feat(etl): housing-class normalization fallback via yandex_realty trigram match (#38) #1911

Merged
bot-backend merged 1 commit from feat/obj-class-normalization-38 into main 2026-06-26 07:44:26 +00:00

1 commit

Author SHA1 Message Date
95b2fb34e2 feat(etl): housing-class normalization fallback via yandex_realty trigram match (#38)
All checks were successful
CI / changes (pull_request) Successful in 6s
CI / frontend-tests (pull_request) Has been skipped
CI / openapi-codegen-check (pull_request) Successful in 2m14s
CI / backend-tests (pull_request) Successful in 11m20s
kn-API не отдаёт obj_class для ~3032 ЖК Свердл, что ломает per-class
velocity/absorption/weight. Добавлен fallback:

- 169_obj_class_fallback.sql: obj_class_fallback + obj_class_source колонки,
  pg_trgm + GIN trgm индексы на LOWER(comm_name) и LOWER(yandex_realty_zk.name).
- app/services/etl/obj_class_backfill.py: trigram-match comm_name ↔
  yandex_realty_zk.name (similarity >= 0.6, CROSS JOIN LATERAL) →
  obj_class_source='yandex_match'; price-inference по
  yandex_realty_class_prices → 'price_inference'. Idempotent, SAVEPOINT per row.
- Консьюмеры (analytics_queries / velocity / competitors): класс-фильтр и
  dominant_class читают COALESCE(obj_class, obj_class_fallback) — реальный
  класс в приоритете, поведение для размеченных ЖК не меняется.

Внутренний backfill — response-схема (parcel.py / OpenAPI) не меняется.

Refs #38
2026-06-26 12:26:13 +05:00