🧰 Tool4Boxology — Hybrid AI Design Toolkit

🔗 Live Interface

Tool4Boxology is a toolkit for designing, validating, and exporting hybrid AI system architectures using the Boxology methodology.

It provides:

• A web-based interface (React + GoJS) for diagramming & validation
• Knowledge Graph (KG) generation from diagrams
• Virtuoso integration for SPARQL queries
• A Draw.io Docker image with preloaded Boxology plugin & libraries

Inspired by:
Frank van Harmelen et al., "Modular Design Patterns for Hybrid Learning and Reasoning Systems" (Web Semantics, 2023).

---

📌 Recommended Entry Point

Start with the Boxology-Interface web app.
It's the main way to design, validate, and export Boxology diagrams and knowledge graphs.

• Visual editor (React + GoJS)
• Real-time validation of Boxology patterns
• Active KG generation + Virtuoso integration
• Multiple export formats (JSON, Styled JSON, DOT, PNG, RDF/Turtle)

The Boxology-Docker submodule (Draw.io image) is optional and provides a Draw.io instance with the Boxology plugin and libraries already installed—useful if you prefer Draw.io or want a lightweight environment for pattern design.

---

📂 Repository Structure

| Folder / Module | Description | |-------------------------|-------------| | Boxology-Interface | Web-based visual editor (React + TypeScript + GoJS) with validation, KG generation, and Virtuoso integration. | | Boxology-Docker | Custom Draw.io Docker image with pre-installed Boxology plugin, libraries, and sidebar images. | | Boxology-plugin | Standalone Draw.io plugin and Boxology shape libraries for manual use in Draw.io. | | kg_creation | Knowledge graph creation logic, RML mappings, RDFizer/SDM-RDFizer integration, and SPARQL utils. | | ElementaryPattern | Elementary patterns in DOT format for modular visualization and reuse. | | Report | Project documentation, notes, and development history. |

---

🚀 Getting Started

There are two main ways to use Tool4Boxology:

1. Boxology-Interface (recommended) – full web editor + KG pipeline
2. Boxology-Docker (Draw.io) – Draw.io with Boxology plugin preloaded

---

1️⃣ Using the Boxology-Interface (Recommended)

This interface is the core visual editor. It supports:

• 📦 Drag-and-drop components
• 🔗 Semantic connectors
• 🎯 Clustering
• 🆔 Stable IDs
• 🧠 KG generation
• 🔍 Virtuoso integration
• 💾 JSON, Styled JSON, RDF, DOT, PNG export
• 🧩 Real-time validation

---

✨ Interface Features

Diagram & Modeling

• Drag-and-drop Boxology components
• Semantically meaningful edges
• Containers & clustering
• Visual customization

Knowledge Graph Features

• Real-time KG generation
• Stable component IDs
• Two JSON export types
• RDF/Turtle export

Virtuoso Integration

• Auto-upload RDF
• One-click SPARQL navigation
• KG browsing

---

🧰 Built With

• React
• TypeScript
• Vite
• GoJS
• Material-UI
• FastAPI
• Virtuoso
• Graphviz
• RDFLib

---

💻 How to Start the Interface

🔧 Prerequisites

• Node.js ≥ 16
• Python ≥ 3.8
• Docker + Docker Compose (for full pipeline)

---

🚀 Option A — Docker Compose (Full Pipeline, Recommended)

From Boxology-Interface root:

docker-compose up -d

Open: http://localhost:5173

---

🚀 Option B — Manual Startup

1. Virtuoso

docker run -d --name virtuoso -p 8890:8890 -p 1111:1111 -e DBA_PASSWORD=dba -e SPARQL_UPDATE=true -v virtuoso_data:/opt/virtuoso-opensource/database openlink/virtuoso-opensource-7:latest

2. Backend

cd Boxology-Interface
cd backend
uvicorn main:app --reload --host 0.0.0.0 --port 8000

3. Frontend

cd Boxology-Interface
npm install
npm run dev

Open: http://localhost:5173

3. Use the Interface

Open the GoJS-based visual editor from the Boxology-interface folder. Instructions are included in its README.

---

📂 Boxology-Interface – Folder Structure

Boxology-Interface
├── src/
│   ├── components/
│   ├── utils/
│   ├── pages/
│   ├── hooks/
│   ├── styles/
│   └── main.tsx
├── backend/
│   └── kg_generator/
├── public/
├── vite.config.ts
└── tsconfig.json

---

🎯 Interface Workflow Example

1. Build architecture
2. Cluster components
3. Validate
4. Generate KG
5. Export JSON/RDF
6. Upload to Virtuoso
7. Run SPARQL queries
8. Export DOT/PNG

---

📘 Example

A sample hybrid AI pipeline using Tool4Boxology.

✅ Check for validation!

---

2️⃣ Using the Boxology-Docker (Draw.io with Plugin)

The Boxology-Docker module provides a Draw.io image extended from fjudith/drawio with:

• Pre-installed Boxology plugin
• Pre-loaded Boxology shape libraries
• Sidebar images
• No manual upload required

---

📦 Boxology-Docker – Getting Started

1. Build + Run

cd Tool4Boxology/Boxology-Docker

docker build -t boxology-drawio .
docker run -p 8080:8080 boxology-drawio

Open:
👉 http://localhost:8080

2. docker-compose (If available)

docker-compose up

---

🧩 What's Inside the Docker Image?

• BoxologyValidation.js plugin
• PatternLib.xml
• ShapeLib.xml
• AnnotationLib.xml
• Sidebar preview icons
• Custom entrypoint script

---

🔍 Boxology-Docker – File Structure

| Path | Description | |------|-------------| | Dockerfile | Builds Boxology-enabled Draw.io | | docker-entrypoint.sh | Configures and runs Tomcat/Draw.io | | js/plugins/ | Plugin logic | | lib/ | XML libraries | | images/ | Sidebar icons |

---

📚 References

• Harmelen et al., Web Semantics 2023

---

📜 License

-CC BY 4.0 — diagrams, documentation, educational assets

-Apache 2.0 — included third-party or extended modules

---

📖 Related Source Paper & Authors

Tool4Boxology is based on the research and methodology described in the source paper (submitted):

Tool4Boxology:
A Semantic Toolbox for Constructing and Analyzing Neuro-Symbolic Architectures

Authors:
Johannes E. Bendler¹, Yashrajsinh Chudasama^2,3, Mahsa Forghani^2,3, Enrique Iglesias^2,4, Disha Purohit^2,3, Jacquiline Roney¹, Annette ten Teije¹, Frank van Harmelen¹, Maria-Esther Vidal^2,3

¹ Department of Computer Science, Vrije Universiteit Amsterdam, The Netherlands
² TIB-Leibniz Information Centre for Science and Technology, Hannover, Germany
³ Leibniz University Hannover, Hannover, Germany
⁴ L3S Research Center Germany, Hannover, Germany

Supervision:
This work is supervised by the above authors and institutions.

Corresponding author:
📧 mahsa.forghani.tehrani@stud.uni-hannover.de Leibniz University Hannover