lekss361/gendesign
1
0
Fork 0
Code Issues 57 Pull requests Projects Releases Packages Wiki Activity Actions

[tradein][analytics] Elite/premium estimate accuracy — data-research findings + roadmap #2002

New issue
Open
opened 2026-06-27 15:22:02 +00:00 by lekss361 · 3 comments
lekss361 commented 2026-06-27 15:22:02 +00:00
Owner
Copy link

Большой data-research (2026-06-27) по систематической недооценке движком бизнес/элит (−21%/−38%). Шипнуто по ходу: гейт #1997, калиброванный 80% PI #2001.

Измерено (hermetic harness + прод-дамп n=2366, held-out)

  • g(es)-рекалибровка — ТУПИК: УХУДШАЕТ хвосты (элит −26%→−36%) — es сжат (R²=0.16), элит неотделим от комфорта при равном es (unidentifiability). НЕ шипить.
  • Премия дорогих — в analog-headline, не в ratio: пул / микро-компсы / ДКП-история / own-building asking ВСЕ занижают элит (0.79 / 0.71 / 0.59 / 0.765); own_ask/pool=0.92. → премия unit-level.
  • Композит (es + year + dkp_ratio + area): R² 0.12→0.31 (held-out, ×3). Помогает эконом/комфорт/бизнес (−22%→−19%, разворачивает вред g(es)); элит/премиум не двигает. year — звезда (corr 0.378 с остатком: движок недооценивает новостройки). floor есть на сделках → measurable, добавить в композит.

ЕКБ ≥20М₽ vs обычные (идентифицируемость элита)

area 115 vs 55 м² (2×), ppm² 254k vs 74k (3.4×). Описания: класс-слова (элит/премиум/бизнес-класс) 25%, вид/набережная 32%, дизайнер 14%. Концентрация по домам: Гоголя 21 (90 листингов), Жукова 16 (66), Энергостроителей 7 (61), Ельцина 6 (36)... → элит идентифицируем: (1) большая площадь [на сделках], (2) data-derived premium-building список, (3) класс-слова описаний (25%).

Источники класса (БД + код + live-браузер)

  • Avito «Класс жилья» — только новостройки (live-verified: вторичка не даёт; в houses.raw_characteristics, не извлекается; потолок ~620 = 4.5%).
  • Yandex по домам — 7 строк (мизер). Вторичный люкс (99% элита) структурного класса не имеет нигде → только description.

Roadmap (приоритет ROI)

  • Композит-PR (year+floor+area) — measurable, помогает эконом/комфорт/бизнес. Валидация через гейт (year/floor/area есть в фикстуре).
  • Description-mining: parse_house_class+infer_repair_state_from_text существуют, но подключены ТОЛЬКО к yandex-новострою — навесить на avito/yandex description → класс/отделка из текста (вторичный люкс). ~25% элит-описаний несут класс явно.
  • Premium-building список (data-derived из концентрации ≥20М) → premium-флаг + частичный uplift.
  • Manual-review флаг high-value/low-conf → ручная оценка (PR-2).
  • Forward-logging прод-оценок — единственный способ валидировать unit-level (отделка/класс/вид): ДКП их не несут.
  • Поднять avito-detail throughput (anti-bot) → наполняет repair/ceiling/finish.

Catch (хард-ограничение)

Ground-truth ДКП несут только floor/rooms/area/year/price. finish/view/class бэктестом НЕ валидируются → measurable-рычаги = year+floor+area; unit-сигналы только через forward-logging.

Bugs найдены попутно — отдельный issue.

