Handling GitHub webhooks#

A common use for a GitHub App integration in Safir/FastAPI applications is to handle webhooks from GitHub. With webhooks, your app can be triggered by events like pushes or pull requests to a repository, and even implement a GitHub Check for pull requests. Gidgethub provides a routing system for handling webhooks. This page shows how to integrate Gidgethub into a Safir/FastAPI application and use the resource models in safir.github.

Adding FastAPI handler for GitHub webhooks#

The first step is to add a FastAPI handler for the webhook endpoint. The URL path for this endpoint corresponds to the webhook callback URL you set up in the GitHub App’s settings.

handlers.py#
import httpx
from fastapi import APIRouter, Depends, Request, Response, status
from gidgethub.sansio import Event
from safir.dependencies.logger import logger_dependency
from structlog.stdlib import BoundLogger

external_router = APIRouter()
"""FastAPI router for all external handlers."""


@external_router.post(
    "/github/webhook",
    summary="GitHub App webhook",
    description=("This endpoint receives webhook events from GitHub"),
    status_code=status.HTTP_200_OK,
)
async def post_github_webhook(
    request: Request,
    logger: BoundLogger = Depends(logger_dependency),
) -> Response:
    """Process GitHub webhook events."""
    webhook_secret = config.github_webhook_secret.get_secret_value()
    body = await request.body()
    event = Event.from_http(request.headers, body, secret=webhook_secret)

    # Bind the X-GitHub-Delivery header to the logger context; this identifies
    # the webhook request in GitHub's API and UI for diagnostics
    logger = logger.bind(github_delivery=event.delivery_id)

    logger.debug("Received GitHub webhook", payload=event.data)
    # Give GitHub some time to reach internal consistency.
    await asyncio.sleep(1)
    await webhook_router.dispatch(event, logger)

    return Response(status_code=status.HTTP_202_ACCEPTED)

Note how a webhook_secret is passed to the Event.from_http contrucstor. You’ll set up this secret in the GitHub App’s settings, and its used to verify that the webhook request is coming from GitHub.

The webhook_router is a Gidgethub Router instance that dispatches webhook events to the appropriate handler in your application, as you’ll see next.

Tip

Notice how the FastAPI handler sleeps for one second before dispatching the webhook event to the webhook_router. This is based on common advice to wait for the GitHub API to reach consistency before processing the webhook event and making API calls to GitHub based on the new event. You can experiement with delay, move it to other parts of the app (for example, to right before submitting a new requests to GitHub) or drop the delay altogether.

Adding a GitHub webhook handler#

Just as your FastAPI application has a “handler” module containing a FastAPI APIRouter and individual FastAPI handler functions, your GitHub App integration will have a “webhook” module containing a Gidgethub Router and individual webhook handlers.

The Gidgethub Router class should be set up as a module-level variable:

webhooks.py#
from gidgethub import routing

webhook_router = routing.Router()

The webhook_router is then used to register webhook handlers for specific events. For example, to handle the push event, you would add a handler function like this:

webhooks.py#
@webhook_router.register("push")
async def handle_push(
    event: gidgethub.sansio.Event, logger: BoundLogger
) -> None:
    """Handle push events."""
    logger.info("Received push event", payload=event.data)

The argument to the register decorator is the name of the GitHub event that the decorated function should handle. GitHub provides a listing of all the events that can be handled by a GitHub App.

Many events have multiple actions associated with them. For example, a pull request could be opened, closed, or merged, among other possibilities. The names of these actions correspond to the action field in the event’s webhook payload. To scope a handler to a specific action, you can pass its name to the action keyword argument of the register decorator:

webhooks.py#
@webhook_router.register("pull_request", action="opened")
async def handle_pull_request_opened(
    event: gidgethub.sansio.Event, logger: BoundLogger
) -> None:
    """Handle pull request opened events."""
    logger.info(
        f"Received {event.event} {event.data.action} event",
        event=event.event,
        action=event.data.action,
        payload=event.data,
    )

Parsing webhook payloads into Pydantic objects#

Safir provides Pydantic models for relevant GitHub event payloads. You can find a listing of these models and their corresponding webhook events in Models for GitHub webhook payloads. You can parse the event.data attribute into a Pydantic model using the parse_obj method:

webhooks.py#
from safir.github.webhooks import GitHubPullRequestEventModel


@webhook_router.register("pull_request", action="opened")
async def handle_pull_request_opened(
    event: gidgethub.sansio.Event, logger: BoundLogger
) -> None:
    """Handle pull request opened events."""
    pull_request_event = GitHubPullRequestEventModel.parse_obj(event.data)
    logger.info(
        f"Received {event.event} {event.data.action} event",
        event=event.event,
        action=event.data.action,
        payload=pull_request_event.dict(),
        number=pull_request_event.number,
    )

Now your application can access the parsed payload as a Pydantic model, with type hints and validation.

Handling webhook events with resiliance#

A good webhook client needs to handle a webhook event quickly and return a response (usually a 202 Accepted) to GitHub as soon as possible, regardless of whether the action triggered by the event succeeded or not. The best way to do this is to have the webhook handler (as discussed in Adding a GitHub webhook handler) simply parse the event data and spawn a background task to handle the event. There are many ways to create a background task.

In-process background task#

The simplest background tasks are handled in-process with asyncio-based task managers. In FastAPI applications, you can create a Starlette BackgroundTask (see also the FastAPI BackgroundTasks documentation). Another option is to use aiojobs, which is a more general-purpose asyncio background task library. The downside of both these approaches is that the background tasks run in the same process as the API server. This can cause the server to become loaded with tasks. And if the server crashes, the tasks are lost.

Distributed queue#

The recommended approach for handling webhooks is to process them in a distributed queue. This has the advantage of decoupling the API server from background task processing. Multiple workers (typically an independent Deployment in Kubernetes) can be configured to process the queue. Further, a persistent store like Redis holds the task queue and its results so that event processing is resiliant to individual pod restarts. For Safir/FastAPI applications, the recommended queue library is arq, which is a Redis-backed queue library that is built on top of asyncio. See the Using the arq Redis queue client and dependency documentation for more information.

Next steps#

More documentation for integrating with GitHub webhooks: