fix(proxy-ui): PR review — key models API clarity and RQ cache keys

- model_checks: simplify empty_models; rename resolved_total_count to
  resolved_config_entry_count; document pattern vs matched counts
- key_management: OpenAPI doc bullets for new response fields
- useGetKeyModels: include userId ?? accessToken in query keys
- networking: drop no-op try/catch in fetchKeyModelCall; update KeyModelResponse
- Tests: mocks/assertions for renamed field

Made-with: Cursor

litellm/proxy/auth/model_checks.py

@@ -534,7 +534,8 @@ def build_key_resolved_model_display_sections(
     only to avoid repeating the full flat list as "Other models".
     """
     sections: List[KeyResolvedModelDisplaySection] = []
-    empty_models: List[str] = [] if compact else []
+    # Always [] — compact still toggles each section's `models=` via `empty_models if compact else ...` below.
+    empty_models: List[str] = []
 
     display_set = set(display_models)
 
@@ -603,6 +604,9 @@ def prepare_key_models_response_payload(
 
     Access-group sections list concrete model names: wildcards in router group metadata
     are expanded against models allowed for this key (resolved ∩ router names).
+
+    ``resolved_config_entry_count`` is len(resolved) (pattern/sentinel entries); ``matched_count`` is
+    the number of concrete router names after expansion and search; they differ when wildcards are used.
     """
     base_concrete = _concrete_models_allowed_by_resolved(resolved, all_router_model_names)
 
@@ -638,7 +642,7 @@ def prepare_key_models_response_payload(
     return {
         "model_display_sections": model_display_sections,
         "source": source,
-        "resolved_total_count": len(resolved),
+        "resolved_config_entry_count": len(resolved),
         "matched_count": matched_count,
         "models_truncated": models_truncated,
         "all_team_models_without_team": all_team_models_without_team,

litellm/proxy/management_endpoints/key_management_endpoints.py

@@ -2828,6 +2828,8 @@ async def key_resolved_models_fn(
     - **search**: Filters the resolved list before truncation and sectioning.
     - **compact**: Metadata-only payload (empty `models` arrays in each section) for fast initial load.
     - **all_team_models_without_team**: True when the key uses `all-team-models` but has no `team_id` (assign a team in key settings).
+    - **resolved_config_entry_count**: Length of the post-sentinel-resolution pattern list (not the same as concrete `matched_count` when wildcards are present).
+    - **matched_count**: Number of concrete router model names after expansion and optional search (before display truncation).
     """
     from litellm.proxy.proxy_server import llm_router, prisma_client
 

tests/test_litellm/proxy/auth/test_model_checks_key_display.py

@@ -254,7 +254,7 @@ def test_prepare_payload_search_and_truncation():
         compact=False,
         all_router_model_names=big,
     )
-    assert out["resolved_total_count"] == len(big)
+    assert out["resolved_config_entry_count"] == len(big)
     assert out["matched_count"] == len(big)
     assert out["models_truncated"] is True
     ung = out["model_display_sections"][0]

