Introduction

ODDToolkit (Ontology Driven Design Toolkit) turns an OWL/RDF ontology into the artefacts that normally sit downstream of a domain model: a SQL schema, Java and TypeScript classes, SHACL shapes, Mermaid class and ER diagrams, a Bikeshed specification and an ODCS data contract.

You maintain one thing — the ontology. Everything else is regenerated from it.

Why it exists

In most projects the domain model is written down several times: once in a diagram, once in a database schema, once in the application's classes, once in a validation rule set, and once in the specification nobody updates. Those copies drift. A column gets added to the database and never reaches the API model; a cardinality changes in the spec and never reaches the validation shapes.

ODDToolkit removes the copies. The ontology is the single source of truth, and each artefact is an output of a generator run. If an artefact disagrees with the ontology, it is stale — you regenerate it rather than patch it. That also means the artefacts stay consistent with each other: the SQL table names, the JPA annotations in the generated Java, the ER diagram and the ODCS contract all come from the same in-memory model.

The pipeline

A run has three phases. First, the ontology (and, optionally, a SKOS concept scheme) is read from disk into an Apache Jena model. Second, a pipeline of adapters walks that model and builds up a normalised, in-memory representation — extracting classes, properties, individuals and URI templates, optionally running a reasoner, pulling in imported external ontologies, and applying the overrides you declared in configuration. Third, a generator renders that representation into a concrete artefact. Adapters do the reading and enriching; generators only render. Every generator shares the same adapter pipeline, which is why the outputs cannot disagree.

You drive the whole thing from one command:

java -jar target/oddtoolkit.jar --generator=all --config-file=config.yml

What you get

Ten generators are registered. They fall into three groups:

• Schema-shaped output — sql, java, er-diagram and odcs all derive a table/column model from the ontology, so they agree on table names, primary keys and join tables.
• Class-shaped output — class-diagram and typescript work at the class level, with interfaces, enumerations and inheritance. class sits underneath them and only builds the in-memory class model; it writes no file of its own.
• Document and data output — shacl, bikeshed and data-frame.

See Generators for the exact output of each one.

Assumptions

• The toolkit is a plain Java 21 command-line application built with Maven; there is no server and no daemon.
• Configuration lives in a YAML or JSON file. Only the ontology and concept-scheme paths can be overridden on the command line.
• Adding a generator or an adapter means editing the source and rebuilding — there is no plugin discovery mechanism. See Extending.

Where to go next

• Installation — build the jar and check your toolchain.
• Quick Start — run your first generator against the bundled example ontology.
• CLI Reference — every option the command line accepts.
• Configuration Reference — the full config file schema.
• Generators — what each generator emits and how to configure it.
• Adapters — the pipeline stages and their settings.
• Ontology & Metadata — the ontology conventions the toolkit reads.
• Generated Examples — real output from the example ontology.
• Architecture — how the toolkit is wired internally.
• Extending — add your own generator or adapter.
• License — GNU GPL v3.