Ontology

Aegir is the canonical owner of the bespoke BFO 2020 / CCO-grounded ontology used by the metadata-tagging stack — the Signals Data Governance (SDG) ontology — and of the ontology-grounded synthetic-data pipeline that produces in-distribution pretraining bytes against it. The ontology, its rigor program, the realized OWL artifact published outward, and the topic layer that measures the corpus against it all live in this chapter. The chapter covers what the ontology is now, how its classes function as the annotation vocabulary, the derive → promote → realize → publish chain that grows and ships it, the quantitative rigor metrics and the formal publish gate every extension must clear, and the disposal membranes that enforce rigor rather than assert it.

The ontology conditions everything downstream — it is the annotation vocabulary for Column Type / Column Property Annotation (CTA/CPA) over wide relational tables. Its classes are not leaf terms but intermediate-depth subsumers: the property-bearing classes a heterogeneous-but-coherent column belongs to. Defining those classes well is building the annotation vocabulary, and the gates exist to keep every term a coherent, grounded annotation target. Putting the ontology next to the model is the only arrangement where these decisions stay coherent.

What the ontology is now

The ontology is content-first and fully derived. The live catalog is src/aegir/ontology/catalog/catalog.json — 433 templates at the time of writing, every one derived from FinePDFs passages (the hand-authored 01–07 seed families were retired in 4d200e4; the catalog is what the derive→promote loop has accreted). Discovery is via schema.catalog_files(), never globs. Each template is a Manchester-syntax axiom skeleton with typed slots ({name:Type} / {name:Type:Bound} — SLOT_DSL.md), bound to an axiom pattern from the pattern library (src/aegir/ontology/patterns.py, four tiers: fhir_minimum / owl2_core / odp / sysmlv2) and carrying provenance — pattern / tier / grounds_ddl / domain / source_span — that downstream stages consume: grounds_ddl drives the deterministic DDL realize profile, domain is the cross-domain tag, source_span is the “informed by the inputs” evidence.

The catalog realizes to a HermiT-certified OWL artifact at corpora/ontology/sdg-ontology.{omn,owl} with a consistency certificate at corpora/ontology/HERMIT_CERTIFICATE.md (828 named classes, 0 unsatisfiable, 10,570 membrane-admitted individuals).

The derive → promote → realize → publish chain

• Derive — the engine (Qwen3.6-35B over vLLM gRPC, src/aegir/engine/) reads aperture-filtered FinePDFs passages and derives axiom-pattern-bound primitives (scripts/derive_ontology.py), staged in catalog.candidate.json.
• Promote — scripts/promote_candidates.py is membrane-gated admission into catalog.json: re-parse (DeepOnto), HermiT consistency against BFO/CCO, and the provenance threading that carries the derivation signals across the boundary (severing grounds_ddl here once left the DDL spine 100% dice-rolled — do not re-sever). combined.json is rebuilt, never edited.
• Realize — scripts/build_realized_ontology.py instantiates templates into concrete OWL over BFO 2020 + CCO (imported as a reasoning authority), HermiT signs the certificate, and unsatisfiability emits justification signals (build/realize_signals.json) that the scripts/reauthor_unsat.py agent loop consumes.
• Publish — aegir.lineup.sync pushes the ontology Data Product to the corpora submodule (zndx/sdg-corpora), hard-gated by the OQuaRE quality gate (sync._gate(); no --push below GREEN).

Membranes with reasons — the doctrine

Everywhere in this chapter the same shape recurs: an agent proposes, a deterministic membrane disposes and returns the reason, and the agent responds and refines — a closed loop, never one-shot propose→drop. Axioms face parse → HermiT (with CCO’s disjointness axioms as the reasoning authority) → OntoClean; authored retrieval surfaces face the annotation membranes M1–M8 (src/aegir/ontology/annotation_membrane.py), including the M8 positive-voice rule (definition-by-negation is an embedding anti-pattern). Rigor is enforced, not asserted, and the two strongest axiom membranes (HermiT and OntoClean) are un-fakeable. The Authors Guide is the canonical reference for every metric, band, gate, and membrane.

The topic layer — the retrieval / measurement face

The same catalog terms are also the corpus’s topic registry: topics ≡ ontology-grounded concept anchors in qdrant (sdg_topics — the 29 SKOS domains plus all 433 live terms; src/aegir/ontology/topic_layer.py). FinePDFs items (anchor-proportional passage windows) map to one topic or none by hierarchical-margin ColBERT MaxSim under the pre-registered, null-calibrated gate τ* = 0.1065; every association is content-addressed to the collection state that adjudicated it. The ambiguous mass is the error signal and the lexicon is the parameter — unaligned input indicts the topics, never the input (the inverted-LDA direction). Registry changes are themselves gated (the M7 basin gate, topic_layer.basin_calibration). On the output side, congruence (src/aegir/ontology/congruence.py) classifies generated chapters against the same ColBERT substrate the harvest classified inputs with — the BERTopic-era R_D reborn. The phase gate is the authoritative record; the BERTopic-era instruments (topic_alignment.py, T_I.pkl, build_topic_model.py) are deprecated, kept for v0.3 reproducibility only.

Scope summary

A separate sibling project (Atelier) consumes Ægir-produced artifacts as an independent pretraining-efficacy gate. Atelier’s own docs describe what it needs from this contract, but those docs are advisory input here, not specification.

Sub-pages

• Authors Guide — metrics & quality gates — canonical: the full quantitative metric suite (IOF rigor dimensions, OntoQA/OQuaRE structural metrics, OntoClean proxies, topic-layer instruments), the OQuaRE publish gate with its [1,5] bands and floors, the disposal membranes, and the pre-registered OQ-Rigor / OQ-Structure objectives, with the exact formulas the tooling enforces
• Charter — Ægir’s internal direction-setter for the ontology scope: provenance discipline, the committed BFO/CCO branch structure, and external-standard anchors
• Migration — authoring history for the initial bespoke vocabulary (dated record)
• Concept brief — RLVR for ontology generation — the design of the long-horizon Signals M4 apparatus: a four-component verifiable reward R(O, I) over OWL artifacts and a GRPO-trained, SAE-instrumented local policy targeting it. That reward is now realized as the deterministic membrane stack (HermiT/CCO, OntoClean, OQuaRE) that the agent-mediated propose/dispose loop — documented in the Authors Guide — is building and proving today
• Semantic engine — authoritative reference — the operational-state description of the SDG ontology, the rigor program, and the closed-loop synthetic-data pipeline
• RLVR for ontology generation — the externally-readable methodological chapter for the long-horizon M4 apparatus: the verifier R(O, I), now realized as the membrane stack, and the SAE-instrumented-Qwen policy that GRPO trains against it to autonomously generate ontology extensions
• Skills Library & Generate→Re-ground→Refine Engine — the dated v0.1 specification of the skills package (src/aegir/ontology/skills/) and the closed refine loop whose shape now lives in the metaflow pipeline