diff --git a/CHANGELOG.md b/CHANGELOG.md
index c499b46f..075c62c4 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -2,7 +2,11 @@
 - `Shout.draft` field added
 - `Draft` entity added
 - `create_draft`, `update_draft`, `delete_draft` mutations and resolvers added
-- `get_shout_drafts` resolver updated
+- `create_shout`, `update_shout`, `delete_shout` mutations removed from GraphQL API
+- `load_drafts` resolver implemented
+- `publish_` and `unpublish_` mutations and resolvers added
+- `create_`, `update_`, `delete_` mutations and resolvers added for `Draft` entity
+- tests with pytest for auth, shouts, drafts
 
 #### [0.4.8] - 2025-02-03
 - `Reaction.deleted_at` filter on `update_reaction` resolver added
diff --git a/auth/identity.py b/auth/identity.py
index 5bbb6030..3a096d9d 100644
--- a/auth/identity.py
+++ b/auth/identity.py
@@ -2,13 +2,13 @@ from binascii import hexlify
 from hashlib import sha256
 
 # from base.exceptions import InvalidPassword, InvalidToken
-from base.orm import local_session
-from jwt import DecodeError, ExpiredSignatureError
+from services.db import local_session
+from auth.exceptions import ExpiredToken, InvalidToken
 from passlib.hash import bcrypt
 
 from auth.jwtcodec import JWTCodec
 from auth.tokenstorage import TokenStorage
-from orm import User
+from orm.user import User
 
 
 class Password:
@@ -79,10 +79,10 @@ class Identity:
             if not await TokenStorage.exist(f"{payload.user_id}-{payload.username}-{token}"):
                 # raise InvalidToken("Login token has expired, please login again")
                 return {"error": "Token has expired"}
-        except ExpiredSignatureError:
+        except ExpiredToken:
             # raise InvalidToken("Login token has expired, please try again")
             return {"error": "Token has expired"}
-        except DecodeError:
+        except InvalidToken:
             # raise InvalidToken("token format error") from e
             return {"error": "Token format error"}
         with local_session() as session:
diff --git a/auth/tokenstorage.py b/auth/tokenstorage.py
index 3ee5e7fd..49fed14d 100644
--- a/auth/tokenstorage.py
+++ b/auth/tokenstorage.py
@@ -1,7 +1,7 @@
 from datetime import datetime, timedelta, timezone
 
-from base.redis import redis
-from validations.auth import AuthInput
+from services.redis import redis
+from auth.validations import AuthInput
 
