fix(tradein): sweep shared-session (SERP+detail) через mobile proxy (#623) (#805)

Co-authored-by: bot-backend <bot-backend@gendsgn.local>
Co-committed-by: bot-backend <bot-backend@gendsgn.local>

tradein-mvp/backend/app/services/scrape_pipeline.py

@@ -34,6 +34,7 @@ from curl_cffi.requests import AsyncSession
 from sqlalchemy import text
 from sqlalchemy.orm import Session
 
+from app.core.config import settings
 from app.services import scrape_runs
 from app.services.scrapers.avito import AvitoScraper
 from app.services.scrapers.avito_detail import fetch_detail, save_detail_enrichment
@@ -66,6 +67,19 @@ _CHROME_HEADERS = {
 }
 
 
+def _avito_proxies() -> dict[str, str] | None:
+    """#623: mobile-proxy egress для shared-сессии sweep'а.
+
+    КРИТИЧНО: sweep создаёт собственную shared AsyncSession и присваивает её
+    `scraper._cffi`, перезатирая проксированную сессию из `AvitoScraper.__aenter__`.
+    Через эту же сессию идёт и SERP, и detail-энричмент (fetch_detail). Без прокси
+    detail-страницы летят с datacenter-IP → HTTP 429 → весь run mark_banned.
+    Пусто (env не задан) → прямое подключение (dev/no-op).
+    """
+    url = settings.avito_proxy_url
+    return {"http": url, "https": url} if url else None
+
+
 # ── Result dataclasses ──────────────────────────────────────────
 
 
@@ -133,6 +147,7 @@ async def run_avito_pipeline(
         impersonate="chrome120",
         timeout=25,
         headers=_CHROME_HEADERS,
+        proxies=_avito_proxies(),
     )
 
     try:
@@ -351,6 +366,7 @@ async def run_avito_city_sweep(
         impersonate="chrome120",
         timeout=25,
         headers=_CHROME_HEADERS,
+        proxies=_avito_proxies(),
     ) as session:
         try:
             for idx, (lat, lon, name) in enumerate(_anchors, start=1):

tradein-mvp/backend/tests/scrapers/test_avito_anti_bot.py

@@ -411,3 +411,22 @@ async def test_fetch_around_blocked_when_rotation_exhausted(
             await scraper.fetch_around(56.84, 60.60, 1000, pages=1)
 
     assert rot.await_count == 1  # одна попытка ротации, потом сдаёмся
+
+
+# ── #623 sweep-pipeline proxy (SERP+detail через mobile proxy) ───
+
+
+def test_avito_proxies_helper(monkeypatch: pytest.MonkeyPatch) -> None:
+    """_avito_proxies: None без env, dict {http,https} при заданном AVITO_PROXY_URL.
+
+    Регрессия: sweep создаёт shared-сессию, через неё идут и SERP, и detail —
+    без прокси detail летит с datacenter-IP → 429 → run banned (#623 followup).
+    """
+    from app.services import scrape_pipeline as sp
+
+    monkeypatch.setattr(sp.settings, "avito_proxy_url", None)
+    assert sp._avito_proxies() is None
+
+    url = "http://u:p@mproxy.site:16928"
+    monkeypatch.setattr(sp.settings, "avito_proxy_url", url)
+    assert sp._avito_proxies() == {"http": url, "https": url}