Official Hugging Face blog, written by the Gradio team · Published 2026-08-25

Gradio adds a visual canvas for multi-step AI workflows

gr.Workflow connects processing steps on a canvas, with step-by-step inspection and API access to the same workflow.

Official Gradio workflow canvas example connecting an image and instruction to an image-editing model and its output — Image: Original image from Gradio · Open full-size image

Connect processing steps as a node graph

gr.Workflow connects Python functions, models and Hugging Face Spaces on a visual canvas to organize multi-step AI workflows. Users can run individual steps and inspect intermediate results.

Gradio: connect nodes and inspect intermediate outputs — Source date 2026-08-25; illustrated: 2026-09-09 · Open full-size image

中文

Call the same workflow through an API

The assembled workflow can also be called through an API by other programs. Official examples cover image editing, voiceovers and dataset analysis, connecting the visual steps to processing tasks.

Gradio: canvas and API access to the same workflow — Source date 2026-08-25; illustrated: 2026-09-09 · Open full-size image

Important additions and corrections

Dates below mark changes to our coverage.

Addition · 06/09/2026, 11:48 Beijing time
Added intermediate-result and debugging details for multi-step workflows, including model endpoints, tokens and compute requirements.

Sources and notes

sources中文

Sources and notes

Read the originalOfficial Hugging Face blog, written by the Gradio team

Retained summary and context

Build a visual graph of an AI workflow, run steps and inspect intermediate results.

These capabilities come from the official tutorial; we have not rerun the examples. Model and Space endpoints may require tokens, paid calls or compute. Deployment and recovery from failed runs still depend on the workflow.

Sources and evidence · 3 links

Each link has a purpose and scope. Link counts do not establish independent confirmation; direct support applies only to the statements specified below.

Official Hugging Face blog, written by the Gradio teamDirect support
Read the August 25 official gr.Workflow tutorial for graph connections, step execution, intermediate results, API access, examples and token configuration. The Space link identifies the tutorial’s demo.
Limits: The demo was not run, costs and deployment were not tested, and debugging time was not measured.
Supporting material
https://huggingface.co/blog/gradio-workflow-guide Full source record (JSON)
Official demo Space: AI media studio (FLUX + removal + TTS + LLM)
Official demo Space: generative art lab (parallel fan-out)

Source timeline

2026-09-01 · Entered the BitShovel radar: 852 entries visible on Hugging Face blog at collection; recheck found no new or metadata-updated entries.

Collection record

First discovered: 2026-09-01

Collection and source publication have distinct dates.

Editorial record

Sources reviewed

06/09/2026, 11:48 · Beijing time (UTC+8)

Reviewed official materials to clarify purpose, changes and usage conditions, distinguishing source descriptions from untested results. Original publication and observation dates are retained; an editorial revision is not a new product release.

Sources and full record

This dates our review of the wording and sources, not a project release or product update.

Official page: Gradio adds a visual canvas for multi-step AI workflows

Full record: before, after and sources (JSON)

Corrections and feedback