Merge branch 'dev' into feat/sv-searching-txtai

:

.gitea/workflows/main.yml

@@ -29,7 +29,16 @@ jobs:
         if: github.ref == 'refs/heads/dev'
         uses: dokku/github-action@master
         with:
-          branch: 'dev'
+          branch: 'main'
           force: true
           git_remote_url: 'ssh://dokku@v2.discours.io:22/core'
           ssh_private_key: ${{ secrets.SSH_PRIVATE_KEY }}
+
+      - name: Push to dokku for staging branch
+        if: github.ref == 'refs/heads/staging'
+        uses: dokku/github-action@master
+        with:
+          branch: 'dev'
+          git_remote_url: 'ssh://dokku@staging.discours.io:22/core'
+          ssh_private_key: ${{ secrets.SSH_PRIVATE_KEY }}
+          git_push_flags: '--force'

.gitignore

@@ -128,6 +128,9 @@ dmypy.json
 .idea
 temp.*
 
+# Debug
+DEBUG.log
+
 discours.key
 discours.crt
 discours.pem
@@ -162,4 +165,4 @@ views.json
 *.crt
 *cache.json
 .cursor
-.devcontainer/devcontainer.json
+.devcontainer/

CHANGELOG.md

@@ -1,14 +1,3 @@
-#### [0.4.20] - 2025-05-03
-- Исправлена ошибка в классе `CacheRevalidationManager`: добавлена инициализация атрибута `_redis`
-- Улучшена обработка соединения с Redis в менеджере ревалидации кэша:
-  - Автоматическое восстановление соединения в случае его потери
-  - Проверка соединения перед выполнением операций с кэшем
-  - Дополнительное логирование для упрощения диагностики проблем
-- Исправлен резолвер `unpublish_shout`:
-  - Корректное формирование синтетического поля `publication` с `published_at: null`
-  - Возвращение полноценного словаря с данными вместо объекта модели
-  - Улучшена загрузка связанных данных (авторы, темы) для правильного формирования ответа
-
 #### [0.4.19] - 2025-04-14
 - dropped `Shout.description` and `Draft.description` to be UX-generated
 - use redis to init views counters after migrator

README.md

@@ -57,7 +57,7 @@ mkcert localhost
 Then, run the server:
 
 ```shell
-python -m granian main:app --interface asgi --host 0.0.0.0 --port 8000
+python server.py dev
 ```
 
 ### Useful Commands

auth/usermodel.py

@@ -0,0 +1,119 @@
+import time
+
+from sqlalchemy import (
+    JSON,
+    Boolean,
+    Column,
+    DateTime,
+    ForeignKey,
+    Integer,
+    String,
+    func,
+)
+from sqlalchemy.orm import relationship
+
+from services.db import Base
+
+
+class Permission(Base):
+    __tablename__ = "permission"
+
+    id = Column(String, primary_key=True, unique=True, nullable=False, default=None)
+    resource = Column(String, nullable=False)
+    operation = Column(String, nullable=False)
+
+
+class Role(Base):
+    __tablename__ = "role"
+
+    id = Column(String, primary_key=True, unique=True, nullable=False, default=None)
+    name = Column(String, nullable=False)
+    permissions = relationship(Permission)
+
+
+class AuthorizerUser(Base):
+    __tablename__ = "authorizer_users"
+
+    id = Column(String, primary_key=True, unique=True, nullable=False, default=None)
+    key = Column(String)
+    email = Column(String, unique=True)
+    email_verified_at = Column(Integer)
+    family_name = Column(String)
+    gender = Column(String)
+    given_name = Column(String)
+    is_multi_factor_auth_enabled = Column(Boolean)
+    middle_name = Column(String)
+    nickname = Column(String)
+    password = Column(String)
+    phone_number = Column(String, unique=True)
+    phone_number_verified_at = Column(Integer)
+    # preferred_username = Column(String, nullable=False)
+    picture = Column(String)
+    revoked_timestamp = Column(Integer)
+    roles = Column(String, default="author,reader")
+    signup_methods = Column(String, default="magic_link_login")
+    created_at = Column(Integer, default=lambda: int(time.time()))
+    updated_at = Column(Integer, default=lambda: int(time.time()))
+
+
+class UserRating(Base):
+    __tablename__ = "user_rating"
+
+    id = None
+    rater: Column = Column(ForeignKey("user.id"), primary_key=True, index=True)
+    user: Column = Column(ForeignKey("user.id"), primary_key=True, index=True)
+    value: Column = Column(Integer)
+
+    @staticmethod
+    def init_table():
+        pass
+
+
+class UserRole(Base):
+    __tablename__ = "user_role"
+
+    id = None
+    user = Column(ForeignKey("user.id"), primary_key=True, index=True)
+    role = Column(ForeignKey("role.id"), primary_key=True, index=True)
+
+
+class User(Base):
+    __tablename__ = "user"
+    default_user = None
+
+    email = Column(String, unique=True, nullable=False, comment="Email")
+    username = Column(String, nullable=False, comment="Login")
+    password = Column(String, nullable=True, comment="Password")
+    bio = Column(String, nullable=True, comment="Bio")  # status description
+    about = Column(String, nullable=True, comment="About")  # long and formatted
+    userpic = Column(String, nullable=True, comment="Userpic")
+    name = Column(String, nullable=True, comment="Display name")
+    slug = Column(String, unique=True, comment="User's slug")
+    links = Column(JSON, nullable=True, comment="Links")
+    oauth = Column(String, nullable=True)
+    oid = Column(String, nullable=True)
+
+    muted = Column(Boolean, default=False)
+    confirmed = Column(Boolean, default=False)
+
+    created_at = Column(DateTime(timezone=True), nullable=False, server_default=func.now(), comment="Created at")
+    updated_at = Column(DateTime(timezone=True), nullable=False, server_default=func.now(), comment="Updated at")
+    last_seen = Column(DateTime(timezone=True), nullable=False, server_default=func.now(), comment="Was online at")
+    deleted_at = Column(DateTime(timezone=True), nullable=True, comment="Deleted at")
+
+    ratings = relationship(UserRating, foreign_keys=UserRating.user)
+    roles = relationship(lambda: Role, secondary=UserRole.__tablename__)
+
+    def get_permission(self):
+        scope = {}
+        for role in self.roles:
+            for p in role.permissions:
+                if p.resource not in scope:
+                    scope[p.resource] = set()
+                scope[p.resource].add(p.operation)
+        print(scope)
+        return scope
+
+
+# if __name__ == "__main__":
+#   print(User.get_permission(user_id=1))

cache/revalidator.py

@@ -8,7 +8,6 @@ from cache.cache import (
     invalidate_cache_by_prefix,
 )
 from resolvers.stat import get_with_stat
-from services.redis import redis
 from utils.logger import root_logger as logger
 
 CACHE_REVALIDATION_INTERVAL = 300  # 5 minutes
@@ -22,19 +21,9 @@ class CacheRevalidationManager:
         self.lock = asyncio.Lock()
         self.running = True
         self.MAX_BATCH_SIZE = 10  # Максимальное количество элементов для поштучной обработки
-        self._redis = redis  # Добавлена инициализация _redis для доступа к Redis-клиенту
 
     async def start(self):
         """Запуск фонового воркера для ревалидации кэша."""
-        # Проверяем, что у нас есть соединение с Redis
-        if not self._redis._client:
-            logger.warning("Redis connection not established. Waiting for connection...")
-            try:
-                await self._redis.connect()
-                logger.info("Redis connection established for revalidation manager")
-            except Exception as e:
-                logger.error(f"Failed to connect to Redis: {e}")
-                
         self.task = asyncio.create_task(self.revalidate_cache())
 
     async def revalidate_cache(self):
@@ -50,10 +39,6 @@ class CacheRevalidationManager:
 
     async def process_revalidation(self):
         """Обновление кэша для всех сущностей, требующих ревалидации."""
-        # Проверяем соединение с Redis
-        if not self._redis._client:
-            return  # Выходим из метода, если не удалось подключиться
-                
         async with self.lock:
             # Ревалидация кэша авторов
             if self.items_to_revalidate["authors"]:

cache/triggers.py

@@ -88,11 +88,7 @@ def after_reaction_handler(mapper, connection, target):
         with local_session() as session:
             shout = (
                 session.query(Shout)
-                .filter(
-                    Shout.id == shout_id,
-                    Shout.published_at.is_not(None),
-                    Shout.deleted_at.is_(None),
-                )
+                .filter(Shout.id == shout_id, Shout.published_at.is_not(None), Shout.deleted_at.is_(None))
                 .first()
             )
 
@@ -112,27 +108,15 @@ def events_register():
 
     event.listen(AuthorFollower, "after_insert", after_follower_handler)
     event.listen(AuthorFollower, "after_update", after_follower_handler)
-    event.listen(
-        AuthorFollower,
-        "after_delete",
-        lambda *args: after_follower_handler(*args, is_delete=True),
-    )
+    event.listen(AuthorFollower, "after_delete", lambda *args: after_follower_handler(*args, is_delete=True))
 
     event.listen(TopicFollower, "after_insert", after_follower_handler)
     event.listen(TopicFollower, "after_update", after_follower_handler)
-    event.listen(
-        TopicFollower,
-        "after_delete",
-        lambda *args: after_follower_handler(*args, is_delete=True),
-    )
+    event.listen(TopicFollower, "after_delete", lambda *args: after_follower_handler(*args, is_delete=True))
 
     event.listen(ShoutReactionsFollower, "after_insert", after_follower_handler)
     event.listen(ShoutReactionsFollower, "after_update", after_follower_handler)
-    event.listen(
-        ShoutReactionsFollower,
-        "after_delete",
-        lambda *args: after_follower_handler(*args, is_delete=True),
-    )
+    event.listen(ShoutReactionsFollower, "after_delete", lambda *args: after_follower_handler(*args, is_delete=True))
 
     event.listen(Reaction, "after_update", mark_for_revalidation)
     event.listen(Author, "after_update", mark_for_revalidation)

docs/caching.md

@@ -147,32 +147,16 @@ await invalidate_topics_cache(456)
 
 ```python
 class CacheRevalidationManager:
-    def __init__(self, interval=CACHE_REVALIDATION_INTERVAL):
-        # ...
-        self._redis = redis  # Прямая ссылка на сервис Redis
-        
-    async def start(self):
-        # Проверка и установка соединения с Redis
-        # ...
-        
+    # ...
     async def process_revalidation(self):
-        # Обработка элементов для ревалидации
         # ...
-        
     def mark_for_revalidation(self, entity_id, entity_type):
-        # Добавляет сущность в очередь на ревалидацию
         # ...
 ```
 
 Менеджер ревалидации работает как асинхронный фоновый процесс, который периодически (по умолчанию каждые 5 минут) проверяет наличие сущностей для ревалидации.
 
-**Взаимодействие с Redis:**
-- CacheRevalidationManager хранит прямую ссылку на сервис Redis через атрибут `_redis`
-- При запуске проверяется наличие соединения с Redis и при необходимости устанавливается новое
-- Включена автоматическая проверка соединения перед каждой операцией ревалидации
-- Система самостоятельно восстанавливает соединение при его потере
-
-**Особенности реализации:**
+Особенности реализации:
 - Для авторов и тем используется поштучная ревалидация каждой записи
 - Для шаутов и реакций используется батчевая обработка, с порогом в 10 элементов
 - При достижении порога система переключается на инвалидацию коллекций вместо поштучной обработки

main.py

@@ -17,7 +17,8 @@ from cache.revalidator import revalidation_manager
 from services.exception import ExceptionHandlerMiddleware
 from services.redis import redis
 from services.schema import create_all_tables, resolvers
-from services.search import search_service
+#from services.search import search_service
+from services.search import search_service, initialize_search_index
 from services.viewed import ViewedStorage
 from services.webhook import WebhookEndpoint, create_webhook_endpoint
 from settings import DEV_SERVER_PID_FILE_NAME, MODE
@@ -34,24 +35,67 @@ async def start():
                 f.write(str(os.getpid()))
     print(f"[main] process started in {MODE} mode")
 
+async def check_search_service():
+    """Check if search service is available and log result"""
+    info = await search_service.info()
+    if info.get("status") in ["error", "unavailable"]:
+        print(f"[WARNING] Search service unavailable: {info.get('message', 'unknown reason')}")
+    else:
+        print(f"[INFO] Search service is available: {info}")
 
+
+# indexing DB data
+# async def indexing():
+#     from services.db import fetch_all_shouts
+#     all_shouts = await fetch_all_shouts()
+#     await initialize_search_index(all_shouts)
 async def lifespan(_app):
     try:
+        print("[lifespan] Starting application initialization")
         create_all_tables()
         await asyncio.gather(
             redis.connect(),
             precache_data(),
             ViewedStorage.init(),
             create_webhook_endpoint(),
-            search_service.info(),
+            check_search_service(),
             start(),
             revalidation_manager.start(),
         )
+        print("[lifespan] Basic initialization complete")
+
+        # Add a delay before starting the intensive search indexing
+        print("[lifespan] Waiting for system stabilization before search indexing...")
+        await asyncio.sleep(10)  # 10-second delay to let the system stabilize
+
+        # Start search indexing as a background task with lower priority
+        asyncio.create_task(initialize_search_index_background())
+
         yield
     finally:
+        print("[lifespan] Shutting down application services")
         tasks = [redis.disconnect(), ViewedStorage.stop(), revalidation_manager.stop()]
         await asyncio.gather(*tasks, return_exceptions=True)
+        print("[lifespan] Shutdown complete")
 
+# Initialize search index in the background
+async def initialize_search_index_background():
+    """Run search indexing as a background task with low priority"""
+    try:
+        print("[search] Starting background search indexing process")
+        from services.db import fetch_all_shouts
+        
+        # Get total count first (optional)
+        all_shouts = await fetch_all_shouts()
+        total_count = len(all_shouts) if all_shouts else 0
+        print(f"[search] Fetched {total_count} shouts for background indexing")
+        
+        # Start the indexing process with the fetched shouts
+        print("[search] Beginning background search index initialization...")
+        await initialize_search_index(all_shouts)
+        print("[search] Background search index initialization complete")
+    except Exception as e:
+        print(f"[search] Error in background search indexing: {str(e)}")
 
 # Создаем экземпляр GraphQL
 graphql_app = GraphQL(schema, debug=True)

orm/author.py

@@ -90,6 +90,7 @@ class Author(Base):
     Модель автора в системе.
 
