operaton-tasks

External task library and worker for https://operaton.org/

Use this package when you want to implement Operaton external service task workers in Python.

Installation

Install as a Python dependency:

pip install operaton-tasks

Install with CLI support:

pip install "operaton-tasks[cli]"

What this package provides

• A decorator API for registering handlers by topic.
• A long-polling worker that fetches, locks, executes, and completes/fails external tasks.
• Optional CLI command to run the worker app.
• Built-in health endpoint at /healthz.

Authentication

Requests to Operaton REST API can be authenticated in two supported ways.

Option 1: Static Authorization header

Set ENGINE_REST_AUTHORIZATION to the full header value.

Example:

export ENGINE_REST_AUTHORIZATION="Basic base64-user-pass"
# or
export ENGINE_REST_AUTHORIZATION="Bearer my-static-token"

Option 2: OAuth2 client credentials

Set the OAuth2 settings:

export OAUTH2_CLIENT_ID="my-client-id"
export OAUTH2_CLIENT_SECRET="my-client-secret"
export OAUTH2_TOKEN_URL="https://idp.example.com/realms/operaton/protocol/openid-connect/token"
export OAUTH2_SCOPES="scope-a scope-b"  # optional, space-separated

When OAuth2 is configured, the worker automatically fetches and refreshes bearer tokens.

Authentication precedence

Authorization is resolved in this order:

1. Explicit authorization argument passed to session/request helpers.
2. OAuth2 client credentials token (if fully configured).
3. ENGINE_REST_AUTHORIZATION.

With OAuth2 enabled, a 401 response triggers one token invalidation and one retry.

Configuration

Main environment variables:

• ENGINE_REST_BASE_URL (default: http://localhost:8080/engine-rest)
• ENGINE_REST_AUTHORIZATION (optional)
• OAUTH2_CLIENT_ID (optional)
• OAUTH2_CLIENT_SECRET (optional)
• OAUTH2_TOKEN_URL (optional)
• OAUTH2_SCOPES (optional)
• ENGINE_REST_TIMEOUT_SECONDS (default: 20)
• ENGINE_REST_POLL_TTL_SECONDS (default: 10)
• ENGINE_REST_LOCK_TTL_SECONDS (default: 30)
• TASKS_WORKER_ID (default: operaton-tasks-client)
• TASKS_HEARTBEAT_TOPIC (default: operaton.tasks.heartbeat)
• TASKS_MODULE (used by CLI/app startup to load your handlers module)
• TASKS_LIMIT (default: 0, stop after processing this many tasks)
• TASKS_RUN_TIMEOUT_SECONDS (default: 0, exit after this many seconds)
• LOG_LEVEL (default: DEBUG)

Usage as standalone worker (CLI)

This is the quickest way to run your handlers as a process.

1. Create a handlers module, for example my_tasks.py.
2. Run operaton-tasks and pass your module path.

Example handler module:

from operaton.tasks.types import CompleteExternalTaskDto
from operaton.tasks.types import ExternalTaskComplete
from operaton.tasks.types import LockedExternalTaskDto
from operaton.tasks.types import VariableValueDto
from operaton.tasks.types import VariableValueType
import operaton.tasks


@operaton.tasks.register("hello-world", localVariables=True)
async def hello_world(task: LockedExternalTaskDto) -> ExternalTaskComplete:
	return ExternalTaskComplete(
		task=task,
		response=CompleteExternalTaskDto(
			workerId=task.workerId,
			localVariables={
				"message": VariableValueDto(
					value="Hello World",
					type=VariableValueType.String,
				),
			},
		),
	)

Run it:

operaton-tasks serve ./my_tasks.py -- --host 0.0.0.0 --port 8000

Pass auth options directly on the CLI if preferred:

operaton-tasks serve ./my_tasks.py \
  --base-url http://localhost:8080/engine-rest \
  --authorization "Basic base64-user-pass"

Or with OAuth2:

operaton-tasks serve ./my_tasks.py \
  --base-url http://localhost:8080/engine-rest \
  --oauth2-client-id my-client-id \
  --oauth2-client-secret my-client-secret \
  --oauth2-token-url https://idp.example.com/realms/operaton/protocol/openid-connect/token \
  --oauth2-scopes "scope-a scope-b"

Note: arguments after -- are passed through to uvicorn.

