> ## Documentation Index
> Fetch the complete documentation index at: https://docs.a2v2.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Upload Files

> Train your agent on PDF, DOCX, and TXT documents.

Uploading files is the most common way to train an agent. Use it for manuals,
policies, product guides, reports, and any document your team already has.

## Prerequisites

* An agent (see [Create an agent](/building/create-agent)).
* A supported file: **PDF**, **DOCX**, or **TXT**.
* Enough credits to process the document (large files cost more — see
  [Credits](/concepts#credits)).

## Upload a file

<Steps>
  <Step title="Open the Knowledge Base">
    In the agent sidebar, go to **Expertise → Knowledge Base**.
  </Step>

  <Step title="Open the Files tab">
    Select **New Knowledge Item**, then open the **Files** tab.
  </Step>

  <Step title="Select your document">
    Pick a PDF, DOCX, or TXT file from your computer. You can add more than one.
  </Step>

  <Step title="Start processing">
    Confirm the upload. The source appears in your list with a **Processing**
    status while it's indexed.
  </Step>

  <Step title="Wait for Completed">
    Once the status turns **Completed**, the document is live and your agent can
    answer from it.
  </Step>
</Steps>

<Frame>
  <img src="https://mintcdn.com/a2v2ai/ItYQQYTHsiiRmr-t/images/knowledge-base/files.png?fit=max&auto=format&n=ItYQQYTHsiiRmr-t&q=85&s=103de36204d9ac91640520dfffc5dcbf" alt="The Files tab of New Knowledge Item with a document selected." width="1440" height="900" data-path="images/knowledge-base/files.png" />
</Frame>

## Supported files & limits

| Item             | Detail                                                                |
| ---------------- | --------------------------------------------------------------------- |
| **File types**   | PDF, DOCX, TXT                                                        |
| **Best results** | Text-based files (selectable text), not scanned images                |
| **Large files**  | Process in the background; very large or image-heavy PDFs take longer |
| **Status**       | Track from the source list — only **Completed** sources are used      |

<Note>
  Scanned PDFs (images of text) extract less reliably than digital PDFs. If a
  scanned document indexes poorly, convert it to text first or add the key points
  as [Q\&A pairs](/knowledge-base/qa).
</Note>

## Tips

<Tip>
  * Give files descriptive names — they help you manage sources later.
  * Split very large manuals into logical documents (e.g. one per chapter) so
    retrieval stays precise.
  * Re-upload or [reprocess](/knowledge-base/manage-sources#reprocess-a-source)
    when the underlying document changes.
</Tip>

## Troubleshooting

<AccordionGroup>
  <Accordion title="The status is stuck on Processing">
    Large or image-heavy files take longer. Give it a few minutes. If it stays
    stuck or turns **Failed**, [reprocess the source](/knowledge-base/manage-sources#reprocess-a-source).
  </Accordion>

  <Accordion title="The agent doesn't use my document">
    Confirm the source shows **Completed**. If it does, the content may not match
    how the question is phrased — try adding a [Q\&A pair](/knowledge-base/qa) for
    that question, or upload supporting material.
  </Accordion>

  <Accordion title="The upload failed">
    Check the file type (PDF, DOCX, TXT only) and that the file isn't corrupted.
    Re-export it and try again.
  </Accordion>
</AccordionGroup>

## Related

<CardGroup cols={2}>
  <Card title="Add a website" icon="globe" href="/knowledge-base/website" />

  <Card title="Add a Q&A" icon="comments" href="/knowledge-base/qa" />

  <Card title="Manage & reprocess sources" icon="arrows-rotate" href="/knowledge-base/manage-sources" />

  <Card title="Test your agent" icon="vial" href="/testing/sandbox" />
</CardGroup>