 from auth.jwtcodec import JWTCodec
 from settings import ONETIME_TOKEN_LIFE_SPAN, SESSION_TOKEN_LIFE_SPAN
diff --git a/auth/validations.py b/auth/validations.py
new file mode 100644
index 00000000..c4a7d253
--- /dev/null
+++ b/auth/validations.py
@@ -0,0 +1,103 @@
+import re
+from datetime import datetime
+from typing import Dict, List, Optional, Union
+from pydantic import BaseModel, Field, field_validator
+
+# RFC 5322 compliant email regex pattern
+EMAIL_PATTERN = r"^[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}$"
+
+class AuthInput(BaseModel):
+    """Base model for authentication input validation"""
+    user_id: str = Field(description="Unique user identifier")
+    username: str = Field(min_length=2, max_length=50)
+    token: str = Field(min_length=32)
+
+    @field_validator('user_id')
+    @classmethod
+    def validate_user_id(cls, v: str) -> str:
+        if not v.strip():
+            raise ValueError("user_id cannot be empty")
+        return v
+
+class UserRegistrationInput(BaseModel):
+    """Validation model for user registration"""
+    email: str = Field(max_length=254)  # Max email length per RFC 5321
+    password: str = Field(min_length=8, max_length=100)
+    name: str = Field(min_length=2, max_length=50)
+    
+    @field_validator('email')
+    @classmethod
+    def validate_email(cls, v: str) -> str:
+        """Validate email format"""
+        if not re.match(EMAIL_PATTERN, v):
+            raise ValueError("Invalid email format")
+        return v.lower()
+    
+    @field_validator('password')
+    @classmethod
+    def validate_password_strength(cls, v: str) -> str:
+        """Validate password meets security requirements"""
+        if not any(c.isupper() for c in v):
+            raise ValueError("Password must contain at least one uppercase letter")
+        if not any(c.islower() for c in v):
+            raise ValueError("Password must contain at least one lowercase letter")
+        if not any(c.isdigit() for c in v):
+            raise ValueError("Password must contain at least one number")
+        if not any(c in "!@#$%^&*()_+-=[]{}|;:,.<>?" for c in v):
+            raise ValueError("Password must contain at least one special character")
+        return v
+
+class UserLoginInput(BaseModel):
+    """Validation model for user login"""
+    email: str = Field(max_length=254)
+    password: str = Field(min_length=8, max_length=100)
+
+    @field_validator('email')
+    @classmethod
+    def validate_email(cls, v: str) -> str:
+        if not re.match(EMAIL_PATTERN, v):
+            raise ValueError("Invalid email format")
+        return v.lower()
+
+class TokenPayload(BaseModel):
+    """Validation model for JWT token payload"""
+    user_id: str
+    username: str
+    exp: datetime
+    iat: datetime
+    scopes: Optional[List[str]] = []
+
+class OAuthInput(BaseModel):
+    """Validation model for OAuth input"""
+    provider: str = Field(pattern='^(google|github|facebook)$')
+    code: str
+    redirect_uri: Optional[str] = None
+
+    @field_validator('provider')
+    @classmethod
+    def validate_provider(cls, v: str) -> str:
+        valid_providers = ['google', 'github', 'facebook']
+        if v.lower() not in valid_providers:
+            raise ValueError(f"Provider must be one of: {', '.join(valid_providers)}")
+        return v.lower()
+
+class AuthResponse(BaseModel):
+    """Validation model for authentication responses"""
+    success: bool
+    token: Optional[str] = None
+    error: Optional[str] = None
+    user: Optional[Dict[str, Union[str, int, bool]]] = None
+
+    @field_validator('error')
+    @classmethod
+    def validate_error_if_not_success(cls, v: Optional[str], info) -> Optional[str]:
+        if not info.data.get('success') and not v:
+            raise ValueError("Error message required when success is False")
+        return v
+
+    @field_validator('token')
+    @classmethod
+    def validate_token_if_success(cls, v: Optional[str], info) -> Optional[str]:
+        if info.data.get('success') and not v:
+            raise ValueError("Token required when success is True")
+        return v
\ No newline at end of file
diff --git a/orm/rbac.py b/orm/rbac.py
new file mode 100644
index 00000000..be22701b
--- /dev/null
+++ b/orm/rbac.py
@@ -0,0 +1,176 @@
+from services.db import REGISTRY, Base, local_session
+from utils.logger import root_logger as logger
+
+from sqlalchemy.types import TypeDecorator
+from sqlalchemy.types import String
+from sqlalchemy import Column, ForeignKey, String, UniqueConstraint
+from sqlalchemy.orm import relationship
+
+class ClassType(TypeDecorator):
+    impl = String
+
+    @property
+    def python_type(self):
+        return NotImplemented
+
+    def process_literal_param(self, value, dialect):
+        return NotImplemented
+
+    def process_bind_param(self, value, dialect):
+        return value.__name__ if isinstance(value, type) else str(value)
+
+    def process_result_value(self, value, dialect):
+        class_ = REGISTRY.get(value)
+        if class_ is None:
+            logger.warn(f"Can't find class <{value}>,find it yourself!", stacklevel=2)
+        return class_
+
+
+class Role(Base):
+    __tablename__ = "role"
+
+    name = Column(String, nullable=False, comment="Role Name")
+    desc = Column(String, nullable=True, comment="Role Description")
+    community = Column(
+        ForeignKey("community.id", ondelete="CASCADE"),
+        nullable=False,
+        comment="Community",
+    )
+    permissions = relationship(lambda: Permission)
+
+    @staticmethod
+    def init_table():
+        with local_session() as session:
+            r = session.query(Role).filter(Role.name == "author").first()
+            if r:
+                Role.default_role = r
+                return
+
+        r1 = Role.create(
+            name="author",
+            desc="Role for an author",
+            community=1,
+        )
+
+        session.add(r1)
+
+        Role.default_role = r1
+
+        r2 = Role.create(
+            name="reader",
+            desc="Role for a reader",
+            community=1,
+        )
+
+        session.add(r2)
+
+        r3 = Role.create(
+            name="expert",
+            desc="Role for an expert",
+            community=1,
+        )
+
+        session.add(r3)
+
+        r4 = Role.create(
+            name="editor",
+            desc="Role for an editor",
