Merge pull request #1162 from CREDO23/feat/vision-autocomplete

[Feat] Multi-suggestion autocomplete, Vision LLM config & Desktop analytics

surfsense_web/.env.example

@@ -7,4 +7,7 @@ NEXT_PUBLIC_ZERO_CACHE_URL=http://localhost:4848
 DATABASE_URL=postgresql://postgres:[YOUR-PASSWORD]@db.sdsf.supabase.co:5432/postgres
 
 # Deployment mode (optional)
-NEXT_PUBLIC_DEPLOYMENT_MODE="self-hosted" or "cloud"
+NEXT_PUBLIC_DEPLOYMENT_MODE="self-hosted" or "cloud"
+
+# PostHog analytics (optional, leave empty to disable)
+NEXT_PUBLIC_POSTHOG_KEY=

surfsense_web/app/dashboard/[search_space_id]/client-layout.tsx

@@ -154,11 +154,14 @@ export function DashboardClientLayout({
 
 		// Sync to Electron store if stored value is null (first navigation)
 		if (electronAPI?.setActiveSearchSpace) {
-			electronAPI.getActiveSearchSpace?.().then((stored) => {
-				if (!stored) {
-					electronAPI.setActiveSearchSpace!(activeSeacrhSpaceId);
-				}
-			}).catch(() => {});
+			electronAPI
+				.getActiveSearchSpace?.()
+				.then((stored) => {
+					if (!stored) {
+						electronAPI.setActiveSearchSpace!(activeSeacrhSpaceId);
+					}
+				})
+				.catch(() => {});
 		}
 	}, [search_space_id, setActiveSearchSpaceIdState, electronAPI]);
 

surfsense_web/app/dashboard/[search_space_id]/user-settings/components/DesktopContent.tsx

@@ -6,12 +6,18 @@ import { toast } from "sonner";
 import { DEFAULT_SHORTCUTS, ShortcutRecorder } from "@/components/desktop/shortcut-recorder";
 import { Card, CardContent, CardDescription, CardHeader, CardTitle } from "@/components/ui/card";
 import { Label } from "@/components/ui/label";
-import { Select, SelectContent, SelectItem, SelectTrigger, SelectValue } from "@/components/ui/select";
+import {
+	Select,
+	SelectContent,
+	SelectItem,
+	SelectTrigger,
+	SelectValue,
+} from "@/components/ui/select";
 import { Spinner } from "@/components/ui/spinner";
 import { Switch } from "@/components/ui/switch";
+import type { SearchSpace } from "@/contracts/types/search-space.types";
 import { useElectronAPI } from "@/hooks/use-platform";
 import { searchSpacesApiService } from "@/lib/apis/search-spaces-api.service";
-import type { SearchSpace } from "@/contracts/types/search-space.types";
 
 export function DesktopContent() {
 	const api = useElectronAPI();
@@ -82,7 +88,10 @@ export function DesktopContent() {
 		await api.setAutocompleteEnabled(checked);
 	};
 
-	const updateShortcut = (key: "generalAssist" | "quickAsk" | "autocomplete", accelerator: string) => {
+	const updateShortcut = (
+		key: "generalAssist" | "quickAsk" | "autocomplete",
+		accelerator: string
+	) => {
 		setShortcuts((prev) => {
 			const updated = { ...prev, [key]: accelerator };
 			api.setShortcuts?.({ [key]: accelerator }).catch(() => {
@@ -110,7 +119,8 @@ export function DesktopContent() {
 				<CardHeader className="px-3 md:px-6 pt-3 md:pt-6 pb-2 md:pb-3">
 					<CardTitle className="text-base md:text-lg">Default Search Space</CardTitle>
 					<CardDescription className="text-xs md:text-sm">
-						Choose which search space General Assist, Quick Assist, and Extreme Assist operate against.
+						Choose which search space General Assist, Quick Assist, and Extreme Assist operate
+						against.
 					</CardDescription>
 				</CardHeader>
 				<CardContent className="px-3 md:px-6 pb-3 md:pb-6">
@@ -128,7 +138,9 @@ export function DesktopContent() {
 							</SelectContent>
 						</Select>
 					) : (
-						<p className="text-sm text-muted-foreground">No search spaces found. Create one first.</p>
+						<p className="text-sm text-muted-foreground">
+							No search spaces found. Create one first.
+						</p>
 					)}
 				</CardContent>
 			</Card>
@@ -143,34 +155,34 @@ export function DesktopContent() {
 				</CardHeader>
 				<CardContent className="px-3 md:px-6 pb-3 md:pb-6">
 					{shortcutsLoaded ? (
-					<div className="flex flex-col gap-3">
-						<ShortcutRecorder
-							value={shortcuts.generalAssist}
-							onChange={(accel) => updateShortcut("generalAssist", accel)}
-							onReset={() => resetShortcut("generalAssist")}
-							defaultValue={DEFAULT_SHORTCUTS.generalAssist}
-							label="General Assist"
-							description="Launch SurfSense instantly from any application"
-							icon={Rocket}
-						/>
-						<ShortcutRecorder
-							value={shortcuts.quickAsk}
-							onChange={(accel) => updateShortcut("quickAsk", accel)}
-							onReset={() => resetShortcut("quickAsk")}
-							defaultValue={DEFAULT_SHORTCUTS.quickAsk}
-							label="Quick Assist"
-							description="Select text anywhere, then ask AI to explain, rewrite, or act on it"
-							icon={Zap}
-						/>
-						<ShortcutRecorder
-							value={shortcuts.autocomplete}
-							onChange={(accel) => updateShortcut("autocomplete", accel)}
-							onReset={() => resetShortcut("autocomplete")}
-							defaultValue={DEFAULT_SHORTCUTS.autocomplete}
-							label="Extreme Assist"
-							description="AI drafts text using your screen context and knowledge base"
-							icon={BrainCog}
-						/>
+						<div className="flex flex-col gap-3">
+							<ShortcutRecorder
+								value={shortcuts.generalAssist}
+								onChange={(accel) => updateShortcut("generalAssist", accel)}
+								onReset={() => resetShortcut("generalAssist")}
+								defaultValue={DEFAULT_SHORTCUTS.generalAssist}
+								label="General Assist"
+								description="Launch SurfSense instantly from any application"
+								icon={Rocket}
+							/>
+							<ShortcutRecorder
+								value={shortcuts.quickAsk}
+								onChange={(accel) => updateShortcut("quickAsk", accel)}
+								onReset={() => resetShortcut("quickAsk")}
+								defaultValue={DEFAULT_SHORTCUTS.quickAsk}
+								label="Quick Assist"
+								description="Select text anywhere, then ask AI to explain, rewrite, or act on it"
+								icon={Zap}
+							/>
+							<ShortcutRecorder
+								value={shortcuts.autocomplete}
+								onChange={(accel) => updateShortcut("autocomplete", accel)}
+								onReset={() => resetShortcut("autocomplete")}
+								defaultValue={DEFAULT_SHORTCUTS.autocomplete}
+								label="Extreme Assist"
+								description="AI drafts text using your screen context and knowledge base"
+								icon={BrainCog}
+							/>
 							<p className="text-[11px] text-muted-foreground">
 								Click a shortcut and press a new key combination to change it.
 							</p>

surfsense_web/app/desktop/login/page.tsx

@@ -139,9 +139,7 @@ export default function DesktopLoginPage() {
 						height={48}
 						priority
 					/>
-					<h1 className="text-lg font-semibold tracking-tight">
-						Welcome to SurfSense Desktop
-					</h1>
+					<h1 className="text-lg font-semibold tracking-tight">Welcome to SurfSense Desktop</h1>
 					<p className="mt-1 text-sm text-muted-foreground">
 						Configure shortcuts, then sign in to get started.
 					</p>

surfsense_web/app/desktop/suggestion/page.tsx

@@ -14,6 +14,10 @@ type SSEEvent =
 	| {
 			type: "data-thinking-step";
 			data: { id: string; title: string; status: string; items: string[] };
+	  }
+	| {
+			type: "data-suggestions";
+			data: { options: string[] };
 	  };
 
 interface AgentStep {
@@ -23,24 +27,32 @@ interface AgentStep {
 	items: string[];
 }
 
-function friendlyError(raw: string | number): string {
+type FriendlyError = { message: string; isSetup?: boolean };
+
+function friendlyError(raw: string | number): FriendlyError {
 	if (typeof raw === "number") {
-		if (raw === 401) return "Please sign in to use suggestions.";
-		if (raw === 403) return "You don\u2019t have permission for this.";
-		if (raw === 404) return "Suggestion service not found. Is the backend running?";
-		if (raw >= 500) return "Something went wrong on the server. Try again.";
-		return "Something went wrong. Try again.";
+		if (raw === 401) return { message: "Please sign in to use suggestions." };
+		if (raw === 403) return { message: "You don\u2019t have permission for this." };
+		if (raw === 404) return { message: "Suggestion service not found. Is the backend running?" };
+		if (raw >= 500) return { message: "Something went wrong on the server. Try again." };
+		return { message: "Something went wrong. Try again." };
 	}
 	const lower = raw.toLowerCase();
 	if (lower.includes("not authenticated") || lower.includes("unauthorized"))
-		return "Please sign in to use suggestions.";
+		return { message: "Please sign in to use suggestions." };
 	if (lower.includes("no vision llm configured") || lower.includes("no llm configured"))
-		return "No Vision LLM configured. Set one in search space settings.";
+		return {
+			message: "Configure a vision-capable model (e.g. GPT-4o, Gemini) to enable autocomplete.",
+			isSetup: true,
+		};
 	if (lower.includes("does not support vision"))
-		return "Selected model doesn\u2019t support vision. Set a vision-capable model in settings.";
+		return {
+			message: "The selected model doesn\u2019t support vision. Choose a vision-capable model.",
+			isSetup: true,
+		};
 	if (lower.includes("fetch") || lower.includes("network") || lower.includes("econnrefused"))
-		return "Can\u2019t reach the server. Check your connection.";
-	return "Something went wrong. Try again.";
+		return { message: "Can\u2019t reach the server. Check your connection." };
+	return { message: "Something went wrong. Try again." };
 }
 
 const AUTO_DISMISS_MS = 3000;
@@ -70,10 +82,11 @@ function StepIcon({ status }: { status: string }) {
 
 export default function SuggestionPage() {
 	const api = useElectronAPI();
-	const [suggestion, setSuggestion] = useState("");
+	const [options, setOptions] = useState<string[]>([]);
 	const [isLoading, setIsLoading] = useState(true);
-	const [error, setError] = useState<string | null>(null);
+	const [error, setError] = useState<FriendlyError | null>(null);
 	const [steps, setSteps] = useState<AgentStep[]>([]);
+	const [expandedOption, setExpandedOption] = useState<number | null>(null);
 	const abortRef = useRef<AbortController | null>(null);
 
 	const isDesktop = !!api?.onAutocompleteContext;
@@ -85,13 +98,21 @@ export default function SuggestionPage() {
 	}, [api]);
 
 	useEffect(() => {
-		if (!error) return;
+		if (!error || error.isSetup) return;
 		const timer = setTimeout(() => {
 			api?.dismissSuggestion?.();
 		}, AUTO_DISMISS_MS);
 		return () => clearTimeout(timer);
 	}, [error, api]);
 
+	useEffect(() => {
+		if (isLoading || error || options.length > 0) return;
+		const timer = setTimeout(() => {
+			api?.dismissSuggestion?.();
+		}, AUTO_DISMISS_MS);
+		return () => clearTimeout(timer);
+	}, [isLoading, error, options, api]);
+
 	const fetchSuggestion = useCallback(
 		async (screenshot: string, searchSpaceId: string, appName?: string, windowTitle?: string) => {
 			abortRef.current?.abort();
@@ -99,9 +120,10 @@ export default function SuggestionPage() {
 			abortRef.current = controller;
 
 			setIsLoading(true);
-			setSuggestion("");
+			setOptions([]);
 			setError(null);
 			setSteps([]);
+			setExpandedOption(null);
 
 			let token = getBearerToken();
 			if (!token) {
@@ -165,8 +187,8 @@ export default function SuggestionPage() {
 
 							try {
 								const parsed: SSEEvent = JSON.parse(data);
-								if (parsed.type === "text-delta") {
-									setSuggestion((prev) => prev + parsed.delta);
+								if (parsed.type === "data-suggestions") {
+									setOptions(parsed.data.options);
 								} else if (parsed.type === "error") {
 									setError(friendlyError(parsed.errorText));
 								} else if (parsed.type === "data-thinking-step") {
@@ -219,14 +241,52 @@ export default function SuggestionPage() {
 	}
 
 	if (error) {
+		if (error.isSetup) {
+			return (
+				<div className="suggestion-tooltip suggestion-setup">
+					<div className="setup-icon">
+						<svg viewBox="0 0 24 24" fill="none" width="28" height="28" aria-hidden="true">
+							<path
+								d="M1 12C1 12 5 4 12 4C19 4 23 12 23 12C23 12 19 20 12 20C5 20 1 12 1 12Z"
+								stroke="#a78bfa"
+								strokeWidth="1.5"
+								strokeLinecap="round"
+								strokeLinejoin="round"
+							/>
+							<circle
+								cx="12"
+								cy="12"
+								r="3"
+								stroke="#a78bfa"
+								strokeWidth="1.5"
+								strokeLinecap="round"
+								strokeLinejoin="round"
+							/>
+						</svg>
+					</div>
+					<div className="setup-content">
+						<span className="setup-title">Vision Model Required</span>
+						<span className="setup-message">{error.message}</span>
+						<span className="setup-hint">Settings → Vision Models</span>
+					</div>
+					<button
+						type="button"
+						className="setup-dismiss"
+						onClick={() => api?.dismissSuggestion?.()}
+					>
+						✕
+					</button>
+				</div>
+			);
+		}
 		return (
 			<div className="suggestion-tooltip suggestion-error">
-				<span className="suggestion-error-text">{error}</span>
+				<span className="suggestion-error-text">{error.message}</span>
 			</div>
 		);
 	}
 
