Hugging Face launches gr.Workflow for drag-and-drop AI pipelines
The new Gradio feature lets developers connect Python functions as a graph, inspect each step’s output, and automatically expose the workflow through a REST API.
Hugging Face has launched `gr.Workflow`, a new Gradio feature for building multi-step AI pipelines as graphs, announced on the company’s blog around August 24–25, 2026.
Developers can wrap Python functions with `gr.Workflow(bind=[...]).launch()`, then arrange their execution order on a drag-and-drop canvas. Each function appears as a connected node, making data flows and intermediate outputs easier to inspect than by reading code alone.
A key feature is that every workflow automatically becomes a REST API, allowing it to connect to websites, apps, and other systems without requiring developers to build the entire API layer separately. It can also be deployed to Hugging Face Spaces with a single command.
This approach does not eliminate coding, as developers must still prepare the Python functions that make up the workflow. However, it brings design, experimentation, step-by-step inspection, and deployment into one process. Launch details were reported consistently by Hugging Face’s official blog and tech publication daily.dev.
Thai development teams can prototype multi-step AI systems and expose them as APIs faster while retaining visibility into each step’s output—ideal for workflows that combine multiple models or functions.