Refs #1966

Большой data-research (2026-06-27) по систематической недооценке движком бизнес/элит (−21%/−38%). Шипнуто по ходу: гейт #1997, калиброванный 80% PI #2001. ## Измерено (hermetic harness + прод-дамп n=2366, held-out) - **g(es)-рекалибровка — ТУПИК**: УХУДШАЕТ хвосты (элит −26%→−36%) — `es` сжат (R²=0.16), элит неотделим от комфорта при равном `es` (unidentifiability). НЕ шипить. - **Премия дорогих — в analog-headline, не в ratio**: пул / микро-компсы / ДКП-история / own-building asking ВСЕ занижают элит (0.79 / 0.71 / 0.59 / 0.765); own_ask/pool=0.92. → премия **unit-level**. - **Композит (es + year + dkp_ratio + area): R² 0.12→0.31 (held-out, ×3)**. Помогает эконом/комфорт/**бизнес** (−22%→−19%, разворачивает вред g(es)); элит/премиум не двигает. `year` — звезда (corr 0.378 с остатком: движок недооценивает новостройки). **`floor` есть на сделках → measurable, добавить в композит.** ## ЕКБ ≥20М₽ vs обычные (идентифицируемость элита) area **115 vs 55 м² (2×)**, ppm² **254k vs 74k (3.4×)**. Описания: класс-слова (элит/премиум/бизнес-класс) **25%**, вид/набережная **32%**, дизайнер 14%. **Концентрация по домам**: Гоголя 21 (90 листингов), Жукова 16 (66), Энергостроителей 7 (61), Ельцина 6 (36)... → элит **идентифицируем**: (1) большая площадь [на сделках], (2) data-derived premium-building список, (3) класс-слова описаний (25%). ## Источники класса (БД + код + live-браузер) - Avito «Класс жилья» — **только новостройки** (live-verified: вторичка не даёт; в `houses.raw_characteristics`, не извлекается; потолок ~620 = 4.5%). - Yandex по домам — 7 строк (мизер). Вторичный люкс (99% элита) структурного класса не имеет нигде → **только `description`**. ## Roadmap (приоритет ROI) - [ ] **Композит-PR** (year+floor+area) — measurable, помогает эконом/комфорт/бизнес. Валидация через гейт (year/floor/area есть в фикстуре). - [ ] **Description-mining**: `parse_house_class`+`infer_repair_state_from_text` существуют, но подключены ТОЛЬКО к yandex-новострою — навесить на avito/yandex `description` → класс/отделка из текста (вторичный люкс). ~25% элит-описаний несут класс явно. - [ ] **Premium-building список** (data-derived из концентрации ≥20М) → premium-флаг + частичный uplift. - [ ] **Manual-review флаг** high-value/low-conf → ручная оценка (PR-2). - [ ] **Forward-logging прод-оценок** — единственный способ валидировать unit-level (отделка/класс/вид): ДКП их не несут. - [ ] Поднять avito-detail throughput (anti-bot) → наполняет repair/ceiling/finish. ## Catch (хард-ограничение) Ground-truth ДКП несут только floor/rooms/area/year/price. finish/view/class **бэктестом НЕ валидируются** → measurable-рычаги = year+floor+area; unit-сигналы только через forward-logging. Bugs найдены попутно — отдельный issue. Refs #1966
lekss361 added the
analytics
priority/p2
scope/backend
tradein
 labels 2026-06-27 15:22:34 +00:00
lekss361 commented 2026-06-27 17:34:00 +00:00
Author
Owner
Copy link

Tier-aware ratio (#928) — measured + diagnosed dead-end (removed)

Ran the live full-spine harness (700 prod deals) with tier_aware_ratio_enabled OFF vs ON:

  • бизнес bias −14.8→−8.5, элит −28.6→−24.0 (helps the tails)
  • BUT эконом +5.8→+18.1, комфорт −4.6→+5.3, overall MAPE 14.6→17.2, bias −2.9→+8.9 (worse)

Root-cause (Monte-Carlo, true sold/asking = const 0.84): the tier derivation bins SOLD deals by sold-ppm² against ASKING-percentile bounds (t33/t66 of listings), then divides within-tier medians of mismatched populations — a ratio-of-truncated-medians artifact. A simulation with a constant 0.84 reproduces the prod tier values (0.94/0.99/0.96) and the deal split (57/27/15% vs listings 33/33/33) almost exactly → the 0.84→0.99 gradient is 100% spurious. Correct binning (both sides by asking) is nearly flat (0.83→0.85); the true price-tier signal is ~2-3pp.

Conclusion: within-price-tier sold/asking is uncomputable from asking-less deals. The valid conditioning is non-price observables — the per-rooms blend (0.84) + the hedonic year+area (#2004), which is the statistically-correct version of price-segment correction. Hybrid (high-tier only) and re-derive are both moot (the high-tier 0.96 is also an artifact; re-derive needs per-deal asking the deals lack).