-	const showLoading = isLoading && !suggestion;
+	const showLoading = isLoading && options.length === 0;
 
 	if (showLoading) {
 		return (
@@ -258,29 +318,63 @@ export default function SuggestionPage() {
 		);
 	}
 
-	const handleAccept = () => {
-		if (suggestion) {
-			api?.acceptSuggestion?.(suggestion);
-		}
+	const handleSelect = (text: string) => {
+		api?.acceptSuggestion?.(text);
 	};
 
 	const handleDismiss = () => {
 		api?.dismissSuggestion?.();
 	};
 
-	if (!suggestion) return null;
+	const TRUNCATE_LENGTH = 120;
+
+	if (options.length === 0) {
+		return (
+			<div className="suggestion-tooltip suggestion-error">
+				<span className="suggestion-error-text">No suggestions available.</span>
+			</div>
+		);
+	}
 
 	return (
 		<div className="suggestion-tooltip">
-			<p className="suggestion-text">{suggestion}</p>
+			<div className="suggestion-options">
+				{options.map((option, index) => {
+					const isExpanded = expandedOption === index;
+					const needsTruncation = option.length > TRUNCATE_LENGTH;
+					const displayText =
+						needsTruncation && !isExpanded ? option.slice(0, TRUNCATE_LENGTH) + "…" : option;
+
+					return (
+						<div
+							key={index}
+							role="button"
+							tabIndex={0}
+							className="suggestion-option"
+							onClick={() => handleSelect(option)}
+							onKeyDown={(e) => {
+								if (e.key === "Enter") handleSelect(option);
+							}}
+						>
+							<span className="option-number">{index + 1}</span>
+							<span className="option-text">{displayText}</span>
+							{needsTruncation && (
+								<button
+									type="button"
+									className="option-expand"
+									onClick={(e) => {
+										e.stopPropagation();
+										setExpandedOption(isExpanded ? null : index);
+									}}
+								>
+									{isExpanded ? "less" : "more"}
+								</button>
+							)}
+						</div>
+					);
+				})}
+			</div>
 			<div className="suggestion-actions">
-				<button
-					type="button"
-					className="suggestion-btn suggestion-btn-accept"
-					onClick={handleAccept}
-				>
-					Accept
-				</button>
 				<button
 					type="button"
 					className="suggestion-btn suggestion-btn-dismiss"

surfsense_web/app/desktop/suggestion/suggestion.css

@@ -117,6 +117,66 @@ body:has(.suggestion-body) {
 	font-size: 12px;
 }
 
+/* --- Setup prompt (vision model not configured) --- */
+
+.suggestion-setup {
+	display: flex;
+	flex-direction: row;
+	align-items: flex-start;
+	gap: 10px;
+	border-color: #3b2d6b;
+	padding: 10px 14px;
+}
+
+.setup-icon {
+	flex-shrink: 0;
+	margin-top: 1px;
+}
+
+.setup-content {
+	display: flex;
+	flex-direction: column;
+	gap: 3px;
+	min-width: 0;
+}
+
+.setup-title {
+	font-size: 13px;
+	font-weight: 600;
+	color: #c4b5fd;
+}
+
+.setup-message {
+	font-size: 11.5px;
+	color: #a1a1aa;
+	line-height: 1.4;
+}
+
+.setup-hint {
+	font-size: 10.5px;
+	color: #7c6dac;
+	margin-top: 2px;
+}
+
+.setup-dismiss {
+	flex-shrink: 0;
+	align-self: flex-start;
+	background: none;
+	border: none;
+	color: #6b6b7b;
+	font-size: 14px;
+	cursor: pointer;
+	padding: 2px 4px;
+	line-height: 1;
+	border-radius: 4px;
+	transition: color 0.15s, background 0.15s;
+}
+
+.setup-dismiss:hover {
+	color: #c4b5fd;
+	background: rgba(124, 109, 172, 0.15);
+}
+
 /* --- Agent activity indicator --- */
 
 .agent-activity {
@@ -127,6 +187,10 @@ body:has(.suggestion-body) {
 	max-height: 340px;
 }
 
+.agent-activity::-webkit-scrollbar {
+	display: none;
+}
+
 .activity-initial {
 	display: flex;
 	align-items: center;
@@ -191,3 +255,96 @@ body:has(.suggestion-body) {
 		transform: rotate(360deg);
 	}
 }
+
+/* --- Suggestion option cards --- */
+
+.suggestion-options {
+	display: flex;
+	flex-direction: column;
+	gap: 4px;
+	overflow-y: auto;
+	flex: 1 1 auto;
+	min-height: 0;
+	margin-bottom: 6px;
+}
+
+.suggestion-options::-webkit-scrollbar {
+	width: 5px;
+}
+
+.suggestion-options::-webkit-scrollbar-track {
+	background: transparent;
+}
+
+.suggestion-options::-webkit-scrollbar-thumb {
+	background: #555;
+	border-radius: 3px;
+}
+
+.suggestion-option {
+	display: flex;
+	align-items: flex-start;
+	gap: 8px;
+	padding: 6px 8px;
+	border-radius: 5px;
+	border: 1px solid #333;
+	background: #262626;
+	cursor: pointer;
+	text-align: left;
+	font-family: inherit;
+	transition:
+		background 0.15s,
+		border-color 0.15s;
+	width: 100%;
+}
+
+.suggestion-option:hover {
+	background: #2a2d3a;
+	border-color: #3b82f6;
+}
+
+.option-number {
+	flex-shrink: 0;
+	width: 18px;
+	height: 18px;
+	border-radius: 50%;
+	background: #3f3f46;
+	color: #d4d4d4;
+	font-size: 10px;
+	font-weight: 600;
+	display: flex;
+	align-items: center;
+	justify-content: center;
+	margin-top: 1px;
+}
+
+.suggestion-option:hover .option-number {
+	background: #2563eb;
+	color: #fff;
+}
+
+.option-text {
+	color: #d4d4d4;
+	font-size: 12px;
+	line-height: 1.45;
+	word-wrap: break-word;
+	white-space: pre-wrap;
+	flex: 1 1 auto;
+	min-width: 0;
+}
+
+.option-expand {
+	flex-shrink: 0;
+	background: none;
+	border: none;
+	color: #71717a;
+	font-size: 10px;
+	cursor: pointer;
+	padding: 0 2px;
+	font-family: inherit;
+	margin-top: 1px;
+}
+
+.option-expand:hover {
+	color: #a1a1aa;
+}

surfsense_web/atoms/vision-llm-config/vision-llm-config-mutation.atoms.ts

@@ -0,0 +1,84 @@
+import { atomWithMutation } from "jotai-tanstack-query";
+import { toast } from "sonner";
+import type {
+	CreateVisionLLMConfigRequest,
+	CreateVisionLLMConfigResponse,
+	DeleteVisionLLMConfigResponse,
+	GetVisionLLMConfigsResponse,
+	UpdateVisionLLMConfigRequest,
+	UpdateVisionLLMConfigResponse,
+} from "@/contracts/types/new-llm-config.types";
+import { visionLLMConfigApiService } from "@/lib/apis/vision-llm-config-api.service";
+import { cacheKeys } from "@/lib/query-client/cache-keys";
+import { queryClient } from "@/lib/query-client/client";
+import { activeSearchSpaceIdAtom } from "../search-spaces/search-space-query.atoms";
+
+export const createVisionLLMConfigMutationAtom = atomWithMutation((get) => {
+	const searchSpaceId = get(activeSearchSpaceIdAtom);
+
+	return {
+		mutationKey: ["vision-llm-configs", "create"],
+		enabled: !!searchSpaceId,
+		mutationFn: async (request: CreateVisionLLMConfigRequest) => {
+			return visionLLMConfigApiService.createConfig(request);
+		},
+		onSuccess: (_: CreateVisionLLMConfigResponse, request: CreateVisionLLMConfigRequest) => {
+			toast.success(`${request.name} created`);
+			queryClient.invalidateQueries({
+				queryKey: cacheKeys.visionLLMConfigs.all(Number(searchSpaceId)),
+			});
+		},
+		onError: (error: Error) => {
+			toast.error(error.message || "Failed to create vision model");
+		},
+	};
+});
+
+export const updateVisionLLMConfigMutationAtom = atomWithMutation((get) => {
+	const searchSpaceId = get(activeSearchSpaceIdAtom);
+
+	return {
+		mutationKey: ["vision-llm-configs", "update"],
+		enabled: !!searchSpaceId,
+		mutationFn: async (request: UpdateVisionLLMConfigRequest) => {
+			return visionLLMConfigApiService.updateConfig(request);
+		},
+		onSuccess: (_: UpdateVisionLLMConfigResponse, request: UpdateVisionLLMConfigRequest) => {
+			toast.success(`${request.data.name ?? "Configuration"} updated`);
+			queryClient.invalidateQueries({
+				queryKey: cacheKeys.visionLLMConfigs.all(Number(searchSpaceId)),
+			});
+			queryClient.invalidateQueries({
+				queryKey: cacheKeys.visionLLMConfigs.byId(request.id),
+			});
+		},
+		onError: (error: Error) => {
+			toast.error(error.message || "Failed to update vision model");
+		},
+	};
+});
+
+export const deleteVisionLLMConfigMutationAtom = atomWithMutation((get) => {
+	const searchSpaceId = get(activeSearchSpaceIdAtom);
+
+	return {
+		mutationKey: ["vision-llm-configs", "delete"],
+		enabled: !!searchSpaceId,
+		mutationFn: async (request: { id: number; name: string }) => {
+			return visionLLMConfigApiService.deleteConfig(request.id);
+		},
+		onSuccess: (_: DeleteVisionLLMConfigResponse, request: { id: number; name: string }) => {
+			toast.success(`${request.name} deleted`);
+			queryClient.setQueryData(
+				cacheKeys.visionLLMConfigs.all(Number(searchSpaceId)),
+				(oldData: GetVisionLLMConfigsResponse | undefined) => {
+					if (!oldData) return oldData;
+					return oldData.filter((config) => config.id !== request.id);
+				}
+			);
+		},
+		onError: (error: Error) => {
+			toast.error(error.message || "Failed to delete vision model");
+		},
+	};
+});

surfsense_web/atoms/vision-llm-config/vision-llm-config-query.atoms.ts

@@ -0,0 +1,27 @@
+import { atomWithQuery } from "jotai-tanstack-query";
+import { visionLLMConfigApiService } from "@/lib/apis/vision-llm-config-api.service";
+import { cacheKeys } from "@/lib/query-client/cache-keys";
+import { activeSearchSpaceIdAtom } from "../search-spaces/search-space-query.atoms";
+
+export const visionLLMConfigsAtom = atomWithQuery((get) => {
+	const searchSpaceId = get(activeSearchSpaceIdAtom);
+
+	return {
+		queryKey: cacheKeys.visionLLMConfigs.all(Number(searchSpaceId)),
+		enabled: !!searchSpaceId,
+		staleTime: 5 * 60 * 1000,
+		queryFn: async () => {
+			return visionLLMConfigApiService.getConfigs(Number(searchSpaceId));
+		},
+	};
+});
+
+export const globalVisionLLMConfigsAtom = atomWithQuery(() => {
+	return {
+		queryKey: cacheKeys.visionLLMConfigs.global(),
+		staleTime: 10 * 60 * 1000,
+		queryFn: async () => {
+			return visionLLMConfigApiService.getGlobalConfigs();
+		},
+	};
+});

surfsense_web/components/TokenHandler.tsx

@@ -2,8 +2,8 @@
 
 import { useEffect } from "react";
 import { useGlobalLoadingEffect } from "@/hooks/use-global-loading";
-import { getAndClearRedirectPath, setBearerToken, setRefreshToken } from "@/lib/auth-utils";
 import { searchSpacesApiService } from "@/lib/apis/search-spaces-api.service";
+import { getAndClearRedirectPath, setBearerToken, setRefreshToken } from "@/lib/auth-utils";
 import { trackLoginSuccess } from "@/lib/posthog/events";
 
 interface TokenHandlerProps {

surfsense_web/components/desktop/shortcut-recorder.tsx

@@ -150,7 +150,9 @@ export function ShortcutRecorder({
 					)}
 				>
 					{recording ? (
-						<span className="text-[11px] text-primary animate-pulse whitespace-nowrap">Press keys…</span>
+						<span className="text-[11px] text-primary animate-pulse whitespace-nowrap">
+							Press keys…
+						</span>
 					) : (
 						<Kbd keys={displayKeys} />
 					)}

surfsense_web/components/new-chat/chat-header.tsx

@@ -3,11 +3,14 @@
 import { useCallback, useState } from "react";
 import { ImageConfigDialog } from "@/components/shared/image-config-dialog";
 import { ModelConfigDialog } from "@/components/shared/model-config-dialog";
+import { VisionConfigDialog } from "@/components/shared/vision-config-dialog";
 import type {
 	GlobalImageGenConfig,
 	GlobalNewLLMConfig,
+	GlobalVisionLLMConfig,
 	ImageGenerationConfig,
 	NewLLMConfigPublic,
+	VisionLLMConfig,
 } from "@/contracts/types/new-llm-config.types";
 import { ModelSelector } from "./model-selector";
 
