fix(frontend): surface model loading failures (#4840)

* fix(frontend): surface model loading failures

* refactor(frontend): reuse model error UI primitives

* fix(frontend): address model banner review feedback

* refactor(frontend): remove unused model fetch state

frontend/src/app/workspace/workspace-content.tsx

@@ -5,6 +5,7 @@ import { QueryClientProvider } from "@/components/query-client-provider";
 import { SidebarInset, SidebarProvider } from "@/components/ui/sidebar";
 import { CommandPalette } from "@/components/workspace/command-palette";
 import { GatewayOfflineBanner } from "@/components/workspace/gateway-offline-banner";
+import { ModelLoadErrorBanner } from "@/components/workspace/model-load-error-banner";
 import { SettingsDialogHost } from "@/components/workspace/settings";
 import { WorkspaceSettingsDeepLink } from "@/components/workspace/workspace-settings-deep-link";
 import { WorkspaceSidebar } from "@/components/workspace/workspace-sidebar";
@@ -35,6 +36,7 @@ export async function WorkspaceContent({
         <WorkspaceSidebar />
         <SidebarInset className="min-w-0">
           <GatewayOfflineBanner gatewayUnavailable={gatewayUnavailable} />
+          <ModelLoadErrorBanner gatewayUnavailable={gatewayUnavailable} />
           {children}
         </SidebarInset>
       </SidebarProvider>

frontend/src/components/workspace/model-load-error-banner.tsx

@@ -0,0 +1,73 @@
+"use client";
+
+import { useState } from "react";
+
+import { Alert, AlertDescription } from "@/components/ui/alert";
+import { Button } from "@/components/ui/button";
+import { UnauthorizedError } from "@/core/api/errors";
+import { useAuth } from "@/core/auth/AuthProvider";
+import { useI18n } from "@/core/i18n/hooks";
+import { useModels } from "@/core/models/hooks";
+
+import { shouldShowOfflineBanner } from "./gateway-offline-banner-helpers";
+
+interface ModelLoadErrorBannerProps {
+  gatewayUnavailable?: boolean;
+}
+
+export function ModelLoadErrorBanner({
+  gatewayUnavailable = false,
+}: ModelLoadErrorBannerProps) {
+  const { t } = useI18n();
+  const { user } = useAuth();
+  const [isRetrying, setIsRetrying] = useState(false);
+  // Observe the shared query without starting it. Model consumers remain in
+  // charge of loading; this single observer only centralizes their feedback.
+  const { error, refetch } = useModels({ enabled: false });
+
+  const retry = async () => {
+    setIsRetrying(true);
+    try {
+      await refetch();
+    } finally {
+      setIsRetrying(false);
+    }
+  };
+
+  // The shared fetcher has already started a login redirect for this error.
+  // Rendering a model-specific warning during navigation would be duplicate
+  // and misleading feedback.
+  if (
+    (!error && !isRetrying) ||
+    error instanceof UnauthorizedError ||
+    shouldShowOfflineBanner(user, gatewayUnavailable)
+  ) {
+    return null;
+  }
+
+  return (
+    <Alert
+      variant="destructive"
+      className="border-destructive/20 bg-destructive/10 rounded-none border-x-0 border-t-0 px-4 py-2"
+    >
+      <AlertDescription className="text-destructive flex w-full items-center justify-between gap-3">
+        <span className="min-w-0">{t.workspace.modelLoadFailed}</span>
+        <Button
+          type="button"
+          variant="outline"
+          size="sm"
+          disabled={isRetrying}
+          aria-busy={isRetrying}
+          onClick={() => {
+            void retry();
+          }}
+          className="border-destructive/30 text-destructive hover:bg-destructive/10 hover:text-destructive dark:hover:bg-destructive/10 h-7 bg-transparent px-3 text-xs shadow-none dark:bg-transparent"
+        >
+          {isRetrying
+            ? t.workspace.modelLoadRetrying
+            : t.workspace.modelLoadRetry}
+        </Button>
+      </AlertDescription>
+    </Alert>
+  );
+}

