> For the complete documentation index, see [llms.txt](https://docs.artific.nl/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.artific.nl/en/for-administrators/knowledge-base/sources.md).

# Sources

The earlier version of the knowledge feature. If your menu shows **Sources** rather than **Knowledge base**, this page describes what you have.

{% hint style="info" %}
An organisation uses either **Sources** or the newer **Knowledge base**, never both. The newer version adds collections, folders, labels and search tools. Your account manager can explain what moving over involves.
{% endhint %}

## How it differs

|                       | Sources                                             | Knowledge base                       |
| --------------------- | --------------------------------------------------- | ------------------------------------ |
| Unit                  | A source: pasted text, a set of files, or a website | Files in folders, inside collections |
| Structure             | One flat list, filtered by labels                   | Collections, folders and labels      |
| Reaching an assistant | Linked directly on the source itself                | Through a search tool                |

The practical difference: with Sources you say *"this assistant may use this source"* on the source. There is no separate step to make it searchable.

## The sources list

Every source with its name, type, processing status and labels. Filter by name, type, label or status.

**Add source** creates one. The row menu deletes one.

## Creating a source

| Field                        | What it does                                         |
| ---------------------------- | ---------------------------------------------------- |
| **Name**                     | How you will recognise it in the list.               |
| **Labels**                   | Tags for filtering. Managed under **Source labels**. |
| **Link AI Assistants**       | Which assistants may use this source.                |
| **Link AI Toolbox elements** | Which elements may use it.                           |
| **Type**                     | Plain text, Document or URL.                         |

{% hint style="warning" %}
The type cannot be changed after the source is created. To switch, create a new source.
{% endhint %}

### Plain text

A **Content** box you type or paste into. Good for short reference material that does not exist as a document: opening hours, a list of internal contacts, a set of standard answers.

### Document

Upload one or more files. Each file gets its own processing status, and hovering it explains any failure: unsupported format, file too large, no text found, and so on.

### URL

Point at a website.

| Setting           | What it does                                                                                    |
| ----------------- | ----------------------------------------------------------------------------------------------- |
| **Website**       | The address, with its protocol.                                                                 |
| **Crawl website** | Follow links rather than reading only that page.                                                |
| **Subpage depth** | How many levels of links to follow.                                                             |
| **Blocked list**  | Paths to skip. `*` works as a wildcard, and `info/` excludes `/info` and everything beneath it. |

## Advanced settings

| Setting             | What it does                                                                                                             |
| ------------------- | ------------------------------------------------------------------------------------------------------------------------ |
| **Only visible to** | Restricts the source to specific user groups.                                                                            |
| **Enrich model**    | Generates extra questions per piece of content to improve how well it is found. The **questions** stepper sets how many. |
| **Blocked list**    | For URL sources, as above.                                                                                               |

**Enrich model** is worth trying on material that is found less often than it should be. It makes processing slower.

## Processing status

A source moves through four stages, and the status names them: **Extracting (1/4)**, **Transforming (2/4)**, **Chunking (3/4)**, **Embedding (4/4)**, ending at **Ready to use**. Each stage has a matching failure state.

For websites you may also see **Website unreachable** or **Website refuses connection**, both of which mean the site did not let the platform read it.

Only a source at **Ready to use** can be used by an assistant.

## Source labels

**Source labels** manages the tags used for filtering. Each has a **Name** and a **Description**. They filter the sources list; unlike knowledge base labels, they do not carry a colour and are not used to scope search.

## During a migration

If your organisation is being moved to the newer knowledge base, sources become temporarily read-only and a banner says so. Adding, editing and deleting are disabled until the migration finishes. Existing sources keep working throughout.