Skip to main content

OntoGraph documentation

OntoGraph keeps canonical meaning stable while storage, transport, and derived read models change underneath it. This site explains the human problems behind that promise and publishes the executable reference contracts supporting those explanations: declared canonical fixtures or governed extensions, versioned requests, and exact expected query outputs paired with the verified results returned by the current runtime.

Choose your starting point​

  • Start with Use cases when you want a recognisable situation, plain-language outcome, and user benefit.
  • Start with Concepts when you want the mental model connecting accepted state, candidate state, authority, and promotion.
  • Start with Design Direction when you want to see how source ingestion, typed gaps, fuzzy candidates, dialogue, and canonical safeguards are intended to fit together—and which parts are not shipped yet.
  • Start with executed proof cases when you want human questions, verified answers, and the exact evidence values returned by the runtime.
  • Start with the reference-query catalogue when you want the exact input, request, expected output, and verified runtime result.

Use cases are an explanatory layer over the evidence. They do not replace or rewrite the canonical query fixtures.

What to inspect​

  1. Open a use case, a concept explanation, an article in Design Direction, an executed proof case, or the reference-query catalogue.
  2. Choose a query family or concrete fixture.
  3. Review the declared input path, kind, and digest.
  4. For a proof case, compare the human answer with its exact evidence checks.
  5. Compare the request, exact expectation, and verified runtime output.
  6. Confirm the execution engine and PASS status.
  7. Follow the source links when you need the executable fixture or product tests.

What these pages prove​

  • the query contract is explicit and machine-comparable;
  • canonical IDs and semantic distinctions survive the read path;
  • lifecycle and governance state are not flattened into a generic status;
  • provenance gaps and adapter loss remain visible; and
  • every published result executed successfully and exactly matched its fixture expectation at the published source commit; and
  • generated documentation drifts when its source fixture changes.

What they do not prove​

Query output is a deterministic read model. It is not canonical truth, model certification, semantic adoption, review approval, a governance decision, promotion, release authority, or proof of equivalence for a production graph backend.

OntoGraph 0.4.0 remains a memory/file-first Developer Preview.