     Attributes:
+        user (str): Идентификатор пользователя в системе авторизации
         name (str): Отображаемое имя
         slug (str): Уникальный строковый идентификатор
         bio (str): Краткая биография/статус
@@ -104,6 +105,8 @@ class Author(Base):
 
     __tablename__ = "author"
 
+    user = Column(String)  # unbounded link with authorizer's User type
+
     name = Column(String, nullable=True, comment="Display name")
     slug = Column(String, unique=True, comment="Author's slug")
     bio = Column(String, nullable=True, comment="Bio")  # status description
@@ -121,14 +124,12 @@ class Author(Base):
 
     # Определяем индексы
     __table_args__ = (
-        # Индекс для быстрого поиска по имени
-        Index("idx_author_name", "name"),
         # Индекс для быстрого поиска по slug
         Index("idx_author_slug", "slug"),
+        # Индекс для быстрого поиска по идентификатору пользователя
+        Index("idx_author_user", "user"),
         # Индекс для фильтрации неудаленных авторов
-        Index(
-            "idx_author_deleted_at", "deleted_at", postgresql_where=deleted_at.is_(None)
-        ),
+        Index("idx_author_deleted_at", "deleted_at", postgresql_where=deleted_at.is_(None)),
         # Индекс для сортировки по времени создания (для новых авторов)
         Index("idx_author_created_at", "created_at"),
         # Индекс для сортировки по времени последнего посещения

orm/draft.py

@@ -26,14 +26,12 @@ class DraftAuthor(Base):
     caption = Column(String, nullable=True, default="")
 
 
-
 class Draft(Base):
     __tablename__ = "draft"
     # required
     created_at: int = Column(Integer, nullable=False, default=lambda: int(time.time()))
-    # Колонки для связей с автором
-    created_by: int = Column("created_by", ForeignKey("author.id"), nullable=False)
-    community: int = Column("community", ForeignKey("community.id"), nullable=False, default=1)
+    created_by: int = Column(ForeignKey("author.id"), nullable=False)
+    community: int = Column(ForeignKey("community.id"), nullable=False, default=1)
 
     # optional
     layout: str = Column(String, nullable=True, default="article")
@@ -51,55 +49,7 @@ class Draft(Base):
     # auto
     updated_at: int | None = Column(Integer, nullable=True, index=True)
     deleted_at: int | None = Column(Integer, nullable=True, index=True)
-    updated_by: int | None = Column("updated_by", ForeignKey("author.id"), nullable=True)
-    deleted_by: int | None = Column("deleted_by", ForeignKey("author.id"), nullable=True)
-    
-    # --- Relationships --- 
-    # Только many-to-many связи через вспомогательные таблицы
-    authors = relationship(Author, secondary="draft_author", lazy="select")
-    topics = relationship(Topic, secondary="draft_topic", lazy="select")
-    
-    # Связь с Community (если нужна как объект, а не ID)
-    # community = relationship("Community", foreign_keys=[community_id], lazy="joined")
-    # Пока оставляем community_id как ID
-
-    # Связь с публикацией (один-к-одному или один-к-нулю)
-    # Загружается через joinedload в резолвере
-    publication = relationship(
-        "Shout", 
-        primaryjoin="Draft.id == Shout.draft", 
-        foreign_keys="Shout.draft", 
-        uselist=False, 
-        lazy="noload", # Не грузим по умолчанию, только через options
-        viewonly=True # Указываем, что это связь только для чтения
-    )
-
-    def dict(self):
-        """
-        Сериализует объект Draft в словарь.
-        Гарантирует, что поля topics и authors всегда будут списками.
-        """
-        return {
-            "id": self.id,
-            "created_at": self.created_at,
-            "created_by": self.created_by,
-            "community": self.community,
-            "layout": self.layout,
-            "slug": self.slug,
-            "title": self.title,
-            "subtitle": self.subtitle,
-            "lead": self.lead,
-            "body": self.body,
-            "media": self.media or [],
-            "cover": self.cover,
-            "cover_caption": self.cover_caption,
-            "lang": self.lang,
-            "seo": self.seo,
-            "updated_at": self.updated_at,
-            "deleted_at": self.deleted_at,
-            "updated_by": self.updated_by,
-            "deleted_by": self.deleted_by,
-            # Гарантируем, что topics и authors всегда будут списками
-            "topics": [topic.dict() for topic in (self.topics or [])],
-            "authors": [author.dict() for author in (self.authors or [])]
-        }
+    updated_by: int | None = Column(ForeignKey("author.id"), nullable=True)
+    deleted_by: int | None = Column(ForeignKey("author.id"), nullable=True)
+    authors = relationship(Author, secondary="draft_author")
+    topics = relationship(Topic, secondary="draft_topic")

pyproject.toml

@@ -1,2 +0,0 @@
-[tool.ruff]
-line-length = 108

requirements.txt

@@ -13,6 +13,10 @@ starlette
 gql
 ariadne
 granian
+
+# NLP and search
+httpx
+
 orjson
 pydantic
 trafilatura

resolvers/__init__.py

@@ -16,11 +16,9 @@ from resolvers.draft import (
     delete_draft,
     load_drafts,
     publish_draft,
+    unpublish_draft,
     update_draft,
 )
-from resolvers.editor import (
-    unpublish_shout,
-)
 from resolvers.feed import (
     load_shouts_coauthored,
     load_shouts_discussed,

resolvers/author.py

@@ -77,45 +77,18 @@ async def get_authors_with_stats(limit=50, offset=0, by: Optional[str] = None):
             base_query = select(Author).where(Author.deleted_at.is_(None))
 
             # Применяем сортировку
-            
-            # vars for statistics sorting
-            stats_sort_field = None
-            stats_sort_direction = "desc"
-            
             if by:
                 if isinstance(by, dict):
-                    logger.debug(f"Processing dict-based sorting: {by}")
                     # Обработка словаря параметров сортировки
-                    from sqlalchemy import asc, desc, func
-                    from orm.shout import ShoutAuthor
-                    from orm.author import AuthorFollower
+                    from sqlalchemy import asc, desc
 
-                    # Checking for order field in the dictionary
-                    if "order" in by:
-                        order_value = by["order"]
-                        logger.debug(f"Found order field with value: {order_value}")
-                        if order_value in ["shouts", "followers", "rating", "comments"]:
-                            stats_sort_field = order_value
-                            stats_sort_direction = "desc"  # По умолчанию убывающая сортировка для статистики
-                            logger.debug(f"Applying statistics-based sorting by: {stats_sort_field}")
-                        elif order_value == "name":
-                            # Sorting by name in ascending order
-                            base_query = base_query.order_by(asc(Author.name))
-                            logger.debug("Applying alphabetical sorting by name")
-                        else:
-                            # If order is not a stats field, treat it as a regular field
-                            column = getattr(Author, order_value, None)
-                            if column:
+                    for field, direction in by.items():
+                        column = getattr(Author, field, None)
+                        if column:
+                            if direction.lower() == "desc":
                                 base_query = base_query.order_by(desc(column))
-                    else:
-                        # Regular sorting by fields
-                        for field, direction in by.items():
-                            column = getattr(Author, field, None)
-                            if column:
-                                if direction.lower() == "desc":
-                                    base_query = base_query.order_by(desc(column))
-                                else:
-                                    base_query = base_query.order_by(column)
+                            else:
+                                base_query = base_query.order_by(column)
                 elif by == "new":
                     base_query = base_query.order_by(desc(Author.created_at))
                 elif by == "active":
@@ -126,55 +99,6 @@ async def get_authors_with_stats(limit=50, offset=0, by: Optional[str] = None):
             else:
                 base_query = base_query.order_by(desc(Author.created_at))
 
-            # If sorting by statistics, modify the query
-            if stats_sort_field == "shouts":
-                # Sorting by the number of shouts
-                from sqlalchemy import func, and_
-                from orm.shout import Shout, ShoutAuthor
-                
-                subquery = (
-                    select(
-                        ShoutAuthor.author,
-                        func.count(func.distinct(Shout.id)).label("shouts_count")
-                    )
-                    .select_from(ShoutAuthor)
-                    .join(Shout, ShoutAuthor.shout == Shout.id)
-                    .where(
-                        and_(
-                            Shout.deleted_at.is_(None),
-                            Shout.published_at.is_not(None)
-                        )
-                    )
-                    .group_by(ShoutAuthor.author)
-                    .subquery()
-                )
-                
-                base_query = (
-                    base_query
-                    .outerjoin(subquery, Author.id == subquery.c.author)
-                    .order_by(desc(func.coalesce(subquery.c.shouts_count, 0)))
-                )
-            elif stats_sort_field == "followers":
-                # Sorting by the number of followers
-                from sqlalchemy import func
-                from orm.author import AuthorFollower
-                
-                subquery = (
-                    select(
-                        AuthorFollower.author,
-                        func.count(func.distinct(AuthorFollower.follower)).label("followers_count")
-                    )
-                    .select_from(AuthorFollower)
-                    .group_by(AuthorFollower.author)
-                    .subquery()
-                )
-                
-                base_query = (
-                    base_query
-                    .outerjoin(subquery, Author.id == subquery.c.author)
-                    .order_by(desc(func.coalesce(subquery.c.followers_count, 0)))
-                )
-
             # Применяем лимит и смещение
             base_query = base_query.limit(limit).offset(offset)
 

resolvers/draft.py

@@ -1,6 +1,8 @@
 import time
+from operator import or_
+
 import trafilatura
-from sqlalchemy.orm import joinedload
+from sqlalchemy.sql import and_
 
 from cache.cache import (
     cache_author,
@@ -10,7 +12,7 @@ from cache.cache import (
     invalidate_shouts_cache,
 )
 from orm.author import Author
-from orm.draft import Draft, DraftAuthor, DraftTopic
+from orm.draft import Draft
 from orm.shout import Shout, ShoutAuthor, ShoutTopic
 from orm.topic import Topic
 from services.auth import login_required
@@ -18,70 +20,34 @@ from services.db import local_session
 from services.notify import notify_shout
 from services.schema import mutation, query
 from services.search import search_service
-from utils.html_wrapper import wrap_html_fragment
 from utils.logger import root_logger as logger
 
+
 def create_shout_from_draft(session, draft, author_id):
-    """
-    Создаёт новый объект публикации (Shout) на основе черновика.
-
-    Args:
-        session: SQLAlchemy сессия (не используется, для совместимости)
-        draft (Draft): Объект черновика
-        author_id (int): ID автора публикации
-
-    Returns:
-        Shout: Новый объект публикации (не сохранённый в базе)
-
-    Пример:
-        >>> from orm.draft import Draft
-        >>> draft = Draft(id=1, title='Заголовок', body='Текст', slug='slug', created_by=1)
-        >>> shout = create_shout_from_draft(None, draft, 1)
-        >>> shout.title
-        'Заголовок'
-        >>> shout.body
-        'Текст'
-        >>> shout.created_by
-        1
-    """
     # Создаем новую публикацию
     shout = Shout(
-        body=draft.body or "",
+        body=draft.body,
         slug=draft.slug,
         cover=draft.cover,
         cover_caption=draft.cover_caption,
         lead=draft.lead,
-        title=draft.title or "",
+        title=draft.title,
         subtitle=draft.subtitle,
-        layout=draft.layout or "article",
-        media=draft.media or [],
-        lang=draft.lang or "ru",
+        layout=draft.layout,
+        media=draft.media,
+        lang=draft.lang,
         seo=draft.seo,
         created_by=author_id,
         community=draft.community,
         draft=draft.id,
         deleted_at=None,
     )
-    
-    # Инициализируем пустые массивы для связей
-    shout.topics = []
-    shout.authors = []
-    
     return shout
 
 
 @query.field("load_drafts")
 @login_required
 async def load_drafts(_, info):
-    """
-    Загружает все черновики, доступные текущему пользователю.
-    
-    Предварительно загружает связанные объекты (topics, authors, publication),
-    чтобы избежать ошибок с отсоединенными объектами при сериализации.
-    
-    Returns:
-        dict: Список черновиков или сообщение об ошибке
-    """
     user_id = info.context.get("user_id")
     author_dict = info.context.get("author", {})
     author_id = author_dict.get("id")
@@ -89,44 +55,13 @@ async def load_drafts(_, info):
     if not user_id or not author_id:
         return {"error": "User ID and author ID are required"}
 
-    try:
-        with local_session() as session:
-            # Предзагружаем authors, topics и связанную publication
-            drafts_query = (
-                session.query(Draft)
-                .options(
-                    joinedload(Draft.topics),
-                    joinedload(Draft.authors),
-                    joinedload(Draft.publication) # Загружаем связанную публикацию
-                )
-                .filter(Draft.authors.any(Author.id == author_id))
-            )
-            drafts = drafts_query.all()
-            
-            # Преобразуем объекты в словари, пока они в контексте сессии
-            drafts_data = []
-            for draft in drafts:
-                draft_dict = draft.dict()
-                # Всегда возвращаем массив для topics, даже если он пустой
-                draft_dict["topics"] = [topic.dict() for topic in (draft.topics or [])]
-                draft_dict["authors"] = [author.dict() for author in (draft.authors or [])]
-                
-                # Добавляем информацию о публикации, если она есть
-                if draft.publication:
-                    draft_dict["publication"] = {
-                        "id": draft.publication.id,
-                        "slug": draft.publication.slug,
-                        "published_at": draft.publication.published_at
-                    }
-                else:
-                     draft_dict["publication"] = None
-                     
-                drafts_data.append(draft_dict)
-                
-            return {"drafts": drafts_data}
-    except Exception as e:
-        logger.error(f"Failed to load drafts: {e}", exc_info=True)
-        return {"error": f"Failed to load drafts: {str(e)}"}
+    with local_session() as session:
+        drafts = (
+            session.query(Draft)
+            .filter(or_(Draft.authors.any(Author.id == author_id), Draft.created_by == author_id))
+            .all()
+        )
+    return {"drafts": drafts}
 
 
 @mutation.field("create_draft")
@@ -180,30 +115,22 @@ async def create_draft(_, info, draft_input):
             # Remove id from input if present since it's auto-generated
             if "id" in draft_input:
                 del draft_input["id"]
-                
-            # Добавляем текущее время создания и ID автора
+
+            if "seo" not in draft_input and not draft_input["seo"]:
+                body_teaser = draft_input.get("body", "")[:300].split("\n")[:-1].join("\n")
+                draft_input["seo"] = draft_input.get("lead", body_teaser)
+
+            # Добавляем текущее время создания
             draft_input["created_at"] = int(time.time())
-            draft_input["created_by"] = author_id
-            draft = Draft(**draft_input)
+
+            draft = Draft(created_by=author_id, **draft_input)
             session.add(draft)
-            session.flush()
-            
-            # Добавляем создателя как автора
-            da = DraftAuthor(shout=draft.id, author=author_id)
-            session.add(da)
-            
             session.commit()
             return {"draft": draft}
     except Exception as e:
         logger.error(f"Failed to create draft: {e}", exc_info=True)
         return {"error": f"Failed to create draft: {str(e)}"}
 
-def generate_teaser(body, limit=300):
-    body_html = wrap_html_fragment(body)
-    body_text = trafilatura.extract(body_html, include_comments=False, include_tables=False)
-    body_teaser = ". ".join(body_text[:limit].split(". ")[:-1])
-    return body_teaser
-
 
 @mutation.field("update_draft")
 @login_required
@@ -212,21 +139,7 @@ async def update_draft(_, info, draft_id: int, draft_input):
 
     Args:
         draft_id: ID черновика для обновления
-        draft_input: Данные для обновления черновика согласно схеме DraftInput:
-            - layout: String
-            - author_ids: [Int!]
-            - topic_ids: [Int!]
-            - main_topic_id: Int
-            - media: [MediaItemInput]
-            - lead: String
-            - subtitle: String
-            - lang: String
-            - seo: String
-            - body: String
-            - title: String
-            - slug: String
-            - cover: String
-            - cover_caption: String
+        draft_input: Данные для обновления черновика
 
