Tool4Boxology is a modular, visual toolkit designed to support the Boxology framework—an approach for modeling and validating AI and knowledge-based systems. It includes a Draw.io plugin, structured vocabulary libraries, and a set of elementary patterns that help users build valid and meaningful Boxology models. This tool is primarily based on the design patterns and taxonomy introduced by Frank van Harmelen and colleagues in their paper "Modular Design Patterns for Hybrid Learning and Reasoning Systems", which serves as the core reference and inspiration for our implementation.
To document the development process, I have also uploaded a detailed progress report. This report outlines how I started building the Tool4Boxology plugin from scratch, including the initial design decisions, implementation milestones, challenges faced, and how the concepts from the foundational Boxology paper were translated into interactive tooling. It serves as a transparent overview of the step-by-step development journey and the rationale behind key components of the toolkit. You can find it here
Boxology is a visual methodology for describing AI system architectures using interconnected conceptual components such as:
- Data
- Symbols
- Models
- Processes (e.g. infer, transform, generate)
Each connection between components is subject to logic constraints. This toolkit helps enforce and validate those constraints.
| Folder | Description |
|---|---|
| ElementaryPattern | Elementary Boxology patterns written in DOT language for visualization |
| BoxologyPlugin | Validation plugin for Draw.io and vocabulary components for modeling |
👉 Each folder includes its own README with specific instructions—we recommend checking them out to understand how to use the contents effectively.
This section includes Boxology representations written in DOT language—a plain text graph description syntax used with Graphviz. These are used only for visualization, not for validation.
You can view and edit them using tools like the Graphviz Visual Editor. Either insert shapes manually or copy from the provided Vocabulary file. Each elementary pattern is stored separately to support modular design.
🔄 Important Notes:
- DOT is case-sensitive (
Symbol≠symbol)- Nodes must have unique identifiers (
Symbol1,Symbol2, etc.)- Use
rankdir=LRorrankdir=TBfor organized layouts
📌 For logic validation, use the Draw.io plugin described below.
A plugin and validation tool for hybrid AI system modeling using the Boxology methodology. This ensures the construction of valid design patterns by enforcing logical flows in both web and desktop app environments.
- Boxology Plugin: Extends Draw.io with validation logic for Boxology patterns
- Vocabulary Library:
.drawiofile containing reusable modular components
- ✅ Validates logical flow and connectivity
- 🔁 Prevents invalid or dangling edges
- ♻️ Automatically merges duplicate nodes
- 🛠️ Adds validation buttons directly into the Draw.io toolbar
📖 See the BoxologyPlugin folder README for setup instructions for both browser-based and desktop usage.