothing gets retrained. 😎

●●○

Detail

▾

Two-step move: search first, answer second. Why everyone's doing it: retraining a model on your company's files is expensive and outdated by Friday, while swapping documents in a folder is free. Bonus: you can check the receipts, because the answer came from passages you can actually read. The catch nobody puts on the slide: if the search grabs the wrong pages, the model doesn't get confused — it confidently explains the wrong pages. Retrieval is the whole game. 😎

●●●

Analogy

▾

It's the new hire who sounds like a ten-year veteran, because you slid them the internal docs under the table before the meeting. Take the docs away and they know nothing about your company. Hand them the wrong docs and they'll confidently present those instead.

AI explanations may contain errors · Not professional advice

Formal definition — The same term, explained the usual way

Retrieval-augmented generation is an architecture in which a language model's input is augmented with documents retrieved from an external corpus at inference time, typically via dense vector similarity search over an indexed store. By conditioning generation on retrieved context rather than solely on parametric knowledge, RAG systems enable domain adaptation without model retraining, support source attribution, and mitigate — without eliminating — hallucination; end-to-end performance remains bounded by retrieval precision and recall.

Want Clicked to explain terms like “RAG” directly in your browser — including on PDFs?

Add to Chrome — Free

50 free Explanations · No credit card required

More from the gallery

What are Vector Embeddings?

Meaning turned into coordinates — how machines find what's similar without matching a single word.

What does an Idempotent API mean?

Why ten clicks charge you once — the property that makes retries safe.

How to actually read a research paper

Skim smart, break the jargon wall, and keep what you learned — one session, done right.

What are “synergies”?

Two merged companies earning more together than apart — the claim, the math, and the donut-shop version.

What is Technical Debt?

Shortcuts and aging systems that bill the future — principal, interest, and repayment.

What is a Zero-Knowledge Proof (ZKP)?

Prove something is true without revealing why — the cryptography behind private verification.