feat(tradein-imv): route IMV backfill через BrowserFetcher.fetch_json за dormant-флагом (#915 Stage 3) #1916

Merged
bot-backend merged 2 commits from feat/915-stage3-imv-browser into main 2026-06-26 09:54:29 +00:00

2 commits

Author SHA1 Message Date
e347303465 docs(tradein): document SCRAPER_FETCH_MODE + AVITO_IMV_USE_BROWSER_FETCHER in .env.example (#915)
All checks were successful
CI / openapi-codegen-check (pull_request) Has been skipped
CI / changes (pull_request) Successful in 6s
CI / backend-tests (pull_request) Has been skipped
CI / frontend-tests (pull_request) Has been skipped
2026-06-26 14:53:57 +05:00
0d010a707c feat(tradein-imv): route IMV backfill через BrowserFetcher.fetch_json за dormant-флагом (#915 Stage 3)
All checks were successful
CI / changes (pull_request) Successful in 7s
CI / backend-tests (pull_request) Has been skipped
CI / frontend-tests (pull_request) Has been skipped
CI / openapi-codegen-check (pull_request) Has been skipped
Транспорт-своп для avito IMV backfill: вместо curl_cffi 4 IMV-запроса
(warm-up → coords → position → get-data) ходят через камуфокс in-page fetch
(/fetch-json sidecar). Same-origin avito.ru + прокси + warmed-cookies +
реальный fingerprint обходят datacenter-403 (#562/#853) без отдельного
IMV-прокси. За dormant-флагом avito_imv_use_browser_fetcher (default False) —
деплой не меняет prod IMV-поведение.

- avito_imv.py: _BrowserResponse + _BrowserSessionAdapter (мимикрия под
  curl_cffi session/response: .get/.post/.close, .status_code/.text/.json/
  .raise_for_status). evaluate_via_imv получает browser_fetcher kwarg; при
  его наличии адаптер вместо curl-сессии (browser-путь не требует curl_cffi).
- config.py: avito_imv_use_browser_fetcher (ENV AVITO_IMV_USE_BROWSER_FETCHER).
- house_imv_backfill.py: backfill_house_imv при флаге открывает ОДИН
  BrowserFetcher(source="avito") на батч и прокидывает в _process_one_house →
  evaluate_via_imv. Флаг OFF → browser_fetcher=None → curl-путь как раньше.
  Backfill-only: estimator и sweep-путь (process_houses_imv_batch) не тронуты.
- browser_fetcher.py: _post_fetch_json защищён от ответа без status/body keys.
- test_server_fetch_json.py: тест crash-retry ветки _do_fetch_json (relaunch
  + второй _fetch_json_once после browser-crash).
- tests: _BrowserResponse/_BrowserSessionAdapter юниты, evaluate_via_imv
  browser happy-path + 403→IMVAuthError, backfill flag on/off проводка.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-26 14:42:26 +05:00