Merge branch 'upstream' into concedo_experimental

# Conflicts:
#	.github/ISSUE_TEMPLATE/010-bug-compilation.yml
#	.github/ISSUE_TEMPLATE/011-bug-results.yml
#	AGENTS.md
#	SECURITY.md
#	ggml/src/ggml-hexagon/htp/flash-attn-ops.c
#	ggml/src/ggml-hexagon/htp/main.c
#	ggml/src/ggml-opencl/CMakeLists.txt
#	ggml/src/ggml-opencl/ggml-opencl.cpp
#	ggml/src/ggml-opencl/kernels/cvt.cl
#	scripts/sync_vendor.py
#	src/unicode.cpp
#	tests/test-backend-ops.cpp
#	tools/cli/cli.cpp

tools/server/webui/src/lib/components/app/chat/ChatMessages/ChatMessage.svelte

@@ -1,5 +1,6 @@
 <script lang="ts">
 	import { goto } from '$app/navigation';
+	import { base } from '$app/paths';
 	import {
 		chatStore,
 		pendingEditMessageId,
@@ -119,7 +120,7 @@
 			const conversationDeleted = await removeSystemPromptPlaceholder(message.id);
 
 			if (conversationDeleted) {
-				goto('/');
+				goto(`${base}/`);
 			}
 
 			return;
@@ -220,7 +221,7 @@
 				const conversationDeleted = await removeSystemPromptPlaceholder(message.id);
 				isEditing = false;
 				if (conversationDeleted) {
-					goto('/');
+					goto(`${base}/`);
 				}
 				return;
 			}

tools/server/webui/src/lib/components/app/chat/ChatMessages/ChatMessageActions.svelte

@@ -5,6 +5,7 @@
 		ChatMessageBranchingControls,
 		DialogConfirmation
 	} from '$lib/components/app';
+	import { Switch } from '$lib/components/ui/switch';
 
 	interface Props {
 		role: 'user' | 'assistant';
@@ -26,6 +27,9 @@
 		onConfirmDelete: () => void;
 		onNavigateToSibling?: (siblingId: string) => void;
 		onShowDeleteDialogChange: (show: boolean) => void;
+		showRawOutputSwitch?: boolean;
+		rawOutputEnabled?: boolean;
+		onRawOutputToggle?: (enabled: boolean) => void;
 	}
 
 	let {
@@ -42,7 +46,10 @@
 		onRegenerate,
 		role,
 		siblingInfo = null,
-		showDeleteDialog
+		showDeleteDialog,
+		showRawOutputSwitch = false,
+		rawOutputEnabled = false,
+		onRawOutputToggle
 	}: Props = $props();
 
 	function handleConfirmDelete() {
@@ -51,9 +58,9 @@
 	}
 </script>
 
-<div class="relative {justify === 'start' ? 'mt-2' : ''} flex h-6 items-center justify-{justify}">
+<div class="relative {justify === 'start' ? 'mt-2' : ''} flex h-6 items-center justify-between">
 	<div
-		class="absolute top-0 {actionsPosition === 'left'
+		class="{actionsPosition === 'left'
 			? 'left-0'
 			: 'right-0'} flex items-center gap-2 opacity-100 transition-opacity"
 	>
@@ -81,6 +88,16 @@
 			<ActionButton icon={Trash2} tooltip="Delete" onclick={onDelete} />
 		</div>
 	</div>
+
+	{#if showRawOutputSwitch}
+		<div class="flex items-center gap-2">
+			<span class="text-xs text-muted-foreground">Show raw output</span>
+			<Switch
+				checked={rawOutputEnabled}
+				onCheckedChange={(checked) => onRawOutputToggle?.(checked)}
+			/>
+		</div>
+	{/if}
 </div>
 
 <DialogConfirmation

tools/server/webui/src/lib/components/app/chat/ChatMessages/ChatMessageAssistant.svelte

@@ -90,6 +90,9 @@
 
 	const processingState = useProcessingState();
 
