> 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/collections.md).

# Collections

Collections hold your documents. Found under **Knowledge base → Collections**.

## The screen

A list of collections on the left, the contents of the selected one on the right. Each collection in the sidebar shows its name and how many documents it holds.

Some collections are created by the platform itself and are marked as system collections, for example the one that holds files users attach inside a chat. They cannot be renamed or deleted.

A collection showing a green **Searchable** badge has a working search tool over it. Anything without that badge is storage only.

## Creating a collection

Choose **Add collection**.

| Field               | What to enter                                                                                       |
| ------------------- | --------------------------------------------------------------------------------------------------- |
| **Name**            | Something an administrator will recognise later: *Contracts*, *Policies*, *Product manuals*.        |
| **Description**     | Optional, but worth writing. It is the only place to record what belongs in here and what does not. |
| **Only visible to** | Restrict the collection to specific user groups. Leave empty for everyone.                          |
| **Type**            | Where the documents come from.                                                                      |

### Type

**Default** is the normal choice: you upload files yourself.

**Website** crawls a public website instead, which is useful for keeping an assistant in step with your published documentation without re-uploading anything.

| Website setting      | What it does                                                    |
| -------------------- | --------------------------------------------------------------- |
| **Root URL**         | Where the crawl starts, for example `https://docs.example.com`. |
| **Max pages**        | An upper bound on how many pages are fetched.                   |
| **Crawl depth**      | How many links deep to follow. Leave empty to stay shallow.     |
| **Allowed patterns** | Only crawl paths matching these, for example `/docs/*, /api/*`. |
| **Blocked patterns** | Skip these, for example `/blog/*, /assets/*`.                   |

Use the two pattern fields. Without them a crawl of a large site will pull in navigation pages, archives and other noise that makes search worse, not better.

{% hint style="info" %}
Other connection types exist for syncing from external document systems. They are set up by the platform team rather than self-service. Ask your account manager if you need one.
{% endhint %}

## Adding documents

Inside a collection you can:

* **Upload**: opens a file picker; select as many files as you like.
* **Drag and drop** files anywhere onto the content area.
* **Add folder**: create a folder, and nest folders inside folders.

Uploads appear in the **Activity** tray, which shows progress and reports any failures with a **Retry** button. You can keep working while it runs.

For a **Website** collection there is no upload button. Instead use **Sync now** to fetch the site, and **Cancel sync** to stop a run in progress.

{% hint style="warning" %}
There is a maximum size per file. A file over the limit fails immediately and appears in the Activity tray with the reason. Very large documents are also worth splitting on their own merits: a focused document retrieves better than a huge one.
{% endhint %}

If you upload a file that is already in that location, the platform recognises it and marks it **Already in this location** rather than creating a duplicate.

## Document status

Uploading is not the end of it. Each document is processed before it can be found.

| Status         | Meaning                                             |
| -------------- | --------------------------------------------------- |
| **New**        | Registered, not yet queued.                         |
| **Pending**    | Waiting to be processed.                            |
| **Processing** | Being prepared for search.                          |
| **Indexed**    | Ready. This is the one you want.                    |
| **Failed**     | Processing failed. Hover the status for the reason. |
| **Stale**      | The document changed and needs reprocessing.        |
| **Skipped**    | Deliberately not processed.                         |

A document only becomes findable at **Indexed**. If everything sits at **New** or **Pending** forever, there is probably no search tool over the collection yet.

## Working with documents

Select a file to open its detail panel, which shows its type, size, when it was updated, its labels, which search configurations include it, and where it came from. From there you can download it, move it or delete it.

Select several rows with the checkboxes to get **Bulk actions**:

| Action     | What it does                                                 |
| ---------- | ------------------------------------------------------------ |
| **Move**   | Move the selection to another folder in the same collection. |
| **Label**  | Add or remove labels across all selected files at once.      |
| **Delete** | Delete the selection. Folders go with all their contents.    |

Files can only be moved within the same collection.

The toolbar filters (by name, type, status and label) persist as you move between folders, which is deliberate: it lets you sweep an entire collection for, say, every failed document.

{% hint style="danger" %}
Deleting a folder deletes everything inside it, and deleting a collection deletes all of its contents. Neither can be undone.
{% endhint %}

## Documents from managed collections

Files pulled in from a website crawl or an external system are marked as managed. You cannot rename, move, relabel or edit them here. The source system owns them. Change them where they live and sync again.

## Next

Documents in a collection still are not searchable. Continue with [Search tools](/en/for-administrators/knowledge-base/search-tools.md).