Infer QUDT/SOSA/PROV annotations for FAIR exports #112
+1,054
−14
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Equipment registered through the wizard stores plain unit strings and a free-text equipment type, so FAIR exports for it lack QUDT, SOSA and PROV-O terms. Add fair_ontology.infer_ontology_annotations(), which derives them from what the owner already entered without overwriting existing values and marks each inferred value under ontology_inferred. Ambiguous or unknown units (A, a.u., blank) are reported, not guessed. Add a read-only report script (api/scripts/fair_ontology_dry_run.py) that lists, per equipment, what would be filled and what still needs an owner, from domains/configs or equipment_metadata.config_json. It also flags NanoM-PER-MIN in etcher.json, which does not exist on qudt.org. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Equipment edits rebuild config_json from the form, so ontology values written back to the database would be lost on the next edit. Infer them at export time instead: _load_domain_config now normalizes and then runs infer_ontology_annotations, so run JSON-LD and RO-Crate bundles for registered equipment carry QUDT units/quantity kinds, a SOSA type and feature PROV direction. Existing values still win; if inference raises, the export proceeds without it. Drop the etcher's qudt_unit NanoM-PER-MIN, which does not exist on qudt.org; the target keeps its LinearVelocity quantity kind and its nm/min unitText. Units QUDT has no individual for are no longer reported as owner work, and the schema example now uses a real unit URI. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A read-only dry run against the 13 registered equipment showed: - Workflow-example equipment store domain as a string and features and targets as name lists. Both the inference and fair_archiver's _get_feature_ontology_map / _normalize_pg_domain_config assumed dicts and raised, so FAIR export for those runs was already failing. - Most unresolved items were not owner work: text/date/flag fields (skipped when their declared type is not numeric), the primary_metric/secondary_metric placeholder targets, percent positions (now DimensionlessRatio), count/unitless/ratio and sub-second times (NUM, UNITLESS, MicroSEC/NanoSEC/PicoSEC; URIs checked). - Equipment typed "other" now get a SOSA type from the equipment name. - Corrupted units (U+FFFD) are reported as such. Unresolved items drop from 322 to 28. The dry-run script also gains --output and keeps going when one config fails. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- Add QUDT DEG_C-PER-MIN (TemperaturePerTime) for the registration form's °C/min option. - Match units case-insensitively only for an explicit list of spellings, so "Mbar" (megabar) is no longer read as millibar. - Carry ontology_inferred into the JSON-LD and RO-Crate exports as a schema.org additionalProperty, so inferred terms can be told apart from owner-entered ones. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Sign in
to join this conversation on GitHub.
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What
Equipment registered through the web form stores plain unit strings ("sccm", "mTorr", "°C/min") and a free-text equipment type. The FAIR exporter could emit QUDT / SOSA / PROV-O terms, but only when the config already carried them, so most registered equipment was exported without them.
api/fair_ontology.pyinfers those terms from what the owner entered:Rules: never overwrite an existing value; never guess an ambiguous unit ("A"), report it instead.
Where it applies
At export time only (
fair_archiver._prepare_domain_config), becausePATCH /equipmentrebuildsconfig_jsonand would drop stored values. If inference fails, the export falls back to the normalized config unchanged.Inferred terms are marked in both the per-run JSON-LD and the RO-Crate manifest with a schema.org
additionalPropertynamedontologyInferred, so consumers can tell them apart from owner-entered terms.api/scripts/fair_ontology_dry_run.pyreports, per equipment, what would be filled in and what still needs the owner. It is read-only.Also handles production config shapes: workflow-example equipment store
domainas a bare id and list features/targets by name.Tests
api/tests/test_fair_ontology.py: 26 tests, including the RO-Crate manifest.test_e2e_platformneeds a live database and was not run.Reviewed by Codex over three rounds, ending in APPROVE.
No schema change and no migration. Only the FAIR export output changes.
🤖 Generated with Claude Code