feat(ui): surface pinned premium quota alerts in chat thread

surfsense_web/app/dashboard/[search_space_id]/new-chat/[[...chat_id]]/page.tsx

@@ -19,6 +19,7 @@ import {
 	currentThreadAtom,
 	setTargetCommentIdAtom,
 } from "@/atoms/chat/current-thread.atom";
+import { setPremiumAlertForThreadAtom } from "@/atoms/chat/premium-alert.atom";
 import {
 	type MentionedDocumentInfo,
 	mentionedDocumentIdsAtom,
@@ -200,6 +201,19 @@ const BASE_TOOLS_WITH_UI = new Set([
 	// "write_todos", // Disabled for now
 ]);
 
+const PINNED_PREMIUM_QUOTA_MESSAGE = "Premium token quota exceeded for this pinned model.";
+
+function getPinnedPremiumQuotaErrorMessage(error: unknown): string | null {
+	if (!(error instanceof Error)) return null;
+	if (!error.message.toLowerCase().includes("premium token quota exceeded")) {
+		return null;
+	}
+	if (!error.message.toLowerCase().includes("pinned model")) {
+		return null;
+	}
+	return error.message || PINNED_PREMIUM_QUOTA_MESSAGE;
+}
+
 export default function NewChatPage() {
 	const params = useParams();
 	const queryClient = useQueryClient();
@@ -226,6 +240,7 @@ export default function NewChatPage() {
 	const setMentionedDocuments = useSetAtom(mentionedDocumentsAtom);
 	const setMessageDocumentsMap = useSetAtom(messageDocumentsMapAtom);
 	const setCurrentThreadState = useSetAtom(currentThreadAtom);
+	const setPremiumAlertForThread = useSetAtom(setPremiumAlertForThreadAtom);
 	const setTargetCommentId = useSetAtom(setTargetCommentIdAtom);
 	const clearTargetCommentId = useSetAtom(clearTargetCommentIdAtom);
 	const closeReportPanel = useSetAtom(closeReportPanelAtom);
@@ -951,6 +966,7 @@ export default function NewChatPage() {
 					return;
 				}
 				console.error("[NewChatPage] Chat error:", error);
+				const premiumQuotaAlertMessage = getPinnedPremiumQuotaErrorMessage(error);
 
 				// Track chat error
 				trackChatError(
@@ -959,7 +975,15 @@ export default function NewChatPage() {
 					error instanceof Error ? error.message : "Unknown error"
 				);
 
-				toast.error("Failed to get response. Please try again.");
+				if (premiumQuotaAlertMessage) {
+					setPremiumAlertForThread({
+						threadId: currentThreadId,
+						message: premiumQuotaAlertMessage,
+					});
+					toast.error(PINNED_PREMIUM_QUOTA_MESSAGE);
+				} else {
+					toast.error("Failed to get response. Please try again.");
+				}
 				// Update assistant message with error
 				setMessages((prev) =>
 					prev.map((m) =>
@@ -969,7 +993,9 @@ export default function NewChatPage() {
 									content: [
 										{
 											type: "text",
-											text: "Sorry, there was an error. Please try again.",
+											text:
+												premiumQuotaAlertMessage ??
+												"Sorry, there was an error. Please try again.",
 										},
 									],
 								}
@@ -998,6 +1024,7 @@ export default function NewChatPage() {
 			pendingUserImageUrls,
 			setPendingUserImageUrls,
 			toolsWithUI,
+			setPremiumAlertForThread,
 		]
 	);
 
@@ -1257,13 +1284,29 @@ export default function NewChatPage() {
 					return;
 				}
 				console.error("[NewChatPage] Resume error:", error);
-				toast.error("Failed to resume. Please try again.");
+				const premiumQuotaAlertMessage = getPinnedPremiumQuotaErrorMessage(error);
+				if (premiumQuotaAlertMessage) {
+					setPremiumAlertForThread({
+						threadId: resumeThreadId,
+						message: premiumQuotaAlertMessage,
+					});
+					toast.error(PINNED_PREMIUM_QUOTA_MESSAGE);
+				} else {
+					toast.error("Failed to resume. Please try again.");
+				}
 			} finally {
 				setIsRunning(false);
 				abortControllerRef.current = null;
 			}
 		},
-		[pendingInterrupt, messages, searchSpaceId, tokenUsageStore, toolsWithUI]
+		[
+			pendingInterrupt,
+			messages,
+			searchSpaceId,
+			tokenUsageStore,
+			toolsWithUI,
+			setPremiumAlertForThread,
+		]
 	);
 
