OWL-SDAOWL Synthetic Data AI-Agent

Generate synthetic RDF instance data from OWL ontologies using Large Language Models.

Get Started

Install

Quick Start

Examples & Benchmarks

Ontology-Driven Generation

Automatically derives SHACL shapes from your OWL ontology and uses them to guide and validate the generated RDF data.

How it works

Multi-Agent Architecture

A supervisor coordinates a configurable pool of parallel worker agents, enabling concurrent generation of large datasets.

Architecture

Built-in Review Loop

A dedicated reviewer agent validates the generated output and requests revisions until the data meets quality standards.

Live Web Dashboard

Start the run with --web-ui to follow it in your browser — agent messages and tool calls, the generated output, the triple store, trend charts, and SHACL violations.

Web UI

Bring Your Own Model

GitHub Copilot, Ollama, or any OpenAI-compatible endpoint — configurable per role, so local workers can run alongside a stronger hosted supervisor.

Configure providers

Reasoner-Backed Shapes

Runs a Jena OWL reasoner (optionally with your own rule file) over the ontology and derives a second set of shapes from the inferred model.

Reasoner

Flexible User Context

Attach local files or live web pages (URLs) as extra context to every agent session, so you can supply domain vocabulary, examples, or specification references.

External Ontology Resolution

Automatically fetches and caches external ontologies referenced via owl:imports, with configurable mirrors and TTL.

External ontologies

Benchmark Snapshots

Capture timestamped snapshots of every generation round including token usage, violations, and context state for full observability.

Benchmarking