+            community=1,
+        )
+
+        session.add(r4)
+
+
+class Operation(Base):
+    __tablename__ = "operation"
+    name = Column(String, nullable=False, unique=True, comment="Operation Name")
+
+    @staticmethod
+    def init_table():
+        with local_session() as session:
+            for name in ["create", "update", "delete", "load"]:
+                """
+                * everyone can:
+                    - load shouts
+                    - load topics
+                    - load reactions
+                    - create an account to become a READER
+                * readers can:
+                    - update and delete their account
+                    - load chats
+                    - load messages
+                    - create reaction of some shout's author allowed kinds
+                    - create shout to become an AUTHOR
+                * authors can:
+                    - update and delete their shout
+                    - invite other authors to edit shout and chat
+                    - manage allowed reactions for their shout
+                * pros can:
+                    - create/update/delete their community
+                    - create/update/delete topics for their community
+
+                """
+                op = session.query(Operation).filter(Operation.name == name).first()
+                if not op:
+                    op = Operation.create(name=name)
+                    session.add(op)
+            session.commit()
+
+
+class Resource(Base):
+    __tablename__ = "resource"
+    resourceClass = Column(String, nullable=False, unique=True, comment="Resource class")
+    name = Column(String, nullable=False, unique=True, comment="Resource name")
+    # TODO: community = Column(ForeignKey())
+
+    @staticmethod
+    def init_table():
+        with local_session() as session:
+            for res in [
+                "shout",
+                "topic",
+                "reaction",
+                "chat",
+                "message",
+                "invite",
+                "community",
+                "user",
+            ]:
+                r = session.query(Resource).filter(Resource.name == res).first()
+                if not r:
+                    r = Resource.create(name=res, resourceClass=res)
+                    session.add(r)
+            session.commit()
+
+
+class Permission(Base):
+    __tablename__ = "permission"
+    __table_args__ = (
+        UniqueConstraint("role", "operation", "resource"),
+        {"extend_existing": True},
+    )
+
+    role: Column = Column(ForeignKey("role.id", ondelete="CASCADE"), nullable=False, comment="Role")
+    operation: Column = Column(
+        ForeignKey("operation.id", ondelete="CASCADE"),
+        nullable=False,
+        comment="Operation",
+    )
+    resource: Column = Column(
+        ForeignKey("resource.id", ondelete="CASCADE"),
+        nullable=False,
+        comment="Resource",
+    )
+
+
+# if __name__ == "__main__":
+#     Base.metadata.create_all(engine)
+#     ops = [
+#         Permission(role=1, operation=1, resource=1),
+#         Permission(role=1, operation=2, resource=1),
+#         Permission(role=1, operation=3, resource=1),
+#         Permission(role=1, operation=4, resource=1),
+#         Permission(role=2, operation=4, resource=1),
+#     ]
+#     global_session.add_all(ops)
+#     global_session.commit()
\ No newline at end of file
diff --git a/orm/user.py b/orm/user.py
new file mode 100644
index 00000000..c31c7913
--- /dev/null
+++ b/orm/user.py
@@ -0,0 +1,105 @@
+from sqlalchemy import JSON as JSONType
+from sqlalchemy import Boolean, Column, DateTime, ForeignKey, Integer, String, func
+from sqlalchemy.orm import relationship
+
+from services.db import Base, local_session
+from orm.rbac import Role
+
+
+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 AuthorFollower(Base):
+    __tablename__ = "author_follower"
+
+    id = None
+    follower: Column = Column(ForeignKey("user.id"), primary_key=True, index=True)
+    author: Column = Column(ForeignKey("user.id"), primary_key=True, index=True)
+    createdAt = Column(
+        DateTime(timezone=True), nullable=False, server_default=func.now(), comment="Created at"
+    )
+    auto = Column(Boolean, nullable=False, default=False)
+
+
+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")
+    muted = Column(Boolean, default=False)
+    emailConfirmed = Column(Boolean, default=False)
+    createdAt = Column(
+        DateTime(timezone=True), nullable=False, server_default=func.now(), comment="Created at"
+    )
+    lastSeen = Column(
+        DateTime(timezone=True), nullable=False, server_default=func.now(), comment="Was online at"
+    )
+    deletedAt = Column(DateTime(timezone=True), nullable=True, comment="Deleted at")
+    links = Column(JSONType, nullable=True, comment="Links")
+    oauth = Column(String, nullable=True)
+    ratings = relationship(UserRating, foreign_keys=UserRating.user)
+    roles = relationship(lambda: Role, secondary=UserRole.__tablename__)
+    oid = Column(String, nullable=True)
+
+    @staticmethod
+    def init_table():
+        with local_session() as session:
+            default = session.query(User).filter(User.slug == "anonymous").first()
+            if not default:
+                default_dict = {
+                    "email": "noreply@discours.io",
+                    "username": "noreply@discours.io",
+                    "name": "Аноним",
+                    "slug": "anonymous",
+                }
+                default = User.create(**default_dict)
+                session.add(default)
+                discours_dict = {
+                    "email": "welcome@discours.io",
+                    "username": "welcome@discours.io",
+                    "name": "Дискурс",
+                    "slug": "discours",
+                }
+                discours = User.create(**discours_dict)
+                session.add(discours)
+                session.commit()
+            User.default_user = default
+
+    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))
\ No newline at end of file
diff --git a/pyproject.toml b/pyproject.toml
index f46a3dc6..ccd0bcf0 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -26,6 +26,8 @@ fakeredis = "^2.25.1"
 pydantic = "^2.9.2"
 jwt = "^1.3.1"
 authlib = "^1.3.2"
+passlib = "^1.7.4"
+bcrypt = "^4.2.1"
 
