fix: remove duplicate index=True from migration 004 columns

improve batch sync

add postgres upsert

fix sync locking

improve listing sync

speed up scraper

clean up project

prepare for github

update docker setup

.env.example

@@ -1,66 +1,57 @@
 IAAI_HEADLESS=true
 IAAI_LOG_LEVEL=INFO
-# IAAI_LOG_FILE=iaai_scraper.log
 
-# Network capture settings.
 IAAI_CAPTURE_SAME_ORIGIN_ONLY=true
 IAAI_MAX_CAPTURED_REQUESTS=40
-IAAI_MAX_CAPTURED_JSON_RESPONSES=30
+IAAI_MAX_CAPTURED_JSON_RESPONSES=20
 
-# Sequential listing-first mode.
 IAAI_CARS_LISTING_URL=https://www.iaai.com/Vehiclelisting/Cars
-IAAI_MAX_PAGES_PER_RUN=10
+IAAI_LISTING_SEGMENTS=auto
-IAAI_MAX_VEHICLES_PER_RUN=30
+IAAI_MAX_PAGES_PER_RUN=999999
-IAAI_PAGE_LINK_LIMIT=100
+IAAI_MAX_VEHICLES_PER_RUN=999999
+IAAI_PAGE_LINK_LIMIT=999999
 IAAI_INCLUDE_PAGINATION=true
 IAAI_COLLECT_CURRENT_PAGE_ONLY=false
 
-# Human-like pacing.
 IAAI_HUMAN_PACE_ENABLED=true
-IAAI_AFTER_LISTING_OPEN_MIN_S=2.5
+IAAI_PARALLEL_TABS=10
-IAAI_AFTER_LISTING_OPEN_MAX_S=4.5
+CELERY_BATCH_SIZE=100
-IAAI_AFTER_FILTER_ACTION_MIN_S=2.0
+IAAI_AFTER_LISTING_OPEN_MIN_S=0.2
-IAAI_AFTER_FILTER_ACTION_MAX_S=4.0
+IAAI_AFTER_LISTING_OPEN_MAX_S=0.5
-IAAI_BEFORE_VEHICLE_OPEN_MIN_S=1.5
+IAAI_AFTER_FILTER_ACTION_MIN_S=0.2
-IAAI_BEFORE_VEHICLE_OPEN_MAX_S=3.0
+IAAI_AFTER_FILTER_ACTION_MAX_S=0.5
-IAAI_AFTER_VEHICLE_OPEN_MIN_S=1.0
+IAAI_BEFORE_VEHICLE_OPEN_MIN_S=0.02
-IAAI_AFTER_VEHICLE_OPEN_MAX_S=2.5
+IAAI_BEFORE_VEHICLE_OPEN_MAX_S=0.08
-IAAI_BETWEEN_VEHICLES_MIN_S=3.0
+IAAI_AFTER_VEHICLE_OPEN_MIN_S=0.02
-IAAI_BETWEEN_VEHICLES_MAX_S=6.0
+IAAI_AFTER_VEHICLE_OPEN_MAX_S=0.08
-IAAI_AFTER_PAGE_CHANGE_MIN_S=3.0
+IAAI_BETWEEN_VEHICLES_MIN_S=0.02
-IAAI_AFTER_PAGE_CHANGE_MAX_S=6.0
+IAAI_BETWEEN_VEHICLES_MAX_S=0.08
+IAAI_AFTER_PAGE_CHANGE_MIN_S=0.2
+IAAI_AFTER_PAGE_CHANGE_MAX_S=0.5
 
-# Sync settings
+IAAI_SYNC_ONLY_NEW=false
-IAAI_SYNC_ONLY_NEW=true
 IAAI_TOKENS_FILE=/data/tokens.json
 IAAI_RUNTIME_CONFIG_FILE=/app/runtime_config.json
 
-# Retry / backoff
+IAAI_MAX_RETRIES=5
-IAAI_RETRY_DELAY_SECONDS=2.5
+IAAI_RETRY_DELAY_SECONDS=4
 IAAI_RETRY_BACKOFF_MULTIPLIER=2.0
-IAAI_RETRY_JITTER_SECONDS=0.25
+IAAI_RETRY_JITTER_SECONDS=0.5
+IAAI_TIMEOUT_MS=90000
+IAAI_FALLBACK_NAV_TIMEOUT_MS=30000
 
-# Proxy (HTTP/HTTPS preferred for Playwright)
-# Chromium does not support SOCKS5 proxy authentication directly.
-# Use residential or mobile USA proxy.
-# IAAI_PROXY_SERVER=http://proxy.example.com:8080
-# IAAI_PROXY_USERNAME=
-# IAAI_PROXY_PASSWORD=
-
-# Database (PostgreSQL).
 IAAI_DATABASE_URL=postgresql+psycopg2://iaai:iaai@postgres:5432/iaai_scraper
 IAAI_DATABASE_ECHO=false
 IAAI_DATABASE_POOL_SIZE=5
-IAAI_DATABASE_MAX_OVERFLOW=10
+IAAI_DATABASE_MAX_OVERFLOW=5
 
-# ─── Redis (Celery broker) ─────────────────────────────────
 IAAI_REDIS_URL=redis://redis:6379/0
 
-# ─── Celery ────────────────────────────────────────────────
 CELERY_BROKER_URL=redis://redis:6379/0
 CELERY_RESULT_BACKEND=redis://redis:6379/0
-CELERY_TASK_SOFT_TIME_LIMIT=600
+CELERY_TASK_SOFT_TIME_LIMIT=86400
-CELERY_TASK_TIME_LIMIT=900
+CELERY_TASK_TIME_LIMIT=86520
+CELERY_BROKER_VISIBILITY_TIMEOUT=90000
 CELERY_WORKER_CONCURRENCY=1
 CELERY_BEAT_SYNC_INTERVAL_MINUTES=60
-CELERY_BEAT_SYNC_LIMIT=26
+CELERY_BEAT_SYNC_LIMIT=0

.gitignore

@@ -2,6 +2,8 @@ __pycache__/
 .pytest_cache/
 .coverage
 coverage.xml
+.env*
+!.env.example
 .venv/
 venv/
 *.pyc
@@ -9,7 +11,7 @@ venv/
 *.pyd
 *.log
 *.db
-.env
+storage_state.json
 .vscode/
 *.egg-info/
 dist/
@@ -17,3 +19,12 @@ build/
 artifacts/
 celerybeat-schedule*
 tokens_data/
+tokens.json
+
+# Local deployment/debug artifacts
+/*.tgz
+/*.tar.gz
+/*.zip
+/NOW()
+/_speed_check.sql
+/.env.bak*

Dockerfile

@@ -3,23 +3,34 @@ FROM mcr.microsoft.com/playwright/python:v1.58.0-noble
 ENV PYTHONDONTWRITEBYTECODE=1 \
     PYTHONUNBUFFERED=1
 
-WORKDIR /app
+RUN groupadd --system app && useradd --system --gid app --create-home app
 
-# Xvfb for headed Chromium in container (IAAI blocks headless)
+WORKDIR /app
-RUN apt-get update -qq && apt-get install -y --no-install-recommends xvfb \
-    && rm -rf /var/lib/apt/lists/*
 
 COPY pyproject.toml uv.lock ./
 RUN pip install --no-cache-dir uv \
     && uv export --format requirements-txt --no-dev --no-hashes --no-emit-project --frozen -o /tmp/requirements.txt \
-    && pip install --no-cache-dir -r /tmp/requirements.txt
+    && pip install --no-cache-dir -r /tmp/requirements.txt \
+    && pip install --no-cache-dir playwright-stealth
+
+# Speed-up libs: orjson (fast JSON parse), brotli (HTTP br decompress).
+# Pinned outside uv.lock to avoid lock-regen churn.
+RUN pip install --no-cache-dir "orjson>=3.10.0" "brotli>=1.1.0"
+
+# Install both Chromium and Firefox. Chromium is the default engine in Docker
+# because it works more reliably with the current IAAI listing page.
+RUN python -m playwright install chromium firefox
 
 COPY . .
+RUN pip install --no-cache-dir -e .
 RUN python -m compileall -q iaai_scraper
 RUN chmod +x entrypoint.sh
+RUN chown -R app:app /app
 
 STOPSIGNAL SIGINT
 
+USER app
+
 # По умолчанию  API, но worker/beat переопределяют CMD в docker-compose
 ENTRYPOINT ["./entrypoint.sh"]
 CMD ["uvicorn", "iaai_scraper.api.app:app", "--host", "0.0.0.0", "--port", "8000"]

README.md

@@ -12,7 +12,94 @@
 
 Поэтому в проекте используются Playwright, паузы между действиями, прокси и сохранение браузерного состояния.
 
-## Запуск
+## Локальный запуск (без Docker)
+
+### Требования
+
+- **Python 3.11+**
+- **Git**
+
+Docker **не нужен**. Данные хранятся в SQLite-файле `iaai_scraper.db` в корне проекта.
+
+### Быстрый старт
+
+```bash
+git clone <repo-url>
+cd iaai_scraper_project
+pip install -e .
+playwright install firefox
+iaai init-db
+```
+
+`iaai init-db` актуален для SQLite/локального CLI-режима. Для PostgreSQL используйте Alembic-миграции (`alembic upgrade head` или сервис `migrate` в Docker).
+
+Готовый `.env` уже в репозитории и настроен на SQLite  ничего менять не нужно.
+
+### Запуск парсера
+
+**Скрапинг одной машины:**
+
+```bash
+iaai sync-vehicle "https://www.iaai.com/VehicleDetail/45089484~US"
+```
+
+**Сбор листинга + скрапинг (например Toyota, 10 штук):**
+
+```bash
+iaai sync-listing --make Toyota --limit 10
+```
+
+**Только ссылки с листинга (без скрапинга):**
+
+```bash
+iaai collect-listing --make Toyota
+```
+
+**С видимым браузером (для отладки):**
+
+```bash
+iaai --headless false sync-vehicle "https://www.iaai.com/VehicleDetail/45089484~US"
+```
+
+**Проверка, что записалось в базу (`iaai_scraper.db`):**
+
+```bash
+python -c "import sqlite3; c=sqlite3.connect('iaai_scraper.db'); q=c.cursor(); print('cars:', q.execute('select count(*) from cars').fetchone()[0]); print('images:', q.execute('select count(*) from images').fetchone()[0]); rows=q.execute('select id, brand, model, year, origin_id from cars order by id desc limit 10').fetchall(); [print(r) for r in rows]"
+```
+
+Результаты сохраняются в `artifacts/json/` и в БД `iaai_scraper.db`.
+
+### Полный стек (API + Worker + Beat)
+
+Для API и автоматического сбора нужны **Redis** и **PostgreSQL**. В `.env` раскомментируй строки Redis/Celery и замени БД на PostgreSQL:
+
+```env
+IAAI_DATABASE_URL=postgresql+psycopg2://iaai:iaai@localhost:5432/iaai_scraper
+IAAI_REDIS_URL=redis://localhost:6379/0
+CELERY_BROKER_URL=redis://localhost:6379/0
+CELERY_RESULT_BACKEND=redis://localhost:6379/0
+```
+
+Применить миграции и запустить в трёх терминалах:
+
+```bash
+alembic upgrade head
+
+# Терминал 1 — API
+uvicorn iaai_scraper.api.app:app --reload --port 8000
+
+# Терминал 2 — Celery Worker
+celery -A iaai_scraper.worker.celery_app worker --loglevel=info --concurrency=1 --pool=solo -Q scraping
+
+# Терминал 3 — Celery Beat (периодический запуск)
+celery -A iaai_scraper.worker.celery_app beat --loglevel=info
+```
+
+API: `http://localhost:8000` · Swagger: `http://localhost:8000/docs`
+
+---
+
+## Запуск через Docker (все сервисы в контейнерах)
 
 ```bash
 cp .env.example .env
@@ -21,16 +108,39 @@ docker compose up -d
 
 Будут запущены сервисы `postgres`, `redis`, `migrate`, `api`, `worker`, `beat`. API доступен на `http://localhost:8000`.
 
-В  Docker-образ дополнительно проверяется Python-синтаксис на этапе сборки (`python -m compileall -q iaai_scraper`), чтобы не выкатывать битый код.
+В Docker-образ дополнительно проверяется Python-синтаксис на этапе сборки (`python -m compileall -q iaai_scraper`), чтобы не выкатывать битый код.
 
 `entrypoint.sh` поднимает SOCKS5→HTTP proxy bridge (если задан `SOCKS5_PROXY_HOST`) и запускает `Xvfb` только для `worker` и CLI scraping-команд.
 
-По умолчанию `beat` запускает сбор листинга **раз в 1 час** и обрабатывает **до 26 машин за запуск** (`CELERY_BEAT_SYNC_INTERVAL_MINUTES=60`, `CELERY_BEAT_SYNC_LIMIT=26`).
+По умолчанию `beat` запускает синхронизацию **раз в 1 час** (`CELERY_BEAT_SYNC_INTERVAL_MINUTES=60`).
+
+Текущая стратегия устойчива к багам пагинации IAAI:
+
+- **первый запуск** делает полный bootstrap через `sitemap`;
+- если bootstrap не завершился, автоматически продолжается с чекпоинта;
+- **после первого полного прохода** каждый час выполняется проход по `sitemap`.
+  По умолчанию включён режим **rolling refresh**:
+  - берутся все URL из sitemap,
+  - новые URL добавляются сразу,
+  - исчезнувшие URL помечаются как `is_sold=true`,
+  - уже записанные авто обновляются **батчами по кругу**, а не все разом.
+
+  Это даёт более стабильную нагрузку и снижает риск не уложиться в час.
+
+  Доступные режимы:
+  - `IAAI_HOURLY_MODE=rolling_refresh` — рекомендуемый продовый режим;
+  - `IAAI_HOURLY_MODE=full_refresh` — перепарсить все активные авто за один hourly цикл;
+  - `IAAI_HOURLY_MODE=diff` — только новые авто + sold.
+
+За счёт этого система:
+
+- не зависит от page 39 / 70 / 100,
+- не уходит в бесконечный цикл пагинации,
+- даёт стабильный hourly refresh уже записанных авто без монолитного полного hourly прохода,
+- остаётся быстрой и устойчивой при hourly sync.
 
 Swagger-документация: `http://localhost:8000/docs`
 
-Проверка состояния сервисов:
-
 ```bash
 docker compose ps
 ```
@@ -74,6 +184,8 @@ iaai sync-vehicle "https://www.iaai.com/VehicleDetail/45089484~US"
 iaai sync-listing --limit 10
 ```
 
