fix(tradein): avito house-link из detail SSR — SERP-карточка JS-only (#871) (#873)

Co-authored-by: lekss361 <lekss361@gendsgn.local>
Co-committed-by: lekss361 <lekss361@gendsgn.local>

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

@@ -268,6 +268,7 @@ async def run_avito_pipeline(
             )
 
             consecutive_blocks = 0
+            detail_house_paths: set[str] = set()
             for idx, row in enumerate(priority_rows):
                 source_url: str = row["source_url"]
                 counters.detail_attempted += 1
@@ -278,6 +279,12 @@ async def run_avito_pipeline(
                     enrichment_detail = await fetch_detail(item_url, cffi_session=session)
                     if save_detail_enrichment(db, enrichment_detail):
                         counters.detail_enriched += 1
+                    # #871: house-link есть в detail SSR (в SERP-карточке — JS-only).
+                    # Копим catalog-пути для house-enrich ниже (Step 5b).
+                    if enrichment_detail.house_catalog_url:
+                        hp = urlparse(enrichment_detail.house_catalog_url).path
+                        if hp:
+                            detail_house_paths.add(hp)
                     consecutive_blocks = 0
                 except (AvitoBlockedError, AvitoRateLimitedError) as e:
                     consecutive_blocks += 1
@@ -310,6 +317,35 @@ async def run_avito_pipeline(
                     jitter = random.uniform(0.8, 1.2)
                     await asyncio.sleep(detail_delay * jitter)
 
+            # ── Step 5b: enrich houses, найденные через detail-страницы (#871) ──
+            # SERP-карточки больше не несут house-link (JS-render) → дома берём из
+            # detail SSR. Дедуп против обработанных в Step 4 (unique_house_paths).
+            new_house_paths = detail_house_paths - unique_house_paths
+            if enrich_houses and new_house_paths:
+                counters.unique_houses += len(new_house_paths)
+                nh_list = list(new_house_paths)
+                for h_idx, house_path in enumerate(nh_list):
+                    try:
+                        enrichment = await fetch_house_catalog(house_path, cffi_session=session)
+                        house_counters = save_house_catalog_enrichment(db, enrichment)
+                        counters.houses_enriched += 1
+                        hid = house_counters.get("house_id")
+                        if hid:
+                            touched_house_ids.add(int(hid))
+                    except (AvitoBlockedError, AvitoRateLimitedError):
+                        logger.error(
+                            "pipeline:house(detail) BLOCKED at %s — propagating", house_path
+                        )
+                        raise
+                    except Exception as e:
+                        logger.warning(
+                            "pipeline:house(detail) enrich failed for %s: %s", house_path, e
+                        )
+                        counters.houses_failed += 1
+                        counters.errors.append(f"house(detail) {house_path}: {e}")
+                    if h_idx < len(nh_list) - 1:
+                        await asyncio.sleep(detail_delay)
+
         logger.info(
             "pipeline:done anchor=(%.4f,%.4f) lots=%d (ins=%d/upd=%d) "
             "houses=%d/%d detail=%d/%d touched_house_ids=%d errors=%d",

tradein-mvp/backend/app/services/scrapers/avito_detail.py

@@ -25,7 +25,7 @@ import re
 from dataclasses import dataclass, field
 from datetime import date
 from typing import Any
-from urllib.parse import urljoin
+from urllib.parse import urljoin, urlparse
 
 from curl_cffi.requests import AsyncSession
 from selectolax.parser import HTMLParser, Node
@@ -319,12 +319,19 @@ def parse_detail_html(html: str, source_url: str) -> DetailEnrichment:
         metro_stations = _extract_metro(description)
 
     # ── House catalog link ───────────────────────────────────────────────────
+    # Новостройки отдают кнопку ЖК (nd-jk-details-button); вторичка — ссылку
+    # «узнать больше о доме» под другим маркером. Fallback на любой /catalog/houses/
+    # якорь: verified 2026-05-31 (#871) — эта ссылка есть в SSR detail-страницы,
+    # даже когда отсутствует в JS-рендеренной SERP-карточке (откуда мы брали 0).
     house_catalog_url: str | None = None
     house_link = tree.css_first("[data-marker='nd-jk-details-button']")
+    if house_link is None:
+        house_link = tree.css_first("a[href*='/catalog/houses/']")
     if house_link is not None:
         href = house_link.attributes.get("href", "")
         if href:
-            house_catalog_url = urljoin(AVITO_BASE, href)
+            # отбрасываем ?context-трекинг → чистый catalog-путь
+            house_catalog_url = urljoin(AVITO_BASE, urlparse(href).path)
 
     # ── Photo gallery ────────────────────────────────────────────────────────
     photo_urls: list[str] = []

tradein-mvp/backend/tests/test_avito_detail_parse.py

@@ -4,6 +4,8 @@
 без сетевых запросов.
 """
 
+# ruff: noqa: E501  # HTML-фикстуры содержат длинные строки разметки
+
 import pytest
 
 from app.services.scrapers.avito_detail import DetailEnrichment, parse_detail_html
@@ -136,6 +138,24 @@ def test_house_catalog_url_full() -> None:
     assert result.house_catalog_url.startswith("https://www.avito.ru")
 
 
+def test_house_catalog_url_secondary_marker_no_nd_jk() -> None:
+    """#871: вторичка без nd-jk-details-button → fallback на /catalog/houses/ якорь;
+    ?context-трекинг отброшен (verified live: house-link есть в detail SSR)."""
+    html = """
+    <html><body>
+    <div data-marker="item-view/item-id">№ 7683141312</div>
+    <span itemprop="price" content="4000000">4 000 000 ₽</span>
+    <a href="/catalog/houses/ekaterinburg/ul_gotvalda_6k1/142792?context=H4sIabc">
+      Узнать больше о доме</a>
+    </body></html>
+    """
+    result = parse_detail_html(html, "https://www.avito.ru/test_7683141312")
+    assert (
+        result.house_catalog_url
+        == "https://www.avito.ru/catalog/houses/ekaterinburg/ul_gotvalda_6k1/142792"
+    )
+
+
 def test_bathroom_type_combined() -> None:
     html = """
     <html><body>