frontend/src/core/api/errors.ts

@@ -1,3 +1,16 @@
+/**
+ * Raised after the shared fetcher has started a login redirect for a 401.
+ *
+ * Callers may use this type to avoid showing a second, misleading API error
+ * while the browser is already navigating to the authentication flow.
+ */
+export class UnauthorizedError extends Error {
+  constructor() {
+    super("Unauthorized");
+    this.name = "UnauthorizedError";
+  }
+}
+
 /**
  * Throw an Error from a failed Gateway REST response.
  *

frontend/src/core/api/fetcher.ts

@@ -1,5 +1,7 @@
 import { buildLoginUrl } from "@/core/auth/types";
 
+import { UnauthorizedError } from "./errors";
+
 /** HTTP methods that the gateway's CSRFMiddleware checks. */
 export type StateChangingMethod = "POST" | "PUT" | "DELETE" | "PATCH";
 
@@ -82,7 +84,7 @@ export async function fetch(
 
   if (res.status === 401) {
     window.location.href = buildLoginUrl(window.location.pathname);
-    throw new Error("Unauthorized");
+    throw new UnauthorizedError();
   }
 
   return res;

frontend/src/core/i18n/locales/en-US.ts

@@ -528,6 +528,10 @@ export const enUS: Translations = {
     logout: "Log out",
     gatewayUnavailable: "Gateway is temporarily unavailable.",
     gatewayUnavailableRetrying: "Retrying in the background…",
+    modelLoadFailed:
+      "Models couldn't be loaded. Model selection and token usage may be unavailable.",
+    modelLoadRetry: "Retry",
+    modelLoadRetrying: "Retrying…",
   },
 
   // Conversation

frontend/src/core/i18n/locales/types.ts

@@ -421,6 +421,9 @@ export interface Translations {
     logout: string;
     gatewayUnavailable: string;
     gatewayUnavailableRetrying: string;
+    modelLoadFailed: string;
+    modelLoadRetry: string;
+    modelLoadRetrying: string;
   };
 
   // Conversation

frontend/src/core/i18n/locales/zh-CN.ts

@@ -503,6 +503,10 @@ export const zhCN: Translations = {
     logout: "退出登录",
     gatewayUnavailable: "网关暂时不可用。",
     gatewayUnavailableRetrying: "正在后台重试…",
+    modelLoadFailed:
+      "模型列表加载失败，模型选择和 Token 用量信息可能暂时不可用。",
+    modelLoadRetry: "重试",
+    modelLoadRetrying: "正在重试…",
   },
 
   // Conversation

frontend/src/core/models/hooks.ts

@@ -2,9 +2,11 @@ import { useQuery } from "@tanstack/react-query";
 
 import { loadModels } from "./api";
 
+export const MODELS_QUERY_KEY = ["models"] as const;
+
 export function useModels({ enabled = true }: { enabled?: boolean } = {}) {
-  const { data, isLoading, error } = useQuery({
+  const { data, isLoading, error, refetch } = useQuery({
-    queryKey: ["models"],
+    queryKey: MODELS_QUERY_KEY,
     queryFn: () => loadModels(),
     enabled,
     refetchOnWindowFocus: false,
@@ -20,5 +22,6 @@ export function useModels({ enabled = true }: { enabled?: boolean } = {}) {
     tokenUsageEnabled: data?.token_usage.enabled ?? false,
     isLoading,
     error,
+    refetch,
   };
 }