+`iaai init-db` используйте для SQLite/локальных тестов. В PostgreSQL-сценарии применяйте миграции Alembic.
+
 ## Структура
 
 ```text
@@ -184,8 +296,6 @@ uv run pytest -q
 - `only_new`
 - `limit`
 
-Так же  используются: `lane`, `only_new`, `limit`.
-
 ### Секция `filters`
 
 Поддерживаются поля:

alembic/versions/001_initial.py

@@ -1,9 +1,4 @@
-"""Initial schema — cars, images, sync_runs
+# Начальная схема: cars, images, sync_runs
-
-Revision ID: 001_initial
-Revises:
-Create Date: 2026-04-08
-"""
 from typing import Sequence, Union
 
 from alembic import op
@@ -16,7 +11,7 @@ depends_on: Union[str, Sequence[str], None] = None
 
 
 def upgrade() -> None:
-    # --- cars ---
+    # Таблица cars — аукционные машины с IAAI
     op.create_table(
         "cars",
         sa.Column("id", sa.BigInteger, primary_key=True, autoincrement=True),
@@ -53,7 +48,7 @@ def upgrade() -> None:
     op.create_index("ix_cars_origin_url", "cars", ["origin_url"])
     op.create_index("ix_cars_origin_id", "cars", ["origin_id"], unique=True)
 
-    # --- images ---
+    # Таблица images — изображения машин
     op.create_table(
         "images",
         sa.Column("id", sa.BigInteger, primary_key=True, autoincrement=True),
@@ -63,7 +58,7 @@ def upgrade() -> None:
         sa.Column("car_id", sa.Integer, sa.ForeignKey("cars.id", ondelete="CASCADE"), nullable=False),
     )
 
-    # --- sync_runs ---
+    # Таблица sync_runs — логирование синхронизаций
     op.create_table(
         "sync_runs",
         sa.Column("id", sa.BigInteger, primary_key=True, autoincrement=True),

alembic/versions/002_add_indexes.py

@@ -1,9 +1,4 @@
-"""Add performance indexes for growing database
+# Индексы производительности
-
-Revision ID: 002_add_indexes
-Revises: 001_initial
-Create Date: 2026-04-09
-"""
 from typing import Sequence, Union
 
 from alembic import op
@@ -15,25 +10,12 @@ depends_on: Union[str, Sequence[str], None] = None
 
 
 def upgrade() -> None:
-    # cars: ускорение фильтрации по бренду в API и статистике
     op.create_index("ix_cars_brand", "cars", ["brand"])
-
-    # cars: составной индекс бренд+модель для комбинированных фильтров
     op.create_index("ix_cars_brand_model", "cars", ["brand", "model"])
-
-    # cars: ускорение фильтрации по году (year_min/year_max)
     op.create_index("ix_cars_year", "cars", ["year"])
-
-    # cars: ускорение фильтрации по статусу продажи
     op.create_index("ix_cars_is_sold", "cars", ["is_sold"])
-
-    # cars: ускорение сортировки ORDER BY last_seen_at DESC (пагинация)
     op.create_index("ix_cars_last_seen_at", "cars", ["last_seen_at"])
-
-    # images: ускорение JOIN/DELETE по car_id (критично при upsert)
     op.create_index("ix_images_car_id", "images", ["car_id"])
-
-    # sync_runs: ускорение поиска stale runs по статусу
     op.create_index("ix_sync_runs_status", "sync_runs", ["status"])
 
 

alembic/versions/003_add_composite_indexes.py

@@ -0,0 +1,38 @@
+# Индексы для масштабированной БД (20k+ записей)
+from typing import Sequence, Union
+
+from alembic import op
+
+revision: str = "003_add_composite_indexes"
+down_revision: Union[str, None] = "002_add_indexes"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    # Partial index для активных машин IAAI (is_sold = FALSE)
+    # Ускоряет поиск при mark_sold операциях
+    op.execute(
+        "CREATE INDEX IF NOT EXISTS ix_cars_active_iaai "
+        "ON cars (origin_url) "
+        "WHERE is_sold = FALSE AND origin_id LIKE 'iaai:%'"
+    )
+
+    # Composite index для проверки дубликатов изображений
+    op.create_index(
+        "ix_images_car_id_fullres",
+        "images",
+        ["car_id", "fullres_image"],
+    )
+
+    # Index для быстрого поиска existing машин по origin_url
+    op.execute(
+        "CREATE INDEX IF NOT EXISTS ix_cars_origin_url_id "
+        "ON cars (origin_url, origin_id)"
+    )
+
+
+def downgrade() -> None:
+    op.execute("DROP INDEX IF EXISTS ix_cars_origin_url_id")
+    op.drop_index("ix_images_car_id_fullres", table_name="images")
+    op.execute("DROP INDEX IF EXISTS ix_cars_active_iaai")

alembic/versions/004_add_ingestion_tables.py

@@ -0,0 +1,77 @@
+# Добавление таблиц для новой архитектуры ingestion: candidates, raw snapshots, parse results, retry queue
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "004_add_ingestion_tables"
+down_revision: Union[str, None] = "003_add_composite_indexes"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    # Таблица vehicle_candidates
+    op.create_table(
+        "vehicle_candidates",
+        sa.Column("id", sa.BigInteger().with_variant(sa.Integer(), "sqlite"), primary_key=True, autoincrement=True),
+        sa.Column("url", sa.String(), nullable=False, unique=True),
+        sa.Column("discovered_at", sa.DateTime(timezone=True), nullable=False, server_default=sa.func.now()),
+        sa.Column("status", sa.String(20), nullable=False, server_default="pending"),
+        sa.Column("priority", sa.Integer(), nullable=False, server_default="0"),
+        sa.Column("last_attempt_at", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("attempts", sa.Integer(), nullable=False, server_default="0"),
+    )
+
+    op.create_index("ix_vehicle_candidates_url", "vehicle_candidates", ["url"])
+    op.create_index("ix_vehicle_candidates_status_discovered", "vehicle_candidates", ["status", "discovered_at"])
+
+    # Таблица vehicle_raw_snapshots
+    op.create_table(
+        "vehicle_raw_snapshots",
+        sa.Column("id", sa.BigInteger().with_variant(sa.Integer(), "sqlite"), primary_key=True, autoincrement=True),
+        sa.Column("candidate_id", sa.Integer(), sa.ForeignKey("vehicle_candidates.id", ondelete="CASCADE"), nullable=False),
+        sa.Column("captured_at", sa.DateTime(timezone=True), nullable=False, server_default=sa.func.now()),
+        sa.Column("method", sa.String(20), nullable=False),
+        sa.Column("success", sa.Boolean(), nullable=False),
+        sa.Column("raw_data", sa.Text(), nullable=True),
+        sa.Column("error_message", sa.Text(), nullable=True),
+    )
+
+    op.create_index("ix_vehicle_raw_snapshots_candidate_id", "vehicle_raw_snapshots", ["candidate_id"])
+    op.create_index("ix_vehicle_raw_snapshots_captured_at", "vehicle_raw_snapshots", ["captured_at"])
+
+    # Таблица vehicle_parse_results
+    op.create_table(
+        "vehicle_parse_results",
+        sa.Column("id", sa.BigInteger().with_variant(sa.Integer(), "sqlite"), primary_key=True, autoincrement=True),
+        sa.Column("snapshot_id", sa.Integer(), sa.ForeignKey("vehicle_raw_snapshots.id", ondelete="CASCADE"), nullable=False),
+        sa.Column("parsed_at", sa.DateTime(timezone=True), nullable=False, server_default=sa.func.now()),
+        sa.Column("success", sa.Boolean(), nullable=False),
+        sa.Column("parsed_data", sa.Text(), nullable=True),
+        sa.Column("error_message", sa.Text(), nullable=True),
+    )
+
+    op.create_index("ix_vehicle_parse_results_snapshot_id", "vehicle_parse_results", ["snapshot_id"])
+    op.create_index("ix_vehicle_parse_results_parsed_at", "vehicle_parse_results", ["parsed_at"])
+
+    # Таблица vehicle_retry_queue
+    op.create_table(
+        "vehicle_retry_queue",
+        sa.Column("id", sa.BigInteger().with_variant(sa.Integer(), "sqlite"), primary_key=True, autoincrement=True),
+        sa.Column("candidate_id", sa.Integer(), sa.ForeignKey("vehicle_candidates.id", ondelete="CASCADE"), nullable=False),
+        sa.Column("reason", sa.String(50), nullable=False),
+        sa.Column("retry_at", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("attempts", sa.Integer(), nullable=False, server_default="0"),
+        sa.Column("max_attempts", sa.Integer(), nullable=False, server_default="3"),
+    )
+
+    op.create_index("ix_vehicle_retry_queue_candidate_id", "vehicle_retry_queue", ["candidate_id"])
+    op.create_index("ix_vehicle_retry_queue_retry_at", "vehicle_retry_queue", ["retry_at"])
+
+
+def downgrade() -> None:
+    op.drop_table("vehicle_retry_queue")
+    op.drop_table("vehicle_parse_results")
+    op.drop_table("vehicle_raw_snapshots")
+    op.drop_table("vehicle_candidates")

deploy_vps.sh

@@ -0,0 +1,60 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+APP_DIR="/opt/iaai_scraper_project"
+REPO_URL="${1:-}"
+
+if [[ -z "${REPO_URL}" ]]; then
+  echo "Usage: ./deploy_vps.sh <git-repo-url>"
+  exit 1
+fi
+
+export DEBIAN_FRONTEND=noninteractive
+
+apt-get update
+apt-get install -y --no-install-recommends \
+  ca-certificates \
+  curl \
+  git \
+  gnupg \
+  lsb-release
+
+install -m 0755 -d /etc/apt/keyrings
+if [[ ! -f /etc/apt/keyrings/docker.gpg ]]; then
+  curl -fsSL https://download.docker.com/linux/ubuntu/gpg | gpg --dearmor -o /etc/apt/keyrings/docker.gpg
+fi
+
+chmod a+r /etc/apt/keyrings/docker.gpg
+ARCH="$(dpkg --print-architecture)"
+CODENAME="$(. /etc/os-release && echo "$VERSION_CODENAME")"
+echo \
+  "deb [arch=${ARCH} signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/ubuntu ${CODENAME} stable" \
+  > /etc/apt/sources.list.d/docker.list
+
+apt-get update
+apt-get install -y --no-install-recommends \
+  docker-ce \
+  docker-ce-cli \
+  containerd.io \
+  docker-buildx-plugin \
+  docker-compose-plugin
+
+mkdir -p /opt
+
+if [[ -d "${APP_DIR}/.git" ]]; then
+  git -C "${APP_DIR}" fetch --all --prune
+  git -C "${APP_DIR}" reset --hard origin/HEAD
+else
+  rm -rf "${APP_DIR}"
+  git clone "${REPO_URL}" "${APP_DIR}"
+fi
+
+cd "${APP_DIR}"
+
+if [[ ! -f .env ]]; then
+  cp .env.vps.example .env
+fi
+
+docker compose down --remove-orphans || true
+docker compose up -d --build postgres redis migrate api worker beat
+docker compose ps

docker-compose.yml

@@ -1,15 +1,27 @@
 x-app-env: &app-env
-  IAAI_DATABASE_URL: ${IAAI_DATABASE_URL:-postgresql+psycopg2://iaai:iaai@postgres:5432/iaai_scraper}
+  # NB: hardcoded to Postgres — .env may contain sqlite for local CLI, don't let it leak here
+  IAAI_DATABASE_URL: postgresql+psycopg2://iaai:iaai@postgres:5432/iaai_scraper
   IAAI_REDIS_URL: ${IAAI_REDIS_URL:-redis://redis:6379/0}
   CELERY_BROKER_URL: ${CELERY_BROKER_URL:-redis://redis:6379/0}
   CELERY_RESULT_BACKEND: ${CELERY_RESULT_BACKEND:-redis://redis:6379/0}
   IAAI_DATABASE_POOL_RECYCLE_SECONDS: ${IAAI_DATABASE_POOL_RECYCLE_SECONDS:-1800}
   CELERY_TASK_SOFT_TIME_LIMIT: ${CELERY_TASK_SOFT_TIME_LIMIT:-900}
   CELERY_TASK_TIME_LIMIT: ${CELERY_TASK_TIME_LIMIT:-1200}
-  CELERY_BROKER_VISIBILITY_TIMEOUT: ${CELERY_BROKER_VISIBILITY_TIMEOUT:-7200}
+  CELERY_BROKER_VISIBILITY_TIMEOUT: ${CELERY_BROKER_VISIBILITY_TIMEOUT:-2400}
-  CELERY_WORKER_MAX_TASKS_PER_CHILD: ${CELERY_WORKER_MAX_TASKS_PER_CHILD:-20}
+  # 0 => без лимита (в коде интерпретируется как None).
+  # После первичного полного прохода hourly-режим должен успевать за всеми новыми авто.
+  CELERY_BEAT_SYNC_LIMIT: ${CELERY_BEAT_SYNC_LIMIT:-0}
+  CELERY_WORKER_MAX_TASKS_PER_CHILD: ${CELERY_WORKER_MAX_TASKS_PER_CHILD:-3}
+  CELERY_BATCH_SIZE: ${CELERY_BATCH_SIZE:-200}
+  IAAI_PARALLEL_TABS: ${IAAI_PARALLEL_TABS:-40}
+  IAAI_BLOCK_RESOURCES: ${IAAI_BLOCK_RESOURCES:-true}
+  IAAI_LISTING_SEGMENTS: ${IAAI_LISTING_SEGMENTS:-auto}
+  IAAI_MAX_PAGES_PER_RUN: ${IAAI_MAX_PAGES_PER_RUN:-9999}
+  IAAI_MAX_VEHICLES_PER_RUN: ${IAAI_MAX_VEHICLES_PER_RUN:-50000}
+  IAAI_HUMAN_PACE_ENABLED: ${IAAI_HUMAN_PACE_ENABLED:-true}
   IAAI_TOKENS_FILE: ${IAAI_TOKENS_FILE:-/data/tokens.json}
   IAAI_RUNTIME_CONFIG_FILE: ${IAAI_RUNTIME_CONFIG_FILE:-/app/runtime_config.json}
+  IAAI_BROWSER_ENGINE: chromium
   TZ: ${TZ:-UTC}
 
 x-env-file: &env-file
@@ -40,7 +52,7 @@ services:
       POSTGRES_PASSWORD: iaai
       POSTGRES_DB: iaai_scraper
     ports:
-      - "5432:5432"
+      - "127.0.0.1:5432:5432"
     volumes:
       - pgdata:/var/lib/postgresql/data
     healthcheck:
@@ -60,7 +72,7 @@ services:
     container_name: iaai-redis
     restart: unless-stopped
     ports:
-      - "6379:6379"
+      - "127.0.0.1:6379:6379"
     healthcheck:
       test: ["CMD", "redis-cli", "ping"]
       interval: 5s
@@ -94,7 +106,7 @@ services:
     container_name: iaai-api
     restart: unless-stopped
     ports:
-      - "8000:8000"
+      - "127.0.0.1:8000:8000"
     depends_on:
       migrate:
         condition: service_completed_successfully
@@ -121,6 +133,9 @@ services:
     <<: *worker-service
     container_name: iaai-worker
     restart: unless-stopped
+    init: true
+    mem_limit: ${IAAI_WORKER_MEM_LIMIT:-2g}
+    memswap_limit: ${IAAI_WORKER_MEM_LIMIT:-2g}
     depends_on:
       migrate:
         condition: service_completed_successfully
@@ -129,14 +144,15 @@ services:
     stop_grace_period: 60s
     command: >
       celery -A iaai_scraper.worker.celery_app worker
-      --loglevel=info --concurrency=1 --pool=prefork
+      --loglevel=info --concurrency=${CELERY_WORKER_CONCURRENCY:-2} --pool=prefork
-      -Q scraping --max-tasks-per-child=20
+      --pidfile=/tmp/celery-worker.pid
+      -Q scraping --max-tasks-per-child=${CELERY_WORKER_MAX_TASKS_PER_CHILD:-3}
     healthcheck:
-      test: ["CMD", "celery", "-A", "iaai_scraper.worker.celery_app", "inspect", "ping", "-d", "celery@$$HOSTNAME"]
+      test: ["CMD-SHELL", "test -f /tmp/celery-worker.pid && kill -0 $(cat /tmp/celery-worker.pid)"]
       interval: 60s
-      timeout: 20s
+      timeout: 10s
       retries: 3
-      start_period: 40s
+      start_period: 90s
     logging:
       driver: json-file
       options:
@@ -148,6 +164,7 @@ services:
     <<: *worker-service
     container_name: iaai-beat
     restart: unless-stopped
+    init: true
     depends_on:
       migrate:
         condition: service_completed_successfully
@@ -156,12 +173,14 @@ services:
     command: >
       celery -A iaai_scraper.worker.celery_app beat
       --loglevel=info
+      --pidfile=/tmp/celery-beat.pid
+      --schedule=/tmp/celerybeat-schedule
     healthcheck:
-      test: ["CMD", "python", "-c", "import pathlib,sys; p=pathlib.Path('/tmp/celerybeat-schedule'); sys.exit(0 if p.exists() else 1)"]
+      test: ["CMD-SHELL", "test -f /tmp/celery-beat.pid && kill -0 $(cat /tmp/celery-beat.pid)"]
       interval: 60s
-      timeout: 20s
+      timeout: 10s
       retries: 3
-      start_period: 60s
+      start_period: 90s
     logging:
       driver: json-file
       options:

entrypoint.sh

@@ -55,27 +55,6 @@ start_proxy_bridge_if_needed() {
   echo "[entrypoint] Proxy bridge started (PID ${BRIDGE_PID})"
 }
 
-start_xvfb_if_needed() {
-  if ! needs_browser_runtime "$@"; then
-    return 0
-  fi
-
-  # IAAI blocks headless Chromium on Linux; run headed via Xvfb virtual display
-  export DISPLAY=:99
-  export IAAI_HEADLESS=false
-
-  if [ -S /tmp/.X11-unix/X99 ] || [ -f /tmp/.X99-lock ]; then
-    echo "[entrypoint] Reusing existing Xvfb on DISPLAY=${DISPLAY}"
-    return 0
-  fi
-
-  Xvfb :99 -screen 0 1920x1080x24 -nolisten tcp &
-  XVFB_PID=$!
-  sleep 0.5
-  echo "[entrypoint] Xvfb started (PID ${XVFB_PID}, DISPLAY=${DISPLAY})"
-}
-
 start_proxy_bridge_if_needed "$@"
-start_xvfb_if_needed "$@"
 
 exec "$@"

iaai_scraper/api/app.py

@@ -12,8 +12,7 @@ from .routes import cars, health, tasks
 @asynccontextmanager
 async def lifespan(app: FastAPI):
     # Жизненный цикл.
-    persistence: PersistenceService = app.state.persistence
+    # Таблицы создаются через Alembic-миграции (сервис migrate).
-    persistence.create_tables()
     yield
 
 

iaai_scraper/api/deps.py

@@ -1,4 +1,4 @@
-# Dependency helpers для FastAPI-роутов.
+# Вспомогательные зависимости для FastAPI-роутов.
 
 from fastapi import Request
 

iaai_scraper/api/routes/health.py

@@ -1,5 +1,7 @@
 # Роут проверки доступности сервиса и соединения с БД.
 
+import logging
+
 from fastapi import APIRouter, Depends
 from sqlalchemy import text
 
@@ -7,6 +9,7 @@ from ..deps import get_persistence
 from ...storage.db import PersistenceService
 
 router = APIRouter()
+logger = logging.getLogger("iaai_scraper.api.health")
 
 
 @router.get("/health")
@@ -18,7 +21,7 @@ def health_check(persistence: PersistenceService = Depends(get_persistence)):
             session.execute(text("SELECT 1"))
             db_ok = True
     except Exception:
-        pass
+        logger.warning("Health DB check failed", exc_info=True)
 
     return {
         "status": "ok" if db_ok else "degraded",

iaai_scraper/api/routes/tasks.py

@@ -7,6 +7,7 @@ from sqlalchemy import select, func
 from ..deps import get_persistence
 from ...storage.db import PersistenceService
 from ...storage.models import SyncRun
+from ...worker.celery_app import celery_app
 from ...worker.tasks import sync_vehicle_task, sync_listing_task
 
 router = APIRouter()
@@ -64,6 +65,25 @@ def start_sync_listing(
     }
 
 
+@router.get("/tasks/{task_id}")
+def get_task_status(task_id: str):
+    result = celery_app.AsyncResult(task_id)
+
+    payload: dict = {
+        "task_id": task_id,
+        "state": result.state,
+    }
+
+    if result.successful():
+        payload["result"] = result.result
+    elif result.failed():
+        payload["error"] = str(result.result)
+    elif result.info is not None:
+        payload["meta"] = result.info
+
+    return payload
+
+
 @router.get("/sync-runs")
 def list_sync_runs(
     page: int = Query(1, ge=1),

iaai_scraper/browser/factory.py

@@ -4,13 +4,18 @@ import random
 
 from playwright.sync_api import Browser, BrowserContext, Playwright
 
+try:
+    from playwright_stealth import stealth_sync
+except ImportError:
+    stealth_sync = None
+
 from ..core.config import Settings
 
 logger = logging.getLogger("iaai_scraper.browser")
 
 
 def _build_init_script() -> str:
-    # JS-патч признаков автоматизации.
+    # Патч признаков автоматизации.
     hardware_concurrency = random.choice([4, 8, 12, 16])
     device_memory = random.choice([4, 8, 16])
     languages = ["en-US", "en"]
@@ -63,17 +68,64 @@ class BrowserFactory:
     def __init__(self, settings: Settings) -> None:
         self.settings = settings
 
+    def _resolve_engine(self) -> str:
+        # Выбор движка браузера.
+        engine = self.settings.browser_engine.strip().lower()
+        if engine == "auto":
+            # Для IAAI в headless-режиме Chromium со stealth-скриптами
+            # значительно стабильнее Firefox по anti-bot.
+            return "chromium"
+        if engine in ("firefox", "chromium"):
+            return engine
+        logger.warning("Unknown IAAI_BROWSER_ENGINE=%r, falling back to auto", engine)
+        return "chromium"
+
     def create_browser(self, playwright: Playwright) -> Browser:
-        launch_kwargs: dict = {
+        engine = self._resolve_engine()
-            "headless": self.settings.headless,
-            "args": [
-                "--disable-blink-features=AutomationControlled",
-                "--no-default-browser-check",
-                "--disable-dev-shm-usage",
-                "--disable-features=IsolateOrigins,site-per-process",
-            ],
-        }
         proxy_dict = self.settings.proxy.to_playwright_dict()
+        logger.info("Resolved browser engine: requested=%s resolved=%s", self.settings.browser_engine, engine)
+
+        if engine == "firefox":
+            launch_kwargs: dict = {"headless": self.settings.headless}
+            if proxy_dict:
+                launch_kwargs["proxy"] = proxy_dict
+                logger.info("Using proxy: %s", self.settings.proxy.server)
+            # Параметры Firefox для headless-режима.
+            launch_kwargs["firefox_user_prefs"] = {
+                "dom.webdriver.enabled": False,
+                "useAutomationExtension": False,
+                # Базовые оптимизации для VPS.
+                "media.autoplay.default": 5,
+                "media.volume_scale": "0.0",
+                "media.audio.playback.standalone": False,
+                "dom.ipc.processCount": 1,
+                "dom.ipc.plugins.enabled": False,
+                "browser.cache.disk.enable": False,
+                "browser.cache.memory.enable": True,
+                "browser.cache.memory.max_entry_size": 8192,
+                "network.prefetch-next": False,
+                "network.dns.disablePrefetch": True,
+                "permissions.default.image": 2,
+                "javascript.options.mem.gc_incremental_mark_slice_ms": 20,
+            }
+            logger.info("Launching Firefox (headless=%s)", self.settings.headless)
+            return playwright.firefox.launch(**launch_kwargs)
+
+        # Путь запуска Chromium.
+        args = [
+            "--disable-blink-features=AutomationControlled",
+            "--no-default-browser-check",
+            "--disable-dev-shm-usage",
+            "--disable-features=IsolateOrigins,site-per-process",
+        ]
+        if self.settings.headless:
+            args.append("--headless=new")
+            pw_headless = False
+            logger.info("Using Chromium new-headless mode (--headless=new)")
+        else:
+            pw_headless = False
+
+        launch_kwargs = {"headless": pw_headless, "args": args}
         if proxy_dict:
             launch_kwargs["proxy"] = proxy_dict
             logger.info("Using proxy: %s", self.settings.proxy.server)
@@ -84,34 +136,80 @@ class BrowserFactory:
             logger.warning("Chrome channel launch failed, falling back to Chromium")
             return playwright.chromium.launch(**launch_kwargs)
 
-    def create_context(self, browser: Browser, storage_state: str | None = None) -> BrowserContext:
+    def create_context(self, browser: Browser) -> BrowserContext:
         viewport = random.choice(self.settings.fingerprint.viewport_presets)
         timezone_id = random.choice(self.settings.fingerprint.timezone_candidates)
         color_scheme = random.choice(["light", "dark"])
 
-        context = browser.new_context(
+        is_firefox = browser.browser_type.name == "firefox"
-            storage_state=storage_state or None,
+
-            user_agent=self.settings.fingerprint.user_agent,
+        ctx_kwargs: dict = {
-            viewport=viewport,
+            "viewport": viewport,
-            screen=viewport,
+            "screen": viewport,
-            device_scale_factor=random.choice([1, 1.25]),
+            "locale": self.settings.fingerprint.locale,
-            is_mobile=False,
+            "timezone_id": timezone_id,
-            has_touch=False,
+            "color_scheme": color_scheme,
-            locale=self.settings.fingerprint.locale,
+            "java_script_enabled": True,
-            timezone_id=timezone_id,
+            "ignore_https_errors": False,
-            color_scheme=color_scheme,
+        }
-            java_script_enabled=True,
+
-            ignore_https_errors=False,
+        if is_firefox:
-            extra_http_headers={
+            # Заголовки и user-agent для Firefox.
+            ctx_kwargs["user_agent"] = (
+                "Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:128.0) "
+                "Gecko/20100101 Firefox/128.0"
+            )
+            ctx_kwargs["extra_http_headers"] = {
+                "Accept-Language": "en-US,en;q=0.5",
+                "DNT": "1",
+                "Upgrade-Insecure-Requests": "1",
+            }
+        else:
+            ctx_kwargs["user_agent"] = self.settings.fingerprint.user_agent
+            ctx_kwargs["device_scale_factor"] = random.choice([1, 1.25])
+            ctx_kwargs["is_mobile"] = False
+            ctx_kwargs["has_touch"] = False
+            ctx_kwargs["extra_http_headers"] = {
                 "Accept-Language": "en-US,en;q=0.9",
                 "DNT": "1",
                 "Upgrade-Insecure-Requests": "1",
                 "Sec-CH-UA": self.settings.fingerprint.sec_ch_ua,
                 "Sec-CH-UA-Mobile": "?0",
                 "Sec-CH-UA-Platform": '"Windows"',
-            },
+            }
-        )
+
+        context = browser.new_context(**ctx_kwargs)
         context.set_default_timeout(self.settings.default_timeout_ms)
         context.set_default_navigation_timeout(self.settings.default_timeout_ms)
-        context.add_init_script(_build_init_script())
+
+        if not is_firefox:
+            # Патчи маскировки для Chromium.
+            context.add_init_script(_build_init_script())
+            if stealth_sync:
+                context.on("page", lambda page: stealth_sync(page))
+                logger.debug("playwright-stealth attached to context")
+
         return context
+
+    @staticmethod
+    def enable_resource_blocking(page) -> None:
+        # Блокируем тяжёлые ресурсы для ускорения загрузки страниц.
+        BLOCKED_TYPES = {"image", "stylesheet", "font", "media"}
+        BLOCKED_URL_PATTERNS = (
+            "google-analytics", "googletagmanager", "facebook.net",
+            "doubleclick.net", "hotjar", "newrelic", ".woff", ".woff2",
+            "analytics", "tracking", "adservice",
+        )
+
+        def _handle_route(route):
+            req = route.request
+            if req.resource_type in BLOCKED_TYPES:
+                route.abort()
+                return
+            url = req.url.lower()
+            if any(pat in url for pat in BLOCKED_URL_PATTERNS):
+                route.abort()
+                return
+            route.continue_()
+
+        page.route("**/*", _handle_route)

iaai_scraper/browser/listing.py

@@ -1,17 +1,29 @@
 import logging
 import re
+import time
 from dataclasses import asdict, dataclass, field
 from typing import Any
-from urllib.parse import urljoin
+from urllib.parse import parse_qsl, urlencode, urljoin, urlparse, urlunparse
 
 from playwright.sync_api import Page
 
 from .pace import HumanPacer
 from ..core.config import Settings
-from ..core.utils import first_non_empty
 
 logger = logging.getLogger("iaai_scraper.listing")
-VEHICLE_HREF_RE = re.compile(r"/VehicleDetail/\d+(?:~[A-Z]{2})?", re.IGNORECASE)
+VEHICLE_HREF_RE = re.compile(r"/VehicleDetail/(\d+)(?:~[A-Z]{2})?", re.IGNORECASE)
+VEHICLE_LINK_SELECTOR = "a[href*='/VehicleDetail/'], a[href*='/vehicledetail/'], a[href*='VehicleDetail'], a[href*='vehicledetail']"
+COOKIE_ACCEPT_SELECTORS: tuple[str, ...] = (
+	"button:has-text('Accept All')",
+	"button:has-text('Accept all')",
+	"button:has-text('I Agree')",
+	"button:has-text('Agree')",
+	"button:has-text('Only necessary')",
+	"button:has-text('Только необходимые')",
+	"button:has-text('Принять все')",
+	"[id*='accept']",
+	"[class*='accept']",
+)
 
 
 @dataclass(slots=True)
@@ -31,94 +43,711 @@ class ListingPageResult:
 
 
 class ListingCollector:
+	_DEEP_PAGINATION_DIRECT_ONLY_FROM_PAGE = 40
+	_NEXT_PAGE_SELECTORS: tuple[str, ...] = (
+		"a[aria-label*='Next']",
+		"button[aria-label*='Next']",
+		"a[aria-label*='next']",
+		"button[aria-label*='next']",
+		"a[title*='Next']",
+		"button[title*='Next']",
+		"a[title*='next']",
+		"button[title*='next']",
+		"a[rel='next']",
+		"link[rel='next']",
+		"a.pagination-next",
+		"button.pagination-next",
+		"a.next",
+		"button.next",
+		"a:has-text('Next')",
+		"button:has-text('Next')",
+		"a:has-text('NEXT')",
+		"button:has-text('NEXT')",
+		"a:has-text('›')",
+		"button:has-text('›')",
+		"a:has-text('»')",
+		"button:has-text('»')",
+		"a:has(img[src*='icon-arrow-right'])",
+		"button:has(img[src*='icon-arrow-right'])",
+		"a:has(img[src*='arrow-right'])",
+		"button:has(img[src*='arrow-right'])",
+	)
+
 	def __init__(self, settings: Settings, pacer: HumanPacer) -> None:
 		self.settings = settings
 		self.pacer = pacer
 
-	def open_cars_listing(self, page: Page) -> None:
+	@staticmethod
-		logger.info("Opening cars listing page: %s", self.settings.listing.cars_url)
+	def _get_current_page_number(page: Page) -> int | None:
-		page.goto(self.settings.listing.cars_url, wait_until="domcontentloaded")
 		try:
-			page.wait_for_load_state("networkidle", timeout=15000)
+			value = page.evaluate(
+				"""
+				() => {
+					const controls = Array.from(document.querySelectorAll('a,button,[role="button"],span,div'));
+					const current = controls.find((el) => {
+						const text = (el.textContent || '').trim();
+						const cls = (el.getAttribute('class') || '').toLowerCase();
+						const ariaCurrent = (el.getAttribute('aria-current') || '').toLowerCase();
+						return /^\d+$/.test(text) && (ariaCurrent === 'page' || cls.includes('active') || cls.includes('current') || cls.includes('selected'));
+					});
+					if (current) {
+						return parseInt((current.textContent || '').trim(), 10);
+					}
+					const match = (document.body?.innerText || '').match(/\b(\d+)\s+of\s+\d+\+?/i);
+					return match ? parseInt(match[1], 10) : null;
+				}
+				"""
+			)
+			return int(value) if value is not None else None
 		except Exception:
-			logger.debug("networkidle timeout on listing page, continuing with current state")
+			return None
+
+	@staticmethod
+	def _get_vehicle_link_fingerprint(page: Page, limit: int = 5) -> tuple[str, ...]:
 		try:
-			page.wait_for_selector("a[href*='/VehicleDetail/']", timeout=20000)
+			values = page.evaluate(
+				"""
+				(limit) => {
+					return Array.from(document.querySelectorAll("a[href*='/VehicleDetail/'], a[href*='/vehicledetail/'], a[href*='VehicleDetail'], a[href*='vehicledetail']"))
+						.map((el) => (el.getAttribute('href') || '').trim())
+						.filter(Boolean)
+						.slice(0, limit);
+				}
+				""",
+				limit,
+			)
+			if not isinstance(values, list):
+				return tuple()
+			return tuple(str(value) for value in values if value)
 		except Exception:
-			logger.warning("Vehicle links did not appear within timeout; page may not have rendered fully")
+			return tuple()
+
+	@staticmethod
+	def _wait_for_navigation_result(
+		page: Page,
+		old_first_href: str,
+		expected_page_number: int | None,
+		*,
+		old_fingerprint: tuple[str, ...] = (),
+	) -> bool:
+		if old_first_href:
+			try:
+				page.wait_for_function(
+					f"""() => {{
+						const a = document.querySelector(\"a[href*='/VehicleDetail/'], a[href*='/vehicledetail/'], a[href*='VehicleDetail'], a[href*='vehicledetail']\");
+						return a && a.getAttribute('href') !== '{old_first_href}';
+					}}""",
+					timeout=12000,
+				)
+				return True
+			except Exception:
+				pass
+
+		new_fingerprint = ListingCollector._get_vehicle_link_fingerprint(page)
+		if expected_page_number is not None:
+			current_page = ListingCollector._get_current_page_number(page)
+			if current_page == expected_page_number and (
+				not old_fingerprint or (new_fingerprint and new_fingerprint != old_fingerprint)
+			):
+				return True
+
+		try:
+			page.wait_for_selector(VEHICLE_LINK_SELECTOR, timeout=3000)
+		except Exception:
+			pass
+
+		if not new_fingerprint:
+			new_fingerprint = ListingCollector._get_vehicle_link_fingerprint(page)
+		current_page = ListingCollector._get_current_page_number(page)
+		if new_fingerprint and old_fingerprint and new_fingerprint != old_fingerprint:
+			return current_page is None or expected_page_number is None or current_page == expected_page_number
+		if expected_page_number is not None and current_page == expected_page_number:
+			return not old_fingerprint
+
+		return not old_first_href and (not old_fingerprint or bool(new_fingerprint))
+
+	def _extract_next_page_href(self, page: Page, expected_page_number: int | None = None) -> str | None:
+		try:
+			href = page.evaluate(
+				"""
+				(expectedPageNumber) => {
+					const visible = (el) => !!(el && (el.offsetWidth || el.offsetHeight || el.getClientRects().length));
+					const disabled = (el) => {
+						if (!el) return true;
+						const cls = (el.getAttribute('class') || '').toLowerCase();
+						const ariaDisabled = (el.getAttribute('aria-disabled') || '').toLowerCase();
+						return el.hasAttribute('disabled') || ariaDisabled === 'true' || cls.includes('disabled');
+					};
+
+					const controls = Array.from(document.querySelectorAll('a,button,[role="button"]'))
+						.filter((el) => visible(el) && !disabled(el));
+
+					const cleanHref = (el) => {
+						const href = (el?.getAttribute('href') || '').trim();
+						if (!href || href.startsWith('javascript:') || href.startsWith('#')) {
+							return '';
+						}
+						return href;
+					};
+
+					if (expectedPageNumber !== null && expectedPageNumber !== undefined) {
+						const numeric = controls.find((el) => {
+							const text = (el.textContent || '').trim();
+							return /^\d+$/.test(text) && parseInt(text, 10) === expectedPageNumber;
+						});
+						const numericHref = cleanHref(numeric);
+						if (numericHref) {
+							return numericHref;
+						}
+					}
+
+					const explicitNext = controls.find((el) => {
+						const text = (el.textContent || '').trim().toLowerCase();
+						const aria = (el.getAttribute('aria-label') || '').trim().toLowerCase();
+						const title = (el.getAttribute('title') || '').trim().toLowerCase();
+						const rel = (el.getAttribute('rel') || '').trim().toLowerCase();
+						const cls = (el.getAttribute('class') || '').trim().toLowerCase();
+						const hasRightArrowIcon = !!el.querySelector('img[src*="icon-arrow-right"], img[src*="arrow-right"]');
+						return rel === 'next' || aria.includes('next') || title.includes('next') || cls.includes('next') || hasRightArrowIcon || ['next', '›', '»', '>'].includes(text);
+					});
+					return cleanHref(explicitNext);
+				}
+				""",
+				expected_page_number,
+			)
+			if not href:
+				return None
+			return urljoin(page.url or self.settings.home_url, str(href))
+		except Exception:
+			return None
+
+	@staticmethod
+	def _build_listing_page_url(url: str, page_number: int) -> str:
+		parsed = urlparse(url)
+		query_items = [
+			(key, value)
+			for key, value in parse_qsl(parsed.query, keep_blank_values=True)
+			if key.lower() not in {"page", "pagenumber", "currentpage"}
+		]
+		query_items.append(("page", str(page_number)))
+		return urlunparse(parsed._replace(query=urlencode(query_items)))
+
+	def open_cars_listing(self, page: Page, *, url_override: str | None = None) -> None:
+		url = url_override or self.settings.listing.cars_url
+		logger.info("Opening cars listing page: %s", url)
+		last_err = None
+		for attempt in range(3):
+			try:
+				page.goto(url, wait_until="commit", timeout=60_000)
+				last_err = None
+				break
+			except Exception as e:
+				last_err = e
+				logger.warning("goto listing attempt %d failed: %s", attempt + 1, e)
+				# Небольшой backoff при ошибках открытия листинга.
+				time.sleep(5 * (attempt + 1))
+		if last_err:
+			logger.warning("All goto attempts failed, trying JS navigation")
+			try:
+				page.evaluate(f"window.location.href = '{url}'")
+			except Exception:
+				pass
+		# Быстрая проверка готовности страницы.
+		try:
+			page.wait_for_load_state("domcontentloaded", timeout=12_000)
+		except Exception:
+			pass
+		self._accept_cookie_banner(page)
+		self._wait_for_listing_content(page)
+		logger.info("Listing page URL: %s", page.url)
 		self.pacer.after_listing_open()
 
-	def apply_filters(self, page: Page, make: str | None = None, model: str | None = None) -> dict[str, str | None]:
+	def _accept_cookie_banner(self, page: Page) -> None:
-		applied = {"make": None, "model": None}
+		for selector in COOKIE_ACCEPT_SELECTORS:
+			locator = page.locator(selector).first
+			try:
+				if locator.count() == 0:
+					continue
+				if not locator.is_visible(timeout=500):
+					continue
+				locator.click(timeout=2_000)
+				logger.info("Accepted cookie banner using selector: %s", selector)
+				try:
+					page.wait_for_load_state("domcontentloaded", timeout=3_000)
+				except Exception:
+					pass
+				return
+			except Exception:
+				continue
+
