mechanic-pwa/docs/superpowers/STATE.md

Premium Mechanic PWA — Implementation State

Last updated: 2026-05-17 (end of session 3 — Phase E COMPLETE)

What is done

Phase	Task	Status	Artifact
0	Git infrastructure (Gitea + repos + feature branches)	✅ DONE	repos exist, ssh keys configured
A	A1: Recon TaxiDashboard backend structure	✅ DONE	docs/backend-layout-notes.md
A	A2: Caddy fragment for mechanic.pptaxi.ru	✅ DONE	ops/Caddyfile.fragment
A	A3: MinIO CORS on pp-inspections bucket	✅ DONE	ops/set-minio-cors.sh, applied, preflight 204
A	A4: SQLAlchemy models (MechanicInspection, MechanicInspectionPhoto, MechanicDamageMarker)	✅ DONE	commit be0553e on feat/mechanic-mvp
B	B1: Mechanic module skeleton + /healthz	✅ DONE	commit d2113a0 — flat backend/app/api/mechanic.py with APIRouter(prefix="/v1/mechanic"), registered in app/main.py with prefix="/api" → final URL /api/v1/mechanic/healthz
B	B2: SQLAlchemy models registered + tests	✅ DONE	already shipped in A4 (be0553e) — 38 metadata tests in backend/tests/test_mechanic_models.py, no extra B2 commit
B	B3: Pydantic schemas	✅ DONE	commit 2739cc8 + fix 2fb978d — 14 schemas + 6 Literals in backend/app/api/mechanic_schemas.py, 15 behavioral tests in backend/tests/test_mechanic_schemas.py; fix added = None defaults to Optional[X] fields and tightened MarkerOut.damage_type/severity to DamageType/Severity Literals
B	B4: Auth + DB session dependencies	✅ DONE	commit f0d7e83 — backend/app/api/mechanic_deps.py re-exports get_db+get_current_user, adds require_mechanic guard (admin OR member of any active dept_type='svc' department); 9 tests in backend/tests/test_mechanic_deps.py use fake-DB stubs (no real session needed)
C	C1: GET /me endpoint	✅ DONE	commit a0bdcaa + signature fix 1e5b822 — appended to backend/app/api/mechanic.py; returns MeOut from require_mechanic user; permissions is ["admin", "mechanic:inspect"] for admins else ["mechanic:inspect"]. Test fixtures (client, override_mechanic, _fake_user) in backend/tests/test_mechanic_endpoints.py use FastAPI TestClient + app.dependency_overrides per Option A — no real DB / no async. 6 endpoint tests.
C	C2: GET /vehicles list with q+limit search	✅ DONE	commit 7be8783 — new backend/app/api/mechanic_service.py with search_vehicles_with_last_inspection (single aggregated subquery + outerjoin to avoid N+1 for last completed inspection per vehicle); endpoint in mechanic.py; new VehicleListResponse schema; 8 tests (6 endpoint + 2 service-stmt) with _FakeSession.execute() returning canned rows. Search hits plate (CarV2.number) + VIN only.
C	C3: GET /vehicles/{id} detail + bound /vehicles limit	✅ DONE	commit 9109383 + style fix 90b8c66 — new VehicleDetail(VehicleSummary) + InspectionSummary(_Base) schemas; service get_vehicle_by_id + recent_mechanic_inspections; endpoint raises 404 when missing, last_inspection_at derived from first completed inspection in the list; applied Query(20, ge=1, le=200) bounds to C2's /vehicles limit param. 7 new tests (5 detail + 2 bounds). Service is now FastAPI-free (404 raise lives in endpoint, matching automations.py convention).
D-pre	B3 schemas widened from vendor recon (a187a92)	✅ DONE	After read-only recon of car_inspections (14k rows, 2yr): InspectionType extended 4→9 (added initial, seizure, equipment-change, pre-repair, post-repair); DamageType fully replaced 7→16 (removed rust/glass/paint which had 0 vendor occurrences; added real-world types from vendor distribution: damaged, chip, scuff, burn, bent, puncture, bulge, not-working, etc.). Severity unchanged (4 values, optional — vendor populates only 3%). 2 new positive-coverage tests. Stale comment in mechanic_damage_marker.py:54 references old DamageType values — fix in Phase F when touching that model.