 
 [tool.poetry.group.dev.dependencies]
@@ -34,13 +36,15 @@ isort = "^5.13.2"
 pydantic = "^2.9.2"
 pytest = "^8.3.4"
 mypy = "^1.15.0"
+pytest-asyncio = "^0.23.5"
+pytest-cov = "^4.1.0"
 
 [build-system]
 requires = ["poetry-core>=1.0.0"]
 build-backend = "poetry.core.masonry.api"
 
 [tool.pyright]
-venvPath = "."
+venvPath = "venv"
 venv = "venv"
 
 [tool.isort]
@@ -49,5 +53,10 @@ include_trailing_comma = true
 force_grid_wrap = 0
 line_length = 120
 
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+pythonpath = ["."]
+venv = "venv"
+
 [tool.ruff]
 line-length = 120
diff --git a/resolvers/draft.py b/resolvers/draft.py
index 9516e1fb..c780de85 100644
--- a/resolvers/draft.py
+++ b/resolvers/draft.py
@@ -1,16 +1,21 @@
 import time
-from importlib import invalidate_caches
-
+from orm.topic import Topic
 from sqlalchemy import select
+from sqlalchemy.sql import and_
 
-from cache.cache import invalidate_shout_related_cache, invalidate_shouts_cache
+from cache.cache import (
+    cache_author, cache_by_id, cache_topic, 
+    invalidate_shout_related_cache, invalidate_shouts_cache
+)
 from orm.author import Author
 from orm.draft import Draft
-from orm.shout import Shout
+from orm.shout import Shout, ShoutAuthor, ShoutTopic
 from services.auth import login_required
 from services.db import local_session
 from services.schema import mutation, query
 from utils.logger import root_logger as logger
+from services.notify import notify_shout
+from services.search import search_service
 
 
 @query.field("load_drafts")
@@ -119,9 +124,10 @@ async def unpublish_draft(_, info, draft_id: int):
 @login_required
 async def publish_shout(_, info, shout_id: int, draft=None):
     """Publish draft as a shout or update existing shout.
-
+    
     Args:
-        session: SQLAlchemy session to use for database operations
+        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", {})
@@ -130,16 +136,25 @@ async def publish_shout(_, info, shout_id: int, draft=None):
         return {"error": "User ID and author ID are required"}
 
     try:
-        # Use proper SQLAlchemy query
         with local_session() as session:
+            # Находим черновик если не передан
             if not draft:
                 find_draft_stmt = select(Draft).where(Draft.shout == shout_id)
                 draft = session.execute(find_draft_stmt).scalar_one_or_none()
+                if not draft:
+                    return {"error": "Draft not found"}
 
             now = int(time.time())
-
+            
+            # Находим существующую публикацию или создаем новую
+            shout = None
+            was_published = False
+            if shout_id:
+                shout = session.query(Shout).filter(Shout.id == shout_id).first()
+                was_published = shout and shout.published_at is not None
+            
             if not shout:
-                # Create new shout from draft
+                # Создаем новую публикацию
                 shout = Shout(
                     body=draft.body,
                     slug=draft.slug,
@@ -155,15 +170,11 @@ async def publish_shout(_, info, shout_id: int, draft=None):
                     seo=draft.seo,
                     created_by=author_id,
                     community=draft.community,
-                    authors=draft.authors.copy(),  # Create copies of relationships
-                    topics=draft.topics.copy(),
                     draft=draft.id,
                     deleted_at=None,
                 )
             else:
-                # Update existing shout
-                shout.authors = draft.authors.copy()
-                shout.topics = draft.topics.copy()
+                # Обновляем существующую публикацию
                 shout.draft = draft.id
                 shout.created_by = author_id
                 shout.title = draft.title
@@ -178,24 +189,78 @@ async def publish_shout(_, info, shout_id: int, draft=None):
                 shout.lang = draft.lang
                 shout.seo = draft.seo
 
