A library for building GitHub Apps and other services that handle GitHub webhooks.
The library provides an http.Handler
implementation that dispatches webhook
events to the correct place, removing boilerplate and letting you focus on the
logic of your application.
- Usage
- Asynchronous Dispatch
- Structured Logging
- GitHub Clients
- Metrics
- Background Jobs and Multi-Organization Operations
- Config Loading
- OAuth2
- Stability and Versioning Guarantees
- Contributing
Most users will implement githubapp.EventHandler
for each webhook
event that needs to be handled. A single implementation can also respond to
multiple event types if they require the same actions:
type CommentHandler struct {
githubapp.ClientCreator
}
func (h *CommentHandler) Handles() []string {
return []string{"issue_comment"}
}
func (h *CommentHandler) Handle(ctx context.Context, eventType, deliveryID string, payload []byte) error {
// from github.com/google/go-github/github
var event github.IssueCommentEvent
if err := json.Unmarshal(payload, &event); err != nil {
return err
}
// do something with the content of the event
}
We recommend embedding githubapp.ClientCreator
in handler implementations as
an easy way to access GitHub clients.
Once you define handlers, register them with an event dispatcher and associate
it with a route in any net/http
-compatible HTTP router:
func registerRoutes(c githubapp.Config) {
cc := githubapp.NewDefaultCachingClientCreator(c)
http.Handle("/api/github/hook", githubapp.NewDefaultEventDispatcher(c,
&CommentHandler{cc},
// ...
))
}
The example package contains a fully functional server
using go-githubapp
. The example app responds to comments on pull requests by
commenting with a copy of the comment body.
To run the app, update example/config.yml
with appropriate secrets and then
run:
./godelw dep
./godelw run example
go-githubapp
has minimal dependencies, but does make some decisions:
- rs/zerolog for logging
- rcrowley/go-metrics for metrics
- google/go-github for v3 (REST) API client
- shurcooL/githubv4 for v4 (GraphQL) API client
Logging and metrics are only active when they are configured (see below). This means you can add your own logging or metrics libraries without conflict, but will miss out on the free built-in support.
GitHub imposes timeouts on webhook delivery responses. If an application does
not respond in time, GitHub closes the connection and marks the delivery as
failed. go-githubapp
optionally supports asynchronous dispatch to help solve
this problem. When enabled, the event dispatcher sends a response to GitHub after
validating the payload and then runs the event handler in a separate goroutine.
To enable, select an appropriate scheduler and configure the event dispatcher to use it:
dispatcher := githubapp.NewEventDispatcher(handlers, secret, githubapp.WithScheduler(
githubapp.AsyncScheduler(),
))
The following schedulers are included in the library:
-
DefaultScheduler
- a synchronous scheduler that runs event handlers in the current goroutine. This is the default mode. -
AsyncScheduler
- an asynchronous scheduler that handles each event in a new goroutine. This is the simplest asynchronous option. -
QueueAsyncScheduler
- an asynchronous scheduler that queues events and handles them with a fixed pool of worker goroutines. This is useful to limit the amount of concurrent work.
AsyncScheduler
and QueueAsyncScheduler
support several additional options
and customizations; see the documentation for details.
go-githubapp
uses rs/zerolog for structured
logging. A logger must be stored in the context.Context
associated with each
http.Request
:
func (d *EventDispatcher) ServeHTTP(w http.ResponseWriter, r *http.Request) {
logger := zerolog.Ctx(r.Context())
...
}
If there is no logger in the context, log output is disabled. It's expected
that HTTP middleware, like that provided by the hlog package, configures
the http.Request
context automatically.
Below are the standard keys used when logging events. They are also exported as constants.
exported constant | key | definition |
---|---|---|
LogKeyEventType |
github_event_type |
the github event type header |
LogKeyDeliveryID |
github_delivery_id |
the github event delivery id header |
LogKeyInstallationID |
github_installation_id |
the installation id the app is authenticating with |
LogKeyRepositoryName |
github_repository_name |
the repository name of the pull request being acted on |
LogKeyRepositoryOwner |
github_repository_owner |
the repository owner of the pull request being acted on |
LogKeyPRNum |
github_pr_num |
the number of the pull request being acted on |
Where appropriate, the library creates derived loggers with the above keys set to the correct values.