@@ -33,6 +36,14 @@ export function ChatHeader({ searchSpaceId, className }: ChatHeaderProps) {
 	const [isImageGlobal, setIsImageGlobal] = useState(false);
 	const [imageDialogMode, setImageDialogMode] = useState<"create" | "edit" | "view">("view");
 
+	// Vision config dialog state
+	const [visionDialogOpen, setVisionDialogOpen] = useState(false);
+	const [selectedVisionConfig, setSelectedVisionConfig] = useState<
+		VisionLLMConfig | GlobalVisionLLMConfig | null
+	>(null);
+	const [isVisionGlobal, setIsVisionGlobal] = useState(false);
+	const [visionDialogMode, setVisionDialogMode] = useState<"create" | "edit" | "view">("view");
+
 	// LLM handlers
 	const handleEditLLMConfig = useCallback(
 		(config: NewLLMConfigPublic | GlobalNewLLMConfig, global: boolean) => {
@@ -79,6 +90,29 @@ export function ChatHeader({ searchSpaceId, className }: ChatHeaderProps) {
 		if (!open) setSelectedImageConfig(null);
 	}, []);
 
+	// Vision model handlers
+	const handleAddVisionModel = useCallback(() => {
+		setSelectedVisionConfig(null);
+		setIsVisionGlobal(false);
+		setVisionDialogMode("create");
+		setVisionDialogOpen(true);
+	}, []);
+
+	const handleEditVisionConfig = useCallback(
+		(config: VisionLLMConfig | GlobalVisionLLMConfig, global: boolean) => {
+			setSelectedVisionConfig(config);
+			setIsVisionGlobal(global);
+			setVisionDialogMode(global ? "view" : "edit");
+			setVisionDialogOpen(true);
+		},
+		[]
+	);
+
+	const handleVisionDialogClose = useCallback((open: boolean) => {
+		setVisionDialogOpen(open);
+		if (!open) setSelectedVisionConfig(null);
+	}, []);
+
 	return (
 		<div className="flex items-center gap-2">
 			<ModelSelector
@@ -86,6 +120,8 @@ export function ChatHeader({ searchSpaceId, className }: ChatHeaderProps) {
 				onAddNewLLM={handleAddNewLLM}
 				onEditImage={handleEditImageConfig}
 				onAddNewImage={handleAddImageModel}
+				onEditVision={handleEditVisionConfig}
+				onAddNewVision={handleAddVisionModel}
 				className={className}
 			/>
 			<ModelConfigDialog
@@ -104,6 +140,14 @@ export function ChatHeader({ searchSpaceId, className }: ChatHeaderProps) {
 				searchSpaceId={searchSpaceId}
 				mode={imageDialogMode}
 			/>
+			<VisionConfigDialog
+				open={visionDialogOpen}
+				onOpenChange={handleVisionDialogClose}
+				config={selectedVisionConfig}
+				isGlobal={isVisionGlobal}
+				searchSpaceId={searchSpaceId}
+				mode={visionDialogMode}
+			/>
 		</div>
 	);
 }

surfsense_web/components/new-chat/model-selector.tsx

@@ -1,7 +1,7 @@
 "use client";
 
 import { useAtomValue } from "jotai";
-import { Bot, Check, ChevronDown, Edit3, ImageIcon, Plus, Search, Zap } from "lucide-react";
+import { Bot, Check, ChevronDown, Edit3, Eye, ImageIcon, Plus, Search, Zap } from "lucide-react";
 import { type UIEvent, useCallback, useMemo, useState } from "react";
 import { toast } from "sonner";
 import {
@@ -15,6 +15,10 @@ import {
 	newLLMConfigsAtom,
 } from "@/atoms/new-llm-config/new-llm-config-query.atoms";
 import { activeSearchSpaceIdAtom } from "@/atoms/search-spaces/search-space-query.atoms";
+import {
+	globalVisionLLMConfigsAtom,
+	visionLLMConfigsAtom,
+} from "@/atoms/vision-llm-config/vision-llm-config-query.atoms";
 import { Badge } from "@/components/ui/badge";
 import { Button } from "@/components/ui/button";
 import {
@@ -32,8 +36,10 @@ import { Tabs, TabsContent, TabsList, TabsTrigger } from "@/components/ui/tabs";
 import type {
 	GlobalImageGenConfig,
 	GlobalNewLLMConfig,
+	GlobalVisionLLMConfig,
 	ImageGenerationConfig,
 	NewLLMConfigPublic,
+	VisionLLMConfig,
 } from "@/contracts/types/new-llm-config.types";
 import { getProviderIcon } from "@/lib/provider-icons";
 import { cn } from "@/lib/utils";
@@ -43,6 +49,8 @@ interface ModelSelectorProps {
 	onAddNewLLM: () => void;
 	onEditImage?: (config: ImageGenerationConfig | GlobalImageGenConfig, isGlobal: boolean) => void;
 	onAddNewImage?: () => void;
+	onEditVision?: (config: VisionLLMConfig | GlobalVisionLLMConfig, isGlobal: boolean) => void;
+	onAddNewVision?: () => void;
 	className?: string;
 }
 
@@ -51,14 +59,18 @@ export function ModelSelector({
 	onAddNewLLM,
 	onEditImage,
 	onAddNewImage,
+	onEditVision,
+	onAddNewVision,
 	className,
 }: ModelSelectorProps) {
 	const [open, setOpen] = useState(false);
-	const [activeTab, setActiveTab] = useState<"llm" | "image">("llm");
+	const [activeTab, setActiveTab] = useState<"llm" | "image" | "vision">("llm");
 	const [llmSearchQuery, setLlmSearchQuery] = useState("");
 	const [imageSearchQuery, setImageSearchQuery] = useState("");
+	const [visionSearchQuery, setVisionSearchQuery] = useState("");
 	const [llmScrollPos, setLlmScrollPos] = useState<"top" | "middle" | "bottom">("top");
 	const [imageScrollPos, setImageScrollPos] = useState<"top" | "middle" | "bottom">("top");
+	const [visionScrollPos, setVisionScrollPos] = useState<"top" | "middle" | "bottom">("top");
 	const handleListScroll = useCallback(
 		(setter: typeof setLlmScrollPos) => (e: UIEvent<HTMLDivElement>) => {
 			const el = e.currentTarget;
@@ -82,8 +94,21 @@ export function ModelSelector({
 		useAtomValue(globalImageGenConfigsAtom);
 	const { data: imageUserConfigs, isLoading: imageUserLoading } = useAtomValue(imageGenConfigsAtom);
 
+	// Vision data
+	const { data: visionGlobalConfigs, isLoading: visionGlobalLoading } = useAtomValue(
+		globalVisionLLMConfigsAtom
+	);
+	const { data: visionUserConfigs, isLoading: visionUserLoading } =
+		useAtomValue(visionLLMConfigsAtom);
+
 	const isLoading =
-		llmUserLoading || llmGlobalLoading || prefsLoading || imageGlobalLoading || imageUserLoading;
+		llmUserLoading ||
+		llmGlobalLoading ||
+		prefsLoading ||
+		imageGlobalLoading ||
+		imageUserLoading ||
+		visionGlobalLoading ||
+		visionUserLoading;
 
 	// ─── LLM current config ───
 	const currentLLMConfig = useMemo(() => {
@@ -116,6 +141,24 @@ export function ModelSelector({
 		);
 	}, [currentImageConfig]);
 
+	// ─── Vision current config ───
+	const currentVisionConfig = useMemo(() => {
+		if (!preferences) return null;
+		const id = preferences.vision_llm_config_id;
+		if (id === null || id === undefined) return null;
+		const globalMatch = visionGlobalConfigs?.find((c) => c.id === id);
+		if (globalMatch) return globalMatch;
+		return visionUserConfigs?.find((c) => c.id === id) ?? null;
+	}, [preferences, visionGlobalConfigs, visionUserConfigs]);
+
+	const isVisionAutoMode = useMemo(() => {
+		return (
+			currentVisionConfig &&
+			"is_auto_mode" in currentVisionConfig &&
+			currentVisionConfig.is_auto_mode
+		);
+	}, [currentVisionConfig]);
+
 	// ─── LLM filtering ───
 	const filteredLLMGlobal = useMemo(() => {
 		if (!llmGlobalConfigs) return [];
@@ -170,6 +213,33 @@ export function ModelSelector({
 
 	const totalImageModels = (imageGlobalConfigs?.length ?? 0) + (imageUserConfigs?.length ?? 0);
 
+	// ─── Vision filtering ───
+	const filteredVisionGlobal = useMemo(() => {
+		if (!visionGlobalConfigs) return [];
+		if (!visionSearchQuery) return visionGlobalConfigs;
+		const q = visionSearchQuery.toLowerCase();
+		return visionGlobalConfigs.filter(
+			(c) =>
+				c.name.toLowerCase().includes(q) ||
+				c.model_name.toLowerCase().includes(q) ||
+				c.provider.toLowerCase().includes(q)
+		);
+	}, [visionGlobalConfigs, visionSearchQuery]);
+
+	const filteredVisionUser = useMemo(() => {
+		if (!visionUserConfigs) return [];
+		if (!visionSearchQuery) return visionUserConfigs;
+		const q = visionSearchQuery.toLowerCase();
+		return visionUserConfigs.filter(
+			(c) =>
+				c.name.toLowerCase().includes(q) ||
+				c.model_name.toLowerCase().includes(q) ||
+				c.provider.toLowerCase().includes(q)
+		);
+	}, [visionUserConfigs, visionSearchQuery]);
+
+	const totalVisionModels = (visionGlobalConfigs?.length ?? 0) + (visionUserConfigs?.length ?? 0);
+
 	// ─── Handlers ───
 	const handleSelectLLM = useCallback(
 		async (config: NewLLMConfigPublic | GlobalNewLLMConfig) => {
@@ -229,6 +299,30 @@ export function ModelSelector({
 		[currentImageConfig, searchSpaceId, updatePreferences]
 	);
 