Dead tier path removed (footgun — flag flip degraded prod): branch fix/tradein-remove-tier-ratio-footgun. Tier tables (mig 098) left inert; optional DROP follow-up.

## Tier-aware ratio (#928) — measured + diagnosed dead-end (removed) Ran the live full-spine harness (700 prod deals) with `tier_aware_ratio_enabled` OFF vs ON: - бизнес bias −14.8→−8.5, элит −28.6→−24.0 (helps the tails) - BUT эконом +5.8→**+18.1**, комфорт −4.6→+5.3, **overall MAPE 14.6→17.2**, bias −2.9→+8.9 (worse) **Root-cause (Monte-Carlo, true sold/asking = const 0.84):** the tier derivation bins SOLD deals by sold-ppm² against ASKING-percentile bounds (t33/t66 of listings), then divides within-tier medians of mismatched populations — a **ratio-of-truncated-medians artifact**. A simulation with a *constant* 0.84 reproduces the prod tier values (0.94/0.99/0.96) and the deal split (57/27/15% vs listings 33/33/33) almost exactly → the 0.84→0.99 gradient is 100% spurious. Correct binning (both sides by asking) is nearly flat (0.83→0.85); the true price-tier signal is ~2-3pp. **Conclusion:** within-price-tier sold/asking is **uncomputable from asking-less deals**. The valid conditioning is non-price observables — the per-rooms blend (0.84) + the hedonic year+area (#2004), which is the statistically-correct version of price-segment correction. Hybrid (high-tier only) and re-derive are both moot (the high-tier 0.96 is also an artifact; re-derive needs per-deal asking the deals lack). Dead tier path removed (footgun — flag flip degraded prod): branch `fix/tradein-remove-tier-ratio-footgun`. Tier tables (mig 098) left inert; optional DROP follow-up.
lekss361 commented 2026-06-27 19:44:48 +00:00
Author
Owner
Copy link

IMV / Tier-S harness-extension — 5th measured dead-end (algo frontier closed)

Extended the harness (--resolve-house-id, #2025) to resolve each deal's house_id and fire the Tier-S canonical analog path + Avito-IMV anchor (prod fires them; the harness hard-coded None). Ran on/off, 700 prod deals:

  • Resolution coverage: 303/700 (43%) resolve a house_id; only 59 (8.4%) are IMV-reachable.
  • expected_sold per-segment OFF vs ON ≈ identical — overall MAPE 14.52→14.49 (noise), бизнес −14.97→−14.97, элит −28.56→−28.56 (zero movement), range-coverage unchanged.
  • Why: Avito-IMV covers mid-market, not elite (doesn't reach the underpriced elite deals); Tier-S canonical feeds the radius-median pool that already underprices elite. House-resolution is inert for accuracy.

Algo frontier — CLOSED (5 measured dead-ends)

g(es) recalibration · price-tier ratio (artifact) · corridor-clamp (inert) · outlier-filter (noise) · IMV/Tier-S (inert). The one shipped accuracy win remains the hedonic year+area (#2004). The residual бизнес/элит underprice is irreducibly unit-level (finish/class/view — not on ДКП deals). No further measurable algorithmic lever exists in the current data; the only paths are structural: forward-logging of live estimates + the premium_building→manual-review routing (#2006/#2019/#2021/#2022, shipped) for the high-value tail.

## IMV / Tier-S harness-extension — 5th measured dead-end (algo frontier closed) Extended the harness (`--resolve-house-id`, #2025) to resolve each deal's house_id and fire the Tier-S canonical analog path + Avito-IMV anchor (prod fires them; the harness hard-coded `None`). Ran on/off, 700 prod deals: - **Resolution coverage: 303/700 (43%) resolve a house_id; only 59 (8.4%) are IMV-reachable.** - **expected_sold per-segment OFF vs ON ≈ identical** — overall MAPE 14.52→14.49 (noise), **бизнес −14.97→−14.97, элит −28.56→−28.56 (zero movement)**, range-coverage unchanged. - Why: Avito-IMV covers mid-market, not elite (doesn't reach the underpriced elite deals); Tier-S canonical feeds the radius-median pool that already underprices elite. House-resolution is **inert for accuracy**. ### Algo frontier — CLOSED (5 measured dead-ends) g(es) recalibration · price-tier ratio (artifact) · corridor-clamp (inert) · outlier-filter (noise) · IMV/Tier-S (inert). The one shipped accuracy win remains the **hedonic year+area** (#2004). The residual бизнес/элит underprice is **irreducibly unit-level** (finish/class/view — not on ДКП deals). No further measurable algorithmic lever exists in the current data; the only paths are structural: forward-logging of live estimates + the premium_building→manual-review routing (#2006/#2019/#2021/#2022, shipped) for the high-value tail.
lekss361 commented 2026-07-04 06:35:56 +00:00
Author
Owner
Copy link