     Returns:
         dict: Обновленный черновик или сообщение об ошибке
@@ -238,89 +151,36 @@ async def update_draft(_, info, draft_id: int, draft_input):
     if not user_id or not author_id:
         return {"error": "Author ID are required"}
 
-    try:
-        with local_session() as session:
-            draft = session.query(Draft).filter(Draft.id == draft_id).first()
-            if not draft:
-                return {"error": "Draft not found"}
+    # Проверяем slug - он должен быть или не пустым, или не передаваться вообще
+    if "slug" in draft_input and (draft_input["slug"] is None or draft_input["slug"] == ""):
+        # Если slug пустой, либо удаляем его из входных данных, либо генерируем временный уникальный
+        # Вариант 1: просто удаляем ключ из входных данных, чтобы оставить старое значение
+        del draft_input["slug"]
+        # Вариант 2 (если нужно обновить): генерируем временный уникальный slug
+        # import uuid
+        # draft_input["slug"] = f"draft-{uuid.uuid4().hex[:8]}"
 
-            # Фильтруем входные данные, оставляя только разрешенные поля
-            allowed_fields = {
-                "layout", "author_ids", "topic_ids", "main_topic_id", 
-                "media", "lead", "subtitle", "lang", "seo", "body", 
-                "title", "slug", "cover", "cover_caption"
-            }
-            filtered_input = {k: v for k, v in draft_input.items() if k in allowed_fields}
+    with local_session() as session:
+        draft = session.query(Draft).filter(Draft.id == draft_id).first()
+        if not draft:
+            return {"error": "Draft not found"}
 
-            # Проверяем slug
-            if "slug" in filtered_input and not filtered_input["slug"]:
-                del filtered_input["slug"]
+        if "seo" not in draft_input and not draft.seo:
+            body_src = draft_input["body"] if "body" in draft_input else draft.body
+            body_text = trafilatura.extract(body_src)
+            lead_src = draft_input["lead"] if "lead" in draft_input else draft.lead
+            lead_text = trafilatura.extract(lead_src)
+            body_teaser = body_text[:300].split(". ")[:-1].join(".\n")
+            draft_input["seo"] = lead_text or body_teaser
 
-            # Обновляем связи с авторами если переданы
-            if "author_ids" in filtered_input:
-                author_ids = filtered_input.pop("author_ids")
-                if author_ids:
-                    # Очищаем текущие связи
-                    session.query(DraftAuthor).filter(DraftAuthor.shout == draft_id).delete()
-                    # Добавляем новые связи
-                    for aid in author_ids:
-                        da = DraftAuthor(shout=draft_id, author=aid)
-                        session.add(da)
+        Draft.update(draft, draft_input)
+        # Set updated_at and updated_by from the authenticated user
+        current_time = int(time.time())
+        draft.updated_at = current_time
+        draft.updated_by = author_id
 
-            # Обновляем связи с темами если переданы
-            if "topic_ids" in filtered_input:
-                topic_ids = filtered_input.pop("topic_ids")
-                main_topic_id = filtered_input.pop("main_topic_id", None)
-                if topic_ids:
-                    # Очищаем текущие связи
-                    session.query(DraftTopic).filter(DraftTopic.shout == draft_id).delete()
-                    # Добавляем новые связи
-                    for tid in topic_ids:
-                        dt = DraftTopic(
-                            shout=draft_id, 
-                            topic=tid,
-                            main=(tid == main_topic_id) if main_topic_id else False
-                        )
-                        session.add(dt)
-
-            # Генерируем SEO если не предоставлено
-            if "seo" not in filtered_input and not draft.seo:
-                body_src = filtered_input.get("body", draft.body)
-                lead_src = filtered_input.get("lead", draft.lead)
-                body_html = wrap_html_fragment(body_src)
-                lead_html = wrap_html_fragment(lead_src)
-                
-                try:
-                    body_text = trafilatura.extract(body_html, include_comments=False, include_tables=False) if body_src else None
-                    lead_text = trafilatura.extract(lead_html, include_comments=False, include_tables=False) if lead_src else None
-                    
-                    body_teaser = generate_teaser(body_text, 300) if body_text else ""
-                    filtered_input["seo"] = lead_text if lead_text else body_teaser
-                except Exception as e:
-                    logger.warning(f"Failed to generate SEO for draft {draft_id}: {e}")
-
-            # Обновляем основные поля черновика
-            for key, value in filtered_input.items():
-                setattr(draft, key, value)
-
-            # Обновляем метаданные
-            draft.updated_at = int(time.time())
-            draft.updated_by = author_id
-
-            session.commit()
-            
-            # Преобразуем объект в словарь для ответа
-            draft_dict = draft.dict()
-            draft_dict["topics"] = [topic.dict() for topic in draft.topics]
-            draft_dict["authors"] = [author.dict() for author in draft.authors]
-            # Добавляем объект автора в updated_by
-            draft_dict["updated_by"] = author_dict
-            
-            return {"draft": draft_dict}
-
-    except Exception as e:
-        logger.error(f"Failed to update draft: {e}", exc_info=True)
-        return {"error": f"Failed to update draft: {str(e)}"}
+        session.commit()
+        return {"draft": draft}
 
 
 @mutation.field("delete_draft")
@@ -340,136 +200,182 @@ async def delete_draft(_, info, draft_id: int):
         return {"draft": draft}
 
 
-def validate_html_content(html_content: str) -> tuple[bool, str]:
-    """
-    Проверяет валидность HTML контента через trafilatura.
-    
-    Args:
-        html_content: HTML строка для проверки
-        
-    Returns:
-        tuple[bool, str]: (валидность, сообщение об ошибке)
-        
-    Example:
-        >>> is_valid, error = validate_html_content("<p>Valid HTML</p>")
-        >>> is_valid
-        True
-        >>> error
-        ''
-        >>> is_valid, error = validate_html_content("Invalid < HTML")
-        >>> is_valid
-        False
-        >>> 'Invalid HTML' in error
-        True
-    """
-    if not html_content or not html_content.strip():
-        return False, "Content is empty"
-        
-    try:
-        html_content = wrap_html_fragment(html_content)
-        extracted = trafilatura.extract(html_content)
-        if not extracted:
-            return False, "Invalid HTML structure or empty content"
-        return True, ""
-    except Exception as e:
-        logger.error(f"HTML validation error: {e}", exc_info=True)
-        return False, f"Invalid HTML content: {str(e)}"
-
-
 @mutation.field("publish_draft")
 @login_required
 async def publish_draft(_, info, draft_id: int):
-    """
-    Публикует черновик, создавая новый Shout или обновляя существующий.
-    
+    user_id = info.context.get("user_id")
+    author_dict = info.context.get("author", {})
+    author_id = author_dict.get("id")
+    if not user_id or not author_id:
+        return {"error": "User ID and author ID are required"}
+
+    with local_session() as session:
+        draft = session.query(Draft).filter(Draft.id == draft_id).first()
+        if not draft:
+            return {"error": "Draft not found"}
+        shout = create_shout_from_draft(session, draft, author_id)
+        session.add(shout)
+        session.commit()
+        return {"shout": shout, "draft": draft}
+
+
+@mutation.field("unpublish_draft")
+@login_required
+async def unpublish_draft(_, info, draft_id: int):
+    user_id = info.context.get("user_id")
+    author_dict = info.context.get("author", {})
+    author_id = author_dict.get("id")
+    if not user_id or not author_id:
+        return {"error": "User ID and author ID are required"}
+
+    with local_session() as session:
+        draft = session.query(Draft).filter(Draft.id == draft_id).first()
+        if not draft:
+            return {"error": "Draft not found"}
+        shout = session.query(Shout).filter(Shout.draft == draft.id).first()
+        if shout:
+            shout.published_at = None
+            session.commit()
+            return {"shout": shout, "draft": draft}
+        return {"error": "Failed to unpublish draft"}
+
+
+@mutation.field("publish_shout")
+@login_required
+async def publish_shout(_, info, shout_id: int):
+    """Publish draft as a shout or update existing shout.
+
     Args:
-        draft_id (int): ID черновика для публикации
-        
-    Returns:
-        dict: Результат публикации с shout или сообщением об ошибке
+        shout_id: ID существующей публикации или 0 для новой
+        draft: Объект черновика (опционально)
     """
     user_id = info.context.get("user_id")
     author_dict = info.context.get("author", {})
     author_id = author_dict.get("id")
-
+    now = int(time.time())
     if not user_id or not author_id:
-        return {"error": "Author ID is required"}
+        return {"error": "User ID and author ID are required"}
 
     try:
         with local_session() as session:
-            # Загружаем черновик со всеми связями
-            draft = (
-                session.query(Draft)
-                .options(
-                    joinedload(Draft.topics),
-                    joinedload(Draft.authors),
-                    joinedload(Draft.publication)
-                )
-                .filter(Draft.id == draft_id)
-                .first()
-            )
-
+            shout = session.query(Shout).filter(Shout.id == shout_id).first()
+            if not shout:
+                return {"error": "Shout not found"}
+            was_published = shout.published_at is not None
+            draft = session.query(Draft).where(Draft.id == shout.draft).first()
             if not draft:
                 return {"error": "Draft not found"}
+            # Находим черновик если не передан
 
-            # Проверка валидности HTML в body
-            is_valid, error = validate_html_content(draft.body)
-            if not is_valid:
-                return {"error": f"Cannot publish draft: {error}"}
-
-            # Проверяем, есть ли уже публикация для этого черновика
-            if draft.publication:
-                shout = draft.publication
-                # Обновляем существующую публикацию
-                for field in ["body", "title", "subtitle", "lead", "cover", "cover_caption", "media", "lang", "seo"]:
-                    if hasattr(draft, field):
-                        setattr(shout, field, getattr(draft, field))
-                shout.updated_at = int(time.time())
-                shout.updated_by = author_id
-            else:
-                # Создаем новую публикацию
+            if not shout:
                 shout = create_shout_from_draft(session, draft, author_id)
-                now = int(time.time())
-                shout.created_at = now
-                shout.published_at = now
-                session.add(shout)
-                session.flush()  # Получаем ID нового шаута
+            else:
+                # Обновляем существующую публикацию
+                shout.draft = draft.id
+                shout.created_by = author_id
+                shout.title = draft.title
+                shout.subtitle = draft.subtitle
+                shout.body = draft.body
+                shout.cover = draft.cover
+                shout.cover_caption = draft.cover_caption
+                shout.lead = draft.lead
+                shout.layout = draft.layout
+                shout.media = draft.media
+                shout.lang = draft.lang
+                shout.seo = draft.seo
 
-            # Очищаем существующие связи
-            session.query(ShoutAuthor).filter(ShoutAuthor.shout == shout.id).delete()
-            session.query(ShoutTopic).filter(ShoutTopic.shout == shout.id).delete()
+                draft.updated_at = now
+                shout.updated_at = now
 
-            # Добавляем авторов
-            for author in (draft.authors or []):
-                sa = ShoutAuthor(shout=shout.id, author=author.id)
+                # Устанавливаем published_at только если была ранее снята с публикации
+                if not was_published:
+                    shout.published_at = now
+
+            # Обрабатываем связи с авторами
+            if (
+                not session.query(ShoutAuthor)
+                .filter(and_(ShoutAuthor.shout == shout.id, ShoutAuthor.author == author_id))
+                .first()
+            ):
+                sa = ShoutAuthor(shout=shout.id, author=author_id)
                 session.add(sa)
 
-            # Добавляем темы
-            for topic in (draft.topics or []):
-                st = ShoutTopic(
-                    topic=topic.id,
-                    shout=shout.id,
-                    main=topic.main if hasattr(topic, "main") else False
-                )
-                session.add(st)
+            # Обрабатываем темы
+            if draft.topics:
+                for topic in draft.topics:
+                    st = ShoutTopic(
+                        topic=topic.id, shout=shout.id, main=topic.main if hasattr(topic, "main") else False
+                    )
+                    session.add(st)
+
+            session.add(shout)
+            session.add(draft)
+            session.flush()
+
+            # Инвалидируем кэш только если это новая публикация или была снята с публикации
+            if not was_published:
+                cache_keys = ["feed", f"author_{author_id}", "random_top", "unrated"]
+
+                # Добавляем ключи для тем
+                for topic in shout.topics:
+                    cache_keys.append(f"topic_{topic.id}")
+                    cache_keys.append(f"topic_shouts_{topic.id}")
+                    await cache_by_id(Topic, topic.id, cache_topic)
+
+                # Инвалидируем кэш
+                await invalidate_shouts_cache(cache_keys)
+                await invalidate_shout_related_cache(shout, author_id)
+
+                # Обновляем кэш авторов
+                for author in shout.authors:
+                    await cache_by_id(Author, author.id, cache_author)
+
+                # Отправляем уведомление о публикации
+                await notify_shout(shout.dict(), "published")
+
+                # Обновляем поисковый индекс
+                search_service.index(shout)
+            else:
+                # Для уже опубликованных материалов просто отправляем уведомление об обновлении
+                await notify_shout(shout.dict(), "update")
 
             session.commit()
-
-            # Инвалидируем кеш
-            invalidate_shouts_cache()
-            invalidate_shout_related_cache(shout.id)
-
-            # Уведомляем о публикации
-            await notify_shout(shout.id)
-
-            # Обновляем поисковый индекс
-            search_service.index_shout(shout)
-
-            logger.info(f"Successfully published shout #{shout.id} from draft #{draft_id}")
-            logger.debug(f"Shout data: {shout.dict()}")
-
             return {"shout": shout}
 
     except Exception as e:
-        logger.error(f"Failed to publish draft {draft_id}: {e}", exc_info=True)
-        return {"error": f"Failed to publish draft: {str(e)}"}
+        logger.error(f"Failed to publish shout: {e}", exc_info=True)
+        if "session" in locals():
+            session.rollback()
+        return {"error": f"Failed to publish shout: {str(e)}"}
+
+
+@mutation.field("unpublish_shout")
+@login_required
+async def unpublish_shout(_, info, shout_id: int):
+    """Unpublish a shout.
+
+    Args:
+        shout_id: The ID of the shout to unpublish
+
+    Returns:
+        dict: The unpublished shout or an error message
+    """
+    author_dict = info.context.get("author", {})
+    author_id = author_dict.get("id")
+    if not author_id:
+        return {"error": "Author ID is required"}
+
+    shout = None
+    with local_session() as session:
+        try:
+            shout = session.query(Shout).filter(Shout.id == shout_id).first()
+            shout.published_at = None
+            session.commit()
+            invalidate_shout_related_cache(shout)
+            invalidate_shouts_cache()
+
+        except Exception:
+            session.rollback()
+            return {"error": "Failed to unpublish shout"}
+
+    return {"shout": shout}

resolvers/editor.py

@@ -3,7 +3,7 @@ import time
 import orjson
 import trafilatura
 from sqlalchemy import and_, desc, select
-from sqlalchemy.orm import joinedload, selectinload
+from sqlalchemy.orm import joinedload
 from sqlalchemy.sql.functions import coalesce
 
 from cache.cache import (
@@ -13,7 +13,6 @@ from cache.cache import (
     invalidate_shouts_cache,
 )
 from orm.author import Author
-from orm.draft import Draft
 from orm.shout import Shout, ShoutAuthor, ShoutTopic
 from orm.topic import Topic
 from resolvers.follower import follow, unfollow
@@ -21,9 +20,8 @@ from resolvers.stat import get_with_stat
 from services.auth import login_required
 from services.db import local_session
 from services.notify import notify_shout
-from services.schema import mutation, query
+from services.schema import query
 from services.search import search_service
-from utils.html_wrapper import wrap_html_fragment
 from utils.logger import root_logger as logger
 
