Knowledge Graph — Project Intelligence

Every Entity, Every Relationship, Mapped Automatically

We built a real-time knowledge graph that maps the entire WinCC OA project as a network of interconnected entities. Every datapoint, every panel, every menu item, every alarm, every archive, and every documentation page becomes a node in the graph — with edges representing their relationships.

The Knowledge Graph visualization showing 129 nodes and 164 edges. The left Explorer panel lists entity types: Panel (8), DataPoint (1), MenuItem (25), Archive (1), PanelObject (79), ManualPage (15). Colored dots represent different entity types — the interconnections reveal the hidden structure of the project.

What It Maps

Entity Type	Count	Source
Panel	8	Extracted from panel XML files
DataPoint	1+	Extracted from WinCC OA DP system
MenuItem	25	Extracted from menu JSON structure
Archive	1	Extracted from archive configuration
PanelObject	79	Extracted from panel elements (IoFields, Buttons, etc.)
ManualPage	15	Extracted from generated documentation

Relationships

The graph doesn't just store entities — it maps how they relate:

• DataPoint BELONGS_TO DPType
• DataPoint DISPLAYED_ON Panel
• Panel OPENS from MenuItem
• MenuItem CHILD_OF MenuItem (parent)
• MenuItem DOCUMENTED_BY ManualPage
• Panel HAS_ALARM Alarm

Interactive Exploration

Click any node to see its properties and all its connections:

Selecting "WATER TANK 1" (orange, MenuItem type) reveals: menu path "Home/Water tank 1", target panel "impianti/acqual.xml", auth level 0, tree level 1, and its backlinks. The radiating orange lines show all entities connected to this item — panels, objects, documentation pages.

Real-Time Sync

We built an Event Bridge — a WinCC OA background script that streams live system events into the graph:

• Navigation tracking — every page transition captured
• Datapoint changes — system-wide monitoring (filtered to exclude internal system DPs)
• Automatic reconnection — graceful degradation if the graph database is temporarily unavailable

This means the graph is always current — it reflects the live state of the project, not a stale snapshot.

29 Extraction Modules

We wrote 29 Python modules that automatically populate the graph from the WinCC OA project structure:

• Menu extractor → MenuItem nodes from menu JSON
• Panel parser → Panel and PanelObject nodes from XML files
• Datapoint extractor → DataPoint nodes from WinCC OA
• DPType extractor → Type definition nodes
• Alarm extractor → Alarm configuration nodes
• Archive extractor → Archive nodes
• Manual extractor → Documentation page nodes

Why It Matters

The Knowledge Graph enables questions that are impossible in standard WinCC OA:

Impact Analysis

"If we change this DPType, which panels, alarms, and menus are affected?" → The graph traces all BELONGS_TO, DISPLAYED_ON, and HAS_ALARM edges instantly.

Natural Language Queries

"Which panels show pump data?" → The integrated AI chat translates this to a graph traversal and returns results.

Tutorial Generation

"How do I get to the Rome plant page?" → The tutorial engine queries the graph as one of 3 parallel search sources, contributing to the confidence score.

The Knowledge Graph is the connective tissue of InteriaLib — the reason every feature can be aware of every other feature.