Authenticated and configured GitHub clients can be retrieved from
githubapp.ClientCreator
implementations. The library provides a basic
implementation and a caching version.
There are three types of clients and two API versions for a total of six distinct clients:
- An application client authenticates as the application and can only call limited APIs that mostly return application metadata.
- An installation client authenticates as an installation and can call any APIs where the has been installed and granted permissions.
- A token client authenticates with a static OAuth2 token associated with a user account.
go-githubapp
also exposes various configuration options for GitHub clients.
These are provided when calling githubapp.NewClientCreator
:
githubapp.WithClientUserAgent
sets aUser-Agent
string for all clientsgithubapp.WithClientTimeout
sets a timeout for requests made by all clientsgithubapp.WithClientCaching
enables response caching for all v3 (REST) clients. The cache can be configured to always validate responses or to respect the cache headers returned by GitHub. Re-validation is useful if data often changes faster than the requested cache duration.githubapp.WithClientMiddleware
allows customization of thehttp.RoundTripper
used by all clients and is useful if you want to log requests or emit metrics about GitHub requests and responses.
The library provides the following middleware:
githubapp.ClientMetrics
emits the standard metrics described belowgithubapp.ClientLogging
logs metadata about all requests and responses
baseHandler, err := githubapp.NewDefaultCachingClientCreator(
config.Github,
githubapp.WithClientUserAgent("example-app/1.0.0"),
githubapp.WithClientCaching(false, func() httpcache.Cache { return httpcache.NewMemoryCache() }),
githubapp.WithClientMiddleware(
githubapp.ClientMetrics(registry),
githubapp.ClientLogging(zerolog.DebugLevel),
),
...
)
go-githubapp
uses rcrowley/go-metrics to provide metrics. Metrics are
optional and disabled by default.
GitHub clients emit the following metrics when configured with the
githubapp.ClientMetrics
middleware:
metric name | type | definition |
---|---|---|
github.requests |
counter |
the count of successfully completed requests made to GitHub |
github.requests.2xx |
counter |
like github.requests , but only counting 2XX status codes |
github.requests.3xx |
counter |
like github.requests , but only counting 3XX status codes |
github.requests.4xx |
counter |
like github.requests , but only counting 4XX status codes |
github.requests.5xx |
counter |
like github.requests , but only counting 5XX status codes |
github.requests.cached |
counter |
the count of successfully cached requests |
github.rate.limit[installation:<id>] |
gauge |
the maximum number of requests permitted to make per hour, tagged with the installation id |
github.rate.remaining[installation:<id>] |
gauge |
the number of requests remaining in the current rate limit window, tagged with the installation id |
When using asynchronous dispatch, the
githubapp.WithSchedulingMetrics
option emits the following metrics:
metric name | type | definition |
---|---|---|
github.event.queue |
gauge |
the number of queued unprocessed event |
github.event.workers |
gauge |
the number of workers actively processing events |
github.event.dropped |
counter |
the number events dropped due to limited queue capacity |
github.event.age |
histogram |
the age (queue time) in milliseconds of events at processing time |
The MetricsErrorCallback
and MetricsAsyncErrorCallback
error callbacks for
the event dispatcher and asynchronous schedulers emit the following metrics:
metric name | type | definition |
---|---|---|
github.handler.error[event:<type>] |
counter |
the number of processing errors, tagged with the GitHub event type |
Note that metrics need to be published in order to be useful. Several publishing options are available or you can implement your own.