+	def _wait_for_listing_content(self, page: Page) -> None:
+		try:
+			page.wait_for_selector(VEHICLE_LINK_SELECTOR, timeout=12_000)
+			return
+		except Exception:
+			pass
+
+		# Fallback: React/SSR разметка может появиться не сразу, даже если <a> ещё нет в DOM.
+		try:
+			page.wait_for_function(
+				"""() => {
+					const html = document.documentElement?.innerHTML || '';
+					const text = document.body?.innerText || '';
+					return html.includes('/VehicleDetail/') || /\\b\d+\s+VEHICLES\b/i.test(text);
+				}""",
+				timeout=12_000,
+			)
+		except Exception:
+			# Короткая пауза вместо длинного sleep.
+			time.sleep(1.0)
+
+	def _get_listing_html(self, page: Page, page_number: int) -> str:
+		try:
+			return page.locator("html").inner_html(timeout=5_000)
+		except Exception as exc:
+			logger.warning("listing html read failed on page %d: %s", page_number, exc)
+			return ""
+
+	def _has_next_page_from_html(self, html: str, current_page_number: int | None = None) -> bool:
+		if not html:
+			return False
+		normalized = html.replace("\\/", "/")
+		if re.search(
+			r"rel\s*=\s*['\"]next['\"]|aria-label\s*=\s*['\"][^'\"]*next|title\s*=\s*['\"][^'\"]*next|class\s*=\s*['\"][^'\"]*next|icon-arrow-right|arrow-right|>\s*next\s*<|>\s*[›»>]\s*<",
+			normalized,
+			re.IGNORECASE,
+		):
+			return True
+		if current_page_number is not None:
+			next_page = current_page_number + 1
+			if re.search(rf">\s*{next_page}\s*<", normalized, re.IGNORECASE):
+				return True
+			if re.search(rf"page={next_page}(?:\D|$)", normalized, re.IGNORECASE):
+				return True
+		return False
+
+	def apply_filters(
+		self,
+		page: Page,
+		make: str | None = None,
+		model: str | None = None,
+		year_min: int | None = None,
+		year_max: int | None = None,
+	) -> dict[str, str | int | None]:
+		applied: dict[str, str | int | None] = {"make": None, "model": None, "year_min": None, "year_max": None}
 		if make and self._try_fill_filter_input(page, ["input[placeholder*='Make']", "input[aria-label*='Make']"], make):
 			applied["make"] = make
 			self.pacer.after_filter_action()
 		if model and self._try_fill_filter_input(page, ["input[placeholder*='Model']", "input[aria-label*='Model']"], model):
 			applied["model"] = model
 			self.pacer.after_filter_action()
+		if year_min is not None or year_max is not None:
+			if self._apply_year_range(page, year_min, year_max):
+				applied["year_min"] = year_min
+				applied["year_max"] = year_max
+				self.pacer.after_filter_action()
 		return applied
 
+	def _apply_year_range(self, page: Page, year_min: int | None, year_max: int | None) -> bool:
+		"""Заполняет поля фильтра Year и нажимает Apply Year."""
+		if year_min is None and year_max is None:
+			return False
+		try:
+			success = page.evaluate(
+				"""([yearMin, yearMax]) => {
+					const inputs = Array.from(document.querySelectorAll('input'));
+					const yearInputs = inputs.filter(inp => {
+						const v = parseInt(inp.value, 10);
+						return !isNaN(v) && v >= 1900 && v <= 2100;
+					});
+					if (yearInputs.length < 2) return false;
+					yearInputs.sort((a, b) => parseInt(a.value) - parseInt(b.value));
+					const setVal = (el, val) => {
+						const setter = Object.getOwnPropertyDescriptor(
+							HTMLInputElement.prototype, 'value'
+						).set;
+						setter.call(el, String(val));
+						el.dispatchEvent(new Event('input', {bubbles: true}));
+						el.dispatchEvent(new Event('change', {bubbles: true}));
+					};
+					if (yearMin !== null) setVal(yearInputs[0], yearMin);
+					if (yearMax !== null) setVal(yearInputs[yearInputs.length - 1], yearMax);
+					const container = yearInputs[0].closest(
+						'[class*="filter"], [class*="year"], section, fieldset'
+					) || yearInputs[0].parentElement.parentElement;
+					if (container) {
+						const btn = Array.from(container.querySelectorAll(
+							'button, a, [role="button"], span[class*="apply"]'
+						)).find(el => /apply|\u043f\u0440\u0438\u043c\u0435\u043d/i.test(el.textContent));
+						if (btn) { btn.click(); return true; }
+					}
+					yearInputs[yearInputs.length - 1].dispatchEvent(
+						new KeyboardEvent('keydown', {
+							key: 'Enter', code: 'Enter', keyCode: 13, bubbles: true
+						})
+					);
+					return true;
+				}""",
+				[year_min, year_max],
+			)
+			if success:
+				try:
+					page.wait_for_load_state("domcontentloaded", timeout=15_000)
+				except Exception:
+					pass
+				self._wait_for_listing_content(page)
+				logger.info("Applied year range filter: %s — %s", year_min, year_max)
+				return True
+		except Exception as exc:
+			logger.warning("Failed to apply year range filter: %s", exc)
+		return False
+
 	def collect_current_page(self, page: Page, page_number: int = 1) -> ListingPageResult:
-		anchors = page.locator("a[href*='/VehicleDetail/']")
+		# Для IAAI HTML/hydration-извлечение стабильнее, чем прямой DOM eval.
-		total = min(anchors.count(), self.settings.listing.page_link_limit)
+		self._accept_cookie_banner(page)
+		self._wait_for_listing_content(page)
 		links: list[ListingVehicleLink] = []
 		seen: set[str] = set()
-		for idx in range(total):
+		html = self._get_listing_html(page, page_number)
-			anchor = anchors.nth(idx)
+		for absolute, lot_number in self._extract_vehicle_links_from_html(html):
-			href = anchor.get_attribute("href") or ""
+			if absolute in seen:
-			match = VEHICLE_HREF_RE.search(href)
-			if not match:
 				continue
+			seen.add(absolute)
+			links.append(ListingVehicleLink(href=absolute, title="", lot_number=lot_number))
+			if len(links) >= self.settings.listing.max_vehicles_per_run:
+				break
+
+		if links:
+			logger.info(
+				"Page %d: recovered %d vehicle links from HTML fallback",
+				page_number,
+				len(links),
+			)
+
+		next_page_detected = self._has_next_page_from_html(html, current_page_number=page_number)
+		if not next_page_detected and not html:
+			next_page_detected = self._has_next_page(page)
+		return ListingPageResult(source_url=page.url, page_number=page_number, vehicle_links=links, pagination_available=next_page_detected, next_page_detected=next_page_detected)
+
+	def _extract_vehicle_links_from_html(self, html: str) -> list[tuple[str, str]]:
+		if not html:
+			return []
+
+		# Частый формат в JSON внутри HTML: "\/VehicleDetail\/12345678~US"
+		normalized = html.replace("\\/", "/")
+		found: list[tuple[str, str]] = []
+		seen: set[str] = set()
+
+		for match in VEHICLE_HREF_RE.finditer(normalized):
+			lot_number = match.group(1)
 			absolute = urljoin(self.settings.home_url, match.group(0))
 			if absolute in seen:
 				continue
 			seen.add(absolute)
-			title = first_non_empty([anchor.get_attribute("title"), anchor.text_content(), ""]) or ""
+			found.append((absolute, lot_number))
-			links.append(ListingVehicleLink(href=absolute, title=str(title).strip()))
+			if len(found) >= self.settings.listing.page_link_limit:
-			if len(links) >= self.settings.listing.max_vehicles_per_run:
 				break
-		next_page_detected = self._has_next_page(page)
-		return ListingPageResult(source_url=page.url, page_number=page_number, vehicle_links=links, pagination_available=next_page_detected, next_page_detected=next_page_detected)
 
-	def go_to_next_page(self, page: Page) -> bool:
+		return found
-		selectors = ["a[aria-label*='Next']", "button[aria-label*='Next']", "a.pagination-next", "button.pagination-next", "a:has-text('Next')", "button:has-text('Next')"]
+
-		for selector in selectors:
+	def go_to_next_page(self, page: Page, expected_page_number: int | None = None) -> bool:
+		# Запоминаем первую ссылку текущей страницы для определения смены контента.
+		old_first_href = ""
+		old_fingerprint: tuple[str, ...] = tuple()
+		try:
+			first_link = page.locator(VEHICLE_LINK_SELECTOR).first
+			if first_link.count() > 0:
+				old_first_href = first_link.get_attribute("href") or ""
+		except Exception:
+			pass
+		old_fingerprint = self._get_vehicle_link_fingerprint(page)
+
+		if expected_page_number is not None and expected_page_number > 1:
+			direct_page_url = self._build_listing_page_url(
+				page.url or self.settings.listing.cars_url,
+				expected_page_number,
+			)
+			try:
+				logger.debug("Navigating directly to listing page %d via URL: %s", expected_page_number, direct_page_url)
+				page.goto(direct_page_url, wait_until="domcontentloaded", timeout=15_000)
+				if self._wait_for_navigation_result(
+					page,
+					old_first_href,
+					expected_page_number,
+					old_fingerprint=old_fingerprint,
+				):
+					self.pacer.after_page_change()
+					return True
+			except Exception as exc:
+				logger.debug("Direct page-number navigation failed for page %d via %s: %s", expected_page_number, direct_page_url, exc)
+
+		next_href = self._extract_next_page_href(page, expected_page_number)
+		if next_href:
+			try:
+				logger.debug("Navigating directly to next listing page: %s", next_href)
+				page.goto(next_href, wait_until="domcontentloaded", timeout=15_000)
+				if self._wait_for_navigation_result(
+					page,
+					old_first_href,
+					expected_page_number,
+					old_fingerprint=old_fingerprint,
+				):
+					self.pacer.after_page_change()
+					return True
+			except Exception as exc:
+				logger.debug("Direct next-page navigation failed for %s: %s", next_href, exc)
+
+		if (
+			expected_page_number is not None
+			and expected_page_number >= self._DEEP_PAGINATION_DIRECT_ONLY_FROM_PAGE
+		):
+			logger.warning(
+				"Deep pagination direct navigation failed for page %d; skipping flaky UI pagination fallbacks",
+				expected_page_number,
+			)
+			return False
+
+		for selector in self._NEXT_PAGE_SELECTORS:
 			locator = page.locator(selector).first
 			if locator.count() == 0:
 				continue
-			disabled = (locator.get_attribute("disabled") or "").lower()
+			try:
-			aria_disabled = (locator.get_attribute("aria-disabled") or "").lower()
+				disabled = (locator.get_attribute("disabled", timeout=1500) or "").lower()
-			classes = (locator.get_attribute("class") or "").lower()
+				aria_disabled = (locator.get_attribute("aria-disabled", timeout=1500) or "").lower()
+				classes = (locator.get_attribute("class", timeout=1500) or "").lower()
+			except Exception:
+				continue
 			if disabled or aria_disabled == "true" or "disabled" in classes:
 				continue
-			self.pacer.move_mouse_to(page, locator)
-			locator.click()
 			try:
-				page.wait_for_load_state("networkidle", timeout=15000)
+				self.pacer.move_mouse_to(page, locator)
+				locator.click(timeout=8000)
 			except Exception:
-				pass
+				continue
-			self.pacer.after_page_change()
+
-			return True
+			if self._wait_for_navigation_result(
+				page,
+				old_first_href,
+				expected_page_number,
+				old_fingerprint=old_fingerprint,
+			):
+				self.pacer.after_page_change()
+				return True
+
+		# Fallback для IAAI: пагинация часто рендерится как набор номеров страниц
+		# + стрелка с иконкой, без явного текста Next.
+		try:
+			clicked = bool(page.evaluate(
+				"""
+				() => {
+					const visible = (el) => !!(el && (el.offsetWidth || el.offsetHeight || el.getClientRects().length));
+					const disabled = (el) => {
+						if (!el) return true;
+						const cls = (el.getAttribute('class') || '').toLowerCase();
+						const ariaDisabled = (el.getAttribute('aria-disabled') || '').toLowerCase();
+						return el.hasAttribute('disabled') || ariaDisabled === 'true' || cls.includes('disabled');
+					};
+
+					const controls = Array.from(document.querySelectorAll('a,button,[role="button"]'))
+						.filter((el) => visible(el) && !disabled(el));
+
+					const current = controls.find((el) => {
+						const text = (el.textContent || '').trim();
+						const cls = (el.getAttribute('class') || '').toLowerCase();
+						const ariaCurrent = (el.getAttribute('aria-current') || '').toLowerCase();
+						return /^\d+$/.test(text) && (ariaCurrent === 'page' || cls.includes('active') || cls.includes('current') || cls.includes('selected'));
+					});
+
+					if (current) {
+						const currentPage = parseInt((current.textContent || '').trim(), 10);
+						const nextNumber = controls.find((el) => {
+							const text = (el.textContent || '').trim();
+							return /^\d+$/.test(text) && parseInt(text, 10) === currentPage + 1;
+						});
+						if (nextNumber) {
+							nextNumber.click();
+							return true;
+						}
+					}
+
+					const iconNext = controls.find((el) => {
+						const text = (el.textContent || '').trim().toLowerCase();
+						const aria = (el.getAttribute('aria-label') || '').trim().toLowerCase();
+						const title = (el.getAttribute('title') || '').trim().toLowerCase();
+						const rel = (el.getAttribute('rel') || '').trim().toLowerCase();
+						const cls = (el.getAttribute('class') || '').trim().toLowerCase();
+						const hasRightArrowIcon = !!el.querySelector('img[src*="icon-arrow-right"], img[src*="arrow-right"]');
+						return hasRightArrowIcon || rel === 'next' || aria.includes('next') || title.includes('next') || cls.includes('next') || ['next', '›', '»', '>'].includes(text);
+					});
+
+					if (iconNext) {
+						iconNext.click();
+						return true;
+					}
+
+					return false;
+				}
+				"""
+			))
+			if clicked:
+				if self._wait_for_navigation_result(
+					page,
+					old_first_href,
+					expected_page_number,
+					old_fingerprint=old_fingerprint,
+				):
+					self.pacer.after_page_change()
+					return True
+		except Exception as exc:
+			logger.debug("Numeric/icon pagination fallback failed: %s", exc)
+
+		# JS fallback: ищем любой видимый pagination-control «next» по атрибутам/тексту.
+		try:
+			clicked = bool(page.evaluate(
+				"""
+				() => {
+					const candidates = Array.from(document.querySelectorAll('a,button,[role="button"]'));
+					for (const el of candidates) {
+						const text = (el.textContent || '').trim().toLowerCase();
+						const aria = (el.getAttribute('aria-label') || '').trim().toLowerCase();
+						const title = (el.getAttribute('title') || '').trim().toLowerCase();
+						const rel = (el.getAttribute('rel') || '').trim().toLowerCase();
+						const cls = (el.getAttribute('class') || '').trim().toLowerCase();
+						const disabled = el.hasAttribute('disabled') || el.getAttribute('aria-disabled') === 'true' || cls.includes('disabled');
+						const visible = !!(el.offsetWidth || el.offsetHeight || el.getClientRects().length);
+						const looksNext = rel === 'next' || aria.includes('next') || title.includes('next') || cls.includes('next') || ['next', '›', '»', '>'].includes(text);
+						if (!disabled && visible && looksNext) {
+							el.click();
+							return true;
+						}
+					}
+					return false;
+				}
+				"""
+			))
+			if clicked:
+				if self._wait_for_navigation_result(
+					page,
+					old_first_href,
+					expected_page_number,
+					old_fingerprint=old_fingerprint,
+				):
+					self.pacer.after_page_change()
+					return True
+		except Exception as exc:
+			logger.debug("JS next-page fallback failed: %s", exc)
+
+		logger.warning("Could not navigate to next page from %s", page.url)
 		return False
 
-	def collect_listing_links(self, page: Page, *, make: str | None = None, model: str | None = None) -> dict[str, Any]:
+	def collect_listing_links(
+		self,
+		page: Page,
+		*,
+		make: str | None = None,
+		model: str | None = None,
+		known_origin_ids: set[str] | None = None,
+		max_duration_seconds: float | None = None,
+	) -> dict[str, Any]:
 		self.open_cars_listing(page)
 		applied_filters = self.apply_filters(page, make=make, model=model)
+		started_at = time.perf_counter()
+		truncated_by_time_budget = False
 		pages: list[dict[str, object]] = []
 		all_links: list[str] = []
+		early_stopped = False
+		threshold = self.settings.listing.early_stop_threshold
+
 		for page_number in range(1, max(1, self.settings.listing.max_pages_per_run) + 1):
+			if max_duration_seconds is not None and max_duration_seconds > 0:
+				elapsed = time.perf_counter() - started_at
+				if elapsed >= max_duration_seconds:
+					truncated_by_time_budget = True
+					logger.warning(
+						"Listing collection stopped by time budget: page=%d elapsed=%.1fs budget=%.1fs",
+						page_number,
+						elapsed,
+						max_duration_seconds,
+					)
+					break
 			page_result = self.collect_current_page(page, page_number=page_number)
-			pages.append({"page_number": page_result.page_number, "source_url": page_result.source_url, "links_found": len(page_result.vehicle_links), "vehicle_links": [asdict(item) for item in page_result.vehicle_links], "next_page_detected": page_result.next_page_detected})
+			pages.append({
+				"page_number": page_result.page_number,
+				"source_url": page_result.source_url,
+				"links_found": len(page_result.vehicle_links),
+				"vehicle_links": [asdict(item) for item in page_result.vehicle_links],
+				"next_page_detected": page_result.next_page_detected,
+			})
 			for item in page_result.vehicle_links:
 				if item.href not in all_links:
 					all_links.append(item.href)
 				if len(all_links) >= self.settings.listing.max_vehicles_per_run:
 					break
-			if len(all_links) >= self.settings.listing.max_vehicles_per_run or self.settings.listing.collect_current_page_only or not self.settings.listing.include_pagination or not page_result.next_page_detected:
+
+			# Ранний останов: если на этой странице много известных И нет новых — дальше нет смысла.
+			# Важно: если есть хоть одна новая машина — продолжаем листать (новые могут быть на любой странице).
+			if (
+				known_origin_ids is not None
+				and threshold > 0.0
+				and page_result.vehicle_links
+			):
+				page_known = sum(
+					1 for item in page_result.vehicle_links
+					if item.lot_number and f"iaai:{item.lot_number}" in known_origin_ids
+				)
+				page_new = len(page_result.vehicle_links) - page_known
+				ratio = page_known / len(page_result.vehicle_links)
+				# Останавливаемся только если нет новых И порог превышен
+				if ratio >= threshold and page_new == 0:
+					logger.info(
+						"Early stop on page %d: %.0f%% known (%d/%d), 0 new >= threshold %.0f%%",
+						page_number, ratio * 100, page_known,
+						len(page_result.vehicle_links), threshold * 100,
+					)
+					early_stopped = True
+					break
+				elif page_new > 0 and ratio >= threshold:
+					logger.info(
+						"Page %d: %.0f%% known but %d new found — продолжаем",
+						page_number, ratio * 100, page_new,
+					)
+
+			if (
+				len(all_links) >= self.settings.listing.max_vehicles_per_run
+				or self.settings.listing.collect_current_page_only
+				or not self.settings.listing.include_pagination
+			):
 				break
-			if not self.go_to_next_page(page):
+
+			blind_page_probe = not page_result.next_page_detected
+			if not self.go_to_next_page(page, expected_page_number=page_number + 1):
+				if blind_page_probe:
+					logger.info(
+						"Stopping pagination on page %d: direct page probe for %d failed and no next-page control was detected",
+						page_number,
+						page_number + 1,
+					)
+					break
 				break
-		return {"listing_url": self.settings.listing.cars_url, "applied_filters": applied_filters, "pages_collected": len(pages), "vehicles_collected": len(all_links), "vehicle_urls": all_links, "pages": pages, "strategy": {"sequential": True, "collect_current_page_only": self.settings.listing.collect_current_page_only, "include_pagination": self.settings.listing.include_pagination, "max_pages_per_run": self.settings.listing.max_pages_per_run, "max_vehicles_per_run": self.settings.listing.max_vehicles_per_run}}
+
+		return {
+			"listing_url": self.settings.listing.cars_url,
+			"applied_filters": applied_filters,
+			"pages_collected": len(pages),
+			"vehicles_collected": len(all_links),
+			"vehicle_urls": all_links,
+			"early_stopped": early_stopped,
+			"truncated_by_time_budget": truncated_by_time_budget,
+			"pages": pages,
+			"strategy": {
+				"sequential": True,
+				"collect_current_page_only": self.settings.listing.collect_current_page_only,
+				"include_pagination": self.settings.listing.include_pagination,
+				"max_pages_per_run": self.settings.listing.max_pages_per_run,
+				"max_vehicles_per_run": self.settings.listing.max_vehicles_per_run,
+				"early_stop_threshold": threshold,
+			},
+		}
 
 	@staticmethod
 	def _try_fill_filter_input(page: Page, selectors: list[str], value: str) -> bool:
@@ -131,7 +760,11 @@ class ListingCollector:
 				locator.fill(value)
 				page.keyboard.press("Enter")
 				try:
-					page.wait_for_load_state("networkidle", timeout=15000)
+					page.wait_for_load_state("domcontentloaded", timeout=15000)
+				except Exception:
+					pass
+				try:
+					page.wait_for_selector(VEHICLE_LINK_SELECTOR, timeout=3000)
 				except Exception:
 					pass
 				return True
@@ -141,7 +774,53 @@ class ListingCollector:
 
 	@staticmethod
 	def _has_next_page(page: Page) -> bool:
-		for selector in ["a[aria-label*='Next']", "button[aria-label*='Next']", "a.pagination-next", "button.pagination-next", "a:has-text('Next')", "button:has-text('Next')"]:
+		for selector in ListingCollector._NEXT_PAGE_SELECTORS:
 			if page.locator(selector).count() > 0:
 				return True
+		try:
+			return bool(page.evaluate(
+				"""
+				() => {
+					const visible = (el) => !!(el && (el.offsetWidth || el.offsetHeight || el.getClientRects().length));
+					const controls = Array.from(document.querySelectorAll('a,button,[role="button"]')).filter((el) => {
+						const cls = (el.getAttribute('class') || '').trim().toLowerCase();
+						const disabled = el.hasAttribute('disabled') || el.getAttribute('aria-disabled') === 'true' || cls.includes('disabled');
+						return !disabled && visible(el);
+					});
+
+					const hasExplicitNext = controls.some((el) => {
+						const text = (el.textContent || '').trim().toLowerCase();
+						const aria = (el.getAttribute('aria-label') || '').trim().toLowerCase();
+						const title = (el.getAttribute('title') || '').trim().toLowerCase();
+						const rel = (el.getAttribute('rel') || '').trim().toLowerCase();
+						const cls = (el.getAttribute('class') || '').trim().toLowerCase();
+						const hasRightArrowIcon = !!el.querySelector('img[src*="icon-arrow-right"], img[src*="arrow-right"]');
+						return rel === 'next' || aria.includes('next') || title.includes('next') || cls.includes('next') || hasRightArrowIcon || ['next', '›', '»', '>'].includes(text);
+					});
+
+					if (hasExplicitNext) {
+						return true;
+					}
+
+					const current = controls.find((el) => {
+						const text = (el.textContent || '').trim();
+						const cls = (el.getAttribute('class') || '').toLowerCase();
+						const ariaCurrent = (el.getAttribute('aria-current') || '').toLowerCase();
+						return /^\d+$/.test(text) && (ariaCurrent === 'page' || cls.includes('active') || cls.includes('current') || cls.includes('selected'));
+					});
+
+					if (!current) {
+						return false;
+					}
+
+					const currentPage = parseInt((current.textContent || '').trim(), 10);
+					return controls.some((el) => {
+						const text = (el.textContent || '').trim();
+						return /^\d+$/.test(text) && parseInt(text, 10) === currentPage + 1;
+					});
+				}
+				"""
+			))
+		except Exception:
+			return False
 		return False

iaai_scraper/browser/network.py

@@ -13,7 +13,7 @@ logger = logging.getLogger("iaai_scraper.network")
 
 @dataclass
 class NetworkCapture:
-    # XHR/fetch перехватчик.
+    # Перехватчик сетевых запросов.
 
     settings: Settings
     requests: list[dict[str, Any]] = field(default_factory=list)
@@ -21,13 +21,21 @@ class NetworkCapture:
     _seen_req: set[str] = field(default_factory=set)
     _seen_resp: set[str] = field(default_factory=set)
     _origin: str | None = None
+    _page: Any = field(default=None)
 
     def attach(self, page: Page, origin_url: str | None = None) -> None:
+        # Снимаем старые подписки перед повторным attach.
+        try:
+            page.remove_listener("request", self._on_request)
+            page.remove_listener("response", self._on_response)
+        except Exception:
+            pass
         # Подписка на сетевые события
         try:
             self._origin = urlparse(origin_url or page.url).netloc.lower() or None
         except Exception:
             self._origin = None
+        self._page = page
         page.on("request", self._on_request)
         page.on("response", self._on_response)
 

iaai_scraper/browser/pace.py

@@ -7,7 +7,7 @@ from ..core.config import Settings
 
 
 class HumanPacer:
-    # Random паузы между действиями.
+    # Случайные паузы между действиями.
 
     def __init__(self, settings: Settings) -> None:
         self.settings = settings

iaai_scraper/cli.py

@@ -38,6 +38,17 @@ def build_parser() -> argparse.ArgumentParser:
     sync_listing_parser.add_argument("--only-new", choices=["true", "false"], default=None)
     sync_listing_parser.add_argument("--output", default=str(default_output_dir / "iaai_sync_listing.json"))
 
+    # Новые команды для ingestion pipeline
+    subparsers.add_parser("discover-vehicles", help="Discover new vehicle URLs from sitemap")
+
+    fetch_parser = subparsers.add_parser("fetch-pending", help="Fetch raw data for pending candidates")
+    fetch_parser.add_argument("--limit", type=int, default=10, help="Max candidates to process")
+
+    enrich_parser = subparsers.add_parser("enrich-snapshots", help="Parse and enrich raw snapshots")
+    enrich_parser.add_argument("--limit", type=int, default=10, help="Max snapshots to process")
+
+    subparsers.add_parser("run-pipeline", help="Run full ingestion pipeline (discover -> fetch -> enrich)")
+
     return parser
 
 
@@ -64,6 +75,39 @@ def main() -> None:
             data = scraper.scrape_vehicle_detail(args.vehicle_url)
         elif args.command == "sync-vehicle":
             data = scraper.sync_vehicle(args.vehicle_url, lane=args.lane)
+        elif args.command == "sync-listing":
+            only_new = None if args.only_new is None else args.only_new == "true"
+            data = scraper.sync_listing(
+                make=args.make,
+                model=args.model,
+                lane=args.lane,
+                limit=args.limit,
+                only_new=only_new,
+            )
+        elif args.command == "discover-vehicles":
+            from .discovery_service import DiscoveryService
+            discovery = DiscoveryService()
+            count = discovery.discover_new_vehicles()
+            print(f"Discovered {count} new vehicle candidates")
+            return
+        elif args.command == "fetch-pending":
+            from .fetch_service import FetchService
+            fetch = FetchService()
+            count = fetch.process_pending_candidates(limit=args.limit)
+            print(f"Successfully fetched {count} candidates")
+            return
+        elif args.command == "enrich-snapshots":
+            from .enrichment_service import EnrichmentService
+            enrichment = EnrichmentService()
+            count = enrichment.process_unparsed_snapshots(limit=args.limit)
+            print(f"Successfully enriched {count} snapshots")
+            return
+        elif args.command == "run-pipeline":
+            from .scheduler_service import SchedulerService
+            scheduler = SchedulerService()
+            scheduler.run_full_pipeline()
+            print("Pipeline completed")
+            return
         else:
             only_new = None if args.only_new is None else args.only_new == "true"
             data = scraper.sync_listing(

iaai_scraper/core/config.py

@@ -1,3 +1,4 @@
+import json
 import os
 from dataclasses import dataclass, field
 from pathlib import Path
@@ -84,18 +85,18 @@ class CaptureConfig:
 @dataclass(slots=True)
 class HumanPaceConfig:
     enabled: bool = _env_bool("IAAI_HUMAN_PACE_ENABLED", True)
-    after_listing_open_min_s: float = _env_float("IAAI_AFTER_LISTING_OPEN_MIN_S", 2.5)
+    after_listing_open_min_s: float = _env_float("IAAI_AFTER_LISTING_OPEN_MIN_S", 0.5)
-    after_listing_open_max_s: float = _env_float("IAAI_AFTER_LISTING_OPEN_MAX_S", 4.5)
+    after_listing_open_max_s: float = _env_float("IAAI_AFTER_LISTING_OPEN_MAX_S", 1.2)
-    after_filter_action_min_s: float = _env_float("IAAI_AFTER_FILTER_ACTION_MIN_S", 2.0)
+    after_filter_action_min_s: float = _env_float("IAAI_AFTER_FILTER_ACTION_MIN_S", 0.5)
-    after_filter_action_max_s: float = _env_float("IAAI_AFTER_FILTER_ACTION_MAX_S", 4.0)
+    after_filter_action_max_s: float = _env_float("IAAI_AFTER_FILTER_ACTION_MAX_S", 1.2)
-    before_vehicle_open_min_s: float = _env_float("IAAI_BEFORE_VEHICLE_OPEN_MIN_S", 0.5)
+    before_vehicle_open_min_s: float = _env_float("IAAI_BEFORE_VEHICLE_OPEN_MIN_S", 0.1)
-    before_vehicle_open_max_s: float = _env_float("IAAI_BEFORE_VEHICLE_OPEN_MAX_S", 1.5)
+    before_vehicle_open_max_s: float = _env_float("IAAI_BEFORE_VEHICLE_OPEN_MAX_S", 0.3)
-    after_vehicle_open_min_s: float = _env_float("IAAI_AFTER_VEHICLE_OPEN_MIN_S", 0.3)
+    after_vehicle_open_min_s: float = _env_float("IAAI_AFTER_VEHICLE_OPEN_MIN_S", 0.05)
-    after_vehicle_open_max_s: float = _env_float("IAAI_AFTER_VEHICLE_OPEN_MAX_S", 1.0)
+    after_vehicle_open_max_s: float = _env_float("IAAI_AFTER_VEHICLE_OPEN_MAX_S", 0.15)
-    between_vehicles_min_s: float = _env_float("IAAI_BETWEEN_VEHICLES_MIN_S", 0.5)
+    between_vehicles_min_s: float = _env_float("IAAI_BETWEEN_VEHICLES_MIN_S", 0.05)
-    between_vehicles_max_s: float = _env_float("IAAI_BETWEEN_VEHICLES_MAX_S", 1.5)
+    between_vehicles_max_s: float = _env_float("IAAI_BETWEEN_VEHICLES_MAX_S", 0.15)
-    after_page_change_min_s: float = _env_float("IAAI_AFTER_PAGE_CHANGE_MIN_S", 3.0)
+    after_page_change_min_s: float = _env_float("IAAI_AFTER_PAGE_CHANGE_MIN_S", 0.8)
-    after_page_change_max_s: float = _env_float("IAAI_AFTER_PAGE_CHANGE_MAX_S", 6.0)
+    after_page_change_max_s: float = _env_float("IAAI_AFTER_PAGE_CHANGE_MAX_S", 1.8)
 
 
 #  Конфиг сбора листинга (URL, лимиты страниц и машин) 
@@ -103,11 +104,104 @@ class HumanPaceConfig:
 @dataclass(slots=True)
 class ListingConfig:
     cars_url: str = _env_str("IAAI_CARS_LISTING_URL", "https://www.iaai.com/Vehiclelisting/Cars")
-    max_pages_per_run: int = _env_int("IAAI_MAX_PAGES_PER_RUN", 5)
+    max_pages_per_run: int = _env_int("IAAI_MAX_PAGES_PER_RUN", 9999)
-    max_vehicles_per_run: int = _env_int("IAAI_MAX_VEHICLES_PER_RUN", 100)
+    max_vehicles_per_run: int = _env_int("IAAI_MAX_VEHICLES_PER_RUN", 50000)
-    page_link_limit: int = _env_int("IAAI_PAGE_LINK_LIMIT", 200)
+    resume_max_nav_pages: int = _env_int("IAAI_LISTING_RESUME_MAX_NAV_PAGES", 90)
+    page_link_limit: int = _env_int("IAAI_PAGE_LINK_LIMIT", 500)
     include_pagination: bool = _env_bool("IAAI_INCLUDE_PAGINATION", True)
     collect_current_page_only: bool = _env_bool("IAAI_COLLECT_CURRENT_PAGE_ONLY", False)
+    # Порог раннего останова: если доля уже известных машин на странице >= этого значения,
+    # прекращаем листать — все новые машины уже найдены. 0 = отключено.
+    early_stop_threshold: float = _env_float("IAAI_EARLY_STOP_THRESHOLD", 0.8)
+    # Сегментация листинга по брендам для обхода лимита пагинации IAAI (~22 600 машин).
+    # JSON-массив объектов: [{"make":"TOYOTA"},{"make":"FORD"},...] или "auto" для авто-списка.
+    # Пустая строка = без сегментации (backward compatible).
+    listing_segments_json: str = _env_str("IAAI_LISTING_SEGMENTS", "")
+
+
+@dataclass(slots=True)
+class DiscoveryConfig:
+    mode: str = _env_str("IAAI_DISCOVERY_MODE", "sitemap")
+    hourly_mode: str = _env_str("IAAI_HOURLY_MODE", "rolling_refresh")
+    hourly_refresh_batch_size: int = _env_int("IAAI_HOURLY_REFRESH_BATCH_SIZE", 3000)
+    sitemap_timeout_seconds: int = _env_int("IAAI_SITEMAP_TIMEOUT_SECONDS", 30)
+    sitemap_retry_attempts: int = _env_int("IAAI_SITEMAP_RETRY_ATTEMPTS", 4)
+    sitemap_retry_backoff_seconds: float = _env_float("IAAI_SITEMAP_RETRY_BACKOFF_SECONDS", 1.25)
+    sitemap_direct_probe_limit: int = _env_int("IAAI_SITEMAP_DIRECT_PROBE_LIMIT", 12)
+    sitemap_direct_probe_stop_after_misses: int = _env_int(
+        "IAAI_SITEMAP_DIRECT_PROBE_STOP_AFTER_MISSES",
+        3,
+    )
+    sitemap_use_curl_cffi: bool = _env_bool("IAAI_SITEMAP_USE_CURL_CFFI", True)
+
+
+# Список брендов IAAI для автоматической сегментации.
+# Покрывает >99% автомобилей на сайте. Порядок: от крупных к мелким.
+IAAI_DEFAULT_MAKES: tuple[str, ...] = (
+    "TOYOTA", "FORD", "CHEVROLET", "HONDA", "NISSAN", "HYUNDAI",
+    "KIA", "DODGE", "JEEP", "BMW", "MERCEDES-BENZ", "SUBARU",
+    "VOLKSWAGEN", "GMC", "MAZDA", "LEXUS", "CHRYSLER", "AUDI",
+    "RAM", "BUICK", "CADILLAC", "ACURA", "INFINITI", "LINCOLN",
+    "MITSUBISHI", "VOLVO", "JAGUAR", "LAND ROVER", "PORSCHE",
+    "MINI", "TESLA", "GENESIS", "FIAT", "ALFA ROMEO", "MASERATI",
+    "SCION", "PONTIAC", "SATURN", "MERCURY", "SAAB", "SUZUKI",
+    "OLDSMOBILE", "ISUZU", "HUMMER", "PLYMOUTH", "SMART",
+    "RIVIAN", "LUCID", "POLESTAR", "FERRARI", "LAMBORGHINI",
+    "BENTLEY", "ROLLS-ROYCE", "ASTON MARTIN", "MCLAREN", "LOTUS",
+    "MAYBACH", "FISKER", "GEO", "DAEWOO", "EAGLE",
+)
+
+# Пагинационный потолок IAAI: ~226 страниц × 100 = 22 600 результатов.
+IAAI_PAGINATION_CEILING = 22_600
+
+# Бренды, потенциально превышающие потолок пагинации — разбиваем по годам.
+_LARGE_MAKES: frozenset[str] = frozenset({
+    "TOYOTA", "FORD", "CHEVROLET", "HONDA", "NISSAN", "HYUNDAI",
+    "KIA", "DODGE", "JEEP",
+})
+_YEAR_SPLITS: tuple[tuple[int, int], ...] = (
+    (1900, 2012),
+    (2013, 2019),
+    (2020, 2027),
+)
+
+
+def parse_listing_segments(raw: str) -> list[dict[str, str | int | None]]:
+    """Парсит IAAI_LISTING_SEGMENTS в список сегментов.
+
+    Каждый сегмент — dict с ключами: make (str), year_min/year_max (int|None).
+    Специальное значение ``"auto"`` генерирует сегменты из IAAI_DEFAULT_MAKES.
+    Крупные бренды автоматически разбиваются по диапазонам годов.
+    """
+    raw = raw.strip()
+    if not raw:
+        return []
+    if raw.lower() == "auto":
+        segments: list[dict[str, str | int | None]] = []
+        for m in IAAI_DEFAULT_MAKES:
+            if m in _LARGE_MAKES:
+                for yr_min, yr_max in _YEAR_SPLITS:
+                    segments.append({"make": m, "year_min": yr_min, "year_max": yr_max})
+            else:
+                segments.append({"make": m, "year_min": None, "year_max": None})
+        return segments
+    try:
+        data = json.loads(raw)
+    except (json.JSONDecodeError, ValueError):
+        return []
+    if not isinstance(data, list):
+        return []
+    segments = []
+    for item in data:
+        if isinstance(item, str):
+            segments.append({"make": item.upper(), "year_min": None, "year_max": None})
+        elif isinstance(item, dict):
+            segments.append({
+                "make": str(item.get("make") or "").upper() or None,
+                "year_min": int(item["year_min"]) if item.get("year_min") is not None else None,
+                "year_max": int(item["year_max"]) if item.get("year_max") is not None else None,
+            })
+    return segments
 
 
 # - Конфиг PostgreSQL (URL, пул соединений, pool_recycle) 
@@ -127,6 +221,9 @@ class DatabaseConfig:
 @dataclass(slots=True)
 class RedisConfig:
     url: str = _env_str("IAAI_REDIS_URL", "redis://localhost:6379/0")
+    socket_timeout_seconds: float = _env_float("IAAI_REDIS_SOCKET_TIMEOUT_SECONDS", 10.0)
+    socket_connect_timeout_seconds: float = _env_float("IAAI_REDIS_SOCKET_CONNECT_TIMEOUT_SECONDS", 5.0)
+    health_check_interval_seconds: int = _env_int("IAAI_REDIS_HEALTH_CHECK_INTERVAL_SECONDS", 30)
 
 
