SemanticSeamsDiagrams are views. The graph is the source of truth.

C4 architecture, IFML interaction, BPMN process, DMN decisions and SHACL data in one RDF graph, navigated by clicking from one notation into the next and queried across layers with SPARQL.

Get started

The seams

Research notes

GitHub

Zooming in is a graph traversal

Click a container and land on the interface it presents. Click an event and land on the process it triggers. Click a rule task and land on the decision table. The renderer changes with the notation of the node you clicked.

The seams are the product

Thirty-seven bridging predicates name the joins between layers, and each one must earn a cross-layer query. Inside a layer, the layer's own vocabulary does the work.

Native notations stay native

BPMN and DMN keep their XML and render in bpmn-js and dmn-js. Only the elements that take part in a seam get an IRI, using the element's own ID as the IRI fragment.

Questions no single tool can answer

Which screens can write this data? What breaks if this server goes away? Which tasks move an order out of this state? Each is one SPARQL query, shown as a button in the viewer.

Interpret behaviour, generate structure

Generators write the htmx screens, the FastAPI routes, the Dockerfile and one compose file per environment from the graph, with a line-level trace behind every output line. A runtime interprets the BPMN and DMN, stores data where the model says, runs timers and more than one replica. The model is edited through one write path.

The method applied to itself

SHACL shapes constrain the seams. CI parses every file, validates the examples, proves the shapes reject broken seams, asserts the exact rows every canned query returns, and regenerates every committed output. The reference pages here are generated from the same files on every build.