 	useEffect(() => {
@@ -1584,18 +1627,34 @@ export default function NewChatPage() {
 				}
 				batcher.dispose();
 				console.error("[NewChatPage] Regeneration error:", error);
+				const premiumQuotaAlertMessage = getPinnedPremiumQuotaErrorMessage(error);
 				trackChatError(
 					searchSpaceId,
 					threadId,
 					error instanceof Error ? error.message : "Unknown error"
 				);
-				toast.error("Failed to regenerate response. Please try again.");
+				if (premiumQuotaAlertMessage) {
+					setPremiumAlertForThread({
+						threadId,
+						message: premiumQuotaAlertMessage,
+					});
+					toast.error(PINNED_PREMIUM_QUOTA_MESSAGE);
+				} else {
+					toast.error("Failed to regenerate response. Please try again.");
+				}
 				setMessages((prev) =>
 					prev.map((m) =>
 						m.id === assistantMsgId
 							? {
 									...m,
-									content: [{ type: "text", text: "Sorry, there was an error. Please try again." }],
+									content: [
+										{
+											type: "text",
+											text:
+												premiumQuotaAlertMessage ??
+												"Sorry, there was an error. Please try again.",
+										},
+									],
 								}
 							: m
 					)
@@ -1605,7 +1664,15 @@ export default function NewChatPage() {
 				abortControllerRef.current = null;
 			}
 		},
-		[threadId, searchSpaceId, messages, disabledTools, tokenUsageStore, toolsWithUI]
+		[
+			threadId,
+			searchSpaceId,
+			messages,
+			disabledTools,
+			tokenUsageStore,
+			toolsWithUI,
+			setPremiumAlertForThread,
+		]
 	);
 
 	// Handle editing a message - truncates history and regenerates with new query

surfsense_web/atoms/chat/premium-alert.atom.ts

@@ -0,0 +1,33 @@
+import { atom } from "jotai";
+
+export type PremiumAlertState = {
+	message: string;
+};
+
+export const premiumAlertByThreadAtom = atom<Record<number, PremiumAlertState>>({});
+
+export const setPremiumAlertForThreadAtom = atom(
+	null,
+	(
+		get,
+		set,
+		payload: {
+			threadId: number;
+			message: string;
+		}
+	) => {
+		const current = get(premiumAlertByThreadAtom);
+		set(premiumAlertByThreadAtom, {
+			...current,
+			[payload.threadId]: { message: payload.message },
+		});
+	}
+);
+
+export const clearPremiumAlertForThreadAtom = atom(null, (get, set, threadId: number) => {
+	const current = get(premiumAlertByThreadAtom);
+	if (!(threadId in current)) return;
+	const next = { ...current };
+	delete next[threadId];
+	set(premiumAlertByThreadAtom, next);
+});

surfsense_web/components/assistant-ui/thread.tsx

@@ -37,10 +37,13 @@ import {
 	toggleToolAtom,
 } from "@/atoms/agent-tools/agent-tools.atoms";
 import { chatSessionStateAtom } from "@/atoms/chat/chat-session-state.atom";
-import {
-	mentionedDocumentsAtom,
-} from "@/atoms/chat/mentioned-documents.atom";
+import { currentThreadAtom } from "@/atoms/chat/current-thread.atom";
+import { mentionedDocumentsAtom } from "@/atoms/chat/mentioned-documents.atom";
 import { pendingUserImageDataUrlsAtom } from "@/atoms/chat/pending-user-images.atom";
+import {
+	clearPremiumAlertForThreadAtom,
+	premiumAlertByThreadAtom,
+} from "@/atoms/chat/premium-alert.atom";
 import { connectorDialogOpenAtom } from "@/atoms/connector-dialog/connector-dialog.atoms";
 import { connectorsAtom } from "@/atoms/connectors/connector-query.atoms";
 import { membersAtom } from "@/atoms/members/members-query.atoms";
@@ -134,6 +137,9 @@ const ThreadContent: FC = () => {
 					style={{ paddingBottom: "max(1rem, env(safe-area-inset-bottom))" }}
 				>
 					<ThreadScrollToBottom />
+					<AuiIf condition={({ thread }) => !thread.isEmpty}>
+						<PremiumQuotaPinnedAlert />
+					</AuiIf>
 					<AuiIf condition={({ thread }) => !thread.isEmpty}>
 						<Composer />
 					</AuiIf>
@@ -143,6 +149,38 @@ const ThreadContent: FC = () => {
 	);
 };
 
+const PremiumQuotaPinnedAlert: FC = () => {
+	const currentThreadState = useAtomValue(currentThreadAtom);
+	const alertsByThread = useAtomValue(premiumAlertByThreadAtom);
+	const clearPremiumAlertForThread = useSetAtom(clearPremiumAlertForThreadAtom);
+
+	const currentThreadId = currentThreadState?.id;
+	if (!currentThreadId) return null;
+
+	const alert = alertsByThread[currentThreadId];
+	if (!alert) return null;
+
+	return (
+		<div className="mx-2 rounded-2xl border border-amber-300/40 bg-amber-500/10 px-4 py-3 text-amber-50 shadow-lg backdrop-blur-sm">
+			<div className="flex items-start gap-2">
+				<AlertCircle className="mt-0.5 size-4 shrink-0 text-amber-300" />
+				<div className="min-w-0 flex-1">
+					<p className="text-sm font-medium">Premium quota exhausted</p>
+					<p className="mt-1 text-xs text-amber-100/90">{alert.message}</p>
+				</div>
+				<button
+					type="button"
+					className="inline-flex size-6 items-center justify-center rounded-md text-amber-200 transition-colors hover:bg-amber-200/20 hover:text-amber-50"
+					aria-label="Dismiss premium quota alert"
+					onClick={() => clearPremiumAlertForThread(currentThreadId)}
+				>
+					<X className="size-4" />
+				</button>
+			</div>
+		</div>
+	);
+};
+
 const ThreadScrollToBottom: FC = () => {
 	return (
 		<ThreadPrimitive.ScrollToBottom asChild>