Hugging Face

The Hugging Face connector tracks your models, datasets and Spaces on the Hub, and the papers linked from them.

Add your account or a repo

Paste any of these on the Sources page, or pass it to areao1 import:

Input	What it tracks
https://huggingface.co/mayachen or hf:mayachen	Every model, dataset and Space the account (or organization) publishes
https://huggingface.co/mayachen/tidy-bert	One model
https://huggingface.co/datasets/mayachen/tidy-corpus	One dataset
https://huggingface.co/spaces/mayachen/tidy-demo	One Space

hf.co links work too.

Private or gated repos

Public repos need no token. For private or gated repos, create a read token:

1. Open huggingface.co/settings/tokens/new?tokenType=read (the Sources page links there as Create one with minimum scopes).
2. Name it, keep the type Read, and create it.
3. Add it on the Sources page (Private repos? Add a read-only token) or with areao1 import https://huggingface.co/mayachen --private.

The token goes to your OS keychain; sources.json stores only the entry's name.

What it reads

Metric	For	Notes
downloads	Models, datasets	Hugging Face's rolling 30-day count
downloads_all_time	Models, datasets
likes	Models, datasets, Spaces
upvotes	Linked papers	From the Hub's Papers page, when the paper is there

Each snapshot is an observation, with claims quoting the exact fields the Hub returned.

What it proposes

• The model, dataset or Space, for original contributions, when it has at least 100 all-time downloads (connectors.min_downloads_for_candidate) or 10 likes. A Space needs 10 likes. The candidate is marked widely adopted at 10,000 all-time downloads, 1,000 in the last 30 days, or 100 likes. For a model, it also counts derivative models (models that name it as their base model) and marks it used by others when there are any.
• A paper, for each arXiv paper the repo is tagged with. The same paper found elsewhere (a scholarly connector, a GitHub README) is proposed only once.

Every proposal waits in your Inbox until you accept it.