fix(seer): use URL-safe GitLab path when building owner/name

src/sentry/seer/agent/tools.py

@@ -57,6 +57,7 @@
     get_retention_boundary,
 )
 from sentry.seer.autofix.autofix import get_all_tags_overview
+from sentry.seer.autofix.utils import get_repo_url_path
 from sentry.seer.seer_setup import get_supported_scm_providers
 from sentry.seer.sentry_data_models import (
     BaselineTagDistributionEntry,
@@ -938,7 +939,9 @@ def get_repository_definition(
         return None
 
     # Use the actual repo name from the database, not the requested name.
-    repo_name_parts = repo.name.split("/")
+    # For GitLab, repo.name is the display name (name_with_namespace, may contain spaces);
+    # get_repo_url_path() returns the URL-safe path_with_namespace instead.
+    repo_name_parts = get_repo_url_path(repo).split("/")
     owner = repo_name_parts[0]
     name = "/".join(repo_name_parts[1:])
 

src/sentry/seer/autofix/utils.py

@@ -498,14 +498,32 @@ def clear_preference_automation_handoff(project: Project) -> None:
     ).delete()
 
 
+def get_repo_url_path(repo: Repository) -> str:
+    """Return the URL-safe owner/name path for a repository.
+
+    For GitLab, ``repo.name`` is ``name_with_namespace`` (the human-readable
+    display name, e.g. ``"My Group / My Project"`` — with spaces).  The
+    URL-safe equivalent is stored in ``repo.config["path"]``
+    (``path_with_namespace``, e.g. ``"my-group/my-project"``).
+
+    For GitHub and all other providers, ``repo.name`` is already the
+    URL-safe ``owner/repo`` string, so we return it unchanged.
+    """
+    if repo.provider == "integrations:gitlab":
+        path = repo.config.get("path")
+        if path:
+            return path
+    return repo.name
+
+
 def build_repo_definition_from_project_repo(
     seer_project_repo: SeerProjectRepository,
 ) -> SeerRepoDefinition | None:
     """Build a SeerRepoDefinition from a SeerProjectRepository with its joined Repository.
 
     Returns None if Repository name is invalid."""
     repo = seer_project_repo.project_repository.repository
-    repo_name_sections = repo.name.split("/")
+    repo_name_sections = get_repo_url_path(repo).split("/")
     if len(repo_name_sections) < 2:
         sentry_sdk.capture_exception(ValueError(f"Invalid repository name format: {repo.name}"))
         return None
@@ -839,7 +857,7 @@ def get_autofix_repos_from_project_code_mappings(
     repos: dict[tuple, dict] = {}
     for code_mapping in code_mappings:
         repo: Repository = code_mapping.project_repository.repository
-        repo_name_sections = repo.name.split("/")
+        repo_name_sections = get_repo_url_path(repo).split("/")
 
         if (
             # We expect a repository name to be in the format of "owner/name" for now.

src/sentry/seer/endpoints/project_seer_repos.py

@@ -25,6 +25,7 @@
 from sentry.models.repository import Repository
 from sentry.seer.autofix.utils import (
     add_seer_project_repos,
+    get_repo_url_path,
     replace_all_seer_project_repos,
 )
 from sentry.seer.models.project_repository import (
@@ -69,9 +70,9 @@ class ProjectRepoResponse(TypedDict):
 
 def _serialize_project_repo(project_repo: SeerProjectRepository) -> ProjectRepoResponse:
     repo = project_repo.project_repository.repository
-    name_parts = repo.name.split("/", 1)
+    name_parts = get_repo_url_path(repo).split("/", 1)
     owner = name_parts[0] if len(name_parts) > 1 else ""
-    name = name_parts[1] if len(name_parts) > 1 else repo.name
+    name = name_parts[1] if len(name_parts) > 1 else get_repo_url_path(repo)
 
     return ProjectRepoResponse(
         id=str(project_repo.id),

tests/sentry/seer/endpoints/test_project_seer_repos.py

@@ -383,6 +383,31 @@ def test_returns_connected_repos(self):
         assert project_repo_relay["branchName"] is None
         assert project_repo_relay["instructions"] is None
 
+    def test_gitlab_repo_uses_path_not_display_name(self):
+        """GitLab repo.name is name_with_namespace (display name with spaces).
+        The serializer must use repo.config['path'] (path_with_namespace) so that
+        owner/name are URL-safe and don't cause 404s in SCM links."""
+        gitlab_integration = self.create_integration(
+            organization=self.organization, provider="gitlab", external_id="gl123"
+        )
+        gitlab_repo = Repository.objects.create(
+            organization_id=self.organization.id,
+            # name_with_namespace: human-readable display name that contains spaces
+            name="My Group / My Project",
+            provider="integrations:gitlab",
+            external_id="gl-999",
+            integration_id=gitlab_integration.id,
+            config={"path": "my-group/my-project"},
+        )
+        self.create_seer_project_repository(self.project, repository=gitlab_repo)
+
+        response = self.get_success_response()
+        assert len(response.data) == 1
+        repo_data = response.data[0]
+        # owner and name must come from the URL-safe path, not the display name
+        assert repo_data["owner"] == "my-group"
+        assert repo_data["name"] == "my-project"
+
     def test_excludes_inactive_repos(self):
         self.create_seer_project_repository(self.project, repository=self.repo1)
         self.repo1.status = ObjectStatus.HIDDEN