Skip to content

The fastest, litest AI Gateway. Rust core with Python SDK. Call 100+ LLM APIs in OpenAI (or native) format with cost tracking, guardrails, load balancing, and logging [Bedrock, Azure, OpenAI, Anthropic, OpenAI, VertexAI, vLLM, Nvidia NIM]

fix(ui): offer access-group models in team per-model rate limit and budget pickers (#47630)

mrinal-berri9026f42710c3
diff --git a/ui/litellm-dashboard/src/components/team/TeamInfo.integration.test.tsx b/ui/litellm-dashboard/src/components/team/TeamInfo.integration.test.tsx
index 7f567a7f936..fc054e51e5b 100644
--- a/ui/litellm-dashboard/src/components/team/TeamInfo.integration.test.tsx
+++ b/ui/litellm-dashboard/src/components/team/TeamInfo.integration.test.tsx
@@ -816,3 +816,44 @@ describe("TeamInfoView - member budget reset prompt", () => {
expect(screen.getByRole("option", { name: "claude-haiku-4-5" })).toBeInTheDocument();
});
});
+
+describe("TeamInfoView - model rate limit picker", () => {
+ const props = {
+ teamId: "123",
+ onUpdate: vi.fn(),
+ onClose: vi.fn(),
+ accessToken: "test-token",
+ is_team_admin: true,
+ is_proxy_admin: true,
+ editTeam: false,
+ };
+
+ beforeEach(() => {
+ seedDefaultMocks();
+ });
+
+ afterEach(() => {
+ vi.clearAllMocks();
+ });
+
+ it("offers access-group models, not no-default-models, in the model rate limit picker", async () => {
+ const user = userEvent.setup({ delay: null });
+ vi.mocked(networking.teamInfoCall).mockResolvedValue(
+ createMockTeamData({
+ models: ["no-default-models"],
+ access_group_models: ["claude-haiku-4-5"],
+ }) as TeamData,
+ );
+
+ renderWithProviders(<TeamInfoView {...props} userModels={["gpt-4o", "claude-haiku-4-5"]} premiumUser />);
+ await waitFor(() => expect(screen.queryAllByText("Test Team").length).toBeGreaterThan(0));
+ await user.click(screen.getByRole("tab", { name: "Settings" }));
+ await user.click(await screen.findByRole("button", { name: /edit settings/i }));
+ await user.click(screen.getByRole("button", { name: /add model limit/i }));
+ await user.click(screen.getByPlaceholderText("Select model"));
+
+ expect(await screen.findByRole("option", { name: "claude-haiku-4-5" })).toBeInTheDocument();
+ expect(screen.queryByRole("option", { name: "no-default-models" })).not.toBeInTheDocument();
+ expect(screen.queryByRole("option", { name: "gpt-4o" })).not.toBeInTheDocument();
+ });
+});
diff --git a/ui/litellm-dashboard/src/components/team/TeamInfo.tsx b/ui/litellm-dashboard/src/components/team/TeamInfo.tsx
index 29c048034f4..0d63f5d0833 100644
--- a/ui/litellm-dashboard/src/components/team/TeamInfo.tsx
+++ b/ui/litellm-dashboard/src/components/team/TeamInfo.tsx
@@ -75,6 +75,7 @@ import {
computeTeamModelBadges,
normalizeTeamModelSelection,
TeamAccessGroupModelGrant,
+ teamPerModelLimitOptions,
TeamModelBadge,
TeamModelBadgeKind,
} from "./teamModelAccess";
@@ -89,7 +90,6 @@ import ModelAliasManager from "../common_components/ModelAliasManager";
import AgentSelector from "../agent_management/AgentSelector";
import DeleteResourceModal from "../common_components/DeleteResourceModal";
import PassThroughRoutesSelector from "../common_components/PassThroughRoutesSelector";
-import { unfurlWildcardModelsInList } from "../key_team_helpers/fetch_available_models_team_key";
import GuardrailSettingsView from "../GuardrailSettingsView";
