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

# Add Data Sources

> **Who can use this?**
>
> * Gallabox users with **Bot Builder access** who want to add documents to their Chat AI Agent's Knowledge Base.
> * Roles: Owners and Admins.

<Info>
  Each data source type has a different upload process. Documents are chunked and indexed after upload — this takes time and consumes AI credits. You can check the indexing status in the Knowledge Base panel.
</Info>

## Plan Availability

| Data source | Basic | Essential | Advanced |
| ----------- | :---: | :-------: | :------: |
| DOC and PDF |   —   |     ✓     |     ✓    |
| URL         |   —   |     —     |     ✓    |

See [Knowledge Base](/ai-agents-and-bots/chat-ai-agent/knowledge-base/overview#plan-availability) for storage limits.

## What Data Sources Can I Add?

| Data source            | Format           | Max size                           | Notes                                               |
| ---------------------- | ---------------- | ---------------------------------- | --------------------------------------------------- |
| PDF                    | .pdf             | 10MB (Essential) / 15MB (Advanced) | Text-based PDFs only — scanned images not supported |
| Word                   | .docx            | 10MB (Essential) / 15MB (Advanced) |                                                     |
| CSV                    | .csv             | 10MB (Essential) / 15MB (Advanced) | UTF-8 encoded                                       |
| Excel                  | .xlsx            | 10MB (Essential) / 15MB (Advanced) |                                                     |
| Text / Markdown / JSON | .txt, .md, .json | 10MB (Essential) / 15MB (Advanced) |                                                     |
| URL (web page)         | web page         | —                                  | Gallabox crawls the page and extracts text          |

> Per-file upload limit: **10MB on Essential**, **15MB on Advanced**. Total Knowledge Base storage is **50MB** per account.

## Chunk Settings

Chunk size and overlap are set **at the Knowledge Base level**, not per upload. Every document you add to a knowledge base inherits its chunk settings:

* **Chunk size** — default **800 tokens**
* **Overlap** — default **400 tokens**

The defaults work well for most content — larger chunks keep more context together, while more overlap prevents information from being lost at chunk boundaries.

## How to Add a PDF

<Steps>
  <Step title="Go to AI Agents and Bots">
    Open it in the left navigation of Gallabox.
  </Step>

  <Step title="Open a Chat AI Agent" />

  <Step title="Click Settings → Knowledge Base" />

  <Step title="Click Add Data Source → PDF" />

  <Step title="Upload your PDF file">
    Drag and drop the file, or click to browse.
  </Step>

  <Step title="Click Upload" />

  <Step title="Wait for indexing to complete">
    The status changes from **Pending** to **Processing** to **Completed**.
  </Step>
</Steps>

**Tips for PDFs:**

* Use text-based PDFs, not scanned images. Scanned documents are not readable by the indexing system.
* Remove unnecessary pages (e.g., cover pages, legal disclaimers) before uploading.
* Structured documents (with clear headings and sections) produce better answers than unstructured text.

## How to Add a CSV or Excel File

<Steps>
  <Step title="Go to AI Agents and Bots → \[Your Agent] → Settings → Knowledge Base" />

  <Step title="Click Add Data Source → CSV or Excel" />

  <Step title="Upload your file" />

  <Step title="Select the sheet">
    Do this for Excel files with multiple sheets.
  </Step>

  <Step title="Map the columns">
    * **Text column** — the column the agent will search in
    * **Metadata columns** — additional context fields (e.g., product name, category)
  </Step>

  <Step title="Click Upload" />

  <Step title="Wait for indexing to complete" />
</Steps>

**Tips for CSV/Excel:**

* First column should be the primary text field (e.g., product description, FAQ question).
* Additional columns are stored as metadata and help the agent contextualise answers.
* Example: For a product catalog, the text column might be "product\_description" and metadata columns might be "product\_name", "price", "category".

## How to Add a URL

<Steps>
  <Step title="Go to AI Agents and Bots → \[Your Agent] → Settings → Knowledge Base" />

  <Step title="Click Add Data Source → URL" />

  <Step title="Enter the URL of the page to crawl" />

  <Step title="Set max pages to crawl">
    Default is **10** — increase it for larger sites.
  </Step>

  <Step title="Click Crawl" />

  <Step title="Review the extracted content">
    Gallabox extracts text from the page(s) and shows a preview.
  </Step>

  <Step title="Click Confirm" />

  <Step title="Wait for indexing to complete" />
</Steps>

**Tips for URL crawling:**

* Start with your FAQ page or help centre — these are the most useful sources for a Chat AI Agent.
* Crawl only the pages you need — crawling entire websites adds noise and consumes more AI credits.
* Gallabox follows internal links on the page but won't crawl the entire site by default.