D	D1: POST /inspections with inline carry-over markers	✅ DONE	commits 39a6579 + correctness fix ee3d370 — service start_inspection + latest_completed_inspection (with NULLS LAST ordering); endpoint returns 201 with full InspectionDetail via re-query with selectinload(photos, markers). Carry-over: copies all unresolved markers from prev completed inspection (resolved=False filter — closed defects don't reopen), each new marker has carried_over_from_id=source.id, photo_id=None, resolved=False. _FakeSession extended with add/flush/commit/refresh write-path methods + rows_per_call queue for multi-SELECT service flows. 9 new tests (6 endpoint + 3 service).
D	D2: GET /inspections/{id} with photos+markers	✅ DONE	commit 1776a81 — service get_inspection_by_id_with_relations (selectinload both relationships, scalar_one_or_none); endpoint raises 404 with "inspection not found" parallel to vehicle's 404. D1's inline re-query block refactored to call the same helper (DRY). 6 new tests (4 endpoint + 2 service introspecting stmt._with_options).
D	D3: PATCH /inspections/{id} status/notes/finished_at	✅ DONE	commit aefe117 — service get_inspection_by_id (thin session.get wrapper) + patch_inspection (auto-sets finished_at=now(UTC) when status→completed AND finished_at not in payload; setattr loop applies all model_dump(exclude_unset=True) fields). Endpoint re-loads via get_inspection_by_id_with_relations for the response. 8 new tests (6 endpoint + 2 service). Endpoint and service both named patch_inspection — disambiguated by module path in code, by as svc_patch alias in tests.
D	Housekeeping: extract test fakes to mechanic_fakes.py	✅ DONE	commit eedbf23 — pure refactor. Moved _FakeResult, _FakeSession, _fake_user, _fake_car, _fake_inspection, _fake_inspection_orm, _fake_marker_orm from test_mechanic_endpoints.py (was 899 lines) to new backend/tests/mechanic_fakes.py (207 lines). Test file dropped to 697 lines. Pytest fixtures (client, override_mechanic, override_db) stayed in the test file. Phase E tests will import from mechanic_fakes directly.
E	E1: MinIO async presigned PUT URL wrapper	✅ DONE	commit fe38981 — added presigned_put_url(bucket, key, *, content_type, expires_sec=300) to app/services/storage.py; PresignedUploadResponse.fields defaults to {}; 3 storage tests.
E	E2: POST /inspections/{id}/photos/upload-url	✅ DONE	commit 111661b — service create_pending_photo + _storage_key_for_photo helper in mechanic_service.py; endpoint in mechanic.py with 415-before-404 guard; UUID4 hex storage keys at mechanic/inspections/{id}/<uuid>.<ext>; 9 new tests (7 endpoint + 2 service).
E	E3: POST /photos/{id}/confirm	✅ DONE	commit 724b209 — service get_photo_by_id + confirm_photo (None-safe dim update, status→ready); endpoint verifies storage.exists, 409 if missing, 404 covers both photo-missing and inspection-mismatch; 9 new tests (7 endpoint + 2 service). _fake_photo_orm factory added to mechanic_fakes.py.
E	E4: GET /photos/{id} + /thumb (302 redirect)	✅ DONE	commit 8361ee1 — auth-guarded 302 to presigned MinIO GET URLs (1h expiry). Thumb endpoint falls back to storage_key when thumb_key is None so galleries degrade gracefully if E5 task hasn't run yet. 8 endpoint tests.
E	E5: async thumbnail generation	✅ DONE	commit b37a068 — generate_and_store_thumbnail worker (download → Pillow resize to 480px → upload → DB update) with try/except at each step (failures logged, return None, never crashes event loop). Triggered via asyncio.create_task from E3 confirm endpoint. _thumb_key_for predictable naming (<key-base>-thumb.jpg). pillow_heif.register_heif_opener() wrapped in try/except for dev-env tolerance. 6 new tests (5 storage unit + 1 endpoint trigger).

