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

# Attach datasets to projects

Attach a folder or dataset version to a project so its files can be labeled there.

Attaching adds files from Data Space to a project without uploading them again. The project gets its own copy of the selected version, so later changes to the dataset do not affect it.

## Before you begin

* The dataset has a published version, or you are ready to publish v1 while attaching.
* The project's [ontology](https://docs.unitlab.ai/documentation/ontologies/ontologies-overview.md) and [workflow](https://docs.unitlab.ai/documentation/workflows/workflows-overview.md) are ready for this type of data.

## Attach data

{% stepper %}
{% step %}

### Open Attach Data

Open the project and select **Datasets**. On the **Upload or attach your data** card, select **Attach**.
{% endstep %}

{% step %}

### Choose a folder or dataset version

Use the **Folders** or **Datasets** tab and search by name. Sources that are already attached are greyed out and labeled **Attached**. This is checked per version, so you can attach v2 after v1. Loose files can't be attached directly; put them in a folder or dataset first.

![Attach Data dialog with the Folders tab, one robotics folder selected and one already attached](https://292810646-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FGjVLUz4wthGkGlRKM6rM%2Fuploads%2Fk1SzprhUlLHowCNeoYAZ%2Fattach-data-dialog.webp?alt=media)
{% endstep %}

{% step %}

### Set video frame sampling

If your selection has videos, **Video frame sampling** appears. Frames are extracted once at full rate. The slider sets how many frames per second are attached to this project, for example 30 of 60 fps.
{% endstep %}

{% step %}

### Attach

Check the summary at the bottom left, such as *Selected: 1 folder*, and select **Attach**. The files appear in the project's **Datasets** page as unassigned work in the **New** status. If a CSV file still needs a type, the **How should this file be annotated?** dialog opens.
{% endstep %}
{% endstepper %}

Attaching does not assign anyone. Assign the new items with **Assign** in the data action bar or from the [Task Queue](https://docs.unitlab.ai/documentation/queues/task-queue.md).

## Good to know

* **Folders**: attaching a folder creates a dataset version for it behind the scenes.
* **Another project's data**: you get only that project's own uploads, not data it attached from elsewhere.
* **Data Groups** arrive as one task each, with their saved layout.

## Detach a source

Each attached version appears as a card in the project's **Datasets** page. Open the card menu:

* **View Source** opens the dataset, folder or version read-only.
* **Detach** removes the source from this project only. The dialog shows how many files and labels are affected.

By default, detaching archives the project's copies and **keeps the labels**. If you attach the same source again, the files and their history come back. Select **Clear annotations** only if you really want to delete the labels; it asks for a second confirmation.

Detaching never changes the dataset itself or other projects.

## Related

[Dataset versions and history](https://docs.unitlab.ai/documentation/datasets/dataset-versions-and-history.md) · [Project data](https://docs.unitlab.ai/documentation/projects/project-data.md) · [Assignment and priority](https://docs.unitlab.ai/documentation/queues/assignment-and-priority.md)