## How to Add Product Catalog Data

The Knowledge Base accepts two source types: **file uploads** and **web pages**. To make product catalog data available to the agent, export your catalog from your e-commerce platform as a CSV or Excel file and upload it as a file data source.

<Check>
  **Pro tip:** The CSV/Excel method lets you hand-pick exactly which product attributes the agent can search — include the columns that matter (name, description, price, category) and drop the rest.
</Check>

## How to Check Indexing Status

<Steps>
  <Step title="Go to AI Agents and Bots → \[Your Agent] → Settings → Knowledge Base" />

  <Step title="Check each data source's status">
    * **Pending** — queued for indexing
    * **Processing** — currently being indexed
    * **Completed** — indexed and available for retrieval
    * **Failed** — indexing failed (check error message)
  </Step>

  <Step title="Click Retry if a source shows Failed">
    This re-attempts indexing.
  </Step>
</Steps>

## How to Re-index Documents

When you update a document, re-index it to update the Knowledge Base:

<Steps>
  <Step title="Go to AI Agents and Bots → \[Your Agent] → Settings → Knowledge Base" />

  <Step title="Find the document you want to update" />

  <Step title="Click the refresh icon next to the document" />

  <Step title="Upload the updated file">
    Or confirm the updated URL.
  </Step>

  <Step title="Wait for re-indexing to complete" />
</Steps>

**Note:** Re-indexing consumes AI credits. Only re-index when necessary — e.g., when product details change or policy documents are updated.

## Example

**Scenario:** You run a software company and want the Chat AI Agent to answer questions about your50-page product manual. The manual is a PDF with clear section headings.

Here's how you'd add it:

<Steps>
  <Step title="Open the PDF in your browser">
    Check that it renders correctly — confirm it's a text-based PDF, not a scanned image.
  </Step>

  <Step title="Go to AI Agents and Bots → \[Your Agent] → Settings → Knowledge Base" />

  <Step title="Click Add Data Source → PDF and upload the file" />

  <Step title="Keep the knowledge base's chunk settings at the defaults">
    The defaults — **800 tokens** chunk size and **400 tokens** overlap — are appropriate for a technical manual: each section contains enough context, and the overlap ensures context from one section doesn't get lost at the boundary. These settings apply to all documents in the knowledge base.
  </Step>

  <Step title="Click Upload and wait for the status to show Completed" />

  <Step title="Test with questions like 'How do I connect the API?' and 'What's the minimum system requirement?'" />
</Steps>

## FAQs

<AccordionGroup>
  <Accordion title="My PDF is not indexing — it just says 'Failed'">
    Common reasons:

    * The PDF is a scanned image (not text-based) — the indexing system can't read scanned documents
    * The PDF is password-protected — remove the password before uploading
    * The PDF is corrupted — try opening it in a browser to verify
    * The file is too large — split the PDF into smaller files and upload separately

    If the PDF is scanned, you can use an OCR tool (like Adobe Acrobat or Google Docs) to convert it to a text-based PDF first.
  </Accordion>

  <Accordion title="How many AI credits does indexing consume?">
    Indexing consumes AI credits based on the number of tokens in your documents — see [AI Wallet](/pricing-and-billing/ai-wallet) for how credits are consumed and monitored.
  </Accordion>

  <Accordion title="Can I crawl an entire website?">
    Yes, but with limits. The URL crawler follows internal links on the page and can crawl multiple pages. Set the **max pages to crawl** to control how many pages are processed. Crawling a large site (100+ pages) will take longer and consume more AI credits. Start with your most important pages first.
  </Accordion>

  <Accordion title="How do I know if the Knowledge Base is being used in answers?">
    Test in the **Test Widget** on the agent. Ask a question that should be answered from the Knowledge Base. If the agent's answer references specific information from your documents, the KB is working. If the agent gives a generic answer, check that: (1) the document status is **Completed**, (2) the agent has the Knowledge Base connected in **Settings → Knowledge Base**, (3) the question is phrased in a way that matches the document content.
  </Accordion>
</AccordionGroup>

## Related Articles

* [Knowledge Base Overview](/ai-agents-and-bots/chat-ai-agent/knowledge-base/overview) — What the Knowledge Base is and how it works
* [Chat AI Agent Channel Behaviour](/ai-agents-and-bots/chat-ai-agent/channel-behaviour) — How agents behave on different channels
* [AI Wallet](/pricing-and-billing/ai-wallet) — AI credits management