Only delete the cache entry on error if the map still holds the same
promise, preventing a stale rejection from evicting a newer entry.
Also switch the test to createDeferred for readability.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
Moves the mcp-server specType registration out of the default catalog
entity model into a separate backend module following the same pattern
as the AiResource module. Types and validators are now alpha exports.
Signed-off-by: benjdlambert <ben@blam.sh>
Registers mcp-server as a specType on v1alpha1/v1beta1 instead of
introducing a new v1alpha2 apiVersion. Adds addKindVersion to the
builder so separate layers can extend existing kinds.
Signed-off-by: benjdlambert <ben@blam.sh>
Adds addKindVersion to CatalogModelLayerBuilder so layers can add
versions to an existing kind without re-declaring its metadata.
Moves v1alpha2 API registration into ApiEntityV1alpha2.ts as a
separate model layer using the new method.
Signed-off-by: benjdlambert <ben@blam.sh>
Disable app/nav in the minimal test shell instead of rendering
legacy nav-item extensions. Use renderTestApp when nav behavior matters.
Signed-off-by: Patrik Oldsberg <poldsberg@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Drop the deprecated NavItemBlueprint from the public API and migrate core
plugins to set title and icon on PageBlueprint instead. AppNav keeps
backward compatibility for legacy nav-item extensions via an internal
core.nav-item.target data ref.
Signed-off-by: Patrik Oldsberg <poldsberg@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
The nav-item extensions were disabled in the template config to prevent
duplicate rendering, but the custom sidebar already handles this via
nav.take(). After #33788 added filtering of disabled nav items from page
discovery, disabling them causes nav.take('page:catalog') to return
nothing, breaking the sidebar navigation.
Signed-off-by: benjdlambert <ben@blam.sh>
The permission backend previously resolved userInfo and minted a plugin
request token sequentially for every authorize request with user
credentials. On high-traffic endpoints this meant two serial internal
HTTP round-trips per request, even when the same user made many
requests in quick succession.
This change:
1. Adds a 5-second TTL cache to DefaultUserInfoService so that repeated
getUserInfo() calls for the same user return the cached result
without an HTTP call to the auth backend.
2. Parallelises the getUserInfo() and getPluginRequestToken() calls in
the permission backend's handleRequest via Promise.all, saving one
sequential round-trip on cache misses.
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
The mock credentials created by mockCredentials.none(), .user(), and
.service() were missing the internal version: 'v1' field. This caused
toInternalBackstageCredentials() to throw when used with mock
credentials in tests.
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
Ben Lambert