Триаж roadmap (tech-analyst, 2026-07-04, git archaeology + code grep):

  • Premium-building список — DONE (139_premium_houses.sql MV + 142_premium_buildings_curated.sql + _is_premium_building() estimator.py:3616)
  • Manual-review флаг — DONE (_manual_review() estimator.py:3666, триггеры premium/high-value/wide-range/elite-ppm2, UI banner)
  • Description-mining — PARTIAL: infer_repair_state_from_text подключён широко (avito/cian/domclick/yandex detail). parse_house_class — только на yandex-новострое; avito использует структурное поле "Класс жилья" (#2003/PR #2359), yandex-secondary/cian/domclick house_class из текста не парсят нигде. Не завожу issue автоматически — фаззи scope (structured-field vs текст, риск false-positive), нужен твой приоритет-колл.
  • Композит year+floor+area — PARTIAL: year+area шипнуты, floor реализован и протестирован, но забыт на неслитой ветке feat/tradein-hedonic-ground-floor (tip fe69d6e7, 356 коммитов позади main) → вынесено в отдельный узкий issue #2382 (ребейз+ревалидация+мерж).
  • Forward-logging — инфраструктура есть (каждая оценка уже пишется в trade_in_estimates с полными inputs/outputs), но консьюмера для forward-валидации против будущих реальных продаж нет — открытый мониторинг без чёткого DoD, не actionable в текущем виде.

Issue оставляю открытым — не весь roadmap закрыт (#2382 + description-mining extension решение за тобой).