+            # Обновляем временные метки
             shout.updated_at = now
-            shout.published_at = now
+            
+            # Устанавливаем published_at только если это новая публикация
+            # или публикация была ранее снята с публикации
+            if not was_published:
+                shout.published_at = now
+            
             draft.updated_at = now
             draft.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)
+
+            # 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()
+            return {"shout": shout}
 
-        invalidate_shout_related_cache(shout)
-        invalidate_shouts_cache()
-        return {"shout": shout}
     except Exception as e:
-        import traceback
-
-        logger.error(f"Failed to publish shout: {e}")
-        logger.error(traceback.format_exc())
-        session.rollback()
-        return {"error": "Failed to publish shout"}
+        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")
diff --git a/schema/type.graphql b/schema/type.graphql
index df4be71a..a9941a62 100644
--- a/schema/type.graphql
+++ b/schema/type.graphql
@@ -188,6 +188,8 @@ type Topic {
 
 type CommonResult {
   error: String
+  drafts: [Draft]
+  draft: Draft
   slugs: [String]
   shout: Shout
   shouts: [Shout]
diff --git a/services/db.py b/services/db.py
index 8b36c406..394de75d 100644
--- a/services/db.py
+++ b/services/db.py
@@ -7,8 +7,7 @@ from typing import Any, Callable, Dict, TypeVar
 
 import sqlalchemy
 from sqlalchemy import JSON, Column, Engine, Integer, create_engine, event, exc, func, inspect
-from sqlalchemy.ext.declarative import declarative_base
-from sqlalchemy.orm import Session, configure_mappers
+from sqlalchemy.orm import Session, configure_mappers, declarative_base
 from sqlalchemy.sql.schema import Table
 