 # --- Конфиг Celery (лимиты задач, concurrency, beat-расписание) ---
@@ -135,13 +232,18 @@ class RedisConfig:
 class CeleryConfig:
     broker_url: str = _env_str("CELERY_BROKER_URL", "")
     result_backend: str = _env_str("CELERY_RESULT_BACKEND", "")
-    task_soft_time_limit: int = _env_int("CELERY_TASK_SOFT_TIME_LIMIT", 600)
+    task_soft_time_limit: int = _env_int("CELERY_TASK_SOFT_TIME_LIMIT", 3300)
-    task_time_limit: int = _env_int("CELERY_TASK_TIME_LIMIT", 900)
+    task_time_limit: int = _env_int("CELERY_TASK_TIME_LIMIT", 3600)
-    worker_concurrency: int = _env_int("CELERY_WORKER_CONCURRENCY", 1)
+    task_stall_timeout_seconds: int = _env_int("CELERY_TASK_STALL_TIMEOUT_SECONDS", 600)
-    worker_max_tasks_per_child: int = _env_int("CELERY_WORKER_MAX_TASKS_PER_CHILD", 20)
+    worker_concurrency: int = _env_int("CELERY_WORKER_CONCURRENCY", 4)
+    worker_max_tasks_per_child: int = _env_int("CELERY_WORKER_MAX_TASKS_PER_CHILD", 5)
     broker_visibility_timeout: int = _env_int("CELERY_BROKER_VISIBILITY_TIMEOUT", 7200)
     beat_sync_interval_minutes: int = _env_int("CELERY_BEAT_SYNC_INTERVAL_MINUTES", 60)
-    beat_sync_limit: int = _env_int("CELERY_BEAT_SYNC_LIMIT", 26)
+    beat_sync_limit: int | None = _env_int("CELERY_BEAT_SYNC_LIMIT", 0) or None
+    batch_size: int = _env_int("CELERY_BATCH_SIZE", 50)
+    parallel_tabs: int = _env_int("IAAI_PARALLEL_TABS", 8)
+    parallel_segments: bool = _env_bool("IAAI_PARALLEL_SEGMENTS", False)
+    block_resources: bool = _env_bool("IAAI_BLOCK_RESOURCES", True)
 
 
 # --- Конфиг прокси (server, username, password) ---
@@ -167,22 +269,26 @@ class ProxyConfig:
         return result
 
 
-# --- Главный объект настроек: собирает все блоки конфигурации ---
+#  Главный объект настроек: собирает все блоки конфигурации 
 
 @dataclass(slots=True)
 class Settings:
     home_url: str = "https://www.iaai.com/"
     default_timeout_ms: int = _env_int("IAAI_TIMEOUT_MS", 45000)
-    network_settle_ms: int = _env_int("IAAI_NETWORK_SETTLE_MS", 800)
+    network_settle_ms: int = _env_int("IAAI_NETWORK_SETTLE_MS", 400)
+    fast_path_timeout_ms: int = _env_int("IAAI_FAST_PATH_TIMEOUT_MS", 5000)
+    fast_path_max_attempts: int = _env_int("IAAI_FAST_PATH_MAX_ATTEMPTS", 1)
+    fallback_navigation_timeout_ms: int = _env_int("IAAI_FALLBACK_NAV_TIMEOUT_MS", 15000)
     max_retries: int = _env_int("IAAI_MAX_RETRIES", 3)
     retry_delay_seconds: float = _env_float("IAAI_RETRY_DELAY_SECONDS", 2.5)
     retry_backoff_multiplier: float = _env_float("IAAI_RETRY_BACKOFF_MULTIPLIER", 2.0)
     retry_jitter_seconds: float = _env_float("IAAI_RETRY_JITTER_SECONDS", 0.25)
     headless: bool = _env_bool("IAAI_HEADLESS", True)
+    browser_engine: str = _env_str("IAAI_BROWSER_ENGINE", "auto")
     log_level: str = _env_str("IAAI_LOG_LEVEL", "INFO")
     log_file: str | None = _env_optional_str("IAAI_LOG_FILE")
     enable_trace_id_logs: bool = _env_bool("IAAI_ENABLE_TRACE_ID_LOGS", True)
-    sync_only_new: bool = _env_bool("IAAI_SYNC_ONLY_NEW", True)
+    sync_only_new: bool = _env_bool("IAAI_SYNC_ONLY_NEW", False)
     raw_output_json: str | None = _env_optional_str("IAAI_RAW_OUTPUT_JSON")
     tokens_file: str | None = _env_path_str("IAAI_TOKENS_FILE")
     runtime_config_file: str | None = _env_path_str("IAAI_RUNTIME_CONFIG_FILE")
@@ -191,10 +297,19 @@ class Settings:
     capture: CaptureConfig = field(default_factory=CaptureConfig)
     pace: HumanPaceConfig = field(default_factory=HumanPaceConfig)
     listing: ListingConfig = field(default_factory=ListingConfig)
+    discovery: DiscoveryConfig = field(default_factory=DiscoveryConfig)
     database: DatabaseConfig = field(default_factory=DatabaseConfig)
     redis: RedisConfig = field(default_factory=RedisConfig)
     celery: CeleryConfig = field(default_factory=CeleryConfig)
     proxy: ProxyConfig = field(default_factory=ProxyConfig)
 
+    @property
+    def parallel_tabs(self) -> int:
+        return self.celery.parallel_tabs
+
+    @property
+    def block_resources(self) -> bool:
+        return self.celery.block_resources
+
 # Глобальный синглтон — используется по умолчанию во всех модулях.
 settings = Settings()

iaai_scraper/core/exceptions.py

@@ -0,0 +1,14 @@
+class ScraperError(Exception):
+    """Базовое исключение скрапера."""
+
+
+class AntiBotDetectedError(ScraperError):
+    """Вызывается, когда сайт блокирует автоматизацию."""
+
+
+class SiteStructureChangedError(ScraperError):
+    """Вызывается, когда структура страницы изменилась и данных не хватает."""
+
+
+class ListingResumeError(ScraperError):
+    """Вызывается, когда resume по checkpoint больше недостижим."""

iaai_scraper/core/logs.py

@@ -2,7 +2,7 @@ import logging
 import sys
 from contextvars import ContextVar
 
-# ContextVar хранит trace_id текущего потока/корутины.
+# Храним trace_id текущего потока/корутины.
 TRACE_ID: ContextVar[str] = ContextVar("trace_id", default="-")
 
 
@@ -18,15 +18,22 @@ def set_trace_id(trace_id: str) -> None:
 
 
 def setup_logging(level: str = "INFO", log_file: str | None = None) -> None:
-    handlers: list[logging.Handler] = [logging.StreamHandler(sys.stdout)]
+    # stderr — Docker и Celery prefork корректно его подхватывают.
+    handlers: list[logging.Handler] = [logging.StreamHandler(sys.stderr)]
     if log_file:
         handlers.append(logging.FileHandler(log_file, encoding="utf-8"))
     trace_filter = TraceIdFilter()
     for handler in handlers:
         handler.addFilter(trace_filter)
-    logging.basicConfig(
+        handler.setLevel(getattr(logging, level.upper(), logging.INFO))
-        level=getattr(logging, level.upper(), logging.INFO),
+    root = logging.getLogger()
-        format="%(asctime)s | %(levelname)s | %(name)s | trace=%(trace_id)s | %(message)s",
+    root.setLevel(getattr(logging, level.upper(), logging.INFO))
-        handlers=handlers,
+    # Убираем старые хендлеры, чтобы не дублировать после fork.
-        force=True,
+    root.handlers.clear()
+    for handler in handlers:
+        root.addHandler(handler)
+    fmt = logging.Formatter(
+        "%(asctime)s | %(levelname)s | %(name)s | trace=%(trace_id)s | %(message)s"
     )
+    for handler in root.handlers:
+        handler.setFormatter(fmt)

iaai_scraper/core/retry.py

@@ -7,6 +7,8 @@ from typing import Any
 
 from playwright.sync_api import Error, TimeoutError as PlaywrightTimeoutError
 
+from .exceptions import AntiBotDetectedError
+
 logger = logging.getLogger("iaai_scraper.retry")
 
 # Типы исключений, при которых retry имеет смысл.
@@ -16,6 +18,7 @@ RETRYABLE_EXCEPTIONS = (
     ConnectionError,
     OSError,
     TimeoutError,
+    AntiBotDetectedError,
 )
 
 

iaai_scraper/core/utils.py

@@ -1,7 +1,7 @@
 import json
 import re
 from pathlib import Path
-from typing import Any, Iterable
+from typing import Any, Callable, Iterable
 
 
 def save_to_json(data: Any, filename: str | Path) -> None:
@@ -17,7 +17,7 @@ def first_non_empty(values: Iterable[Any]) -> Any | None:
     return None
 
 
-# Regex для VIN, lot, price.
+# Регулярные выражения для VIN, lot и price.
 VIN_RE = re.compile(r"\b([A-HJ-NPR-Z0-9]{17})\b", re.IGNORECASE)
 LOT_RE = re.compile(r"\b(\d{7,10})\b")
 PRICE_RE = re.compile(r"\$\s?([\d,]+(?:\.\d{1,2})?)")
@@ -37,3 +37,36 @@ def deep_find_key(obj, target_keys: set[str], max_depth: int = 64, _depth: int =
         for item in obj:
             found.extend(deep_find_key(item, target_keys, max_depth=max_depth, _depth=_depth + 1))
     return found
+
+
+def deep_find_all_keys(
+    payloads: list,
+    field_map: dict[str, set[str]],
+    max_depth: int = 64,
+) -> dict[str, list]:
+    """Извлекает все нужные поля за один проход по JSON."""
+    # Готовим обратную карту: нормализованный ключ -> имя поля.
+    reverse: dict[str, str] = {}
+    for field_name, keys in field_map.items():
+        for k in keys:
+            reverse[k.lower()] = field_name
+
+    result: dict[str, list] = {f: [] for f in field_map}
+
+    def _recurse(obj: Any, depth: int) -> None:
+        if depth >= max_depth:
+            return
+        if isinstance(obj, dict):
+            for k, v in obj.items():
+                field = reverse.get(k.lower())
+                if field is not None:
+                    result[field].append(v)
+                _recurse(v, depth + 1)
+        elif isinstance(obj, list):
+            for item in obj:
+                _recurse(item, depth + 1)
+
+    for payload in payloads:
+        _recurse(payload, 0)
+
+    return result

iaai_scraper/discovery/__init__.py

@@ -0,0 +1,17 @@
+from .sitemap import (
+    DEFAULT_SITEMAP_INDEX_URL,
+    SitemapDiscoveryError,
+    SitemapBlockedError,
+    SitemapDiscoveryResult,
+    discover_vehicle_urls_from_sitemap,
+    discover_vehicle_urls_from_sitemap_with_stats,
+)
+
+__all__ = [
+    "DEFAULT_SITEMAP_INDEX_URL",
+    "SitemapBlockedError",
+    "SitemapDiscoveryError",
+    "SitemapDiscoveryResult",
+    "discover_vehicle_urls_from_sitemap",
+    "discover_vehicle_urls_from_sitemap_with_stats",
+]

iaai_scraper/discovery/sitemap.py

@@ -0,0 +1,441 @@
+from __future__ import annotations
+
+import gzip
+import io
+import logging
+import random
+import re
+import time
+import xml.etree.ElementTree as ET
+from dataclasses import dataclass
+from typing import Iterable
+from urllib.error import HTTPError, URLError
+from urllib.parse import urljoin, urlsplit, urlunsplit
+from urllib.request import Request, urlopen
+
+from ..core.config import Settings
+
+try:
+    from curl_cffi import requests as curl_requests
+except ImportError:  # pragma: no cover - optional runtime dependency guard
+    curl_requests = None
+
+logger = logging.getLogger("iaai_scraper.discovery.sitemap")
+
+DEFAULT_SITEMAP_INDEX_URL = "https://www.iaai.com/Xj9rDOVMEi0hc38S/sitemap_index.xml"
+_LOC_TAG_RE = re.compile(rb"<loc>\s*(.*?)\s*</loc>", re.IGNORECASE | re.DOTALL)
+_XML_PREFIX_RE = re.compile(rb"^\s*(<\?xml\b.*?\?>)?\s*<", re.IGNORECASE | re.DOTALL)
+_BLOCK_MARKERS = (
+    b"pardon our interruption",
+    b"incapsula",
+    b"please stand by",
+    b"_incapsula_resource",
+    b"as you were browsing something about your browser",
+)
+_HTML_MARKERS = (b"<html", b"<!doctype html", b"<script", b"document.getelementsbyclassname")
+_CURL_IMPERSONATE_CHOICES = ("chrome136", "chrome133", "chrome131", "safari184")
+
+
+class SitemapDiscoveryError(RuntimeError):
+    pass
+
+
+class SitemapBlockedError(SitemapDiscoveryError):
+    pass
+
+
+@dataclass(slots=True)
+class SitemapFetchResult:
+    url: str
+    payload: bytes
+    source: str
+    blocked: bool = False
+    status_code: int | None = None
+    content_type: str | None = None
+
+
+@dataclass(slots=True)
+class SitemapDiscoveryStats:
+    transport: str
+    fetched_sitemaps: int = 0
+    blocked_sitemaps: int = 0
+    malformed_sitemaps: int = 0
+    direct_probe_hits: int = 0
+    direct_probe_misses: int = 0
+
+
+@dataclass(slots=True)
+class SitemapDiscoveryResult:
+    vehicle_urls: list[str]
+    stats: SitemapDiscoveryStats
+
+
+_DEFAULT_HEADERS = {
+    "Accept": "application/xml,text/xml,application/xhtml+xml,text/html;q=0.9,*/*;q=0.8",
+    "Accept-Encoding": "gzip, deflate, br",
+    "Accept-Language": "en-US,en;q=0.9",
+    "Cache-Control": "no-cache",
+    "Pragma": "no-cache",
+    "Upgrade-Insecure-Requests": "1",
+}
+
+
+class _SitemapDownloader:
+    def __init__(self, settings: Settings) -> None:
+        self.settings = settings
+        self.discovery = settings.discovery
+        self.proxy_url = settings.proxy.server
+        self.proxy_auth = None
+        if settings.proxy.username:
+            password = settings.proxy.password or ""
+            self.proxy_auth = (settings.proxy.username, password)
+        self._transport_name = self._resolve_transport_name()
+
+    @property
+    def transport_name(self) -> str:
+        return self._transport_name
+
+    def _resolve_transport_name(self) -> str:
+        if self.discovery.sitemap_use_curl_cffi and curl_requests is not None:
+            return "curl_cffi"
+        return "urllib"
+
+    def fetch(self, url: str) -> SitemapFetchResult:
+        attempts = max(1, int(self.discovery.sitemap_retry_attempts))
+        last_exc: Exception | None = None
+        for attempt in range(1, attempts + 1):
+            try:
+                result = self._fetch_once(url)
+                if result.blocked:
+                    raise SitemapBlockedError(f"Anti-bot page returned for {url}")
+                return result
+            except SitemapBlockedError as exc:
+                last_exc = exc
+                logger.warning(
+                    "Sitemap fetch blocked (attempt %d/%d, transport=%s): %s",
+                    attempt,
+                    attempts,
+                    self.transport_name,
+                    url,
+                )
+            except Exception as exc:
+                last_exc = exc
+                logger.warning(
+                    "Sitemap fetch failed (attempt %d/%d, transport=%s): %s -> %s",
+                    attempt,
+                    attempts,
+                    self.transport_name,
+                    url,
+                    exc,
+                )
+            if attempt >= attempts:
+                break
+            time.sleep(self._backoff_delay(attempt))
+        if last_exc is None:
+            raise SitemapDiscoveryError(f"Failed to fetch sitemap: {url}")
+        if isinstance(last_exc, SitemapDiscoveryError):
+            raise last_exc
+        raise SitemapDiscoveryError(f"Failed to fetch sitemap {url}: {last_exc}") from last_exc
+
+    def _fetch_once(self, url: str) -> SitemapFetchResult:
+        if self.transport_name == "curl_cffi":
+            return self._fetch_with_curl_cffi(url)
+        return self._fetch_with_urllib(url)
+
+    def _build_headers(self) -> dict[str, str]:
+        headers = dict(_DEFAULT_HEADERS)
+        headers["User-Agent"] = self.settings.fingerprint.user_agent
+        return headers
+
+    def _fetch_with_curl_cffi(self, url: str) -> SitemapFetchResult:
+        assert curl_requests is not None
+        response = curl_requests.get(
+            url,
+            headers=self._build_headers(),
+            timeout=self.discovery.sitemap_timeout_seconds,
+            impersonate=random.choice(_CURL_IMPERSONATE_CHOICES),
+            proxies={"http": self.proxy_url, "https": self.proxy_url} if self.proxy_url else None,
+            proxy_auth=self.proxy_auth,
+            allow_redirects=True,
+        )
+        payload = self._decode_payload(
+            payload=response.content,
+            encoding=(response.headers.get("Content-Encoding") or ""),
+            url=url,
+        )
+        blocked = _looks_like_block_page(payload)
+        return SitemapFetchResult(
+            url=url,
+            payload=payload,
+            source="curl_cffi",
+            blocked=blocked,
+            status_code=int(response.status_code),
+            content_type=response.headers.get("Content-Type"),
+        )
+
+    def _fetch_with_urllib(self, url: str) -> SitemapFetchResult:
+        request = Request(url, headers=self._build_headers())
+        try:
+            with urlopen(request, timeout=self.discovery.sitemap_timeout_seconds) as response:
+                payload = response.read()
+                decoded = self._decode_payload(
+                    payload=payload,
+                    encoding=(response.headers.get("Content-Encoding") or ""),
+                    url=url,
+                )
+                return SitemapFetchResult(
+                    url=url,
+                    payload=decoded,
+                    source="urllib",
+                    blocked=_looks_like_block_page(decoded),
+                    status_code=getattr(response, "status", None),
+                    content_type=response.headers.get("Content-Type"),
+                )
+        except HTTPError as exc:
+            payload = exc.read() if hasattr(exc, "read") else b""
+            decoded = self._decode_payload(payload=payload, encoding=exc.headers.get("Content-Encoding", ""), url=url)
+            blocked = exc.code in {403, 429} or _looks_like_block_page(decoded)
+            if blocked:
+                raise SitemapBlockedError(f"HTTP {exc.code} for {url}") from exc
+            raise SitemapDiscoveryError(f"HTTP {exc.code} for {url}") from exc
+        except URLError as exc:
+            raise SitemapDiscoveryError(f"Network error for {url}: {exc}") from exc
+
+    @staticmethod
+    def _decode_payload(*, payload: bytes, encoding: str, url: str) -> bytes:
+        normalized = (encoding or "").lower().strip()
+        if normalized == "gzip" or url.lower().endswith(".gz"):
+            try:
+                return gzip.GzipFile(fileobj=io.BytesIO(payload)).read()
+            except OSError:
+                return payload
+        return payload
+
+    def _backoff_delay(self, attempt: int) -> float:
+        base = max(0.1, float(self.discovery.sitemap_retry_backoff_seconds))
+        jitter = random.uniform(0.05, 0.35)
+        return base * (2 ** (attempt - 1)) + jitter
+
+
+def _is_vehicle_sitemap_url(url: str) -> bool:
+    lowered = url.strip().lower()
+    if "sitemapbranches" in lowered or "sitemapauctions" in lowered:
+        return False
+    return lowered.endswith(".xml") or lowered.endswith(".xml.gz")
+
+
+def _normalize_vehicle_url(url: str) -> str:
+    parts = urlsplit(url.strip())
+    return urlunsplit((parts.scheme, parts.netloc, parts.path, "", ""))
+
+
+def _local_name(tag: str) -> str:
+    if "}" in tag:
+        return tag.rsplit("}", 1)[1]
+    return tag
+
+
+def _looks_like_block_page(payload: bytes) -> bool:
+    sample = payload[:8192].lower()
+    if any(marker in sample for marker in _BLOCK_MARKERS):
+        return True
+    if any(marker in sample for marker in _HTML_MARKERS) and b"<loc>" not in sample:
+        return True
+    return False
+
+
+def _iter_loc_values_fallback(xml_bytes: bytes) -> Iterable[str]:
+    for match in _LOC_TAG_RE.finditer(xml_bytes):
+        try:
+            value = match.group(1).decode("utf-8", errors="ignore").strip()
+        except Exception:
+            continue
+        if value:
+            yield value
+
+
+def _iter_loc_values(xml_bytes: bytes) -> Iterable[str]:
+    if _looks_like_block_page(xml_bytes):
+        raise SitemapBlockedError("Anti-bot content returned instead of sitemap XML")
+    if not _XML_PREFIX_RE.search(xml_bytes[:256]):
+        fallback_values = list(_iter_loc_values_fallback(xml_bytes))
+        if fallback_values:
+            yield from fallback_values
+            return
+    try:
+        root = ET.fromstring(xml_bytes)
+    except ET.ParseError as exc:
+        fallback_values = list(_iter_loc_values_fallback(xml_bytes))
+        if fallback_values:
+            logger.warning(
+                "Falling back to regex sitemap loc extraction after XML parse error: %s",
+                exc,
+            )
+            yield from fallback_values
+            return
+        raise SitemapDiscoveryError(f"Invalid sitemap XML: {exc}") from exc
+
+    for element in root.iter():
+        if _local_name(element.tag) != "loc":
+            continue
+        if not element.text:
+            continue
+        value = element.text.strip()
+        if value:
+            yield value
+
+
+def _filter_vehicle_urls(urls: Iterable[str]) -> list[str]:
+    result: list[str] = []
+    seen: set[str] = set()
+    for url in urls:
+        normalized = _normalize_vehicle_url(url)
+        if "/VehicleDetail/" not in normalized and "/vehicledetail/" not in normalized:
+            continue
+        if normalized in seen:
+            continue
+        seen.add(normalized)
+        result.append(normalized)
+    return result
+
+
+def _looks_like_vehicle_detail_sitemap(urls: list[str]) -> bool:
+    return any("/vehicledetail/" in url.lower() for url in urls)
+
+
+def _derive_direct_probe_urls(index_url: str, limit: int) -> list[str]:
+    base_dir = index_url.rsplit("/", 1)[0] + "/"
+    return [urljoin(base_dir, f"sitemap{idx}.xml") for idx in range(1, max(1, limit) + 1)]
+
+
+def _discover_sitemap_urls(index_url: str, downloader: _SitemapDownloader, stats: SitemapDiscoveryStats) -> list[str]:
+    logger.info("Downloading sitemap index: %s", index_url)
+    try:
+        index_result = downloader.fetch(index_url)
+        sitemap_urls = [url for url in _iter_loc_values(index_result.payload) if _is_vehicle_sitemap_url(url)]
+        if sitemap_urls:
+            return sitemap_urls
+    except SitemapBlockedError as exc:
+        stats.blocked_sitemaps += 1
+        logger.warning("Sitemap index blocked, switching to direct probe: %s", exc)
+    except SitemapDiscoveryError as exc:
+        stats.malformed_sitemaps += 1
+        logger.warning("Sitemap index unusable, switching to direct probe: %s", exc)
+
+    logger.warning("Sitemap index returned no usable sitemap URLs; switching to direct probe")
+    return _probe_direct_sitemap_urls(index_url, downloader, stats)
+
+
+def _probe_direct_sitemap_urls(
+    index_url: str,
+    downloader: _SitemapDownloader,
+    stats: SitemapDiscoveryStats,
+) -> list[str]:
+    discovered: list[str] = []
+    consecutive_misses = 0
+    probe_urls = _derive_direct_probe_urls(index_url, downloader.discovery.sitemap_direct_probe_limit)
+
+    for sitemap_url in probe_urls:
+        try:
+            result = downloader.fetch(sitemap_url)
+            raw_urls = list(_iter_loc_values(result.payload))
+        except SitemapBlockedError:
+            stats.blocked_sitemaps += 1
+            consecutive_misses += 1
+            stats.direct_probe_misses += 1
+            if consecutive_misses >= downloader.discovery.sitemap_direct_probe_stop_after_misses:
+                break
+            continue
+        except SitemapDiscoveryError:
+            consecutive_misses += 1
+            stats.direct_probe_misses += 1
+            if consecutive_misses >= downloader.discovery.sitemap_direct_probe_stop_after_misses:
+                break
+            continue
+
+        if not raw_urls:
+            consecutive_misses += 1
+            stats.direct_probe_misses += 1
+            if consecutive_misses >= downloader.discovery.sitemap_direct_probe_stop_after_misses:
+                break
+            continue
+
+        consecutive_misses = 0
+        stats.direct_probe_hits += 1
+        discovered.append(sitemap_url)
+
+    return discovered
+
+
+def _collect_vehicle_urls_from_sitemap(
+    sitemap_url: str,
+    downloader: _SitemapDownloader,
+    stats: SitemapDiscoveryStats,
+) -> list[str]:
+    logger.info("Downloading sitemap: %s", sitemap_url)
+    try:
+        result = downloader.fetch(sitemap_url)
+        raw_urls = list(_iter_loc_values(result.payload))
+    except SitemapBlockedError as exc:
+        stats.blocked_sitemaps += 1
+        logger.warning("Skipping blocked sitemap %s: %s", sitemap_url, exc)
+        return []
+    except SitemapDiscoveryError as exc:
+        stats.malformed_sitemaps += 1
+        logger.warning("Skipping malformed sitemap %s: %s", sitemap_url, exc)
+        return []
+
+    stats.fetched_sitemaps += 1
+    vehicle_urls = _filter_vehicle_urls(raw_urls)
+    if raw_urls and not vehicle_urls and not _looks_like_vehicle_detail_sitemap(raw_urls):
+        logger.info("Skipping non-vehicle sitemap %s", sitemap_url)
+        return []
+
+    logger.info("Sitemap %s yielded %d vehicle URLs", sitemap_url, len(vehicle_urls))
+    return vehicle_urls
+
+
+def discover_vehicle_urls_from_sitemap(
+    index_url: str = DEFAULT_SITEMAP_INDEX_URL,
+    *,
+    settings: Settings | None = None,
+) -> list[str]:
+    result = discover_vehicle_urls_from_sitemap_with_stats(index_url=index_url, settings=settings)
+    return result.vehicle_urls
+
+
+
+def discover_vehicle_urls_from_sitemap_with_stats(
+    index_url: str = DEFAULT_SITEMAP_INDEX_URL,
+    *,
+    settings: Settings | None = None,
+) -> SitemapDiscoveryResult:
+    runtime_settings = settings or Settings()
+    downloader = _SitemapDownloader(runtime_settings)
+    stats = SitemapDiscoveryStats(transport=downloader.transport_name)
+
+    sitemap_urls = _discover_sitemap_urls(index_url, downloader, stats)
+    if not sitemap_urls:
+        raise SitemapDiscoveryError("Sitemap discovery found no sitemap URLs")
+
+    all_vehicle_urls: list[str] = []
+    for sitemap_url in sitemap_urls:
+        all_vehicle_urls.extend(_collect_vehicle_urls_from_sitemap(sitemap_url, downloader, stats))
+
+    deduped = _filter_vehicle_urls(all_vehicle_urls)
+    if not deduped:
+        raise SitemapDiscoveryError(
+            "No vehicle detail URLs discovered from vehicle sitemaps; likely anti-bot or upstream sitemap issue"
+        )
+
+    logger.info(
+        "Sitemap discovery done: %d vehicle URLs (transport=%s fetched=%d blocked=%d malformed=%d direct_hits=%d direct_misses=%d)",
+        len(deduped),
+        stats.transport,
+        stats.fetched_sitemaps,
+        stats.blocked_sitemaps,
+        stats.malformed_sitemaps,
+        stats.direct_probe_hits,
+        stats.direct_probe_misses,
+    )
+    return SitemapDiscoveryResult(vehicle_urls=deduped, stats=stats)

iaai_scraper/discovery_service.py

@@ -0,0 +1,129 @@
+import logging
+from datetime import datetime, timezone
+from typing import List
+
+from sqlalchemy.orm import Session
+
+from .core.config import settings
+from .discovery import discover_vehicle_urls_from_sitemap_with_stats, SitemapDiscoveryError
+from .storage.db import get_db_session
+from .storage.models import VehicleCandidate
+
+logger = logging.getLogger("iaai_scraper.discovery_service")
+
+
+class DiscoveryService:
+    """Сервис для обнаружения новых автомобилей через sitemap IAAI."""
+
+    def __init__(self):
+        self.db: Session = get_db_session()
+
+    def discover_new_vehicles(self, max_urls: int = None) -> int:
+        """
+        Обнаруживает новые URL автомобилей и добавляет их в vehicle_candidates.
+
+        Args:
+            max_urls: Максимальное количество URL для обработки (None = все)
+
+        Returns:
+            Количество добавленных кандидатов
+        """
+        try:
+            logger.info("Starting vehicle discovery from sitemap")
+            result = discover_vehicle_urls_from_sitemap_with_stats()
+
+            if not result.urls:
+                logger.info("No new URLs discovered")
+                return 0
+
+            # Ограничиваем количество URL
+            urls_to_process = result.urls[:max_urls] if max_urls else result.urls
+            logger.info(f"Discovered {len(urls_to_process)} potential vehicle URLs (from {len(result.urls)} total)")
+
+            # Фильтруем уже существующие кандидаты
+            existing_urls = self._get_existing_candidate_urls(urls_to_process)
+            new_urls = [url for url in urls_to_process if url not in existing_urls]
+
+            if not new_urls:
+                logger.info("All discovered URLs already exist as candidates")
+                return 0
+
+            # Добавляем новых кандидатов
+            added_count = self._add_candidates(new_urls)
+            logger.info(f"Added {added_count} new vehicle candidates")
+
+            return added_count
+
+        except SitemapDiscoveryError as e:
+            logger.error(f"Sitemap discovery failed: {e}")
+            raise
+        except Exception as e:
+            logger.error(f"Unexpected error during discovery: {e}")
+            raise
+
+    def _get_existing_candidate_urls(self, urls: List[str]) -> set:
+        """Получает множество уже существующих URL кандидатов."""
+        if not urls:
+            return set()
+
+        # Разбиваем на батчи для эффективности
+        batch_size = 1000
+        existing = set()
+
+        for i in range(0, len(urls), batch_size):
+            batch = urls[i:i + batch_size]
+            result = self.db.query(VehicleCandidate.url).filter(
+                VehicleCandidate.url.in_(batch)
+            ).all()
+            existing.update(row[0] for row in result)
+
+        return existing
+
+    def _add_candidates(self, urls: List[str]) -> int:
+        """Добавляет новые кандидаты в базу данных."""
+        now = datetime.now(timezone.utc)
+        candidates = [
+            VehicleCandidate(
+                url=url,
+                discovered_at=now,
+                status="pending",
+                priority=0
+            )
+            for url in urls
+        ]
+
+        self.db.add_all(candidates)
+        self.db.commit()
+
+        return len(candidates)
+
+    def get_pending_candidates(self, limit: int = 100) -> List[VehicleCandidate]:
+        """Получает кандидатов в статусе 'pending' для обработки."""
+        return self.db.query(VehicleCandidate).filter(
+            VehicleCandidate.status == "pending"
+        ).order_by(VehicleCandidate.priority.desc(), VehicleCandidate.discovered_at).limit(limit).all()
+
+    def mark_candidate_processing(self, candidate_id: int):
+        """Помечает кандидата как обрабатываемого."""
+        self.db.query(VehicleCandidate).filter(
+            VehicleCandidate.id == candidate_id
+        ).update({
+            "status": "processing",
+            "last_attempt_at": datetime.now(timezone.utc),
+            "attempts": VehicleCandidate.attempts + 1
+        })
+        self.db.commit()
+
+    def mark_candidate_processed(self, candidate_id: int):
+        """Помечает кандидата как обработанного."""
+        self.db.query(VehicleCandidate).filter(
+            VehicleCandidate.id == candidate_id
+        ).update({"status": "processed"})
+        self.db.commit()
+
+    def mark_candidate_failed(self, candidate_id: int):
+        """Помечает кандидата как неудачного."""
+        self.db.query(VehicleCandidate).filter(
+            VehicleCandidate.id == candidate_id
+        ).update({"status": "failed"})
+        self.db.commit()

iaai_scraper/enrichment_service.py

@@ -0,0 +1,140 @@
+import json
+import logging
+from datetime import datetime, timezone
+from typing import Optional
+
+from sqlalchemy.orm import Session
+
+from .core.config import settings
+from .parsing.mapper import CarMapper
+from .parsing.parser import VehicleParser
+from .storage.db import get_db_session
+from .storage.models import VehicleRawSnapshot, VehicleParseResult, Car
+from .storage.schemas import CarRecord
+
+logger = logging.getLogger("iaai_scraper.enrichment_service")
+
+
+class EnrichmentService:
+    """Сервис для парсинга сырых данных и обогащения автомобилей."""
+
+    def __init__(self):
+        self.db: Session = get_db_session()
+        self.parser = VehicleParser()
+        self.mapper = CarMapper()
+
+    def enrich_vehicle(self, snapshot: VehicleRawSnapshot) -> Optional[VehicleParseResult]:
+        """
+        Парсит snapshot и сохраняет результат.
+
+        Args:
+            snapshot: Raw snapshot для парсинга
+
+        Returns:
+            VehicleParseResult если парсинг успешен
+        """
+        logger.info(f"Enriching snapshot {snapshot.id} for candidate {snapshot.candidate_id}")
+
+        if not snapshot.success or not snapshot.raw_data:
+            logger.warning(f"Snapshot {snapshot.id} has no data to parse")
+            return self._save_parse_result(snapshot.id, False, None, "No raw data available")
+
+        try:
+            # Парсим данные
+            parsed_data = self._parse_raw_data(snapshot.raw_data)
+            if not parsed_data:
+                return self._save_parse_result(snapshot.id, False, None, "Parsing failed")
+
+            # Маппим в CarRecord
+            car_record = self._map_to_car_record(parsed_data)
+            if not car_record:
+                return self._save_parse_result(snapshot.id, False, None, "Mapping failed")
+
+            # Сохраняем в cars таблицу
+            self._save_car(car_record)
+
+            # Сохраняем успешный результат парсинга
+            return self._save_parse_result(snapshot.id, True, json.dumps(parsed_data))
+
+        except Exception as e:
+            error_msg = f"Unexpected error during enrichment: {str(e)}"
+            logger.error(f"Error enriching snapshot {snapshot.id}: {error_msg}")
+            return self._save_parse_result(snapshot.id, False, None, error_msg)
+
+    def _parse_raw_data(self, raw_data: str) -> Optional[dict]:
+        """Парсит сырые данные в словарь."""
+        try:
+            # Если это JSON от browser capture
+            if raw_data.strip().startswith('{'):
+                data = json.loads(raw_data)
+                # Извлекаем vehicle_summary из scrape result
+                return data.get("vehicle_summary", {})
+
+            # Если HTML, используем VehicleParser
+            parsed = self.parser.parse_vehicle_page(raw_data, "dummy_url")
+            return parsed.get("vehicle_summary", {})
+
+        except json.JSONDecodeError:
+            # Если не JSON, пробуем как HTML
+            try:
+                parsed = self.parser.parse_vehicle_page(raw_data, "dummy_url")
+                return parsed.get("vehicle_summary", {})
+            except Exception:
+                return None
+
+    def _map_to_car_record(self, parsed_data: dict) -> Optional[CarRecord]:
+        """Маппит parsed data в CarRecord."""
+        try:
+            # Используем CarMapper
+            payload_insights = {}  # TODO: extract from raw data if needed
+            return self.mapper.map_to_car_record("dummy_url", parsed_data, payload_insights)
+        except Exception as e:
+            logger.error(f"Mapping failed: {e}")
+            return None
+
+    def _save_car(self, car_record: CarRecord):
+        """Сохраняет CarRecord в базу данных."""
+        # Используем PersistenceService
+        from .storage.db import PersistenceService
+        from .core.config import settings
+        persistence = PersistenceService(settings)
+        persistence.create_tables()
+        upsert_result = persistence.upsert_car(car_record)
+        logger.info(f"Car upserted: {upsert_result}")
+
+    def _save_parse_result(self, snapshot_id: int, success: bool,
+                          parsed_data: Optional[str], error_message: Optional[str] = None) -> VehicleParseResult:
+        """Сохраняет результат парсинга."""
+        result = VehicleParseResult(
+            snapshot_id=snapshot_id,
+            success=success,
+            parsed_data=parsed_data,
+            error_message=error_message
+        )
+        self.db.add(result)
+        self.db.commit()
+        self.db.refresh(result)
+        return result
+
+    def process_unparsed_snapshots(self, limit: int = 10) -> int:
+        """
+        Обрабатывает snapshots без результатов парсинга.
+
+        Returns:
+            Количество успешно обогащенных snapshots
+        """
+        # Находим snapshots без parse results
+        snapshots = self.db.query(VehicleRawSnapshot).outerjoin(
+            VehicleParseResult, VehicleRawSnapshot.id == VehicleParseResult.snapshot_id
+        ).filter(
+            VehicleRawSnapshot.success == True,
+            VehicleParseResult.id.is_(None)
+        ).limit(limit).all()
+
+        enriched_count = 0
+        for snapshot in snapshots:
+            result = self.enrich_vehicle(snapshot)
+            if result and result.success:
+                enriched_count += 1
+
+        return enriched_count