+	const handleSelectVision = useCallback(
+		async (configId: number) => {
+			if (currentVisionConfig?.id === configId) {
+				setOpen(false);
+				return;
+			}
+			if (!searchSpaceId) {
+				toast.error("No search space selected");
+				return;
+			}
+			try {
+				await updatePreferences({
+					search_space_id: Number(searchSpaceId),
+					data: { vision_llm_config_id: configId },
+				});
+				toast.success("Vision model updated");
+				setOpen(false);
+			} catch {
+				toast.error("Failed to switch vision model");
+			}
+		},
+		[currentVisionConfig, searchSpaceId, updatePreferences]
+	);
+
 	return (
 		<Popover open={open} onOpenChange={setOpen}>
 			<PopoverTrigger asChild>
@@ -282,6 +376,23 @@ export function ModelSelector({
 							) : (
 								<ImageIcon className="size-4 text-muted-foreground" />
 							)}
+
+							{/* Divider */}
+							<div className="h-4 w-px bg-border/60 dark:bg-white/10 mx-0.5" />
+
+							{/* Vision section */}
+							{currentVisionConfig ? (
+								<>
+									{getProviderIcon(currentVisionConfig.provider, {
+										isAutoMode: isVisionAutoMode ?? false,
+									})}
+									<span className="max-w-[80px] md:max-w-[100px] truncate hidden md:inline">
+										{currentVisionConfig.name}
+									</span>
+								</>
+							) : (
+								<Eye className="size-4 text-muted-foreground" />
+							)}
 						</>
 					)}
 					<ChevronDown className="h-3.5 w-3.5 text-muted-foreground ml-1 shrink-0" />
@@ -295,25 +406,32 @@ export function ModelSelector({
 			>
 				<Tabs
 					value={activeTab}
-					onValueChange={(v) => setActiveTab(v as "llm" | "image")}
+					onValueChange={(v) => setActiveTab(v as "llm" | "image" | "vision")}
 					className="w-full"
 				>
 					<div className="border-b border-border/80 dark:border-neutral-800">
-						<TabsList className="w-full grid grid-cols-2 rounded-none rounded-t-lg bg-transparent h-11 p-0 gap-0">
+						<TabsList className="w-full grid grid-cols-3 rounded-none rounded-t-lg bg-transparent h-11 p-0 gap-0">
 							<TabsTrigger
 								value="llm"
-								className="gap-2 text-sm font-medium rounded-none text-muted-foreground transition-all duration-200 h-full bg-transparent data-[state=active]:bg-transparent shadow-none data-[state=active]:shadow-none border-b-[1.5px] border-transparent data-[state=active]:border-foreground dark:data-[state=active]:border-white data-[state=active]:text-foreground"
+								className="gap-1.5 text-sm font-medium rounded-none text-muted-foreground transition-all duration-200 h-full bg-transparent data-[state=active]:bg-transparent shadow-none data-[state=active]:shadow-none border-b-[1.5px] border-transparent data-[state=active]:border-foreground dark:data-[state=active]:border-white data-[state=active]:text-foreground"
 							>
-								<Zap className="size-4" />
+								<Zap className="size-3.5" />
 								LLM
 							</TabsTrigger>
 							<TabsTrigger
 								value="image"
-								className="gap-2 text-sm font-medium rounded-none text-muted-foreground transition-all duration-200 h-full bg-transparent data-[state=active]:bg-transparent shadow-none data-[state=active]:shadow-none border-b-[1.5px] border-transparent data-[state=active]:border-foreground dark:data-[state=active]:border-white data-[state=active]:text-foreground"
+								className="gap-1.5 text-sm font-medium rounded-none text-muted-foreground transition-all duration-200 h-full bg-transparent data-[state=active]:bg-transparent shadow-none data-[state=active]:shadow-none border-b-[1.5px] border-transparent data-[state=active]:border-foreground dark:data-[state=active]:border-white data-[state=active]:text-foreground"
 							>
-								<ImageIcon className="size-4" />
+								<ImageIcon className="size-3.5" />
 								Image
 							</TabsTrigger>
+							<TabsTrigger
+								value="vision"
+								className="gap-1.5 text-sm font-medium rounded-none text-muted-foreground transition-all duration-200 h-full bg-transparent data-[state=active]:bg-transparent shadow-none data-[state=active]:shadow-none border-b-[1.5px] border-transparent data-[state=active]:border-foreground dark:data-[state=active]:border-white data-[state=active]:text-foreground"
+							>
+								<Eye className="size-3.5" />
+								Vision
+							</TabsTrigger>
 						</TabsList>
 					</div>
 
@@ -676,6 +794,174 @@ export function ModelSelector({
 							</CommandList>
 						</Command>
 					</TabsContent>
+
+					{/* ─── Vision Tab ─── */}
+					<TabsContent value="vision" className="mt-0">
+						<Command
+							shouldFilter={false}
+							className="rounded-none rounded-b-lg dark:bg-neutral-900 [&_[data-slot=command-input-wrapper]]:border-0 [&_[data-slot=command-input-wrapper]]:px-0 [&_[data-slot=command-input-wrapper]]:gap-2"
+						>
+							{totalVisionModels > 3 && (
+								<div className="px-2 md:px-3 py-1.5 md:py-2">
+									<CommandInput
+										placeholder="Search vision models"
+										value={visionSearchQuery}
+										onValueChange={setVisionSearchQuery}
+										className="h-7 md:h-8 w-full text-xs md:text-sm border-0 bg-transparent focus:ring-0 placeholder:text-muted-foreground/60"
+									/>
+								</div>
+							)}
+							<CommandList
+								className="max-h-[300px] md:max-h-[400px] overflow-y-auto"
+								onScroll={handleListScroll(setVisionScrollPos)}
+								style={{
+									maskImage: `linear-gradient(to bottom, ${visionScrollPos === "top" ? "black" : "transparent"}, black 16px, black calc(100% - 16px), ${visionScrollPos === "bottom" ? "black" : "transparent"})`,
+									WebkitMaskImage: `linear-gradient(to bottom, ${visionScrollPos === "top" ? "black" : "transparent"}, black 16px, black calc(100% - 16px), ${visionScrollPos === "bottom" ? "black" : "transparent"})`,
+								}}
+							>
+								<CommandEmpty className="py-8 text-center">
+									<div className="flex flex-col items-center gap-2">
+										<Search className="size-8 text-muted-foreground" />
+										<p className="text-sm text-muted-foreground">No vision models found</p>
+										<p className="text-xs text-muted-foreground/60">Try a different search term</p>
+									</div>
+								</CommandEmpty>
+
+								{filteredVisionGlobal.length > 0 && (
+									<CommandGroup>
+										<div className="flex items-center gap-2 px-3 py-2 text-xs font-semibold text-muted-foreground tracking-wider">
+											Global Vision Models
+										</div>
+										{filteredVisionGlobal.map((config) => {
+											const isSelected = currentVisionConfig?.id === config.id;
+											const isAuto = "is_auto_mode" in config && config.is_auto_mode;
+											return (
+												<CommandItem
+													key={`vis-g-${config.id}`}
+													value={`vis-g-${config.id}`}
+													onSelect={() => handleSelectVision(config.id)}
+													className={cn(
+														"mx-2 rounded-lg mb-1 cursor-pointer group transition-all hover:bg-accent/50 dark:hover:bg-white/[0.06]",
+														isSelected && "bg-accent/80 dark:bg-white/[0.06]"
+													)}
+												>
+													<div className="flex items-center gap-3 min-w-0 flex-1">
+														<div className="shrink-0">
+															{getProviderIcon(config.provider, { isAutoMode: isAuto })}
+														</div>
+														<div className="min-w-0 flex-1">
+															<div className="flex items-center gap-2">
+																<span className="font-medium truncate">{config.name}</span>
+																{isAuto && (
+																	<Badge
+																		variant="secondary"
+																		className="text-[9px] px-1 py-0 h-3.5 bg-violet-800 text-white dark:bg-violet-800 dark:text-white border-0"
+																	>
+																		Recommended
+																	</Badge>
+																)}
+																{isSelected && <Check className="size-3.5 text-primary shrink-0" />}
+															</div>
+															<span className="text-xs text-muted-foreground truncate block">
+																{isAuto ? "Auto Mode" : config.model_name}
+															</span>
+														</div>
+														{onEditVision && !isAuto && (
+															<Button
+																variant="ghost"
+																size="icon"
+																className="size-7 shrink-0 rounded-md hover:bg-muted opacity-0 group-hover:opacity-100 transition-opacity"
+																onClick={(e) => {
+																	e.stopPropagation();
+																	setOpen(false);
+																	onEditVision(config as VisionLLMConfig, true);
+																}}
+															>
+																<Edit3 className="size-3.5 text-muted-foreground" />
+															</Button>
+														)}
+													</div>
+												</CommandItem>
+											);
+										})}
+									</CommandGroup>
+								)}
+
+								{filteredVisionUser.length > 0 && (
+									<>
+										{filteredVisionGlobal.length > 0 && (
+											<CommandSeparator className="my-1 mx-4 bg-border/60" />
+										)}
+										<CommandGroup>
+											<div className="flex items-center gap-2 px-3 py-2 text-xs font-semibold text-muted-foreground tracking-wider">
+												Your Vision Models
+											</div>
+											{filteredVisionUser.map((config) => {
+												const isSelected = currentVisionConfig?.id === config.id;
+												return (
+													<CommandItem
+														key={`vis-u-${config.id}`}
+														value={`vis-u-${config.id}`}
+														onSelect={() => handleSelectVision(config.id)}
+														className={cn(
+															"mx-2 rounded-lg mb-1 cursor-pointer group transition-all hover:bg-accent/50 dark:hover:bg-white/[0.06]",
+															isSelected && "bg-accent/80 dark:bg-white/[0.06]"
+														)}
+													>
+														<div className="flex items-center gap-3 min-w-0 flex-1">
+															<div className="shrink-0">{getProviderIcon(config.provider)}</div>
+															<div className="min-w-0 flex-1">
+																<div className="flex items-center gap-2">
+																	<span className="font-medium truncate">{config.name}</span>
+																	{isSelected && (
+																		<Check className="size-3.5 text-primary shrink-0" />
+																	)}
+																</div>
+																<span className="text-xs text-muted-foreground truncate block">
+																	{config.model_name}
+																</span>
+															</div>
+															{onEditVision && (
+																<Button
+																	variant="ghost"
+																	size="icon"
+																	className="h-7 w-7 shrink-0 opacity-0 group-hover:opacity-100 transition-opacity"
+																	onClick={(e) => {
+																		e.stopPropagation();
+																		setOpen(false);
+																		onEditVision(config, false);
+																	}}
+																>
+																	<Edit3 className="size-3.5 text-muted-foreground" />
+																</Button>
+															)}
+														</div>
+													</CommandItem>
+												);
+											})}
+										</CommandGroup>
+									</>
+								)}
+
+								{onAddNewVision && (
+									<div className="p-2 bg-muted/20 dark:bg-neutral-900">
+										<Button
+											variant="ghost"
+											size="sm"
+											className="w-full justify-start gap-2 h-9 rounded-lg hover:bg-accent/50 dark:hover:bg-white/[0.06]"
+											onClick={() => {
+												setOpen(false);
+												onAddNewVision();
+											}}
+										>
+											<Plus className="size-4 text-primary" />
+											<span className="text-sm font-medium">Add Vision Model</span>
+										</Button>
+									</div>
+								)}
+							</CommandList>
+						</Command>
+					</TabsContent>
 				</Tabs>
 			</PopoverContent>
 		</Popover>

surfsense_web/components/settings/llm-role-manager.tsx

@@ -24,6 +24,10 @@ import {
 	llmPreferencesAtom,
 	newLLMConfigsAtom,
 } from "@/atoms/new-llm-config/new-llm-config-query.atoms";
+import {
+	globalVisionLLMConfigsAtom,
+	visionLLMConfigsAtom,
+} from "@/atoms/vision-llm-config/vision-llm-config-query.atoms";
 import { Alert, AlertDescription } from "@/components/ui/alert";
 import { Badge } from "@/components/ui/badge";
 import { Button } from "@/components/ui/button";
@@ -77,8 +81,8 @@ const ROLE_DESCRIPTIONS = {
 		description: "Vision-capable model for screenshot analysis and context extraction",
 		color: "text-amber-600 dark:text-amber-400",
 		bgColor: "bg-amber-500/10",
-		prefKey: "vision_llm_id" as const,
-		configType: "llm" as const,
+		prefKey: "vision_llm_config_id" as const,
+		configType: "vision" as const,
 	},
 };
 
@@ -112,6 +116,18 @@ export function LLMRoleManager({ searchSpaceId }: LLMRoleManagerProps) {
 		error: globalImageConfigsError,
 	} = useAtomValue(globalImageGenConfigsAtom);
 
+	// Vision LLM configs
+	const {
+		data: userVisionConfigs = [],
+		isFetching: visionConfigsLoading,
+		error: visionConfigsError,
+	} = useAtomValue(visionLLMConfigsAtom);
+	const {
+		data: globalVisionConfigs = [],
+		isFetching: globalVisionConfigsLoading,
+		error: globalVisionConfigsError,
+	} = useAtomValue(globalVisionLLMConfigsAtom);
+
 	// Preferences
 	const {
 		data: preferences = {},
@@ -125,7 +141,7 @@ export function LLMRoleManager({ searchSpaceId }: LLMRoleManagerProps) {
 		agent_llm_id: preferences.agent_llm_id ?? "",
 		document_summary_llm_id: preferences.document_summary_llm_id ?? "",
 		image_generation_config_id: preferences.image_generation_config_id ?? "",
-		vision_llm_id: preferences.vision_llm_id ?? "",
+		vision_llm_config_id: preferences.vision_llm_config_id ?? "",
 	}));
 
 	const [savingRole, setSavingRole] = useState<string | null>(null);
@@ -137,14 +153,14 @@ export function LLMRoleManager({ searchSpaceId }: LLMRoleManagerProps) {
 				agent_llm_id: preferences.agent_llm_id ?? "",
 				document_summary_llm_id: preferences.document_summary_llm_id ?? "",
 				image_generation_config_id: preferences.image_generation_config_id ?? "",
-				vision_llm_id: preferences.vision_llm_id ?? "",
+				vision_llm_config_id: preferences.vision_llm_config_id ?? "",
 			});
 		}
 	}, [
 		preferences?.agent_llm_id,
 		preferences?.document_summary_llm_id,
 		preferences?.image_generation_config_id,
-		preferences?.vision_llm_id,
+		preferences?.vision_llm_config_id,
 	]);
 
 	const handleRoleAssignment = useCallback(
@@ -181,6 +197,14 @@ export function LLMRoleManager({ searchSpaceId }: LLMRoleManagerProps) {
 		...(userImageConfigs ?? []).filter((config) => config.id && config.id.toString().trim() !== ""),
 	];
 