import LoggingSettingsView from "../logging_settings_view";
import MCPServerSelector from "../mcp_server_management/MCPServerSelector";
@@ -666,20 +666,9 @@ const TeamInfoView: React.FC<TeamInfoProps> = ({
[mcpServersLoading || mcpToolsetsLoading || accessGroupsLoading, "the MCP server inventory is still loading"],
] as const
).find(([failed]) => failed)?.[1] ?? null;
- const availableRateLimitModels = useMemo(() => {
+ const perModelLimitModels = useMemo(() => {
const selected = (isEditing ? watchedModels : undefined) ?? teamData?.team_info?.models ?? [];
- if (selected.includes("all-proxy-models") || selected.includes("all-team-models")) {
- return userModels;
- }
- return unfurlWildcardModelsInList(selected, userModels);
- }, [isEditing, watchedModels, teamData, userModels]);
- const availableModelBudgetModels = useMemo(() => {
- const selected = (isEditing ? watchedModels : undefined) ?? teamData?.team_info?.models ?? [];
- if (selected.length === 0 || selected.includes("all-proxy-models") || selected.includes("all-team-models")) {
- return userModels;
- }
- const granted = [...selected, ...(teamData?.team_info?.access_group_models ?? [])];
- return unfurlWildcardModelsInList(Array.from(new Set(granted)), userModels);
+ return teamPerModelLimitOptions(selected, teamData?.team_info?.access_group_models ?? [], userModels);
}, [isEditing, watchedModels, teamData, userModels]);
const teamEditAccess = useMemo(() => parseTeamEditAccess(teamData?.team_info?.caller_edit_access), [teamData]);
@@ -1724,7 +1713,7 @@ const TeamInfoView: React.FC<TeamInfoProps> = ({
premiumUser={premiumUser}
value={teamMemberModelMaxBudget}
onChange={setTeamMemberModelMaxBudget}
- availableModels={availableModelBudgetModels}
+ availableModels={perModelLimitModels}
hint="Set a default per-model spend cap for each team member, with an independent reset window."
/>
<FormField
@@ -1853,7 +1842,7 @@ const TeamInfoView: React.FC<TeamInfoProps> = ({
premiumUser={premiumUser}
value={teamModelMaxBudget}
onChange={setTeamModelMaxBudget}
- availableModels={availableRateLimitModels}
+ availableModels={perModelLimitModels}
usage={info.model_max_budget_usage}
hint="Cap this team's spend on individual models, each with its own reset window. Every key on the team shares the cap unless the key sets its own budget for that model."
/>
@@ -1906,7 +1895,7 @@ const TeamInfoView: React.FC<TeamInfoProps> = ({
inputId={id}
value={value ?? ""}
onValueChange={onChange}
- options={availableRateLimitModels.map((model) => ({
+ options={perModelLimitModels.map((model) => ({
label: model,
value: model,
}))}
@@ -2644,7 +2633,7 @@ const TeamInfoView: React.FC<TeamInfoProps> = ({
</span>
),
type: "model-max-budget" as const,
- availableModels: availableModelBudgetModels,
+ availableModels: perModelLimitModels,
premiumUser,
},
{
diff --git a/ui/litellm-dashboard/src/components/team/teamModelAccess.test.ts b/ui/litellm-dashboard/src/components/team/teamModelAccess.test.ts
index 3f0b4e20407..5ca42a9391b 100644
--- a/ui/litellm-dashboard/src/components/team/teamModelAccess.test.ts
+++ b/ui/litellm-dashboard/src/components/team/teamModelAccess.test.ts
@@ -1,5 +1,10 @@
import { describe, expect, it } from "vitest";
-import { computeTeamModelBadges, normalizeTeamModelSelection, TeamAccessGroupModelGrant } from "./teamModelAccess";
