Skip to main content

stage-events-client

PyPI - Version PyPI - Python Version GitHub Actions Workflow Status Code coverage

Python client for sending structured-mode CloudEvents 1.0.2 to the Stage Events API.

The package provides Pydantic models for the supported workflow events and synchronous and asynchronous clients built on HTTPX.

Requirements

  • Python 3.10 or newer
  • A Stage Events API endpoint
  • A bearer token when authentication is enabled

Installation

python -m pip install stage-events-client

The command-line interface dependencies are optional. Enable the CLI explicitly when installing the package:

python -m pip install 'stage-events-client[cli]'

Command-line interface

The send-stage-event executable provides one command for every supported CloudEvent model:

send-stage-event --help
send-stage-event submitted --help

Pass the complete destination URL as the command argument. There is no separate base URL or endpoint-path setting:

send-stage-event submitted \
  https://events.example.com/hooks/cloud-events \
  --source workflows:example-process:submit \
  --subject workflows:2f660c57:example-workflow \
  --data '{"namespace":"workflows","time":"2026-07-18T12:00:00Z"}' \
  --x-kafka-topic workflows.2f660c57.submitted \
  --token your-bearer-token

The available commands are calendar, submitted, dismissed, prepared, completed, failed, piped, staged, and ordered. Each command validates --data against its corresponding Pydantic model before sending the request.

--data accepts inline JSON, a JSON file prefixed with @, or - to read from standard input:

send-stage-event prepared https://events.example.com/cloud-events \
  --source workflows:example-process:prepare \
  --subject workflows:2f660c57:example-workflow \
  --data @prepared-data.json

cat submitted-data.json | send-stage-event submitted \
  https://events.example.com/cloud-events \
  --source workflows:example-process:submit \
  --subject workflows:2f660c57:example-workflow \
  --data -

The partition key defaults to the subject. Override it with --partition-key when required. The --x-kafka-topic option is optional.

The bearer token can be supplied through --token or the STAGE_EVENTS_TOKEN environment variable. Run a command with --help for TLS, timeout, and all other options.

Quick start

Create a client, construct an event, and send it with the required Kafka topic:

from datetime import datetime, timezone
from uuid import uuid4

from stage_events_client import AuthenticatedClient
from stage_events_client.api.default import send_cloud_event
from stage_events_client.models import SubmittedCloudEvent, SubmittedData

client = AuthenticatedClient(
    base_url="https://events.example.com",
    token="your-bearer-token",
)

subject = "workflows:2f660c57:example-workflow"
event = SubmittedCloudEvent(
    source="workflows:example-process:submit",
    subject=subject,
    partitionkey=subject,
    specversion="1.0",
    id=str(uuid4()),
    data=SubmittedData(
        namespace="workflows",
        time=datetime.now(timezone.utc),
    ),
)

with client:
    result = send_cloud_event.sync(
        client=client,
        body=event,
        x_kafka_topic="workflows.2f660c57.submitted",
    )

print(result)

Use Client instead when the API does not require authentication:

from stage_events_client import Client

client = Client(base_url="https://events.example.com")

Detailed responses

sync returns the parsed response body. Use sync_detailed when the status, headers, or raw response body are also needed:

response = send_cloud_event.sync_detailed(
    client=client,
    body=event,
    x_kafka_topic="workflows.2f660c57.submitted",
)

print(response.status_code)
print(response.headers)
print(response.content)
print(response.parsed)

A successful 200 response is parsed as a string. Documented 400 responses are parsed into the corresponding problem-details model. An undocumented status returns None unless raise_on_unexpected_status=True is set on the client, in which case stage_events_client.errors.UnexpectedStatus is raised.

Async usage

Each endpoint has equivalent asyncio and asyncio_detailed functions:

import asyncio


async def main() -> None:
    async with client:
        result = await send_cloud_event.asyncio(
            client=client,
            body=event,
            x_kafka_topic="workflows.2f660c57.submitted",
        )
    print(result)


asyncio.run(main())

Do not use the same client instance in synchronous and asynchronous context managers at the same time.