ui/litellm-dashboard/src/components/key_team_helpers/KeyModelList.test.tsx

@@ -27,7 +27,7 @@ describe("KeyModelList", () => {
             { title: "grp-a", section_kind: "access_group", models: ["m1"] },
           ],
           source: "all-proxy-models",
-          resolved_total_count: 2,
+          resolved_config_entry_count: 2,
           matched_count: 2,
           models_truncated: false,
           all_team_models_without_team: false,
@@ -62,7 +62,7 @@ describe("KeyModelList", () => {
             { title: "All team models", section_kind: "all_team_models", models: ["a"] },
           ],
           source: "all-team-models",
-          resolved_total_count: 1,
+          resolved_config_entry_count: 1,
           matched_count: 1,
           models_truncated: false,
           all_team_models_without_team: true,
@@ -92,7 +92,7 @@ describe("KeyModelList", () => {
             { title: "Other models", section_kind: "ungrouped", models: ["x"] },
           ],
           source: "no-default-models",
-          resolved_total_count: 1,
+          resolved_config_entry_count: 1,
           matched_count: 1,
           models_truncated: false,
           all_team_models_without_team: false,
@@ -124,7 +124,7 @@ describe("KeyModelList", () => {
             { title: "Other models", section_kind: "ungrouped", models: ["gpt-4"] },
           ],
           source: "no-default-models",
-          resolved_total_count: 1,
+          resolved_config_entry_count: 1,
           matched_count: 1,
           models_truncated: false,
           all_team_models_without_team: false,
@@ -152,7 +152,7 @@ describe("KeyModelList", () => {
         data: {
           model_display_sections: [],
           source: "no-default-models",
-          resolved_total_count: 5,
+          resolved_config_entry_count: 5,
           matched_count: 0,
           models_truncated: false,
           all_team_models_without_team: false,
@@ -184,7 +184,7 @@ describe("KeyModelList", () => {
             { title: "grp-b", section_kind: "access_group", models: ["m2", "m1"] },
           ],
           source: "all-proxy-models",
-          resolved_total_count: 2,
+          resolved_config_entry_count: 2,
           matched_count: 2,
           models_truncated: false,
           all_team_models_without_team: false,

ui/litellm-dashboard/src/components/networking.test.ts

@@ -423,7 +423,7 @@ describe("fetchKeyModelCall", () => {
       json: vi.fn().mockResolvedValue({
         model_display_sections: [],
         source: "no-default-models",
-        resolved_total_count: 0,
+        resolved_config_entry_count: 0,
         matched_count: 0,
         models_truncated: false,
         all_team_models_without_team: false,

ui/litellm-dashboard/src/components/networking.tsx

@@ -3323,7 +3323,8 @@ export interface KeyModelDisplaySection {
 export interface KeyModelResponse {
   model_display_sections: KeyModelDisplaySection[];
   source: string;
-  resolved_total_count: number;
+  /** Number of entries in the post–sentinel-resolution list (`resolved`), not concrete model cardinality. */
+  resolved_config_entry_count: number;
   matched_count: number;
   models_truncated: boolean;
   all_team_models_without_team: boolean;
@@ -3339,38 +3340,34 @@ export const fetchKeyModelCall = async (
   key_id: string,
   options?: FetchKeyModelCallOptions
 ): Promise<KeyModelResponse> => {
-  try {
-    const params = new URLSearchParams();
-    if (options?.search !== undefined && options.search.trim() !== "") {
-      params.set("search", options.search.trim());
-    }
-    if (options?.compact === true) {
-      params.set("compact", "true");
-    }
-    const qs = params.toString();
-    const base = proxyBaseUrl ? `${proxyBaseUrl}/key/${key_id}/models` : `/key/${key_id}/models`;
-    let url = qs ? `${base}?${qs}` : base;
-
-    const response = await fetch(url, {
-      method: "GET",
-      headers: {
-        [globalLitellmHeaderName]: `Bearer ${accessToken}`,
-        "Content-Type": "application/json",
-      },
-    });
-
-    if (!response.ok) {
-      const errorData = await response.json();
-      const errorMessage = deriveErrorMessage(errorData);
-      handleError(errorMessage);
-      throw new Error(errorMessage);
-    }
-
-    const data = await response.json();
-    return data;
-  } catch (error) {
-    throw error;
+  const params = new URLSearchParams();
+  if (options?.search !== undefined && options.search.trim() !== "") {
+    params.set("search", options.search.trim());
   }
+  if (options?.compact === true) {
+    params.set("compact", "true");
+  }
+  const qs = params.toString();
+  const base = proxyBaseUrl ? `${proxyBaseUrl}/key/${key_id}/models` : `/key/${key_id}/models`;
+  const url = qs ? `${base}?${qs}` : base;
+
+  const response = await fetch(url, {
+    method: "GET",
+    headers: {
+      [globalLitellmHeaderName]: `Bearer ${accessToken}`,
+      "Content-Type": "application/json",
+    },
+  });
+
+  if (!response.ok) {
+    const errorData = await response.json();
+    const errorMessage = deriveErrorMessage(errorData);
+    handleError(errorMessage);
+    throw new Error(errorMessage);
+  }
+
+  const data = await response.json();
+  return data;
 };
 
