What is RAG? Retrieval-Augmented Generation Explained Simply

You've probably seen demos of AI answering questions from a PDF or a company knowledge base. That's RAG — Retrieval-Augmented Generation. It's one of the most useful patterns in modern AI, and it's simpler than it sounds.

In this article, I'll break it down exactly — what the problem is, how RAG solves it, and the code pattern behind every "chat with your PDF" app you've seen.

The Problem RAG Solves

LLMs like Gemini or GPT know a lot — but they were trained on data up to a certain date, and they don't know your data. If you ask ChatGPT about your company's internal docs, it has no idea. It'll either hallucinate an answer or say it doesn't know.

How RAG Actually Works

There are two phases: indexing (one-time setup) and querying (every time a user asks something).

Phase 1 — Indexing Your Documents

Before a user can ask anything, you need to process your documents into a searchable format:

1. Extract text from the PDF or document
2. Split it into small chunks (e.g. every 500 words)
3. Convert each chunk to an embedding — a list of numbers that captures the meaning
4. Store those embeddings in a vector database

// Step 1: Extract text from PDF
const pdfText = await extractText(pdfBuffer);

// Step 2: Split into chunks
const chunks = splitIntoChunks(pdfText, {
  size: 500,
  overlap: 50
});

// Step 3: Generate embeddings
const embeddings = await Promise.all(
  chunks.map(chunk => gemini.embedContent(chunk))
);

// Step 4: Store in vector database
await vectorDB.upsert(
  chunks.map((chunk, i) => ({
    id: generateId(),
    vector: embeddings[i],
    text: chunk,
  }))
);

Phase 2 — Answering a Question

When a user asks a question, you don't send the entire document to the AI. Instead:

async function answerQuestion(userQuestion) {
  // 1. Embed the question
  const qEmbed = await gemini.embedContent(userQuestion);

  // 2. Find the most relevant chunks
  const results = await vectorDB.search({
    vector: qEmbed,
    limit: 3,
  });

  // 3. Build the prompt with context
  const prompt = `
    Use the following context to answer.
    Context: ${results.map(r => r.text).join('\n\n')}
    Question: ${userQuestion}
  `;

  // 4. Get the answer
  return await gemini.generateContent(prompt);
}

RAG vs Fine-Tuning