+import {
+ computeTeamModelBadges,
+ normalizeTeamModelSelection,
+ TeamAccessGroupModelGrant,
+ teamPerModelLimitOptions,
+} from "./teamModelAccess";
const GRANTS: TeamAccessGroupModelGrant[] = [
{ access_group_id: "ag-1", access_group_name: "shared", models: ["haiku", "gpt-4o-mini"] },
@@ -84,3 +89,35 @@ describe("computeTeamModelBadges", () => {
]);
});
});
+
+describe("teamPerModelLimitOptions", () => {
+ const USER_MODELS = ["gpt-4o", "haiku", "openai/gpt-4o-mini", "openai/gpt-4.1"];
+
+ it("offers access-group models, not the sentinel, for a no-default-models team", () => {
+ expect(teamPerModelLimitOptions(["no-default-models"], ["gpt-4o", "haiku"], USER_MODELS)).toEqual([
+ "gpt-4o",
+ "haiku",
+ ]);
+ });
+
+ it("merges direct and access-group models without duplicates", () => {
+ expect(teamPerModelLimitOptions(["gpt-4o"], ["gpt-4o", "haiku"], USER_MODELS)).toEqual(["gpt-4o", "haiku"]);
+ });
+
+ it("offers every user model when the team covers all models", () => {
+ expect(teamPerModelLimitOptions([], [], USER_MODELS)).toEqual(USER_MODELS);
+ expect(teamPerModelLimitOptions(["all-proxy-models"], [], USER_MODELS)).toEqual(USER_MODELS);
+ expect(teamPerModelLimitOptions(["all-team-models"], [], USER_MODELS)).toEqual(USER_MODELS);
+ });
+
+ it("expands wildcards to concrete names and drops the pattern itself", () => {
+ expect(teamPerModelLimitOptions(["no-default-models"], ["openai/*"], USER_MODELS)).toEqual([
+ "openai/gpt-4o-mini",
+ "openai/gpt-4.1",
+ ]);
+ });
+
+ it("offers nothing when the team has no models it can call", () => {
+ expect(teamPerModelLimitOptions(["no-default-models"], [], USER_MODELS)).toEqual([]);
+ });
+});
diff --git a/ui/litellm-dashboard/src/components/team/teamModelAccess.ts b/ui/litellm-dashboard/src/components/team/teamModelAccess.ts
index d5d9245ed51..3027fbb8686 100644
--- a/ui/litellm-dashboard/src/components/team/teamModelAccess.ts
+++ b/ui/litellm-dashboard/src/components/team/teamModelAccess.ts
@@ -1,3 +1,5 @@
+import { unfurlWildcardModelsInList } from "../key_team_helpers/fetch_available_models_team_key";
+
export const ALL_PROXY_MODELS = "all-proxy-models";
export const NO_DEFAULT_MODELS = "no-default-models";
@@ -18,6 +20,24 @@ export interface TeamModelBadge {
tooltip: string;
}
+const ALL_TEAM_MODELS = "all-team-models";
+const MODEL_SENTINELS: ReadonlySet<string> = new Set([ALL_PROXY_MODELS, ALL_TEAM_MODELS, NO_DEFAULT_MODELS]);
+
+// Per-model rate limits and budgets are enforced on the model name a request sends,
+// so only concrete names are offered: sentinels and wildcard patterns never match one.
+export function teamPerModelLimitOptions(
+ teamModels: string[],
+ accessGroupModels: string[],
+ userModels: string[],
+): string[] {
+ const coversAllModels =
+ teamModels.length === 0 || teamModels.includes(ALL_PROXY_MODELS) || teamModels.includes(ALL_TEAM_MODELS);
+ const granted = coversAllModels ? userModels : [...new Set([...teamModels, ...accessGroupModels])];
+ return unfurlWildcardModelsInList(granted, userModels).filter(
+ (model) => !MODEL_SENTINELS.has(model) && !model.endsWith("*"),
+ );
+}
+
export function normalizeTeamModelSelection(models: string[] | undefined): string[] {
return models && models.length > 0 ? models : [NO_DEFAULT_MODELS];
}