TT Claude Memory

project_zeta_q3_eps_iota_S7_mondo_precedent_2026_05_02

Properties1
descriptionMondo SSSOM-canonical at production scale (109K rows + 73 TSVs + ODKfull docker CI + Jenkins) validates ω.η Q-004 + A-22 single-file lock; SSSOM 1.1 curation_rule slot NOT USED anywhere — INHERIT v2 ζ-Q5 ψ.γ' lock is FORWARD-LOOKING ahead of Mondo precedent; combined-file canonical + per-source-files derived as INHERIT v2 reference pattern

3 min read

ζ-Q3 ε.ι S7 Mondo Disease Ontology precedent inspection — outcome-VALIDATED-WITH-NOTE 2026-05-02

Outcome

outcome-VALIDATED-WITH-NOTE — Mondo is a production-grade SSSOM precedent at scale (363,132 mapping rows across 73 TSV files; 109,384 rows in combined mondo.sssom.tsv; ODKfull-docker CI on Jenkins + GitHub Actions). Several patterns transfer cleanly to INHERIT v2’s ω.η lock; some patterns explicitly do not transfer.

Key surprise: Mondo does NOT use the SSSOM 1.1 curation_rule slot anywhere in 73 TSVs. INHERIT v2 ζ-Q5 ψ.γ’ lock is AHEAD of Mondo’s production usage. INHERIT v2’s curation_rule design is FORWARD-LOOKING; Mondo’s mapping_justification (e.g., semapv:ManualMappingCuration) is the current-production baseline.

Findings

AspectFinding
SSSOM TSV count73 (1 combined + 71 per-source-per-predicate + 1 generated merged view)
Total mapping rows363,132
Combined mondo.sssom.tsv109,384 rows
Predicate distribution (combined)99.92% skos:exactMatch + 0.08% skos:broadMatch
Mapping_justification (combined)99.95% semapv:UnspecifiedMatching (loses curation attribution that per-source preserves)
Mapping_justification (per-source)semapv:ManualMappingCuration (curated)
Schema columnsMinimal 5-6: subject_id / subject_label / predicate_id / object_id / object_label / mapping_justification
curation_rule slotNOT USED anywhere
confidence columnNOT USED
creator_id columnNOT USED
extension_definitionsNOT USED
CI runtimeobolibrary/odkfull:v1.6 container
GitHub Actions17 workflows; PR-blocking on master via main.yaml
JenkinsJenkinsfile runs make test with 100 GB RAM + 128 GB ROBOT_JAVA_ARGS
Per-source split distribution38 hasdbxref + 14 exactmatch + 5 each (broadmatch / narrowmatch / relatedmatch) + 4 closematch + 1 combined
LicenseCC0-1.0

Transferable patterns (5)

  1. SSSOM-canonical pattern at production scale (109K mappings; CC0; CI-validated)
  2. Curie_map metadata block in TSV header (per-file scoped curie maps; SSSOM 1.0/1.1 spec compliant)
  3. PR-blocking SSSOM validation in CI (main.yaml runs validation on every PR to master)
  4. Combined-file canonical + per-source-files derived (Mondo: mondo.sssom.tsv canonical + 71 per-source generated views)
  5. mapping_set_version pinned to release date (NEW Phase-1 Sprint S2 task candidate ~½ day)

NOT-transferable patterns (5)

A. Minimal schema (no curation_rule / confidence / creator_id) — INHERIT v2 retains 7-column schema for forward-looking traceability B. semapv:UnspecifiedMatching in combined file (loses per-mapping curation attribution) C. ODKfull docker container (massively over-engineered at INHERIT v2’s 30-50 alignment scale) D. Jenkins-based CI redundancy (over-belt-and-braces; PR-blocking GitHub Actions sufficient) E. NO SSSOM 1.1 curation_rule slot — INHERIT v2 ζ-Q5 ψ.γ’ lock STANDS but acknowledges Mondo precedent gap; INHERIT v2 is forward-looking

Implications for Phase E Task 13 ε.ι lock-decision

  1. ω.η Q-004 lock STANDS — Mondo precedent confirms SSSOM-canonical pattern at production scale (109K mappings).
  2. A-22 single-file canonical lock STANDS — Mondo’s hybrid combined+per-source pattern shows single-file canonical is right for INHERIT v2’s 30-50 alignment scale.
  3. NEW Phase-1 Sprint S2 task candidate (lock-time): pin mapping_set_version to release date.
  4. NEW A-21 CI gate candidate (lock-time): G-SSSOM-PR-BLOCKING (already implied by branch protection but worth explicit gate-row).
  5. ζ-Q5 ψ.γ’ curation_rule lock STANDS but acknowledged forward-looking — Mondo precedent does NOT use curation_rule.
  6. Phase-1.5 trigger candidate: if INHERIT v2 alignment count exceeds 100-200 (Year-2+), evaluate ODKfull adoption + per-source-split.

Cross-references

  • T-file: ~/off-github/library/projects/inherit/T-spike-eps-iota-S7-mondo-precedent-2026-05-02.md v1.0 (22 KB, 208 lines)
  • arch-state v3.27 → v3.28 §11 S7 row LANDED + Changelog row
  • Q-003 v1.9 → v1.10 §10 S7 row LANDED + CHANGELOG entry
  • MEMORY.md +1 entry
  • Active-work-log entry
  • Companion artefacts: sample-and-grep on /tmp/spike-s7-mondo/ (1.1 GB; 0.007% sample sufficient per plan §5)

Methodological observations

  1. First negative-finding from precedent inspection (Mondo doesn’t use SSSOM 1.1 curation_rule). NEGATIVE findings are first-class deliverables — they identify INHERIT v2 design choices that are FORWARD-LOOKING ahead of current production.
  2. Sample-and-grep scope discipline followed per plan §5 (1.1 GB corpus; 0.007% sample sufficient for precedent assessment).
  3. 12th spike consecutively (S2+S2.5+S3+S2.10+S2.6+S4+S2.9+S2.9b+S5+S8+S6+S7) with logging-contract closed within same session.

Graph View

Backlinks