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

# Managing Your Knowledge Base

> Upload documents, add web sources, and create text snippets to train your AI agent.

Your agent's knowledge base is the content it uses to answer questions. You can add three types of sources: documents, web sources, and text snippets.

## Documents

Upload files that your agent should learn from. When you upload a document, Mind-e converts it to text, splits it into chunks, and stores embeddings for retrieval.

### Supported file types

| Format   | Extension | Max file size | Max content          |
| -------- | --------- | ------------- | -------------------- |
| PDF      | `.pdf`    | 50 MB         | 500 pages            |
| Word     | `.docx`   | 50 MB         | 300 pages            |
| Text     | `.txt`    | 10 MB         | 5,000,000 characters |
| Markdown | `.md`     | 10 MB         | 5,000,000 characters |

<Note>
  Each agent has a **50 MB total storage limit** across all sources (documents, web sources, and snippets combined). Your overall storage limit also depends on your [plan](/account/plans-and-pricing).
</Note>

### Upload a document

<Steps>
  <Step title="Open sources">
    Go to your agent's **Sources** tab in the dashboard.
  </Step>

  <Step title="Upload">
    Click **Upload Document** and select a file. The file goes through validation (size, type, content) before uploading.
  </Step>

  <Step title="Wait for processing">
    The document progresses through these stages:

    1. **Uploading** — file is being transferred
    2. **Converting** — content is extracted and converted to markdown
    3. **Chunking** — text is split into searchable chunks
    4. **Embedded** — chunks are converted to embeddings and stored

    Once the status shows **Embedded**, your agent can answer questions from this document.
  </Step>
</Steps>

<Warning>
  If a document gets stuck in an error state, try re-uploading it. Common causes include corrupted files, scanned PDFs without text layers, or files exceeding content limits.
</Warning>

### Delete a document

Click the delete icon next to a document in the Sources tab. This removes the document, its chunks, and its embeddings from the knowledge base.

## Web sources

Add content from websites by providing a URL. Mind-e crawls the page and indexes the content.

### Add a web source

<Steps>
  <Step title="Enter a URL">
    In the **Sources** tab, switch to the **Web Sources** section and enter a URL.
  </Step>

  <Step title="Discover pages (optional)">
    Click **Discover** to find related pages on the same domain. This helps you index an entire website or section.
  </Step>

  <Step title="Train">
    Click **Train** to crawl the URL, extract content, chunk it, and generate embeddings. The process is similar to document processing.
  </Step>
</Steps>

## Text snippets

Write custom text content directly — useful for FAQs, product details, or information that isn't in a document or on a website.

### Create a text snippet

<Steps>
  <Step title="Open text snippets">
    In the **Sources** tab, switch to the **Text Snippets** section.
  </Step>

  <Step title="Add content">
    Click **Add Snippet**, enter a title and the content. The content is processed and indexed just like documents.
  </Step>
</Steps>

## How retrieval works

When a user asks a question, your agent:

1. Converts the question into an embedding
2. Searches all sources (documents, web sources, snippets) for the most relevant chunks
3. Reranks the results to find the best matches
4. Uses the top results as context to generate an answer

The agent cites its sources in the response, so users can see where the information came from. See [Core Concepts](/core-concepts#rag-retrieval-augmented-generation) for a deeper explanation.

## Tips

* **Quality over quantity** — upload relevant, well-structured content. A few good documents perform better than many irrelevant ones.
* **Keep content up to date** — delete outdated documents and re-upload updated versions.
* **Use text snippets for FAQs** — if you find your agent struggles with specific questions, add a text snippet with the answer.
* **Check document status** — if a document shows an error, the agent cannot use it. Re-upload or check the file format.