Repository navigation
Conversation
The published spec describes only published methods, so it can't serve as a contract for the private and experimental endpoints the frontend calls. sentry.apidocs.internal builds one that does, reading undeclared responses from Response[T] annotations and omitting nothing. Its generator switches this on for its own build through hook parameters and a type-hint policy whose defaults are the published build's, so the published spec is unchanged. Extracted from the internal-spec half of #129315.
Consumers that filter operations by the permissions guarding them need the class names, which only the view classes have. Each operation the internal build admits now carries its view's permission_classes as x-sentry-permissions, with ones composed by &, | and ~ unwrapped to the classes they compose.
…s view DRF's ViewInspector.__set__ binds an assigned schema to the view instance, the same way drf-spectacular's create_view swaps in @extend_schema classes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Issue endpoints declare IssueParams.ISSUES_OR_GROUPS, a `var` path
parameter for the published build's `{var}` spelling of (?:issues|groups),
which the published postprocessing drops along with the `{var}`. The
internal build spells the path `issues/` and kept the parameter, so 16
issue operations declared a required `var` that is not in their path.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ations drf-spectacular appends APPEND_PATHS over whole path items, so a hand-written path item dropped every method its view serves that the item does not describe. The published build only describes the hand-written ones anyway; the internal one dropped POST on organization sentry-app-installations. The internal build now appends them method by method, hand-written still winning. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…nd-written one
The published build appends APPEND_PATHS, then _fix_issue_paths moves the
generated issue operations onto the hand-written paths, so the published
reference describes GET issues/{issue_id}/tags/{key}/values/ with its
generated operation. The internal build spells that path issues/ up front,
and the hand-written operation won instead, with another operationId,
parameters and response schema. A published method's generated operation
now wins here too; hand-written ones still stand in for the rest.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… over a hand-written one The internal build appended the hand-written api-docs operations over every unpublished method's generated one, so their inlined response bodies hid the named components drf-spectacular reads from the backend's own types: ListServiceHooks and ServiceHook, ListTagValuesResponse, ListSentryAppInstallations and PlatformExternalIssueResponse. A generated operation that declares a success body now stands over the hand-written one too. The hand-written one still stands in where the generated response is undocumented. The published build is unchanged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Give every response the internal spec spelled out inline a named component, so the generated frontend contracts can name it too: - The service hook details GET and PUT declare the ServiceHook component registering one already returns, so all four service hook responses share it. - The project user feedback, project issues GET and project issues PUT handlers declare what they return, with Response[T] annotations or, for the issues GET, whose @extend_schema declared only errors, a 200. - track_slo_response keeps the handler it wraps visible, so the internal build can read the issues handlers' annotations. - The uptime response capture endpoints describe their bodies with TypedDicts instead of `dict`. - Symbol sources, whose responses are JSON schemas, name them with inline_sentry_response_schema, a JSON-schema counterpart to inline_sentry_response_serializer. The published operations are unchanged; the published spec gains the SymbolSource and ListSymbolSources components they now refer to. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
github-actions
Bot
added
the
Scope: Backend
billyvg
force-pushed
the
billyvg/feat/apidocs-internal-spec
branch
from
November 5, 2026 16:55
1aa8b71 to
fca4a0c
Compare
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why?
The frontend API contracts generated from the internal spec spell out any response the spec inlines, which review on the contracts PR flagged. The generator now declares each one as
<OperationId>Response, but the names are better where the backend declares them. 22 of the responses the contracts map weren't components:api-docs/JSON, which the internal build appended over every unpublished method's generated operation. For service hooks, tag values, Sentry App installations and external issues, that hid named components drf-spectacular had already read from the backend's types.@extend_schemaresponses given asdictor as a raw JSON schema.Stacked on:
For:
How?
In the internal build, a generated operation that declares a success body now stands over the hand-written one, which still stands in where the generated response is undocumented. The endpoints behind the rest now declare their responses as components. Symbol sources, whose responses are JSON schemas, use a new
inline_sentry_response_schema, the JSON-schema counterpart toinline_sentry_response_serializer.20 of the 22 are now components. The four service-hook responses share
ServiceHook. The published operations are unchanged; the published spec gains theSymbolSourceandListSymbolSourcescomponents they now refer to.Decisions
DeployResponse[],CustomInboundFilter[]). Naming the list itself would mean replacing their documented DRF serializers with TypedDicts and losing the public field descriptions.track_slo_responsenow usesfunctools.wraps. Without it, the internal build couldn't read the annotations of the issue handlers it wraps. The published spec built with it is byte-identical.Generated with Claude Code
🤖 Generated with Claude Code