 from settings import DB_URL
diff --git a/services/pretopic.py b/services/pretopic.py
new file mode 100644
index 00000000..b6cd60c8
--- /dev/null
+++ b/services/pretopic.py
@@ -0,0 +1,179 @@
+import concurrent.futures
+from typing import Dict, Tuple, List
+from txtai.embeddings import Embeddings
+from services.logger import root_logger as logger
+
+class TopicClassifier:
+    def __init__(self, shouts_by_topic: Dict[str, str], publications: List[Dict[str, str]]):
+        """
+        Инициализация классификатора тем и поиска публикаций.
+        Args:
+            shouts_by_topic: Словарь {тема: текст_всех_публикаций}
+            publications: Список публикаций с полями 'id', 'title', 'text'
+        """
+        self.shouts_by_topic = shouts_by_topic
+        self.topics = list(shouts_by_topic.keys())
+        self.publications = publications
+        self.topic_embeddings = None  # Для классификации тем
+        self.search_embeddings = None  # Для поиска публикаций
+        self._initialization_future = None
+        self._executor = concurrent.futures.ThreadPoolExecutor(max_workers=1)
+
+    def initialize(self) -> None:
+        """
+        Асинхронная инициализация векторных представлений.
+        """
+        if self._initialization_future is None:
+            self._initialization_future = self._executor.submit(self._prepare_embeddings)
+            logger.info("Векторизация текстов начата в фоновом режиме...")
+
+    def _prepare_embeddings(self) -> None:
+        """
+        Подготавливает векторные представления для тем и поиска.
+        """
+        logger.info("Начинается подготовка векторных представлений...")
+        
+        # Модель для русского языка
+        # TODO: model local caching
+        model_path = "sentence-transformers/paraphrase-multilingual-mpnet-base-v2"
+        
+        # Инициализируем embeddings для классификации тем
+        self.topic_embeddings = Embeddings(path=model_path)
+        topic_documents = [
+            (topic, text) 
+            for topic, text in self.shouts_by_topic.items()
+        ]
+        self.topic_embeddings.index(topic_documents)
+        
+        # Инициализируем embeddings для поиска публикаций
+        self.search_embeddings = Embeddings(path=model_path)
+        search_documents = [
+            (str(pub['id']), f"{pub['title']} {pub['text']}")
+            for pub in self.publications
+        ]
+        self.search_embeddings.index(search_documents)
+        
+        logger.info("Подготовка векторных представлений завершена.")
+
+    def predict_topic(self, text: str) -> Tuple[float, str]:
+        """
+        Предсказывает тему для заданного текста из известного набора тем.
+        Args:
+            text: Текст для классификации
+        Returns:
+            Tuple[float, str]: (уверенность, тема)
+        """
+        if not self.is_ready():
+            logger.error("Векторные представления не готовы. Вызовите initialize() и дождитесь завершения.")
+            return 0.0, "unknown"
+        
+        try:
+            # Ищем наиболее похожую тему
+            results = self.topic_embeddings.search(text, 1)
+            if not results:
+                return 0.0, "unknown"
+            
+            score, topic = results[0]
+            return float(score), topic
+
+        except Exception as e:
+            logger.error(f"Ошибка при определении темы: {str(e)}")
+            return 0.0, "unknown"
+
+    def search_similar(self, query: str, limit: int = 5) -> List[Dict[str, any]]:
+        """
+        Ищет публикации похожие на поисковый запрос.
+        Args:
+            query: Поисковый запрос
+            limit: Максимальное количество результатов
+        Returns:
+            List[Dict]: Список найденных публикаций с оценкой релевантности
+        """
+        if not self.is_ready():
+            logger.error("Векторные представления не готовы. Вызовите initialize() и дождитесь завершения.")
+            return []
+        
+        try:
+            # Ищем похожие публикации
+            results = self.search_embeddings.search(query, limit)
+            
+            # Формируем результаты
+            found_publications = []
+            for score, pub_id in results:
+                # Находим публикацию по id
+                publication = next(
+                    (pub for pub in self.publications if str(pub['id']) == pub_id),
+                    None
+                )
+                if publication:
+                    found_publications.append({
+                        **publication,
+                        'relevance': float(score)
+                    })
+            
+            return found_publications
+
+        except Exception as e:
+            logger.error(f"Ошибка при поиске публикаций: {str(e)}")
+            return []
+
+    def is_ready(self) -> bool:
+        """
+        Проверяет, готовы ли векторные представления.
+        """
+        return self.topic_embeddings is not None and self.search_embeddings is not None
+
+    def wait_until_ready(self) -> None:
+        """
+        Ожидает завершения подготовки векторных представлений.
+        """
+        if self._initialization_future:
+            self._initialization_future.result()
+
+    def __del__(self):
+        """
+        Очистка ресурсов при удалении объекта.
+        """
+        if self._executor:
+            self._executor.shutdown(wait=False)
+
+# Пример использования:
+"""
+shouts_by_topic = {
+    "Спорт": "... большой текст со всеми спортивными публикациями ...",
+    "Технологии": "... большой текст со всеми технологическими публикациями ...",
+    "Политика": "... большой текст со всеми политическими публикациями ..."
+}
+
+publications = [
+    {
+        'id': 1,
+        'title': 'Новый процессор AMD',
+        'text': 'Компания AMD представила новый процессор...'
+    },
+    {
+        'id': 2,
+        'title': 'Футбольный матч',
+        'text': 'Вчера состоялся решающий матч...'
+    }
+]
+
+# Создание классификатора
+classifier = TopicClassifier(shouts_by_topic, publications)
+classifier.initialize()
+classifier.wait_until_ready()
+
+# Определение темы текста
+text = "Новый процессор показал высокую производительность"
+score, topic = classifier.predict_topic(text)
+print(f"Тема: {topic} (уверенность: {score:.4f})")
+
+# Поиск похожих публикаций
+query = "процессор AMD производительность"
+similar_publications = classifier.search_similar(query, limit=3)
+for pub in similar_publications:
+    print(f"\nНайдена публикация (релевантность: {pub['relevance']:.4f}):")
+    print(f"Заголовок: {pub['title']}")
+    print(f"Текст: {pub['text'][:100]}...")
+"""
+
diff --git a/settings.py b/settings.py
index a2a18efb..2760140b 100644
--- a/settings.py
+++ b/settings.py
@@ -5,7 +5,7 @@ PORT = 8000
 DB_URL = (
     environ.get("DATABASE_URL", "").replace("postgres://", "postgresql://")
     or environ.get("DB_URL", "").replace("postgres://", "postgresql://")
-    or "sqlite:///discoursio-db.sqlite3"
+    or "sqlite:///discoursio.db"
 )
 REDIS_URL = environ.get("REDIS_URL") or "redis://127.0.0.1"
 AUTH_URL = environ.get("AUTH_URL") or ""
@@ -15,3 +15,9 @@ MODE = "development" if "dev" in sys.argv else "production"
 
