dbos-inc/dbosify-py

Source: https://github.com/dbos-inc/dbosify-py

DBOSify

DBOSify is a drop-in replacement for Temporal Python that uses Postgres (through DBOS Transact) instead of a Temporal server. This lets you run durable workflows, activities, signals, updates, retries, and recovery without needing any infrastructure except Postgres.

To use this library, import dbosify instead of temporalio and connect your workers and clients to a Postgres database:

Usage

To install:

pip install dbosify

This is a drop-in replacement: simply import dbosify instead of temporalio and connect your clients and workers to a Postgres database instead of a Temporal server. Further documentation here.

import asyncio
import os
from datetime import timedelta

from dbosify import activity, workflow
from dbosify.client import Client
from dbosify.worker import Worker

# Set this to a connection string to your Postgres database
DB_URL = os.environ.get("DBOS_SYSTEM_DATABASE_URL")


@activity.defn
async def compose_greeting(name: str) -> str:
    return f"Hello, {name}!"


@workflow.defn
class GreetingWorkflow:
    @workflow.run
    async def run(self, name: str) -> str:
        return await workflow.execute_activity(
            compose_greeting, name, start_to_close_timeout=timedelta(seconds=10)
        )


async def main() -> None:
    worker = Worker(
        DB_URL,
        task_queue="greetings",
        workflows=[GreetingWorkflow],
        activities=[compose_greeting],
    )
    async with worker:
        async with await Client.connect(DB_URL) as client:
            result = await client.execute_workflow(
                GreetingWorkflow.run, "World", id="greeting-1", task_queue="greetings"
            )
            print(result)  # Hello, World!


if __name__ == "__main__":
    asyncio.run(main())

How It Works

DBOSify runs each Temporal workflow as a Postgres-backed DBOS workflow. A deterministic interpreter runs the workflow (both its main coroutine and its signal, update, and query handlers) on a virtual event loop that only advances when an event arrives. Using DBOS steps and workflow communication primitives, all nondeterministic actions are checkpointed in Postgres before the workflow observes them.

• Activities and timers become DBOS steps and durable sleeps, each checkpointed on completion.
• Signals, updates, and cancellations are durable messages delivered through Postgres using LISTEN/NOTIFY.
• Recovery re-runs the workflow on a new worker: the interpreter replays the same sequence of operations against the recorded checkpoints, so execution resumes where it left off and completes exactly once.
• Namespaces each map to their own Postgres schema; a Client wraps a DBOS client and a Worker wraps the DBOS runtime.

How It’s Tested

As DBOSify is a drop-in replacement for Temporal, its tests cover both correctness and conformance with Temporal using the following strategies:

• Ports of all relevant Temporal Python unit and integration tests
• Ports of relevant Temporal Python sample applications, verifying DBOSify is a drop-in replacement
• New unit and integration tests, with an emphasis on kill-and-recover tests verifying deterministic failure recovery
• Signature parity tests mechanically asserting the public APIs of these libraries are identical (with documented exceptions)

What This Is Not

• No wire protocol compatibility. There is no gRPC wire compatibility. Temporal SDKs in other languages cannot connect. This replaces the Temporal server and Python SDK altogether for Python-only applications.
• No Temporal Web UI, temporal CLI, or tctl. You operate workflows with DBOS’s workflow-management APIs and DBOS Conductor instead.

See this documentation for information on architectural differences and feature compatibility.