iaai_scraper/fetch_service.py

@@ -0,0 +1,119 @@
+import logging
+from datetime import datetime, timezone
+from typing import Optional
+
+from sqlalchemy.orm import Session
+
+from .core.config import settings
+from .scraper import IAAIScraper
+from .storage.db import get_db_session
+from .storage.models import VehicleCandidate, VehicleRawSnapshot
+
+logger = logging.getLogger("iaai_scraper.fetch_service")
+
+
+class FetchService:
+    """Сервис для захвата сырых данных автомобилей (HTTP/browser fallback)."""
+
+    def __init__(self):
+        self.db: Session = get_db_session()
+        self.scraper = IAAIScraper()
+
+    def fetch_vehicle_data(self, candidate: VehicleCandidate) -> Optional[VehicleRawSnapshot]:
+        """
+        Захватывает данные для кандидата автомобиля.
+
+        Args:
+            candidate: Кандидат для обработки
+
+        Returns:
+            VehicleRawSnapshot если захват успешен, None если неудача
+        """
+        logger.info(f"Fetching data for candidate {candidate.id}: {candidate.url}")
+
+        try:
+            # Сначала пытаемся HTTP fast-path
+            raw_data = self._try_http_capture(candidate.url)
+            if raw_data:
+                return self._save_snapshot(candidate.id, "http", True, raw_data)
+
+            # Если HTTP не сработал, используем browser fallback
+            logger.info(f"HTTP failed for {candidate.url}, trying browser fallback")
+            raw_data = self._try_browser_capture(candidate.url)
+            if raw_data:
+                return self._save_snapshot(candidate.id, "browser", True, raw_data)
+
+            # Оба метода failed
+            logger.warning(f"Both HTTP and browser capture failed for {candidate.url}")
+            self._save_snapshot(candidate.id, "browser", False, None, "Both capture methods failed")
+            return None
+
+        except Exception as e:
+            error_msg = f"Unexpected error during capture: {str(e)}"
+            logger.error(f"Error fetching {candidate.url}: {error_msg}")
+            self._save_snapshot(candidate.id, "unknown", False, None, error_msg)
+            return None
+
+    def _try_http_capture(self, url: str) -> Optional[str]:
+        """Пытается захватить данные через HTTP."""
+        try:
+            # Используем существующий метод из scraper
+            # Но нам нужно инициализировать scraper с сессией
+            # Пока заглушка - интегрируем позже
+            # Для теста вернем None, чтобы использовать browser
+            return None
+        except Exception as e:
+            logger.debug(f"HTTP capture failed for {url}: {e}")
+            return None
+
+    def _try_browser_capture(self, url: str) -> Optional[str]:
+        """Пытается захватить данные через browser."""
+        try:
+            # Используем scrape_vehicle_detail из scraper
+            with IAAIScraper() as scraper:
+                result = scraper.scrape_vehicle_detail(url)
+                # Возвращаем HTML или JSON данные
+                return result.get("raw_html") or json.dumps(result)
+        except Exception as e:
+            logger.debug(f"Browser capture failed for {url}: {e}")
+            return None
+
+    def _save_snapshot(self, candidate_id: int, method: str, success: bool,
+                      raw_data: Optional[str], error_message: Optional[str] = None) -> VehicleRawSnapshot:
+        """Сохраняет snapshot в базу данных."""
+        snapshot = VehicleRawSnapshot(
+            candidate_id=candidate_id,
+            method=method,
+            success=success,
+            raw_data=raw_data,
+            error_message=error_message
+        )
+        self.db.add(snapshot)
+        self.db.commit()
+        self.db.refresh(snapshot)
+        return snapshot
+
+    def process_pending_candidates(self, limit: int = 10) -> int:
+        """
+        Обрабатывает ожидающих кандидатов.
+
+        Returns:
+            Количество успешно обработанных кандидатов
+        """
+        from .discovery_service import DiscoveryService
+        discovery = DiscoveryService()
+
+        candidates = discovery.get_pending_candidates(limit)
+        processed_count = 0
+
+        for candidate in candidates:
+            discovery.mark_candidate_processing(candidate.id)
+
+            snapshot = self.fetch_vehicle_data(candidate)
+            if snapshot and snapshot.success:
+                discovery.mark_candidate_processed(candidate.id)
+                processed_count += 1
+            else:
+                discovery.mark_candidate_failed(candidate.id)
+
+        return processed_count

iaai_scraper/parsing/mapper.py

@@ -1,5 +1,7 @@
+import hashlib
 import re
 from datetime import datetime, timezone
+from string import ascii_letters, digits
 from typing import Any
 from urllib.parse import urlparse
 
@@ -18,7 +20,7 @@ from ..storage.schemas import CarRecord, ImageRecord
 
 
 class CarMapper:
-	# IAAI data → CarRecord.
+	# Преобразование данных IAAI в CarRecord.
 
 	BODY_MAP = {
 		"sedan": "SEDAN", "coupe": "COUPE", "hatchback": "HATCHBACK", "sport utility": "SUV",
@@ -56,7 +58,7 @@ class CarMapper:
 		images = payload_insights.get("images", {})
 
 		origin_id = self._build_origin_id(vehicle_url, vehicle_summary, core)
-		parser_id = f"iaai:{origin_id}"
+		parser_id = self._generate_parser_id(origin_id)
 		brand = self._as_str(first_non_empty([core.get("make"), vehicle_summary.get("make")])) or "UNKNOWN"
 		model = self._as_str(first_non_empty([core.get("model"), vehicle_summary.get("model")])) or "UNKNOWN"
 		year = self._to_year(first_non_empty([core.get("year"), vehicle_summary.get("year")]))
@@ -70,12 +72,18 @@ class CarMapper:
 			],
 			self._to_money_int,
 		)
-		mileage = self._first_parsed_int(
+		mileage = self._parse_odometer(
-			[core.get("odometer"), vehicle_summary.get("odometer"), 0],
+			first_non_empty([core.get("odometer"), vehicle_summary.get("odometer")])
-			self._to_int,
+		)
-		) or 0
 		color = self._normalize_color(first_non_empty([core.get("color"), vehicle_summary.get("color"), "other"]))
 		drive = self._normalize_drive(first_non_empty([core.get("drive"), vehicle_summary.get("drive")]))
+		# Пытаемся определить привод из строки двигателя.
+		if not drive or drive == "NA":
+			engine_text = self._as_str(first_non_empty([core.get("engine"), vehicle_summary.get("engine")]))
+			if engine_text:
+				inferred_drive = self._normalize_drive(engine_text)
+				if inferred_drive and inferred_drive != "NA":
+					drive = inferred_drive
 		gearbox = self._normalize_gearbox(first_non_empty([core.get("gearbox"), vehicle_summary.get("gearbox")]))
 		steering = self._normalize_steering(first_non_empty([core.get("steering_wheel"), vehicle_summary.get("steering_wheel")])) or "LEFT"
 		body_type = self._normalize_body_type(first_non_empty([core.get("body_type"), vehicle_summary.get("body_type")]))
@@ -105,7 +113,7 @@ class CarMapper:
 				]
 			)
 		)
-		slug = self._slugify(" ".join(filter(None, [str(year or ""), brand, model, origin_id])))
+		slug = self._slugify(" ".join(filter(None, [brand, model, str(year or "")])))
 		images_records = self._build_images(images.get("urls") or vehicle_summary.get("image_urls") or [])
 		origin = "IAAI"
 
@@ -204,6 +212,25 @@ class CarMapper:
 			return parsed
 		return None
 
+	@staticmethod
+	def _parse_odometer(value: Any) -> int:
+		# Разбор пробега из строк IAAI.
+		if value is None:
+			return 0
+		text = str(value).strip()
+		if not text:
+			return 0
+		lowered = text.lower()
+		if any(skip in lowered for skip in ["not required", "exempt", "n/a", "na", "unknown", "tbd"]):
+			return 0
+		# Извлекаем число из строкового формата одометра.
+		numbers = re.findall(r"[\d,]+", text)
+		for num_str in numbers:
+			clean = num_str.replace(",", "")
+			if clean.isdigit() and int(clean) > 0:
+				return int(clean)
+		return 0
+
 	def _to_engine_cc(self, value: Any) -> int | None:
 		# Поддержка литров и cc.
 		text = str(value).lower().strip() if value is not None else ""
@@ -342,16 +369,34 @@ class CarMapper:
 				return preview
 		return url
 
+	# Формирование parser_id.
+
+	_PARSER_ID_ALPHABET = ascii_letters + digits
+
+	@classmethod
+	def _generate_parser_id(cls, origin_id: str) -> str:
+		# Стабильный parser_id по origin_id.
+		digest = hashlib.sha256(origin_id.encode()).digest()
+		alphabet = cls._PARSER_ID_ALPHABET
+		base = len(alphabet)
+		num = int.from_bytes(digest[:17], "big")  # 17 байт = 136 бит, хватает на 22 символа
+		chars: list[str] = []
+		for _ in range(22):
+			num, idx = divmod(num, base)
+			chars.append(alphabet[idx])
+		return "car-" + "".join(chars)
+
 	def _build_origin_id(self, vehicle_url: str, vehicle_summary: dict[str, Any], core: dict[str, Any]) -> str:
-		# Берем lot_number, затем vin, затем хвост URL.
+		# Формат: iaai:{lot_number} (аналог copart:94323985).
-		for value in [core.get("lot_number"), vehicle_summary.get("lot_number"), vehicle_summary.get("vin")]:
+		for value in [core.get("lot_number"), vehicle_summary.get("lot_number")]:
 			text = self._as_str(value)
 			if text:
-				return text
+				return f"iaai:{text}"
 		tail = urlparse(vehicle_url).path.rstrip("/").split("/")[-1]
 		if "~" in tail:
 			tail = tail.split("~")[0]
-		return tail or self._slugify(vehicle_url)
+		raw = tail or self._slugify(vehicle_url)
+		return f"iaai:{raw}"
 
 	@staticmethod
 	def _slugify(value: str) -> str:

iaai_scraper/parsing/parser.py

@@ -4,16 +4,22 @@ import logging
 import re
 from typing import Any
 
-from ..core.utils import LOT_RE, PRICE_RE, VIN_RE, deep_find_key, first_non_empty
+from ..core.utils import LOT_RE, PRICE_RE, deep_find_all_keys, deep_find_key, first_non_empty
 
 logger = logging.getLogger("iaai_scraper.parsers")
 
 
 class VehicleParser:
-	# DOM + JSON парсер страницы авто.
+	# Парсер данных страницы автомобиля.
+
+	# Регулярные выражения для парсинга и детекции защиты.
+	_BUY_NOW_RE = re.compile(r"Buy\s+Now[:\s]*\$\s*([\d,]+(?:\.\d{1,2})?)", re.IGNORECASE)
+	_CAPTCHA_TOKENS = frozenset(["captcha", "verify you are human", "i am human", "recaptcha", "cloudflare"])
+	_ANTIBOT_TOKENS = frozenset(["incapsula", "access denied", "request unsuccessful", "bot detection"])
+	_CAPTCHA_RE = re.compile(r"captcha|recaptcha|robot|are you human|security check", re.IGNORECASE)
+	_ANTIBOT_RE = re.compile(r"incapsula|imperva|ddos.guard|cloudflare|access denied|forbidden", re.IGNORECASE)
 
 	SUMMARY_KEY_MAP = {
-		"vin": {"vin", "vehicleidentificationnumber"},
 		"lot_number": {"lotnumber", "stockno", "itemid", "lotid", "itemnumber"},
 		"year": {"year"},
 		"make": {"make", "manufacturer", "brand"},
@@ -32,10 +38,10 @@ class VehicleParser:
 		"seller": {"seller", "sellername"},
 		"location": {"location", "branchname", "auctionlocation", "branch"},
 		"auction_date": {"auctiondate", "saledate", "liveauctiondate"},
-		"body_type": {"bodytype", "bodystyle"},
+		"body_type": {"bodytype", "bodystyle", "vehicletype", "bodyclass"},
-		"drive": {"driveline", "drive"},
+		"drive": {"driveline", "drive", "drivelinetype", "drivetype", "drivetrain"},
-		"gearbox": {"transmission", "gearbox"},
+		"gearbox": {"transmission", "gearbox", "transmissiontype"},
-		"engine": {"engine", "enginevolume"},
+		"engine": {"engine", "enginevolume", "enginetype", "enginedescription"},
 		"fuel_type": {"fueltype", "fuel"},
 		"cylinders": {"cylinders", "cylindercount"},
 		"color": {"color", "primarycolor", "exteriorcolor"},
@@ -43,24 +49,35 @@ class VehicleParser:
 
 	DOM_LABEL_MAP: dict[str, str] = {
 		"stock #": "lot_number",
-		"vin (status)": "vin",
+		"stock": "lot_number",
-		"vin": "vin",
 		"primary damage": "primary_damage",
 		"secondary damage": "secondary_damage",
 		"odometer": "odometer",
+		"odometer (miles)": "odometer",
+		"mileage": "odometer",
 		"body style": "body_type",
+		"body type": "body_type",
+		"vehicle type": "body_type",
 		"engine": "engine",
+		"engine type": "engine",
 		"transmission": "gearbox",
 		"drive line type": "drive",
+		"driveline type": "drive",
+		"drive line": "drive",
+		"driveline": "drive",
+		"drive type": "drive",
 		"fuel type": "fuel_type",
+		"fuel": "fuel_type",
 		"cylinders": "cylinders",
 		"exterior/interior": "color",
 		"exterior color": "color",
+		"color": "color",
 		"model": "model",
 		"series": "trim",
 		"selling branch": "location",
 		"vehicle location": "vehicle_location",
 		"auction date and time": "auction_date",
+		"sale date": "auction_date",
 		"lane/run #": "lane",
 		"actual cash value": "actual_cash_value",
 		"estimated repair cost": "estimated_repair_cost",
@@ -69,6 +86,7 @@ class VehicleParser:
 		"title/sale doc brand": "title_brand",
 		"start code": "run_and_drive",
 		"key": "keys",
+		"keys": "keys",
 		"manufactured in": "manufactured_in",
 		"vehicle class": "vehicle_class",
 	}
@@ -79,15 +97,41 @@ class VehicleParser:
 			return result
 		lines = [line.strip() for line in dom_text.split("\n") if line.strip()]
 		known_labels = set(self.DOM_LABEL_MAP.keys())
+		skip_values = {"more actions", "view", "print", "share", "back to results", "all images", "view all images"}
+		max_fields = len(set(self.DOM_LABEL_MAP.values()))
+
 		for i, line in enumerate(lines):
-			clean = line.rstrip(":").lower().strip()
+			# Ранний выход, когда уже нашли все поля.
-			if clean in known_labels and i + 1 < len(lines):
+			if len(result) >= max_fields:
+				break
+
+			# Сценарий 1: "метка: значение" в одной строке.
+			colon_pos = line.find(":")
+			if colon_pos > 0:
+				label_part = line[:colon_pos].strip().lower()
+				value_part = line[colon_pos + 1:].strip()
+				if label_part in known_labels and value_part and value_part.lower() not in skip_values:
+					field_name = self.DOM_LABEL_MAP[label_part]
+					if field_name not in result or not result[field_name]:
+						result[field_name] = value_part
+						continue
+
+			# Сценарий 2: метка и значение на соседних строках.
+			clean = line.rstrip(":").strip().lower()
+			clean_alt = clean.rstrip("#").strip()
+			matched_label = None
+			if clean in known_labels:
+				matched_label = clean
+			elif clean_alt in known_labels:
+				matched_label = clean_alt
+
+			if matched_label and i + 1 < len(lines):
 				value = lines[i + 1].strip()
 				if value.rstrip(":").lower().strip() in known_labels:
 					continue
-				if value.lower() in ("more actions", "view", "print", "share", "back to results"):
+				if value.lower() in skip_values:
 					continue
-				field_name = self.DOM_LABEL_MAP[clean]
+				field_name = self.DOM_LABEL_MAP[matched_label]
 				if field_name not in result or not result[field_name]:
 					result[field_name] = value
 		return result
@@ -120,11 +164,11 @@ class VehicleParser:
 			page_title = title_match.group(1).strip()
 		title_parsed = self._parse_title_for_year_make_model(page_title, dom_text)
 
+		# Один проход по payload для всех полей.
+		all_found = deep_find_all_keys(payloads, self.SUMMARY_KEY_MAP)
+
 		summary: dict[str, Any] = {"source_url": vehicle_url}
-		for field, candidate_keys in self.SUMMARY_KEY_MAP.items():
+		for field, values in all_found.items():
-			values: list[Any] = []
-			for payload in payloads:
-				values.extend(deep_find_key(payload, candidate_keys))
 			if field in dom_kv:
 				values.append(dom_kv[field])
 			summary[field] = first_non_empty(values)
@@ -133,11 +177,6 @@ class VehicleParser:
 		summary["make"] = summary.get("make") or title_parsed.get("make")
 		summary["model"] = summary.get("model") or title_parsed.get("model")
 		summary["trim"] = summary.get("trim") or dom_kv.get("trim")
-		summary["vin"] = summary.get("vin") or self._extract_vin(dom_text) or self._extract_vin(page_html)
-		if summary.get("vin") and isinstance(summary["vin"], str):
-			vin_clean = re.sub(r"\s*\(.*?\)\s*$", "", summary["vin"]).strip()
-			vin_match = VIN_RE.search(vin_clean)
-			summary["vin"] = vin_match.group(1) if vin_match else vin_clean
 		summary["lot_number"] = summary.get("lot_number") or self._extract_lot_number(dom_text)
 		summary["image_urls"] = self._extract_image_urls(payloads, page_html, vehicle_url)
 		for dom_field, dom_value in dom_kv.items():
@@ -147,7 +186,7 @@ class VehicleParser:
 		if not summary.get("actual_cash_value") and prices:
 			summary["actual_cash_value"] = prices[0]
 		if not summary.get("buy_now"):
-			buy_now_match = re.search(r"Buy\s+Now[:\s]*\$\s*([\d,]+(?:\.\d{1,2})?)", dom_text or "", re.IGNORECASE)
+			buy_now_match = self._BUY_NOW_RE.search(dom_text or "")
 			if buy_now_match:
 				summary["buy_now"] = buy_now_match.group(1)
 			elif prices:
@@ -160,25 +199,30 @@ class VehicleParser:
 			p = item.get("payload")
 			if isinstance(p, (dict, list)):
 				payloads.append(p)
-				for field, candidate_keys in self.SUMMARY_KEY_MAP.items():
+				# Дополнительный проход по встроенному JSON.
+				extra = deep_find_all_keys([p], self.SUMMARY_KEY_MAP)
+				for field, vals in extra.items():
 					if not summary.get(field):
-						val = first_non_empty(deep_find_key(p, candidate_keys))
+						v = first_non_empty(vals)
-						if val:
+						if v:
-							summary[field] = val
+							summary[field] = v
 
+		# Передаём image_urls без повторного извлечения.
+		image_urls = summary.get("image_urls") or []
 		return {
 			"vehicle_summary": summary,
-			"payload_insights": self._build_payload_insights(summary, responses, payloads, vehicle_url),
+			"payload_insights": self._build_payload_insights(summary, responses, payloads, vehicle_url, image_urls=image_urls),
 			"embedded_json": embedded,
 			"dom_hints": self._dom_hints(dom_text),
 			"access_notes": self._build_access_notes(summary, responses),
 		}
 
-	def _build_payload_insights(self, summary: dict[str, Any], responses: list[dict[str, Any]], payloads: list[Any], vehicle_url: str = "") -> dict[str, Any]:
+	def _build_payload_insights(self, summary: dict[str, Any], responses: list[dict[str, Any]], payloads: list[Any], vehicle_url: str = "", image_urls: list[str] | None = None) -> dict[str, Any]:
-		image_urls = self._extract_image_urls(payloads, "", vehicle_url)
+		if image_urls is None:
+			image_urls = self._extract_image_urls(payloads, "", vehicle_url)
 		return {
 			"vehicle_core": {
-				"vin": summary.get("vin"), "lot_number": summary.get("lot_number"), "year": summary.get("year"),
+				"lot_number": summary.get("lot_number"), "year": summary.get("year"),
 				"make": summary.get("make"), "model": summary.get("model"), "trim": summary.get("trim"),
 				"odometer": summary.get("odometer"), "run_and_drive": summary.get("run_and_drive"),
 				"seller": summary.get("seller"), "location": summary.get("location"), "title": summary.get("title"),
@@ -236,12 +280,12 @@ class VehicleParser:
 
 	def _build_access_notes(self, summary: dict[str, Any], responses: list[dict[str, Any]]) -> dict[str, Any]:
 		endpoints = [item.get("url", "") for item in responses]
+		dom_hints = self._dom_hints(" ".join(str(value) for value in summary.values() if value is not None))
 		return {
-			"vin_visible": bool(summary.get("vin")),
 			"images_visible": bool(summary.get("image_urls")),
 			"network_json_count": len(responses),
-			"possible_captcha": False,
+			"possible_captcha": bool(dom_hints.get("has_captcha_text")),
-			"possible_antibot": False,
+			"possible_antibot": bool(dom_hints.get("has_antibot_text")),
 			"observed_endpoints": endpoints[:20],
 		}
 
@@ -265,11 +309,6 @@ class VehicleParser:
 				return "JPY"
 		return "USD"
 
-	@staticmethod
-	def _extract_vin(text: str) -> str | None:
-		match = VIN_RE.search(text or "")
-		return match.group(1) if match else None
-
 	@staticmethod
 	def _extract_lot_number(text: str) -> str | None:
 		match = LOT_RE.search(text or "")
@@ -286,6 +325,9 @@ class VehicleParser:
 		for script_text in scripts:
 			if "{" not in script_text and "[" not in script_text:
 				continue
+			# Пропускаем слишком большие минифицированные блоки.
+			if len(script_text) > 51_200:
+				continue
 			try:
 				parsed = json.loads(script_text.strip())
 			except Exception:
@@ -360,8 +402,7 @@ class VehicleParser:
 			"has_buy_now_text": "buy now" in lowered,
 			"has_run_and_drive_text": "run & drive" in lowered or "run and drive" in lowered,
 			"has_damage_text": "damage" in lowered,
-			"has_vin_text": "vin" in lowered,
 			"has_title_text": "title" in lowered,
-			"has_captcha_text": any(token in lowered for token in ["captcha", "verify you are human", "i am human", "recaptcha", "cloudflare"]),
+			"has_captcha_text": any(token in lowered for token in VehicleParser._CAPTCHA_TOKENS),
-			"has_antibot_text": any(token in lowered for token in ["incapsula", "access denied", "request unsuccessful", "bot detection"]),
+			"has_antibot_text": any(token in lowered for token in VehicleParser._ANTIBOT_TOKENS),
 		}

iaai_scraper/scheduler_service.py

@@ -0,0 +1,69 @@
+import logging
+import time
+from datetime import datetime, timezone, timedelta
+
+from .core.config import settings
+from .discovery_service import DiscoveryService
+from .fetch_service import FetchService
+from .enrichment_service import EnrichmentService
+
+logger = logging.getLogger("iaai_scraper.scheduler_service")
+
+
+class SchedulerService:
+    """Сервис для планирования и координации ingestion pipeline."""
+
+    def __init__(self):
+        self.discovery = DiscoveryService()
+        self.fetch = FetchService()
+        self.enrichment = EnrichmentService()
+
+    def run_full_pipeline(self):
+        """Запускает полный цикл ingestion: discovery -> fetch -> enrichment."""
+        logger.info("Starting full ingestion pipeline")
+
+        try:
+            # 1. Discovery phase
+            logger.info("Phase 1: Discovery")
+            discovered_count = self.discovery.discover_new_vehicles()
+            logger.info(f"Discovered {discovered_count} new candidates")
+
+            # 2. Fetch phase
+            logger.info("Phase 2: Fetch")
+            fetched_count = self.fetch.process_pending_candidates(limit=50)
+            logger.info(f"Successfully fetched {fetched_count} candidates")
+
+            # 3. Enrichment phase
+            logger.info("Phase 3: Enrichment")
+            enriched_count = self.enrichment.process_unparsed_snapshots(limit=50)
+            logger.info(f"Successfully enriched {enriched_count} snapshots")
+
+            logger.info("Ingestion pipeline completed")
+
+        except Exception as e:
+            logger.error(f"Pipeline failed: {e}")
+            raise
+
+    def run_continuous_pipeline(self, interval_minutes: int = 30):
+        """Запускает непрерывный цикл ingestion с интервалом."""
+        logger.info(f"Starting continuous ingestion pipeline with {interval_minutes}min intervals")
+
+        while True:
+            try:
+                self.run_full_pipeline()
+            except Exception as e:
+                logger.error(f"Pipeline iteration failed: {e}")
+
+            logger.info(f"Sleeping for {interval_minutes} minutes")
+            time.sleep(interval_minutes * 60)
+
+    def run_targeted_enrichment(self):
+        """Запускает только enrichment для существующих snapshots."""
+        logger.info("Running targeted enrichment")
+        enriched_count = self.enrichment.process_unparsed_snapshots(limit=100)
+        logger.info(f"Enriched {enriched_count} snapshots")
+
+    def cleanup_old_data(self, days_to_keep: int = 30):
+        """Очищает старые данные (опционально)."""
+        # TODO: implement if needed
+        pass

iaai_scraper/storage/db.py

@@ -1,9 +1,10 @@
 import logging
 from contextlib import contextmanager
 from datetime import datetime, timezone
-from typing import Iterator
+from typing import Any, Iterator
 
-from sqlalchemy import create_engine, or_, select
+from sqlalchemy import create_engine, delete, or_, select, text, update
+from sqlalchemy.dialects.postgresql import insert as pg_insert
 from sqlalchemy.orm import Session, sessionmaker
 
 from ..core.config import Settings
@@ -13,11 +14,20 @@ from .schemas import CarRecord
 logger = logging.getLogger("iaai_scraper.db")
 
 
+def get_db_session() -> Session:
+    """Получить новую сессию базы данных."""
+    settings = Settings()
+    persistence = PersistenceService(settings)
+    return persistence.session_factory()
+
+
 CAR_DB_FIELDS = {
 	col.key for col in Car.__table__.columns
 	if col.key not in ("id",)
 }
 
+_IN_CHUNK_SIZE = 5000
+
 
 class PersistenceService:
 
@@ -99,21 +109,250 @@ class PersistenceService:
 			rows = session.execute(select(Car.origin_id).where(Car.origin_id.in_(origin_ids))).all()
 			return {str(row[0]) for row in rows if row and row[0]}
 
+	def get_existing_urls_and_ids(
+		self, origin_urls: list[str], origin_ids: list[str],
+	) -> tuple[set[str], set[str]]:
+		# Загрузка существующих URL и origin_id.
+		if not origin_urls and not origin_ids:
+			return set(), set()
+		urls: set[str] = set()
+		ids: set[str] = set()
+		with self.session_scope() as session:
+			max_len = max(len(origin_urls), len(origin_ids), 1)
+			for i in range(0, max_len, _IN_CHUNK_SIZE):
+				url_chunk = origin_urls[i:i + _IN_CHUNK_SIZE]
+				id_chunk = origin_ids[i:i + _IN_CHUNK_SIZE]
+				conditions = []
+				if url_chunk:
+					conditions.append(Car.origin_url.in_(url_chunk))
+				if id_chunk:
+					conditions.append(Car.origin_id.in_(id_chunk))
+				if not conditions:
+					continue
+				rows = session.execute(
+					select(Car.origin_url, Car.origin_id).where(or_(*conditions))
+				).all()
+				for r in rows:
+					if r[0]:
+						urls.add(str(r[0]))
+					if r[1]:
+						ids.add(str(r[1]))
+		return urls, ids
+
 	@staticmethod
 	def _add_images(session: Session, car_id: int, images: list[dict[str, object]]) -> None:
-		for image_payload in images:
+		if not images:
-			session.add(Image(fullres_image=str(image_payload["fullres_image"]), preview_image=str(image_payload["preview_image"]), order_index=int(image_payload.get("order_index", 0)), car_id=car_id))
+			return
+		session.add_all([
+			Image(
+				fullres_image=str(img["fullres_image"]),
+				preview_image=str(img["preview_image"]),
+				order_index=int(img.get("order_index", 0)),
+				car_id=car_id,
+			)
+			for img in images
+		])
 
 	@staticmethod
 	def _car_payload(record: CarRecord) -> dict[str, object]:
 		payload = record.model_dump(mode="python")
 		return {key: value for key, value in payload.items() if key in CAR_DB_FIELDS}
 
+	def _is_postgres(self) -> bool:
+		return self.engine.dialect.name == "postgresql"
+
+	def _load_existing_cars(
+		self,
+		session: Session,
+		origin_ids: list[str],
+		origin_urls: list[str],
+	) -> tuple[dict[str, Car], dict[str, Car]]:
+		existing_by_id: dict[str, Car] = {}
+		existing_by_url: dict[str, Car] = {}
+		if not origin_ids and not origin_urls:
+			return existing_by_id, existing_by_url
+
+		existing_cars: list[Car] = []
+		max_len = max(len(origin_ids), len(origin_urls), 1)
+		for i in range(0, max_len, _IN_CHUNK_SIZE):
+			id_chunk = origin_ids[i:i + _IN_CHUNK_SIZE]
+			url_chunk = origin_urls[i:i + _IN_CHUNK_SIZE]
+			conditions = []
+			if id_chunk:
+				conditions.append(Car.origin_id.in_(id_chunk))
+			if url_chunk:
+				conditions.append(Car.origin_url.in_(url_chunk))
+			if not conditions:
+				continue
+			rows = session.execute(
+				select(Car).where(or_(*conditions))
+			).scalars().all()
+			existing_cars.extend(rows)
+
+		for car in existing_cars:
+			if car.origin_id:
+				existing_by_id[car.origin_id] = car
+			if car.origin_url:
+				existing_by_url[car.origin_url] = car
+		return existing_by_id, existing_by_url
+
+	def _load_existing_image_urls(self, session: Session, car_ids: set[int]) -> dict[int, set[str]]:
+		existing_images_map: dict[int, set[str]] = {}
+		if not car_ids:
+			return existing_images_map
+
+		car_id_list = list(car_ids)
+		for i in range(0, len(car_id_list), _IN_CHUNK_SIZE):
+			chunk = car_id_list[i:i + _IN_CHUNK_SIZE]
+			img_rows = session.execute(
+				select(Image.car_id, Image.fullres_image).where(Image.car_id.in_(chunk))
+			).all()
+			for cid, furl in img_rows:
+				existing_images_map.setdefault(int(cid), set()).add(str(furl))
+		return existing_images_map
+
+	@staticmethod
+	def _postgres_upsert_set_map(insert_stmt) -> dict[str, object]:
+		return {
+			key: getattr(insert_stmt.excluded, key)
+			for key in CAR_DB_FIELDS
+		}
+
+	def _replace_images_for_car(
+		self,
+		session: Session,
+		car_id: int,
+		images: list[dict[str, object]],
+		origin_id: str,
+	) -> int:
+		nested = session.begin_nested()
+		try:
+			session.execute(delete(Image).where(Image.car_id == car_id))
+			self._add_images(session, car_id, images)
+			session.flush()
+			nested.commit()
+			return len(images)
+		except Exception:
+			nested.rollback()
+			logger.warning("Image replacement failed for car %s, keeping old images", origin_id, exc_info=True)
+			return 0
+
+	def _upsert_cars_batch_postgres(self, records: list[CarRecord]) -> dict[str, int]:
+		inserted = 0
+		updated = 0
+		images_total = 0
+
+		with self.session_scope() as session:
+			origin_ids = [r.origin_id for r in records if r.origin_id]
+			origin_urls = [r.origin_url for r in records if r.origin_url]
+			existing_by_id, existing_by_url = self._load_existing_cars(session, origin_ids, origin_urls)
+
+			entries: list[dict[str, object]] = []
+			upsert_payloads: list[dict[str, object]] = []
+			for record in records:
+				payload = self._car_payload(record)
+				images = [image.model_dump(mode="python") for image in record.images]
+				car_by_id = existing_by_id.get(record.origin_id)
+				car_by_url = existing_by_url.get(record.origin_url)
+				entry: dict[str, object] = {"record": record, "images": images, "car_id": None}
+
+				if car_by_url is not None and car_by_url.origin_id != record.origin_id and car_by_id is None:
+					for key, value in payload.items():
+						setattr(car_by_url, key, value)
+					car_by_url.last_seen_at = record.last_seen_at
+					entry["car_id"] = int(car_by_url.id)
+					updated += 1
+				else:
+					upsert_payloads.append(payload)
+					if car_by_id is not None:
+						updated += 1
+					else:
+						inserted += 1
+				entries.append(entry)
+
+			session.flush()
+
+			if upsert_payloads:
+				insert_stmt = pg_insert(Car).values(upsert_payloads)
+				upsert_stmt = insert_stmt.on_conflict_do_update(
+					index_elements=[Car.origin_id],
+					set_=self._postgres_upsert_set_map(insert_stmt),
+				).returning(Car.id, Car.origin_id)
+				rows = session.execute(upsert_stmt).all()
+				car_ids_by_origin_id = {str(origin_id): int(car_id) for car_id, origin_id in rows}
+				for entry in entries:
+					if entry["car_id"] is not None:
+						continue
+					record = entry["record"]
+					car_id = car_ids_by_origin_id.get(record.origin_id)
+					if car_id is None:
+						raise RuntimeError(f"PostgreSQL upsert did not return car_id for {record.origin_id}")
+					entry["car_id"] = car_id
+
+			car_ids = {int(entry["car_id"]) for entry in entries if entry["car_id"] is not None}
+			existing_images_map = self._load_existing_image_urls(session, car_ids)
+			images_by_car_id: dict[int, list[dict[str, object]]] = {}
+			replace_ids: list[int] = []
+
+			for entry in entries:
+				car_id = int(entry["car_id"])
+				images = entry["images"]
+				new_image_urls = {
+					str(img.get("fullres_image", ""))
+					for img in images
+					if img.get("fullres_image")
+				}
+				old_image_urls = existing_images_map.get(car_id, set())
+				if new_image_urls != old_image_urls:
+					replace_ids.append(car_id)
+					images_by_car_id[car_id] = images
+				else:
+					images_total += len(old_image_urls)
+
+			if replace_ids:
+				for i in range(0, len(replace_ids), _IN_CHUNK_SIZE):
+					chunk = replace_ids[i:i + _IN_CHUNK_SIZE]
+					session.execute(delete(Image).where(Image.car_id.in_(chunk)))
+				for car_id in replace_ids:
+					images = images_by_car_id[car_id]
+					self._add_images(session, car_id, images)
+					images_total += len(images)
+
+		return {"inserted": inserted, "updated": updated, "images_upserted": images_total}
+
 	def upsert_car(self, record: CarRecord):
-		# Insert/update автомобиля по origin_id/origin_url.
+		# Вставка или обновление автомобиля по origin_id/origin_url.
 		payload = self._car_payload(record)
 		images = [image.model_dump(mode="python") for image in record.images]
 		with self.session_scope() as session:
+			if self._is_postgres():
+				car_by_url = session.execute(
+					select(Car).where(Car.origin_url == record.origin_url)
+				).scalar_one_or_none()
+				if car_by_url is not None and car_by_url.origin_id != record.origin_id:
+					for key, value in payload.items():
+						setattr(car_by_url, key, value)
+					car_by_url.last_seen_at = record.last_seen_at
+					session.flush()
+					car_id = int(car_by_url.id)
+					action = "updated"
+				else:
+					existed = session.execute(
+						select(Car.id).where(Car.origin_id == record.origin_id)
+					).scalar_one_or_none() is not None
+					insert_stmt = pg_insert(Car).values(**payload)
+					upsert_stmt = insert_stmt.on_conflict_do_update(
+						index_elements=[Car.origin_id],
+						set_=self._postgres_upsert_set_map(insert_stmt),
+					).returning(Car.id)
+					car_id = int(session.execute(upsert_stmt).scalar_one())
+					action = "updated" if existed or car_by_url is not None else "inserted"
+
+				images_upserted = self._replace_images_for_car(
+					session, car_id, images, record.origin_id,
+				)
+				return {"car_id": car_id, "images_upserted": images_upserted, "action": action}
+
 			car = session.execute(
 				select(Car).where(or_(Car.origin_id == record.origin_id, Car.origin_url == record.origin_url))
 			).scalar_one_or_none()
@@ -128,19 +367,273 @@ class PersistenceService:
 					setattr(car, key, value)
 				car.last_seen_at = record.last_seen_at
 				session.flush()
-				nested = session.begin_nested()
+				images_upserted = self._replace_images_for_car(session, int(car.id), images, record.origin_id)
-				try:
+				return {"car_id": int(car.id), "images_upserted": images_upserted, "action": action}
-					for image in list(car.images):
-						session.delete(image)
-					session.flush()
-					self._add_images(session, int(car.id), images)
-					session.flush()
-					nested.commit()
-				except Exception:
-					nested.rollback()
-					logger.warning("Image replacement failed for car %s, keeping old images", record.origin_id)
-					images = []
-				return {"car_id": int(car.id), "images_upserted": len(images), "action": action}
 			self._add_images(session, int(car.id), images)
 			session.flush()
 			return {"car_id": int(car.id), "images_upserted": len(images), "action": action}
+	def upsert_cars_batch(self, records: list[CarRecord]) -> dict[str, int]:
+		"""Пакетный upsert нескольких автомобилей в одной транзакции.
+
+		Оптимизации:
+		- Дедупликация записей по origin_id перед вставкой.
+		- Chunked IN-queries для больших списков (обход лимита PG параметров).
+		- Пропуск перезаписи изображений, если набор URL не изменился.
+		- Один DELETE по car_id IN (...) вместо удаления по одному.
+		- Fallback на по-одному upsert если batch commit упал.
+		"""
+		# ── Дедупликация записей внутри батча ──
+		seen_ids: dict[str, int] = {}
+		unique_records: list[CarRecord] = []
+		for idx, r in enumerate(records):
+			key = r.origin_id or r.origin_url
+			if key in seen_ids:
+				logger.debug("Dedup: skipping duplicate record %s (idx %d vs %d)", key, idx, seen_ids[key])
+				continue
+			seen_ids[key] = idx
+			unique_records.append(r)
+
+		if len(unique_records) < len(records):
+			logger.info("Deduped batch: %d → %d records", len(records), len(unique_records))
