WING-196 requirement gap matrix¶
Status: the d1f2d5e query foundation is tested, but the WING-196 epic is not complete.
The foundation currently supports keyword, work model, fit level, source, ATS kind, posted dates,
four job buckets, self-excluding facets, deterministic sorting, paging, deduplication, and a
redacted result DTO.
| Requirement | Model / stored evidence needed | Query or service work | UI work | Required tests | Status / owner lane |
|---|---|---|---|---|---|
| Career tier, confidence, corrections | tier, confidence, corrected value, reason, source version | corrected/effective value predicates and facets | tier/confidence filters; correction explanation | correction precedence; unknown confidence; facet counts | Missing — model/query |
| Discipline | normalized discipline taxonomy plus source | multi-select discipline facet | discipline selector and chips | taxonomy aliases; multi-select; unknown | Missing — model/query |
| Skill intent and proof | owned/growing/want/avoid; claim vs evidence/proof | intent and evidence-strength predicates | intent filters; claim/proof distinction | contradictory intent; claim without proof; version freshness | Missing — model/query |
| Remote geography | remote scope, eligible states/countries, residency limits | geographic eligibility evaluator | remote scope and eligibility explanation | eligible/ineligible/unknown locations | Missing — location |
| Place and commute | city/state/ZIP/metro; multi-location; coordinates; radius/distance/commute assumptions | location resolver and deterministic distance/commute service | place, radius, distance, commute controls | ambiguous place; multi-location; boundary; stale/unknown route | Missing — location |
| Original and repost history | canonical job identity, first seen, repost chain | original/repost predicates and history projection | original/repost filter and history label | repost dedup; first-seen stability; conflicting sources | Missing — model/query |
| Salary, compensation, stock | pay range/currency/period; bonus; equity type/value assumptions | normalized compensation evaluator and facets | pay/total-comp/stock controls with assumptions | annualization; unknown equity; currency; no double count | Missing — model/query |
| ATS support level | ATS kind plus support/readiness level and reason | support-level facet | support filter and explanation | unsupported/partial/full/unknown | Partial — model/query |
| Company type, industry, cohorts | normalized company type, industry, cohort membership and provenance | company predicates/facets | company filters and cohort labels | aliases; multi-cohort; unknown | Missing — adapters |
| Company health and layoffs | health signal, layoffs, observed/effective dates, source, freshness, unknown state | freshness-aware health service | sourced health/layoff filters and stale/unknown labels | stale/fresh/unknown; conflicting sources | Missing — adapters |
| ERGs, insurance, benefits | value, coverage/detail, provenance, observed/effective dates | provenance-aware benefit predicates | benefit filters with source/freshness | absent vs not offered vs unknown; stale source | Missing — adapters |
| Federal path | hiring path, clearance, citizenship/eligibility, grade/series evidence | federal eligibility evaluator | path/clearance/eligibility filters | eligible/ineligible/unknown; no inferred clearance | Missing — model/query |
| Application readiness | required artifacts, missing fields, blockers, consent state | readiness evaluator | readiness filter and missing-step explanation | ready/blocked/unknown; sensitive data boundaries | Missing — adapters |
| Cross-app reuse | canonical query contract and adapter capability metadata | desktop, YC, Flow, Studio tenant-safe adapters | one shared filter rail; capability-aware disabled states | parity contract; tenant isolation; unsupported adapter | Missing — adapters |
| Office/location resolution | office identity, address provenance, ambiguity and confidence | office-to-location resolver | resolved office plus correction flow | ambiguous/unresolved/corrected office | Missing — location |
| Total tax burden in comparisons | filing/household/home/vehicle/spending/locality assumptions; effective dates; authoritative sources; explicit not-taxed/unknown/estimated states | deterministic federal/state/county/city income/payroll, property, sales/grocery, fuel, vehicle, and other-local-tax rules; embedded-tax deduplication | editable assumptions and sourced breakdown in compare/compare-net | locality and effective-date scenarios; unknown rules; embedded-tax no-double-count; deterministic totals | Missing — location/comparison; financial-sensitive |
Integration rules¶
- The model/query lane owns additive domain records, predicates, and facets. It does not edit location intelligence, Angular templates, or deployment composition.
- The location lane owns normalized place, office, distance, commute, and total-tax rule services. Financial rules must be deterministic, sourced, effective-dated, and independently reviewed.
- The adapters lane owns tenant-safe composition and cross-surface projections. It does not duplicate query semantics in a client.
- The existing frontend filter rail is the only UI integration target. UI work starts after the three backend contracts are committed to avoid parallel edits to hot files.
- Every unknown value remains unknown. Absence of evidence must not become a negative claim.