Gradio's WorkflowCanvas 0.6.0 adds model endpoints and vision-language support
New release brings HTML modality, auto-launch browser links, and image-text-to-text model nodes.
Gradio, the popular open-source Python library from Hugging Face for building machine learning demos, has released version 0.6.0 of its WorkflowCanvas component. This update brings several notable features aimed at streamlining the creation of complex AI pipelines. The star addition is model endpoint integration (PR #13558 by @hannahblair), allowing developers to connect any model endpoint directly within the visual workflow canvas, making it easier to compose multi-step AI applications.
Other major improvements include HTML modality support (#13668) for richer content rendering, and vision-language model integration through chat completions so that image-text-to-text model nodes work natively. The team also improved canvas authoring and added more examples (#13654 by @abidlabs), along with UX enhancements like automatically opening the write-access link in a browser tab when a gr.Workflow is launched locally. Several bug fixes address error banners, textarea visibility, and trailing null value handling. The release also bumps dependency versions (@gradio/client@2.4.0, @gradio/utils@0.14.0, etc.). With these additions, Gradio continues to lower the barrier for building and sharing interactive machine learning workflows.
- Model endpoint integration lets users add any endpoint directly in the workflow canvas
- New HTML modality and vision-language support via chat completions enable image-text-to-text nodes
- Auto-launch browser tab for write-access links and improved canvas authoring with more examples
Why It Matters
Gradio makes building multi-step AI workflows visual and code-free, accelerating prototyping and sharing of complex ML pipelines.