 ADMIN_SECRET = environ.get("AUTH_SECRET") or "nothing"
 WEBHOOK_SECRET = environ.get("WEBHOOK_SECRET") or "nothing-else"
+
+# own auth
+ONETIME_TOKEN_LIFE_SPAN = 60 * 60 * 24 * 3  # 3 days
+SESSION_TOKEN_LIFE_SPAN = 60 * 60 * 24 * 30  # 30 days
+JWT_ALGORITHM = "HS256"
+JWT_SECRET_KEY = environ.get("JWT_SECRET") or "nothing-else-jwt-secret-matters"
\ No newline at end of file
diff --git a/tests/conftest.py b/tests/conftest.py
new file mode 100644
index 00000000..b45ef526
--- /dev/null
+++ b/tests/conftest.py
@@ -0,0 +1,55 @@
+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 settings import DB_URL
+
+# Use SQLite for testing
+TEST_DB_URL = "sqlite:///test.db"
+
+@pytest.fixture(scope="session")
+def event_loop():
+    """Create an instance of the default event loop for the test session."""
+    loop = asyncio.get_event_loop_policy().new_event_loop()
+    yield loop
+    loop.close()
+
+@pytest.fixture(scope="session")
+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_engine):
+    """Create a new database session for a test."""
+    connection = test_engine.connect()
+    transaction = connection.begin()
+    session = Session(bind=connection)
+    
+    yield session
+    
+    session.close()
+    transaction.rollback()
+    connection.close()
+
+@pytest.fixture
+async def redis_client():
+    """Create a test Redis client."""
+    await redis.connect()
+    yield redis
+    await redis.disconnect()
+
+@pytest.fixture
+def test_client():
+    """Create a TestClient instance."""
+    return TestClient(app) 
\ No newline at end of file
diff --git a/tests/test_drafts.py b/tests/test_drafts.py
new file mode 100644
index 00000000..5cd44a76
--- /dev/null
+++ b/tests/test_drafts.py
@@ -0,0 +1,95 @@
+import pytest
+from orm.shout import Shout
+from orm.author import Author
+
+@pytest.fixture
+def test_author(db_session):
+    """Create a test author."""
+    author = Author(
+        name="Test Author",
+        slug="test-author",
+        user="test-user-id"
+    )
+    db_session.add(author)
+    db_session.commit()
+    return author
+
+@pytest.fixture
+def test_shout(db_session):
+    """Create test shout with required fields."""
+    author = Author(name="Test Author", slug="test-author", user="test-user-id")
+    db_session.add(author)
+    db_session.flush()
+
+    shout = Shout(
+        title="Test Shout",
+        slug="test-shout",
+        created_by=author.id,  # Обязательное поле
+        body="Test body",
+        layout="article",
+        lang="ru"
+    )
+    db_session.add(shout)
+    db_session.commit()
+    return shout
+
+@pytest.mark.asyncio
+async def test_create_shout(test_client, db_session, test_author):
+    """Test creating a new shout."""
+    response = test_client.post(
+        "/",
+        json={
+            "query": """
+            mutation CreateDraft($input: DraftInput!) {
+                create_draft(input: $input) {
+                    error
+                    draft {
+                        id
+                        title
+                        body
+                    }
+                }
+            }
+            """,
+            "variables": {
+                "input": {
+                    "title": "Test Shout",
+                    "body": "This is a test shout",
+                }
+            }
+        }
+    )
+    
+    assert response.status_code == 200
+    data = response.json()
+    assert "errors" not in data
+    assert data["data"]["create_draft"]["draft"]["title"] == "Test Shout"
+
+@pytest.mark.asyncio
+async def test_load_drafts(test_client, db_session):
+    """Test retrieving a shout."""
+    response = test_client.post(
+        "/",
+        json={
+            "query": """
+            query {
+                load_drafts {
+                    error
+                    drafts {
+                        id
+                        title
+                        body
+                    }
+                }
+            }
+            """,
+            "variables": {
+                "slug": "test-shout"
+            }
+        }
+    )
+    
+    assert response.status_code == 200
+    data = response.json()
+    assert "errors" not in data
+    assert data["data"]["load_drafts"]["drafts"] == [] 
\ No newline at end of file
diff --git a/tests/test_reactions.py b/tests/test_reactions.py
new file mode 100644
index 00000000..0a37a20d
--- /dev/null
+++ b/tests/test_reactions.py
@@ -0,0 +1,64 @@
+import pytest
+from orm.reaction import Reaction, ReactionKind
+from orm.shout import Shout
+from orm.author import Author
+from datetime import datetime
+
+@pytest.fixture
+def test_setup(db_session):
+    """Set up test data."""
