> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ardie.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# What is RAG?

> Retrieval-Augmented Generation allows AI to retrieve information from a knowledge base and generate responses grounded in real documents, improving accuracy, transparency, and control.

## What is Retrieval-Augmented Generation (RAG)?

Retrieval-Augmented Generation, usually shortened to **RAG**, is a way of combining your own knowledge with AI language models so answers are based on real information, not guesswork.

Instead of asking an AI to rely only on what it remembers from training, RAG allows the AI to **look things up in a trusted knowledge base first**, then generate a response using that information.

<Note>
  **In simple terms:** RAG lets AI read your documents before it answers.
</Note>

## Why RAG Exists

Large language models are very good at writing text, but they have three important limitations:

* They **do not automatically know** your documents
* They can **confidently say things that are wrong**
* They **cannot easily show** where information came from

RAG was developed to solve these problems by grounding AI responses in retrieved source material rather than memory alone.

## How RAG Works (Step by Step)

At a high level, RAG follows a simple process:

<Steps>
  <Step title="You create a knowledge base">
    Documents such as PDFs, reports, policies, or articles are uploaded and converted into a searchable format.
  </Step>

  <Step title="The system retrieves relevant information">
    When a question is asked, the system searches the knowledge base to find the most relevant passages.
  </Step>

  <Step title="The AI generates an answer">
    The AI uses the retrieved passages as context and writes a response based on them.
  </Step>

  <Step title="Sources can be shown">
    Because the answer is based on retrieved documents, it is possible to show where the information came from.
  </Step>
</Steps>

## How RAG is Different from a Normal Chatbot

<Columns cols={2}>
  <Card title="Standard Chatbot" icon="robot">
    * Relies mostly on its training data
    * Cannot see your private documents
    * May sound confident even when incorrect
  </Card>

  <Card title="RAG-Based System" icon="brain-circuit">
    * Actively retrieves relevant documents
    * Grounds answers in real sources
    * Can be updated by changing the knowledge base
    * Reduces hallucinations and factual errors
  </Card>
</Columns>

This difference is important for **research, policy, education, legal, and organisational use**, where accuracy matters.

## What Makes RAG Powerful

<CardGroup cols={2}>
  <Card title="Uses Up-to-Date Information" icon="clock">
    You can update or replace documents without retraining the AI model.
  </Card>

  <Card title="Improves Accuracy" icon="bullseye">
    Answers are constrained by retrieved evidence, which significantly reduces fabricated responses.
  </Card>

  <Card title="Supports Transparency" icon="eye">
    Retrieved sources can be inspected, cited, or audited.
  </Card>

  <Card title="Separates Knowledge from AI" icon="layer-group">
    The same knowledge base can be reused across websites, internal tools, chat interfaces, research workflows, and automated systems.
  </Card>
</CardGroup>

This separation is a core advantage highlighted in the original RAG research.

## What RAG is Not

<Warning>
  RAG is **not**:

  * A chatbot product
  * A single AI model
  * A replacement for human expertise

  RAG is a **method** for connecting AI systems to structured knowledge in a controlled way.
</Warning>

## Why Ardie Uses RAG

Ardie is built around the idea that knowledge should be **reusable, governable, and independent** of any single interface.

Using RAG allows Ardie to:

* Turn documents into durable knowledge bases
* Serve that knowledge through APIs
* Let users connect their own AI tools
* Enable sharing or selling access without exposing infrastructure

**RAG is the technical foundation that makes this possible.**

## When RAG is the Right Choice

RAG is especially useful when:

| Scenario                                 | Why RAG Helps                            |
| ---------------------------------------- | ---------------------------------------- |
| **Accuracy matters**                     | Answers are grounded in real documents   |
| **Information must be traceable**        | Sources can be cited and audited         |
| **Knowledge changes over time**          | Update documents without retraining      |
| **Same information used in many places** | One knowledge base, many applications    |
| **You want control over AI responses**   | Constrain what the AI can and cannot say |

This is why RAG is now widely used in research systems, enterprise tools, and knowledge platforms.

## Learn More

<CardGroup cols={2}>
  <Card title="Knowledge Bases" icon="database" href="/concepts/knowledge-bases">
    Learn how Ardie structures and stores knowledge for retrieval.
  </Card>

  <Card title="API Reference" icon="code" href="/api-reference/introduction">
    See how to query knowledge bases programmatically.
  </Card>
</CardGroup>