Google Analytics, Vercel Web Analytics, Leadsy and the TikTok Pixel stay off until you accept. The help chat starts once you have chosen. The scheduling button starts once you have chosen. Read our Privacy Policy.
Hugging Face hosts models, datasets, Spaces, and papers. Once you connect it, your assistants can search the Hub and read one repository, including the private repositories a read token can see.
In Hugging Face, open Settings, then Access Tokens, and create a fine-grained token that can read the repositories this assistant should see. A read token is enough, so leave the Jobs and write permissions off. It starts with hf_.
Open Integrations in StellarFirm and pick Hugging Face.
Paste the token, switch on the live option, and press Connect.
Nothing here waits for an Approve because nothing here changes data.
Public Hub search can answer without a token. The token is what lets an assistant see private repositories and which account is signed in.
A fine-grained token without the Jobs permission is the safest choice when you only want reads, since a token that can start Jobs could be used for paid compute elsewhere.
Revoke the token in Hugging Face whenever you want to cut access.
This connection reads the Hub. To use a model you trained, add it in Settings, Models, as your own model or from Hugging Face.
Starting a paid Job and uploading to the Hub are never run through this connection. StellarFirm assistants have no GPU, so the Coder writes the project and a RUN.md, and you start the real run on your own machine or your own provider account. If you have given a Coder's own computer a Hugging Face token, a Job it tries to start waits for your Approve as a spend, and an upload waits as a publish.
Never paste a token into the chat or a file in your repository.