data-validation

name	data-validation
description	Validate integrity of all public/data JSON files. Use when: after editing candidates, constituencies, or districts JSON; before committing data changes; checking coverage completeness; debugging data mismatches.

Data Validation

When to Use

• After creating or editing any file in public/data/
• Before committing data changes
• When troubleshooting missing candidates or broken district pages

Procedure

Run npm run check:e2e-lite (or trigger ingestion-validator-agent) and inspect:

• research/swarm/control/validation-state.json
• If research/constituency-outcomes/ or research/constituency-outcome-reviews/ changed, confirm the validator reports zero constituency-outcome gate defects before commit.
• For Tamil Nadu missing-party or thin-seat questions, also run npm run check:live-candidate-coverage -- --state tamil-nadu plus node tools/tamil-nadu-party-coverage-report.mjs --json so literal missing DMK / AIADMK / TVK / NTK rows, alliance-covered literal gaps, source-confirmed official absences, and missing SPA / NDA / TVK / NTK fronts come from current runtime data rather than screenshots.
• For portrait fixes, rerun node tools/candidate-photo-coverage-report.mjs --ref HEAD (or a scoped candidate check) and keep district-row + linked-detail photo parity on the same verified candidate-local URL.

What It Checks

1. District counts — constituencyCount in districts.json matches actual count in constituencies.json
2. Sequential numbering — Constituency numbers are 1 through N with no gaps
3. Packet resolution — Coverage packet headings resolve cleanly to the district constituency registry
4. Structured seat coverage — Every constituency live-field JSON has at least one candidate
5. ID validity — All constituencyId values in candidate files exist in constituencies.json
6. Full staged backing — Every public candidate row is backed by staged facts/timelines
7. Public parity — Every structured seat candidate also appears in the generated public district rows
8. E2E-lite push gate — Constituency coverage, candidate/detail id mapping, and timeline visible copy/source constraints
9. Outcome packet gate — When outcome packet/review files exist, passed seats need one primary plus three independent approvals and cannot leak runtime projections early

Required Fields Per Candidate

Every candidate row must have: id, name, frontId, frontLabel, party, partyFull, constituencyId. Thin rows set: age: null, education: null, photo: null, timeline: []. Trust fields: verificationStatus, confidence, lastReviewed, sourceNote.