 
@@ -181,11 +179,9 @@ async def create_shout(_, info, inp):
                 # Создаем публикацию без topics
                 body = inp.get("body", "")
                 lead = inp.get("lead", "")
-                body_html = wrap_html_fragment(body)
-                lead_html = wrap_html_fragment(lead)
-                body_text = trafilatura.extract(body_html)
-                lead_text = trafilatura.extract(lead_html)
-                seo = inp.get("seo", lead_text.strip() or body_text.strip()[:300].split(". ")[:-1].join(". "))
+                body_text = trafilatura.extract(body)
+                lead_text = trafilatura.extract(lead)
+                seo = inp.get("seo", lead_text or body_text[:300].split(". ")[:-1].join(". "))
                 new_shout = Shout(
                     slug=slug,
                     body=body,
@@ -649,178 +645,39 @@ def get_main_topic(topics):
     """Get the main topic from a list of ShoutTopic objects."""
     logger.info(f"Starting get_main_topic with {len(topics) if topics else 0} topics")
     logger.debug(
-        f"Topics data: {[(t.slug, getattr(t, 'main', False)) for t in topics] if topics else []}"
+        f"Topics data: {[(t.topic.slug if t.topic else 'no-topic', t.main) for t in topics] if topics else []}"
     )
 
     if not topics:
         logger.warning("No topics provided to get_main_topic")
         return {"id": 0, "title": "no topic", "slug": "notopic", "is_main": True}
 
-    # Проверяем, является ли topics списком объектов ShoutTopic или Topic
-    if hasattr(topics[0], 'topic') and topics[0].topic:
-        # Для ShoutTopic объектов (старый формат)
-        # Find first main topic in original order
-        main_topic_rel = next((st for st in topics if getattr(st, 'main', False)), None)
-        logger.debug(
-            f"Found main topic relation: {main_topic_rel.topic.slug if main_topic_rel and main_topic_rel.topic else None}"
-        )
+    # Find first main topic in original order
+    main_topic_rel = next((st for st in topics if st.main), None)
+    logger.debug(
+        f"Found main topic relation: {main_topic_rel.topic.slug if main_topic_rel and main_topic_rel.topic else None}"
+    )
 
-        if main_topic_rel and main_topic_rel.topic:
-            result = {
-                "slug": main_topic_rel.topic.slug,
-                "title": main_topic_rel.topic.title,
-                "id": main_topic_rel.topic.id,
-                "is_main": True,
-            }
-            logger.info(f"Returning main topic: {result}")
-            return result
+    if main_topic_rel and main_topic_rel.topic:
+        result = {
+            "slug": main_topic_rel.topic.slug,
+            "title": main_topic_rel.topic.title,
+            "id": main_topic_rel.topic.id,
+            "is_main": True,
+        }
+        logger.info(f"Returning main topic: {result}")
+        return result
 
-        # If no main found but topics exist, return first
-        if topics and topics[0].topic:
-            logger.info(f"No main topic found, using first topic: {topics[0].topic.slug}")
-            result = {
-                "slug": topics[0].topic.slug,
-                "title": topics[0].topic.title,
-                "id": topics[0].topic.id,
-                "is_main": True,
-            }
-            return result
-    else:
-        # Для Topic объектов (новый формат из selectinload)
-        # После смены на selectinload у нас просто список Topic объектов
-        if topics:
-            logger.info(f"Using first topic as main: {topics[0].slug}")
-            result = {
-                "slug": topics[0].slug,
-                "title": topics[0].title,
-                "id": topics[0].id,
-                "is_main": True,
-            }
-            return result
+    # If no main found but topics exist, return first
+    if topics and topics[0].topic:
+        logger.info(f"No main topic found, using first topic: {topics[0].topic.slug}")
+        result = {
+            "slug": topics[0].topic.slug,
+            "title": topics[0].topic.title,
+            "id": topics[0].topic.id,
+            "is_main": True,
+        }
+        return result
 
     logger.warning("No valid topics found, returning default")
     return {"slug": "notopic", "title": "no topic", "id": 0, "is_main": True}
-
-@mutation.field("unpublish_shout")
-@login_required
-async def unpublish_shout(_, info, shout_id: int):
-    """Снимает публикацию (shout) с публикации.
-
-    Предзагружает связанный черновик (draft) и его авторов/темы, чтобы избежать
-    ошибок при последующем доступе к ним в GraphQL.
-
-    Args:
-        shout_id: ID публикации для снятия с публикации
-
-    Returns:
-        dict: Снятая с публикации публикация или сообщение об ошибке
-    """
-    author_dict = info.context.get("author", {})
-    author_id = author_dict.get("id")
-    if not author_id:
-        # В идеале нужна проверка прав, имеет ли автор право снимать публикацию
-        return {"error": "Author ID is required"}
-
-    shout = None
-    with local_session() as session:
-        try:
-            # Загружаем Shout со всеми связями для правильного формирования ответа
-            shout = (
-                session.query(Shout)
-                .options(
-                    joinedload(Shout.authors),
-                    selectinload(Shout.topics)
-                )
-                .filter(Shout.id == shout_id)
-                .first()
-            )
-            
-            if not shout:
-                 logger.warning(f"Shout not found for unpublish: ID {shout_id}")
-                 return {"error": "Shout not found"} 
-            
-            # Если у публикации есть связанный черновик, загружаем его с relationships
-            if shout.draft:
-                # Отдельно загружаем черновик с его связями
-                draft = (
-                    session.query(Draft)
-                    .options(
-                        selectinload(Draft.authors),
-                        selectinload(Draft.topics)
-                    )
-                    .filter(Draft.id == shout.draft)
-                    .first()
-                )
-                
-                # Связываем черновик с публикацией вручную для доступа через API
-                if draft:
-                    shout.draft_obj = draft
-
-            # TODO: Добавить проверку прав доступа, если необходимо
-            # if author_id not in [a.id for a in shout.authors]: # Требует selectinload(Shout.authors) выше
-            #    logger.warning(f"Author {author_id} denied unpublishing shout {shout_id}")
-            #    return {"error": "Access denied"}
-
-            # Запоминаем старый slug и id для формирования поля publication
-            shout_slug = shout.slug
-            shout_id_for_publication = shout.id
-
-            # Снимаем с публикации (устанавливаем published_at в None)
-            shout.published_at = None
-            session.commit()
-            
-            # Формируем полноценный словарь для ответа
-            shout_dict = shout.dict()
-            
-            # Добавляем связанные данные
-            shout_dict["topics"] = (
-                [
-                    {"id": topic.id, "slug": topic.slug, "title": topic.title}
-                    for topic in shout.topics
-                ]
-                if shout.topics
-                else []
-            )
-            
-            # Добавляем main_topic 
-            shout_dict["main_topic"] = get_main_topic(shout.topics)
-            
-            # Добавляем авторов
-            shout_dict["authors"] = (
-                [
-                    {"id": author.id, "name": author.name, "slug": author.slug}
-                    for author in shout.authors
-                ]
-                if shout.authors
-                else []
-            )
-            
-            # Важно! Обновляем поле publication, отражая состояние "снят с публикации"
-            shout_dict["publication"] = {
-                "id": shout_id_for_publication,
-                "slug": shout_slug,
-                "published_at": None  # Ключевое изменение - устанавливаем published_at в None
-            }
-
-            # Инвалидация кэша
-            try:
-                cache_keys = [
-                    "feed",  # лента
-                    f"author_{author_id}",  # публикации автора
-                    "random_top",  # случайные топовые
-                    "unrated",  # неоцененные
-                ]
-                await invalidate_shout_related_cache(shout, author_id) 
-                await invalidate_shouts_cache(cache_keys)
-                logger.info(f"Cache invalidated after unpublishing shout {shout_id}")
-            except Exception as cache_err:
-                 logger.error(f"Failed to invalidate cache for unpublish shout {shout_id}: {cache_err}")
-
-        except Exception as e: 
-            session.rollback()
-            logger.error(f"Failed to unpublish shout {shout_id}: {e}", exc_info=True) 
-            return {"error": f"Failed to unpublish shout: {str(e)}"}
-
-    # Возвращаем сформированный словарь вместо объекта
-    logger.info(f"Shout {shout_id} unpublished successfully by author {author_id}")
-    return {"shout": shout_dict}

resolvers/pyrightconfig.json

@@ -0,0 +1,25 @@
+{
+    "include": [
+      "."
+    ],
+    "exclude": [
+      "**/node_modules",
+      "**/__pycache__",
+      "**/.*"
+    ],
+    "defineConstant": {
+      "DEBUG": true
+    },
+    "venvPath": ".",
+    "venv": ".venv",
+    "pythonVersion": "3.11",
+    "typeCheckingMode": "strict",
+    "reportMissingImports": true,
+    "reportMissingTypeStubs": false,
+    "reportUnknownMemberType": false,
+    "reportUnknownParameterType": false,
+    "reportUnknownVariableType": false,
+    "reportUnknownArgumentType": false,
+    "reportPrivateUsage": false,
+    "reportUntypedFunctionDecorator": false
+}

resolvers/reaction.py

@@ -487,16 +487,12 @@ def apply_reaction_filters(by, q):
     shout_slug = by.get("shout")
     if shout_slug:
         q = q.filter(Shout.slug == shout_slug)
-    
-    shout_id = by.get("shout_id")
-    if shout_id:
-        q = q.filter(Shout.id == shout_id)
 
     shouts = by.get("shouts")
     if shouts:
         q = q.filter(Shout.slug.in_(shouts))
 
-    created_by = by.get("created_by", by.get("author_id"))
+    created_by = by.get("created_by")
     if created_by:
         q = q.filter(Author.id == created_by)
 

resolvers/reader.py

@@ -10,7 +10,7 @@ from orm.shout import Shout, ShoutAuthor, ShoutTopic
 from orm.topic import Topic
 from services.db import json_array_builder, json_builder, local_session
 from services.schema import query
-from services.search import search_text
+from services.search import search_text, get_search_count
 from services.viewed import ViewedStorage
 from utils.logger import root_logger as logger
 
@@ -187,12 +187,10 @@ def get_shouts_with_links(info, q, limit=20, offset=0):
     """
     shouts = []
     try:
-        # logger.info(f"Starting get_shouts_with_links with limit={limit}, offset={offset}")
         q = q.limit(limit).offset(offset)
 
         with local_session() as session:
             shouts_result = session.execute(q).all()
-            # logger.info(f"Got {len(shouts_result) if shouts_result else 0} shouts from query")
 
             if not shouts_result:
                 logger.warning("No shouts found in query result")
@@ -203,7 +201,6 @@ def get_shouts_with_links(info, q, limit=20, offset=0):
                     shout = None
                     if hasattr(row, "Shout"):
                         shout = row.Shout
-                        # logger.debug(f"Processing shout#{shout.id} at index {idx}")
                     if shout:
                         shout_id = int(f"{shout.id}")
                         shout_dict = shout.dict()
@@ -231,20 +228,16 @@ def get_shouts_with_links(info, q, limit=20, offset=0):
                         topics = None
                         if has_field(info, "topics") and hasattr(row, "topics"):
                             topics = orjson.loads(row.topics) if isinstance(row.topics, str) else row.topics
-                            # logger.debug(f"Shout#{shout_id} topics: {topics}")
                             shout_dict["topics"] = topics
 
                         if has_field(info, "main_topic"):
                             main_topic = None
                             if hasattr(row, "main_topic"):
-                                # logger.debug(f"Raw main_topic for shout#{shout_id}: {row.main_topic}")
                                 main_topic = (
                                     orjson.loads(row.main_topic) if isinstance(row.main_topic, str) else row.main_topic
                                 )
-                                # logger.debug(f"Parsed main_topic for shout#{shout_id}: {main_topic}")
 
                             if not main_topic and topics and len(topics) > 0:
-                                # logger.info(f"No main_topic found for shout#{shout_id}, using first topic from list")
                                 main_topic = {
                                     "id": topics[0]["id"],
                                     "title": topics[0]["title"],
@@ -252,10 +245,8 @@ def get_shouts_with_links(info, q, limit=20, offset=0):
                                     "is_main": True,
                                 }
                             elif not main_topic:
-                                logger.warning(f"No main_topic and no topics found for shout#{shout_id}")
                                 main_topic = {"id": 0, "title": "no topic", "slug": "notopic", "is_main": True}
                             shout_dict["main_topic"] = main_topic
-                            # logger.debug(f"Final main_topic for shout#{shout_id}: {main_topic}")
 
                         if has_field(info, "authors") and hasattr(row, "authors"):
                             shout_dict["authors"] = (
@@ -282,7 +273,6 @@ def get_shouts_with_links(info, q, limit=20, offset=0):
         logger.error(f"Fatal error in get_shouts_with_links: {e}", exc_info=True)
         raise
     finally:
-        logger.info(f"Returning {len(shouts)} shouts from get_shouts_with_links")
         return shouts
 
 
@@ -401,8 +391,17 @@ async def load_shouts_search(_, info, text, options):
     """
     limit = options.get("limit", 10)
     offset = options.get("offset", 0)
+    
     if isinstance(text, str) and len(text) > 2:
+        # Get search results with pagination
         results = await search_text(text, limit, offset)
+        
+        # If no results, return empty list
+        if not results:
+            logger.info(f"No search results found for '{text}'")
+            return []
+        
+        # Extract IDs and scores
         scores = {}
         hits_ids = []
         for sr in results:
@@ -412,22 +411,42 @@ async def load_shouts_search(_, info, text, options):
                 scores[shout_id] = sr.get("score")
                 hits_ids.append(shout_id)
 
-        q = (
-            query_with_stat(info)
-            if has_field(info, "stat")
-            else select(Shout).filter(and_(Shout.published_at.is_not(None), Shout.deleted_at.is_(None)))
-        )
+        # Query DB for only the IDs in the current page
+        q = query_with_stat(info)
         q = q.filter(Shout.id.in_(hits_ids))
-        q = apply_filters(q, options)
-        q = apply_sorting(q, options)
-        shouts = get_shouts_with_links(info, q, limit, offset)
+        q = apply_filters(q, options.get("filters", {}))
+
+        #
+        shouts = get_shouts_with_links(info, q, len(hits_ids), 0)
+        
+        # Add scores from search results
         for shout in shouts:
-            shout.score = scores[f"{shout.id}"]
-        shouts.sort(key=lambda x: x.score, reverse=True)
+            shout_id = str(shout['id'])
+            shout["score"] = scores.get(shout_id, 0)
+
+        # Re-sort by search score to maintain ranking
+        shouts.sort(key=lambda x: scores.get(str(x['id']), 0), reverse=True)
+        
         return shouts
     return []
 