+		records = unique_records
+
+		try:
+			return self._upsert_cars_batch_inner(records)
+		except Exception as exc:
+			logger.warning("Batch upsert failed (%s), falling back to individual upserts", exc)
+			return self._upsert_cars_individually(records)
+
+	def _upsert_cars_batch_inner(self, records: list[CarRecord]) -> dict[str, int]:
+		"""Внутренняя реализация batched upsert (одна транзакция)."""
+		if self._is_postgres():
+			return self._upsert_cars_batch_postgres(records)
+
+		inserted = 0
+		updated = 0
+		images_total = 0
+
+		with self.session_scope() as session:
+			# Получаем существующие записи chunked-запросами.
+			origin_ids = [r.origin_id for r in records if r.origin_id]
+			origin_urls = [r.origin_url for r in records if r.origin_url]
+
+			existing_by_id, existing_by_url = self._load_existing_cars(session, origin_ids, origin_urls)
+
+			# Предзагружаем ВСЕ изображения для обновляемых машин одним запросом.
+			existing_car_ids = set()
+			for record in records:
+				car = existing_by_id.get(record.origin_id) or existing_by_url.get(record.origin_url)
+				if car is not None:
+					existing_car_ids.add(int(car.id))
+
+			# Строим маппинг car_id → set(image_urls) для сравнения.
+			existing_images_map = self._load_existing_image_urls(session, existing_car_ids)
+
+			new_cars: list[tuple[Car, list[dict]]] = []
+			update_cars_needing_images: list[tuple[Car, list[dict]]] = []
+
+			for record in records:
+				payload = self._car_payload(record)
+				images = [image.model_dump(mode="python") for image in record.images]
+
+				car = existing_by_id.get(record.origin_id) or existing_by_url.get(record.origin_url)
+				if car is None:
+					car = Car(**payload)
+					session.add(car)
+					inserted += 1
+					new_cars.append((car, images))
+				else:
+					for key, value in payload.items():
+						setattr(car, key, value)
+					car.last_seen_at = record.last_seen_at
+					updated += 1
+
+					# Проверяем, изменились ли изображения.
+					new_image_urls = {img.get("fullres_image", "") for img in images}
+					old_image_urls = existing_images_map.get(int(car.id), set())
+					if new_image_urls != old_image_urls:
+						update_cars_needing_images.append((car, images))
+					else:
+						images_total += len(old_image_urls)
+
+			# Один flush для всех вставок.
+			session.flush()
+
+			# Добавляем изображения для новых автомобилей.
+			for car, images in new_cars:
+				self._add_images(session, int(car.id), images)
+				images_total += len(images)
+
+			# Массово обновляем изображения только для машин с изменёнными картинками.
+			if update_cars_needing_images:
+				update_ids = [int(car.id) for car, _ in update_cars_needing_images]
+				for i in range(0, len(update_ids), _IN_CHUNK_SIZE):
+					chunk = update_ids[i:i + _IN_CHUNK_SIZE]
+					session.execute(delete(Image).where(Image.car_id.in_(chunk)))
+				for car, images in update_cars_needing_images:
+					self._add_images(session, int(car.id), images)
+					images_total += len(images)
+
+		return {"inserted": inserted, "updated": updated, "images_upserted": images_total}
+
+	def _upsert_cars_individually(self, records: list[CarRecord]) -> dict[str, int]:
+		# Fallback на поштучный upsert.
+		inserted = 0
+		updated = 0
+		images_total = 0
+		for record in records:
+			try:
+				result = self.upsert_car(record)
+				action = result.get("action", "inserted")
+				if action == "inserted":
+					inserted += 1
+				else:
+					updated += 1
+				images_total += int(result.get("images_upserted", 0))
+			except Exception as exc:
+				logger.error("Individual upsert failed for %s: %s", record.origin_id, exc)
+		return {"inserted": inserted, "updated": updated, "images_upserted": images_total}
+
+	def mark_sold_not_in_listing(self, active_origin_ids: set[str], lane: str = "iaai") -> int:
+		"""Помечает авто как проданные, если их нет в активном листинге (по origin_id)."""
+		if not active_origin_ids:
+			return 0
+		with self.session_scope() as session:
+			stmt = (
+				update(Car)
+				.where(Car.origin_id.notin_(active_origin_ids))
+				.where(Car.is_sold == False)  # noqa: E712
+				.where(Car.origin_id.like("iaai:%"))
+				.values(is_sold=True)
+			)
+			result = session.execute(stmt)
+			count = result.rowcount or 0
+			if count:
+				logger.info("Marked %d cars as sold (no longer in listing)", count)
+			return count
+
+	def mark_sold_not_in_listing_by_urls(self, active_origin_urls: set[str], lane: str = "iaai") -> int:
+		"""Помечает авто как проданные, если их URL нет в активном листинге.
+
+		Для PostgreSQL использует временную таблицу + LEFT JOIN вместо NOT IN,
+		что кардинально быстрее при больших объёмах (100K+ URLs).
+		"""
+		if not active_origin_urls:
+			return 0
+
+		is_postgres = "postgresql" in self.settings.database.url
+
+		with self.session_scope() as session:
+			# Страховка от ложного mark_sold при битом/неполном full-scan:
+			# если текущий список активных URL аномально мал относительно уже активных машин в БД,
+			# ничего не помечаем проданным.
+			active_db_count = int(session.execute(
+				select(func.count())
+				.select_from(Car)
+				.where(Car.is_sold == False)  # noqa: E712
+				.where(Car.origin_id.like(f"{lane}:%"))
+			).scalar_one() or 0)
+
+			current_active_count = len(active_origin_urls)
+			if active_db_count >= 1000 and current_active_count < max(500, int(active_db_count * 0.25)):
+				logger.warning(
+					"Skipping mark_sold: suspiciously small active set (%d URLs vs %d active in DB)",
+					current_active_count,
+					active_db_count,
+				)
+				return 0
+
+			if is_postgres:
+				# Создаём временную таблицу с активными URL.
+				session.execute(text("CREATE TEMP TABLE IF NOT EXISTS _active_urls (url TEXT PRIMARY KEY) ON COMMIT DROP"))
+				session.execute(text("TRUNCATE _active_urls"))
+
+				# Вставляем активные URL чанками через executemany.
+				url_list = list(active_origin_urls)
+				for i in range(0, len(url_list), _IN_CHUNK_SIZE):
+					chunk = url_list[i:i + _IN_CHUNK_SIZE]
+					session.execute(
+						text("INSERT INTO _active_urls (url) VALUES (:url) ON CONFLICT DO NOTHING"),
+						[{"url": url} for url in chunk],
+					)
+
+				# Массовая пометка проданных в PostgreSQL.
+				result = session.execute(text("""
+					UPDATE cars
+					SET is_sold = TRUE
+					FROM (
+						SELECT c.id
+						FROM cars c
+						LEFT JOIN _active_urls a ON c.origin_url = a.url
+						WHERE a.url IS NULL
+						AND c.is_sold = FALSE
+						AND c.origin_id LIKE :lane_prefix
+					) sub
+					WHERE cars.id = sub.id
+				"""), {"lane_prefix": f"{lane}:%"})
+				count = result.rowcount or 0
+			else:
+				# Упрощённый путь для SQLite.
+				stmt = (
+					update(Car)
+					.where(Car.origin_url.notin_(active_origin_urls))
+					.where(Car.is_sold == False)  # noqa: E712
+					.where(Car.origin_id.like(f"{lane}:%"))
+					.values(is_sold=True)
+				)
+				result = session.execute(stmt)
+				count = result.rowcount or 0
+
+			if count:
+				logger.info("Marked %d cars as sold by URL (no longer in listing)", count)
+			return count
+
+	def get_all_origin_ids_for_lane(self, prefix: str = "iaai:") -> set[str]:
+		"""Возвращает все известные origin_id для заданного префикса.
+
+		Использует yield_per для потоковой загрузки при большом количестве записей.
+		"""
+		with self.session_scope() as session:
+			result = session.execute(
+				select(Car.origin_id).where(Car.origin_id.like(f"{prefix}%")).execution_options(yield_per=10000)
+			)
+			return {str(row[0]) for row in result if row and row[0]}
+
+	def get_all_active_origin_urls_for_lane(self, prefix: str = "iaai:") -> set[str]:
+		"""Возвращает все активные (не sold) origin_url для указанного lane/prefix."""
+		with self.session_scope() as session:
+			result = session.execute(
+				select(Car.origin_url)
+				.where(Car.origin_id.like(f"{prefix}%"))
+				.where(Car.is_sold == False)  # noqa: E712
+				.execution_options(yield_per=10000)
+			)
+			return {str(row[0]) for row in result if row and row[0]}
+
+	def get_active_origin_urls_batch_for_refresh(
+		self,
+		*,
+		prefix: str = "iaai:",
+		offset: int = 0,
+		limit: int = 3000,
+	) -> list[str]:
+		"""Возвращает батч активных origin_url для циклического hourly refresh."""
+		with self.session_scope() as session:
+			rows = session.execute(
+				select(Car.origin_url)
+				.where(Car.origin_id.like(f"{prefix}%"))
+				.where(Car.is_sold == False)  # noqa: E712
+				.order_by(Car.last_seen_at.asc(), Car.id.asc())
+				.offset(max(0, int(offset)))
+				.limit(max(1, int(limit)))
+			).all()
+			return [str(row[0]) for row in rows if row and row[0]]
+
+	def count_active_cars_for_lane(self, prefix: str = "iaai:") -> int:
+		with self.session_scope() as session:
+			return int(session.execute(
+				select(func.count())
+				.select_from(Car)
+				.where(Car.origin_id.like(f"{prefix}%"))
+				.where(Car.is_sold == False)  # noqa: E712
+			).scalar_one() or 0)

iaai_scraper/storage/enums.py

@@ -18,15 +18,6 @@ BODY_TYPE_ENUM_VALUES = (
 )
 COUNTRY_ENUM_VALUES = ("JP", "KR", "US", "CA", "NA")
 ORIGIN_ENUM_VALUES = (
-	"TAU",
-	"CARSENSOR",
-	"HANAMARU",
-	"ENCAR",
-	"KURUMA_TRADER",
-	"ASNET",
-	"KABABA",
-	"ACV",
-	"COPART",
 	"IAAI",
 	"NA",
 )

iaai_scraper/storage/models.py

@@ -23,6 +23,7 @@ class Car(Base):
 	__tablename__ = "cars"
 	__table_args__ = (
 		Index("ix_cars_brand_model", "brand", "model"),
+		Index("ix_cars_origin_id_not_sold", "origin_id", "is_sold"),
 	)
 	id: Mapped[int] = mapped_column(BigInteger().with_variant(Integer, "sqlite"), primary_key=True, autoincrement=True)
 	parser_id: Mapped[str] = mapped_column(String(50), nullable=False, unique=True)
@@ -30,21 +31,21 @@ class Car(Base):
 	model: Mapped[str] = mapped_column(String(50), nullable=False)
 	year: Mapped[int | None] = mapped_column(Integer, nullable=True, index=True)
 	price: Mapped[int | None] = mapped_column(BigInteger, nullable=True)
-	currency: Mapped[str] = mapped_column(Enum(*CURRENCY_ENUM_VALUES, name="currencyenum", native_enum=True, create_constraint=False), nullable=False, default="USD")
+	currency: Mapped[str] = mapped_column(Enum(*CURRENCY_ENUM_VALUES, name="currencyenum", native_enum=False, create_constraint=False), nullable=False, default="USD")
 	mileage: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
-	country: Mapped[str] = mapped_column(Enum(*COUNTRY_ENUM_VALUES, name="countryenum", native_enum=True, create_constraint=False), nullable=False, default="NA")
+	country: Mapped[str] = mapped_column(Enum(*COUNTRY_ENUM_VALUES, name="countryenum", native_enum=False, create_constraint=False), nullable=False, default="NA")
 	is_sold: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False, index=True)
 	color: Mapped[str] = mapped_column(String(), nullable=False, default="other")
-	drive: Mapped[str | None] = mapped_column(Enum(*DRIVE_ENUM_VALUES, name="driveenum", native_enum=True, create_constraint=False), nullable=True)
+	drive: Mapped[str | None] = mapped_column(Enum(*DRIVE_ENUM_VALUES, name="driveenum", native_enum=False, create_constraint=False), nullable=True)
-	gearbox: Mapped[str | None] = mapped_column(Enum(*GEARBOX_ENUM_VALUES, name="gearboxenum", native_enum=True, create_constraint=False), nullable=True)
+	gearbox: Mapped[str | None] = mapped_column(Enum(*GEARBOX_ENUM_VALUES, name="gearboxenum", native_enum=False, create_constraint=False), nullable=True)
-	steering_wheel: Mapped[str | None] = mapped_column(Enum(*STEERING_WHEEL_ENUM_VALUES, name="steeringwheelenum", native_enum=True, create_constraint=False), nullable=True)
+	steering_wheel: Mapped[str | None] = mapped_column(Enum(*STEERING_WHEEL_ENUM_VALUES, name="steeringwheelenum", native_enum=False, create_constraint=False), nullable=True)
-	body_type: Mapped[str] = mapped_column(Enum(*BODY_TYPE_ENUM_VALUES, name="bodytypeenum", native_enum=True, create_constraint=False), nullable=False, default="OTHER")
+	body_type: Mapped[str] = mapped_column(Enum(*BODY_TYPE_ENUM_VALUES, name="bodytypeenum", native_enum=False, create_constraint=False), nullable=False, default="OTHER")
 	engine_volume: Mapped[int | None] = mapped_column(Integer, nullable=True)
-	selling_type: Mapped[str] = mapped_column(Enum(*SELLING_TYPE_ENUM_VALUES, name="sellingtypeenum", native_enum=True, create_constraint=False), nullable=False, default="NA")
+	selling_type: Mapped[str] = mapped_column(Enum(*SELLING_TYPE_ENUM_VALUES, name="sellingtypeenum", native_enum=False, create_constraint=False), nullable=False, default="NA")
 	one_owner: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False)
 	new_car: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False)
 	is_hidden: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False)
-	origin: Mapped[str] = mapped_column(Enum(*ORIGIN_ENUM_VALUES, name="originenum", native_enum=True, create_constraint=False), nullable=False, default="NA")
+	origin: Mapped[str] = mapped_column(Enum(*ORIGIN_ENUM_VALUES, name="originenum", native_enum=False, create_constraint=False), nullable=False, default="NA")
 	origin_url: Mapped[str] = mapped_column(String(), nullable=False, index=True)
 	origin_id: Mapped[str] = mapped_column(String(), nullable=False, unique=True, index=True)
 	is_damaged: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False)
@@ -79,3 +80,65 @@ class SyncRun(Base):
 	cars_failed: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
 	images_upserted: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
 	error_summary: Mapped[str | None] = mapped_column(Text, nullable=True)
+
+
+class VehicleCandidate(Base):
+	__tablename__ = "vehicle_candidates"
+	__table_args__ = (
+		Index("ix_vehicle_candidates_url", "url"),
+		Index("ix_vehicle_candidates_status_discovered", "status", "discovered_at"),
+	)
+	id: Mapped[int] = mapped_column(BigInteger().with_variant(Integer, "sqlite"), primary_key=True, autoincrement=True)
+	url: Mapped[str] = mapped_column(String(), nullable=False, unique=True, index=True)
+	discovered_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False, default=func.now(), index=True)
+	status: Mapped[str] = mapped_column(String(20), nullable=False, default="pending", index=True)  # pending, processing, processed, failed
+	priority: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
+	last_attempt_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True), nullable=True)
+	attempts: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
+	raw_snapshots: Mapped[list["VehicleRawSnapshot"]] = relationship("VehicleRawSnapshot", back_populates="candidate", cascade="all, delete-orphan")
+
+
+class VehicleRawSnapshot(Base):
+	__tablename__ = "vehicle_raw_snapshots"
+	__table_args__ = (
+		Index("ix_vehicle_raw_snapshots_candidate_id", "candidate_id"),
+		Index("ix_vehicle_raw_snapshots_captured_at", "captured_at"),
+	)
+	id: Mapped[int] = mapped_column(BigInteger().with_variant(Integer, "sqlite"), primary_key=True, autoincrement=True)
+	candidate_id: Mapped[int] = mapped_column(Integer, ForeignKey("vehicle_candidates.id", ondelete="CASCADE"), nullable=False, index=True)
+	candidate: Mapped[VehicleCandidate] = relationship("VehicleCandidate", back_populates="raw_snapshots")
+	captured_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False, default=func.now(), index=True)
+	method: Mapped[str] = mapped_column(String(20), nullable=False)  # http, browser
+	success: Mapped[bool] = mapped_column(Boolean, nullable=False)
+	raw_data: Mapped[str | None] = mapped_column(Text, nullable=True)  # HTML or JSON content
+	error_message: Mapped[str | None] = mapped_column(Text, nullable=True)
+	parse_results: Mapped[list["VehicleParseResult"]] = relationship("VehicleParseResult", back_populates="snapshot", cascade="all, delete-orphan")
+
+
+class VehicleParseResult(Base):
+	__tablename__ = "vehicle_parse_results"
+	__table_args__ = (
+		Index("ix_vehicle_parse_results_snapshot_id", "snapshot_id"),
+		Index("ix_vehicle_parse_results_parsed_at", "parsed_at"),
+	)
+	id: Mapped[int] = mapped_column(BigInteger().with_variant(Integer, "sqlite"), primary_key=True, autoincrement=True)
+	snapshot_id: Mapped[int] = mapped_column(Integer, ForeignKey("vehicle_raw_snapshots.id", ondelete="CASCADE"), nullable=False, index=True)
+	snapshot: Mapped[VehicleRawSnapshot] = relationship("VehicleRawSnapshot", back_populates="parse_results")
+	parsed_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False, default=func.now(), index=True)
+	success: Mapped[bool] = mapped_column(Boolean, nullable=False)
+	parsed_data: Mapped[str | None] = mapped_column(Text, nullable=True)  # JSON with parsed vehicle data
+	error_message: Mapped[str | None] = mapped_column(Text, nullable=True)
+
+
+class VehicleRetryQueue(Base):
+	__tablename__ = "vehicle_retry_queue"
+	__table_args__ = (
+		Index("ix_vehicle_retry_queue_candidate_id", "candidate_id"),
+		Index("ix_vehicle_retry_queue_retry_at", "retry_at"),
+	)
+	id: Mapped[int] = mapped_column(BigInteger().with_variant(Integer, "sqlite"), primary_key=True, autoincrement=True)
+	candidate_id: Mapped[int] = mapped_column(Integer, ForeignKey("vehicle_candidates.id", ondelete="CASCADE"), nullable=False, index=True)
+	reason: Mapped[str] = mapped_column(String(50), nullable=False)  # parse_failed, capture_failed, etc.
+	retry_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False, index=True)
+	attempts: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
+	max_attempts: Mapped[int] = mapped_column(Integer, nullable=False, default=3)

iaai_scraper/worker/celery_app.py

@@ -1,8 +1,31 @@
 # Инициализация Celery-приложения и периодических задач.
 
+import logging
+
 from celery import Celery
+from celery.signals import worker_process_init, worker_ready, setup_logging as celery_setup_logging
+from redis import Redis
 
 from ..core.config import settings
+from ..core.logs import setup_logging
+
+logger = logging.getLogger("iaai_scraper.worker.celery_app")
+STARTUP_SYNC_DISPATCH_KEY = "iaai:state:startup_sync_dispatched"
+
+
+@celery_setup_logging.connect
+def _configure_logging(loglevel=None, **kwargs):
+    # Перехватываем логирование Celery и пишем только в stderr (Docker logs).
+    level = settings.log_level if settings.log_level else "INFO"
+    setup_logging(level, None)
+
+
+@worker_process_init.connect
+def _on_worker_process_init(**kwargs):
+    # Повторно настраиваем логирование в каждом дочернем prefork-процессе,
+    # чтобы StreamHandler(stderr) корректно работал после fork.
+    level = settings.log_level if settings.log_level else "INFO"
+    setup_logging(level, None)
 
 
 def _broker_url() -> str:
@@ -19,14 +42,27 @@ celery_app = Celery(
     backend=_result_backend(),
 )
 
+# Auto-clamp: если hard limit слишком далёк от soft (> soft + 120),
+# ограничиваем, чтобы зависший worker не жил вечно.
+_soft = settings.celery.task_soft_time_limit
+_hard = settings.celery.task_time_limit
+_max_hard = _soft + 120 if _soft else _hard
+if _hard > _max_hard:
+    logger.warning(
+        "CELERY_TASK_TIME_LIMIT=%d too far from CELERY_TASK_SOFT_TIME_LIMIT=%d; "
+        "clamping hard limit to %d",
+        _hard, _soft, _max_hard,
+    )
+    _hard = _max_hard
+
 celery_app.conf.update(
     task_serializer="json",
     accept_content=["json"],
     result_serializer="json",
     timezone="UTC",
     enable_utc=True,
-    task_soft_time_limit=settings.celery.task_soft_time_limit,
+    task_soft_time_limit=_soft,
-    task_time_limit=settings.celery.task_time_limit,
+    task_time_limit=_hard,
     task_acks_late=True,
     task_reject_on_worker_lost=True,
     task_track_started=True,
@@ -39,12 +75,14 @@ celery_app.conf.update(
         "visibility_timeout": settings.celery.broker_visibility_timeout,
     },
     result_expires=86400,
+    worker_redirect_stdouts=False,
+    worker_hijack_root_logger=False,
     beat_schedule={
         "periodic-sync-listing": {
             "task": "iaai_scraper.worker.tasks.sync_listing_task",
             "schedule": settings.celery.beat_sync_interval_minutes * 60.0,
             "args": (),
-            "kwargs": {"limit": settings.celery.beat_sync_limit},
+            "kwargs": {"limit": settings.celery.beat_sync_limit, "only_new": False},
             "options": {"queue": "scraping"},
         }
     },
@@ -54,3 +92,33 @@ celery_app.conf.update(
 )
 
 celery_app.autodiscover_tasks(["iaai_scraper.worker"])
+
+
+@worker_ready.connect
+def _on_worker_ready(**kwargs):
+    """Сразу при старте worker отправляем первую задачу sync_listing,
+    чтобы не ждать час до первого beat-цикла."""
+    try:
+        redis_client = Redis.from_url(
+            settings.redis.url,
+            decode_responses=True,
+            socket_connect_timeout=settings.redis.socket_connect_timeout_seconds,
+            socket_timeout=settings.redis.socket_timeout_seconds,
+            health_check_interval=settings.redis.health_check_interval_seconds,
+            retry_on_timeout=True,
+        )
+        should_dispatch = bool(redis_client.set(STARTUP_SYNC_DISPATCH_KEY, "1", nx=True, ex=600))
+    except Exception:
+        logger.warning("Worker ready startup sync dedupe check failed; skipping immediate dispatch", exc_info=True)
+        return
+
+    if not should_dispatch:
+        logger.info("Worker ready immediate sync already dispatched recently; skipping duplicate enqueue")
+        return
+
+    logger.info("Worker ready — dispatching initial sync_listing task")
+    celery_app.send_task(
+        "iaai_scraper.worker.tasks.sync_listing_task",
+        kwargs={"limit": settings.celery.beat_sync_limit, "only_new": False},
+        queue="scraping",
+    )

pyproject.toml

@@ -11,6 +11,7 @@ requires-python = ">=3.11"
 dependencies = [
   "alembic>=1.14.0",
   "celery>=5.4.0",
+  "curl-cffi>=0.11.0",
   "fastapi>=0.115.0",
   "playwright>=1.53.0",
   "psycopg2-binary>=2.9.9",
@@ -19,6 +20,9 @@ dependencies = [
   "redis>=5.2.0",
   "sqlalchemy>=2.0.32",
   "uvicorn>=0.34.0",
+  "urllib3>=2.0.0",
+  "orjson>=3.10.0",
+  "brotli>=1.1.0",
 ]
 
 [project.optional-dependencies]

runtime_config.json

@@ -6,8 +6,8 @@
     "ids_max_pages": null,
     "condition_check_enabled": false,
     "lane": "iaai_cars",
-    "only_new": true,
+    "only_new": false,
-    "limit": 50
+    "limit": null
   },
   "filters": {
     "price": {

tests/test_db.py

@@ -92,16 +92,6 @@ class TestPersistenceServiceIntegration(unittest.TestCase):
         self.assertEqual(len(images), 1)
         self.assertIn("imageKeys=2", images[0].fullres_image)
 
-    def test_persistence_ignores_non_db_fields(self) -> None:
-        record = self._record("1000")
-
-        result = self.persistence.upsert_car(record)
-
-        self.assertEqual(result["action"], "inserted")
-        with self.persistence.session_scope() as session:
-            car = session.execute(select(Car).where(Car.origin_id == "1000")).scalar_one()
-        self.assertEqual(car.origin_id, "1000")
-
     def test_start_sync_run_marks_stale_running_runs_as_failed(self) -> None:
         first_run_id = self.persistence.start_sync_run("lane-a")
         second_run_id = self.persistence.start_sync_run("lane-b")

tests/test_listing.py

@@ -8,34 +8,178 @@ from iaai_scraper.browser.listing import ListingCollector
 
 
 class _FakePage:
-    def __init__(self, counts: dict[str, int]) -> None:
+    def __init__(self, counts: dict[str, int], attrs: dict[str, dict[str, str]] | None = None) -> None:
         self._counts = counts
+        self._attrs = attrs or {}
+        self._evaluate_result = False
+        self._evaluate_values: list[object] = []
+        self._html = ""
+        self.url = "https://www.iaai.com/Vehiclelisting/Cars"
+        self._current_page_number: int | None = None
+        self._vehicle_hrefs: list[str] = []
+        self._wait_for_function_error: Exception | None = None
+        self._clicks: dict[str, int] = {}
+        self._goto_calls: list[str] = []
 
     class _Locator:
-        def __init__(self, count_value: int) -> None:
+        def __init__(self, page: "_FakePage", selector: str, count_value: int) -> None:
+            self._page = page
+            self._selector = selector
             self._count_value = count_value
 
+        @property
+        def first(self) -> "_FakePage._Locator":
+            return self
+
         def count(self) -> int:
             return self._count_value
 
-    def locator(self, selector: str) -> "_FakePage._Locator":
+        def is_visible(self, timeout: int | None = None) -> bool:
-        return _FakePage._Locator(self._counts.get(selector, 0))
+            _ = timeout
+            return False
+
+        def get_attribute(self, name: str, timeout: int | None = None) -> str | None:
+            _ = timeout
+            return self._page._attrs.get(self._selector, {}).get(name)
+
+        def click(self, timeout: int | None = None) -> None:
+            _ = timeout
+            self._page._clicks[self._selector] = self._page._clicks.get(self._selector, 0) + 1
+            return None
+
+    class _HtmlLocator:
+        def __init__(self, html: str) -> None:
+            self._html = html
+
+        def inner_html(self, timeout: int | None = None) -> str:
+            _ = timeout
+            return self._html
+
+    def locator(self, selector: str):
+        if selector == "html":
+            return _FakePage._HtmlLocator(self._html)
+        return _FakePage._Locator(self, selector, self._counts.get(selector, 0))
+
+    def evaluate(self, script: str, *args):
+        _ = args
+        if self._evaluate_values:
+            return self._evaluate_values.pop(0)
+        if "querySelectorAll" in script and "VehicleDetail" in script:
+            return list(self._vehicle_hrefs)
+        if "document.body?.innerText" in script and "aria-current" in script and "parseInt" in script:
+            return self._current_page_number if self._current_page_number is not None else self._evaluate_result
+        return self._evaluate_result
+
+    def wait_for_selector(self, selector: str, timeout: int | None = None) -> None:
+        _ = selector, timeout
+        return None
+
+    def wait_for_function(self, script: str, timeout: int | None = None) -> None:
+        _ = script, timeout
+        if self._wait_for_function_error is not None:
+            raise self._wait_for_function_error
+        return None
+
+    def goto(self, url: str, wait_until: str | None = None, timeout: int | None = None) -> None:
+        _ = wait_until, timeout
+        self._goto_calls.append(url)
+        self.url = url
 
 
 class TestListingUnit(unittest.TestCase):
-    def test_has_next_page_true_for_known_selector(self) -> None:
+    def test_pagination_detection_and_page_number(self) -> None:
-        page = _FakePage({"a[aria-label*='Next']": 1})
+        # Есть кнопка Next → True.
-        self.assertTrue(ListingCollector._has_next_page(page))
+        self.assertTrue(ListingCollector._has_next_page(_FakePage({"a[aria-label*='Next']":  1})))
-
+        # Нет селекторов → False.
-    def test_has_next_page_false_when_no_selectors(self) -> None:
+        self.assertFalse(ListingCollector._has_next_page(_FakePage({})))
+        # Числовая пагинация через JS → True только если evaluate явно нашёл next.
         page = _FakePage({})
-        self.assertFalse(ListingCollector._has_next_page(page))
+        page._evaluate_result = True
+        self.assertTrue(ListingCollector._has_next_page(page))
+        # Номер текущей страницы.
+        page2 = _FakePage({})
+        page2._evaluate_values = [5]
+        self.assertEqual(ListingCollector._get_current_page_number(page2), 5)
 
-    def test_constructor_with_settings_and_pacer(self) -> None:
+    def test_extract_vehicle_links_from_html_finds_detail_urls(self) -> None:
-        settings = Settings()
+        collector = ListingCollector(Settings(), HumanPacer(Settings()))
-        pacer = HumanPacer(settings)
+        html = """
-        collector = ListingCollector(settings, pacer)
+        <div>
-        self.assertIsNotNone(collector)
+            <h4><a href=\"/VehicleDetail/45184893~US\">Car 1</a></h4>
+            <script>window.__data = {\"href\":\"\\/VehicleDetail\\/45171480~US\"}</script>
+        </div>
+        """
+
+        links = collector._extract_vehicle_links_from_html(html)
+
+        self.assertEqual(
+            links,
+            [
+                ("https://www.iaai.com/VehicleDetail/45184893~US", "45184893"),
+                ("https://www.iaai.com/VehicleDetail/45171480~US", "45171480"),
+            ],
+        )
+
+    def test_has_next_page_from_html(self) -> None:
+        collector = ListingCollector(Settings(), HumanPacer(Settings()))
+        html = '<a class="pagination-next" href="/Vehiclelisting/Cars?page=43">Next</a>'
+        self.assertTrue(collector._has_next_page_from_html(html, current_page_number=42))
+        self.assertFalse(collector._has_next_page_from_html("<div>done</div>", current_page_number=42))
+
+    def test_build_listing_page_url_replaces_existing_page_parameter(self) -> None:
+        collector = ListingCollector(Settings(), HumanPacer(Settings()))
+
+        url = collector._build_listing_page_url(
+            "https://www.iaai.com/Vehiclelisting/Cars?Make=TOYOTA&page=43&foo=bar",
+            44,
+        )
+
+        self.assertEqual(
+            url,
+            "https://www.iaai.com/Vehiclelisting/Cars?Make=TOYOTA&foo=bar&page=44",
+        )
+
+    def test_collect_current_page_uses_html_first(self) -> None:
+        collector = ListingCollector(Settings(), HumanPacer(Settings()))
+        page = _FakePage({})
+        page._html = '<a href="/VehicleDetail/555~US">Car 555</a><a class="pagination-next" href="/Vehiclelisting/Cars?page=2">Next</a>'
+
+        result = collector.collect_current_page(page, page_number=1)
+
+        self.assertEqual(len(result.vehicle_links), 1)
+        self.assertEqual(result.vehicle_links[0].lot_number, "555")
+        self.assertTrue(result.next_page_detected)
+
+    def test_wait_for_navigation_result_rejects_duplicate_content_even_when_page_number_matches(self) -> None:
+        page = _FakePage({})
+        page._wait_for_function_error = RuntimeError("same content")
+        page._current_page_number = 41
+        page._vehicle_hrefs = ["/VehicleDetail/111~US", "/VehicleDetail/222~US"]
+
+        self.assertFalse(
+            ListingCollector._wait_for_navigation_result(
+                page,
+                "/VehicleDetail/111~US",
+                41,
+                old_fingerprint=("/VehicleDetail/111~US", "/VehicleDetail/222~US"),
+            )
+        )
+
+    def test_go_to_next_page_skips_flaky_ui_fallbacks_on_deep_pages(self) -> None:
+        collector = ListingCollector(Settings(), HumanPacer(Settings()))
+        page = _FakePage(
+            {ListingCollector._NEXT_PAGE_SELECTORS[0]: 1, "a[href*='/VehicleDetail/'], a[href*='/vehicledetail/'], a[href*='VehicleDetail'], a[href*='vehicledetail']": 1},
+            attrs={
+                "a[href*='/VehicleDetail/'], a[href*='/vehicledetail/'], a[href*='VehicleDetail'], a[href*='vehicledetail']": {"href": "/VehicleDetail/111~US"},
+            },
+        )
+        page._wait_for_function_error = RuntimeError("same content")
+        page._current_page_number = 41
+        page._vehicle_hrefs = ["/VehicleDetail/111~US", "/VehicleDetail/222~US"]
+
+        self.assertFalse(collector.go_to_next_page(page, expected_page_number=41))
+        self.assertGreaterEqual(len(page._goto_calls), 1)
+        self.assertEqual(page._clicks.get(ListingCollector._NEXT_PAGE_SELECTORS[0], 0), 0)
 
 
 if __name__ == "__main__":

tests/test_mappers.py

@@ -41,30 +41,27 @@ class TestCarMapper(unittest.TestCase):
 
         self.assertFalse(record.is_damaged)
 
-    def test_unknown_and_empty_values_fallbacks(self) -> None:
+    def test_unknown_empty_values_and_normalization(self) -> None:
         record = self.mapper.map_to_car_record(
             vehicle_url="https://www.iaai.com/VehicleDetail/999~US",
             vehicle_summary={"make": "  ", "model": None, "drive": "???", "gearbox": "unknown"},
             payload_insights={"vehicle_core": {}, "pricing": {}, "damage": {}, "auction": {}, "images": {}},
         )
-
         self.assertEqual(record.brand, "UNKNOWN")
         self.assertEqual(record.model, "UNKNOWN")
-        self.assertIsNone(record.steering_wheel if record.steering_wheel not in {"LEFT", None} else None)
         self.assertEqual(record.drive, "NA")
         self.assertEqual(record.gearbox, "NA")
 
-    def test_mapper_handles_case_and_spaces(self) -> None:
+        # Нормализация регистра и пробелов.
-        record = self.mapper.map_to_car_record(
+        record2 = self.mapper.map_to_car_record(
             vehicle_url="https://www.iaai.com/VehicleDetail/888~US",
             vehicle_summary={"make": "Honda", "model": "Civic", "drive": "  Front Wheel Drive  ", "gearbox": " AUTOMATIC "},
             payload_insights={"vehicle_core": {}, "pricing": {}, "damage": {}, "auction": {}, "images": {}},
         )
+        self.assertEqual(record2.drive, "FWD")
+        self.assertEqual(record2.gearbox, "AT")
 
-        self.assertEqual(record.drive, "FWD")
+    def test_price_and_currency_parsing(self) -> None:
-        self.assertEqual(record.gearbox, "AT")
-
-    def test_price_parsing_dirty_formats(self) -> None:
         record = self.mapper.map_to_car_record(
             vehicle_url="https://www.iaai.com/VehicleDetail/777~US",
             vehicle_summary={"make": "Toyota", "model": "Corolla"},
@@ -76,11 +73,9 @@ class TestCarMapper(unittest.TestCase):
                 "images": {},
             },
         )
-
         self.assertEqual(record.price, 5200)
 
-    def test_currency_detection_from_symbol(self) -> None:
+        record2 = self.mapper.map_to_car_record(
-        record = self.mapper.map_to_car_record(
             vehicle_url="https://www.iaai.com/VehicleDetail/778~US",
             vehicle_summary={"make": "Toyota", "model": "Corolla"},
             payload_insights={
@@ -91,9 +86,8 @@ class TestCarMapper(unittest.TestCase):
                 "images": {},
             },
         )
-
+        self.assertEqual(record2.currency, "EUR")
-        self.assertEqual(record.currency, "EUR")
+        self.assertEqual(record2.price, 4500)
-        self.assertEqual(record.price, 4500)
 
 
 if __name__ == "__main__":

tests/test_parser.py

@@ -9,54 +9,46 @@ class TestVehicleParserUnit(unittest.TestCase):
     def setUp(self) -> None:
         self.parser = VehicleParser()
 
-    def test_parse_dom_key_value_pairs_extracts_known_fields(self) -> None:
+    def test_parse_dom_pairs_and_title(self) -> None:
         dom_text = """
         Stock #:
         45089484
-        VIN (Status):
-        1HGCM82633A123456 (OK)
         Primary Damage:
         Front End
+        Odometer:
+        50,123 mi (Actual)
         """
         result = self.parser._parse_dom_key_value_pairs(dom_text)
         self.assertEqual(result.get("lot_number"), "45089484")
-        self.assertEqual(result.get("vin"), "1HGCM82633A123456 (OK)")
         self.assertEqual(result.get("primary_damage"), "Front End")
+        self.assertEqual(result.get("odometer"), "50,123 mi (Actual)")
 
-    def test_parse_title_for_year_make_model(self) -> None:
         parsed = self.parser._parse_title_for_year_make_model("2014 TOYOTA CAMRY for sale", "")
         self.assertEqual(parsed["year"], "2014")
         self.assertEqual(parsed["make"], "TOYOTA")
         self.assertEqual(parsed["model"], "CAMRY")
 
