Add GitHub App service

docs/dev/install.rst

@@ -262,10 +262,30 @@ For others, the webhook will simply fail to connect when there are new commits t
 
 * Configure the applications on GitHub, Bitbucket, and GitLab.
   For each of these, the callback URI is ``http://devthedocs.org/accounts/<provider>/login/callback/``
-  where ``<provider>`` is one of ``github``, ``gitlab``, or ``bitbucket_oauth2``.
+  where ``<provider>`` is one of ``github``, ``githubapp``, ``gitlab``, or ``bitbucket_oauth2``.
   When setup, you will be given a "Client ID" (also called an "Application ID" or just "Key") and a "Secret".
 * Take the "Client ID" and "Secret" for each service and set them as :ref:`environment variables <settings:Allauth secrets>`.
 
+Configuring GitHub App
+~~~~~~~~~~~~~~~~~~~~~~
+
+- Create a new GitHub app from https://github.com/settings/apps/new.
+- Callback URL should be ``http://dev.readthedocs.org/accounts/githubapp/login/callback/``.
+- Keep marked "Expire user authorization tokens"
+- Activate the webhook, and set the URL to one provided by a service like `Webhook.site <https://docs.webhook.site/cli.html>`__ to forward all incoming webhooks to your local development instance.
+  You should forward all events to ``http://dev.readthedocs.org/webhook/githubapp/``.
+- In permissions, select the following:
+  - Repository permissions: Commit statuses (read and write, so we can create commit statuses),
+    Contents (read only, so we can clone repos with a token),
+    Metadata (read only, so we read the repo collaborators),
+    Pull requests (read and write, so we can post a comment on PRs in the future).
+  - Organization permissions: Members (read only so we can read the organization members).
+  - Account permissions: Email addresses (read only, so allauth can fetch all verified emails).
+- Subscribe to the following events: Installation target, Member, Organization, Membership, Pull request, Push, and Repository.
+- Copy the "Client ID" and "Client Secret" and set them as :ref:`environment variables <settings:Allauth secrets>`.
+- Generate a webhook secret and a private key from the GitHub App settings,
+  and set them as :ref:`environment variables <settings:GitHub App secrets>`.
+
 Troubleshooting
 ---------------
 

docs/dev/settings.rst

@@ -167,6 +167,16 @@ providers using the following environment variables:
 .. envvar:: RTD_SOCIALACCOUNT_PROVIDERS_GOOGLE_CLIENT_ID
 .. envvar:: RTD_SOCIALACCOUNT_PROVIDERS_GOOGLE_SECRET
 
+GitHub App
+~~~~~~~~~~
+
+You can use the following environment variables to set the settings used by the GitHub App:
+
+.. envvar:: RTD_GITHUB_APP_ID
+.. envvar:: RTD_GITHUB_APP_NAME
+.. envvar:: RTD_GITHUB_PRIVATE_KEY
+.. envvar:: RTD_GITHUB_APP_WEBHOOK_SECRET
+
 Stripe secrets
 ~~~~~~~~~~~~~~
 

readthedocs/oauth/clients.py

@@ -1,7 +1,10 @@
 from datetime import datetime
 
 import structlog
+from django.conf import settings
 from django.utils import timezone
+from github import Auth
+from github import GithubIntegration
 from requests_oauthlib import OAuth2Session
 
 
@@ -70,3 +73,15 @@ def get_oauth2_client(account):
         token_updater=_get_token_updater(token),
     )
     return session
+
+
+def get_gh_app_client() -> GithubIntegration:
+    """Return a client authenticated as the GitHub App to interact with the API."""
+    app_auth = Auth.AppAuth(
+        app_id=settings.GITHUB_APP_CLIENT_ID,
+        private_key=settings.GITHUB_APP_PRIVATE_KEY,
+        # 10 minutes is the maximum allowed by GitHub.
+        # PyGithub will handle the token expiration and renew it automatically.
+        jwt_expiry=60 * 10,
+    )
+    return GithubIntegration(auth=app_auth)

readthedocs/oauth/models.py

@@ -1,5 +1,7 @@
 """OAuth service models."""
 
+from functools import cached_property
+
 import structlog
 from allauth.socialaccount.models import SocialAccount
 from django.contrib.auth.models import User
@@ -12,6 +14,7 @@
 from readthedocs.projects.constants import REPO_CHOICES
 from readthedocs.projects.models import Project
 
+from .constants import GITHUB_APP
 from .constants import VCS_PROVIDER_CHOICES
 from .querysets import RemoteOrganizationQuerySet
 from .querysets import RemoteRepositoryQuerySet
@@ -84,6 +87,78 @@ class GitHubAppInstallation(TimeStampedModel):
     class Meta(TimeStampedModel.Meta):
         verbose_name = _("GitHub app installation")
 
+    @cached_property
+    def service(self):
+        """Return the service for this installation."""
+        from readthedocs.oauth.services import GitHubAppService
+
+        return GitHubAppService(self)
+
+    def delete(self, *args, **kwargs):
+        """Override delete method to remove orphaned organizations."""
+        self.delete_repositories()
+        return super().delete(*args, **kwargs)
+
+    def delete_repositories(self, repository_ids: list[int] | None = None):
+        """
+        Delete repositories linked to this installation.
+        When an installation is deleted, we delete all its remote repositories
+        and relations, users will need to manually link the projects to each repository again.
+        We also remove organizations that don't have any repositories after removing the repositories.
+        :param repository_ids: List of repository ids (remote ID) to delete.
+         If None, all repositories will be considered for deletion.
+        """
+        if repository_ids is not None and not repository_ids:
+            log.info("No repositories to delete")
+            return
+
+        remote_organizations = RemoteOrganization.objects.filter(
+            repositories__github_app_installation=self,
+            vcs_provider=GITHUB_APP,
+        )
+        remote_repositories = self.repositories.filter(vcs_provider=GITHUB_APP)
+        if repository_ids:
+            remote_organizations = remote_organizations.filter(
+                repositories__remote_id__in=repository_ids
+            )
+            remote_repositories = remote_repositories.filter(remote_id__in=repository_ids)
+
+        # Fetch all IDs before deleting the repositories, so we can filter the organizations later.
+        remote_organizations_ids = list(remote_organizations.values_list("id", flat=True))
+
+        count, deleted = remote_repositories.delete()
+        log.info(
+            "Deleted repositories projects",
+            count=count,
+            deleted=deleted,
+            installation_id=self.installation_id,
+        )
+
+        count, deleted = RemoteOrganization.objects.filter(
+            id__in=remote_organizations_ids,
+            repositories=None,
+        ).delete()
+        log.info(
+            "Deleted orphaned organizations",
+            count=count,
+            deleted=deleted,
+            installation_id=self.installation_id,
+        )
+
+    def delete_organization(self, organization_id: int):
+        """Delete an organization and all its repositories and relations from the database."""
+        count, deleted = RemoteOrganization.objects.filter(
+            remote_id=str(organization_id),
+            vcs_provider=GITHUB_APP,
+        ).delete()
+        log.info(
+            "Deleted organization",
+            count=count,
+            deleted=deleted,
+            organization_id=organization_id,
+            installation_id=self.installation_id,
+        )
+
 
 class RemoteOrganization(TimeStampedModel):
     """

readthedocs/oauth/querysets.py

@@ -3,6 +3,7 @@
 from django.db import models
 
 from readthedocs.core.querysets import NoReprQuerySet
+from readthedocs.oauth.constants import GITHUB_APP
 
 
 class RelatedUserQuerySet(NoReprQuerySet, models.QuerySet):
@@ -12,7 +13,10 @@ def api(self, user=None):
         """Return objects for user."""
         if not user.is_authenticated:
             return self.none()
-        return self.filter(users=user)
+        queryset = self.filter(users=user)
+        # Exclude repositories/organizations from the GitHub App for now to avoid confusions.
+        queryset = queryset.exclude(vcs_provider=GITHUB_APP)
+        return queryset
 
     def api_v2(self, *args, **kwargs):
         # API v2 is the same as API v3 for .org, but it's
@@ -28,10 +32,13 @@ def for_project_linking(self, user):
         Repositories can be linked to a project only if the user has admin access
         to the repository on the VCS service.
         """
-        return self.filter(
+        queryset = self.filter(
             remote_repository_relations__user=user,
             remote_repository_relations__admin=True,
-        ).distinct()
+        )
+        # Exclude repositories from the GitHub App for now to avoid confusions.
+        queryset = queryset.exclude(vcs_provider=GITHUB_APP)
+        return queryset.distinct()
 
 
 class RemoteOrganizationQuerySet(RelatedUserQuerySet):

readthedocs/oauth/services/__init__.py

@@ -4,6 +4,16 @@
 from readthedocs.oauth.services import bitbucket
 from readthedocs.oauth.services import github
 from readthedocs.oauth.services import gitlab
+from readthedocs.oauth.services.githubapp import GitHubAppService
+
+
+__all__ = [
+    "GitHubService",
+    "BitbucketService",
+    "GitLabService",
+    "GitHubAppService",
+    "registry",
+]
 
 
 class GitHubService(SettingsOverrideObject):
@@ -21,4 +31,4 @@ class GitLabService(SettingsOverrideObject):
     _override_setting = "OAUTH_GITLAB_SERVICE"
 
 
-registry = [GitHubService, BitbucketService, GitLabService]
+registry = [GitHubService, BitbucketService, GitLabService, GitHubAppService]

readthedocs/oauth/services/base.py

@@ -34,7 +34,7 @@ class Service:
     vcs_provider_slug: str
     allauth_provider = type[OAuth2Provider]
 
-    url_pattern: re.Pattern | None
+    url_pattern: re.Pattern | None = None
     default_user_avatar_url = settings.OAUTH_AVATAR_USER_DEFAULT_URL
     default_org_avatar_url = settings.OAUTH_AVATAR_ORG_DEFAULT_URL
     supports_build_status = False
@@ -106,6 +106,10 @@ def send_build_status(self, *, build, commit, status):
         """
         raise NotImplementedError
 
+    def get_clone_token(self, project):
+        """Get a token used for cloning the repository."""
+        raise NotImplementedError
+
     @classmethod
     def is_project_service(cls, project):
         """
@@ -326,3 +330,7 @@ def sync_repositories(self):
 
     def sync_organizations(self):
         raise NotImplementedError
+
+    def get_clone_token(self, project):
+        """User services make use of SSH keys only for cloning."""
+        return None