ui : gate llama.cpp-only features by backend capabilities

Assisted-by: pi:llama.cpp/DeepSeek-V4.1-Flash

tools/ui/src/lib/constants/backend.constants.ts

@@ -1,4 +1,4 @@
-import type { BackendPreset, BackendProtocol } from '$lib/types';
+import type { BackendCapabilities, BackendPreset, BackendProtocol } from '$lib/types';
 
 /** Version sent with the Anthropic Messages API. */
 export const ANTHROPIC_API_VERSION = '2023-06-01';
@@ -18,6 +18,34 @@ export const DEFAULT_BACKEND_MODELS_PATH = '/v1/models';
 /** Id of the built-in backend that points at the server serving this UI. */
 export const LOCAL_BACKEND_ID = 'local';
 
+/** Capabilities of a full llama.cpp server. */
+const LLAMA_CPP_CAPABILITIES: BackendCapabilities = {
+	corsProxy: true,
+	loadUnload: true,
+	props: true,
+	router: true,
+	slots: true,
+	statusFeed: true,
+	tools: true
+};
+/** Capabilities of a plain OpenAI- or Anthropic-compatible endpoint. */
+const COMPATIBLE_CAPABILITIES: BackendCapabilities = {
+	corsProxy: false,
+	loadUnload: false,
+	props: false,
+	router: false,
+	slots: false,
+	statusFeed: false,
+	tools: false
+};
+
+/** Capabilities per backend protocol. */
+export const BACKEND_CAPABILITIES: Record<BackendProtocol, BackendCapabilities> = {
+	anthropic: COMPATIBLE_CAPABILITIES,
+	'llama.cpp': LLAMA_CPP_CAPABILITIES,
+	openai: COMPATIBLE_CAPABILITIES
+};
+
 /**
  * Ready-made endpoints offered when adding a backend. `custom` intentionally
  * carries no URL so the user starts from an empty form.

tools/ui/src/lib/services/chat.service.ts

@@ -33,6 +33,7 @@ import {
 	StreamConnectionState
 } from '$lib/enums';
 import { modelsStore } from '$lib/stores/models/index.svelte';
+import { serverStore } from '$lib/stores/server.svelte';
 import { settingsStore } from '$lib/stores/settings/index.svelte';
 import type { DatabaseMessageExtraMcpPrompt, DatabaseMessageExtraMcpResource } from '$lib/types';
 import type {
@@ -87,6 +88,9 @@ export class ChatService {
 	 * @returns {Promise<boolean>} Promise that resolves to true if all slots are idle, false if any is processing
 	 */
 	static async areAllSlotsIdle(model?: string | null, signal?: AbortSignal): Promise<boolean> {
+		// the /slots endpoint only exists on llama.cpp servers
+		if (!serverStore.capabilities.slots) return true;
+
 		try {
 			const url = model ? `${API_SLOTS.LIST}?model=${encodeURIComponent(model)}` : API_SLOTS.LIST;
 			const res = await fetch(apiUrl(url), { signal });

tools/ui/src/lib/stores/server.svelte.ts

@@ -6,9 +6,13 @@
  * PropsService for the /props fetch.
  */
 
+import { BACKEND_CAPABILITIES } from '$lib/constants';
 import { ServerRole } from '$lib/enums';
 import { PropsService } from '$lib/services/props.service';
+import type { BackendCapabilities } from '$lib/types';
 import { ApiError } from '$lib/utils';
+import { getBackend } from '$lib/utils/api-base';
+import { getBackendCapabilities } from '$lib/utils/backend';
 
 const LOADING_RETRY_INTERVAL_MS = 1000;
 
@@ -21,6 +25,13 @@ class ServerStore {
 	private fetchPromise: Promise<void> | null = null;
 	private retryTimer: ReturnType<typeof setTimeout> | null = null;
 
+	/** Features of the active backend. Defaults to full llama.cpp support. */
+	get capabilities(): BackendCapabilities {
+		const backend = getBackend();
+
+		return backend ? getBackendCapabilities(backend) : BACKEND_CAPABILITIES['llama.cpp'];
+	}
+
 	get contextSize(): number | null {
 		const nCtx = this.props?.default_generation_settings?.n_ctx;
 
@@ -63,6 +74,15 @@ class ServerStore {
 
 		this.clearRetryTimer();
 
