* feat(site-finder): P1 geometry suitability score (#45)
Backend (parcels.py):
- _polygon_suitability() — Shapely metrics on parcel WGS84 polygon:
area_ha/m2 (projected via cos(lat) к UTM-like meters), perimeter_m,
aspect_ratio (long/short side of MABR), convex_hull_ratio (area / hull area),
min_inscribed_rect_dim_m.
- Composite suitability score 0..1:
base = area_subscore (0 при <0.2 ha, linear → 1 при ≥0.5 ha)
−0.3 если aspect_ratio > 5 (вытянутый)
−0.3 если convex_hull_ratio < 0.65 (изрезанный)
−0.5 если min_inscribed_rect_dim_m < 30 (узкий)
- Label: микро / подходящий / сложная форма / слабо подходит
- Recommendation: "для МКД 16+ нужно >0.3 ha + минимум 40м"
- В analyze response → geometry_suitability.
Frontend:
- Новый GeometrySuitabilityBlock.tsx с color-coded badge + метрики grid +
penalties + recommendation
- Добавлен в LandTab (выше "Зонирование (ПЗЗ)")
- TS типы расширены: GeometrySuitability
Closes#45.
* fix(site-finder): address PR #89 auto-review minor feedback
Backend (parcels.py):
1. Фактическая ошибка фикс: recommendation теперь различает "строительный
минимум 30м" (physical penalty trigger) и "комфорт МКД 12-16эт 40м"
(recommendation level). Пользователю чётко видны 2 порога.
2. Lazy imports → module-level: `from shapely import wkt as _shp_wkt`,
`from shapely.geometry import Polygon`.
3. MABR inner exception теперь логирует: `logger.debug("MABR computation
failed, falling back to sqrt(area): %s", mabr_err)`.
4. Magic numbers вынесены в константы: _GEOM_MIN_AREA_HA, _GEOM_AREA_SCORE_FULL_HA,
_GEOM_ASPECT_PENALTY_THRESHOLD, _GEOM_CONVEX_PENALTY_THRESHOLD,
_GEOM_MIN_WIDTH_PHYSICAL_M, _GEOM_MIN_WIDTH_COMFORT_M, _GEOM_LABEL_MICRO_HA,
_GEOM_LABEL_GOOD, _GEOM_LABEL_MEDIUM, плюс penalty константы.
5. Label "микро" теперь комбинируется с penalties: "микро, узкий" — UI видит
обе проблемы. Pure "микро" остаётся когда нет penalty.
Frontend (GeometrySuitabilityBlock.tsx + types):
6. Заменил emoji ⚠ (U+26A0) на текстовый <strong>"Проблемы формы:"</strong> —
стабильнее в WeasyPrint PDF-экспорте и cross-platform.
- TS типы: GeometrySuitabilityBaseLabel экспортирован, label расширен до
`string` для допуска combo-labels; helper colorForLabel() парсит base часть.
Per auto-review on f4e7491.
* fix(site-finder): P1 #45 — extract 0.3 ha to _GEOM_AREA_COMFORT_HA
Auto-review нашёл: recommendation упоминала "от 0.3 га" но в constants block
не было 0.3 — magic literal в строке. Тот же класс проблемы что 30/40м inconsistency
из прошлого review, только в другом поле.
Fix: новая константа `_GEOM_AREA_COMFORT_HA = 0.3` с комментарием
"рекомендуемая комфортная площадь МКД (recommendation)". Размещена между
_GEOM_MIN_AREA_HA (физический минимум) и _GEOM_AREA_SCORE_FULL_HA (premium) —
третий semantic threshold. Recommendation теперь f-string использует константу.
Per auto-review on ae5e8de.
---------
Co-authored-by: lekss361 <claudestars@proton.me>
* feat(site-finder): P2 cad_buildings соседи + overlap check (#46)
Backend (parcels.py):
- _parse_floors() helper для TEXT column (cad_buildings.floors хранится как
строка, могут быть диапазоны "5-7"). Возвращает верхнюю границу.
- _neighbors_summary(db, geom_wkt, our_cad) — query соседей в 100м (GIST):
cad_num, building_name, floors, year_built, cost_value, area, address, distance.
Aggregate: avg_floors_100m, max_floors_100m, median_cost_per_m2_100m,
count_buildings_100m. Outliers cost/m² фильтруются (1k < x < 500k).