 
+@query.field("get_search_results_count")
+async def get_search_results_count(_, info, text):
+    """
+    Returns the total count of search results for a search query.
+    
+    :param _: Root query object (unused)
+    :param info: GraphQL context information
+    :param text: Search query text
+    :return: Total count of results
+    """
+    if isinstance(text, str) and len(text) > 2:
+        count = await get_search_count(text)
+        return {"count": count}
+    return {"count": 0}
+
+
 @query.field("load_shouts_unrated")
 async def load_shouts_unrated(_, info, options):
     """

resolvers/topic.py

@@ -10,7 +10,6 @@ from cache.cache import (
 )
 from orm.author import Author
 from orm.topic import Topic
-from orm.reaction import ReactionKind
 from resolvers.stat import get_with_stat
 from services.auth import login_required
 from services.db import local_session
@@ -113,7 +112,7 @@ async def get_topics_with_stats(limit=100, offset=0, community_id=None, by=None)
             shouts_stats_query = f"""
             SELECT st.topic, COUNT(DISTINCT s.id) as shouts_count
             FROM shout_topic st
-            JOIN shout s ON st.shout = s.id AND s.deleted_at IS NULL AND s.published_at IS NOT NULL
+            JOIN shout s ON st.shout = s.id AND s.deleted_at IS NULL
             WHERE st.topic IN ({",".join(map(str, topic_ids))})
             GROUP BY st.topic
             """
@@ -122,7 +121,7 @@ async def get_topics_with_stats(limit=100, offset=0, community_id=None, by=None)
             # Запрос на получение статистики по подписчикам для выбранных тем
             followers_stats_query = f"""
             SELECT topic, COUNT(DISTINCT follower) as followers_count
-            FROM topic_followers tf
+            FROM topic_followers
             WHERE topic IN ({",".join(map(str, topic_ids))})
             GROUP BY topic
             """
@@ -144,8 +143,7 @@ async def get_topics_with_stats(limit=100, offset=0, community_id=None, by=None)
             SELECT st.topic, COUNT(DISTINCT r.id) as comments_count
             FROM shout_topic st
             JOIN shout s ON st.shout = s.id AND s.deleted_at IS NULL AND s.published_at IS NOT NULL
-            JOIN reaction r ON r.shout = s.id AND r.kind = '{ReactionKind.COMMENT.value}' AND r.deleted_at IS NULL
-            JOIN author a ON r.created_by = a.id AND a.deleted_at IS NULL
+            JOIN reaction r ON r.shout = s.id
             WHERE st.topic IN ({",".join(map(str, topic_ids))})
             GROUP BY st.topic
             """

schema/input.graphql

@@ -92,14 +92,12 @@ input LoadShoutsOptions {
 
 input ReactionBy {
   shout: String
-  shout_id: Int
   shouts: [String]
   search: String
   kinds: [ReactionKind]
   reply_to: Int # filter
   topic: String
   created_by: Int
-  author_id: Int
   author: String
   after: Int
   sort: ReactionSort # sort

schema/query.graphql

@@ -33,6 +33,7 @@ type Query {
   get_shout(slug: String, shout_id: Int): Shout
   load_shouts_by(options: LoadShoutsOptions): [Shout]
   load_shouts_search(text: String!, options: LoadShoutsOptions): [SearchResult]
+  get_search_results_count(text: String!): CountResult!
   load_shouts_bookmarked(options: LoadShoutsOptions): [Shout]
 
   # rating

schema/type.graphql

@@ -107,12 +107,6 @@ type Shout {
   score: Float
 }
 
-type PublicationInfo {
-  id: Int!
-  slug: String!
-  published_at: Int
-}
-
 type Draft {
   id: Int!
   created_at: Int!
@@ -135,9 +129,9 @@ type Draft {
   deleted_at: Int
   updated_by: Author
   deleted_by: Author
-  authors: [Author]!
-  topics: [Topic]!
-  publication: PublicationInfo
+  authors: [Author]
+  topics: [Topic]
+
 }
 
 type Stat {
@@ -213,6 +207,7 @@ type CommonResult {
 }
 
 type SearchResult {
+    id: Int!
     slug: String!
     title: String!
     cover: String
@@ -280,3 +275,7 @@ type MyRateComment {
   my_rate: ReactionKind
 }
 
+type CountResult {
+  count: Int!
+}
+

server.py

@@ -0,0 +1,34 @@
+import sys
+from pathlib import Path
+
+from granian.constants import Interfaces
+from granian.log import LogLevels
+from granian.server import Server
+
+from settings import PORT
+from utils.logger import root_logger as logger
+
+if __name__ == "__main__":
+    logger.info("started")
+    try:
+
+        granian_instance = Server(
+            "main:app",
+            address="0.0.0.0",
+            port=PORT,
+            interface=Interfaces.ASGI,
+            workers=1,
+            websockets=False,
+            log_level=LogLevels.debug,
+            backlog=2048,
+        )
+
+        if "dev" in sys.argv:
+            logger.info("dev mode, building ssl context")
+            granian_instance.build_ssl_context(cert=Path("localhost.pem"), key=Path("localhost-key.pem"), password=None)
+        granian_instance.serve()
+    except Exception as error:
+        logger.error(error, exc_info=True)
+        raise
+    finally:
+        logger.info("stopped")

services/__init__.py

@@ -1 +0,0 @@
-# This file makes services a Python package

services/db.py

@@ -259,3 +259,27 @@ def get_json_builder():
 
 # Используем их в коде
 json_builder, json_array_builder, json_cast = get_json_builder()
+
+async def fetch_all_shouts(session=None):
+    """Fetch all published shouts for search indexing"""
+    from orm.shout import Shout
+    
+    close_session = False
+    if session is None:
+        session = local_session()
+        close_session = True
+    
+    try:
+        # Fetch only published and non-deleted shouts
+        query = session.query(Shout).filter(
+            Shout.published_at.is_not(None),
+            Shout.deleted_at.is_(None)
+        )
+        shouts = query.all()
+        return shouts
+    except Exception as e:
+        logger.error(f"Error fetching shouts for search indexing: {e}")
+        return []
+    finally:
+        if close_session:
+            session.close()

services/notify.py

@@ -53,71 +53,3 @@ async def notify_follower(follower: dict, author_id: int, action: str = "follow"
     except Exception as e:
         # Log the error and re-raise it
         logger.error(f"Failed to publish to channel {channel_name}: {e}")
-
-
-async def notify_draft(draft_data, action: str = "publish"):
-    """
-    Отправляет уведомление о публикации или обновлении черновика.
-
-    Функция гарантирует, что данные черновика сериализуются корректно, включая
-    связанные атрибуты (topics, authors).
-
-    Args:
-        draft_data (dict): Словарь с данными черновика. Должен содержать минимум id и title
-        action (str, optional): Действие ("publish", "update"). По умолчанию "publish"
-
-    Returns:
-        None
-
-    Examples:
-        >>> draft = {"id": 1, "title": "Тестовый черновик", "slug": "test-draft"}
-        >>> await notify_draft(draft, "publish")
-    """
-    channel_name = "draft"
-    try:
-        # Убеждаемся, что все необходимые данные присутствуют
-        # и объект не требует доступа к отсоединенным атрибутам
-        if isinstance(draft_data, dict):
-            draft_payload = draft_data
-        else:
-            # Если это ORM объект, преобразуем его в словарь с нужными атрибутами
-            draft_payload = {
-                "id": getattr(draft_data, "id", None),
-                "slug": getattr(draft_data, "slug", None),
-                "title": getattr(draft_data, "title", None),
-                "subtitle": getattr(draft_data, "subtitle", None),
-                "media": getattr(draft_data, "media", None),
-                "created_at": getattr(draft_data, "created_at", None),
-                "updated_at": getattr(draft_data, "updated_at", None),
-            }
-
-            # Если переданы связанные атрибуты, добавим их
-            if hasattr(draft_data, "topics") and draft_data.topics is not None:
-                draft_payload["topics"] = [
-                    {"id": t.id, "name": t.name, "slug": t.slug}
-                    for t in draft_data.topics
-                ]
-
-            if hasattr(draft_data, "authors") and draft_data.authors is not None:
-                draft_payload["authors"] = [
-                    {
-                        "id": a.id,
-                        "name": a.name,
-                        "slug": a.slug,
-                        "pic": getattr(a, "pic", None),
-                    }
-                    for a in draft_data.authors
-                ]
-
-        data = {"payload": draft_payload, "action": action}
-
-        # Сохраняем уведомление
-        save_notification(action, channel_name, data.get("payload"))
-
-        # Публикуем в Redis
-        json_data = orjson.dumps(data)
-        if json_data:
-            await redis.publish(channel_name, json_data)
-
-    except Exception as e:
-        logger.error(f"Failed to publish to channel {channel_name}: {e}")

services/pyrightconfig.json

@@ -1,19 +0,0 @@
-{
-  "include": ["."],
-  "exclude": ["**/node_modules", "**/__pycache__", "**/.*"],
-  "defineConstant": {
-    "DEBUG": true
-  },
-  "venvPath": ".",
-  "venv": ".venv",
-  "pythonVersion": "3.11",
-  "typeCheckingMode": "strict",
-  "reportMissingImports": true,
-  "reportMissingTypeStubs": false,
-  "reportUnknownMemberType": false,
-  "reportUnknownParameterType": false,
-  "reportUnknownVariableType": false,
-  "reportUnknownArgumentType": false,
-  "reportPrivateUsage": false,
-  "reportUntypedFunctionDecorator": false
-}

services/schema.py

@@ -1,15 +1,14 @@
 from asyncio.log import logger
 
 import httpx
-from ariadne import MutationType, ObjectType, QueryType
+from ariadne import MutationType, QueryType
 
 from services.db import create_table_if_not_exists, local_session
 from settings import AUTH_URL
 
 query = QueryType()
 mutation = MutationType()
-type_draft = ObjectType("Draft")
-resolvers = [query, mutation, type_draft]
+resolvers = [query, mutation]
 
 
 async def request_graphql_data(gql, url=AUTH_URL, headers=None):

services/search.py

@@ -2,231 +2,683 @@ import asyncio
 import json
 import logging
 import os
+import httpx
+import time
+from collections import defaultdict
+from datetime import datetime, timedelta
 
-import orjson
-from opensearchpy import OpenSearch
-
-from services.redis import redis
-from utils.encoders import CustomJSONEncoder
-
-# Set redis logging level to suppress DEBUG messages
+# Set up proper logging
 logger = logging.getLogger("search")
-logger.setLevel(logging.WARNING)
+logger.setLevel(logging.INFO)  # Change to INFO to see more details
 
-ELASTIC_HOST = os.environ.get("ELASTIC_HOST", "").replace("https://", "")
-ELASTIC_USER = os.environ.get("ELASTIC_USER", "")
-ELASTIC_PASSWORD = os.environ.get("ELASTIC_PASSWORD", "")
-ELASTIC_PORT = os.environ.get("ELASTIC_PORT", 9200)
-ELASTIC_URL = os.environ.get(
-    "ELASTIC_URL",
-    f"https://{ELASTIC_USER}:{ELASTIC_PASSWORD}@{ELASTIC_HOST}:{ELASTIC_PORT}",
-)
-REDIS_TTL = 86400  # 1 день в секундах
+# Configuration for search service
+SEARCH_ENABLED = bool(os.environ.get("SEARCH_ENABLED", "true").lower() in ["true", "1", "yes"])
+TXTAI_SERVICE_URL = os.environ.get("TXTAI_SERVICE_URL", "none")
+MAX_BATCH_SIZE = int(os.environ.get("SEARCH_MAX_BATCH_SIZE", "25"))
 
-index_settings = {
-    "settings": {
-        "index": {"number_of_shards": 1, "auto_expand_replicas": "0-all"},
-        "analysis": {
-            "analyzer": {
-                "ru": {
-                    "tokenizer": "standard",
-                    "filter": ["lowercase", "ru_stop", "ru_stemmer"],
-                }
-            },
-            "filter": {
-                "ru_stemmer": {"type": "stemmer", "language": "russian"},
-                "ru_stop": {"type": "stop", "stopwords": "_russian_"},
-            },
-        },
-    },
-    "mappings": {
-        "properties": {
-            "body": {"type": "text", "analyzer": "ru"},
-            "title": {"type": "text", "analyzer": "ru"},
-            "subtitle": {"type": "text", "analyzer": "ru"},
-            "lead": {"type": "text", "analyzer": "ru"},
-            "media": {"type": "text", "analyzer": "ru"},
-        }
-    },
-}
+# Search cache configuration
+SEARCH_CACHE_ENABLED = bool(os.environ.get("SEARCH_CACHE_ENABLED", "true").lower() in ["true", "1", "yes"])
+SEARCH_CACHE_TTL_SECONDS = int(os.environ.get("SEARCH_CACHE_TTL_SECONDS", "900"))  # Default: 15 minutes
+SEARCH_MIN_SCORE = float(os.environ.get("SEARCH_MIN_SCORE", "0.1"))
+SEARCH_PREFETCH_SIZE = int(os.environ.get("SEARCH_PREFETCH_SIZE", "200"))
+SEARCH_USE_REDIS = bool(os.environ.get("SEARCH_USE_REDIS", "true").lower() in ["true", "1", "yes"])
 
-expected_mapping = index_settings["mappings"]
+search_offset = 0
 
-# Создание цикла событий
-search_loop = asyncio.get_event_loop()
+# Import Redis client if Redis caching is enabled
+if SEARCH_USE_REDIS:
+    try:
+        from services.redis import redis
+        logger.info("Redis client imported for search caching")
+    except ImportError:
+        logger.warning("Redis client import failed, falling back to memory cache")
+        SEARCH_USE_REDIS = False
 
