Hugging Face releases Daggr for code-first visual AI pipelines
hugging face released daggr, a python library for building and debugging multi-step ai workflows. unlike traditional gui tools, daggr uses a code-first approach where users define pipelines in python, and the system generates a visual canvas automatically. it integrates with gradio spaces, supports three node types, and allows inspecting and rerunning specific steps without re-executing the entire pipeline. state persistence is included by default.
Hugging Face released Daggr, a Python library designed to construct and inspect multi-step AI workflows programmatically. The tool addresses the difficulties of debugging complex chains by allowing developers to define pipelines in Python while automatically generating a visual canvas. Key features include the ability to inspect and re-run individual steps without executing the full pipeline, first-class integration with Gradio Spaces (run locally or via API), and state persistence for inputs and cached results. The library supports GradioNode, FnNode for custom functions, and InferenceNode for Hugging Face model providers.