-    def test_extract_image_urls_filters_other_vehicle(self) -> None:
+    def test_extract_image_urls_filters_and_deduplicates(self) -> None:
-        vehicle_url = "https://www.iaai.com/VehicleDetail/45089484~US"
-        payloads = [
-            {
-                "imageUrls": [
-                    "https://vis.iaai.com/resizer?imageKeys=45089484~SID1&width=845&height=633",
-                    "https://vis.iaai.com/resizer?imageKeys=99999999~SID2&width=845&height=633",
-                ]
-            }
-        ]
-        urls = self.parser._extract_image_urls(payloads, "", vehicle_url)
-        self.assertEqual(len(urls), 1)
-        self.assertIn("45089484", urls[0])
-
-    def test_extract_image_urls_deduplicates_same_url(self) -> None:
         vehicle_url = "https://www.iaai.com/VehicleDetail/45089484~US"
         payloads = [{"imageUrls": [
             "https://vis.iaai.com/resizer?imageKeys=45089484~SID1&width=845&height=633",
             "https://vis.iaai.com/resizer?imageKeys=45089484~SID1&width=845&height=633",
+            "https://vis.iaai.com/resizer?imageKeys=99999999~SID2&width=845&height=633",
         ]}]
         urls = self.parser._extract_image_urls(payloads, "", vehicle_url)
         self.assertEqual(len(urls), 1)
+        self.assertIn("45089484", urls[0])
 
-    def test_dom_hints_detect_captcha_and_antibot(self) -> None:
+    def test_dom_hints_and_access_notes_detect_antibot(self) -> None:
         hints = self.parser._dom_hints("Please verify you are human. CAPTCHA. Incapsula access denied.")
         self.assertTrue(hints["has_captcha_text"])
         self.assertTrue(hints["has_antibot_text"])
 
+        summary = {"vin": "", "image_urls": [], "note": "Incapsula access denied. Verify you are human."}
+        notes = self.parser._build_access_notes(summary, [])
+        self.assertTrue(notes["possible_captcha"])
+        self.assertTrue(notes["possible_antibot"])
+
 
 if __name__ == "__main__":
     unittest.main()

tests/test_scraper.py

@@ -1,9 +1,13 @@
 from __future__ import annotations
 
 import unittest
+from types import SimpleNamespace
 from unittest.mock import MagicMock
 
 from iaai_scraper.core.config import Settings
+from iaai_scraper.core.exceptions import AntiBotDetectedError, SiteStructureChangedError
+from iaai_scraper.core.exceptions import ListingResumeError
+from iaai_scraper.discovery import SitemapDiscoveryError
 from iaai_scraper.scraper import IAAIScraper
 from iaai_scraper.storage.schemas import CarRecord
 
@@ -26,128 +30,410 @@ class TestScraperSync(unittest.TestCase):
         s.database.url = "sqlite://"
         return IAAIScraper(s)
 
-    def test_sync_vehicle_uses_db_record_without_remapping(self) -> None:
+    def test_sync_vehicle_uses_db_record(self) -> None:
         scraper = self._make_scraper()
-
         scraper.persistence.create_tables = MagicMock()
         scraper.persistence.start_sync_run = MagicMock(return_value=1)
         scraper.persistence.finish_sync_run = MagicMock()
         scraper.persistence.upsert_car = MagicMock(return_value={"action": "inserted", "images_upserted": 0})
-
         scraper.scrape_vehicle_detail = MagicMock(return_value={"db_record": make_db_record("111")})
-        scraper.car_mapper.map_to_car_record = MagicMock(side_effect=AssertionError("should not be called"))
 
         result = scraper.sync_vehicle("https://www.iaai.com/VehicleDetail/111~US")
 
         self.assertEqual(result["status"], "success")
         self.assertIn("trace_id", result)
-        self.assertIn("elapsed_seconds", result)
+        scraper.persistence.upsert_car.assert_called_once()
-        self.assertEqual(scraper.persistence.upsert_car.call_count, 1)
 
-    def test_sync_listing_uses_db_record_without_remapping(self) -> None:
+    def test_only_new_routing_legacy_and_streaming(self) -> None:
+        # Legacy path: only_new без listing_url.
         scraper = self._make_scraper()
-
         scraper.persistence.create_tables = MagicMock()
         scraper.persistence.start_sync_run = MagicMock(return_value=2)
         scraper.persistence.finish_sync_run = MagicMock()
-        scraper.persistence.upsert_car = MagicMock(return_value={"action": "inserted", "images_upserted": 1})
+        scraper.persistence.get_existing_urls_and_ids = MagicMock(return_value=(
-        scraper.persistence.get_existing_origin_urls = MagicMock(return_value=set())
+            {"https://www.iaai.com/VehicleDetail/111~US"}, {"iaai:222"},
-        scraper.persistence.get_existing_origin_ids = MagicMock(return_value=set())
+        ))
-
-        scraper.collect_listing = MagicMock(return_value={"vehicle_urls": ["https://www.iaai.com/VehicleDetail/222~US"]})
-        page = MagicMock()
-        scraper._scrape_on_page = MagicMock(return_value={"db_record": make_db_record("222")})
-        scraper._get_page = MagicMock(return_value=page)
-        scraper.car_mapper.map_to_car_record = MagicMock(side_effect=AssertionError("should not be called"))
-
-        result = scraper.sync_listing()
-
-        self.assertEqual(result["cars_upserted"], 1)
-        self.assertEqual(result["cars_failed"], 0)
-        self.assertIn("trace_id", result)
-        self.assertIn("elapsed_seconds", result)
-        self.assertEqual(scraper.persistence.upsert_car.call_count, 1)
-        page.close.assert_called_once()
-
-    def test_sync_listing_respects_limit(self) -> None:
-        scraper = self._make_scraper()
-
-        scraper.persistence.create_tables = MagicMock()
-        scraper.persistence.start_sync_run = MagicMock(return_value=3)
-        scraper.persistence.finish_sync_run = MagicMock()
-        scraper.persistence.upsert_car = MagicMock(return_value={"action": "inserted", "images_upserted": 1})
-        scraper.persistence.get_existing_origin_urls = MagicMock(return_value=set())
-        scraper.persistence.get_existing_origin_ids = MagicMock(return_value=set())
         scraper.collect_listing = MagicMock(return_value={
             "vehicle_urls": [
+                "https://www.iaai.com/VehicleDetail/111~US",
                 "https://www.iaai.com/VehicleDetail/222~US",
                 "https://www.iaai.com/VehicleDetail/333~US",
             ]
         })
-        scraper._scrape_on_page = MagicMock(return_value={"db_record": make_db_record("222")})
+        scraper.sync_batch = MagicMock(return_value={
-        scraper._get_page = MagicMock(return_value=MagicMock())
+            "cars_upserted": 1, "cars_failed": 0, "images_upserted": 0, "failures": [],
+        })
+        result = scraper.sync_listing(only_new=True)
+        self.assertEqual(result["skipped_existing"], 2)
+        self.assertEqual(result["cars_upserted"], 1)
 
-        scraper.sync_listing(limit=1)
+        # Streaming path: only_new + listing_url.
+        scraper2 = self._make_scraper()
+        scraper2.persistence.create_tables = MagicMock()
+        scraper2.persistence.start_sync_run = MagicMock(return_value=6)
+        scraper2.persistence.finish_sync_run = MagicMock()
+        scraper2.collect_listing = MagicMock(side_effect=AssertionError("legacy path should not be used"))
+        scraper2._sync_listing_streaming = MagicMock(return_value={
+            "listing": {"vehicles_collected": 10, "early_stopped": False, "truncated_by_time_budget": False},
+            "total": 10, "skipped_existing": 0, "cars_upserted": 10, "cars_failed": 0,
+            "images_upserted": 20, "failures": [], "all_listing_origin_urls": set(),
+        })
+        result2 = scraper2.sync_listing(
+            only_new=True,
+            listing_url="https://www.iaai.com/Vehiclelisting/Cars?Make=TOYOTA",
+            year_min=2020, year_max=2027,
+        )
+        self.assertEqual(result2["cars_upserted"], 10)
+        scraper2._sync_listing_streaming.assert_called_once()
+        scraper2.collect_listing.assert_not_called()
 
-        self.assertEqual(scraper._scrape_on_page.call_count, 1)
+    def test_full_scan_uses_sitemap_discovery_path(self) -> None:
-
-    def test_sync_listing_does_not_count_skipped_as_upserted(self) -> None:
         scraper = self._make_scraper()
-
+        scraper.settings.discovery.mode = "listing"
         scraper.persistence.create_tables = MagicMock()
-        scraper.persistence.start_sync_run = MagicMock(return_value=4)
+        scraper.persistence.start_sync_run = MagicMock(return_value=88)
         scraper.persistence.finish_sync_run = MagicMock()
-        scraper.persistence.upsert_car = MagicMock(return_value={"action": "skipped", "images_upserted": 0})
+        scraper.persistence.mark_sold_not_in_listing_by_urls = MagicMock(return_value=0)
-        scraper.persistence.get_existing_origin_urls = MagicMock(return_value=set())
+        scraper._sync_listing_via_sitemap = MagicMock(return_value={
-        scraper.persistence.get_existing_origin_ids = MagicMock(return_value=set())
+            "listing": {
-        scraper.collect_listing = MagicMock(return_value={"vehicle_urls": ["https://www.iaai.com/VehicleDetail/444~US"]})
+                "vehicles_collected": 123,
-        scraper._scrape_on_page = MagicMock(return_value={"db_record": make_db_record("444")})
+                "early_stopped": False,
-        scraper._get_page = MagicMock(return_value=MagicMock())
+                "truncated_by_time_budget": False,
+                "pagination_interrupted": False,
+                "source": "sitemap",
+            },
+            "total": 123,
+            "skipped_existing": 0,
+            "cars_upserted": 120,
+            "cars_failed": 3,
+            "images_upserted": 500,
+            "failures": [{"vehicle_url": "v", "error": "e"}],
+            "all_listing_origin_urls": {"https://www.iaai.com/VehicleDetail/111~US"},
+        })
+        scraper._sync_listing_streaming = MagicMock(side_effect=AssertionError("pagination path should not be used"))
 
         result = scraper.sync_listing()
 
+        scraper._sync_listing_via_sitemap.assert_called_once()
+        scraper._sync_listing_streaming.assert_not_called()
+        self.assertEqual(result["cars_upserted"], 120)
+        self.assertTrue(result["full_scan_completed"])
+
+    def test_full_scan_does_not_fallback_to_pagination_when_sitemap_fails(self) -> None:
+        scraper = self._make_scraper()
+        scraper.persistence.create_tables = MagicMock()
+        scraper.persistence.start_sync_run = MagicMock(return_value=89)
+        scraper.persistence.finish_sync_run = MagicMock()
+        scraper.persistence.mark_sold_not_in_listing_by_urls = MagicMock(return_value=0)
+        scraper._sync_listing_via_sitemap = MagicMock(side_effect=SitemapDiscoveryError("sitemap down"))
+        result = scraper.sync_listing()
+
+        scraper._sync_listing_via_sitemap.assert_called_once()
+        self.assertEqual(result["status"], "failed")
         self.assertEqual(result["cars_upserted"], 0)
 
-    def test_sync_listing_only_new_filters_existing_by_url_and_origin_id(self) -> None:
+    def test_segmented_sync_calls_per_segment_and_resumes(self) -> None:
         scraper = self._make_scraper()
-
         scraper.persistence.create_tables = MagicMock()
-        scraper.persistence.start_sync_run = MagicMock(return_value=5)
+        scraper.persistence.start_sync_run = MagicMock(return_value=3)
         scraper.persistence.finish_sync_run = MagicMock()
-        scraper.persistence.upsert_car = MagicMock(return_value={"action": "inserted", "images_upserted": 0})
-        scraper.persistence.get_existing_origin_urls = MagicMock(return_value={"https://www.iaai.com/VehicleDetail/111~US"})
-        scraper.persistence.get_existing_origin_ids = MagicMock(return_value={"222"})
 
-        scraper.collect_listing = MagicMock(return_value={
+        call_args_log: list[dict] = []
-            "vehicle_urls": [
+        def _fake_sync_listing(**kwargs):
-                "https://www.iaai.com/VehicleDetail/111~US",  # exists by URL
+            call_args_log.append(kwargs)
-                "https://www.iaai.com/VehicleDetail/222~US",  # exists by ID
+            return {
-                "https://www.iaai.com/VehicleDetail/333~US",  # new
+                "status": "success", "cars_upserted": 5, "cars_failed": 0,
+                "images_upserted": 10, "skipped_existing": 0,
+                "listing": {"vehicles_collected": 50}, "failures": [],
+            }
+
+        scraper.sync_listing = MagicMock(side_effect=_fake_sync_listing)
+        segments = [
+            {"make": "TOYOTA", "year_min": 2020, "year_max": 2027},
+            {"make": "FORD", "year_min": None, "year_max": None},
+            {"make": "HONDA", "year_min": None, "year_max": None},
+        ]
+
+        # Resume: пропускаем TOYOTA, начинаем сразу с FORD.
+        result = scraper.sync_listing_segmented(
+            segments=segments, start_segment=1,
+        )
+
+        self.assertEqual(result["segments_completed"], 2)  # FORD + HONDA
+        self.assertEqual(result["cars_upserted"], 10)
+        # URL содержит бренд.
+        self.assertIn("FORD", call_args_log[0]["listing_url"])
+        self.assertIn("HONDA", call_args_log[1]["listing_url"])
+
+    def test_segmented_sync_does_not_mark_full_scan_completed_when_segment_failed(self) -> None:
+        scraper = self._make_scraper()
+        scraper.sync_listing = MagicMock(side_effect=[
+            {
+                "status": "failed",
+                "full_scan_completed": False,
+                "cars_upserted": 0,
+                "cars_failed": 0,
+                "images_upserted": 0,
+                "skipped_existing": 0,
+                "listing": {"vehicles_collected": 0},
+                "failures": [{"vehicle_url": "segment", "error": "resume failed"}],
+            },
+            {
+                "status": "success",
+                "full_scan_completed": True,
+                "cars_upserted": 1,
+                "cars_failed": 0,
+                "images_upserted": 0,
+                "skipped_existing": 0,
+                "listing": {"vehicles_collected": 10},
+                "failures": [],
+            },
+        ])
+
+        result = scraper.sync_listing_segmented(
+            segments=[
+                {"make": "EAGLE", "year_min": None, "year_max": None},
+                {"make": "FORD", "year_min": None, "year_max": None},
             ]
-        })
+        )
-        page = MagicMock()
-        scraper._get_page = MagicMock(return_value=page)
-        scraper._scrape_on_page = MagicMock(return_value={"db_record": make_db_record("333")})
 
-        result = scraper.sync_listing(only_new=True)
+        self.assertFalse(result["full_scan_completed"])
+        self.assertEqual(result["status"], "partial_success")
 
-        self.assertEqual(result["skipped_existing"], 2)
+    def test_build_segment_listing_url(self) -> None:
-        self.assertEqual(result["cars_upserted"], 1)
+        base = "https://www.iaai.com/Vehiclelisting/Cars"
-        self.assertEqual(scraper._scrape_on_page.call_count, 1)
+        self.assertEqual(
-        scraper.persistence.get_existing_origin_urls.assert_called_once()
+            IAAIScraper._build_segment_listing_url(base, "TOYOTA"),
-        scraper.persistence.get_existing_origin_ids.assert_called_once()
+            "https://www.iaai.com/Vehiclelisting/Cars?Make=TOYOTA",
+        )
+        self.assertEqual(
+            IAAIScraper._build_segment_listing_url(base, "LAND ROVER"),
+            "https://www.iaai.com/Vehiclelisting/Cars?Make=LAND%20ROVER",
+        )
+        self.assertEqual(IAAIScraper._build_segment_listing_url(base, None), base)
+        self.assertEqual(IAAIScraper._build_segment_listing_url(base, ""), base)
+
+    def test_guard_and_protection_detection(self) -> None:
+        with self.assertRaises(AntiBotDetectedError):
+            IAAIScraper._raise_if_blocked_or_incomplete(
+                {"dom_hints": {"has_captcha_text": True, "has_antibot_text": False},
+                 "access_notes": {}, "vehicle_summary": {}},
+                "https://www.iaai.com/VehicleDetail/999~US",
+            )
+        with self.assertRaises(SiteStructureChangedError):
+            IAAIScraper._raise_if_blocked_or_incomplete(
+                {"dom_hints": {"has_captcha_text": False, "has_antibot_text": False},
+                 "access_notes": {"possible_captcha": False, "possible_antibot": False},
+                 "vehicle_summary": {}},
+                "https://www.iaai.com/VehicleDetail/999~US",
+            )
+        self.assertTrue(IAAIScraper._is_protection_or_network_error(RuntimeError("NS_ERROR_NET_INTERRUPT")))
+        self.assertTrue(IAAIScraper._is_protection_or_network_error(RuntimeError("captcha challenge")))
+        self.assertFalse(IAAIScraper._is_protection_or_network_error(RuntimeError("plain validation error")))
 
     def test_close_resets_browser_state(self) -> None:
         scraper = self._make_scraper()
+        http_pool = MagicMock()
+        scraper._http_pool = http_pool
         scraper.context = MagicMock()
         scraper.browser = MagicMock()
         scraper.playwright = MagicMock()
-
         scraper.close()
-
+        http_pool.clear.assert_called_once()
         self.assertIsNone(scraper.context)
         self.assertIsNone(scraper.browser)
-        self.assertIsNone(scraper.playwright)
+
+    def test_recover_empty_listing_page(self) -> None:
+        scraper = self._make_scraper()
+        page = MagicMock()
+        page_result = SimpleNamespace(
+            vehicle_links=[SimpleNamespace(href="https://www.iaai.com/VehicleDetail/123~US")],
+        )
+        scraper.listing_collector.collect_current_page = MagicMock(return_value=page_result)
+        scraper.listing_collector.open_cars_listing = MagicMock()
+
+        # Страница > 1: reload.
+        _, urls = scraper._recover_empty_listing_page(
+            page, page_number=5, all_raw_urls=[], seen_urls=set(),
+        )
+        page.reload.assert_called_once()
+        self.assertEqual(len(urls), 1)
+
+        # Страница 1: переоткрытие листинга.
+        page.reset_mock()
+        _, urls = scraper._recover_empty_listing_page(
+            page, page_number=1, all_raw_urls=[], seen_urls=set(),
+        )
+        scraper.listing_collector.open_cars_listing.assert_called_once()
+        page.reload.assert_not_called()
+
+    def test_sync_listing_always_starts_from_page_one(self) -> None:
+        scraper = self._make_scraper()
+        scraper.settings.celery.batch_size = 1
+
+        page = MagicMock()
+        page_result = SimpleNamespace(
+            page_number=1,
+            vehicle_links=[SimpleNamespace(href="https://www.iaai.com/VehicleDetail/999~US", lot_number="999")],
+            next_page_detected=False,
+        )
+
+        scraper._get_page_with_warmup = MagicMock(return_value=page)
+        # Pagination-resume убран: _reopen_listing_and_resume не должен вызываться.
+        scraper._reopen_listing_and_resume = MagicMock(
+            side_effect=AssertionError("pagination resume must not be used")
+        )
+        scraper.listing_collector.open_cars_listing = MagicMock()
+        scraper.listing_collector.apply_filters = MagicMock(return_value={
+            "make": None,
+            "model": None,
+            "year_min": None,
+            "year_max": None,
+        })
+        scraper.listing_collector.collect_current_page = MagicMock(return_value=page_result)
+        scraper._extract_page_urls = MagicMock(return_value=["https://www.iaai.com/VehicleDetail/999~US"])
+        scraper.sync_batch = MagicMock(return_value={
+            "cars_upserted": 1,
+            "cars_failed": 0,
+            "images_upserted": 0,
+            "failures": [],
+        })
+
+        result = scraper._sync_listing_streaming(
+            make=None,
+            model=None,
+            lane="iaai_cars",
+            limit=None,
+            effective_only_new=False,
+            started_at=0.0,
+            listing_url="https://www.iaai.com/Vehiclelisting/Cars?Make=EAGLE",
+        )
+
+        scraper.listing_collector.open_cars_listing.assert_called_once()
+        scraper._reopen_listing_and_resume.assert_not_called()
+        scraper.sync_batch.assert_called_once()
+        self.assertEqual(result["cars_upserted"], 1)
+        self.assertEqual(result["total"], 1)
+
+    def test_sync_listing_marks_pagination_interrupted_when_next_page_resume_fails(self) -> None:
+        scraper = self._make_scraper()
+        scraper.settings.celery.batch_size = 1000
+
+        page = MagicMock()
+        page_result = SimpleNamespace(
+            page_number=1,
+            vehicle_links=[SimpleNamespace(href="https://www.iaai.com/VehicleDetail/999~US", lot_number="999")],
+            next_page_detected=True,
+        )
+
+        scraper._get_page_with_warmup = MagicMock(return_value=page)
+        scraper.listing_collector.open_cars_listing = MagicMock()
+        scraper.listing_collector.apply_filters = MagicMock(return_value={
+            "make": None,
+            "model": None,
+            "year_min": None,
+            "year_max": None,
+        })
+        scraper.listing_collector.collect_current_page = MagicMock(return_value=page_result)
+        scraper.listing_collector.go_to_next_page = MagicMock(return_value=False)
+        scraper._extract_page_urls = MagicMock(return_value=["https://www.iaai.com/VehicleDetail/999~US"])
+        scraper._reopen_listing_and_resume = MagicMock(side_effect=ListingResumeError("resume failed"))
+        scraper.sync_batch = MagicMock(return_value={
+            "cars_upserted": 0,
+            "cars_failed": 0,
+            "images_upserted": 0,
+            "failures": [],
+        })
+
+        result = scraper._sync_listing_streaming(
+            make=None,
+            model=None,
+            lane="iaai_cars",
+            limit=None,
+            effective_only_new=False,
+            started_at=0.0,
+            listing_url="https://www.iaai.com/Vehiclelisting/Cars?Make=TOYOTA",
+        )
+
+        self.assertTrue(result["listing"]["pagination_interrupted"])
+
+    def test_sync_listing_attempts_next_page_even_without_detected_next_control(self) -> None:
+        scraper = self._make_scraper()
+        scraper.settings.celery.batch_size = 1000
+
+        page = MagicMock()
+        first_page = SimpleNamespace(
+            page_number=1,
+            vehicle_links=[SimpleNamespace(href="https://www.iaai.com/VehicleDetail/111~US", lot_number="111")],
+            next_page_detected=False,
+        )
+        second_page = SimpleNamespace(
+            page_number=2,
+            vehicle_links=[SimpleNamespace(href="https://www.iaai.com/VehicleDetail/222~US", lot_number="222")],
+            next_page_detected=False,
+        )
+
+        scraper._get_page_with_warmup = MagicMock(return_value=page)
+        scraper.listing_collector.open_cars_listing = MagicMock()
+        scraper.listing_collector.apply_filters = MagicMock(return_value={
+            "make": None,
+            "model": None,
+            "year_min": None,
+            "year_max": None,
+        })
+        scraper.listing_collector.collect_current_page = MagicMock(side_effect=[first_page, second_page])
+        scraper.listing_collector.go_to_next_page = MagicMock(side_effect=[True, False])
+        scraper._extract_page_urls = MagicMock(side_effect=[
+            ["https://www.iaai.com/VehicleDetail/111~US"],
+            ["https://www.iaai.com/VehicleDetail/222~US"],
+        ])
+        scraper.sync_batch = MagicMock(return_value={
+            "cars_upserted": 2,
+            "cars_failed": 0,
+            "images_upserted": 0,
+            "failures": [],
+        })
+
+        result = scraper._sync_listing_streaming(
+            make=None,
+            model=None,
+            lane="iaai_cars",
+            limit=None,
+            effective_only_new=False,
+            started_at=0.0,
+            listing_url="https://www.iaai.com/Vehiclelisting/Cars?Make=TOYOTA",
+        )
+
+        self.assertEqual(scraper.listing_collector.go_to_next_page.call_count, 2)
+        scraper.listing_collector.go_to_next_page.assert_any_call(page, expected_page_number=2)
+        self.assertEqual(result["listing"]["pages_collected"], 2)
+
+    def test_sync_listing_treats_pagination_interrupted_as_partial_scan(self) -> None:
+        scraper = self._make_scraper()
+        scraper.persistence.create_tables = MagicMock()
+        scraper.persistence.start_sync_run = MagicMock(return_value=77)
+        scraper.persistence.finish_sync_run = MagicMock()
+        scraper.persistence.mark_sold_not_in_listing_by_urls = MagicMock()
+        scraper._sync_listing_streaming = MagicMock(return_value={
+            "listing": {
+                "vehicles_collected": 10,
+                "early_stopped": False,
+                "truncated_by_time_budget": False,
+                "pagination_interrupted": True,
+            },
+            "total": 10,
+            "skipped_existing": 0,
+            "cars_upserted": 10,
+            "cars_failed": 0,
+            "images_upserted": 0,
+            "failures": [],
+            "all_listing_origin_urls": {"https://www.iaai.com/VehicleDetail/999~US"},
+        })
+
+        result = scraper.sync_listing(listing_url="https://www.iaai.com/Vehiclelisting/Cars?Make=TOYOTA")
+
+        self.assertTrue(result["full_scan_completed"] is False)
+        scraper.persistence.mark_sold_not_in_listing_by_urls.assert_not_called()
 
 
 if __name__ == "__main__":

tests/test_sitemap_discovery.py

@@ -0,0 +1,217 @@
+from __future__ import annotations
+
+import unittest
+from unittest.mock import patch
+
+from iaai_scraper.core.config import Settings
+from iaai_scraper.discovery.sitemap import (
+    SitemapBlockedError,
+    SitemapDiscoveryError,
+    SitemapFetchResult,
+    _filter_vehicle_urls,
+    discover_vehicle_urls_from_sitemap,
+    discover_vehicle_urls_from_sitemap_with_stats,
+)
+
+
+class TestSitemapDiscovery(unittest.TestCase):
+    @staticmethod
+    def _fetch_result(url: str, payload: bytes, *, source: str = "curl_cffi") -> SitemapFetchResult:
+        return SitemapFetchResult(url=url, payload=payload, source=source)
+
+    def test_filter_vehicle_urls_dedupes_and_normalizes(self) -> None:
+        urls = _filter_vehicle_urls([
+            "https://www.iaai.com/VehicleDetail/111~US?foo=1",
+            "https://www.iaai.com/VehicleDetail/111~US?bar=2",
+            "https://www.iaai.com/VehicleDetail/222~US",
+            "https://www.iaai.com/about",
+        ])
+
+        self.assertEqual(
+            urls,
+            [
+                "https://www.iaai.com/VehicleDetail/111~US",
+                "https://www.iaai.com/VehicleDetail/222~US",
+            ],
+        )
+
+    def test_discover_vehicle_urls_from_sitemap(self) -> None:
+        index_xml = b"""
+        <sitemapindex xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
+          <sitemap><loc>https://www.iaai.com/sitemap-a.xml</loc></sitemap>
+          <sitemap><loc>https://www.iaai.com/sitemap-b.xml</loc></sitemap>
+                    <sitemap><loc>https://www.iaai.com/sitemapauctions1.xml</loc></sitemap>
+        </sitemapindex>
+        """
+        sitemap_a = b"""
+        <urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
+          <url><loc>https://www.iaai.com/VehicleDetail/111~US</loc></url>
+          <url><loc>https://www.iaai.com/VehicleDetail/222~US?x=1</loc></url>
+        </urlset>
+        """
+        sitemap_b = b"""
+        <urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
+          <url><loc>https://www.iaai.com/VehicleDetail/222~US?y=2</loc></url>
+          <url><loc>https://www.iaai.com/VehicleDetail/333~US</loc></url>
+        </urlset>
+        """
+
+        def _fake_fetch(url: str) -> SitemapFetchResult:
+            if url.endswith("sitemap_index.xml"):
+                return self._fetch_result(url, index_xml)
+            if url.endswith("sitemap-a.xml"):
+                return self._fetch_result(url, sitemap_a)
+            if url.endswith("sitemap-b.xml"):
+                return self._fetch_result(url, sitemap_b)
+            raise AssertionError(f"unexpected url: {url}")
+
+        with patch("iaai_scraper.discovery.sitemap._SitemapDownloader.fetch", side_effect=_fake_fetch):
+            result = discover_vehicle_urls_from_sitemap("https://www.iaai.com/sitemap_index.xml")
+
+        self.assertEqual(
+            result,
+            [
+                "https://www.iaai.com/VehicleDetail/111~US",
+                "https://www.iaai.com/VehicleDetail/222~US",
+                "https://www.iaai.com/VehicleDetail/333~US",
+            ],
+        )
+
+    def test_discover_vehicle_urls_raises_on_empty_index(self) -> None:
+        empty_index = b"<sitemapindex xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\"></sitemapindex>"
+        with patch(
+            "iaai_scraper.discovery.sitemap._SitemapDownloader.fetch",
+            return_value=self._fetch_result("https://www.iaai.com/sitemap_index.xml", empty_index),
+        ):
+            with self.assertRaises(SitemapDiscoveryError):
+                discover_vehicle_urls_from_sitemap("https://www.iaai.com/sitemap_index.xml")
+
+    def test_discover_vehicle_urls_skips_malformed_and_non_vehicle_sitemaps(self) -> None:
+        sitemap_vehicle = b"""
+        <urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
+            <url><loc>https://www.iaai.com/VehicleDetail/111~US</loc></url>
+        </urlset>
+        """
+        sitemap_non_vehicle = b"""
+        <urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
+            <url><loc>https://www.iaai.com/SalesList/111~US/04222026</loc></url>
+        </urlset>
+        """
+        index_xml = b"""
+        <sitemapindex xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
+            <sitemap><loc>https://www.iaai.com/sitemap1.xml</loc></sitemap>
+            <sitemap><loc>https://www.iaai.com/sitemapbranches1.xml</loc></sitemap>
+            <sitemap><loc>https://www.iaai.com/sitemap2.xml</loc></sitemap>
+            <sitemap><loc>https://www.iaai.com/sitemap3.xml</loc></sitemap>
+        </sitemapindex>
+        """
+
+        def _fake_fetch(url: str) -> SitemapFetchResult:
+            if url.endswith("sitemap_index.xml"):
+                return self._fetch_result(url, index_xml)
+            if url.endswith("sitemap1.xml"):
+                return self._fetch_result(url, sitemap_vehicle)
+            if url.endswith("sitemap2.xml"):
+                return self._fetch_result(url, sitemap_non_vehicle)
+            if url.endswith("sitemap3.xml"):
+                raise SitemapDiscoveryError("broken sitemap")
+            raise AssertionError(f"unexpected url: {url}")
+
+        with patch("iaai_scraper.discovery.sitemap._SitemapDownloader.fetch", side_effect=_fake_fetch):
+            result = discover_vehicle_urls_from_sitemap("https://www.iaai.com/sitemap_index.xml")
+
+        self.assertEqual(result, ["https://www.iaai.com/VehicleDetail/111~US"])
+
+    def test_discover_vehicle_urls_falls_back_to_regex_loc_extraction(self) -> None:
+        malformed_index = (
+            b"<sitemapindex>"
+            b"<sitemap><loc>https://www.iaai.com/sitemap1.xml</loc></sitemap>"
+            b"<broken"
+        )
+        sitemap_vehicle = b"""
+        <urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
+          <url><loc>https://www.iaai.com/VehicleDetail/111~US</loc></url>
+        </urlset>
+        """
+
+        def _fake_fetch(url: str) -> SitemapFetchResult:
+            if url.endswith("sitemap_index.xml"):
+                return self._fetch_result(url, malformed_index)
+            if url.endswith("sitemap1.xml"):
+                return self._fetch_result(url, sitemap_vehicle)
+            raise AssertionError(f"unexpected url: {url}")
+
+        with patch("iaai_scraper.discovery.sitemap._SitemapDownloader.fetch", side_effect=_fake_fetch):
+            result = discover_vehicle_urls_from_sitemap("https://www.iaai.com/sitemap_index.xml")
+
+        self.assertEqual(result, ["https://www.iaai.com/VehicleDetail/111~US"])
+
+    def test_discovery_uses_direct_probe_when_index_has_no_urls(self) -> None:
+        index_xml = b"<sitemapindex xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\"></sitemapindex>"
+        sitemap_one = b"""
+        <urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
+          <url><loc>https://www.iaai.com/VehicleDetail/111~US</loc></url>
+        </urlset>
+        """
+
+        settings = Settings()
+        settings.discovery.sitemap_direct_probe_limit = 2
+        settings.discovery.sitemap_direct_probe_stop_after_misses = 1
+
+        def _fake_fetch(url: str) -> SitemapFetchResult:
+            if url.endswith("sitemap_index.xml"):
+                return self._fetch_result(url, index_xml)
+            if url.endswith("sitemap1.xml"):
+                return self._fetch_result(url, sitemap_one)
+            raise SitemapDiscoveryError("not found")
+
+        with patch("iaai_scraper.discovery.sitemap._SitemapDownloader.fetch", side_effect=_fake_fetch):
+            result = discover_vehicle_urls_from_sitemap_with_stats(
+                "https://www.iaai.com/sitemap_index.xml",
+                settings=settings,
+            )
+
+        self.assertEqual(result.vehicle_urls, ["https://www.iaai.com/VehicleDetail/111~US"])
+        self.assertEqual(result.stats.direct_probe_hits, 1)
+
+    def test_discovery_stats_report_direct_probe_hits(self) -> None:
+        index_xml = b"<sitemapindex xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\"></sitemapindex>"
+        sitemap1 = b"""
+        <urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
+          <url><loc>https://www.iaai.com/VehicleDetail/111~US</loc></url>
+        </urlset>
+        """
+
+        def _fake_fetch(url: str):
+            if url.endswith("sitemap_index.xml"):
+                return self._fetch_result(url, index_xml)
+            if url.endswith("sitemap1.xml"):
+                return self._fetch_result(url, sitemap1)
+            raise SitemapDiscoveryError("not found")
+
+        with patch("iaai_scraper.discovery.sitemap._SitemapDownloader.fetch", side_effect=_fake_fetch):
+            result = discover_vehicle_urls_from_sitemap_with_stats("https://www.iaai.com/sitemap_index.xml")
+
+        self.assertEqual(result.vehicle_urls, ["https://www.iaai.com/VehicleDetail/111~US"])
+        self.assertEqual(result.stats.transport, "curl_cffi")
+        self.assertEqual(result.stats.direct_probe_hits, 1)
+        self.assertGreaterEqual(result.stats.direct_probe_misses, 1)
+
+    def test_block_page_raises_discovery_error(self) -> None:
+        settings = Settings()
+        settings.discovery.sitemap_direct_probe_limit = 1
+        settings.discovery.sitemap_direct_probe_stop_after_misses = 1
+
+        with patch(
+            "iaai_scraper.discovery.sitemap._SitemapDownloader.fetch",
+            side_effect=SitemapBlockedError("Anti-bot page returned for https://www.iaai.com/sitemap_index.xml"),
+        ):
+            with self.assertRaises(SitemapDiscoveryError):
+                discover_vehicle_urls_from_sitemap(
+                    "https://www.iaai.com/sitemap_index.xml",
+                    settings=settings,
+                )
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_utils.py

