hexastack-db

SQLAlchemy 2.0+ persistence layer, generic repositories, Unit of Work, declarative mixins, and Alembic migrations for Hexastack.

1. Overview & Capabilities

hexastack-db provides a robust, decoupled database layer supporting both synchronous and asynchronous workflows:

• Generic Repositories: SqlAlchemyRepository[T, ID] and AsyncSqlAlchemyRepository[T, ID] implementing Repository[E, ID] and AsyncRepository[E, ID].
• Unit of Work: SqlAlchemyUnitOfWork and AsyncSqlAlchemyUnitOfWork providing contextual transaction boundaries (commit(), rollback()).
• Declarative Base & Mixins:
  • HexastackBase: Shared DeclarativeBase for unified metadata management across applications and migration autogen.
  • UuidPrimaryKeyMixin: Portable Python-side uuid4() stored as a 36-character string primary key.
  • TimestampMixin: Automatic UTC created_at and updated_at timestamps.
• Alembic Migration Engine: Programmatic migration execution (upgrade, downgrade, revision, current, history, stamp) and CLI integration without requiring manual alembic.ini authoring.
• Auto Table Creation: Schema discovery and create_all() execution on bootstrap via register_metadata().

2. Package Anatomy & Key Components

hexastack_db/
├── domain/          # DatabaseError, EntityNotFoundError, UniqueConstraintViolationError
├── adapters/        # SqlAlchemyRepository, AsyncSqlAlchemyRepository, SqlAlchemyUnitOfWork, AsyncSqlAlchemyUnitOfWork
└── infra/
    ├── bootstrap.py # DatabaseBootstrapper (order=15)
    ├── config.py    # HexastackDatabaseConfig
    ├── engine.py    # Engine and sessionmaker factory functions
    ├── mixins.py    # HexastackBase, UuidPrimaryKeyMixin, TimestampMixin
    ├── migrations.py# Alembic migration helpers
    └── registries/  # metadata.py (register_metadata, get_registered_metadata)

Key Exports

3. Monorepo & Sibling Relationships

graph TD
    subgraph SiblingConsumers ["Sibling Consumers"]
        CQRS["hexastack-cqrs (UnitOfWorkMiddleware)"]
        FASTAPI["hexastack-fastapi (DbSessionMiddleware)"]
        CLI["hexastack (hexastack db <cmd>)"]
    end

    subgraph DatabasePackage ["hexastack-db"]
        BOOT["DatabaseBootstrapper (order=15)"]
        UOW["SqlAlchemyUnitOfWork / AsyncSqlAlchemyUnitOfWork"]
        ENG["Engine & sessionmaker"]
        MIGR["Alembic Migration Engine"]
    end

    subgraph CoreContracts ["hexastack-core"]
        UOW_PORT["UnitOfWorkPort"]
        REPO_PORT["RepositoryPort"]
    end

    BOOT --> UOW
    BOOT --> ENG
    UOW -->|implements| UOW_PORT

    CQRS -. consumes UnitOfWorkPort from DI .-> UOW
    FASTAPI -. consumes sessionmaker from DI .-> ENG
    CLI -. invokes migration helpers .-> MIGR

Explicit Dependencies (Direct)

• hexastack-core: Core ports (UnitOfWorkPort, Repository), exception types, and DI container.
• sqlalchemy>=2.0.38: Core ORM and SQL toolkit.

Implied / Behavioral Relationships (DI-Mediated)

• Unit of Work Binding: Binds UnitOfWorkPort into the DI container at order=15 (before CQRS order=20), allowing UnitOfWorkMiddleware to manage transactions automatically.
• Session Injection: Injects sessionmaker / async_sessionmaker into DI, which hexastack-fastapi resolves in DbSessionMiddleware without requiring a direct package dependency.
• Migration CLI Integration: The hexastack umbrella package exposes CLI subcommands (hexastack db init/upgrade/revision/...) backed by hexastack_db.infra.migrations.

Optional Integrations (Extras)

• [sqlite]: Installs aiosqlite>=0.20.0 for async SQLite support.
• [postgresql]: Installs asyncpg>=0.30.0 and psycopg[binary]>=3.2.0.
• [pgvector]: Installs pgvector>=0.3.0 and PostgreSQL drivers for VectorStorePort support.
• [migrations]: Installs alembic>=1.13.0 for migration commands and autogeneration.
• [all]: Installs all drivers, pgvector, and Alembic.

4. Installation

# Standalone install with SQLite async support
pip install "hexastack-db[sqlite]"

# Standalone with PostgreSQL and Alembic migrations
pip install "hexastack-db[postgresql,migrations]"

# With PostgreSQL vector search (pgvector)
pip install "hexastack-db[pgvector]"

# Via umbrella package
pip install "hexastack[db]"

5. Configuration Reference

[hexastack.db]
# Global Settings
url = "sqlite:///app.db" # "postgresql+asyncpg://user:pass@localhost:5432/dbname"
async_mode = false # Set to true for async engine & sessions
auto_create_tables = false # Runs create_all() on registered metadata at bootstrap
echo = false # Enables raw SQLAlchemy SQL logging
pool_size = 5
max_overflow = 10
pool_timeout = 30
pool_recycle = 1800

# SQLite-Specific Dialect Settings
[hexastack.db.sqlite]
foreign_keys = true # PRAGMA foreign_keys = ON
journal_mode = "WAL" # PRAGMA journal_mode = WAL
busy_timeout_ms = 5000 # PRAGMA busy_timeout = 5000
synchronous = "NORMAL"

# PostgreSQL-Specific Dialect Settings
[hexastack.db.postgres]
search_path = "public"
ssl_mode = "require"
server_side_cursors = false

# PostgreSQL pgvector Configuration
[hexastack.db.vector]
enabled = false # Auto-binds VectorStorePort in DI container
table_name = "hexastack_vectors"
dimension = 1536
distance_strategy = "cosine" # "cosine", "l2", "inner_product"
index_type = "hnsw" # "hnsw", "ivfflat", "none"
m = 16
ef_construction = 64

6. Quickstart Example

from sqlalchemy.orm import Mapped, mapped_column
from hexastack_core.infra.bootstrap import bootstrap
from hexastack_core.ports.unit_of_work import UnitOfWorkPort
from hexastack_db.infra.mixins import HexastackBase, UuidPrimaryKeyMixin, TimestampMixin
from hexastack_db.infra.registries.metadata import register_metadata
from hexastack_db.adapters.repository import SqlAlchemyRepository


# 1. Define Model
class UserRecord(UuidPrimaryKeyMixin, TimestampMixin, HexastackBase):
    __tablename__ = "users"
    email: Mapped[str] = mapped_column(unique=True)


register_metadata(HexastackBase.metadata)

# 2. Bootstrap with auto-table creation
runtime = bootstrap(
    config_overrides={
        "db": {
            "url": "sqlite:///:memory:",
            "auto_create_tables": True,
        }
    }
)

# 3. Use Unit of Work and Repository
uow = runtime.container.get(UnitOfWorkPort)
with uow:
    repo = SqlAlchemyRepository(session=uow.session, model_cls=UserRecord)
    user = UserRecord(email="alice@hexastack.dev")
    repo.add(user)
    uow.commit()

print(f"Created user {user.id} at {user.created_at}")