+	// Combine global and custom vision LLM configs
+	const allVisionConfigs = [
+		...globalVisionConfigs.map((config) => ({ ...config, is_global: true })),
+		...(userVisionConfigs ?? []).filter(
+			(config) => config.id && config.id.toString().trim() !== ""
+		),
+	];
+
 	const isAssignmentComplete =
 		allLLMConfigs.some((c) => c.id === assignments.agent_llm_id) &&
 		allLLMConfigs.some((c) => c.id === assignments.document_summary_llm_id) &&
@@ -191,13 +215,17 @@ export function LLMRoleManager({ searchSpaceId }: LLMRoleManagerProps) {
 		preferencesLoading ||
 		globalConfigsLoading ||
 		imageConfigsLoading ||
-		globalImageConfigsLoading;
+		globalImageConfigsLoading ||
+		visionConfigsLoading ||
+		globalVisionConfigsLoading;
 	const hasError =
 		configsError ||
 		preferencesError ||
 		globalConfigsError ||
 		imageConfigsError ||
-		globalImageConfigsError;
+		globalImageConfigsError ||
+		visionConfigsError ||
+		globalVisionConfigsError;
 	const hasAnyConfigs = allLLMConfigs.length > 0 || allImageConfigs.length > 0;
 
 	return (
@@ -291,15 +319,27 @@ export function LLMRoleManager({ searchSpaceId }: LLMRoleManagerProps) {
 				<div className="grid gap-4 grid-cols-1 lg:grid-cols-2">
 					{Object.entries(ROLE_DESCRIPTIONS).map(([key, role]) => {
 						const IconComponent = role.icon;
-						const isImageRole = role.configType === "image";
 						const currentAssignment = assignments[role.prefKey as keyof typeof assignments];
 
 						// Pick the right config lists based on role type
-						const roleGlobalConfigs = isImageRole ? globalImageConfigs : globalConfigs;
-						const roleUserConfigs = isImageRole
-							? (userImageConfigs ?? []).filter((c) => c.id && c.id.toString().trim() !== "")
-							: newLLMConfigs.filter((c) => c.id && c.id.toString().trim() !== "");
-						const roleAllConfigs = isImageRole ? allImageConfigs : allLLMConfigs;
+						const roleGlobalConfigs =
+							role.configType === "image"
+								? globalImageConfigs
+								: role.configType === "vision"
+									? globalVisionConfigs
+									: globalConfigs;
+						const roleUserConfigs =
+							role.configType === "image"
+								? (userImageConfigs ?? []).filter((c) => c.id && c.id.toString().trim() !== "")
+								: role.configType === "vision"
+									? (userVisionConfigs ?? []).filter((c) => c.id && c.id.toString().trim() !== "")
+									: newLLMConfigs.filter((c) => c.id && c.id.toString().trim() !== "");
+						const roleAllConfigs =
+							role.configType === "image"
+								? allImageConfigs
+								: role.configType === "vision"
+									? allVisionConfigs
+									: allLLMConfigs;
 
 						const assignedConfig = roleAllConfigs.find((config) => config.id === currentAssignment);
 						const isAssigned = !!assignedConfig;

surfsense_web/components/settings/search-space-settings-dialog.tsx

@@ -1,7 +1,7 @@
 "use client";
 
 import { useAtom } from "jotai";
-import { Bot, Brain, FileText, Globe, ImageIcon, MessageSquare, Shield } from "lucide-react";
+import { Bot, Brain, Eye, FileText, Globe, ImageIcon, MessageSquare, Shield } from "lucide-react";
 import { useTranslations } from "next-intl";
 import type React from "react";
 import { searchSpaceSettingsDialogAtom } from "@/atoms/settings/settings-dialog.atoms";
@@ -13,6 +13,7 @@ import { ModelConfigManager } from "@/components/settings/model-config-manager";
 import { PromptConfigManager } from "@/components/settings/prompt-config-manager";
 import { RolesManager } from "@/components/settings/roles-manager";
 import { SettingsDialog } from "@/components/settings/settings-dialog";
+import { VisionModelManager } from "@/components/settings/vision-model-manager";
 
 interface SearchSpaceSettingsDialogProps {
 	searchSpaceId: number;
@@ -31,6 +32,11 @@ export function SearchSpaceSettingsDialog({ searchSpaceId }: SearchSpaceSettings
 			label: t("nav_image_models"),
 			icon: <ImageIcon className="h-4 w-4" />,
 		},
+		{
+			value: "vision-models",
+			label: t("nav_vision_models"),
+			icon: <Eye className="h-4 w-4" />,
+		},
 		{ value: "team-roles", label: t("nav_team_roles"), icon: <Shield className="h-4 w-4" /> },
 		{
 			value: "prompts",
@@ -45,6 +51,7 @@ export function SearchSpaceSettingsDialog({ searchSpaceId }: SearchSpaceSettings
 		models: <ModelConfigManager searchSpaceId={searchSpaceId} />,
 		roles: <LLMRoleManager searchSpaceId={searchSpaceId} />,
 		"image-models": <ImageModelManager searchSpaceId={searchSpaceId} />,
+		"vision-models": <VisionModelManager searchSpaceId={searchSpaceId} />,
 		"team-roles": <RolesManager searchSpaceId={searchSpaceId} />,
 		prompts: <PromptConfigManager searchSpaceId={searchSpaceId} />,
 		"public-links": <PublicChatSnapshotsManager searchSpaceId={searchSpaceId} />,

