[DOCS] "Select Small" in Space Settings no longer exists — persistent storage instructions outdated

[karolinaranjo]

Which page or section is this issue related to?

Documentation Report

Page: https://docs.argilla.io/latest/getting_started/how-to-configure-argilla-on-huggingface/#persistent-storage

---

What the docs say

The persistent storage guide instructs users to:

"Once you have stored all your data safely, go to your Space Settings Tab and select Small."

---

What actually happens

The "Small" option does not exist in the HuggingFace Space Settings tab.
The HuggingFace UI has changed and this step cannot be completed as documented.
There is currently no documented path to downgrade or manage storage after
annotation work is complete.

---

Why this is urgent

We are a research team that:

• Completed a full annotation cycle on an annotated dataset
• Purchased a paid HuggingFace Space to try figured out data storage during annotation , but it seems to not be a solution.
• Are now actively paying for a Space with no documented path to safely
  export and downgrade

This affects any team using Argilla on HuggingFace for time-bounded annotation
projects — complete work, export data, stop paying. That workflow is currently
undocumented and broken.

---

Steps to reproduce

1. Deploy Argilla on a paid HuggingFace Space with persistent storage enabled
2. Complete annotation work
3. Follow the docs: Space Settings tab → select "Small"
4. "Small" option is not present — workflow cannot continue

---

Expected behavior

Documentation should provide accurate, up-to-date instructions that match the current HuggingFace Spaces UI, including:

• The correct location to manage persistent storage after annotation is complete
• A verified step-by-step guide to safely export data and downgrade the Space
• Screenshots reflecting the current HF interface

---

Suggested fix

• Update the persistent storage section to reflect the current HuggingFace UI
• If HF removed this option, document the alternative workflow (e.g., pause
  Space, delete and recreate, or contact HF support)
• Clarify which export steps guarantee data safety before any Space restart
  or configuration change

[stadlerb]

We had the same problem - you can still add storage via the API as described in https://huggingface.co/docs/huggingface_hub/guides/manage-spaces ("6. Add persistent storage to your Space").

[JiwaniZakir]

The docs reference a "Small" persistent storage tier that HuggingFace appears to have removed or renamed in their Space Settings UI — this is a straightforward case of HF changing their storage options without a corresponding update to the Argilla docs. The relevant section is in docs/getting_started/how-to-configure-argilla-on-huggingface.md under the persistent storage instructions. The fix likely involves verifying the current HF Space storage tier names (they may now use different sizing labels or a different UI flow entirely) and updating the step-by-step instructions to match. Until the docs are updated, the workaround is to navigate to Space Settings and look for any storage downgrade or tier-management option available in the current HF UI, as the underlying functionality likely still exists under a different label.

[stadlerb]

Actually persistent storage has been discontinued and replaced by Storage Buckets. I have not tested yet whether they work properly with Argilla, but what I do know is that our previous persistent storages have been migrated to such buckets (and I have been able to download them locally using the current version of the HF Hub CLI utility).

Also, apparently Argilla itself is not being actively developed anymore at least since May 2025: 829af12 - not sure if there are any drop-in replacements or forks that are still being actively developed.

[karolinaranjo]

Thanks for the context, both of you. This is helpful.

To confirm: yes, persistent storage has been discontinued and replaced by Storage Buckets. For Argilla on HuggingFace Spaces, the path forward is HF Mount, which lets you mount Hub repositories or Storage Buckets directly as volumes into a Space.

As the HF team mentioned, it's currently in beta and only available via the API — UI support is coming soon. In the meantime, you can get a feel for how it works from the Jobs documentation (same underlying mechanism): https://huggingface.co/docs/hub/jobs-configuration#volumes

I hope they update the Argilla docs with proper HF Mount instructions for Spaces once it's out of beta. If anyone tests it before then, happy to hear how it goes.

[TheMrguiller]

Hi,

I have generated the HF Space with the mounted and the argilla space breaks completely. Is there a solution? I would like to use it as part of my ongoing research. What other options do i have to correctly have it in hugging face?

Thank you in advance.

[stadlerb]

It seems that buckets are not a 100% replacement because they are not a proper filesystem even if they can be mounted with HF mount: https://github.com/huggingface/hf-mount?tab=readme-ov-file#best-for--not-for

We have Argilla instances which had their (SQLite) database on persistent storage.
Gladly, the contents were migrated to buckets. After backing the bucket up locally, we tried mounting them to the Space via the UI, but Argilla did not work properly.

Maybe mounting buckets either via UI or via hf-mount would not have been a good idea anyway, as the filesystem possibly does not deliver the guarantees databases need in order to prevent data loss (cf. How To Corrupt An SQLite Database File). We are using a cloud SQL provider now, so nothing has to be stored in the HF space or buckets.

In any case you can copy the contents of a bucket to your local filesystem using a current version if the HF hub library: https://huggingface.co/docs/huggingface_hub/en/guides/buckets#download-a-directory-with-the-cli
That way, you can at save your data and maybe migrate them elsewhere.

[finiteautomata]

Is argilla development ongoing? Seems like last commits were just some months ago, and docs are outdated.

So sad because it is such a great tool.

[stadlerb]

@finiteautomata No, it's not. There's a notice at the top of the README.md of this Github repository.

The original authors have moved on to exciting new projects! The codebase is mature and stable, having served users reliably for years. While we won't be adding new features going forward, we're committed to solve bug fixes and publish patches as needed. If you're interested in helping maintain or extend this project, we'd love to hear from you! Please open an issue to discuss becoming a maintainer - we're looking for dedicated contributors who can take ownership of the project's future development.