If you want the CLI to process a bounded number of tasks and then exit, pass the worker controls directly:

operaton-tasks serve ./my_tasks.py --limit 10 --run-timeout 60

When either --limit or --run-timeout is set, the CLI loads the handler module, runs the worker in one-shot mode, and exits when the limit or timeout is reached.

Usage as a library/dependency

Use this mode when you already have your own Python app/runtime and want to embed the worker.

A minimal worker process without CLI

import asyncio
import operaton.tasks
from operaton.tasks.api import external_task_worker
from operaton.tasks.api import handlers
from operaton.tasks.types import CompleteExternalTaskDto
from operaton.tasks.types import ExternalTaskComplete
from operaton.tasks.types import LockedExternalTaskDto


@operaton.tasks.task("hello-world")
async def hello_world(task: LockedExternalTaskDto) -> ExternalTaskComplete:
	return ExternalTaskComplete(
		task=task,
		response=CompleteExternalTaskDto(workerId=task.workerId),
	)


if __name__ == "__main__":
	asyncio.run(external_task_worker(handlers))

Configure auth and connection through environment variables, or set values programmatically through operaton.tasks.settings before starting the worker.

Embed into an existing FastAPI app

If you want to reuse built-in routes such as /healthz and run the worker in your own FastAPI service:

import asyncio
from contextlib import asynccontextmanager
from fastapi import FastAPI
from operaton.tasks.api import external_task_worker
from operaton.tasks.api import handlers
from operaton.tasks.api import router


@asynccontextmanager
async def lifespan(app: FastAPI):
	asyncio.create_task(external_task_worker(handlers))
	yield


app = FastAPI(lifespan=lifespan)
app.include_router(router)

Handler contract

• Handlers are async functions registered with @task(topic) or @register(topic).
• Input is LockedExternalTaskDto.
• Return ExternalTaskComplete for success (with CompleteExternalTaskDto or ExternalTaskBpmnError).
• If your handler raises an exception, it is converted to ExternalTaskFailure automatically.

Health endpoint

GET /healthz is included on the built-in router.

• Without heartbeat activity, it checks Operaton engine reachability.
• With heartbeat activity, it verifies recent heartbeat timestamps.

Development

This project uses devenv.sh for reproducible development environments with Nix.

Local Development

Enter the development shell:

devenv shell
# or
make shell

Running Tests Locally

Run all checks (linting, type checking, unit tests):

make test

Run individual checks:

make check           # treefmt, flake8, mypy
make test-pytest     # Unit tests only
make devenv-test     # Integration tests with Operaton + Keycloak services

Start background services (Operaton + Keycloak) for manual testing:

make devenv-up       # Start services
make devenv-down     # Stop services

Watch mode for development:

make watch           # Run app in reload mode
make watch-tests     # Continuously run mypy + pytest

CI/CD Pipelines

This project includes GitHub Actions and GitLab CI pipelines configured with devenv.sh best practices.

GitHub Actions (.github/workflows/):

• ci.yml: Runs on push, pull requests, and manual dispatch. Executes linting, type checking, and integration tests.
• release.yml: Publishes to PyPI on version tags (v*).

GitLab CI (.gitlab-ci.yml):

• Lint stage: treefmt, flake8, mypy
• Test stage: Unit and integration tests with live Operaton + Keycloak services
• Build & Publish stages: Triggered on tags

Both pipelines use:

• Nix for reproducible environments (via devenv)
• Magic Nix Cache (GitHub) or Nix cache directories (GitLab) for faster builds
• devenv shell -- make <target> to ensure identical environments between local development and CI

For PyPI publishing, configure:

• GitHub Actions: Use trusted publishers (OIDC) or set PYPI_API_TOKEN secret
• GitLab CI: Set PYPI_TOKEN CI/CD variable with your PyPI token