- Overlap check: ST_Intersects + ST_Area(ST_Intersection) > 50 m² (transformed
to UTM 32641 для метров). Если есть → has_existing_buildings: true +
overlap_buildings list.
- В response → neighbors_summary.
Frontend:
- Новый NeighborsBlock.tsx: hard red warn block для overlap (с building names +
overlap_m2 + "Инвестиции невозможны без сноса"); summary metrics (avg/max
floors, median price); toggle "Показать N ближайших" → таблица.
- Border меняется на красный при has_existing_buildings — visual cue.
- Добавлен в LandTab выше "Зонирование (ПЗЗ)".
- TS типы: NeighborBuilding, OverlapBuilding, NeighborsSummary.
Closes#46. Closes#21 (cad_buildings в Site Finder фильтрах).
* fix(site-finder): address PR #91 auto-review minor feedback
Backend (parcels.py):
1. (medium) Aggregation loop _neighbors_summary теперь обёрнут в try/except
(ValueError, TypeError) с fallback к data_available:False + log warning.
Защищает от non-numeric cost_value/area придёт в строке (e.g. "N/A") —
ранее весь endpoint падал 500.
2. Magic numbers вынесены: _COST_PER_M2_MIN=1000, _COST_PER_M2_MAX=500_000.
3. _parse_floors docstring + inline note про malformed parts ("5а-7" filter,
multi-range "1-2-3" max acceptable degradation).
Frontend (NeighborsBlock.tsx):
5. Русский plural fix: pluralBuildings(n) helper — 1 здание, 2-4 здания,
5+/11-14 зданий. Раньше "3 зданий" — теперь "3 здания".
Не сделано (defer):
4. ST_Area для overlap query — практически 0-5 buildings в ЕКБ, GIST scan OK.
6. Discriminated union для NeighborsSummary — refactor а не bug.
7. Vault entry для P2 — добавится batch'ем после merge всех текущих PR.
Per auto-review on 60d53bb.
---------
Co-authored-by: lekss361 <claudestars@proton.me>
Backend (parcels.py):
- _compute_confidence() composite score 0..1 from 7 subscores: poi_freshness,
geom_source (parcel vs quarter), district, market_trend (rosreestr_deals depth),
competitors, environment (noise/air/weather availability), zoning (placeholder
до G1).
- confidence_label: high (>0.75) / medium (0.4-0.75) / low (<0.4)
- confidence_caveats: list of конкретных проблем для UI
- confidence_breakdown: per-subscore 0..1 для прозрачности
Это stub-версия (полная — после G1/G2/D1/D2). Использует только текущие сигналы.
Frontend:
- Новый ConfidenceBadge.tsx — color-coded (green/yellow/red) badge с %
- Caveats для low — показываются сразу; для medium/high — под toggle
- Toggle "Подробнее" → breakdown per-subscore + полный список caveats
- Размещён в начале OverviewTab (выше "Район")
- TS типы расширены: confidence, confidence_label, confidence_breakdown, confidence_caveats
Closes#48.
Co-authored-by: lekss361 <claudestars@proton.me>
* feat(site-finder): deep-link tabs via ?tab=env|land|market URL param
Tab state now syncs with URL — можно поделиться ссылкой на конкретный
таб (например /site-finder?tab=market для тренда цен по последнему cad).
- useRouter + useSearchParams (Next 15 app router, client-component)
- router.replace (не push) чтобы tab switches не засоряли history
- useEffect on searchParams → back/forward работают
- overview = default, не пишется в URL (clean)
- Type-guard isTabId(): только valid values из URL переключают state
* fix(site-finder): address PR #86 auto-review minor recommendations
- Use usePathname() for router.replace fallback (avoid dangling "?" in Next 15)
- Remove stale-closure guard in useEffect — React bails Object.is, no eslint-disable
- Wrap page in <Suspense> for useSearchParams (Next 15 App Router req for static)
- Lazy useState initializer — skip wasted CPU on re-renders
Per auto-review on c47e56b.