frontend/tests/unit/components/workspace/model-load-error-banner.dom.test.tsx

@@ -0,0 +1,221 @@
+import { afterEach, beforeEach, describe, expect, it, rs } from "@rstest/core";
+import { QueryClient, QueryClientProvider } from "@tanstack/react-query";
+import {
+  cleanup,
+  fireEvent,
+  render,
+  screen,
+  waitFor,
+} from "@testing-library/react";
+import type { PropsWithChildren } from "react";
+
+rs.mock("@/core/models/api", () => ({
+  loadModels: rs.fn(),
+}));
+
+rs.mock("@/core/i18n/hooks", () => ({
+  useI18n: () => ({
+    locale: "en-US",
+    t: {
+      workspace: {
+        modelLoadFailed:
+          "Models couldn't be loaded. Model selection and token usage may be unavailable.",
+        modelLoadRetry: "Retry",
+        modelLoadRetrying: "Retrying…",
+      },
+    },
+    changeLocale: rs.fn(),
+  }),
+}));
+
+rs.mock("@/core/auth/AuthProvider", () => ({
+  useAuth: rs.fn(),
+}));
+
+import { ModelLoadErrorBanner } from "@/components/workspace/model-load-error-banner";
+import { UnauthorizedError } from "@/core/api/errors";
+import { useAuth } from "@/core/auth/AuthProvider";
+import type { User } from "@/core/auth/types";
+import { loadModels } from "@/core/models/api";
+import { MODELS_QUERY_KEY, useModels } from "@/core/models/hooks";
+import type { ModelsResponse } from "@/core/models/types";
+
+const mockedLoadModels = rs.mocked(loadModels);
+const mockedUseAuth = rs.mocked(useAuth);
+const fakeUser = {} as User;
+
+function createAuthState(user: User | null): ReturnType<typeof useAuth> {
+  return {
+    user,
+    isAuthenticated: user !== null,
+    isLoading: false,
+    logout: rs.fn(),
+    refreshUser: rs.fn(),
+    applyUser: rs.fn(),
+  };
+}
+
+function createDeferred<T>() {
+  let resolve!: (value: T) => void;
+  const promise = new Promise<T>((next) => {
+    resolve = next;
+  });
+  return { promise, resolve };
+}
+
+beforeEach(() => {
+  mockedUseAuth.mockReturnValue(createAuthState(fakeUser));
+});
+
+afterEach(() => {
+  cleanup();
+  mockedLoadModels.mockReset();
+  mockedUseAuth.mockReset();
+});
+
+function createWrapper() {
+  const queryClient = new QueryClient({
+    defaultOptions: {
+      queries: { retry: false },
+    },
+  });
+
+  function QueryWrapper({ children }: PropsWithChildren) {
+    return (
+      <QueryClientProvider client={queryClient}>{children}</QueryClientProvider>
+    );
+  }
+
+  return { queryClient, QueryWrapper };
+}
+
+function ModelConsumer() {
+  useModels();
+  return null;
+}
+
+describe("ModelLoadErrorBanner", () => {
+  it("observes model failures without starting an extra request", async () => {
+    const { QueryWrapper } = createWrapper();
+    render(<ModelLoadErrorBanner />, { wrapper: QueryWrapper });
+
+    await Promise.resolve();
+
+    expect(mockedLoadModels).not.toHaveBeenCalled();
+    expect(screen.queryByRole("alert")).toBeNull();
+  });
+
+  it("shows one actionable error for all model consumers and clears after retry", async () => {
+    const retryResult = createDeferred<ModelsResponse>();
+    mockedLoadModels
+      .mockRejectedValueOnce(new Error("Gateway returned 503"))
+      .mockImplementationOnce(() => retryResult.promise);
+    const { QueryWrapper } = createWrapper();
+
+    render(
+      <>
+        <ModelLoadErrorBanner />
+        <ModelConsumer />
+        <ModelConsumer />
+      </>,
+      { wrapper: QueryWrapper },
+    );
+
+    const alert = await screen.findByRole("alert");
+    expect(alert.textContent).toContain("Models couldn't be loaded");
+    expect(alert.textContent).not.toContain("Gateway returned 503");
+    expect(screen.getAllByRole("alert")).toHaveLength(1);
+    expect(mockedLoadModels).toHaveBeenCalledTimes(1);
+
+    fireEvent.click(screen.getByRole("button", { name: "Retry" }));
