Ontology Quality Suite

Python tools for checking that an OWL2 ontology stays internally consistent across versions, and that TARQL/oxi-gen CONSTRUCT-query transformation files stay aligned with the ontology(ies) they triplify against -- namespaces, class/property usage -- with suggested diffs for automated repair, either against local files or a live Fuseki (SPARQL 1.1 Protocol) triplestore.

This package carries forward the full ontology_suite pipeline from the sibling consolidated_ontology_suite checkout (OntoQA/OQuaRE metrics, SHACL+SPARQL registry checks, OWL2 profile/reasoner-backed consistency, docgen -- the same logic is also ported to TypeScript in consolidated_ontology_suite_webapp, the VS Code extension) and adds three things that didn't exist in Python before:

• Ontology-version consistency + rename detection (ontology_suite.versioning.rename_detection) -- pairs up a version diff's removed/added classes and properties into probable renames, using either an explicit owl:equivalentClass/dcterms:isReplacedBy migration annotation (high confidence) or local-name similarity (fallback).
• Automated repair suggestions (ontology_suite.repair, ontology_suite.checks.repair) -- turns a TARQL/ontology misalignment or a registry-check finding into a concrete unified diff: a PREFIX line fix, an IRI-rename substitution across a query file, an ontology declaration stub, or a real SPARQL 1.1 Update -- reviewable as a .patch file, or applied directly.
• Live triplestore support (ontology_suite.remote) -- runs the same checks against a Fuseki (or any SPARQL 1.1 Protocol) dataset, named-graph aware: diff two ontology versions held as separate named graphs, run the existing registry .rq checks scoped to exactly the graphs they should see, or three-way check a live triplified named graph against its ontology graph and its local TARQL source file.

Every --ontology/--data/query-source argument throughout accepts a local file, an http(s) URL, or either gzip-compressed, transparently (see ontology_suite/io_utils.py and docs/ARCHITECTURE.md's "Loading files" section).

New here? Start with docs/PRIMER.md -- a task-oriented guide to which command answers which question, worked examples for every major use case (auditing an ontology, validating a CSV-to-RDF pipeline, releasing a new version, auto-repairing drift, taxonomy governance, live-triplestore checks, CI wiring), and a suggested adoption path. New to RDF/OWL? It opens with a short, plain-language glossary. Prefer running things over reading them? docs/primer.ipynb is a Jupyter notebook covering most of the same ground with live, executable cells against this repo's own examples/ fixtures (validated on every push by .github/workflows/notebook.yml).

See docs/CONSISTENCY_AND_REPAIR.md for the local-file workflow, docs/FUSEKI.md for the live-triplestore one, docs/ARCHITECTURE.md for the inherited pipeline this all sits on top of, and docs/UPSTREAM_README.md for that pipeline's own original README.

Working on TARQL/oxi-gen queries? docs/TESTING_TARQL.md is the guide to testing a query folder: what goes wrong in a CONSTRUCT query, which check catches each thing, what none of them catch, and a review order that puts the cheapest checks first. ontology-quality-suite sketch --queries ... needs no ontology and no CSV to run the BIND review half of it. It also covers writing your own query-source checks: the run publishes what it parsed as bind-facts.ttl, so a TARQL check is a .rq file and a registry entry -- which can live in your project rather than in this package.

Want to know what the suite actually checks, and how to add checks of your own? docs/check-registry.html is a standalone, self-contained page covering all the registered checks -- what each one asserts, its severity and remediation, and whether it is implemented in SPARQL, SHACL or native Python -- plus the result contract a new check has to satisfy and how to run a project-local panel via --registry/--shapes/--sparql. It is generated from registry.json by docs/generate_check_registry.py, so it cannot drift from the suite; tests/test_check_registry_doc.py fails if the committed copy goes stale. docs/CHECKS.md is the terser Markdown reference over the same registry.

registry.json, the shapes, the queries and the repair templates are shared data rather than this package's private tables: the same files ship inside the VS Code extension, which loads them at runtime and can be pointed at a checkout of this repo instead (ontologySuite.checksRegistryPath). So the registry declares every check any implementation produces -- three of them the CLI cannot run, marked as such in docs/CHECKS.md -- and tests/test_shared_registry_parity.py fails if the two copies stop agreeing, whenever a machine has both checked out.

Want one continuous, realistic worked example instead of task-by-task fixtures? docs/ACME_ROBOTICS_WALKTHROUGH.md runs nearly every stage above against a single small org chart built on two real external vocabularies (W3C Organization Ontology, FOAF) -- with two companion notebooks, docs/acme_robotics_lifecycle.ipynb and docs/acme_robotics_data_pipeline.ipynb (also validated on every push).

Quick start

uv sync
uv run ontology-quality-suite consistency --new domain-v2.ttl --old domain-v1.ttl --queries queries/

from ontology_suite import consistency

report = consistency.check_consistency(
    "domain-v2.ttl", old_ontology="domain-v1.ttl", tarql_sources=["queries/"],
)
print(consistency.format_consistency_report(report))
consistency.apply_repairs(report.repairs, min_confidence=0.7)

CLI

ontology-quality-suite ontology            --ontology domain.ttl
ontology-quality-suite checks              --ontology domain.ttl [--data data.ttl]
ontology-quality-suite sketch              --queries queries/ [--ontology domain.ttl]
ontology-quality-suite triplify            --csv-dir csv/ --queries queries/
ontology-quality-suite data                data.ttl [more.ttl ...] [--ontology domain.ttl]
ontology-quality-suite docgen              --ontology domain.ttl
ontology-quality-suite run                 whichever of --ontology/--queries/--csv-dir/--data apply
ontology-quality-suite version-diff        old.ttl new.ttl
ontology-quality-suite consistency         --new domain.ttl [--old domain-v1.ttl] [--queries queries/] [--apply-repairs]
ontology-quality-suite consistency-remote  --query-endpoint URL --manifest graphs.json

Tests

uv run pytest