+    now = int(datetime.now().timestamp())
+    author = Author(name="Test Author", slug="test-author", user="test-user-id")
+    db_session.add(author)
+    db_session.flush()
+    
+    shout = Shout(
+        title="Test Shout", 
+        slug="test-shout",
+        created_by=author.id,
+        body="This is a test shout",
+        layout="article",
+        lang="ru",
+        community=1,
+        created_at=now,
+        updated_at=now
+    )
+    db_session.add_all([author, shout])
+    db_session.commit()
+    return {"author": author, "shout": shout}
+
+@pytest.mark.asyncio
+async def test_create_reaction(test_client, db_session, test_setup):
+    """Test creating a reaction on a shout."""
+    response = test_client.post(
+        "/",
+        json={
+            "query": """
+            mutation CreateReaction($reaction: ReactionInput!) {
+                create_reaction(reaction: $reaction) {
+                    error
+                    reaction {
+                        id
+                        kind
+                        body
+                        created_by {
+                            name
+                        }
+                    }
+                }
+            }
+            """,
+            "variables": {
+                    "reaction": {
+                        "shout": test_setup["shout"].id,
+                        "kind": ReactionKind.LIKE.value,
+                        "body": "Great post!"
+                    }
+            }
+        }
+    )
+    
+    assert response.status_code == 200
+    data = response.json()
+    assert "error" not in data
+    assert data["data"]["create_reaction"]["reaction"]["kind"] == ReactionKind.LIKE.value 
\ No newline at end of file
diff --git a/tests/test_shouts.py b/tests/test_shouts.py
new file mode 100644
index 00000000..66f3b1e9
--- /dev/null
+++ b/tests/test_shouts.py
@@ -0,0 +1,83 @@
+import pytest
+from orm.author import Author
+from orm.shout import Shout
+from datetime import datetime
+
+@pytest.fixture
+def test_shout(db_session):
+    """Create test shout with required fields."""
+    now = int(datetime.now().timestamp())
+    author = Author(name="Test Author", slug="test-author", user="test-user-id")
+    db_session.add(author)
+    db_session.flush()
+    
+    now = int(datetime.now().timestamp())
+
+    shout = Shout(
+        title="Test Shout",
+        slug="test-shout",
+        created_by=author.id,
+        body="Test body",
+        layout="article",
+        lang="ru",
+        community=1,
+        created_at=now,
+        updated_at=now
+    )
+    db_session.add(shout)
+    db_session.commit()
+    return shout
+
+@pytest.mark.asyncio
+async def test_get_shout(test_client, db_session):
+    """Test retrieving a shout."""
+    # Создаем автора
+    author = Author(name="Test Author", slug="test-author", user="test-user-id")
+    db_session.add(author)
+    db_session.flush()
+    now = int(datetime.now().timestamp())
+    
+    # Создаем публикацию со всеми обязательными полями
+    shout = Shout(
+        title="Test Shout",
+        body="This is a test shout",
+        slug="test-shout",
+        created_by=author.id,
+        layout="article",
+        lang="ru",
+        community=1,
+        created_at=now,
+        updated_at=now
+    )
+    db_session.add(shout)
+    db_session.commit()
+    
+    response = test_client.post(
+        "/",
+        json={
+            "query": """
+            query GetShout($slug: String!) {
+                get_shout(slug: $slug) {
+                    id
+                    title
+                    body
+                    created_at
+                    updated_at
+                    created_by {
+                        id
+                        name
+                        slug
+                    }
+                }
+            }
+            """,
+            "variables": {
+                "slug": "test-shout"
+            }
+        }
+    )
+    
+    data = response.json()
+    assert response.status_code == 200
+    assert "errors" not in data
+    assert data["data"]["get_shout"]["title"] == "Test Shout" 
\ No newline at end of file
diff --git a/tests/test_validations.py b/tests/test_validations.py
new file mode 100644
index 00000000..2cffae7a
--- /dev/null
+++ b/tests/test_validations.py
@@ -0,0 +1,101 @@
+import pytest
+from datetime import datetime, timedelta
+from pydantic import ValidationError
+
+from auth.validations import (
+    AuthInput,
+    UserRegistrationInput,
+    UserLoginInput,
+    TokenPayload,
+    OAuthInput,
+    AuthResponse
+)
+
+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"}
+            ) 
\ No newline at end of file