+
+    const retryingButton = await screen.findByRole("button", {
+      name: "Retrying…",
+    });
+    expect((retryingButton as HTMLButtonElement).disabled).toBe(true);
+
+    retryResult.resolve({
+      models: [],
+      token_usage: { enabled: false },
+    });
+
+    await waitFor(() => {
+      expect(screen.queryByRole("alert")).toBeNull();
+    });
+    expect(mockedLoadModels).toHaveBeenCalledTimes(2);
+  });
+
+  it("does not duplicate the login redirect with a model warning", async () => {
+    mockedLoadModels.mockRejectedValueOnce(new UnauthorizedError());
+    const { queryClient, QueryWrapper } = createWrapper();
+
+    render(
+      <>
+        <ModelLoadErrorBanner />
+        <ModelConsumer />
+      </>,
+      { wrapper: QueryWrapper },
+    );
+
+    await waitFor(() => {
+      expect(queryClient.getQueryState(MODELS_QUERY_KEY)?.status).toBe("error");
+    });
+    expect(screen.queryByRole("alert")).toBeNull();
+  });
+
+  it("suppresses a model symptom only while the gateway banner is visible", async () => {
+    mockedUseAuth.mockReturnValue(createAuthState(null));
+    mockedLoadModels.mockRejectedValueOnce(new Error("Gateway returned 503"));
+    const { queryClient, QueryWrapper } = createWrapper();
+
+    const renderView = () => (
+      <>
+        <ModelLoadErrorBanner gatewayUnavailable />
+        <ModelConsumer />
+      </>
+    );
+    const { rerender } = render(renderView(), { wrapper: QueryWrapper });
+
+    await waitFor(() => {
+      expect(queryClient.getQueryState(MODELS_QUERY_KEY)?.status).toBe("error");
+    });
+    expect(screen.queryByRole("alert")).toBeNull();
+
+    mockedUseAuth.mockReturnValue(createAuthState(fakeUser));
+    rerender(renderView());
+
+    expect(await screen.findByRole("alert")).not.toBeNull();
+  });
+
+  it("does not show manual retry progress for a shared background refetch", async () => {
+    const backgroundResult = createDeferred<ModelsResponse>();
+    mockedLoadModels
+      .mockRejectedValueOnce(new Error("Gateway returned 503"))
+      .mockImplementationOnce(() => backgroundResult.promise);
+    const { queryClient, QueryWrapper } = createWrapper();
+
+    render(
+      <>
+        <ModelLoadErrorBanner />
+        <ModelConsumer />
+      </>,
+      { wrapper: QueryWrapper },
+    );
+
+    await screen.findByRole("alert");
+    const backgroundRefetch = queryClient.refetchQueries({
+      queryKey: MODELS_QUERY_KEY,
+    });
+    await waitFor(() => {
+      expect(mockedLoadModels).toHaveBeenCalledTimes(2);
+    });
+
+    expect(screen.queryByRole("button", { name: "Retrying…" })).toBeNull();
+
+    backgroundResult.resolve({
+      models: [],
+      token_usage: { enabled: false },
+    });
+    await backgroundRefetch;
+  });
+});

frontend/tests/unit/core/models/api.test.ts

@@ -1,5 +1,7 @@
 import { afterEach, expect, test, rs } from "@rstest/core";
 
+import { UnauthorizedError } from "@/core/api/errors";
+
 afterEach(() => {
   rs.unstubAllGlobals();
 });
@@ -53,6 +55,20 @@ test("loadModels rejects unsuccessful gateway responses", async () => {
   await expect(loadModels()).rejects.toThrow("Model registry unavailable");
 });
 
+test("loadModels exposes the typed 401 redirect error", async () => {
+  const location = { href: "", pathname: "/workspace/chats" };
+  rs.stubGlobal("window", { location });
+  rs.stubGlobal(
+    "fetch",
+    rs.fn(async () => new Response(null, { status: 401 })),
+  );
+
+  const { loadModels } = await import("@/core/models/api");
+
+  await expect(loadModels()).rejects.toBeInstanceOf(UnauthorizedError);
+  expect(location.href).toBe("/login?next=%2Fworkspace%2Fchats");
+});
+
 test("loadModels includes the status code when statusText is empty", async () => {
   rs.stubGlobal(
     "fetch",