While applications will mostly operate on the installation IDs provided in webhook payloads, sometimes there is a need to run background jobs or make API calls against multiple organizations. In these cases, use an application client to look up specific installations of the application and then construct an installation client to make API calls:
func getOrganizationClient(cc githubapp.ClientCreator, org string) (*github.Client, error) {
// create a client to perform actions as the application
appClient, err := cc.NewAppClient()
if err != nil {
return nil, err
}
// look up the installation ID for a particular organization
installations := githubapp.NewInstallationsService(appClient)
install := installations.GetByOwner(context.Background(), org)
// create a client to perform actions on that specific organization
return cc.NewInstallationClient(install.ID)
}
The appconfig
package provides a flexible configuration loader for finding
repository configuration. It supports repository-local files, files containing
remote references, and organization-level defaults.
By default, the loader will:
- Try a list of paths in the repository
- If a file exists at a path, load its contents
- If the contents define a remote reference, load the remote file. Otherwise, return the contents.
- If no files exist in the repository, try a list of paths in a
.github
repository owned by the same owner.
Users can customize the paths, the remote reference encoding, whether remote references are enabled, the name of the owner-level default repository, and whether the owner-level default is enabled.
The standard remote reference encoding is YAML:
remote: owner/repo
path: config/app.yml
ref: develop
Usage is straightforward:
func loadConfig(ctx context.Context, client *github.Client, owner, repo, ref string) (*AppConfig, error) {
loader := appconfig.NewLoader([]string{".github/app.yml"})
c, err := loader.LoadConfig(ctx, client, onwer, repo, ref)
if err != nil {
return nil, err
}
if c.IsUndefined() {
return nil, nil
}
var appConfig AppConfig
if err := yaml.Unmarshal(c.Content, &appConfig); err != nil {
return nil, err
}
return &appConfig, nil
}
The oauth2
package provides an http.Handler
implementation that simplifies
OAuth2 authentication with GitHub. When a user visits the endpoint, they are
redirected to GitHub to authorize the application. GitHub redirects back to the
same endpoint, which performs the code exchange and obtains a token for the
user. The token is passed to a callback for further processing.
func registerOAuth2Handler(c githubapp.Config) {
http.Handle("/api/auth/github", oauth2.NewHandler(
oauth2.GetConfig(c, []string{"user:email"}),
// force generated URLs to use HTTPS; useful if the app is behind a reverse proxy
oauth2.ForceTLS(true),
// set the callback for successful logins
oauth2.OnLogin(func(w http.ResponseWriter, r *http.Request, login *oauth2.Login) {
// look up the current user with the authenticated client
client := github.NewClient(login.Client)
user, _, err := client.Users.Get(r.Context(), "")
// handle error, save the user, ...
// redirect the user back to another page
http.Redirect(w, r, "/dashboard", http.StatusFound)
}),
))
}
Production applications should also use the oauth2.WithStore
option to set a
secure StateStore
implementation. oauth2.SessionStateStore
is a good choice
that uses alexedwards/scs to store the
state in a session.
For most applications, the default responses should be sufficient: they use correct status codes and include enough information to match up GitHub delivery records with request logs. If your application has additional requirements for responses, two methods are provided for customization:
-
Error responses can be modified with a custom error callback. Use the
WithErrorCallback
option when creating an event dispatcher. -
Non-error responses can be modified with a custom response callback. Use the
WithResponseCallback
option when creating an event dispatcher. -
Individual hook responses can be modified by calling the
SetResponder
function before the handler returns. Note that if you register a custom response handler as described above, you must make it aware of handler-level responders if you want to keep usingSetResponder
. See the default response callback for an example of how to implement this.
While we've used this library to build multiple applications internally, there's still room for API tweaks and improvements as we find better ways to solve problems. These will be backwards compatible when possible and should require only minor changes when not.
Releases will be tagged periodically and will follow semantic versioning, with new major versions tagged after any backwards-incompatible changes. Still, we recommend vendoring this library to avoid surprises.
In general, fixes will only be applied to trunk and future releases, not backported to older versions.
Contributions and issues are welcome. For new features or large contributions, we prefer discussing the proposed change on a GitHub issue prior to a PR.
New functionality should avoid adding new dependencies if possible and should be broadly useful. Feature requests that are specific to certain uses will likely be declined unless they can be redesigned to be generic or optional.
Before submitting a pull request, please run tests and style checks:
./godelw verify
This library is made available under the Apache 2.0 License.