TT Claude Memory

project_zeta_q3_eps_iota_S2_2026_05_02

Properties1
descriptionSecond of 10 derisking spikes for ε.ι Phase-2 lock; schema-automator OwlImportEngine fails on all three production ontologies; manual fallback path validates LinkML codegen end-to-end

3 min read

Outcome

outcome-MITIGATED (kill condition met on bulk-import pathway; predefined fallback path validated end-to-end).

Key findings

  • schema-automator OwlImportEngine is not production-ready for arbitrary OWL inputs. Three distinct hard-fail modes on three distinct ontologies in 21s combined wall-clock:

    • CCO (1431 classes): funowl AssertionError: Quote expected in string after ~10000 progress dots — rdflib turtle-parser literal-validation chokes on apostrophe-containing literals embedded in CCO content.
    • IAO (v2026-03-30): funowl Unknown function: Variable → NotImplementedType is not callable. IAO uses an OWL functional-syntax feature funowl’s dispatcher does not implement.
    • IOF Core (iofoundry main): ~30+ cannot handle anon child/parent classes warnings on industrial-ontology equivalent-class definitions (ObjectIntersectionOf/ObjectUnionOf in SubClassOf), then UnboundLocalError in owl_import_engine.py:291.

  • The plan’s predefined fallback path (manual LinkML authoring without structured_imports) works end-to-end. Pilot inherit-pilot.yaml with 2 hand-authored classes + 1 enum produces:

    • gen-pydantic → 164-line Pydantic 2 module; instantiation passes for both classes.
    • gen-typescript → 39-line TS; valid per node 24.14.1 --experimental-strip-types --check.
    • gen-json-schema → 71-line JSON Schema 2020-12 with correct $defs, enum, pattern (ISO 6166 ISIN regex preserved).

  • Implication for ε.ι Layer 1 lock-time framing: hand-curated alignment annotations (class_uri: + per-attribute SSSOM-mediated mappings via inherit.sssom.tsv per ω.η lock); no structured_imports of CCO/IAO/IOF in Phase 1; bulk-import deferred Year-2+ when schema-automator + funowl mature (or are replaced — owlready2 is the obvious candidate).

  • Operational consistency with ω.η: the fallback approach is the SAME pattern ω.η already commits us to for upper-ontology alignment — we do not OWL-import PROV-O / IAO under ω.η, we mediate via SSSOM TSV. ε.ι Layer 1 simply joins the same pattern. This is good cross-decision consistency and reduces tool-chain surface area (one pattern, not two).

  • Cost-row update for ε.ι at lock-time: 10h hand-curation one-shot for 11 CCO mid-level + 9 i-ζ classes (£1.5K junior-engineer or ~½ Paul-day at acquirer-DD-equivalent rates). Rounding-error vs acquirer-narrative-value per feedback_bold_front_loaded_synthesis_preference.

  • NEW richard-task candidate (logged in T-file §5; will land formally at Phase E Task 13 lock-time): annual schema-automator + funowl ecosystem-health review + thin owlready2 wrapper evaluation as Year-2+ uplift trigger. Currently pre-lock; not yet richard-tasks-tt registered.

  • Methodological pattern noted: T-file authored 2026-05-02T10:58 BST + logging-contract closure (arch-state §11 row + Q-003 §10 row + this memory entry) within ~10 min — this avoids the S1-anti-pattern flagged in arch-state v3.16 changelog row (“T-file authoring without immediate logging-contract closure leaves substantive evidence orphaned”). Codification candidate for refined-prompt v3.6.

Cross-references

  • T-file: ~/off-github/library/projects/inherit/T-spike-eps-iota-S2-linkml-imports-2026-05-02.md (off-repo; not in git)
  • arch-state row: inherit-v2-architecture-state.md v3.17 §11 + Changelog v3.17 row
  • Q-003 row: 2026-04-29-multi-phase-audit/answered-questions/Q-003-zeta-asset-taxonomy-CCO-BFO-rooted-9-classes-locked.md v1.2 §10
  • Plan: docs-strategy/docs/superpowers/plans/2026-05-02-zeta-q3-eps-iota-derisking-spikes.md §2 Task 2

Phase-2 status post-spike

ε.ι 10-spike suite: 2 of 10 spikes complete (S1 VALIDATED + S2 MITIGATED). 8 remaining: S3 OntoUML / S4 UK&W NRB pipeline / S5 FIBO SSSOM / S6 AM-CDM precedent / S7 Mondo precedent / S8 Catala HMRC golden / S9 OntoGPT F1 (likely subsumed by S4) / S10 partner conversation Mon 11 May. ε.ι lockable signal: net positive so far — both spikes confirm the foundational pattern works (S1: SEED is high-quality; S2: codegen pipeline works in fallback mode).

  • feedback_universal_production_pipeline_sequence — Rich-directive 2026-05-02 BST canonical 3-layer SEED→LLM-REFINE→partner-REVIEW
  • project_zeta_q3_eps_iota_S1_2026_05_02 — sibling S1 spike VALIDATED outcome
  • feedback_iri_verification_before_lock — discipline applied here (CCO/IAO/IOF probed via OFN file inspection)
  • feedback_actively_use_t_files_in_scorecard_authoring — verbatim error excerpts captured in T-file §2
  • feedback_bold_front_loaded_synthesis_preference — applied to cost-row update reasoning (~£1.5K rounding-error vs acquirer-narrative-value)

Graph View

Backlinks