Supported events

The request body can be any of the following models from stage_events_client.models:

Event type Model
calendar-event CalendarCloudEvent
submitted SubmittedCloudEvent
dismissed DismissedCloudEvent
prepared PreparedCloudEvent
completed CompletedCloudEvent
failed FailedCloudEvent
piped PipedCloudEvent
staged StagedCloudEvent
ordered OrderedCloudEvent

The models validate event-specific payloads, timezone-aware timestamps, semantic versions, and GeoJSON structures where applicable. Additional standard CloudEvent attributes such as id and specversion are preserved by the models.

Event addressing

The API expects these values:

  • source: three colon-separated components, normally {namespace}:{process-id}:{step-name}.
  • subject: three colon-separated components, normally {namespace}:{workflow-uid}:{workflow-name}.
  • partitionkey: usually the same value as subject.
  • x_kafka_topic: a topic in the form {namespace}.{workflow-uid}.{event-suffix}, such as workflows.2f660c57.submitted.

Calendar topics use a duration followed by .calendar, for example workflows.2f660c57.10m.calendar.

Client configuration

Both Client and AuthenticatedClient accept shared headers, cookies, timeout, TLS verification, redirect handling, and additional HTTPX options:

import httpx

from stage_events_client import AuthenticatedClient

client = AuthenticatedClient(
    base_url="https://events.example.com",
    token="your-bearer-token",
    headers={"X-Correlation-ID": "request-id"},
    timeout=httpx.Timeout(30.0),
    verify_ssl="/path/to/ca-bundle.pem",
    follow_redirects=True,
    httpx_args={"proxy": "http://proxy.example.com:8080"},
    raise_on_unexpected_status=True,
)

TLS certificate verification is enabled by default. Setting verify_ssl=False disables server certificate validation and should only be used in controlled development environments.

Clients can be copied with updated settings:

client = client.with_headers({"X-Correlation-ID": "new-request-id"})
client = client.with_cookies({"session": "value"})
client = client.with_timeout(httpx.Timeout(10.0))

An existing httpx.Client or httpx.AsyncClient can also be supplied with set_httpx_client or set_async_httpx_client. Doing so overrides the generated client configuration, so the HTTPX instance must define its own base URL and other required settings.

Development

Install Hatch and run the unit tests:

hatch run test:test

Other useful checks are:

hatch run test:cov
hatch run types:check
hatch run dev:check
hatch run dev:lint

License

License

Metadata

Release files for stage-events-client 1.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for stage-events-client 1.1.0
File Size Uploaded
stage_events_client-1.1.0.tar.gz 60.7 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for stage-events-client 1.1.0
File Interpreter ABI Platform
stage_events_client-1.1.0-py3-none-any.whl Python 3 none any Details

Total release size: 86.8 kB

Release files / stage_events_client-1.1.0.tar.gz

Download URL stage_events_client-1.1.0.tar.gz
Size 60.7 kB
Tags Source
SHA-256 checksum
How to use checksums
fe886e69f33942a699541b337dc3bef1cf56798000ac10d5efffd234355f2fd9
BLAKE2b-256 checksum
How to use checksums
fcf572ea842b77b3af26075e12ee0b23a1e4ac0df1d35cc09baffb3f497c24af
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 2, 2026.

Transparency log

Release files / stage_events_client-1.1.0-py3-none-any.whl

Download URL stage_events_client-1.1.0-py3-none-any.whl
Size 26.1 kB
Tags Python 3
SHA-256 checksum
How to use checksums
d969c153317f72c4e8982f6875d9bafb5326a7bfc2f1a5ac76de071b0fa7d9ff
BLAKE2b-256 checksum
How to use checksums
f0f7d0e5304cd2fb8584889ce5ed743c7e4c8b21666be245f09763e464bef039
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 2, 2026.

Transparency log

Release history Release notifications | RSS feed

1.2.0

2 release files

This release

1.1.0 This release

2 release files

1.0.5

2 release files

1.0.4

2 release files

1.0.3

2 release files

1.0.2

2 release files

1.0.1

2 release files

1.0.0

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page