+		// External backends expose no /props endpoint. Keep MODEL-mode defaults so
+		// role detection and generation defaults degrade instead of failing.
+		if (!this.capabilities.props) {
+			this.clear();
+			this.role = ServerRole.MODEL;
+
+			return;
+		}
+
 		if (!background) {
 			this.loading = true;
 		}

tools/ui/src/lib/stores/tools.svelte.ts

@@ -30,6 +30,7 @@ import { ToolsService } from '$lib/services/tools.service';
 // direct imports between stores, not via the barrel, to avoid circular deps
 import { mcpStore } from '$lib/stores/mcp/index.svelte';
 import { modelsStore } from '$lib/stores/models/index.svelte';
+import { serverStore } from '$lib/stores/server.svelte';
 import { settingsStore } from '$lib/stores/settings/index.svelte';
 import type { OpenAIToolDefinition, ToolEntry, ToolGroup } from '$lib/types';
 import { ApiError, buildSandboxToolDefinition } from '$lib/utils';
@@ -232,6 +233,14 @@ class ToolsStore {
 	}
 
 	async fetchServerTools(): Promise<void> {
+		// the /tools endpoint only exists on llama.cpp servers
+		if (!serverStore.capabilities.tools) {
+			this._serverTools = [];
+			this.cwdAwareTools = new SvelteSet();
+
+			return;
+		}
+
 		if (this._loading) return;
 
 		this._loading = true;

tools/ui/src/lib/types/backend.d.ts

@@ -10,6 +10,28 @@
 /** Request/response shape a backend speaks. */
 export type BackendProtocol = 'llama.cpp' | 'openai' | 'anthropic';
 
+/**
+ * Features a backend supports. A llama.cpp server exposes extra endpoints on
+ * top of the OpenAI-compatible API; plain OpenAI- and Anthropic-compatible
+ * endpoints only provide chat and model listing.
+ */
+export interface BackendCapabilities {
+	/** llama-server's /cors-proxy endpoint for cross-origin MCP requests. */
+	corsProxy: boolean;
+	/** Router-mode model load/unload. */
+	loadUnload: boolean;
+	/** The /props endpoint with server role and generation defaults. */
+	props: boolean;
+	/** Multi-model router mode. */
+	router: boolean;
+	/** The /slots introspection endpoint. */
+	slots: boolean;
+	/** The /models/sse load and download progress feed. */
+	statusFeed: boolean;
+	/** The /tools listing and execution endpoint. */
+	tools: boolean;
+}
+
 /** One configured API endpoint. */
 export interface Backend {
 	/** Bearer token / API key used for this backend. */

tools/ui/src/lib/types/index.ts

@@ -36,7 +36,7 @@ export type {
 } from './api';
 
 // Backend types
-export type { Backend, BackendPreset, BackendProtocol } from './backend';
+export type { Backend, BackendCapabilities, BackendPreset, BackendProtocol } from './backend';
 
 // HuggingFace types
 export type {

tools/ui/src/lib/utils/backend.ts

@@ -7,13 +7,14 @@
  */
 
 import {
+	BACKEND_CAPABILITIES,
 	BACKEND_ID_PREFIX,
 	BACKEND_PROTOCOLS,
 	DEFAULT_BACKEND_CHAT_PATH,
 	DEFAULT_BACKEND_MODELS_PATH,
 	LOCAL_BACKEND_ID
 } from '$lib/constants';
-import type { Backend, BackendProtocol } from '$lib/types';
+import type { Backend, BackendCapabilities, BackendProtocol } from '$lib/types';
 
 /** Absolute chat completions URL for a backend. */
 export function backendChatUrl(backend: Backend): string {
@@ -25,6 +26,11 @@ export function backendModelsUrl(backend: Backend): string {
 	return joinBackendUrl(backend.baseUrl, backend.modelsPath ?? DEFAULT_BACKEND_MODELS_PATH);
 }
 
+/** Features a backend supports, derived from its protocol. */
+export function getBackendCapabilities(backend: Backend): BackendCapabilities {
+	return BACKEND_CAPABILITIES[backend.protocol] ?? BACKEND_CAPABILITIES.openai;
+}
+
 /** The built-in backend pointing at the server that serves this UI. */
 export function createLocalBackend(apiKey?: string, enabled = true): Backend {
 	return {

tools/ui/src/lib/utils/index.ts

@@ -21,6 +21,7 @@ export {
 	backendChatUrl,
 	backendModelsUrl,
 	createLocalBackend,
+	getBackendCapabilities,
 	parseBackendsSettings
 } from './backend';