@@ -3,28 +3,72 @@ from __future__ import annotations
 import unittest
 
 from iaai_scraper.core.utils import deep_find_key
+from iaai_scraper.core.config import parse_listing_segments, IAAI_DEFAULT_MAKES, _LARGE_MAKES, _YEAR_SPLITS
 
 
 class TestDeepFindKey(unittest.TestCase):
-    def test_finds_key_in_nested_structure(self) -> None:
+    def test_finds_nested_and_respects_depth(self) -> None:
         payload = {
             "root": {
                 "target": "a",
                 "nested": [{"target": "b"}, {"x": 1}],
             }
         }
+        self.assertEqual(deep_find_key(payload, {"target"}), ["a", "b"])
 
-        found = deep_find_key(payload, {"target"})
+        deep_payload = {"l1": {"l2": {"l3": {"target": "value"}}}}
-        self.assertEqual(found, ["a", "b"])
+        self.assertEqual(deep_find_key(deep_payload, {"target"}, max_depth=2), [])
+        self.assertEqual(deep_find_key(deep_payload, {"target"}, max_depth=8), ["value"])
 
-    def test_respects_max_depth(self) -> None:
-        payload = {"l1": {"l2": {"l3": {"target": "value"}}}}
 
-        found_too_shallow = deep_find_key(payload, {"target"}, max_depth=2)
+class TestParseListingSegments(unittest.TestCase):
-        found_enough_depth = deep_find_key(payload, {"target"}, max_depth=8)
+    def test_empty_and_invalid_return_empty(self) -> None:
+        self.assertEqual(parse_listing_segments(""), [])
+        self.assertEqual(parse_listing_segments("  "), [])
+        self.assertEqual(parse_listing_segments("invalid"), [])
 
-        self.assertEqual(found_too_shallow, [])
+    def test_auto_segments_complete_coverage(self) -> None:
-        self.assertEqual(found_enough_depth, ["value"])
+        """auto: все бренды покрыты, крупные разбиты по годам, годы без дыр."""
+        segs = parse_listing_segments("auto")
+
+        # Точное число сегментов.
+        expected = len(_LARGE_MAKES) * len(_YEAR_SPLITS) + (len(IAAI_DEFAULT_MAKES) - len(_LARGE_MAKES))
+        self.assertEqual(len(segs), expected)
+
+        # Все бренды из списка присутствуют.
+        makes_in_segments = {s["make"] for s in segs}
+        for make in IAAI_DEFAULT_MAKES:
+            self.assertIn(make, makes_in_segments)
+
+        # Крупные бренды разбиты на 3 сегмента с годами.
+        toyota = [s for s in segs if s["make"] == "TOYOTA"]
+        self.assertEqual(len(toyota), 3)
+        for s in toyota:
+            self.assertIsNotNone(s["year_min"])
+
+        # Мелкие бренды без годов.
+        lexus = [s for s in segs if s["make"] == "LEXUS"]
+        self.assertEqual(len(lexus), 1)
+        self.assertIsNone(lexus[0]["year_min"])
+
+        # Годовые диапазоны покрывают 1950-2027.
+        years = set()
+        for yr_min, yr_max in _YEAR_SPLITS:
+            years.update(range(yr_min, yr_max + 1))
+        for year in range(1950, 2027):
+            self.assertIn(year, years)
+
+    def test_json_input_formats(self) -> None:
+        # Массив строк.
+        segs = parse_listing_segments('["toyota", "ford"]')
+        self.assertEqual(len(segs), 2)
+        self.assertEqual(segs[0]["make"], "TOYOTA")
+
+        # Массив объектов с годами.
+        segs = parse_listing_segments('[{"make":"BMW","year_min":2020,"year_max":2025}]')
+        self.assertEqual(segs[0]["make"], "BMW")
+        self.assertEqual(segs[0]["year_min"], 2020)
+        self.assertEqual(segs[0]["year_max"], 2025)
 
 
 if __name__ == "__main__":

tests/test_worker_tasks.py

@@ -0,0 +1,661 @@
+from __future__ import annotations
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from iaai_scraper.worker import tasks
+
+
+def make_settings_stub(*, parallel_segments: bool = False) -> MagicMock:
+    settings_stub = MagicMock()
+    settings_stub.celery.parallel_segments = parallel_segments
+    settings_stub.celery.task_soft_time_limit = 3300
+    settings_stub.celery.task_time_limit = 3600
+    settings_stub.celery.task_stall_timeout_seconds = 600
+    settings_stub.listing.listing_segments_json = "ignored"
+    settings_stub.discovery.mode = "listing"
+    settings_stub.discovery.hourly_mode = "rolling_refresh"
+    return settings_stub
+
+
+class TestWorkerTaskLockHelpers(unittest.TestCase):
+    def test_sync_listing_task_uses_hourly_sitemap_even_when_mode_is_not_sitemap(self) -> None:
+        settings_stub = make_settings_stub(parallel_segments=False)
+        settings_stub.discovery.mode = "listing"
+        settings_stub.discovery.hourly_mode = "rolling_refresh"
+
+        with patch.object(tasks, "_get_persistence") as get_persistence, \
+             patch.object(tasks, "_get_redis") as get_redis, \
+             patch.object(tasks, "Settings", return_value=settings_stub), \
+             patch.object(tasks, "_acquire_lock", return_value=True), \
+             patch.object(tasks, "_is_full_scan_done", return_value=True), \
+             patch.object(tasks, "_start_lock_heartbeat") as start_heartbeat, \
+             patch.object(tasks, "_release_lock_if_owner") as release_lock, \
+             patch.object(tasks, "_hourly_sitemap_rolling_refresh_sync", return_value={
+                 "status": "success",
+                 "run_id": None,
+                 "cars_upserted": 25,
+                 "cars_failed": 0,
+                 "images_upserted": 50,
+                 "skipped_existing": 100,
+                 "elapsed_seconds": None,
+                 "failures": [],
+                 "hourly_mode": "sitemap_rolling_refresh",
+                 "discovered_urls": 105,
+                 "new_urls": 5,
+                 "refresh_urls": 20,
+                 "sold_marked": 2,
+             }) as hourly_sync, \
+             patch.object(tasks.sync_listing_task, "update_state"):
+            get_persistence.return_value = MagicMock()
+            redis_client = MagicMock()
+            redis_client.get.return_value = None
+            get_redis.return_value = redis_client
+            start_heartbeat.return_value = (MagicMock(), MagicMock())
+
+            tasks.sync_listing_task.push_request(id="task-hourly-force-sitemap")
+            try:
+                result = tasks.sync_listing_task.run()
+            finally:
+                tasks.sync_listing_task.pop_request()
+
+        self.assertEqual(result["status"], "success")
+        self.assertEqual(result["hourly_mode"], "sitemap_rolling_refresh")
+        hourly_sync.assert_called_once()
+        release_lock.assert_called_once()
+
+    def test_sync_listing_task_uses_hourly_sitemap_rolling_refresh_after_bootstrap(self) -> None:
+        settings_stub = make_settings_stub(parallel_segments=False)
+        settings_stub.discovery.mode = "sitemap"
+        settings_stub.discovery.hourly_mode = "rolling_refresh"
+
+        with patch.object(tasks, "_get_persistence") as get_persistence, \
+             patch.object(tasks, "_get_redis") as get_redis, \
+             patch.object(tasks, "Settings", return_value=settings_stub), \
+             patch.object(tasks, "_acquire_lock", return_value=True), \
+             patch.object(tasks, "_is_full_scan_done", return_value=True), \
+             patch.object(tasks, "_start_lock_heartbeat") as start_heartbeat, \
+             patch.object(tasks, "_release_lock_if_owner") as release_lock, \
+             patch.object(tasks, "_hourly_sitemap_rolling_refresh_sync", return_value={
+                 "status": "success",
+                 "run_id": None,
+                 "cars_upserted": 25,
+                 "cars_failed": 0,
+                 "images_upserted": 50,
+                 "skipped_existing": 100,
+                 "elapsed_seconds": None,
+                 "failures": [],
+                 "hourly_mode": "sitemap_rolling_refresh",
+                 "discovered_urls": 105,
+                 "new_urls": 5,
+                 "refresh_urls": 20,
+                 "sold_marked": 2,
+             }) as hourly_sync, \
+             patch.object(tasks.sync_listing_task, "update_state"):
+            get_persistence.return_value = MagicMock()
+            redis_client = MagicMock()
+            redis_client.get.return_value = None
+            get_redis.return_value = redis_client
+            start_heartbeat.return_value = (MagicMock(), MagicMock())
+
+            tasks.sync_listing_task.push_request(id="task-hourly")
+            try:
+                result = tasks.sync_listing_task.run()
+            finally:
+                tasks.sync_listing_task.pop_request()
+
+        self.assertEqual(result["status"], "success")
+        self.assertEqual(result["hourly_mode"], "sitemap_rolling_refresh")
+        hourly_sync.assert_called_once()
+        release_lock.assert_called_once()
+
+    def test_sync_listing_task_can_use_hourly_sitemap_diff_when_configured(self) -> None:
+        settings_stub = make_settings_stub(parallel_segments=False)
+        settings_stub.discovery.mode = "sitemap"
+        settings_stub.discovery.hourly_mode = "diff"
+
+        with patch.object(tasks, "_get_persistence") as get_persistence, \
+             patch.object(tasks, "_get_redis") as get_redis, \
+             patch.object(tasks, "Settings", return_value=settings_stub), \
+             patch.object(tasks, "_acquire_lock", return_value=True), \
+             patch.object(tasks, "_is_full_scan_done", return_value=True), \
+             patch.object(tasks, "_start_lock_heartbeat") as start_heartbeat, \
+             patch.object(tasks, "_release_lock_if_owner") as release_lock, \
+             patch.object(tasks, "_hourly_sitemap_diff_sync", return_value={
+                 "status": "success",
+                 "run_id": None,
+                 "cars_upserted": 5,
+                 "cars_failed": 0,
+                 "images_upserted": 10,
+                 "skipped_existing": 100,
+                 "elapsed_seconds": None,
+                 "failures": [],
+                 "hourly_mode": "sitemap_diff",
+                 "discovered_urls": 105,
+                 "new_urls": 5,
+                 "sold_marked": 2,
+             }) as hourly_sync, \
+             patch.object(tasks.sync_listing_task, "update_state"):
+            get_persistence.return_value = MagicMock()
+            redis_client = MagicMock()
+            redis_client.get.return_value = None
+            get_redis.return_value = redis_client
+            start_heartbeat.return_value = (MagicMock(), MagicMock())
+
+            tasks.sync_listing_task.push_request(id="task-hourly-diff")
+            try:
+                result = tasks.sync_listing_task.run()
+            finally:
+                tasks.sync_listing_task.pop_request()
+
+        self.assertEqual(result["status"], "success")
+        self.assertEqual(result["hourly_mode"], "sitemap_diff")
+        hourly_sync.assert_called_once()
+        release_lock.assert_called_once()
+
+    def test_sync_listing_task_forces_sitemap_full_scan_in_bootstrap(self) -> None:
+        settings_stub = make_settings_stub(parallel_segments=False)
+        settings_stub.discovery.mode = "listing"
+
+        with patch.object(tasks, "_get_persistence") as get_persistence, \
+             patch.object(tasks, "_get_redis") as get_redis, \
+             patch.object(tasks, "Settings", return_value=settings_stub), \
+             patch.object(tasks, "_acquire_lock", return_value=True), \
+             patch.object(tasks, "_is_full_scan_done", return_value=False), \
+             patch.object(tasks, "_start_lock_heartbeat") as start_heartbeat, \
+             patch.object(tasks, "_release_lock_if_owner") as release_lock, \
+             patch.object(tasks, "_run_browser_job", side_effect=lambda fn: fn()), \
+             patch.object(tasks.sync_listing_task, "update_state"), \
+             patch("iaai_scraper.worker.tasks.parse_listing_segments", return_value=[{"make": "HONDA"}]):
+            get_persistence.return_value = MagicMock()
+            redis_client = MagicMock()
+            redis_client.get.return_value = None
+            get_redis.return_value = redis_client
+            start_heartbeat.return_value = (MagicMock(), MagicMock())
+
+            sync_listing_mock = MagicMock(return_value={
+                "run_id": 99,
+                "status": "success",
+                "full_scan_completed": True,
+                "cars_upserted": 10,
+                "cars_failed": 0,
+                "images_upserted": 20,
+                "skipped_existing": 0,
+                "elapsed_seconds": 1.0,
+                "failures": [],
+            })
+            sync_listing_segmented_mock = MagicMock(side_effect=AssertionError("segmented path should not be used"))
+            scraper = MagicMock()
+            scraper.sync_listing = sync_listing_mock
+            scraper.sync_listing_segmented = sync_listing_segmented_mock
+            scraper_ctx = MagicMock()
+            scraper_ctx.__enter__.return_value = scraper
+            scraper_ctx.__exit__.return_value = None
+
+            with patch.object(tasks, "IAAIScraper", return_value=scraper_ctx):
+                tasks.sync_listing_task.push_request(id="task-bootstrap-sitemap")
+                try:
+                    result = tasks.sync_listing_task.run()
+                finally:
+                    tasks.sync_listing_task.pop_request()
+
+        self.assertEqual(result["status"], "success")
+        sync_listing_mock.assert_called_once_with(
+            make=None,
+            model=None,
+            lane="iaai_cars",
+            limit=None,
+            only_new=False,
+        )
+        sync_listing_segmented_mock.assert_not_called()
+        release_lock.assert_called_once()
+
+    def test_lock_acquire_refresh_release(self) -> None:
+        redis_client = MagicMock()
+
+        # Acquire.
+        redis_client.set.return_value = True
+        self.assertTrue(tasks._acquire_lock(redis_client, "lock:key", "owner-token", 120))
+        redis_client.set.assert_called_once_with("lock:key", "owner-token", nx=True, ex=120)
+
+        # Refresh.
+        redis_client.eval.return_value = 1
+        self.assertTrue(tasks._refresh_lock_if_owner(redis_client, "lock:key", "owner-token", 120))
+
+        # Release.
+        redis_client.eval.reset_mock()
+        tasks._release_lock_if_owner(redis_client, "lock:key", "owner-token")
+        args = redis_client.eval.call_args[0]
+        self.assertEqual(args[2], "lock:key")
+        self.assertEqual(args[3], "owner-token")
+
+    def test_sync_listing_task_skips_when_lock_not_acquired(self) -> None:
+        with patch.object(tasks, "_get_persistence") as get_persistence, \
+             patch.object(tasks, "_get_redis") as get_redis, \
+             patch.object(tasks, "_acquire_lock", return_value=False):
+            persistence = MagicMock()
+            get_persistence.return_value = persistence
+            get_redis.return_value = MagicMock()
+
+            tasks.sync_listing_task.push_request(id="task-123")
+            try:
+                result = tasks.sync_listing_task.run()
+            finally:
+                tasks.sync_listing_task.pop_request()
+
+        persistence.create_tables.assert_called_once()
+        self.assertEqual(result["status"], "skipped")
+        self.assertEqual(result["reason"], "sync_already_running")
+
+    def test_sync_listing_task_releases_owned_lock(self) -> None:
+        with patch.object(tasks, "_get_persistence") as get_persistence, \
+             patch.object(tasks, "_get_redis") as get_redis, \
+             patch.object(tasks, "_acquire_lock", return_value=True), \
+             patch.object(tasks, "_is_full_scan_done", return_value=True), \
+             patch.object(tasks, "_start_lock_heartbeat") as start_heartbeat, \
+             patch.object(tasks, "_release_lock_if_owner") as release_lock, \
+             patch.object(tasks.sync_listing_task, "update_state"), \
+             patch.object(tasks, "_run_browser_job", return_value={
+                 "run_id": 7,
+                 "cars_upserted": 2,
+                 "cars_failed": 0,
+                 "images_upserted": 4,
+                 "skipped_existing": 1,
+                 "elapsed_seconds": 1.25,
+             }):
+            persistence = MagicMock()
+            get_persistence.return_value = persistence
+            redis_client = MagicMock()
+            redis_client.get.return_value = None
+            get_redis.return_value = redis_client
+            stop_event = MagicMock()
+            heartbeat_thread = MagicMock()
+            start_heartbeat.return_value = (stop_event, heartbeat_thread)
+
+            tasks.sync_listing_task.push_request(id="task-123")
+            try:
+                result = tasks.sync_listing_task.run(make="Toyota")
+            finally:
+                tasks.sync_listing_task.pop_request()
+
+        self.assertEqual(result["status"], "success")
+        stop_event.set.assert_called_once()
+        heartbeat_thread.join.assert_called_once()
+        release_lock.assert_called_once()
+
+    def test_clear_orphan_sync_listing_lock(self) -> None:
+        # Нет запущенных задач → удаляет.
+        redis_client = MagicMock()
+        redis_client.get.return_value = "owner-token"
+        redis_client.ttl.return_value = 120
+        celery_app = MagicMock()
+        inspector = MagicMock()
+        inspector.active.return_value = {"worker@node": []}
+        inspector.reserved.return_value = {"worker@node": []}
+        inspector.scheduled.return_value = {"worker@node": []}
+        celery_app.control.inspect.return_value = inspector
+
+        self.assertTrue(tasks._clear_orphan_sync_listing_lock(redis_client, celery_app))
+        redis_client.delete.assert_called_once_with(tasks.SYNC_LISTING_LOCK_KEY)
+
+        # Задача активна → не удаляет.
+        redis_client2 = MagicMock()
+        redis_client2.get.return_value = "owner-token"
+        inspector2 = MagicMock()
+        inspector2.active.return_value = {"worker@node": [{"name": tasks.SYNC_LISTING_TASK_NAME}]}
+        inspector2.reserved.return_value = {"worker@node": []}
+        inspector2.scheduled.return_value = {"worker@node": []}
+        celery_app2 = MagicMock()
+        celery_app2.control.inspect.return_value = inspector2
+
+        self.assertFalse(tasks._clear_orphan_sync_listing_lock(redis_client2, celery_app2))
+        redis_client2.delete.assert_not_called()
+
+    def test_sync_listing_task_recovers_orphan_lock_and_runs(self) -> None:
+        with patch.object(tasks, "_get_persistence") as get_persistence, \
+             patch.object(tasks, "_get_redis") as get_redis, \
+             patch.object(tasks, "_acquire_lock", side_effect=[False, True]) as acquire_lock, \
+             patch.object(tasks, "_clear_orphan_sync_listing_lock", return_value=True) as clear_orphan, \
+             patch.object(tasks, "_is_full_scan_done", return_value=True), \
+             patch.object(tasks, "_start_lock_heartbeat") as start_heartbeat, \
+             patch.object(tasks, "_release_lock_if_owner") as release_lock, \
+             patch.object(tasks.sync_listing_task, "update_state"), \
+             patch.object(tasks, "_run_browser_job", return_value={
+                 "run_id": 9,
+                 "cars_upserted": 3,
+                 "cars_failed": 0,
+                 "images_upserted": 5,
+                 "skipped_existing": 0,
+                 "elapsed_seconds": 2.0,
+             }):
+            persistence = MagicMock()
+            get_persistence.return_value = persistence
+            redis_client = MagicMock()
+            redis_client.get.return_value = None
+            get_redis.return_value = redis_client
+            stop_event = MagicMock()
+            heartbeat_thread = MagicMock()
+            start_heartbeat.return_value = (stop_event, heartbeat_thread)
+
+            tasks.sync_listing_task.push_request(id="task-456")
+            try:
+                result = tasks.sync_listing_task.run(make="Honda")
+            finally:
+                tasks.sync_listing_task.pop_request()
+
+        self.assertEqual(result["status"], "success")
+        clear_orphan.assert_called_once()
+        self.assertEqual(acquire_lock.call_count, 2)
+        release_lock.assert_called_once()
+
+    def test_checkpoint_save_load_roundtrip(self) -> None:
+        storage: dict[str, str] = {}
+
+        def _fake_set(key, value, ex=None):
+            storage[key] = value
+            return True
+
+        redis_client = MagicMock()
+        redis_client.set.side_effect = _fake_set
+        redis_client.get.side_effect = lambda key: storage.get(key)
+
+        tasks._save_last_completed_segment(redis_client, 12)
+
+        self.assertEqual(tasks._load_last_completed_segment(redis_client), 12)
+        self.assertEqual(redis_client.set.call_args.kwargs["ex"], tasks.SYNC_LISTING_CHECKPOINT_TTL_SECONDS)
+
+    def test_load_checkpoint_clears_invalid_payload(self) -> None:
+        redis_client = MagicMock()
+        redis_client.get.return_value = "{not-a-number"
+
+        self.assertIsNone(tasks._load_last_completed_segment(redis_client))
+        redis_client.delete.assert_called_once_with(tasks.SYNC_LISTING_CHECKPOINT_KEY)
+
+    def test_load_checkpoint_empty_when_missing(self) -> None:
+        redis_client = MagicMock()
+        redis_client.get.return_value = None
+
+        self.assertIsNone(tasks._load_last_completed_segment(redis_client))
+        redis_client.delete.assert_not_called()
+
+    def test_sync_listing_bootstrap_prefers_sitemap_over_segment_resume(self) -> None:
+        segments = [
+            {"make": "ACURA"},
+            {"make": "AUDI"},
+            {"make": "BMW"},
+            {"make": "EAGLE"},
+        ]
+        settings_stub = make_settings_stub()
+        with patch.object(tasks, "_get_persistence") as get_persistence, \
+             patch.object(tasks, "_get_redis") as get_redis, \
+             patch.object(tasks, "Settings", return_value=settings_stub), \
+             patch.object(tasks, "_acquire_lock", return_value=True), \
+             patch.object(tasks, "_is_full_scan_done", return_value=False), \
+             patch.object(tasks, "_start_lock_heartbeat") as start_heartbeat, \
+             patch.object(tasks, "_release_lock_if_owner") as release_lock, \
+             patch.object(tasks, "_run_browser_job", side_effect=lambda fn: fn()), \
+             patch.object(tasks.sync_listing_task, "update_state"), \
+             patch("iaai_scraper.worker.tasks.parse_listing_segments", return_value=segments):
+            get_persistence.return_value = MagicMock()
+            redis_client = MagicMock()
+            redis_client.get.side_effect = lambda key: (
+                "1" if key == tasks.SYNC_LISTING_CHECKPOINT_KEY else None
+            )
+            get_redis.return_value = redis_client
+            start_heartbeat.return_value = (MagicMock(), MagicMock())
+
+            sync_segmented_mock = MagicMock(side_effect=AssertionError("segmented path should not be used"))
+            sync_listing_mock = MagicMock(return_value={
+                "run_id": 11, "status": "success", "full_scan_completed": True,
+                "cars_upserted": 1, "cars_failed": 0, "images_upserted": 0,
+                "skipped_existing": 0, "elapsed_seconds": 1.0, "failures": [],
+            })
+            scraper_ctx = MagicMock()
+            scraper_ctx.__enter__.return_value.sync_listing_segmented = sync_segmented_mock
+            scraper_ctx.__enter__.return_value.sync_listing = sync_listing_mock
+            scraper_ctx.__exit__.return_value = None
+
+            with patch.object(tasks, "IAAIScraper", return_value=scraper_ctx):
+                tasks.sync_listing_task.push_request(id="task-resume-seg")
+                try:
+                    result = tasks.sync_listing_task.run()
+                finally:
+                    tasks.sync_listing_task.pop_request()
+
+        self.assertEqual(result["status"], "success")
+        sync_listing_mock.assert_called_once_with(
+            make=None,
+            model=None,
+            lane="iaai_cars",
+            limit=None,
+            only_new=False,
+        )
+        sync_segmented_mock.assert_not_called()
+        release_lock.assert_called_once()
+
+    def test_sync_listing_hourly_path_ignores_checkpoint_after_full_scan_completed(self) -> None:
+        segments = [{"make": "ACURA"}, {"make": "AUDI"}]
+        settings_stub = make_settings_stub()
+        with patch.object(tasks, "_get_persistence") as get_persistence, \
+             patch.object(tasks, "_get_redis") as get_redis, \
+             patch.object(tasks, "Settings", return_value=settings_stub), \
+             patch.object(tasks, "_acquire_lock", return_value=True), \
+             patch.object(tasks, "_is_full_scan_done", return_value=True), \
+             patch.object(tasks, "_start_lock_heartbeat") as start_heartbeat, \
+             patch.object(tasks, "_release_lock_if_owner") as release_lock, \
+             patch.object(tasks, "_clear_sync_checkpoint") as clear_checkpoint, \
+             patch.object(tasks, "_hourly_sitemap_rolling_refresh_sync", return_value={
+                 "status": "success",
+                 "run_id": None,
+                 "cars_upserted": 1,
+                 "cars_failed": 0,
+                 "images_upserted": 0,
+                 "skipped_existing": 0,
+                 "elapsed_seconds": None,
+                 "failures": [],
+                 "hourly_mode": "sitemap_rolling_refresh",
+                 "discovered_urls": 10,
+                 "new_urls": 1,
+                 "refresh_urls": 1,
+                 "sold_marked": 0,
+             }) as hourly_sync, \
+             patch.object(tasks.sync_listing_task, "update_state"), \
+             patch("iaai_scraper.worker.tasks.parse_listing_segments", return_value=segments):
+            get_persistence.return_value = MagicMock()
+            redis_client = MagicMock()
+            # Оставшийся чекпоинт не должен использоваться.
+            redis_client.get.side_effect = lambda key: (
+                "0" if key == tasks.SYNC_LISTING_CHECKPOINT_KEY else None
+            )
+            get_redis.return_value = redis_client
+            start_heartbeat.return_value = (MagicMock(), MagicMock())
+
+            tasks.sync_listing_task.push_request(id="task-792")
+            try:
+                result = tasks.sync_listing_task.run()
+            finally:
+                tasks.sync_listing_task.pop_request()
+
+        self.assertEqual(result["status"], "success")
+        self.assertEqual(result["hourly_mode"], "sitemap_rolling_refresh")
+        hourly_sync.assert_called_once()
+        clear_checkpoint.assert_called()
+        release_lock.assert_called_once()
+
+    def test_sync_listing_bootstrap_ignores_beyond_segment_checkpoint(self) -> None:
+        segments = [{"make": "ACURA"}, {"make": "AUDI"}]
+        settings_stub = make_settings_stub()
+        with patch.object(tasks, "_get_persistence") as get_persistence, \
+             patch.object(tasks, "_get_redis") as get_redis, \
+             patch.object(tasks, "Settings", return_value=settings_stub), \
+             patch.object(tasks, "_acquire_lock", return_value=True), \
+             patch.object(tasks, "_is_full_scan_done", return_value=False), \
+             patch.object(tasks, "_start_lock_heartbeat") as start_heartbeat, \
+             patch.object(tasks, "_release_lock_if_owner") as release_lock, \
+             patch.object(tasks, "_run_browser_job", side_effect=lambda fn: fn()), \
+             patch.object(tasks.sync_listing_task, "update_state"), \
+             patch("iaai_scraper.worker.tasks.parse_listing_segments", return_value=segments):
+            get_persistence.return_value = MagicMock()
+            redis_client = MagicMock()
+            redis_client.get.side_effect = lambda key: (
+                "99" if key == tasks.SYNC_LISTING_CHECKPOINT_KEY else None
+            )
+            get_redis.return_value = redis_client
+            start_heartbeat.return_value = (MagicMock(), MagicMock())
+
+            sync_segmented_mock = MagicMock(side_effect=AssertionError("segmented path should not be used"))
+            sync_listing_mock = MagicMock(return_value={
+                "run_id": 15, "status": "success", "full_scan_completed": True,
+                "cars_upserted": 0, "cars_failed": 0, "images_upserted": 0,
+                "skipped_existing": 0, "elapsed_seconds": 1.0, "failures": [],
+            })
+            scraper_ctx = MagicMock()
+            scraper_ctx.__enter__.return_value.sync_listing_segmented = sync_segmented_mock
+            scraper_ctx.__enter__.return_value.sync_listing = sync_listing_mock
+            scraper_ctx.__exit__.return_value = None
+
+            with patch.object(tasks, "IAAIScraper", return_value=scraper_ctx):
+                tasks.sync_listing_task.push_request(id="task-beyond")
+                try:
+                    result = tasks.sync_listing_task.run()
+                finally:
+                    tasks.sync_listing_task.pop_request()
+
+        self.assertEqual(result["status"], "success")
+        sync_listing_mock.assert_called_once_with(
+            make=None,
+            model=None,
+            lane="iaai_cars",
+            limit=None,
+            only_new=False,
+        )
+        sync_segmented_mock.assert_not_called()
+        release_lock.assert_called_once()
+
+    def test_sync_listing_task_clears_checkpoint_on_hourly_run(self) -> None:
+        with patch.object(tasks, "_get_persistence") as get_persistence, \
+             patch.object(tasks, "_get_redis") as get_redis, \
+             patch.object(tasks, "_acquire_lock", return_value=True), \
+             patch.object(tasks, "_is_full_scan_done", return_value=True), \
+             patch.object(tasks, "_start_lock_heartbeat") as start_heartbeat, \
+             patch.object(tasks, "_release_lock_if_owner") as release_lock, \
+             patch.object(tasks, "_clear_sync_checkpoint") as clear_checkpoint, \
+             patch.object(tasks, "_run_browser_job", return_value={
+                 "run_id": 13,
+                 "status": "partial_success",
+                 "full_scan_completed": True,
+                 "cars_upserted": 2,
+                 "cars_failed": 1,
+                 "images_upserted": 3,
+                 "skipped_existing": 0,
+                 "elapsed_seconds": 4.0,
+                 "failures": [{"vehicle_url": "v", "error": "e"}],
+             }), \
+             patch.object(tasks.sync_listing_task, "update_state"):
+            get_persistence.return_value = MagicMock()
+            redis_client = MagicMock()
+            redis_client.get.return_value = None
+            get_redis.return_value = redis_client
+            start_heartbeat.return_value = (MagicMock(), MagicMock())
+
+            tasks.sync_listing_task.push_request(id="task-791")
+            try:
+                result = tasks.sync_listing_task.run(make="Toyota")
+            finally:
+                tasks.sync_listing_task.pop_request()
+
+        self.assertEqual(result["status"], "partial_success")
+        # Hourly-ветка (full_scan_done=True) всегда удаляет оставшийся чекпоинт
+        # до браузерного job + после. Главное — вызов произошёл.
+        clear_checkpoint.assert_called()
+        release_lock.assert_called_once()
+
+    def test_try_set_followup_pending_deduplicates(self) -> None:
+        redis_client = MagicMock()
+        redis_client.set.side_effect = [True, False]
+
+        self.assertTrue(tasks._try_set_followup_pending(redis_client, ttl_seconds=120))
+        self.assertFalse(tasks._try_set_followup_pending(redis_client, ttl_seconds=120))
+
+    def test_bump_bootstrap_failure_streak_opens_circuit_breaker(self) -> None:
+        redis_client = MagicMock()
+        redis_client.incr.return_value = tasks.SYNC_LISTING_BOOTSTRAP_FAILURE_STREAK_LIMIT
+
+        streak, should_enqueue = tasks._bump_bootstrap_failure_streak(
+            redis_client,
+            reason="max_retries_exceeded",
+        )
+
+        self.assertEqual(streak, tasks.SYNC_LISTING_BOOTSTRAP_FAILURE_STREAK_LIMIT)
+        self.assertFalse(should_enqueue)
+        redis_client.expire.assert_called_once_with(
+            tasks.SYNC_LISTING_BOOTSTRAP_FAILURE_STREAK_KEY,
+            tasks.SYNC_LISTING_BOOTSTRAP_FAILURE_STREAK_TTL_SECONDS,
+        )
+
+    def test_bump_bootstrap_failure_streak_allows_retry_before_limit(self) -> None:
+        redis_client = MagicMock()
+        redis_client.incr.return_value = 1
+
+        streak, should_enqueue = tasks._bump_bootstrap_failure_streak(
+            redis_client,
+            reason="bootstrap_not_completed",
+        )
+
+        self.assertEqual(streak, 1)
+        self.assertTrue(should_enqueue)
+
+    def test_sync_listing_task_does_not_enqueue_followup_after_bootstrap_error_limit(self) -> None:
+        settings_stub = make_settings_stub(parallel_segments=False)
+        with patch.object(tasks, "_get_persistence") as get_persistence, \
+            patch.object(tasks, "_get_redis") as get_redis, \
+            patch.object(tasks, "Settings", return_value=settings_stub), \
+            patch.object(tasks, "_acquire_lock", return_value=True), \
+            patch.object(tasks, "_is_full_scan_done", return_value=False), \
+            patch.object(tasks, "_start_lock_heartbeat") as start_heartbeat, \
+            patch.object(tasks, "_release_lock_if_owner") as release_lock, \
+            patch.object(tasks, "_try_set_followup_pending", return_value=True), \
+            patch.object(
+                tasks,
+                "_bump_bootstrap_failure_streak",
+                return_value=(tasks.SYNC_LISTING_BOOTSTRAP_FAILURE_STREAK_LIMIT, False),
+            ) as bump_streak, \
+            patch.object(tasks, "_clear_followup_pending") as clear_pending, \
+            patch.object(tasks.sync_listing_task, "update_state"), \
+            patch.object(
+                tasks,
+                "_run_browser_job",
+                return_value={
+                    "run_id": 99,
+                    "status": "failed",
+                    "full_scan_completed": False,
+                    "cars_upserted": 0,
+                    "cars_failed": 0,
+                    "images_upserted": 0,
+                    "skipped_existing": 0,
+                    "elapsed_seconds": 1.0,
+                    "failures": [{"vehicle_url": "listing", "error": "bad resume"}],
+                    "listing": {"vehicles_collected": 0},
+                },
+            ):
+            get_persistence.return_value = MagicMock()
+            redis_client = MagicMock()
+            redis_client.get.return_value = None
+            get_redis.return_value = redis_client
+            start_heartbeat.return_value = (MagicMock(), MagicMock())
+
+            with patch.object(tasks.sync_listing_task, "app", new=MagicMock()) as task_app:
+                tasks.sync_listing_task.push_request(id="task-900")
+                try:
+                    result = tasks.sync_listing_task.run()
+                finally:
+                    tasks.sync_listing_task.pop_request()
+
+        self.assertEqual(result["status"], "failed")
+        bump_streak.assert_called_once()
+        clear_pending.assert_called()
+        task_app.send_task.assert_not_called()
+
+
+if __name__ == "__main__":
+    unittest.main()