Branch state: feat/mechanic-mvp at b37a068. Full stack: A4 → B1 → B3 + fix → B4 → C1 + fix → C2 → C3 + fix → B3-vendor-align → D1 + fix → D2 → D3 → test-refactor → E1 → E2 → E3 → E4 → E5. All pushed to origin (Gitea). See git log for full SHAs.

Test count on branch: 144 mechanic tests pass (38 models + 17 schemas + 9 deps + 9 storage + 71 endpoints) in ~2.5s, no real DB needed. Pillow image-resize tests use real PIL in-memory (verify JPEG magic header + dimensions); thumbnail task end-to-end uses mocked I/O.

Live mechanic API surface (11 routes):

• GET /api/v1/mechanic/healthz
• GET /api/v1/mechanic/me
• GET /api/v1/mechanic/vehicles?q=&limit=
• GET /api/v1/mechanic/vehicles/{vehicle_id}
• POST /api/v1/mechanic/inspections — create with carry-over for type=return
• GET /api/v1/mechanic/inspections/{inspection_id} — full detail with photos+markers
• PATCH /api/v1/mechanic/inspections/{inspection_id} — partial update + auto finished_at
• POST /api/v1/mechanic/inspections/{inspection_id}/photos/upload-url — create pending photo row + presigned PUT URL (5min expiry)
• POST /api/v1/mechanic/inspections/{inspection_id}/photos/{photo_id}/confirm — verify upload, mark ready, schedule thumbnail task
• GET /api/v1/mechanic/photos/{photo_id} — auth-guarded 302 to presigned GET (1h)
• GET /api/v1/mechanic/photos/{photo_id}/thumb — same with thumb_key, falls back to original

Phase E architecture decisions (RESOLVED 2026-05-17)

Decision	Choice	Rationale
Upload pattern	PUT (single presigned URL) not POST/FormData	Simpler PWA code: fetch(url, {method:'PUT', body:blob}). PresignedUploadResponse.fields defaults to {} for future POST migration.
Storage key naming	mechanic/inspections/{insp_id}/<uuid4hex>.<ext>	Human-readable path + unguessable suffix. UUID4 avoids enumeration.
Allowed content types	image/{jpeg,png,heic,webp}	Whitelist; 415 on others.
Confirm status semantics	status="ready" set immediately on confirm; thumb generated async	Frontend doesn't need to poll. Galleries fall back to original via E4 endpoint.
Thumbnail backend	asyncio.create_task + Pillow (no Celery)	Pillow + 15+ existing fire-and-forget call sites; no broker setup needed. Thumb resize is ~50ms for typical photo — acceptable on event loop.
Thumb size	480px max-edge, JPEG q=85	Matches services/download_wazzup_media.py precedent.
Auth model for read URLs	Auth-required on /photos/{id} endpoint; presigned URLs themselves bearer-free (1h TTL)	Same as existing media_proxy.py convention.

Vendor data reality (from recon 2026-05-17 — 14,013 inspections, 2 years)

Metric	Value	Implication
Inspections with damage	1,941 / 14,013 = 13.9%	Most inspections have empty marker list — UI optimize for that
Median markers per inspection	2	Inline carry-over copy is cheap
p95 markers	10	Confirms no async needed
Max markers	34	Worst case still ~50 inserts on commit
Median photos per inspection	11 (all_photo_ids)	Includes 9 sides + damage photos
Max photos per inspection	30	Set MinIO presigned batch size accordingly
Carry-over in vendor	None — denormalized snapshot	Our prev_inspection_id+carried_over_from_id FK approach is net improvement
Vendor inspection status field	None — write-once when completed	Our in_progress/completed/cancelled is greenfield, no conflicts