Триаж roadmap (tech-analyst, 2026-07-04, git archaeology + code grep): - **Premium-building список** — DONE (`139_premium_houses.sql` MV + `142_premium_buildings_curated.sql` + `_is_premium_building()` estimator.py:3616) - **Manual-review флаг** — DONE (`_manual_review()` estimator.py:3666, триггеры premium/high-value/wide-range/elite-ppm2, UI banner) - **Description-mining** — PARTIAL: `infer_repair_state_from_text` подключён широко (avito/cian/domclick/yandex detail). `parse_house_class` — только на yandex-новострое; avito использует структурное поле "Класс жилья" (#2003/PR #2359), yandex-secondary/cian/domclick house_class из текста не парсят нигде. Не завожу issue автоматически — фаззи scope (structured-field vs текст, риск false-positive), нужен твой приоритет-колл. - **Композит year+floor+area** — PARTIAL: year+area шипнуты, **floor реализован и протестирован, но забыт на неслитой ветке** `feat/tradein-hedonic-ground-floor` (tip fe69d6e7, 356 коммитов позади main) → вынесено в отдельный узкий issue #2382 (ребейз+ревалидация+мерж). - **Forward-logging** — инфраструктура есть (каждая оценка уже пишется в `trade_in_estimates` с полными inputs/outputs), но консьюмера для forward-валидации против будущих реальных продаж нет — открытый мониторинг без чёткого DoD, не actionable в текущем виде. Issue оставляю открытым — не весь roadmap закрыт (#2382 + description-mining extension решение за тобой).
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
main
fix/2013-avito-cohort-null
fix/rbac-test-user2-expired
fix/expire-brusnika-trial-access
feat/tradein-domclick-house-sku-schema
chore/design-sync-tradein
fix/tradein-deploy-scraper-paths-filter
fix/tradein-no-numberless-house-insert
fix/tradein-dedup-fias-pass-no-geoguard
feat/tradein-fias-estimate-frontend
fix/tradein-dedup-by-fias
feat/tradein-fias-tier-matching
fix/tradein-v2-uiux-p1
fix/objective-lots-dedup-1964
fix/audit-followups-1953
feat/concept-house-picker-1965-stage3b
feat/concept-house-catalog-1965-stage3a
feat/concept-live-massing-1965-stage2b
feat/concept-recompute-1965-stage2a
feat/concept-section-1965-stage1
fix/best-layouts-per-object-snapshot-1956
fix/deficit-segmentation-1959
feat/tradein-hedonic-ground-floor
fix/report-map-points-1961
fix/report-microcopy-1963
fix/report-nspd-zoning-1962
fix/report-frontend-display-1953
perf/tradein-estimate-audit
perf/best-layouts-objects-first
fix/171-migration-path
perf/171-objective-crm-report-month-idx
fix/1926-app-log-stdout
fix/estimator-tiebreaker-nanalogs
fix/1917-fetch-json-network-retry
fix/1926-scrape-runs-total-seen
fix/1781-house-fk-source-identity
feat/39-osrm-per-category
feat/1746-utility-infra-map
fix/test-849-savepoint-count
fix/parking-economics-by-class
fix/zouit-engineering-area-gate
fix/zoning-resolution-decouple
feat/dcf-velocity-schedule
feat/zone-regulation-backfill
chore/mock-financial-estimate
fix/finmodel-split-disclosure-threshold
feat/finmodel-financing
feat/finmodel-sitefinder-bridge
feat/finmodel-dcf
feat/finmodel-price-calibration
feat/finmodel-cost-cascade
audit/1871-p2-estimator-dedup-split
audit/1871-p2-avito-coord
audit/1871-p3-uiux
audit/1871-p2-invest-moi
audit/1871-p2-velocity-market
audit/1871-p1-3-isreal-stubs
audit/1871-p1-2-ghost-anchor-guard
audit/1871-p1-scenarios-collapse
revert/site-finder-v2
fix/v2-remove-resource-legend
fix/v2-map-legend-position
feat/site-finder-v2-map-light-polish
rollback/site-finder-home-light
feat/site-finder-v2-cockpit
rollback/sf-old-analysis-design
feat/ptica-tabs-6-6-reports-params
fix/ptica-vri-row
feat/ptica-map-glyphs-sitefinder-dark
fix/ptica-3d-preview-product-verdict
fix/ptica-button-fontsize
fix/ptica-parity-data-mapping
chore/ptica-iso-preview-spin
fix/ptica-inter-map-product
fix/ptica-visual-parity
fix/ptica-left-rail
feat/ptica-cockpit-complete-bottom
feat/ptica-default-and-rich-map
feat/ptica-real-ksit-frontend
feat/tradein-ekb-geoportal-buildings
fix/sf-feedback-batch
feat/avito-newbuilding-sweep
feat/browser-concurrent-pool
fix/backfill-kadastr-column
fix/avito-proactive-ip-rotate
feat/activate-domclick-sweep
feat/avito-newbuilding-capture
fix/house-id-fk-realtime-link
fix/migration-162-table-if-exists
feat/riasurt-sverdl-layers-108
fix/sber-dataset-paths-902
fix/conflict-resolution-freshness-1539
feat/objective-lots-history-writer-307-obj6