---------
Co-authored-by: lekss361 <claudestars@proton.me>
* docs(claude): mandate branch+PR workflow, ban direct push to main
* docs(claude): sync TL;DR + Don't-do-these with new PR workflow
Address auto-review feedback:
- TL;DR rule 5 → branch+PR mandate (was: "user commits themselves")
- TL;DR rule 6 → no --no-verify/--force/--amend (split from rule 5)
- Don't-do-these → drop conflicting "no auto-commit" line, add
"no direct push to main" + "no merge without approval"
- Critical workflow rules: add explicit rule 2 — worker-agents don't
commit themselves; main session commits on feature branch
- Restore "Squash on merge" wording in PR section (was dropped)
- Auto-mode section: clarify that main session does commits on feature
branch (worker-agents leave changes staged)
* fix(claude): Code-reviewer diagram had stale 'push origin HEAD:main' — replace with branch+PR flow
---------
Co-authored-by: lekss361 <claudestars@proton.me>
- nspd_geo: add _save_parcel() for thematic_id=1 → cad_parcels_geom (UPSERT,
ST_Transform from Web Mercator); _persist_target now handles 1/2/5
- parcels.py: analyze endpoint geom lookup extended with cad_parcels_geom as
3rd fallback source (after cad_quarters_geom, cad_buildings); both SELECT
and WKT subqueries updated
- parcels.py: POI score_breakdown items now include lat/lon for map markers
- poi_loader: OSM_CATEGORIES expanded — college+university→school,
hypermarket→shop_supermarket; coverage +3 tag pairs
Two UX nits:
- Caveats badge had whiteSpace:nowrap on a 250-char data_caveat → overflowed
visibly outside the section in the screenshot. Drop nowrap, keep small
font + light grey background. Reads as a compact note.
- Headline KPIs (369 млн ₽ · 19.4 мес · темп 6.0 · ликв 97/100) are
computed by backend at price_factor=1.0, while live KPI cards below
recompute with the current slider position. When user moves price to
+40% the discrepancy is confusing. Append small grey '(базовая цена)'
marker after headline with a tooltip explaining baseline vs live.
The comparables block on /analytics/recommend was showing the same ЖК
multiple times (ЖК СТАРТ × 3, ЭХО ЛЕСА × 2 etc) because domrf_kn_objects
has ~3 historical snapshots per obj_id. The previous query joined all
rows and LIMIT 5 cherry-picked duplicates by flat_count.
Fix: wrap base in CTE `latest_obj` with `DISTINCT ON (obj_id) ORDER BY
obj_id, snapshot_date DESC` to pick only the latest snapshot per object
before sorting by flat_count.
Previously `docker image prune -af --filter until=72h` only removed
images older than 72h. With multiple deploys/day we accumulate many
fresh (<72h) tagged images — each gendesign-backend / worker / frontend
SHA-tag is 500MB-1GB, filling /var/lib/docker.
New logic: for each ghcr.io/lekss361/gendesign-* repo, list tags sorted
by Created DESC, skip :latest, keep top 2, remove the rest. Still runs
the standard prune after for dangling layers + build cache.
Root cause of "auto-resume never fired": the kn_scrape_runs resume section
hit `if not rows: return` (and similar `return` in except) before reaching
the nspd_geo_jobs resume section. Whenever there were no zombie kn-runs
(the normal case), the handler bailed out and geo jobs stayed forever
'running' with stale heartbeats — users had to manual cancel/resume after
every deploy.
Fix: don't return early. Initialize `ids = []`, only run UPDATE if rows
exist, drop the inner `return` from exception branch. The for-loop over
ids becomes a no-op when empty, and execution falls through to the geo
section. Same pattern as the breadcrumb above — fail soft, continue.
cleanup_zombies beat task (added in caa467f) stays as belt-and-suspenders
in case worker_ready signal ever misbehaves again.
worker_ready signal handler was NOT firing in our setup (verified via
DB breadcrumb after 3 deploys — zero rows of stage='worker_ready' in
nspd_geo_log). Root cause of unreliability unknown — possibly Celery
internals, possibly compose recreate timing. Either way, after every
redeploy users had to manually cancel/resume jobs to keep them moving.
Replace signal-based resume with periodic beat task:
- cleanup_zombies runs every minute (* * * * *)
- Finds nspd_geo_jobs in status running/paused with heartbeat >2 min stale
- Sets status='queued' + apply_async with queue=geo
- Idempotent — if no zombies, no-op
worker_ready handler kept (with FK-fix breadcrumb on NULL job_id) for
diagnostic purposes — if signal ever does fire, we'll have evidence.
