hamid
5839b3508f
Add the discovery layer: the denormalized nurse_search_index read model (one row per bookable variant x covered service area), maintained inline inside each source write's transaction, plus the single public search query behind the INurseSearch seam. - Entity + EF config + migration (search schema): covering search index, filtered-unique (variant_id, city_id, district_id) pair with NULL district participating, nurse_id index, soft-delete. - ISearchIndexMaintainer (write seam) + SearchIndexMaintainer: reindex variant / nurse / fan-out / remove-area / full rebuild, staged in the owning source write's unit of work; wired into the b3/b4/b5/b6 handlers. - INurseSearch (read seam) + SqlNurseSearch (real MVP backend): reads only is_searchable=1, category/city/district(NULL-aware)/gender/price filters, rating sort, pagination. Elasticsearch deferred (config Search:Backend). - SearchNursesQuery (+ validator) and RebuildSearchIndexCommand; public SearchController (GET search/nurses) + admin AdminSearchController (POST admin_search/rebuild_index). - Tests: 9 DB-backed maintainer/search + 4 WebApplicationFactory; updated affected b3/b4/b5/b6 handler tests. Build clean, 167 tests green. - Docs: server CLAUDE.md project map, contract search.md, swagger refresh, handoff, report, mocks-registry rows, STATUS. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Shared working context — the parallel-agent handoff
This folder lets a backend agent and a frontend agent work at the same time without ever
editing the same file. It is the running, append-only record of what each side has done and what it
needs from the other. (Stable API shapes live in ../contracts/; this folder
is the running coordination on top of them.)
Lane ownership — the one rule that keeps it safe
Each lane writes only its own files. Neither lane edits the other's.
| Lane | Writes (only) | Reads |
|---|---|---|
| Backend | backend/STATUS.md, backend/handoff/after-backend-phase-N.md (new file per phase), reports/backend-phase-N-report.md, reports/mocks-registry.md |
frontend/requests/for-backend.md, prior backend handoffs |
| Frontend | frontend/STATUS.md, frontend/requests/for-backend.md (append), reports/frontend-phase-N-report.md |
backend/handoff/*, ../contracts/*, prior frontend reports |
Because each handoff is a new file per phase and each STATUS/requests file is append-only, two agents can run concurrently and only ever append — no merge conflicts, no clobbering.
Layout
shared-working-context/
├── backend/
│ ├── STATUS.md # append: one block per backend phase (what shipped, gate status)
│ └── handoff/
│ └── after-backend-phase-N.md # "context for frontend after backend phase N" (one per phase)
├── frontend/
│ ├── STATUS.md # append: one block per frontend phase
│ └── requests/
│ └── for-backend.md # append: contract gaps / shape requests the backend should fulfil
└── reports/
├── README.md # the per-phase report template
├── mocks-registry.md # master list of every mock/seam + how to make it real (backend-owned)
├── backend-phase-N-report.md # what was built / testable / mocked (one per backend phase)
└── frontend-phase-N-report.md # one per frontend phase
The handoff note (backend → frontend), per phase
backend/handoff/after-backend-phase-N.md should answer, for the frontend agent:
- Which endpoints/contracts are now live (link the
contracts/domains/*doc + the swagger snapshot). - What the frontend can now build because of this phase.
- What is mocked server-side (so the frontend knows responses are fake-but-shaped) and what's real.
- Any auth/enum/format detail the frontend must mirror.
- Open questions / things the frontend should not assume yet.
The request note (frontend → backend)
frontend/requests/for-backend.md is where the frontend appends: missing endpoints, fields it needs,
shape mismatches, pagination/filter needs — anything that should land in a later backend change. The
backend agent reads this at the start of each phase. The frontend never edits backend code to "fix" it.