🌊 hexastack-flow

Hexagonal workflow engine adapter integrating hexaflow with Hexastack CQRS, Relational Persistence, and Event Streaming.

Part of the Hexastack Framework · Powered by Hexaflow · Governed by Hexaqual.

🎯 Architectural Intent

hexastack-flow provides first-class hexagonal adapters connecting the lightweight hexaflow workflow engine to Hexastack:

1. CQRS Bridge: CommandStep and QueryStep allow workflow steps to dispatch typed commands and queries directly onto the CommandBus and QueryBus, inheriting all CQRS middleware (auth, tracing, logging, metrics, retries).
2. Durable Distributed Sagas: Workflows natively subsume the Saga pattern by pairing forward command execution stages with compensating rollback stages, backed by durable checkpoints.
3. Enterprise Persistence: SqlAlchemyWorkflowStore persists workflow execution records and step checkpoints directly to PostgreSQL/SQLite/MySQL within transactional boundaries.
4. Event Streaming: Emits domain events (WorkflowStartedEvent, WorkflowStepCompletedEvent, WorkflowSuspendedEvent, WorkflowCompletedEvent, WorkflowAbortedEvent) wrapped in standard CloudEventEnvelopes through hexastack-events.
5. DevTools Visualizer: Backing engine for NiceGUI and Textual DAG monitoring.

🔄 Implementing the Saga Pattern as a Workflow

Historically, the Saga Pattern structures a distributed business transaction across multiple microservices or bounded contexts as a sequence of local transactions:

• Forward transactions: $T_1, T_2, \dots, T_n$
• Compensating transactions: $C_{k-1}, \dots, C_1$ executed in reverse (LIFO) order if step $k$ fails.

In hexastack-flow, you do not need a separate, restricted saga engine. A Saga is simply a workflow DAG that dispatches CQRS commands and executes compensating commands on failure.

Example: Trip Booking Saga with CQRS & Hexaflow

from hexaflow import Workflow
from hexastack_cqrs.ports.buses import CommandBusPort
from hexastack_flow.adapters.cqrs.runner import CqrsWorkflowRunner
from hexastack_flow.adapters.cqrs.steps import as_command_step

# 1. Define CQRS commands for forward actions & compensations
from my_domain.commands import (
    BookFlightCommand,
    CancelFlightCommand,
    BookHotelCommand,
    CancelHotelCommand,
    ChargePaymentCommand,
)


def build_trip_booking_workflow(bus: CommandBusPort, customer_id: str) -> Workflow:
    """Construct a durable distributed saga as a hexaflow Workflow."""
    workflow = (
        Workflow("trip-booking-saga")
        # Step 1: Reserve Flight
        .stage(
            "flight",
            as_command_step(
                lambda ctx: BookFlightCommand(
                    customer_id=customer_id, destination="JFK"
                ),
                bus,
                name="book_flight",
            ),
        )
        # Step 2: Reserve Hotel (runs after flight is confirmed)
        .stage(
            "hotel",
            as_command_step(
                lambda ctx: BookHotelCommand(
                    customer_id=customer_id, room_type="deluxe"
                ),
                bus,
                name="book_hotel",
            ),
            depends_on=["flight"],
        )
        # Step 3: Process Payment
        .stage(
            "payment",
            as_command_step(
                lambda ctx: ChargePaymentCommand(
                    customer_id=customer_id, amount_cents=50000
                ),
                bus,
                name="charge_payment",
            ),
            depends_on=["hotel"],
        )
    )
    return workflow


# 2. Execute via CqrsWorkflowRunner (with durable DB checkpoints)
runner = CqrsWorkflowRunner(engine=engine)
result = await runner.run(workflow.build())

Why Workflows are Superior to Linear Sagas

📦 Installation

pip install hexastack[flow]

Or standalone:

pip install hexastack-flow

📄 License

Apache 2.0