 export const userDailyActivityAggregatedCall = async (accessToken: string, startTime: Date, endTime: Date, userId: string | null = null) => {

ui/litellm-dashboard/src/components/templates/key_info_view.test.tsx

@@ -36,7 +36,7 @@ vi.mock("../networking", () => ({
       },
     ],
     source: "default",
-    resolved_total_count: 1,
+    resolved_config_entry_count: 1,
     matched_count: 1,
     models_truncated: false,
     all_team_models_without_team: false,

ui/litellm-dashboard/src/hooks/keys/useGetKeyModels.test.tsx

@@ -13,13 +13,14 @@ vi.mock("@/components/networking", () => ({
 vi.mock("@/app/(dashboard)/hooks/useAuthorized", () => ({
   default: vi.fn(() => ({
     accessToken: "test-token-456",
+    userId: "test-user-id",
   })),
 }));
 
 const emptyKeyModelResponse = {
   model_display_sections: [],
   source: "no-default-models",
-  resolved_total_count: 0,
+  resolved_config_entry_count: 0,
   matched_count: 0,
   models_truncated: false,
   all_team_models_without_team: false,
@@ -43,13 +44,14 @@ describe("useGetKeyModels", () => {
     const useAuthorizedModule = await import("@/app/(dashboard)/hooks/useAuthorized");
     vi.mocked(useAuthorizedModule.default).mockReturnValue({
       accessToken: "test-token-456",
+      userId: "test-user-id",
     } as any);
   });
 
   it("should load default full model list without compact", async () => {
     vi.mocked(networking.fetchKeyModelCall).mockResolvedValue({
       ...emptyKeyModelResponse,
-      resolved_total_count: 3,
+      resolved_config_entry_count: 3,
       model_display_sections: [
         { title: "Other models", section_kind: "ungrouped", models: ["a", "b", "c"] },
       ],
@@ -62,6 +64,6 @@ describe("useGetKeyModels", () => {
 
     await waitFor(() => expect(result.current.defaultModelsQuery.isSuccess).toBe(true));
     expect(networking.fetchKeyModelCall).toHaveBeenCalledWith("test-token-456", "test-key-id");
-    expect(result.current.defaultModelsQuery.data?.resolved_total_count).toBe(3);
+    expect(result.current.defaultModelsQuery.data?.resolved_config_entry_count).toBe(3);
   });
 });

ui/litellm-dashboard/src/hooks/keys/useGetKeyModels.ts

@@ -8,7 +8,8 @@ import useAuthorized from '@/app/(dashboard)/hooks/useAuthorized';
 const SEARCH_DEBOUNCE_MS = 450;
 
 export const useGetKeyModels = (key_id: string) => {
-  const { accessToken } = useAuthorized();
+  const { accessToken, userId } = useAuthorized();
+  const cachePrincipal = userId ?? accessToken ?? '';
   const [searchInput, setSearchInputState] = useState('');
   const [debouncedSearch, setDebouncedSearch] = useDebouncedState('', {
     wait: SEARCH_DEBOUNCE_MS,
@@ -28,7 +29,7 @@ export const useGetKeyModels = (key_id: string) => {
   const isSearchDebouncing = searchInput.trim() !== trimmedDebounced;
 
   const defaultModelsQuery = useQuery({
-    queryKey: ['keyModelsDefault', key_id],
+    queryKey: ['keyModelsDefault', key_id, cachePrincipal],
     queryFn: () => {
       if (!accessToken) throw new Error('Access Token required');
       return fetchKeyModelCall(accessToken, key_id);
@@ -37,7 +38,7 @@ export const useGetKeyModels = (key_id: string) => {
   });
 
   const searchQuery = useQuery({
-    queryKey: ['keyModelsSearch', key_id, trimmedDebounced],
+    queryKey: ['keyModelsSearch', key_id, trimmedDebounced, cachePrincipal],
     queryFn: () => {
       if (!accessToken) throw new Error('Access Token required');
       return fetchKeyModelCall(accessToken, key_id, { search: trimmedDebounced });