-# В начале файла добавим флаг
-SEARCH_ENABLED = bool(os.environ.get("ELASTIC_HOST", ""))
-
-
-def get_indices_stats():
-    indices_stats = search_service.client.cat.indices(format="json")
-    for index_info in indices_stats:
-        index_name = index_info["index"]
-        if not index_name.startswith("."):
-            index_health = index_info["health"]
-            index_status = index_info["status"]
-            pri_shards = index_info["pri"]
-            rep_shards = index_info["rep"]
-            docs_count = index_info["docs.count"]
-            docs_deleted = index_info["docs.deleted"]
-            store_size = index_info["store.size"]
-            pri_store_size = index_info["pri.store.size"]
-
-            logger.info(f"Index: {index_name}")
-            logger.info(f"Health: {index_health}")
-            logger.info(f"Status: {index_status}")
-            logger.info(f"Primary Shards: {pri_shards}")
-            logger.info(f"Replica Shards: {rep_shards}")
-            logger.info(f"Documents Count: {docs_count}")
-            logger.info(f"Deleted Documents: {docs_deleted}")
-            logger.info(f"Store Size: {store_size}")
-            logger.info(f"Primary Store Size: {pri_store_size}")
+class SearchCache:
+    """Cache for search results to enable efficient pagination"""
+    
+    def __init__(self, ttl_seconds=SEARCH_CACHE_TTL_SECONDS, max_items=100):
+        self.cache = {}  # Maps search query to list of results
+        self.last_accessed = {}  # Maps search query to last access timestamp
+        self.ttl = ttl_seconds
+        self.max_items = max_items
+        self._redis_prefix = "search_cache:"
+    
+    async def store(self, query, results):
+        """Store search results for a query"""
+        normalized_query = self._normalize_query(query)
+        
+        if SEARCH_USE_REDIS:
+            try:
+                serialized_results = json.dumps(results)
+                await redis.set(
+                    f"{self._redis_prefix}{normalized_query}", 
+                    serialized_results,
+                    ex=self.ttl
+                )
+                logger.info(f"Stored {len(results)} search results for query '{query}' in Redis")
+                return True
+            except Exception as e:
+                logger.error(f"Error storing search results in Redis: {e}")
+                # Fall back to memory cache if Redis fails
+        
+        # First cleanup if needed for memory cache
+        if len(self.cache) >= self.max_items:
+            self._cleanup()
+            
+        # Store results and update timestamp
+        self.cache[normalized_query] = results
+        self.last_accessed[normalized_query] = time.time()
+        logger.info(f"Cached {len(results)} search results for query '{query}' in memory")
+        return True
+    
+    async def get(self, query, limit=10, offset=0):
+        """Get paginated results for a query"""
+        normalized_query = self._normalize_query(query)
+        all_results = None
 
+        # Try to get from Redis first
+        if SEARCH_USE_REDIS:
+            try:
+                cached_data = await redis.get(f"{self._redis_prefix}{normalized_query}")
+                if cached_data:
+                    all_results = json.loads(cached_data)
+                    logger.info(f"Retrieved search results for '{query}' from Redis")
+            except Exception as e:
+                logger.error(f"Error retrieving search results from Redis: {e}")
+        
+        # Fall back to memory cache if not in Redis
+        if all_results is None and normalized_query in self.cache:
+            all_results = self.cache[normalized_query]
+            self.last_accessed[normalized_query] = time.time()
+            logger.info(f"Retrieved search results for '{query}' from memory cache")
+        
+        # If not found in any cache
+        if all_results is None:
+            logger.info(f"Cache miss for query '{query}'")
+            return None
+        
+        # Return paginated subset
+        end_idx = min(offset + limit, len(all_results))
+        if offset >= len(all_results):
+            logger.warning(f"Requested offset {offset} exceeds result count {len(all_results)}")
+            return []
+            
+        logger.info(f"Cache hit for '{query}': serving {offset}:{end_idx} of {len(all_results)} results")
+        return all_results[offset:end_idx]
+    
+    async def has_query(self, query):
+        """Check if query exists in cache"""
+        normalized_query = self._normalize_query(query)
+        
+        # Check Redis first
+        if SEARCH_USE_REDIS:
+            try:
+                exists = await redis.get(f"{self._redis_prefix}{normalized_query}")
+                if exists:
+                    return True
+            except Exception as e:
+                logger.error(f"Error checking Redis for query existence: {e}")
+        
+        # Fall back to memory cache
+        return normalized_query in self.cache
+    
+    async def get_total_count(self, query):
+        """Get total count of results for a query"""
+        normalized_query = self._normalize_query(query)
+        
+        # Check Redis first
+        if SEARCH_USE_REDIS:
+            try:
+                cached_data = await redis.get(f"{self._redis_prefix}{normalized_query}")
+                if cached_data:
+                    all_results = json.loads(cached_data)
+                    return len(all_results)
+            except Exception as e:
+                logger.error(f"Error getting result count from Redis: {e}")
+        
+        # Fall back to memory cache
+        if normalized_query in self.cache:
+            return len(self.cache[normalized_query])
+            
+        return 0
+    
+    def _normalize_query(self, query):
+        """Normalize query string for cache key"""
+        if not query:
+            return ""
+        # Simple normalization - lowercase and strip whitespace
+        return query.lower().strip()
+    
+    def _cleanup(self):
+        """Remove oldest entries if memory cache is full"""
+        now = time.time()
+        # First remove expired entries
+        expired_keys = [
+            key for key, last_access in self.last_accessed.items() 
+            if now - last_access > self.ttl
+        ]
+        
+        for key in expired_keys:
+            if key in self.cache:
+                del self.cache[key]
+            if key in self.last_accessed:
+                del self.last_accessed[key]
+        
+        logger.info(f"Cleaned up {len(expired_keys)} expired search cache entries")
+        
+        # If still above max size, remove oldest entries
+        if len(self.cache) >= self.max_items:
+            # Sort by last access time
+            sorted_items = sorted(self.last_accessed.items(), key=lambda x: x[1])
+            # Remove oldest 20%
+            remove_count = max(1, int(len(sorted_items) * 0.2))
+            for key, _ in sorted_items[:remove_count]:
+                if key in self.cache:
+                    del self.cache[key]
+                if key in self.last_accessed:
+                    del self.last_accessed[key]
+            logger.info(f"Removed {remove_count} oldest search cache entries")
 
 class SearchService:
-    def __init__(self, index_name="search_index"):
-        logger.info("Инициализируем поиск...")
-        self.index_name = index_name
-        self.client = None
-        self.lock = asyncio.Lock()
-
-        # Инициализация клиента OpenSearch только если поиск включен
-        if SEARCH_ENABLED:
-            try:
-                self.client = OpenSearch(
-                    hosts=[{"host": ELASTIC_HOST, "port": ELASTIC_PORT}],
-                    http_compress=True,
-                    http_auth=(ELASTIC_USER, ELASTIC_PASSWORD),
-                    use_ssl=True,
-                    verify_certs=False,
-                    ssl_assert_hostname=False,
-                    ssl_show_warn=False,
-                )
-                logger.info("Клиент OpenSearch.org подключен")
-                search_loop.create_task(self.check_index())
-            except Exception as exc:
-                logger.warning(f"Поиск отключен из-за ошибки подключения: {exc}")
-                self.client = None
-        else:
-            logger.info("Поиск отключен (ELASTIC_HOST не установлен)")
-
+    def __init__(self):
+        logger.info(f"Initializing search service with URL: {TXTAI_SERVICE_URL}")
+        self.available = SEARCH_ENABLED
+        # Use different timeout settings for indexing and search requests
+        self.client = httpx.AsyncClient(timeout=30.0, base_url=TXTAI_SERVICE_URL)
+        self.index_client = httpx.AsyncClient(timeout=120.0, base_url=TXTAI_SERVICE_URL)
+        # Initialize search cache
+        self.cache = SearchCache() if SEARCH_CACHE_ENABLED else None
+        
+        if not self.available:
+            logger.info("Search disabled (SEARCH_ENABLED = False)")
+        
+        if SEARCH_CACHE_ENABLED:
+            cache_location = "Redis" if SEARCH_USE_REDIS else "Memory"
+            logger.info(f"Search caching enabled using {cache_location} cache with TTL={SEARCH_CACHE_TTL_SECONDS}s")
+            logger.info(f"Minimum score filter: {SEARCH_MIN_SCORE}, prefetch size: {SEARCH_PREFETCH_SIZE}")
+            
     async def info(self):
-        if not SEARCH_ENABLED:
+        """Return information about search service"""
+        if not self.available:
             return {"status": "disabled"}
-
         try:
-            return get_indices_stats()
+            response = await self.client.get("/info")
+            response.raise_for_status()
+            result = response.json()
+            logger.info(f"Search service info: {result}")
+            return result
         except Exception as e:
             logger.error(f"Failed to get search info: {e}")
             return {"status": "error", "message": str(e)}
-
-    def delete_index(self):
-        if self.client:
-            logger.warning(f"[!!!] Удаляем индекс {self.index_name}")
-            self.client.indices.delete(index=self.index_name, ignore_unavailable=True)
-
-    def create_index(self):
-        if self.client:
-            logger.info(f"Создается индекс: {self.index_name}")
-            self.client.indices.create(index=self.index_name, body=index_settings)
-            logger.info(f"Индекс {self.index_name} создан")
-
-    async def check_index(self):
-        if self.client:
-            logger.info(f"Проверяем индекс {self.index_name}...")
-            if not self.client.indices.exists(index=self.index_name):
-                self.create_index()
-                self.client.indices.put_mapping(index=self.index_name, body=expected_mapping)
-            else:
-                logger.info(f"Найден существующий индекс {self.index_name}")
-                # Проверка и обновление структуры индекса, если необходимо
-                result = self.client.indices.get_mapping(index=self.index_name)
-                if isinstance(result, str):
-                    result = orjson.loads(result)
-                if isinstance(result, dict):
-                    mapping = result.get(self.index_name, {}).get("mappings")
-                    logger.info(f"Найдена структура индексации: {mapping['properties'].keys()}")
-                    expected_keys = expected_mapping["properties"].keys()
-                    if mapping and mapping["properties"].keys() != expected_keys:
-                        logger.info(f"Ожидаемая структура индексации: {expected_mapping}")
-                        logger.warning("[!!!] Требуется переиндексация всех данных")
-                        self.delete_index()
-                        self.client = None
-        else:
-            logger.error("клиент не инициализован, невозможно проверить индекс")
-
+    
+    def is_ready(self):
+        """Check if service is available"""
+        return self.available
+    
+    async def verify_docs(self, doc_ids):
+        """Verify which documents exist in the search index"""
+        if not self.available:
+            return {"status": "disabled"}
+            
+        try:
+            logger.info(f"Verifying {len(doc_ids)} documents in search index")
+            response = await self.client.post(
+                "/verify-docs",
+                json={"doc_ids": doc_ids},
+                timeout=60.0  # Longer timeout for potentially large ID lists
+            )
+            response.raise_for_status()
+            result = response.json()
+            
+            # Log summary of verification results
+            missing_count = len(result.get("missing", []))
+            logger.info(f"Document verification complete: {missing_count} missing out of {len(doc_ids)} total")
+            
+            return result
+        except Exception as e:
+            logger.error(f"Document verification error: {e}")
+            return {"status": "error", "message": str(e)}
+    
     def index(self, shout):
-        if not SEARCH_ENABLED:
+        """Index a single document"""
+        if not self.available:
             return
+        logger.info(f"Indexing post {shout.id}")
+        # Start in background to not block
+        asyncio.create_task(self.perform_index(shout))
 
-        if self.client:
-            logger.info(f"Индексируем пост {shout.id}")
-            index_body = {
-                "body": shout.body,
-                "title": shout.title,
-                "subtitle": shout.subtitle,
-                "lead": shout.lead,
-                "media": shout.media,
-            }
-            asyncio.create_task(self.perform_index(shout, index_body))
+    async def perform_index(self, shout):
+        """Actually perform the indexing operation"""
+        if not self.available:
+            return
+        
+        try:
+            # Combine all text fields
+            text = " ".join(filter(None, [
+                shout.title or "", 
+                shout.subtitle or "", 
+                shout.lead or "", 
+                shout.body or "",
+                shout.media or ""
+            ]))
+            
+            if not text.strip():
+                logger.warning(f"No text content to index for shout {shout.id}")
+                return
+                
+            logger.info(f"Indexing document: ID={shout.id}, Text length={len(text)}")
+            
+            # Send to txtai service
+            response = await self.client.post(
+                "/index",
+                json={"id": str(shout.id), "text": text}
+            )
+            response.raise_for_status()
+            result = response.json()
+            logger.info(f"Post {shout.id} successfully indexed: {result}")
+        except Exception as e:
+            logger.error(f"Indexing error for shout {shout.id}: {e}")
 