Where things live

Repos (Gitea: http://100.64.0.9:3001/tremble7681)

Repo	Local clone	Working branch
mechanic-pwa	c:\NewProject\PremiumDriverApp\	feat/mechanic-mvp
taxi-dashboard	c:\NewProject\taxi-dashboard\	feat/mechanic-mvp

taxi-dashboard clone has two remotes: origin → Gitea, vds → SSH to /opt/sites/taxi-dashboard/ on 100.64.0.12 (for pulling back-and-forth with prod).

Reference paths (real, verified — see docs/backend-layout-notes.md)

• Base, TimestampMixin → app.models.base
• get_db → app.db (file: app/db/__init__.py)
• get_current_user → app.auth.deps
• User → app.models.user
• Auth flow: OAuth2 password + Bearer JWT (not PIN as plan originally said)
• DB init: Base.metadata.create_all at startup (not Alembic) + inline ALTER TABLE ADD COLUMN IF NOT EXISTS in app/main.py lifespan
• Backend Docker container: taxi-backend-green / taxi-backend-blue (blue/green)
• MinIO container: taxi-minio
• Models registered in: app/models/__init__.py (imports + __all__)

Naming clash decision

Existing app/api/inspections.py + model CarInspection (table car_inspections) hold vendor-imported read-only data from NaughtySoft. Our new tables use the mechanic_ prefix to coexist:

• mechanic_inspections (model MechanicInspection)
• mechanic_inspection_photos (model MechanicInspectionPhoto)
• mechanic_damage_markers (model MechanicDamageMarker)

Plan-vs-reality overrides applied in Phase B (Authoritative)

Three places where the original plan (docs/superpowers/plans/2026-05-16-premium-mechanic-phase1.md) was overridden during execution because Phase A1 recon showed the codebase doesn't match the plan's assumptions:

Plan said	Actual implementation	Why
backend/app/mechanic/router.py (subpackage)	backend/app/api/mechanic.py (flat)	TaxiDashboard uses flat app/api/*.py files; no subpackages
backend/app/mechanic/schemas.py (subpackage)	backend/app/api/mechanic_schemas.py (flat sibling)	Same convention
backend/app/mechanic/deps.py (subpackage)	backend/app/api/mechanic_deps.py (flat sibling)	Same convention
Router prefix /api/v1/mechanic	Router prefix /v1/mechanic, with app.include_router(..., prefix="/api") in main.py	Matches existing convention (inspections.py → /v2/inspections + include with /api). Final URL is identical: /api/v1/mechanic/...
require_mechanic checks user.dept.lower() in (...)	Queries UserDepartment ⨝ Department WHERE dept_type='svc' AND is_active, admin shortcuts before DB hit	User has no dept attribute; departments are M2M via user_departments. DEPT_TYPES=['ops','svc','adm'], svc = service-type (мастерская)

When dispatching Phase C–F implementers, copy this override map into their context so they don't follow the stale subpackage / dept paths from the plan.

Phase C testing-infrastructure decision (RESOLVED — Option A in force)

Phase C runs against FastAPI TestClient (sync) + app.dependency_overrides[require_mechanic] and app.dependency_overrides[get_db] to inject fake user / fake session. Fake session is a SimpleNamespace-flavored _FakeSession class in tests/test_mechanic_endpoints.py that supports .execute(stmt) (returns _FakeResult seeded with rows) and .get(model, pk) (returns seeded object or None).

No conftest.py, no pytest-asyncio, no real DB. SQL correctness is deliberately deferred to Phase J e2e.

If Phase D/E/F needs a more complex query shape (e.g. .scalars().one() / .first()), extend _FakeResult.scalars() accordingly — currently it returns self and supports only .all(). See test_mechanic_endpoints.py:_FakeResult docstring.