feat/velocity-mv-61
fix/objective-mapping-is-reviewed-gate-307-obj2
fix/cian-detail-rich-fields
fix/avito-novostroyki-segment
feat/yandex-novostroyki-newflat
fix/ci-dedup-runner-storm-1709
fix/emiss-pk-granularity-1687
feat/listings-segment-param-1188
fix/inn-regex-word-boundary-1682
fix/excel-deficit-horizon-1590
fix/admin-scrape-validation
fix/admin-leads-interval-cast-1694
fix/dkp-corridor-doc-p10p90-1695
fix/weather-wind-d-allnone
test/residential-zone-rawprops-1681
fix/domrf-badge-regex-narrow-1686
feat/yandex-camoufox-transport
fix/score-sat-factor-1509
fix/base-pace-room-bucket-1593
fix/gate-verdict-no-blocker-1617
fix/confidence-deal-count-window-1637
fix/match-economics-1511-1512
fix/listings-partial-indexes-1398
fix/smart-suggestions-threshold-1503
fix/dkp-corridor-percentile-1520
fix/enrich-economics-window-1513
fix/leads-stats-window-1383
fix/poi-score-scale-1486
fix/geo-weight-floor-1633
fix/inn-regex-context-1640
fix/domrf-status-classify-1609
fix/emiss-period-dedup-1606
fix/cian-paginate-tail-1393
fix/normalize-zero-fill-guard-1638
fix/supply-view-devgroup-key-1650
fix/residential-zone-detect-1353
fix/score-label-threshold-1357
fix/trend-ratio-vprior-zero-1508
fix/district-median-dedup-1352
fix/supply-layers-overlap-1355
fix/pdf-expires-at-1530
fix/quartirography-area-bucket-1406
fix/cad-zones-multipolygon-1344
fix/sold-pct-market-verdict-1407
fix/week-review-finish
fix/week-review-backend-v2
feat/plan-vectorization-spike-299
fix/1668-analyze-confidence
fix/1667-metro-block
fix/1666-layout-velocity-beat
chore/remove-n1-scraper
fix/cian-kitchen-area
feat/yandex-detail-backfill
feat/avito-detail-backfill
fix/best-layouts-bucket-normalization
fix/gknspecial-incremental-commit-per-prefix
fix/okn-objects-full-registry-load
fix/parcel-ird-overlap-except-expand
fix/report-assembler-confounded-keys
fix/job-settings-json-persist-v2
fix/nspd-geo-soft-time-limit-passthrough
fix/nspd-soft-time-limit-v2
fix/pilot-request-modal-validation
fix/recommend-mix-saturation-latest-snapshot
fix/scrape-cadastre-fail-counter-v2
fix/izyatie-ocr-cad-normalize
fix/job-settings-panel-json-state-persist
fix/noise-loader-savepoint-per-row
fix/objective-sync-config-bare-except
fix/parcel-snapshot-mdash-double-escape
fix/poi-loader-leap-year-date-replace
fix/scrape-cadastre-permanent-fail-counter
fix/usermenu-analyst-role-label
feat/location-948b-r
feat/location-entity-948b
feat/insight-entity-948a
feat/analyst-role-audit-962
feat/cbr-inflation-macro-946
feat/cian-zhk-resolve-972
perf/forecast-parallel-horizons-1129
feat/newbuilding-enrich-backfill-972
feat/emiss-macro-indicators-946
feat/rosstat-emiss-macro-946
ci/frontend-vitest-job
chore/frontend-vitest-wkt-tests
feat/forecast-meta-footer
feat/forecast-future-supply-competitors
feat/forecast-docx-pptx-buttons
feat/forecast-scoring-panel
feat/forecast-product-tz-block
feat/market-map-layers-999
feat/forecast-chart-section6
feat/forecast-export-pptx-959
feat/forecast-export-docx-959
feat/forecast-export-ui-959
feat/forecast-export-telegram-959
feat/forecast-export-md-959
feat/analyze-response-contract-961
feat/competitor-velocity-spatial-fallback-968
fix/frontend-healthcheck-hostname
feat/analyze-market-coords-999
fix/site-finder-entry-crash-1001
feat/horizon-selector-996
feat/forecast-section-998
feat/forecast-endpoint-horizon-enqueue
feat/forecast-celery-task-persist
feat/forecast-orchestrator-service
feat/demand-supply-months-of-inventory
feat/wire-district-resolver
feat/district-vocab-reconciliation
fix/affordability-market-spread
fix/backtest-sig-gate
feat/wire-978-almon-rate-regime
feat/backtest-oos-almon-seasonal
ci/forgejo-pytest-gate
fix/tradein-praktika-logo-form-scroll
feat/yandex-exhaustive-price-room
feat/794-sber-time-adjust
chore/bot-harness-tuning
docs/agent-usage-throttle-qa-terminal
fix/avito-sweep-single-run-guard
chore/analyst-inbound-queue-tradein-db
feat/801-preview-full-wt
chore/bot-safety-followups
feat/tradein-avito-imv-fix
feat/tradein-repair-coef-calibration
feat/tradein-avito-houses-fix
feat/tradein-address-precision
feat/tradein-backfill-prioritize-listings
feat/tradein-geocode-deals-nominatim
feat/tradein-geocode-deals-from-houses
feat/tradein-same-building-match
feat/claude-envvars-and-stale-cron
No results found.
Labels
Clear labels   
admin