-    async def perform_index(self, shout, index_body):
-        if self.client:
+    async def bulk_index(self, shouts):
+        """Index multiple documents at once with adaptive batch sizing"""
+        if not self.available or not shouts:
+            logger.warning(f"Bulk indexing skipped: available={self.available}, shouts_count={len(shouts) if shouts else 0}")
+            return
+        
+        start_time = time.time()
+        logger.info(f"Starting bulk indexing of {len(shouts)} documents")
+
+        MAX_TEXT_LENGTH = 4000  # Maximum text length to send in a single request
+        max_batch_size = MAX_BATCH_SIZE
+        total_indexed = 0
+        total_skipped = 0
+        total_truncated = 0
+        total_retries = 0
+        
+        # Group documents by size to process smaller documents in larger batches
+        small_docs = []
+        medium_docs = []
+        large_docs = []
+        
+        # First pass: prepare all documents and categorize by size
+        for shout in shouts:
             try:
-                await asyncio.wait_for(
-                    self.client.index(index=self.index_name, id=str(shout.id), body=index_body), timeout=40.0
-                )
-            except asyncio.TimeoutError:
-                logger.error(f"Indexing timeout for shout {shout.id}")
+                text_fields = []
+                for field_name in ['title', 'subtitle', 'lead', 'body']:
+                    field_value = getattr(shout, field_name, None)
+                    if field_value and isinstance(field_value, str) and field_value.strip():
+                        text_fields.append(field_value.strip())
+                
+                # Media field processing remains the same
+                media = getattr(shout, 'media', None)
+                if media:
+                    if isinstance(media, str):
+                        try:
+                            media_json = json.loads(media)
+                            if isinstance(media_json, dict):
+                                if 'title' in media_json:
+                                    text_fields.append(media_json['title'])
+                                if 'body' in media_json:
+                                    text_fields.append(media_json['body'])
+                        except json.JSONDecodeError:
+                            text_fields.append(media)
+                    elif isinstance(media, dict):
+                        if 'title' in media:
+                            text_fields.append(media['title'])
+                        if 'body' in media:
+                            text_fields.append(media['body'])
+                
+                text = " ".join(text_fields)
+                
+                if not text.strip():
+                    total_skipped += 1
+                    continue
+
+                # Truncate text if it exceeds the maximum length
+                original_length = len(text)
+                if original_length > MAX_TEXT_LENGTH:
+                    text = text[:MAX_TEXT_LENGTH]
+                    total_truncated += 1
+                
+                document = {
+                    "id": str(shout.id),
+                    "text": text
+                }
+                
+                # Categorize by size
+                text_len = len(text)
+                if text_len > 5000:
+                    large_docs.append(document)
+                elif text_len > 2000:
+                    medium_docs.append(document)
+                else:
+                    small_docs.append(document)
+                
+                total_indexed += 1
+                
             except Exception as e:
-                logger.error(f"Indexing error for shout {shout.id}: {e}")
+                logger.error(f"Error processing shout {getattr(shout, 'id', 'unknown')} for indexing: {e}")
+                total_skipped += 1
+        
+        # Process each category with appropriate batch sizes
+        logger.info(f"Documents categorized: {len(small_docs)} small, {len(medium_docs)} medium, {len(large_docs)} large")
+        
+        # Process small documents (larger batches)
+        if small_docs:
+            batch_size = min(max_batch_size, 15)
+            await self._process_document_batches(small_docs, batch_size, "small")
+            
+        # Process medium documents (medium batches)
+        if medium_docs:
+            batch_size = min(max_batch_size, 10)
+            await self._process_document_batches(medium_docs, batch_size, "medium")
+            
+        # Process large documents (small batches)
+        if large_docs:
+            batch_size = min(max_batch_size, 3)
+            await self._process_document_batches(large_docs, batch_size, "large")
+        
+        elapsed = time.time() - start_time
+        logger.info(f"Bulk indexing completed in {elapsed:.2f}s: {total_indexed} indexed, {total_skipped} skipped, {total_truncated} truncated, {total_retries} retries")
+
+    async def _process_document_batches(self, documents, batch_size, size_category):
+            """Process document batches with retry logic"""
+            # Check for possible database corruption before starting
+            db_error_count = 0
+            
+            for i in range(0, len(documents), batch_size):
+                batch = documents[i:i+batch_size]
+                batch_id = f"{size_category}-{i//batch_size + 1}"
+                logger.info(f"Processing {size_category} batch {batch_id} of {len(batch)} documents")
+                
+                retry_count = 0
+                max_retries = 3
+                success = False
+                
+                # Process with retries
+                while not success and retry_count < max_retries:
+                    try:
+                        logger.info(f"Sending batch {batch_id} of {len(batch)} documents to search service (attempt {retry_count+1})")
+                        response = await self.index_client.post(
+                            "/bulk-index",
+                            json=batch,
+                            timeout=120.0  # Explicit longer timeout for large batches
+                        )
+                        
+                        # Handle 422 validation errors - these won't be fixed by retrying
+                        if response.status_code == 422:
+                            error_detail = response.json()
+                            truncated_error = self._truncate_error_detail(error_detail)
+                            logger.error(f"Validation error from search service for batch {batch_id}: {truncated_error}")
+                            break
+                        
+                        # Handle 500 server errors - these might be fixed by retrying with smaller batches
+                        elif response.status_code == 500:
+                            db_error_count += 1
+                            
+                            # If we've seen multiple 500s, log a critical error
+                            if db_error_count >= 3:
+                                logger.critical(f"Multiple server errors detected (500). The search service may need manual intervention. Stopping batch {batch_id} processing.")
+                                break
+                                
+                            # Try again with exponential backoff
+                            if retry_count < max_retries - 1:
+                                retry_count += 1
+                                wait_time = (2 ** retry_count) + (random.random() * 0.5)  # Exponential backoff with jitter
+                                await asyncio.sleep(wait_time)
+                                continue
+                            
+                            # Final retry, split the batch
+                            elif len(batch) > 1:
+                                mid = len(batch) // 2
+                                await self._process_single_batch(batch[:mid], f"{batch_id}-A")
+                                await self._process_single_batch(batch[mid:], f"{batch_id}-B")
+                                break
+                            else:
+                                # Can't split a single document
+                                break
+                        
+                        # Normal success case
+                        response.raise_for_status()
+                        success = True
+                        db_error_count = 0  # Reset error counter on success
+                        
+                    except Exception as e:
+                        error_str = str(e).lower()
+                        if "duplicate key" in error_str or "unique constraint" in error_str or "nonetype" in error_str:
+                            db_error_count += 1
+                            if db_error_count >= 2:
+                                logger.critical(f"Potential database corruption detected: {error_str}. The search service may need manual intervention. Stopping batch {batch_id} processing.")
+                                break
+                        
+                        if retry_count < max_retries - 1:
+                            retry_count += 1
+                            wait_time = (2 ** retry_count) + (random.random() * 0.5)
+                            await asyncio.sleep(wait_time)
+                        else:
+                            if len(batch) > 1:
+                                mid = len(batch) // 2
+                                await self._process_single_batch(batch[:mid], f"{batch_id}-A")
+                                await self._process_single_batch(batch[mid:], f"{batch_id}-B")
+                            break
+
+    async def _process_single_batch(self, documents, batch_id):
+        """Process a single batch with maximum reliability"""
+        max_retries = 3
+        retry_count = 0
+        
+        while retry_count < max_retries:
+            try:
+                if not documents:
+                    return
+                    
+                response = await self.index_client.post(
+                    "/bulk-index",
+                    json=documents,
+                    timeout=90.0
+                )
+                response.raise_for_status()
+                return  # Success, exit the retry loop
+                
+            except Exception as e:
+                error_str = str(e).lower()
+                retry_count += 1
+                
+                if "dictionary changed size" in error_str or "transaction error" in error_str:
+                    wait_time = (2 ** retry_count) + (random.random() * 0.5)
+                    await asyncio.sleep(wait_time)  # Wait for txtai to recover
+                    continue
+                
+                if retry_count >= max_retries and len(documents) > 1:
+                    for i, doc in enumerate(documents):
+                        try:
+                            resp = await self.index_client.post("/index", json=doc, timeout=30.0)
+                            resp.raise_for_status()
+                        except Exception as e2:
+                            pass
+                    return  # Exit after individual processing attempt
+
+    def _truncate_error_detail(self, error_detail):
+        """Truncate error details for logging"""
+        truncated_detail = error_detail.copy() if isinstance(error_detail, dict) else error_detail
+        
+        if isinstance(truncated_detail, dict) and 'detail' in truncated_detail and isinstance(truncated_detail['detail'], list):
+            for i, item in enumerate(truncated_detail['detail']):
+                if isinstance(item, dict) and 'input' in item:
+                    if isinstance(item['input'], dict) and any(k in item['input'] for k in ['documents', 'text']):
+                        if 'documents' in item['input'] and isinstance(item['input']['documents'], list):
+                            for j, doc in enumerate(item['input']['documents']):
+                                if 'text' in doc and isinstance(doc['text'], str) and len(doc['text']) > 100:
+                                    item['input']['documents'][j]['text'] = f"{doc['text'][:100]}... [truncated, total {len(doc['text'])} chars]"
+                        
+                        if 'text' in item['input'] and isinstance(item['input']['text'], str) and len(item['input']['text']) > 100:
+                            item['input']['text'] = f"{item['input']['text'][:100]}... [truncated, total {len(item['input']['text'])} chars]"
+        
+        return truncated_detail
 
     async def search(self, text, limit, offset):
-        if not SEARCH_ENABLED:
+        """Search documents"""
+        if not self.available:
             return []
+        
+        if not isinstance(text, str) or not text.strip():
+            return []
+            
+        logger.info(f"Searching for: '{text}' (limit={limit}, offset={offset})")
+        
+        # Check if we can serve from cache
+        if SEARCH_CACHE_ENABLED:
+            has_cache = await self.cache.has_query(text)
+            if has_cache:
+                cached_results = await self.cache.get(text, limit, offset)
+                if cached_results is not None:
+                    return cached_results
+        
+        # Not in cache or cache disabled, perform new search
+        try:
+            search_limit = limit
+            search_offset = offset
 
-        logger.info(f"Ищем: {text} {offset}+{limit}")
-        search_body = {
-            "query": {"multi_match": {"query": text, "fields": ["title", "lead", "subtitle", "body", "media"]}}
-        }
-
-        if self.client:
-            search_response = self.client.search(
-                index=self.index_name,
-                body=search_body,
-                size=limit,
-                from_=offset,
-                _source=False,
-                _source_excludes=["title", "body", "subtitle", "media", "lead", "_index"],
+            if SEARCH_CACHE_ENABLED:
+                search_limit = SEARCH_PREFETCH_SIZE
+                search_offset = 0
+            else:
+                search_limit = limit
+                search_offset = offset
+                
+            response = await self.client.post(
+                "/search",
+                json={"text": text, "limit": search_limit, "offset": search_offset}
             )
-            hits = search_response["hits"]["hits"]
-            results = [{"id": hit["_id"], "score": hit["_score"]} for hit in hits]
-
-            # если результаты не пустые
-            if results:
-                # Кэширование в Redis с TTL
-                redis_key = f"search:{text}:{offset}+{limit}"
-                await redis.execute(
-                    "SETEX",
-                    redis_key,
-                    REDIS_TTL,
-                    json.dumps(results, cls=CustomJSONEncoder),
-                )
-            return results
-        return []
-
+            response.raise_for_status()
+            
+            result = response.json()
+            
+            formatted_results = result.get("results", [])
+            
+            valid_results = []
+            for item in formatted_results:
+                doc_id = item.get("id")
+                if doc_id and doc_id.isdigit():
+                    valid_results.append(item)
+            
+            if len(valid_results) != len(formatted_results):
+                formatted_results = valid_results
+                
+            if SEARCH_MIN_SCORE > 0:
+                initial_count = len(formatted_results)
+                formatted_results = [r for r in formatted_results if r.get("score", 0) >= SEARCH_MIN_SCORE]
+            
+            if SEARCH_CACHE_ENABLED:
+                await self.cache.store(text, formatted_results)
+                end_idx = offset + limit
+                page_results = formatted_results[offset:end_idx]
+                return page_results
+                    
+            return formatted_results
+        except Exception as e:
+            logger.error(f"Search error for '{text}': {e}", exc_info=True)
+            return []
+    
+    async def check_index_status(self):
+        """Get detailed statistics about the search index health"""
+        if not self.available:
+            return {"status": "disabled"}
+            
+        try:
+            response = await self.client.get("/index-status")
+            response.raise_for_status()
+            result = response.json()
+            
+            if result.get("consistency", {}).get("status") != "ok":
+                null_count = result.get("consistency", {}).get("null_embeddings_count", 0)
+                if null_count > 0:
+                    logger.warning(f"Found {null_count} documents with NULL embeddings")
+                    
+            return result
+        except Exception as e:
+            logger.error(f"Failed to check index status: {e}")
+            return {"status": "error", "message": str(e)}
 
+# Create the search service singleton
 search_service = SearchService()
 
-
+# API-compatible function to perform a search
 async def search_text(text: str, limit: int = 50, offset: int = 0):
     payload = []
-    if search_service.client:
-        # Использование метода search_post из OpenSearchService
+    if search_service.available:
         payload = await search_service.search(text, limit, offset)
     return payload
 
+async def get_search_count(text: str):
+    """Get total count of results for a query without fetching all results"""
+    if search_service.available and SEARCH_CACHE_ENABLED:
+        if await search_service.cache.has_query(text):
+            return await search_service.cache.get_total_count(text)
+    results = await search_text(text, SEARCH_PREFETCH_SIZE, 0)
+    return len(results)
 
-# Проверить что URL корректный
-OPENSEARCH_URL = os.getenv("OPENSEARCH_URL", "rc1a-3n5pi3bhuj9gieel.mdb.yandexcloud.net")
+async def initialize_search_index(shouts_data):
+    """Initialize search index with existing data during application startup"""
+    if not SEARCH_ENABLED:
+        return
+        
+    if not shouts_data:
+        return
+        
+    info = await search_service.info()
+    if info.get("status") in ["error", "unavailable", "disabled"]:
+        return
+    
+    index_stats = info.get("index_stats", {})
+    indexed_doc_count = index_stats.get("document_count", 0)
+
+    index_status = await search_service.check_index_status()
+    if index_status.get("status") == "inconsistent":
+        problem_ids = index_status.get("consistency", {}).get("null_embeddings_sample", [])
+                
+        if problem_ids:
+            problem_docs = [shout for shout in shouts_data if str(shout.id) in problem_ids]
+            if problem_docs:
+                await search_service.bulk_index(problem_docs)
+    
+    db_ids = [str(shout.id) for shout in shouts_data]
+    
+    try:
+        numeric_ids = [int(sid) for sid in db_ids if sid.isdigit()]
+        if numeric_ids:
+            min_id = min(numeric_ids)
+            max_id = max(numeric_ids)
+            id_range = max_id - min_id + 1
+    except Exception as e:
+        pass
+    
+    if abs(indexed_doc_count - len(shouts_data)) > 10:
+        doc_ids = [str(shout.id) for shout in shouts_data]
+        
+        verification = await search_service.verify_docs(doc_ids)
+        
+        if verification.get("status") == "error":
+            return
+            
+        missing_ids = verification.get("missing", [])
+        if missing_ids:
+            missing_docs = [shout for shout in shouts_data if str(shout.id) in missing_ids]
+            await search_service.bulk_index(missing_docs)
+    else:
+        pass
+        
+    try:
+        test_query = "test"
+        test_results = await search_text(test_query, 5)
+        
+        if test_results:
+            categories = set()
+            for result in test_results:
+                result_id = result.get("id")
+                matching_shouts = [s for s in shouts_data if str(s.id) == result_id]
+                if matching_shouts and hasattr(matching_shouts[0], 'category'):
+                    categories.add(getattr(matching_shouts[0], 'category', 'unknown'))
+    except Exception as e:
+        pass

tests/auth/conftest.py

@@ -1,25 +0,0 @@
-import pytest
-from typing import Dict
-
-
-@pytest.fixture
-def oauth_settings() -> Dict[str, Dict[str, str]]:
-    """Тестовые настройки OAuth"""
-    return {
-        "GOOGLE": {"id": "test_google_id", "key": "test_google_secret"},
-        "GITHUB": {"id": "test_github_id", "key": "test_github_secret"},
-        "FACEBOOK": {"id": "test_facebook_id", "key": "test_facebook_secret"},
-    }
-
-
-@pytest.fixture
-def frontend_url() -> str:
-    """URL фронтенда для тестов"""
-    return "https://localhost:3000"
-
-
-@pytest.fixture(autouse=True)
-def mock_settings(monkeypatch, oauth_settings, frontend_url):
-    """Подменяем настройки для тестов"""
-    monkeypatch.setattr("auth.oauth.OAUTH_CLIENTS", oauth_settings)
-    monkeypatch.setattr("auth.oauth.FRONTEND_URL", frontend_url)