surfsense_web/components/settings/vision-model-manager.tsx

@@ -0,0 +1,401 @@
+"use client";
+
+import { useAtomValue } from "jotai";
+import { AlertCircle, Dot, Edit3, Info, RefreshCw, Trash2 } from "lucide-react";
+import { useMemo, useState } from "react";
+import { membersAtom, myAccessAtom } from "@/atoms/members/members-query.atoms";
+import { deleteVisionLLMConfigMutationAtom } from "@/atoms/vision-llm-config/vision-llm-config-mutation.atoms";
+import {
+	globalVisionLLMConfigsAtom,
+	visionLLMConfigsAtom,
+} from "@/atoms/vision-llm-config/vision-llm-config-query.atoms";
+import { VisionConfigDialog } from "@/components/shared/vision-config-dialog";
+import { Alert, AlertDescription } from "@/components/ui/alert";
+import {
+	AlertDialog,
+	AlertDialogAction,
+	AlertDialogCancel,
+	AlertDialogContent,
+	AlertDialogDescription,
+	AlertDialogFooter,
+	AlertDialogHeader,
+	AlertDialogTitle,
+} from "@/components/ui/alert-dialog";
+import { Avatar, AvatarFallback, AvatarImage } from "@/components/ui/avatar";
+import { Button } from "@/components/ui/button";
+import { Card, CardContent } from "@/components/ui/card";
+import { Skeleton } from "@/components/ui/skeleton";
+import { Spinner } from "@/components/ui/spinner";
+import { Tooltip, TooltipContent, TooltipProvider, TooltipTrigger } from "@/components/ui/tooltip";
+import type { VisionLLMConfig } from "@/contracts/types/new-llm-config.types";
+import { useMediaQuery } from "@/hooks/use-media-query";
+import { getProviderIcon } from "@/lib/provider-icons";
+import { cn } from "@/lib/utils";
+
+interface VisionModelManagerProps {
+	searchSpaceId: number;
+}
+
+function getInitials(name: string): string {
+	const parts = name.trim().split(/\s+/);
+	if (parts.length >= 2) {
+		return (parts[0][0] + parts[1][0]).toUpperCase();
+	}
+	return name.slice(0, 2).toUpperCase();
+}
+
+export function VisionModelManager({ searchSpaceId }: VisionModelManagerProps) {
+	const isDesktop = useMediaQuery("(min-width: 768px)");
+
+	const {
+		mutateAsync: deleteConfig,
+		isPending: isDeleting,
+		error: deleteError,
+	} = useAtomValue(deleteVisionLLMConfigMutationAtom);
+
+	const {
+		data: userConfigs,
+		isFetching: configsLoading,
+		error: fetchError,
+		refetch: refreshConfigs,
+	} = useAtomValue(visionLLMConfigsAtom);
+	const { data: globalConfigs = [], isFetching: globalLoading } = useAtomValue(
+		globalVisionLLMConfigsAtom
+	);
+
+	const { data: members } = useAtomValue(membersAtom);
+	const memberMap = useMemo(() => {
+		const map = new Map<string, { name: string; email?: string; avatarUrl?: string }>();
+		if (members) {
+			for (const m of members) {
+				map.set(m.user_id, {
+					name: m.user_display_name || m.user_email || "Unknown",
+					email: m.user_email || undefined,
+					avatarUrl: m.user_avatar_url || undefined,
+				});
+			}
+		}
+		return map;
+	}, [members]);
+
+	const { data: access } = useAtomValue(myAccessAtom);
+	const canCreate = useMemo(() => {
+		if (!access) return false;
+		if (access.is_owner) return true;
+		return access.permissions?.includes("vision_configs:create") ?? false;
+	}, [access]);
+	const canDelete = useMemo(() => {
+		if (!access) return false;
+		if (access.is_owner) return true;
+		return access.permissions?.includes("vision_configs:delete") ?? false;
+	}, [access]);
+	const canUpdate = canCreate;
+	const isReadOnly = !canCreate && !canDelete;
+
+	const [isDialogOpen, setIsDialogOpen] = useState(false);
+	const [editingConfig, setEditingConfig] = useState<VisionLLMConfig | null>(null);
+	const [configToDelete, setConfigToDelete] = useState<VisionLLMConfig | null>(null);
+
+	const isLoading = configsLoading || globalLoading;
+	const errors = [deleteError, fetchError].filter(Boolean) as Error[];
+
+	const openEditDialog = (config: VisionLLMConfig) => {
+		setEditingConfig(config);
+		setIsDialogOpen(true);
+	};
+
+	const openNewDialog = () => {
+		setEditingConfig(null);
+		setIsDialogOpen(true);
+	};
+
+	const handleDelete = async () => {
+		if (!configToDelete) return;
+		try {
+			await deleteConfig({ id: configToDelete.id, name: configToDelete.name });
+			setConfigToDelete(null);
+		} catch {
+			// Error handled by mutation
+		}
+	};
+
+	return (
+		<div className="space-y-4 md:space-y-6">
+			<div className="flex flex-col space-y-4 sm:flex-row sm:items-center sm:justify-between sm:space-y-0">
+				<Button
+					variant="secondary"
+					size="sm"
+					onClick={() => refreshConfigs()}
+					disabled={isLoading}
+					className="gap-2"
+				>
+					<RefreshCw className={cn("h-3.5 w-3.5", configsLoading && "animate-spin")} />
+					Refresh
+				</Button>
+				{canCreate && (
+					<Button
+						variant="outline"
+						onClick={openNewDialog}
+						className="gap-2 bg-white text-black hover:bg-neutral-100 dark:bg-white dark:text-black dark:hover:bg-neutral-200"
+					>
+						Add Vision Model
+					</Button>
+				)}
+			</div>
+
+			{errors.map((err) => (
+				<div key={err?.message}>
+					<Alert variant="destructive" className="py-3">
+						<AlertCircle className="h-3 w-3 md:h-4 md:w-4 shrink-0" />
+						<AlertDescription className="text-xs md:text-sm">{err?.message}</AlertDescription>
+					</Alert>
+				</div>
+			))}
+
+			{access && !isLoading && isReadOnly && (
+				<div>
+					<Alert className="bg-muted/50 py-3 md:py-4">
+						<Info className="h-3 w-3 md:h-4 md:w-4 shrink-0" />
+						<AlertDescription className="text-xs md:text-sm">
+							You have <span className="font-medium">read-only</span> access to vision model
+							configurations. Contact a space owner to request additional permissions.
+						</AlertDescription>
+					</Alert>
+				</div>
+			)}
+			{access && !isLoading && !isReadOnly && (!canCreate || !canDelete) && (
+				<div>
+					<Alert className="bg-muted/50 py-3 md:py-4">
+						<Info className="h-3 w-3 md:h-4 md:w-4 shrink-0" />
+						<AlertDescription className="text-xs md:text-sm">
+							You can{" "}
+							{[canCreate && "create and edit", canDelete && "delete"]
+								.filter(Boolean)
+								.join(" and ")}{" "}
+							vision model configurations
+							{!canDelete && ", but cannot delete them"}.
+						</AlertDescription>
+					</Alert>
+				</div>
+			)}
+
+			{globalConfigs.filter((g) => !("is_auto_mode" in g && g.is_auto_mode)).length > 0 && (
+				<Alert className="bg-muted/50 py-3">
+					<Info className="h-3 w-3 md:h-4 md:w-4 shrink-0" />
+					<AlertDescription className="text-xs md:text-sm">
+						<p>
+							<span className="font-medium">
+								{globalConfigs.filter((g) => !("is_auto_mode" in g && g.is_auto_mode)).length}{" "}
+								global vision{" "}
+								{globalConfigs.filter((g) => !("is_auto_mode" in g && g.is_auto_mode)).length === 1
+									? "model"
+									: "models"}
+							</span>{" "}
+							available from your administrator. Use the model selector to view and select them.
+						</p>
+					</AlertDescription>
+				</Alert>
+			)}
+
+			{isLoading && (
+				<div className="space-y-4 md:space-y-6">
+					<div className="space-y-4">
+						<div className="flex items-center justify-between">
+							<Skeleton className="h-6 md:h-7 w-40 md:w-48" />
+							<Skeleton className="h-8 md:h-9 w-32 md:w-36 rounded-md" />
+						</div>
+						<div className="grid gap-3 grid-cols-1 sm:grid-cols-2 xl:grid-cols-3">
+							{["skeleton-a", "skeleton-b", "skeleton-c"].map((key) => (
+								<Card key={key} className="border-border/60">
+									<CardContent className="p-4 flex flex-col gap-3">
+										<div className="flex items-start justify-between gap-2">
+											<div className="space-y-1.5 flex-1 min-w-0">
+												<Skeleton className="h-4 w-28 md:w-32" />
+												<Skeleton className="h-3 w-40 md:w-48" />
+											</div>
+										</div>
+										<div className="flex items-center gap-2">
+											<Skeleton className="h-5 w-16 rounded-full" />
+											<Skeleton className="h-5 w-24 rounded-md" />
+										</div>
+										<div className="flex items-center gap-2 pt-2 border-t border-border/40">
+											<Skeleton className="h-3 w-20" />
+											<Skeleton className="h-4 w-4 rounded-full" />
+											<Skeleton className="h-3 w-16" />
+										</div>
+									</CardContent>
+								</Card>
+							))}
+						</div>
+					</div>
+				</div>
+			)}
+
+			{!isLoading && (
+				<div className="space-y-4 md:space-y-6">
+					{(userConfigs?.length ?? 0) === 0 ? (
+						<Card className="border-0 bg-transparent shadow-none">
+							<CardContent className="flex flex-col items-center justify-center py-10 md:py-16 text-center">
+								<h3 className="text-sm md:text-base font-semibold mb-2">No Vision Models Yet</h3>
+								<p className="text-[11px] md:text-xs text-muted-foreground max-w-sm mb-4">
+									{canCreate
+										? "Add your own vision-capable model (GPT-4o, Claude, Gemini, etc.)"
+										: "No vision models have been added to this space yet. Contact a space owner to add one."}
+								</p>
+							</CardContent>
+						</Card>
+					) : (
+						<div className="grid gap-3 grid-cols-1 sm:grid-cols-2 xl:grid-cols-3">
+							{userConfigs?.map((config) => {
+								const member = config.user_id ? memberMap.get(config.user_id) : null;
+
+								return (
+									<div key={config.id}>
+										<Card className="group relative overflow-hidden transition-all duration-200 border-border/60 hover:shadow-md h-full">
+											<CardContent className="p-4 flex flex-col gap-3 h-full">
+												<div className="flex items-start justify-between gap-2">
+													<div className="min-w-0 flex-1">
+														<h4 className="text-sm font-semibold tracking-tight truncate">
+															{config.name}
+														</h4>
+														{config.description && (
+															<p className="text-[11px] text-muted-foreground/70 truncate mt-0.5">
+																{config.description}
+															</p>
+														)}
+													</div>
+													{(canUpdate || canDelete) && (
+														<div className="flex items-center gap-0.5 shrink-0 sm:opacity-0 sm:group-hover:opacity-100 transition-opacity duration-150">
+															{canUpdate && (
+																<TooltipProvider>
+																	<Tooltip open={isDesktop ? undefined : false}>
+																		<TooltipTrigger asChild>
+																			<Button
+																				variant="ghost"
+																				size="icon"
+																				onClick={() => openEditDialog(config)}
+																				className="h-7 w-7 text-muted-foreground hover:text-foreground"
+																			>
+																				<Edit3 className="h-3 w-3" />
+																			</Button>
+																		</TooltipTrigger>
+																		<TooltipContent>Edit</TooltipContent>
+																	</Tooltip>
+																</TooltipProvider>
+															)}
+															{canDelete && (
+																<TooltipProvider>
+																	<Tooltip open={isDesktop ? undefined : false}>
+																		<TooltipTrigger asChild>
+																			<Button
+																				variant="ghost"
+																				size="icon"
+																				onClick={() => setConfigToDelete(config)}
+																				className="h-7 w-7 text-muted-foreground hover:text-destructive"
+																			>
+																				<Trash2 className="h-3 w-3" />
+																			</Button>
+																		</TooltipTrigger>
+																		<TooltipContent>Delete</TooltipContent>
+																	</Tooltip>
+																</TooltipProvider>
+															)}
+														</div>
+													)}
+												</div>
+
+												<div className="flex items-center gap-2 flex-wrap">
+													{getProviderIcon(config.provider, {
+														className: "size-3.5 shrink-0",
+													})}
+													<code className="text-[11px] font-mono text-muted-foreground bg-muted/60 px-2 py-0.5 rounded-md truncate max-w-[160px]">
+														{config.model_name}
+													</code>
+												</div>
+
+												<div className="flex items-center gap-2 pt-2 border-t border-border/40 mt-auto">
+													<span className="text-[11px] text-muted-foreground/60">
+														{new Date(config.created_at).toLocaleDateString(undefined, {
+															year: "numeric",
+															month: "short",
+															day: "numeric",
+														})}
+													</span>
+													{member && (
+														<>
+															<Dot className="h-4 w-4 text-muted-foreground/30" />
+															<TooltipProvider>
+																<Tooltip open={isDesktop ? undefined : false}>
+																	<TooltipTrigger asChild>
+																		<div className="flex items-center gap-1.5 cursor-default">
+																			<Avatar className="size-4.5 shrink-0">
+																				{member.avatarUrl && (
+																					<AvatarImage src={member.avatarUrl} alt={member.name} />
+																				)}
+																				<AvatarFallback className="text-[9px]">
+																					{getInitials(member.name)}
+																				</AvatarFallback>
+																			</Avatar>
+																			<span className="text-[11px] text-muted-foreground/60 truncate max-w-[120px]">
+																				{member.name}
+																			</span>
+																		</div>
+																	</TooltipTrigger>
+																	<TooltipContent side="bottom">
+																		{member.email || member.name}
+																	</TooltipContent>
+																</Tooltip>
+															</TooltipProvider>
+														</>
+													)}
+												</div>
+											</CardContent>
+										</Card>
+									</div>
+								);
+							})}
+						</div>
+					)}
+				</div>
+			)}
+
+			<VisionConfigDialog
+				open={isDialogOpen}
+				onOpenChange={(open) => {
+					setIsDialogOpen(open);
+					if (!open) setEditingConfig(null);
+				}}
+				config={editingConfig}
+				isGlobal={false}
+				searchSpaceId={searchSpaceId}
+				mode={editingConfig ? "edit" : "create"}
+			/>
+
+			<AlertDialog
+				open={!!configToDelete}
+				onOpenChange={(open) => !open && setConfigToDelete(null)}
+			>
+				<AlertDialogContent className="select-none">
+					<AlertDialogHeader>
+						<AlertDialogTitle>Delete Vision Model</AlertDialogTitle>
+						<AlertDialogDescription>
+							Are you sure you want to delete{" "}
+							<span className="font-semibold text-foreground">{configToDelete?.name}</span>?
+						</AlertDialogDescription>
+					</AlertDialogHeader>
+					<AlertDialogFooter>
+						<AlertDialogCancel disabled={isDeleting}>Cancel</AlertDialogCancel>
+						<AlertDialogAction
+							onClick={handleDelete}
+							disabled={isDeleting}
+							className="relative bg-destructive text-destructive-foreground hover:bg-destructive/90"
+						>
+							<span className={isDeleting ? "opacity-0" : ""}>Delete</span>
+							{isDeleting && <Spinner size="sm" className="absolute" />}
+						</AlertDialogAction>
+					</AlertDialogFooter>
+				</AlertDialogContent>
+			</AlertDialog>
+		</div>
+	);
+}

