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

# Database (RAG)

> The **Database** (RAG, Retrieval-Augmented Generation) enriches your Volubile agents with your own knowledge. Import documents, index web pages, then refine the content fragment by fragment for accurate, up-to-date answers.

<Frame>
  <img src="https://mintcdn.com/weengage/_3vYDlD6yxxxMu7I/images/rag/volubile-rag.png?fit=max&auto=format&n=_3vYDlD6yxxxMu7I&q=85&s=7188ee797917fadea23cb4270d7c3575" alt="Database (RAG)" width="1568" height="765" data-path="images/rag/volubile-rag.png" />
</Frame>

## Overview

The **Database** section, available from the sidebar, centralizes the knowledge made available to your agents. It is organized into two complementary sources:

| Source        | Description                                                             |
| ------------- | ----------------------------------------------------------------------- |
| **Documents** | Files you import (PDF, DOCX, PPTX, XLSX, CSV).                          |
| **Web pages** | Website pages indexed automatically by crawling, with optional refresh. |

Each source is split into **fragments**: the text excerpts the agent searches and cites to build its answers.

<Note>
  A single knowledge base is shared across the customer. Agents draw the information they need from it according to
  their configuration.
</Note>

## Documents

### Import a file

Drop your file into the dedicated area, or click it to browse your computer.

<Frame>
  <img src="https://mintcdn.com/weengage/_3vYDlD6yxxxMu7I/images/rag/volubile-rag-documents.png?fit=max&auto=format&n=_3vYDlD6yxxxMu7I&q=85&s=0482e0faf4152fc3a4b083b1421890de" alt="Importing documents into the Database" width="1201" height="530" data-path="images/rag/volubile-rag-documents.png" />
</Frame>

| Specification         | Description                               |
| --------------------- | ----------------------------------------- |
| **Supported formats** | `PDF`, `DOCX`, `PPTX`, `XLSX`, `CSV`      |
| **Maximum size**      | 10 MB per file                            |
| **Row limit**         | Up to 10,000 rows for Excel and CSV files |
| **Fragment size**     | Up to 6,000 characters                    |

Once processed, the file appears in the **Indexed documents** list, with its name, size and import date. Indexing may take a moment: a status indicator shows its progress.

### Document actions

The <Icon icon="ellipsis-vertical" iconType="solid" /> menu on the right of each document gives access to the following actions:

| Action           | Description                                                        |
| ---------------- | ------------------------------------------------------------------ |
| **Download**     | Exports the document content to Excel (limited to 10,000 records). |
| **Replace file** | Imports a new file in place of the current one.                    |
| **Delete**       | Removes the document and all its fragments from the base.          |

### Replace a file

Replacing re-indexes the document entirely from the new file.

<Warning>
  All existing fragments for this document will be **deleted** and replaced with the content of the new file. Any
  manual edits made to the fragments will be lost.
</Warning>

<Tip>
  If you import a file with the same name as an existing document, Volubile offers to **replace** the existing
  document rather than create a duplicate.
</Tip>

## Managing fragments

Click a document to open the side panel and review its **fragments**.

<Frame>
  <img src="https://mintcdn.com/weengage/_3vYDlD6yxxxMu7I/images/rag/volubile-rag-fragments.png?fit=max&auto=format&n=_3vYDlD6yxxxMu7I&q=85&s=7ee55557e7dfaa2be9048f3ba005ecba" alt="Managing a document's fragments" width="672" height="1230" data-path="images/rag/volubile-rag-fragments.png" />
</Frame>

### Searching fragments

Two search modes are available:

| Mode          | Description                                                                                     |
| ------------- | ----------------------------------------------------------------------------------------------- |
| **Full-text** | Classic text search over fragment content.                                                      |
| **Vector**    | Semantic search: the fragments closest in meaning to your query, ranked by relevance **score**. |

### Editing content

Each fragment can be managed individually:

| Action                                                  | Description                                                         |
| ------------------------------------------------------- | ------------------------------------------------------------------- |
| **Add**                                                 | The **Add** button creates a new fragment (up to 6,000 characters). |
| **Edit** <Icon icon="pen-to-square" iconType="solid" /> | Edits the fragment content inline.                                  |
| **Delete** <Icon icon="trash" iconType="solid" />       | Permanently removes the fragment.                                   |

<Tip>
  Manually refining fragments is the best way to improve answer quality: rephrase, split or enrich excerpts so they
  match the real questions your users ask.
</Tip>

## Web pages

In addition to files, you can index the content of web pages. Enter a URL, then click **Add**.

<Frame>
  <img src="https://mintcdn.com/weengage/_3vYDlD6yxxxMu7I/images/rag/volubile-rag-webpages.png?fit=max&auto=format&n=_3vYDlD6yxxxMu7I&q=85&s=e5eb1634257b2d88e8c966f92279673b" alt="Indexing web pages" width="1224" height="357" data-path="images/rag/volubile-rag-webpages.png" />
</Frame>

### Crawl configuration

Adding a URL opens the crawl configuration window.

<Frame>
  <img src="https://mintcdn.com/weengage/_3vYDlD6yxxxMu7I/images/rag/volubile-rag-crawl-config.png?fit=max&auto=format&n=_3vYDlD6yxxxMu7I&q=85&s=da3efe29a92c7ff9c55a35ff653f86a6" alt="Web page crawl configuration" width="601" height="888" data-path="images/rag/volubile-rag-crawl-config.png" />
</Frame>

<Steps>
  <Step title="Source title">
    Give the web source a recognizable name (required).
  </Step>

  <Step title="Root URL">
    The site's root URL. Discovery and manual URL additions are restricted to sub-paths under this root URL.
  </Step>

  <Step title="URLs to retrieve">
    Use **Discover** to automatically detect pages on the domain, or add sub-paths manually. Up to 100 URLs per source.
  </Step>

  <Step title="Automatic refresh">
    Enable a recurrence (**daily**, **weekly** or **monthly**) so the pages are re-indexed regularly.
  </Step>
</Steps>

<Warning>
  Automatic refresh re-indexes **all** pages of the source. If you manually edited some fragments from these pages,
  those changes will be lost.
</Warning>

### Tracking indexed pages

Each web source displays its title, URL, page count and indexing date, along with a status indicator:

| Status                                                | Meaning                          |
| ----------------------------------------------------- | -------------------------------- |
| <Icon icon="circle-check" iconType="solid" /> Success | Indexing completed successfully. |
| <Icon icon="spinner" iconType="solid" /> In progress  | The crawl is being processed.    |
| <Icon icon="circle-xmark" iconType="solid" /> Error   | Indexing failed.                 |

The <Icon icon="ellipsis-vertical" iconType="solid" /> menu on each source lets you **configure** the crawl, **refresh** it, or **delete** the source. As with documents, click a source to review its crawled pages and their fragments (**Fragments** and **Markdown** views).

## Deleting a source

Deleting a document or a web source is **irreversible**.

<Warning>
  If the source is used by one or more agents, deletion is blocked. Volubile shows the list of affected agents:
  remove the source from their configuration first before you can delete it.
</Warning>