> For the complete documentation index, see [llms.txt](https://docs.katara.ai/about/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.katara.ai/about/documents-and-collections/collections.md).

# Collections

A collection selects corpus content for collection-backed [Q\&A](https://docs.katara.ai/about/agents-and-answering/qna.md). It is implemented by a synchronizer agent: saving its filter does not immediately make every matching document searchable. It is not a Brain, an upload destination, or a way to attach existing documents to a Brain.

## Before you start

You need access to **Content Management** and permission to create or edit collections. Running synchronization requires run permission. Upload the intended [documents](https://docs.katara.ai/about/documents-and-collections/documents.md) first, and prepare a reusable filter in **Filter management** if you need a restricted selection.

## Create and synchronize

1. Open **Content Management → Collections** and select **Create a collection**.
2. Enter **Name** and, optionally, **Description**.
3. Choose **Filter** for the intended corpus selection. Do not treat an unset filter as a deliberately restricted collection.
4. Choose **Syncing**. **Manual only** leaves the collection enabled without an automatic schedule. **Disabled** disables it; it is not another name for manual synchronization. Scheduled choices include **Every hour**, **Every 8 hours**, **Every day**, **Every week**, and **Every month**.
5. Select **Create collection**, then open the collection.
6. Use **Sync now** when available. Check **Logs** for the run result and **Documents** for content; **Overview** shows collection configuration and status.

The expected outcome is a saved collection followed by a completed synchronization, not merely an accepted run request. Select this collection when configuring a Q\&A agent, then test in [Prompt Lab](https://docs.katara.ai/about/agents-and-answering/prompt-lab.md).

## Understand synchronization

Normal synchronization processes changes since the previous completed run. **Rebuild collection** clears the collection's existing search records before a full synchronization. It does not delete the original corpus documents, but search results can be incomplete while rebuilding or after a failed rebuild. Use it deliberately when a full reconstruction is needed, including after changing selection criteria; do not assume incremental synchronization removes every record excluded by a new filter.

A run can process some documents successfully and fail others. Read its message, not only its success indicator.

## Recover from problems

If **Sync now** is disabled, check whether another run is active, the collection is disabled, or its filter was deleted. Restore a valid filter or enabled syncing mode and retry when authorized. For missing answers, check the upload outcome, filter, synchronization logs, and selected Q\&A target in that order. Missing indexing infrastructure requires operator help; repeated clicks do not repair it.