surfsense_web/components/shared/vision-config-dialog.tsx

@@ -0,0 +1,381 @@
+"use client";
+
+import { useAtomValue } from "jotai";
+import { AlertCircle } from "lucide-react";
+import { useCallback, useEffect, useRef, useState } from "react";
+import { toast } from "sonner";
+import { updateLLMPreferencesMutationAtom } from "@/atoms/new-llm-config/new-llm-config-mutation.atoms";
+import {
+	createVisionLLMConfigMutationAtom,
+	updateVisionLLMConfigMutationAtom,
+} from "@/atoms/vision-llm-config/vision-llm-config-mutation.atoms";
+import { Alert, AlertDescription } from "@/components/ui/alert";
+import { Badge } from "@/components/ui/badge";
+import { Button } from "@/components/ui/button";
+import { Dialog, DialogContent, DialogTitle } from "@/components/ui/dialog";
+import { Input } from "@/components/ui/input";
+import { Label } from "@/components/ui/label";
+import {
+	Select,
+	SelectContent,
+	SelectItem,
+	SelectTrigger,
+	SelectValue,
+} from "@/components/ui/select";
+import { Separator } from "@/components/ui/separator";
+import { Spinner } from "@/components/ui/spinner";
+import { VISION_PROVIDERS } from "@/contracts/enums/vision-providers";
+import type {
+	GlobalVisionLLMConfig,
+	VisionLLMConfig,
+	VisionProvider,
+} from "@/contracts/types/new-llm-config.types";
+
+interface VisionConfigDialogProps {
+	open: boolean;
+	onOpenChange: (open: boolean) => void;
+	config: VisionLLMConfig | GlobalVisionLLMConfig | null;
+	isGlobal: boolean;
+	searchSpaceId: number;
+	mode: "create" | "edit" | "view";
+}
+
+const INITIAL_FORM = {
+	name: "",
+	description: "",
+	provider: "",
+	model_name: "",
+	api_key: "",
+	api_base: "",
+	api_version: "",
+};
+
+export function VisionConfigDialog({
+	open,
+	onOpenChange,
+	config,
+	isGlobal,
+	searchSpaceId,
+	mode,
+}: VisionConfigDialogProps) {
+	const [isSubmitting, setIsSubmitting] = useState(false);
+	const [formData, setFormData] = useState(INITIAL_FORM);
+	const [scrollPos, setScrollPos] = useState<"top" | "middle" | "bottom">("top");
+	const scrollRef = useRef<HTMLDivElement>(null);
+
+	useEffect(() => {
+		if (open) {
+			if (mode === "edit" && config && !isGlobal) {
+				setFormData({
+					name: config.name || "",
+					description: config.description || "",
+					provider: config.provider || "",
+					model_name: config.model_name || "",
+					api_key: (config as VisionLLMConfig).api_key || "",
+					api_base: config.api_base || "",
+					api_version: (config as VisionLLMConfig).api_version || "",
+				});
+			} else if (mode === "create") {
+				setFormData(INITIAL_FORM);
+			}
+			setScrollPos("top");
+		}
+	}, [open, mode, config, isGlobal]);
+
+	const { mutateAsync: createConfig } = useAtomValue(createVisionLLMConfigMutationAtom);
+	const { mutateAsync: updateConfig } = useAtomValue(updateVisionLLMConfigMutationAtom);
+	const { mutateAsync: updatePreferences } = useAtomValue(updateLLMPreferencesMutationAtom);
+
+	const handleScroll = useCallback((e: React.UIEvent<HTMLDivElement>) => {
+		const el = e.currentTarget;
+		const atTop = el.scrollTop <= 2;
+		const atBottom = el.scrollHeight - el.scrollTop - el.clientHeight <= 2;
+		setScrollPos(atTop ? "top" : atBottom ? "bottom" : "middle");
+	}, []);
+
+	const getTitle = () => {
+		if (mode === "create") return "Add Vision Model";
+		if (isGlobal) return "View Global Vision Model";
+		return "Edit Vision Model";
+	};
+
+	const getSubtitle = () => {
+		if (mode === "create") return "Set up a new vision-capable LLM provider";
+		if (isGlobal) return "Read-only global configuration";
+		return "Update your vision model settings";
+	};
+
+	const handleSubmit = useCallback(async () => {
+		setIsSubmitting(true);
+		try {
+			if (mode === "create") {
+				const result = await createConfig({
+					name: formData.name,
+					provider: formData.provider as VisionProvider,
+					model_name: formData.model_name,
+					api_key: formData.api_key,
+					api_base: formData.api_base || undefined,
+					api_version: formData.api_version || undefined,
+					description: formData.description || undefined,
+					search_space_id: searchSpaceId,
+				});
+				if (result?.id) {
+					await updatePreferences({
+						search_space_id: searchSpaceId,
+						data: { vision_llm_config_id: result.id },
+					});
+				}
+				onOpenChange(false);
+			} else if (!isGlobal && config) {
+				await updateConfig({
+					id: config.id,
+					data: {
+						name: formData.name,
+						description: formData.description || undefined,
+						provider: formData.provider as VisionProvider,
+						model_name: formData.model_name,
+						api_key: formData.api_key,
+						api_base: formData.api_base || undefined,
+						api_version: formData.api_version || undefined,
+					},
+				});
+				onOpenChange(false);
+			}
+		} catch (error) {
+			console.error("Failed to save vision config:", error);
+			toast.error("Failed to save vision model");
+		} finally {
+			setIsSubmitting(false);
+		}
+	}, [
+		mode,
+		isGlobal,
+		config,
+		formData,
+		searchSpaceId,
+		createConfig,
+		updateConfig,
+		updatePreferences,
+		onOpenChange,
+	]);
+
+	const handleUseGlobalConfig = useCallback(async () => {
+		if (!config || !isGlobal) return;
+		setIsSubmitting(true);
+		try {
+			await updatePreferences({
+				search_space_id: searchSpaceId,
+				data: { vision_llm_config_id: config.id },
+			});
+			toast.success(`Now using ${config.name}`);
+			onOpenChange(false);
+		} catch (error) {
+			console.error("Failed to set vision model:", error);
+			toast.error("Failed to set vision model");
+		} finally {
+			setIsSubmitting(false);
+		}
+	}, [config, isGlobal, searchSpaceId, updatePreferences, onOpenChange]);
+
+	const isFormValid = formData.name && formData.provider && formData.model_name && formData.api_key;
+	const selectedProvider = VISION_PROVIDERS.find((p) => p.value === formData.provider);
+
+	return (
+		<Dialog open={open} onOpenChange={onOpenChange}>
+			<DialogContent
+				className="max-w-lg h-[85vh] flex flex-col p-0 gap-0 overflow-hidden"
+				onOpenAutoFocus={(e) => e.preventDefault()}
+			>
+				<DialogTitle className="sr-only">{getTitle()}</DialogTitle>
+
+				<div className="flex items-start justify-between px-6 pt-6 pb-4 pr-14">
+					<div className="space-y-1">
+						<div className="flex items-center gap-2">
+							<h2 className="text-lg font-semibold tracking-tight">{getTitle()}</h2>
+							{isGlobal && mode !== "create" && (
+								<Badge variant="secondary" className="text-[10px]">
+									Global
+								</Badge>
+							)}
+						</div>
+						<p className="text-sm text-muted-foreground">{getSubtitle()}</p>
+						{config && mode !== "create" && (
+							<p className="text-xs font-mono text-muted-foreground/70">{config.model_name}</p>
+						)}
+					</div>
+				</div>
+
+				<div
+					ref={scrollRef}
+					onScroll={handleScroll}
+					className="flex-1 overflow-y-auto px-6 py-5"
+					style={{
+						maskImage: `linear-gradient(to bottom, ${scrollPos === "top" ? "black" : "transparent"}, black 16px, black calc(100% - 16px), ${scrollPos === "bottom" ? "black" : "transparent"})`,
+						WebkitMaskImage: `linear-gradient(to bottom, ${scrollPos === "top" ? "black" : "transparent"}, black 16px, black calc(100% - 16px), ${scrollPos === "bottom" ? "black" : "transparent"})`,
+					}}
+				>
+					{isGlobal && config && (
+						<>
+							<Alert className="mb-5 border-amber-500/30 bg-amber-500/5">
+								<AlertCircle className="size-4 text-amber-500" />
+								<AlertDescription className="text-sm text-amber-700 dark:text-amber-400">
+									Global configurations are read-only. To customize, create a new model.
+								</AlertDescription>
+							</Alert>
+							<div className="space-y-4">
+								<div className="grid gap-4 sm:grid-cols-2">
+									<div className="space-y-1.5">
+										<div className="text-xs font-medium text-muted-foreground uppercase tracking-wider">
+											Name
+										</div>
+										<p className="text-sm font-medium">{config.name}</p>
+									</div>
+									{config.description && (
+										<div className="space-y-1.5">
+											<div className="text-xs font-medium text-muted-foreground uppercase tracking-wider">
+												Description
+											</div>
+											<p className="text-sm text-muted-foreground">{config.description}</p>
+										</div>
+									)}
+								</div>
+								<Separator />
+								<div className="grid gap-4 sm:grid-cols-2">
+									<div className="space-y-1.5">
+										<div className="text-xs font-medium text-muted-foreground uppercase tracking-wider">
+											Provider
+										</div>
+										<p className="text-sm font-medium">{config.provider}</p>
+									</div>
+									<div className="space-y-1.5">
+										<div className="text-xs font-medium text-muted-foreground uppercase tracking-wider">
+											Model
+										</div>
+										<p className="text-sm font-medium font-mono">{config.model_name}</p>
+									</div>
+								</div>
+							</div>
+						</>
+					)}
+
+					{(mode === "create" || (mode === "edit" && !isGlobal)) && (
+						<div className="space-y-4">
+							<div className="space-y-2">
+								<Label className="text-sm font-medium">Name *</Label>
+								<Input
+									placeholder="e.g., My GPT-4o Vision"
+									value={formData.name}
+									onChange={(e) => setFormData((p) => ({ ...p, name: e.target.value }))}
+								/>
+							</div>
+
+							<div className="space-y-2">
+								<Label className="text-sm font-medium">Description</Label>
+								<Input
+									placeholder="Optional description"
+									value={formData.description}
+									onChange={(e) => setFormData((p) => ({ ...p, description: e.target.value }))}
+								/>
+							</div>
+
+							<Separator />
+
+							<div className="space-y-2">
+								<Label className="text-sm font-medium">Provider *</Label>
+								<Select
+									value={formData.provider}
+									onValueChange={(val) =>
+										setFormData((p) => ({ ...p, provider: val, model_name: "" }))
+									}
+								>
+									<SelectTrigger>
+										<SelectValue placeholder="Select a provider" />
+									</SelectTrigger>
+									<SelectContent>
+										{VISION_PROVIDERS.map((p) => (
+											<SelectItem key={p.value} value={p.value} description={p.example}>
+												{p.label}
+											</SelectItem>
+										))}
+									</SelectContent>
+								</Select>
+							</div>
+
+							<div className="space-y-2">
+								<Label className="text-sm font-medium">Model Name *</Label>
+								<Input
+									placeholder={selectedProvider?.example?.split(",")[0]?.trim() || "e.g., gpt-4o"}
+									value={formData.model_name}
+									onChange={(e) => setFormData((p) => ({ ...p, model_name: e.target.value }))}
+								/>
+							</div>
+
+							<div className="space-y-2">
+								<Label className="text-sm font-medium">API Key *</Label>
+								<Input
+									type="password"
+									placeholder="sk-..."
+									value={formData.api_key}
+									onChange={(e) => setFormData((p) => ({ ...p, api_key: e.target.value }))}
+								/>
+							</div>
+
+							<div className="space-y-2">
+								<Label className="text-sm font-medium">API Base URL</Label>
+								<Input
+									placeholder={selectedProvider?.apiBase || "Optional"}
+									value={formData.api_base}
+									onChange={(e) => setFormData((p) => ({ ...p, api_base: e.target.value }))}
+								/>
+							</div>
+
+							{formData.provider === "AZURE_OPENAI" && (
+								<div className="space-y-2">
+									<Label className="text-sm font-medium">API Version (Azure)</Label>
+									<Input
+										placeholder="2024-02-15-preview"
+										value={formData.api_version}
+										onChange={(e) => setFormData((p) => ({ ...p, api_version: e.target.value }))}
+									/>
+								</div>
+							)}
+						</div>
+					)}
+				</div>
+
+				<div className="shrink-0 px-6 py-4 flex items-center justify-end gap-3">
+					<Button
+						type="button"
+						variant="secondary"
+						onClick={() => onOpenChange(false)}
+						disabled={isSubmitting}
+						className="text-sm h-9"
+					>
+						Cancel
+					</Button>
+					{mode === "create" || (mode === "edit" && !isGlobal) ? (
+						<Button
+							onClick={handleSubmit}
+							disabled={isSubmitting || !isFormValid}
+							className="relative text-sm h-9 min-w-[120px]"
+						>
+							<span className={isSubmitting ? "opacity-0" : ""}>
+								{mode === "edit" ? "Save Changes" : "Add Model"}
+							</span>
+							{isSubmitting && <Spinner size="sm" className="absolute" />}
+						</Button>
+					) : isGlobal && config ? (
+						<Button
+							className="relative text-sm h-9"
+							onClick={handleUseGlobalConfig}
+							disabled={isSubmitting}
+						>
+							<span className={isSubmitting ? "opacity-0" : ""}>Use This Model</span>
+							{isSubmitting && <Spinner size="sm" className="absolute" />}
+						</Button>
+					) : null}
+				</div>
+			</DialogContent>
+		</Dialog>
+	);
+}

surfsense_web/contracts/enums/vision-providers.ts

@@ -0,0 +1,102 @@
+export interface VisionProviderInfo {
+	value: string;
+	label: string;
+	example: string;
+	description: string;
+	apiBase?: string;
+}
+
+export const VISION_PROVIDERS: VisionProviderInfo[] = [
+	{
+		value: "OPENAI",
+		label: "OpenAI",
+		example: "gpt-4o, gpt-4o-mini",
+		description: "GPT-4o vision models",
+	},
+	{
+		value: "ANTHROPIC",
+		label: "Anthropic",
+		example: "claude-sonnet-4-20250514",
+		description: "Claude vision models",
+	},
+	{
+		value: "GOOGLE",
+		label: "Google AI Studio",
+		example: "gemini-2.5-flash, gemini-2.0-flash",
+		description: "Gemini vision models",
+	},
+	{
+		value: "AZURE_OPENAI",
+		label: "Azure OpenAI",
+		example: "azure/gpt-4o",
+		description: "OpenAI vision models on Azure",
+	},
+	{
+		value: "VERTEX_AI",
+		label: "Google Vertex AI",
+		example: "vertex_ai/gemini-2.5-flash",
+		description: "Gemini vision models on Vertex AI",
+	},
+	{
+		value: "BEDROCK",
+		label: "AWS Bedrock",
+		example: "bedrock/anthropic.claude-sonnet-4-20250514-v1:0",
+		description: "Vision models on AWS Bedrock",
+	},
+	{
+		value: "XAI",
+		label: "xAI",
+		example: "grok-2-vision",
+		description: "Grok vision models",
+	},
+	{
+		value: "OPENROUTER",
+		label: "OpenRouter",
+		example: "openrouter/openai/gpt-4o",
+		description: "Vision models via OpenRouter",
+	},
+	{
+		value: "OLLAMA",
+		label: "Ollama",
+		example: "llava, bakllava",
+		description: "Local vision models via Ollama",
+		apiBase: "http://localhost:11434",
+	},
+	{
+		value: "GROQ",
+		label: "Groq",
+		example: "llama-4-scout-17b-16e-instruct",
+		description: "Vision models on Groq",
+	},
+	{
+		value: "TOGETHER_AI",
+		label: "Together AI",
+		example: "meta-llama/Llama-4-Scout-17B-16E-Instruct",
+		description: "Vision models on Together AI",
+	},
+	{
+		value: "FIREWORKS_AI",
+		label: "Fireworks AI",
+		example: "fireworks_ai/phi-3-vision-128k-instruct",
+		description: "Vision models on Fireworks AI",
+	},
+	{
+		value: "DEEPSEEK",
+		label: "DeepSeek",
+		example: "deepseek-chat",
+		description: "DeepSeek vision models",
+		apiBase: "https://api.deepseek.com",
+	},
+	{
+		value: "MISTRAL",
+		label: "Mistral",
+		example: "pixtral-large-latest",
+		description: "Pixtral vision models",
+	},
+	{
+		value: "CUSTOM",
+		label: "Custom Provider",
+		example: "custom/my-vision-model",
+		description: "Custom OpenAI-compatible vision endpoint",
+	},
+];

surfsense_web/contracts/types/new-llm-config.types.ts

@@ -252,23 +252,99 @@ export const globalImageGenConfig = z.object({
 
 export const getGlobalImageGenConfigsResponse = z.array(globalImageGenConfig);
 