Next task

Task F1 — Markers CRUD (Phase F starts; only 2 tasks)

(From docs/superpowers/plans/2026-05-16-premium-mechanic-phase1.md Phase F.)

Working dir: c:\NewProject\taxi-dashboard (branch feat/mechanic-mvp, HEAD b37a068)

Phase F adds the marker endpoints — finally exposes the carry-over data we've been carefully copying since D1. After F, backend Phase 1 MVP is feature-complete and Phase G begins frontend work.

Two endpoints:

• POST /inspections/{inspection_id}/markers — create a new damage marker (request: MarkerCreate from B3)
• PATCH /markers/{marker_id} — update marker (request: MarkerPatch from B3)

The plan likely also wants DELETE /markers/{marker_id} — verify in the plan. There may also be a "resolve marker" endpoint that's just a PATCH with resolved=True.

Important context:

• MechanicDamageMarker has a stale comment at line 54 (refs old DamageType values like rust/glass/paint) — fix in F1 while touching this model area.
• MarkerCreate and MarkerOut schemas already exist in mechanic_schemas.py from B3.
• The carry-over markers (carried_over_from_id IS NOT NULL) deserve a thought: PATCH should be allowed to update them (mechanic confirms the existing damage is still there with current photo); DELETE should probably be NOT allowed on carry-overs (only on freshly-created markers) to preserve audit trail. Verify this in the plan.

Phase F implementer pattern: apply the same Phase D path overrides (flat app/api/, MechanicDamageMarker not DamageMarker, etc.). Reuse _FakeSession + _fake_marker_orm from mechanic_fakes.py.

Phase D architecture decisions (RESOLVED 2026-05-17)

1. Carry-over: inline в POST /inspections (atomic in the same transaction). Until carry-over copy completes, the new inspection isn't "ready" for the UI — async would create race conditions and complicate testing without a meaningful payoff for ~10-20 markers per copy.

2. _FakeSession write-path extension. Add the following to the existing _FakeSession in tests/test_mechanic_endpoints.py:

   • add(obj) — appends to self.added list; auto-assigns obj.id if not set, popping from _next_ids queue if provided, else incrementing _next_id_counter (starts at 1000 to avoid collision with test-seeded ids in the 1-99 range)
   • async flush() — no-op + self.flush_count += 1 (ids already assigned in add)
   • async commit() — no-op + self.commit_count += 1
   • async refresh(obj) — no-op + self.refresh_count += 1
   • rows_per_call: Optional[list[Sequence]] kwarg — if provided, each execute() call pops one item from this queue and returns it as the result rows. This handles multi-query service flows (e.g. start_inspection does select(prev_inspection) then select(prev_markers)).
   • Existing execute(stmt) and get(model, pk) keep their behavior.
   • All extensions backward-compatible with current 83 tests.

3. Test infrastructure: stay on Option A, do NOT add real DB. SQL specifics (selectinload, cascade, JSONB) will be validated in Phase J e2e. aiosqlite ≠ Postgres (JSONB mismatch), testcontainers is too heavy for 3 endpoints, two-paradigm test suite is more pain than gain. Extend the mock instead (see point 2).

Phase D path overrides (apply to D1, D2, D3)

Same flat-file pattern as Phase B-C; map plan's stale paths:

Plan says	Use instead
backend/app/mechanic/router.py	backend/app/api/mechanic.py (append)
backend/app/mechanic/service.py	backend/app/api/mechanic_service.py (append)
backend/app/mechanic/schemas.py	backend/app/api/mechanic_schemas.py (already has all needed shapes from B3)
backend/tests/test_mechanic/test_inspections.py	backend/tests/test_mechanic_endpoints.py (append)
Inspection model	MechanicInspection (app/models/mechanic_inspection)
DamageMarker model	MechanicDamageMarker (app/models/mechanic_damage_marker)
InspectionPhoto model	MechanicInspectionPhoto (app/models/mechanic_inspection_photo)
service.get_vehicle_or_404 (plan)	service.get_vehicle_by_id + endpoint-level None-check + HTTPException(404) raise (C3 fix)
from fastapi import HTTPException inline in service	At endpoint level only; service stays FastAPI-free
404 raise inside endpoint via inline from fastapi import HTTPException	Use top-of-file import (already there from C3)

