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

# Ask Your Own Documents

> Upload files, then ask the AI questions about what's inside them.

## What is this?

Normally the AI answers from what it already knows. But you can also **give it your own documents** — a PDF, a report, meeting notes — and then ask questions about *those*. The AI reads your files and answers based on what's actually in them.

The technical name for this is **RAG** (retrieval-augmented generation), but you don't need to remember that. Just think of it as:

> **"Read this, then answer my questions about it."**

## When is this useful?

* "What does our return policy say about damaged items?" (from a policy PDF)
* "Summarize the key decisions from these meeting notes."
* "Find every mention of pricing in this contract."
* "What are the main points across these 5 documents?"

Instead of reading everything yourself, you ask, and the AI finds the answer inside your files.

## How to use it

<Steps>
  <Step title="Upload your document">
    Go to the place where you add knowledge — this is usually inside an **App** (on its **Knowledge** or **Data** section) or your **Storage** area. Click **Upload** and choose your file.

    <Frame caption="Uploading a document so the AI can read and answer questions about it.">
      <img src="https://mintcdn.com/sundaypyjamas/s9bw5Uhxs58-xnp0/images/user-guide/knowledge-upload.png?fit=max&auto=format&n=s9bw5Uhxs58-xnp0&q=85&s=3db674ed609680de289d29ae866e7d13" alt="An upload area for adding documents to the AI's knowledge" width="1440" height="900" data-path="images/user-guide/knowledge-upload.png" />
    </Frame>
  </Step>

  <Step title="Wait for it to be ready">
    After uploading, the AI needs a moment to "read" the document (you might see a status like *processing* or *ready*). This usually takes a few seconds to a minute.
  </Step>

  <Step title="Ask a question">
    Now ask about the content in plain words:

    > *According to the uploaded handbook, how many vacation days do new employees get?*

    <Frame caption="Asking a question and getting an answer drawn from your document.">
      <img src="https://mintcdn.com/sundaypyjamas/s9bw5Uhxs58-xnp0/images/user-guide/knowledge-answer.png?fit=max&auto=format&n=s9bw5Uhxs58-xnp0&q=85&s=7a993e96cde8278cb6ebcff45e4aad10" alt="A chat answer that cites the uploaded document" width="1440" height="900" data-path="images/user-guide/knowledge-answer.png" />
    </Frame>
  </Step>

  <Step title="Check the source">
    Good answers often show **where** they came from in your document, so you can trust and verify them. Look for a citation or a highlighted snippet.
  </Step>
</Steps>

## Tips for good results

* **Upload clean files.** Clear, text-based PDFs work best. Very blurry scans are harder to read.
* **Ask specific questions.** "What's the refund window?" beats "tell me about refunds."
* **Add more documents** for broader questions — the AI can look across all of them.
* **Ask for the source.** "Which section is that from?" helps you double-check.

<Note>
  Your uploaded documents stay in your workspace. They're used to answer *your* questions, not shared publicly.
</Note>

## Next step

<Card title="Build your own AI app" icon="wand-magic-sparkles" href="/user-guide/build-an-app">
  Package all of this — a model, instructions, and your documents — into a reusable helper.
</Card>