tests/auth/test_oauth.py

@@ -1,222 +0,0 @@
-import pytest
-from unittest.mock import AsyncMock, MagicMock, patch
-from starlette.responses import JSONResponse, RedirectResponse
-
-from auth.oauth import get_user_profile, oauth_login, oauth_callback
-
-# Подменяем настройки для тестов
-with (
-    patch("auth.oauth.FRONTEND_URL", "https://localhost:3000"),
-    patch(
-        "auth.oauth.OAUTH_CLIENTS",
-        {
-            "GOOGLE": {"id": "test_google_id", "key": "test_google_secret"},
-            "GITHUB": {"id": "test_github_id", "key": "test_github_secret"},
-            "FACEBOOK": {"id": "test_facebook_id", "key": "test_facebook_secret"},
-        },
-    ),
-):
-
-    @pytest.fixture
-    def mock_request():
-        """Фикстура для мока запроса"""
-        request = MagicMock()
-        request.session = {}
-        request.path_params = {}
-        request.query_params = {}
-        return request
-
-    @pytest.fixture
-    def mock_oauth_client():
-        """Фикстура для мока OAuth клиента"""
-        client = AsyncMock()
-        client.authorize_redirect = AsyncMock()
-        client.authorize_access_token = AsyncMock()
-        client.get = AsyncMock()
-        return client
-
-    @pytest.mark.asyncio
-    async def test_get_user_profile_google():
-        """Тест получения профиля из Google"""
-        client = AsyncMock()
-        token = {
-            "userinfo": {
-                "sub": "123",
-                "email": "test@gmail.com",
-                "name": "Test User",
-                "picture": "https://lh3.googleusercontent.com/photo=s96",
-            }
-        }
-
-        profile = await get_user_profile("google", client, token)
-
-        assert profile["id"] == "123"
-        assert profile["email"] == "test@gmail.com"
-        assert profile["name"] == "Test User"
-        assert profile["picture"] == "https://lh3.googleusercontent.com/photo=s600"
-
-    @pytest.mark.asyncio
-    async def test_get_user_profile_github():
-        """Тест получения профиля из GitHub"""
-        client = AsyncMock()
-        client.get.side_effect = [
-            MagicMock(
-                json=lambda: {
-                    "id": 456,
-                    "login": "testuser",
-                    "name": "Test User",
-                    "avatar_url": "https://github.com/avatar.jpg",
-                }
-            ),
-            MagicMock(
-                json=lambda: [
-                    {"email": "other@github.com", "primary": False},
-                    {"email": "test@github.com", "primary": True},
-                ]
-            ),
-        ]
-
-        profile = await get_user_profile("github", client, {})
-
-        assert profile["id"] == "456"
-        assert profile["email"] == "test@github.com"
-        assert profile["name"] == "Test User"
-        assert profile["picture"] == "https://github.com/avatar.jpg"
-
-    @pytest.mark.asyncio
-    async def test_get_user_profile_facebook():
-        """Тест получения профиля из Facebook"""
-        client = AsyncMock()
-        client.get.return_value = MagicMock(
-            json=lambda: {
-                "id": "789",
-                "name": "Test User",
-                "email": "test@facebook.com",
-                "picture": {"data": {"url": "https://facebook.com/photo.jpg"}},
-            }
-        )
-
-        profile = await get_user_profile("facebook", client, {})
-
-        assert profile["id"] == "789"
-        assert profile["email"] == "test@facebook.com"
-        assert profile["name"] == "Test User"
-        assert profile["picture"] == "https://facebook.com/photo.jpg"
-
-    @pytest.mark.asyncio
-    async def test_oauth_login_success(mock_request, mock_oauth_client):
-        """Тест успешного начала OAuth авторизации"""
-        mock_request.path_params["provider"] = "google"
-
-        # Настраиваем мок для authorize_redirect
-        redirect_response = RedirectResponse(url="http://example.com")
-        mock_oauth_client.authorize_redirect.return_value = redirect_response
-
-        with patch("auth.oauth.oauth.create_client", return_value=mock_oauth_client):
-            response = await oauth_login(mock_request)
-
-            assert isinstance(response, RedirectResponse)
-            assert mock_request.session["provider"] == "google"
-            assert "code_verifier" in mock_request.session
-            assert "state" in mock_request.session
-
-            mock_oauth_client.authorize_redirect.assert_called_once()
-
-    @pytest.mark.asyncio
-    async def test_oauth_login_invalid_provider(mock_request):
-        """Тест с неправильным провайдером"""
-        mock_request.path_params["provider"] = "invalid"
-
-        response = await oauth_login(mock_request)
-
-        assert isinstance(response, JSONResponse)
-        assert response.status_code == 400
-        assert "Invalid provider" in response.body.decode()
-
-    @pytest.mark.asyncio
-    async def test_oauth_callback_success(mock_request, mock_oauth_client):
-        """Тест успешного OAuth callback"""
-        mock_request.session = {
-            "provider": "google",
-            "code_verifier": "test_verifier",
-            "state": "test_state",
-        }
-        mock_request.query_params["state"] = "test_state"
-
-        mock_oauth_client.authorize_access_token.return_value = {
-            "userinfo": {"sub": "123", "email": "test@gmail.com", "name": "Test User"}
-        }
-
-        with (
-            patch("auth.oauth.oauth.create_client", return_value=mock_oauth_client),
-            patch("auth.oauth.local_session") as mock_session,
-            patch("auth.oauth.TokenStorage.create_session", return_value="test_token"),
-        ):
-            # Мокаем сессию базы данных
-            session = MagicMock()
-            session.query.return_value.filter.return_value.first.return_value = None
-            mock_session.return_value.__enter__.return_value = session
-
-            response = await oauth_callback(mock_request)
-
-            assert isinstance(response, RedirectResponse)
-            assert response.status_code == 307
-            assert "auth/success" in response.headers["location"]
-
-            # Проверяем cookie
-            cookies = response.headers.getlist("set-cookie")
-            assert any("session_token=test_token" in cookie for cookie in cookies)
-            assert any("httponly" in cookie.lower() for cookie in cookies)
-            assert any("secure" in cookie.lower() for cookie in cookies)
-
-            # Проверяем очистку сессии
-            assert "code_verifier" not in mock_request.session
-            assert "provider" not in mock_request.session
-            assert "state" not in mock_request.session
-
-    @pytest.mark.asyncio
-    async def test_oauth_callback_invalid_state(mock_request):
-        """Тест с неправильным state параметром"""
-        mock_request.session = {"provider": "google", "state": "correct_state"}
-        mock_request.query_params["state"] = "wrong_state"
-
-        response = await oauth_callback(mock_request)
-
-        assert isinstance(response, JSONResponse)
-        assert response.status_code == 400
-        assert "Invalid state" in response.body.decode()
-
-    @pytest.mark.asyncio
-    async def test_oauth_callback_existing_user(mock_request, mock_oauth_client):
-        """Тест OAuth callback с существующим пользователем"""
-        mock_request.session = {
-            "provider": "google",
-            "code_verifier": "test_verifier",
-            "state": "test_state",
-        }
-        mock_request.query_params["state"] = "test_state"
-
-        mock_oauth_client.authorize_access_token.return_value = {
-            "userinfo": {"sub": "123", "email": "test@gmail.com", "name": "Test User"}
-        }
-
-        with (
-            patch("auth.oauth.oauth.create_client", return_value=mock_oauth_client),
-            patch("auth.oauth.local_session") as mock_session,
-            patch("auth.oauth.TokenStorage.create_session", return_value="test_token"),
-        ):
-            # Мокаем существующего пользователя
-            existing_user = MagicMock()
-            session = MagicMock()
-            session.query.return_value.filter.return_value.first.return_value = existing_user
-            mock_session.return_value.__enter__.return_value = session
-
-            response = await oauth_callback(mock_request)
-
-            assert isinstance(response, RedirectResponse)
-            assert response.status_code == 307
-
-            # Проверяем обновление существующего пользователя
-            assert existing_user.name == "Test User"
-            assert existing_user.oauth == "google:123"
-            assert existing_user.email_verified is True

tests/auth/test_settings.py

@@ -1,9 +0,0 @@
-"""Тестовые настройки для OAuth"""
-
-FRONTEND_URL = "https://localhost:3000"
-
-OAUTH_CLIENTS = {
-    "GOOGLE": {"id": "test_google_id", "key": "test_google_secret"},
-    "GITHUB": {"id": "test_github_id", "key": "test_github_secret"},
-    "FACEBOOK": {"id": "test_facebook_id", "key": "test_facebook_secret"},
-}

tests/conftest.py

@@ -1,7 +1,17 @@
 import asyncio
+import os
+
 import pytest
+from sqlalchemy import create_engine
+from sqlalchemy.orm import Session
+from starlette.testclient import TestClient
+
+from main import app
+from services.db import Base
 from services.redis import redis
-from tests.test_config import get_test_client
+
+# Use SQLite for testing
+TEST_DB_URL = "sqlite:///test.db"
 
 
 @pytest.fixture(scope="session")
@@ -13,36 +23,38 @@ def event_loop():
 
 
 @pytest.fixture(scope="session")
-def test_app():
-    """Create a test client and session factory."""
-    client, SessionLocal = get_test_client()
-    return client, SessionLocal
+def test_engine():
+    """Create a test database engine."""
+    engine = create_engine(TEST_DB_URL)
+    Base.metadata.create_all(engine)
+    yield engine
+    Base.metadata.drop_all(engine)
+    os.remove("test.db")
 
 
 @pytest.fixture
-def db_session(test_app):
+def db_session(test_engine):
     """Create a new database session for a test."""
-    _, SessionLocal = test_app
-    session = SessionLocal()
+    connection = test_engine.connect()
+    transaction = connection.begin()
+    session = Session(bind=connection)
 
     yield session
 
-    session.rollback()
     session.close()
-
-
-@pytest.fixture
-def test_client(test_app):
-    """Get the test client."""
-    client, _ = test_app
-    return client
+    transaction.rollback()
+    connection.close()
 
 
 @pytest.fixture
 async def redis_client():
     """Create a test Redis client."""
     await redis.connect()
-    await redis.flushall()  # Очищаем Redis перед каждым тестом
     yield redis
-    await redis.flushall()  # Очищаем после теста
     await redis.disconnect()
+
+
+@pytest.fixture
+def test_client():
+    """Create a TestClient instance."""
+    return TestClient(app)

tests/test_config.py

@@ -1,67 +0,0 @@
-"""
-Конфигурация для тестов
-"""
-
-from sqlalchemy import create_engine
-from sqlalchemy.orm import sessionmaker
-from sqlalchemy.pool import StaticPool
-from starlette.applications import Starlette
-from starlette.middleware import Middleware
-from starlette.middleware.base import BaseHTTPMiddleware
-from starlette.testclient import TestClient
-
-# Используем in-memory SQLite для тестов
-TEST_DB_URL = "sqlite:///:memory:"
-
-
-class DatabaseMiddleware(BaseHTTPMiddleware):
-    """Middleware для внедрения сессии БД"""
-
-    def __init__(self, app, session_maker):
-        super().__init__(app)
-        self.session_maker = session_maker
-
-    async def dispatch(self, request, call_next):
-        session = self.session_maker()
-        request.state.db = session
-        try:
-            response = await call_next(request)
-        finally:
-            session.close()
-        return response
-
-
-def create_test_app():
-    """Create a test Starlette application."""
-    from services.db import Base
-
-    # Создаем движок и таблицы
-    engine = create_engine(
-        TEST_DB_URL,
-        connect_args={"check_same_thread": False},
-        poolclass=StaticPool,
-        echo=False,
-    )
-    Base.metadata.drop_all(bind=engine)
-    Base.metadata.create_all(bind=engine)
-
-    # Создаем фабрику сессий
-    SessionLocal = sessionmaker(bind=engine)
-
-    # Создаем middleware для сессий
-    middleware = [Middleware(DatabaseMiddleware, session_maker=SessionLocal)]
-
-    # Создаем тестовое приложение
-    app = Starlette(
-        debug=True,
-        middleware=middleware,
-        routes=[],  # Здесь можно добавить тестовые маршруты если нужно
-    )
-
-    return app, SessionLocal
-
-
-def get_test_client():
-    """Get a test client with initialized database."""
-    app, SessionLocal = create_test_app()
-    return TestClient(app), SessionLocal

tests/test_reactions.py

@@ -53,11 +53,7 @@ async def test_create_reaction(test_client, db_session, test_setup):
             }
             """,
             "variables": {
-                "reaction": {
-                    "shout": test_setup["shout"].id,
-                    "kind": ReactionKind.LIKE.value,
-                    "body": "Great post!",
-                }
+                "reaction": {"shout": test_setup["shout"].id, "kind": ReactionKind.LIKE.value, "body": "Great post!"}
             },
         },
     )

tests/test_validations.py

@@ -0,0 +1,70 @@
+from datetime import datetime, timedelta
+
+import pytest
+from pydantic import ValidationError
+
+from auth.validations import (
+    AuthInput,
+    AuthResponse,
+    TokenPayload,
+    UserRegistrationInput,
+)
+
+
+class TestAuthValidations:
+    def test_auth_input(self):
+        """Test basic auth input validation"""
+        # Valid case
+        auth = AuthInput(user_id="123", username="testuser", token="1234567890abcdef1234567890abcdef")
+        assert auth.user_id == "123"
+        assert auth.username == "testuser"
+
+        # Invalid cases
+        with pytest.raises(ValidationError):
+            AuthInput(user_id="", username="test", token="x" * 32)
+
+        with pytest.raises(ValidationError):
+            AuthInput(user_id="123", username="t", token="x" * 32)
+
+    def test_user_registration(self):
+        """Test user registration validation"""
+        # Valid case
+        user = UserRegistrationInput(email="test@example.com", password="SecurePass123!", name="Test User")
+        assert user.email == "test@example.com"
+        assert user.name == "Test User"
+
+        # Test email validation
+        with pytest.raises(ValidationError) as exc:
+            UserRegistrationInput(email="invalid-email", password="SecurePass123!", name="Test")
+        assert "Invalid email format" in str(exc.value)
+
+        # Test password validation
+        with pytest.raises(ValidationError) as exc:
+            UserRegistrationInput(email="test@example.com", password="weak", name="Test")
+        assert "String should have at least 8 characters" in str(exc.value)
+
+    def test_token_payload(self):
+        """Test token payload validation"""
+        now = datetime.utcnow()
+        exp = now + timedelta(hours=1)
+
+        payload = TokenPayload(user_id="123", username="testuser", exp=exp, iat=now)
+        assert payload.user_id == "123"
+        assert payload.username == "testuser"
+        assert payload.scopes == []  # Default empty list
+
+    def test_auth_response(self):
+        """Test auth response validation"""
+        # Success case
+        success_resp = AuthResponse(success=True, token="valid_token", user={"id": "123", "name": "Test"})
+        assert success_resp.success is True
+        assert success_resp.token == "valid_token"
+
+        # Error case
+        error_resp = AuthResponse(success=False, error="Invalid credentials")
+        assert error_resp.success is False
+        assert error_resp.error == "Invalid credentials"
+
+        # Invalid case - отсутствует обязательное поле token при success=True
+        with pytest.raises(ValidationError):
+            AuthResponse(success=True, user={"id": "123", "name": "Test"})

utils/__init__.py

@@ -1 +0,0 @@
- 

utils/html_wrapper.py

@@ -1,38 +0,0 @@
-"""
-Модуль для обработки HTML-фрагментов
-"""
-
-def wrap_html_fragment(fragment: str) -> str:
-    """
-    Оборачивает HTML-фрагмент в полную HTML-структуру для корректной обработки.
-    
-    Args:
-        fragment: HTML-фрагмент для обработки
-        
-    Returns:
-        str: Полный HTML-документ
-        
-    Example:
-        >>> wrap_html_fragment("<p>Текст параграфа</p>")
-        '<!DOCTYPE html><html><head><meta charset="utf-8"></head><body><p>Текст параграфа</p></body></html>'
-    """
-    if not fragment or not fragment.strip():
-        return fragment
-        
-    # Проверяем, является ли контент полным HTML-документом
-    is_full_html = fragment.strip().startswith('<!DOCTYPE') or fragment.strip().startswith('<html')
-    
-    # Если это фрагмент, оборачиваем его в полный HTML-документ
-    if not is_full_html:
-        return f"""<!DOCTYPE html>
-<html>
-<head>
-    <meta charset="utf-8">
-    <title></title>
-</head>
-<body>
-{fragment}
-</body>
-</html>"""
-    
-    return fragment