Admin UI / internal tools / pipeline tracker

  
analytics

Аналитика данных / matching

  
auth

Authentication / authorization / multi-tenancy

  
automation

Автоматизация / cron / scheduler

  
bug

Something isn't working

  
business

Коммерческая воронка / пилоты / pipeline / реестр ПО

  
chore

Operational housekeeping

  
ci

Forgejo Actions / build pipeline

  
compliance

152-ФЗ / реестр ПО / legal / TOS

  
data

Data sources, parsers, ETL

  
data-moat

Уникальные данные — sales tracker, CRM aggregation

  
docs

README / runbooks / onboarding

  
duplicate

This issue or pull request already exists

  
dx

Developer experience / build speed

  
enhancement

New feature or request

  
Fable 5 ревью

Находки полного мульти-агентного ревью (Claude Fable 5, 2026-06-09)

  
feedback/max

Feedback from Max (Senior partner)

  
generative

Generative Design — concept generation engine

  
GG-форсайт

Новое ТЗ от Ген Геныча (июнь 2026): Site Finder v2 — ИИ-прогноз продуктовой стратегии девелопера. Актуальный приоритет.

  
needs-discussion

Pending team sync / decision

  
needs-human

Human review/decision required

  
observability

Monitoring / metrics / logs / Sentry / Uptime

  
pause-bots

Kill-switch — all autonomous bots pause

  
performance

Latency / throughput / DB query tuning

  
priority/p0

Critical / production-blocker

  
priority/p1

High priority

  
priority/p2

Normal priority

  
priority/p3

Низкий приоритет (follow-up, tech-debt)

  
scope/backend

Backend Python — FastAPI, Celery, scrapers

  
scope/db

DB migrations, indexes, SQL

  
scope/devops

docker-compose, Caddy, workflows

  
scope/frontend

Frontend — Next.js, React, TypeScript

  
scope/qa

Tests, smoke, Playwright

  
scrapers

Скрапперы (Avito/Cian/Yandex/Rosreestr)

  
security

Security audit / secrets / vulnerabilities

  
site-finder

Site Finder scoring quality & data layer

  
stage/1

Stage 1: Generative MVP (нед 2-7)

  
stage/2

Stage 2: Site Finder MVP (нед 8-12)

  
status/blocked

Blocked — needs human intervention

  
status/done

Closed, completed

  
status/needs-analysis

Raw/fuzzy task filed by human — auto-analyst подхватывает: code-archeology + декомпозиция в actionable scope/* + status/ready (или split на под-issues)

  
status/needs-fix

Reviewer/QA вернул worker на доработку — worker сам чинит свой PR (fixup), НЕ требует human (в отличие от status/blocked)

  
status/qa

PR merged, awaiting smoke test

  
status/ready

Ready for pickup by worker bot

  
status/review

PR open, awaiting code review

  
status/wip

Work in progress (worker claimed)

  
tech-debt

Технический долг

  
tradein

Trade-In MVP проект

  
ux

User experience / responsive / conversion

  
week ревью 1

Week review 1 — находки multi-agent аудита проекта (2026-06-14)

  
wontfix

This will not be worked on

  
вторичка

Trade-In Estimator — проект оценки вторичного жилья

  
ИРД

Исходно-разрешительная документация — слой ограничений участка для analyze-таблицы

No labels
admin
analytics
auth
automation
bug
business
chore
ci
compliance
data
data-moat
docs
duplicate
dx
enhancement
Fable 5 ревью
feedback/max
generative
GG-форсайт
needs-discussion
needs-human
observability
pause-bots
performance
priority/p0
priority/p1
priority/p2
priority/p3
scope/backend
scope/db
scope/devops
scope/frontend
scope/qa
scrapers
security
site-finder
stage/1
stage/2
status/blocked
status/done
status/needs-analysis
status/needs-fix
status/qa
status/ready
status/review
status/wip
tech-debt
tradein
ux
week ревью 1
wontfix
вторичка
ИРД
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: lekss361/gendesign#2002
No description provided.
Delete branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?