A Hugging Face connection lets the Hugging Face steps run open models — to answer prompts, classify, summarise and translate text, make embeddings and images, and transcribe audio — through Hugging Face's Inference Providers, using your company's own account. Bizomate does not pay for or resell AI use: Hugging Face bills your account, and each run's output shows what it used.

Who can do this

Workspace Admins and Editors, on every plan — Hugging Face is not a premium connection. Viewers see the connection but cannot add or change it.

Before you start

A Hugging Face account. Free accounts get a small monthly allowance of inference; beyond it, Hugging Face bills the account (a PRO or organisation plan, or pay as you go).

Steps

In Hugging Face, make the token:

  1. Sign in at huggingface.co.
  2. Open Settings › Access Tokens and select Create new token.
  3. Choose Fine-grained, name it — for example Bizomate — and tick Make calls to Inference Providers. To bill an organisation instead of yourself, give the token that permission for the organisation too.
  4. Select Create token and copy it. It starts with hf_. Hugging Face shows it only once.

A Read token works as well, but gives more access than the steps need.

In Bizomate, add the connection:

  1. Select Connections in the bar at the top, then + Add connection.
  2. Under AI, select Hugging Face. It reads "Stored credential".
  3. In Connection name, type a name — for example Consulace Hugging Face.
  4. Paste the token in Access token. "Stored encrypted. It is never shown again after you save."
  5. Optional: if the token has an end date, choose it in This key expires on · optional.
  6. Leave Test the connection before saving ticked. Bizomate asks Hugging Face whose token it is before keeping it.
  7. Select Save connection. The button reads Testing, then the dialog closes.

The toast "Connection added" says "“Consulace Hugging Face” is ready. Every workflow in this workspace can use it."

What happens next

  • Test in the connection's panel answers with the account and the kind of token — Reached @consulace · fine-grained token. It costs nothing.
  • In the designer, the Hugging Face steps list it under Connection: Ask a model, Classify text, Summarise text, Translate text, Create embeddings, Create an image and Transcribe audio.

Good to know

  • To use a new token: open the connection's panel and select Replace key, paste the new token and select Replace key. The old one is used until then. With Test the connection before saving ticked, nothing changes if Hugging Face refuses it. The toast says "Key replaced".
  • A token without Inference Providers passes the test but fails in steps — the test reads only who you are.

If something goes wrong

What you see Why What to do
"Fill in Access token." The box is empty. Paste the token.
Hugging Face refused it — "Hugging Face refused: … Check the connection's details on Connections. Nothing was saved." The token is wrong, deleted or cut short. Make a new token and paste the whole of it.
"Hugging Face refused: … Check the connection's details on Connections." (in a run) The token lacks Make calls to Inference Providers, or was deleted. Edit the token's permissions in Hugging Face, or make a new one.
"Hugging Face refused: …" (in a run) The monthly allowance is used up, a rate limit, or a model Inference Providers cannot run. Read Hugging Face's reason; check billing, or set the step's If this step fails to try again.
"Could not reach Hugging Face: …" or "Hugging Face did not answer within 30 seconds." Hugging Face could not be reached, or was slow. Try again in a few minutes.