Visual Blocks is a JavaScript-based visual programming framework designed for no-code creation of machine learning pipelines. It targets developers and platforms seeking to accelerate ML workflow prototyping, experimentation, and evaluation through an intuitive drag-and-drop interface. The framework offers a low-code editing experience and a separate library for embedding this functionality into applications.

How It Works

The core of Visual Blocks consists of an Angular-based node graph editor that accepts JSON specifications for nodes (inputs, outputs, properties) and outputs the graph structure. It includes a library of ML-specific nodes for models, data I/O, and visualizations. A runtime component traverses the graph, determines execution order, and dynamically loads Angular components for each node, executing their defined run functions.

Quick Start & Requirements

• Install via pip: !pip install visualblocks
• Requires Python for Colab integration.
• Official documentation and example notebooks are available in the repository.

Highlighted Details

• Integrates with Google Colaboratory via a Python package, allowing registration of Python functions as ML nodes.
• Offers a "Save to Colab" feature to persist pipelines directly within notebooks.
• Provides example pipelines for applications like 3D photos and LLM-powered manuscript helpers.
• Includes academic publications detailing its use and development.

Maintenance & Community

This project is not accepting contributions to the core library; the Visual Blocks team manages development. Community contributions of pipelines are welcomed. The project is experimental and not an officially supported Google product, with no guarantee of long-term maintenance.

Licensing & Compatibility

The repository does not explicitly state a license. Given its origin and lack of explicit mention, users should assume it is proprietary or consult Google for licensing terms before commercial use.

Limitations & Caveats

The project is experimental and explicitly states a lack of guarantee for long-term maintenance. Contributions to the core library are not accepted, limiting external development on the framework itself.