Pydantic schemas (already shipped in B3):

• InspectionCreate (vehicle_id, type, driver_id) — request body for POST
• InspectionPatch (status, notes, finished_at) — request body for PATCH
• InspectionDetail (full response with photos + markers list) — response model

Phase B3 schemas use model_config = ConfigDict(from_attributes=True) via _Base, so SQLAlchemy ORM instances serialize automatically with response_model=InspectionDetail.

How to resume in a new session

In a new Claude Code chat in c:\NewProject\, paste this prompt:

Продолжаем проект Premium Mechanic PWA.

Прочитай PremiumDriverApp/docs/superpowers/STATE.md — там полное состояние и следующая задача.

Phase E полностью завершён (E1-E5, последний коммит b37a068, 144/144 тестов зелёных).
Backend MVP API готов: 11 routes (полный flow inspection + photos + thumbnails).
Следующая фаза — Phase F (markers CRUD, 2 задачи) из плана.

После F backend Phase 1 MVP feature-complete и начинается Phase G (frontend PWA scaffold).

Используй superpowers:subagent-driven-development.

That prompt is the single entry point for the next session — Claude will read STATE.md, ground itself, and dispatch the F1 implementer.

Remaining roadmap (Phase 1 MVP)

• Phase B (4 tasks) ✅ done — module skeleton, models, schemas, auth deps
• Phase C (3 tasks) ✅ done — GET /me, GET /vehicles, GET /vehicles/{id}
• Phase D (3 tasks) ✅ done — POST /inspections (+ carry-over), GET /inspections/{id}, PATCH /inspections/{id}
• Phase E (5 tasks) ✅ done — presigned PUT URL, photo upload-url, confirm, GET /photos/{id}+thumb, async thumb generation
• Phase F (2 tasks): markers create/update/delete ← NEXT
• Phase G (5 tasks): Vite + React + Tailwind + shadcn + PWA scaffold
• Phase H (3 tasks): Login + Home + VehicleCard pages
• Phase I (6 tasks): CameraCapture + PhotoSlot + VehicleScheme + InspectionEditor + InspectionReview
• Phase J (4 tasks): e2e test + deploy backend + deploy frontend + smoke test

≈ 20 tasks left until MVP live on mechanic.pptaxi.ru. Backend MVP is feature-complete after F.

≈ 25 tasks left until MVP live on mechanic.pptaxi.ru.

Known external dependencies

• MinIO pp-inspections bucket — CORS already configured, presigned uploads will work
• Caddy on 100.64.0.12 — fragment ready, will be wired during Phase J deploy
• Docker blue/green deployment on VDS — backend re-deployment will need a touch of docker compose (Phase J)
• Frontend deployment target: /opt/sites/mechanic-pwa/dist/ on VDS (Phase J)

Conventions reminder

• All new code in feat/mechanic-mvp branches in both repos
• All new tables use mechanic_ prefix
• Don't touch existing app/api/inspections.py or app/models/inspection.py (vendor-imported)
• All new mechanic source lives in flat files under backend/app/api/ with mechanic_* naming (mechanic.py, mechanic_schemas.py, mechanic_deps.py). No app/mechanic/ subpackage — see "Plan-vs-reality overrides" table above.
• Router prefix is /v1/mechanic; /api is added by include_router(..., prefix="/api") in main.py. Final URLs always /api/v1/mechanic/*.
• Phase C+ endpoint tests: pending the test-infrastructure decision (Option A / B / C above). Do NOT introduce pytest-asyncio, conftest.py, or a real DB without first resolving that decision.
• require_mechanic allows admins + members of any active Department with dept_type='svc'. Use it as the default auth dep for all mechanic endpoints.