Apps › AI

Hugging Face automation with Bizomate

An open model is often the right size for one job: a classifier that labels every review, a translation model for the languages your customers write in. Bizomate runs them through Hugging Face's Inference Providers, one step per job, for each item, and passes the answer on. It runs on your company's own Hugging Face account, and every run's record shows the model and the provider that ran it.

Start free

On every plan, Free included.

At a glance

Label	Value
Plans	Every plan
Connect	A Hugging Face access token
Starts runs	No trigger — starts from another app, a schedule or a webhook
Kind	Connected app · AI

What a workflow can do with it

Ask a model
Ask an open model, get text or fields

Classify text
Gives text a label — positive or negative, or labels you name

Name the labels to choose from, such as delivery, billing, product, or leave Labels empty for a model with its own, such as sentiment.

Summarise text
Shortens a long text

Set the Longest summary in words, or leave it to the model.

Translate text
Turns text into another language

Choose To, and From for the models that need it, such as NLLB and mBART.

Create embeddings
Turns text into vectors, for search and matching

A text, or a list of up to 512 texts from an earlier step, each with its own vector, in order.

Create an image
Draws a picture from words with an open model

A prompt, what to Leave out, and a width and height from 64 to 2,048 pixels if you want them; the image comes back as a file.

Transcribe audio
Turns a recording into text with an open model

How you connect it

Make a fine-grained access token under Settings › Access Tokens on Hugging Face, with Make calls to Inference Providers ticked, and paste it into Bizomate under Access token. Bizomate tests the token before saving it by asking Hugging Face whose it is. The test does not check the Inference Providers permission; a step run without it says so.

The token is your company's own, so Hugging Face bills your account: Bizomate does not pay for or resell AI use. When a step runs, Bizomate sends Hugging Face what that step is given — the prompt, the text, the recording — under your account, and Hugging Face passes it to the provider that runs the model; your account's terms with Hugging Face apply to it.

The token is encrypted before it is saved, used only when a step runs or a test calls Hugging Face, and never shown again — not to you, your colleagues or Bizomate support. It belongs to one workspace, the Code step cannot read it, and revoking the connection wipes it at once.

In the help centre: Connect Hugging Face, Ask, classify, summarise or translate with Hugging Face and Send items down named paths with Switch.

Questions about Hugging Face

Which models can I use?

Each step's Model list shows models Hugging Face's Inference Providers run for that job. You can also type any other owner/name; left empty, each step uses a default model for its job — for Transcribe audio, Whisper.

Why does a step fail when the connection test passed?

The test reads only whose token it is. If the token was made without Make calls to Inference Providers, the test passes but the steps cannot run models, and the step says so. Make a new fine-grained token with that permission ticked.

Can I classify text with my own labels?

Yes. In Classify text, list the labels to choose from — delivery, billing, product — and the step gives back the top label and its score, and every label from highest to lowest.

Who pays for the models?

Your company does: Hugging Face bills your own account. Bizomate does not pay for or resell AI use, and each run's output shows the model and the provider used.