+// =============================================================================
+// Vision LLM Config (separate table for vision-capable models)
+// =============================================================================
+
+export const visionProviderEnum = z.enum([
+	"OPENAI",
+	"ANTHROPIC",
+	"GOOGLE",
+	"AZURE_OPENAI",
+	"VERTEX_AI",
+	"BEDROCK",
+	"XAI",
+	"OPENROUTER",
+	"OLLAMA",
+	"GROQ",
+	"TOGETHER_AI",
+	"FIREWORKS_AI",
+	"DEEPSEEK",
+	"MISTRAL",
+	"CUSTOM",
+]);
+
+export type VisionProvider = z.infer<typeof visionProviderEnum>;
+
+export const visionLLMConfig = z.object({
+	id: z.number(),
+	name: z.string().max(100),
+	description: z.string().max(500).nullable().optional(),
+	provider: visionProviderEnum,
+	custom_provider: z.string().max(100).nullable().optional(),
+	model_name: z.string().max(100),
+	api_key: z.string(),
+	api_base: z.string().max(500).nullable().optional(),
+	api_version: z.string().max(50).nullable().optional(),
+	litellm_params: z.record(z.string(), z.any()).nullable().optional(),
+	created_at: z.string(),
+	search_space_id: z.number(),
+	user_id: z.string(),
+});
+
+export const createVisionLLMConfigRequest = visionLLMConfig.omit({
+	id: true,
+	created_at: true,
+	user_id: true,
+});
+
+export const createVisionLLMConfigResponse = visionLLMConfig;
+
+export const getVisionLLMConfigsResponse = z.array(visionLLMConfig);
+
+export const updateVisionLLMConfigRequest = z.object({
+	id: z.number(),
+	data: visionLLMConfig
+		.omit({ id: true, created_at: true, search_space_id: true, user_id: true })
+		.partial(),
+});
+
+export const updateVisionLLMConfigResponse = visionLLMConfig;
+
+export const deleteVisionLLMConfigResponse = z.object({
+	message: z.string(),
+	id: z.number(),
+});
+
+export const globalVisionLLMConfig = z.object({
+	id: z.number(),
+	name: z.string(),
+	description: z.string().nullable().optional(),
+	provider: z.string(),
+	custom_provider: z.string().nullable().optional(),
+	model_name: z.string(),
+	api_base: z.string().nullable().optional(),
+	api_version: z.string().nullable().optional(),
+	litellm_params: z.record(z.string(), z.any()).nullable().optional(),
+	is_global: z.literal(true),
+	is_auto_mode: z.boolean().optional().default(false),
+});
+
+export const getGlobalVisionLLMConfigsResponse = z.array(globalVisionLLMConfig);
+
 // =============================================================================
 // LLM Preferences (Role Assignments)
 // =============================================================================
 
-/**
- * LLM Preferences schemas - for role assignments
- * image_generation uses image_generation_config_id (not llm_id)
- */
 export const llmPreferences = z.object({
 	agent_llm_id: z.union([z.number(), z.null()]).optional(),
 	document_summary_llm_id: z.union([z.number(), z.null()]).optional(),
 	image_generation_config_id: z.union([z.number(), z.null()]).optional(),
-	vision_llm_id: z.union([z.number(), z.null()]).optional(),
+	vision_llm_config_id: z.union([z.number(), z.null()]).optional(),
 	agent_llm: z.union([z.record(z.string(), z.unknown()), z.null()]).optional(),
 	document_summary_llm: z.union([z.record(z.string(), z.unknown()), z.null()]).optional(),
 	image_generation_config: z.union([z.record(z.string(), z.unknown()), z.null()]).optional(),
-	vision_llm: z.union([z.record(z.string(), z.unknown()), z.null()]).optional(),
+	vision_llm_config: z.union([z.record(z.string(), z.unknown()), z.null()]).optional(),
 });
 
 /**
@@ -289,7 +365,7 @@ export const updateLLMPreferencesRequest = z.object({
 		agent_llm_id: true,
 		document_summary_llm_id: true,
 		image_generation_config_id: true,
-		vision_llm_id: true,
+		vision_llm_config_id: true,
 	}),
 });
 
@@ -341,6 +417,15 @@ export type UpdateImageGenConfigResponse = z.infer<typeof updateImageGenConfigRe
 export type DeleteImageGenConfigResponse = z.infer<typeof deleteImageGenConfigResponse>;
 export type GlobalImageGenConfig = z.infer<typeof globalImageGenConfig>;
 export type GetGlobalImageGenConfigsResponse = z.infer<typeof getGlobalImageGenConfigsResponse>;
+export type VisionLLMConfig = z.infer<typeof visionLLMConfig>;
+export type CreateVisionLLMConfigRequest = z.infer<typeof createVisionLLMConfigRequest>;
+export type CreateVisionLLMConfigResponse = z.infer<typeof createVisionLLMConfigResponse>;
+export type GetVisionLLMConfigsResponse = z.infer<typeof getVisionLLMConfigsResponse>;
+export type UpdateVisionLLMConfigRequest = z.infer<typeof updateVisionLLMConfigRequest>;
+export type UpdateVisionLLMConfigResponse = z.infer<typeof updateVisionLLMConfigResponse>;
+export type DeleteVisionLLMConfigResponse = z.infer<typeof deleteVisionLLMConfigResponse>;
+export type GlobalVisionLLMConfig = z.infer<typeof globalVisionLLMConfig>;
+export type GetGlobalVisionLLMConfigsResponse = z.infer<typeof getGlobalVisionLLMConfigsResponse>;
 export type LLMPreferences = z.infer<typeof llmPreferences>;
 export type GetLLMPreferencesRequest = z.infer<typeof getLLMPreferencesRequest>;
 export type GetLLMPreferencesResponse = z.infer<typeof getLLMPreferencesResponse>;

surfsense_web/lib/apis/vision-llm-config-api.service.ts

@@ -0,0 +1,58 @@
+import {
+	type CreateVisionLLMConfigRequest,
+	createVisionLLMConfigRequest,
+	createVisionLLMConfigResponse,
+	deleteVisionLLMConfigResponse,
+	getGlobalVisionLLMConfigsResponse,
+	getVisionLLMConfigsResponse,
+	type UpdateVisionLLMConfigRequest,
+	updateVisionLLMConfigRequest,
+	updateVisionLLMConfigResponse,
+} from "@/contracts/types/new-llm-config.types";
+import { ValidationError } from "../error";
+import { baseApiService } from "./base-api.service";
+
+class VisionLLMConfigApiService {
+	getGlobalConfigs = async () => {
+		return baseApiService.get(
+			`/api/v1/global-vision-llm-configs`,
+			getGlobalVisionLLMConfigsResponse
+		);
+	};
+
+	createConfig = async (request: CreateVisionLLMConfigRequest) => {
+		const parsed = createVisionLLMConfigRequest.safeParse(request);
+		if (!parsed.success) {
+			const msg = parsed.error.issues.map((i) => i.message).join(", ");
+			throw new ValidationError(`Invalid request: ${msg}`);
+		}
+		return baseApiService.post(`/api/v1/vision-llm-configs`, createVisionLLMConfigResponse, {
+			body: parsed.data,
+		});
+	};
+
+	getConfigs = async (searchSpaceId: number) => {
+		const params = new URLSearchParams({
+			search_space_id: String(searchSpaceId),
+		}).toString();
+		return baseApiService.get(`/api/v1/vision-llm-configs?${params}`, getVisionLLMConfigsResponse);
+	};
+
+	updateConfig = async (request: UpdateVisionLLMConfigRequest) => {
+		const parsed = updateVisionLLMConfigRequest.safeParse(request);
+		if (!parsed.success) {
+			const msg = parsed.error.issues.map((i) => i.message).join(", ");
+			throw new ValidationError(`Invalid request: ${msg}`);
+		}
+		const { id, data } = parsed.data;
+		return baseApiService.put(`/api/v1/vision-llm-configs/${id}`, updateVisionLLMConfigResponse, {
+			body: data,
+		});
+	};
+
+	deleteConfig = async (id: number) => {
+		return baseApiService.delete(`/api/v1/vision-llm-configs/${id}`, deleteVisionLLMConfigResponse);
+	};
+}
+
+export const visionLLMConfigApiService = new VisionLLMConfigApiService();

surfsense_web/lib/query-client/cache-keys.ts

@@ -39,6 +39,11 @@ export const cacheKeys = {
 		byId: (configId: number) => ["image-gen-configs", "detail", configId] as const,
 		global: () => ["image-gen-configs", "global"] as const,
 	},
+	visionLLMConfigs: {
+		all: (searchSpaceId: number) => ["vision-llm-configs", searchSpaceId] as const,
+		byId: (configId: number) => ["vision-llm-configs", "detail", configId] as const,
+		global: () => ["vision-llm-configs", "global"] as const,
+	},
 	auth: {
 		user: ["auth", "user"] as const,
 	},

surfsense_web/messages/en.json

@@ -738,6 +738,8 @@
 		"nav_role_assignments_desc": "Assign configs to agent roles",
 		"nav_image_models": "Image Models",
 		"nav_image_models_desc": "Configure image generation models",
+		"nav_vision_models": "Vision Models",
+		"nav_vision_models_desc": "Configure vision-capable LLM models",
 		"nav_system_instructions": "System Instructions",
 		"nav_system_instructions_desc": "SearchSpace-wide AI instructions",
 		"nav_public_links": "Public Chat Links",

surfsense_web/messages/es.json

@@ -738,6 +738,8 @@
 		"nav_role_assignments_desc": "Asignar configuraciones a roles de agente",
 		"nav_image_models": "Modelos de imagen",
 		"nav_image_models_desc": "Configurar modelos de generación de imágenes",
+		"nav_vision_models": "Modelos de visión",
+		"nav_vision_models_desc": "Configurar modelos LLM con capacidad de visión",
 		"nav_system_instructions": "Instrucciones del sistema",
 		"nav_system_instructions_desc": "Instrucciones de IA a nivel del espacio de búsqueda",
 		"nav_public_links": "Enlaces de chat públicos",

surfsense_web/messages/hi.json

@@ -738,6 +738,8 @@
 		"nav_role_assignments_desc": "एजेंट भूमिकाओं को कॉन्फ़िगरेशन असाइन करें",
 		"nav_image_models": "इमेज मॉडल",
 		"nav_image_models_desc": "इमेज जनरेशन मॉडल कॉन्फ़िगर करें",
+		"nav_vision_models": "विज़न मॉडल",
+		"nav_vision_models_desc": "विज़न-सक्षम LLM मॉडल कॉन्फ़िगर करें",
 		"nav_system_instructions": "सिस्टम निर्देश",
 		"nav_system_instructions_desc": "सर्च स्पेस-व्यापी AI निर्देश",
 		"nav_public_links": "सार्वजनिक चैट लिंक",

surfsense_web/messages/pt.json

@@ -738,6 +738,8 @@
 		"nav_role_assignments_desc": "Atribuir configurações a funções do agente",
 		"nav_image_models": "Modelos de imagem",
 		"nav_image_models_desc": "Configurar modelos de geração de imagens",
+		"nav_vision_models": "Modelos de visão",
+		"nav_vision_models_desc": "Configurar modelos LLM com capacidade de visão",
 		"nav_system_instructions": "Instruções do sistema",
 		"nav_system_instructions_desc": "Instruções de IA em nível do espaço de pesquisa",
 		"nav_public_links": "Links de chat públicos",

surfsense_web/messages/zh.json

@@ -722,6 +722,8 @@
 		"nav_role_assignments_desc": "为代理角色分配配置",
 		"nav_image_models": "图像模型",
 		"nav_image_models_desc": "配置图像生成模型",
+		"nav_vision_models": "视觉模型",
+		"nav_vision_models_desc": "配置具有视觉能力的LLM模型",
 		"nav_system_instructions": "系统指令",
 		"nav_system_instructions_desc": "搜索空间级别的 AI 指令",
 		"nav_public_links": "公开聊天链接",