+	// Local state for raw output toggle (per message)
+	let showRawOutput = $state(false);
+
 	let currentConfig = $derived(config());
 	let isRouter = $derived(isRouterMode());
 	let displayedModel = $derived((): string | null => {
@@ -238,7 +241,7 @@
 			</div>
 		</div>
 	{:else if message.role === 'assistant'}
-		{#if config().disableReasoningFormat}
+		{#if showRawOutput}
 			<pre class="raw-output">{messageContent || ''}</pre>
 		{:else}
 			<MarkdownContent content={messageContent || ''} />
@@ -352,6 +355,9 @@
 			{onConfirmDelete}
 			{onNavigateToSibling}
 			{onShowDeleteDialogChange}
+			showRawOutputSwitch={currentConfig.showRawOutputSwitch}
+			rawOutputEnabled={showRawOutput}
+			onRawOutputToggle={(enabled) => (showRawOutput = enabled)}
 		/>
 	{/if}
 </div>

tools/server/webui/src/lib/components/app/chat/ChatMessages/ChatMessageStatistics.svelte

@@ -3,6 +3,7 @@
 	import { BadgeChatStatistic } from '$lib/components/app';
 	import * as Tooltip from '$lib/components/ui/tooltip';
 	import { ChatMessageStatsView } from '$lib/enums';
+	import { formatPerformanceTime } from '$lib/utils/formatters';
 
 	interface Props {
 		predictedTokens?: number;
@@ -57,8 +58,8 @@
 	);
 
 	let tokensPerSecond = $derived(hasGenerationStats ? (predictedTokens! / predictedMs!) * 1000 : 0);
-	let timeInSeconds = $derived(
-		predictedMs !== undefined ? (predictedMs / 1000).toFixed(2) : '0.00'
+	let formattedTime = $derived(
+		predictedMs !== undefined ? formatPerformanceTime(predictedMs) : '0s'
 	);
 
 	let promptTokensPerSecond = $derived(
@@ -67,15 +68,15 @@
 			: undefined
 	);
 
-	let promptTimeInSeconds = $derived(
-		promptMs !== undefined ? (promptMs / 1000).toFixed(2) : undefined
+	let formattedPromptTime = $derived(
+		promptMs !== undefined ? formatPerformanceTime(promptMs) : undefined
 	);
 
 	let hasPromptStats = $derived(
 		promptTokens !== undefined &&
 			promptMs !== undefined &&
 			promptTokensPerSecond !== undefined &&
-			promptTimeInSeconds !== undefined
+			formattedPromptTime !== undefined
 	);
 
 	// In live mode, generation tab is disabled until we have generation stats
@@ -142,7 +143,7 @@
 			<BadgeChatStatistic
 				class="bg-transparent"
 				icon={Clock}
-				value="{timeInSeconds}s"
+				value={formattedTime}
 				tooltipLabel="Generation time"
 			/>
 			<BadgeChatStatistic
@@ -161,7 +162,7 @@
 			<BadgeChatStatistic
 				class="bg-transparent"
 				icon={Clock}
-				value="{promptTimeInSeconds}s"
+				value={formattedPromptTime ?? '0s'}
 				tooltipLabel="Prompt processing time"
 			/>
 			<BadgeChatStatistic

tools/server/webui/src/lib/components/app/chat/ChatScreen/ChatScreen.svelte

@@ -21,6 +21,7 @@
 		chatStore,
 		errorDialog,
 		isLoading,
+		isChatStreaming,
 		isEditing,
 		getAddFilesHandler
 	} from '$lib/stores/chat.svelte';
@@ -81,7 +82,7 @@
 	let isServerLoading = $derived(serverLoading());
 	let hasPropsError = $derived(!!serverError());
 
-	let isCurrentConversationLoading = $derived(isLoading());
+	let isCurrentConversationLoading = $derived(isLoading() || isChatStreaming());
 
 	let isRouter = $derived(isRouterMode());
 

tools/server/webui/src/lib/components/app/chat/ChatSettings/ChatSettings.svelte

@@ -254,8 +254,13 @@
 					type: 'checkbox'
 				},
 				{
-					key: 'disableReasoningFormat',
-					label: 'Show raw LLM output',
+					key: 'disableReasoningParsing',
+					label: 'Disable reasoning content parsing',
+					type: 'checkbox'
+				},
+				{
+					key: 'showRawOutputSwitch',
+					label: 'Enable raw output toggle',
 					type: 'checkbox'
 				},
 				{

tools/server/webui/src/lib/components/app/misc/DropdownMenuSearchable.svelte

@@ -0,0 +1,88 @@
+<script lang="ts">
+	import type { Snippet } from 'svelte';
+	import * as DropdownMenu from '$lib/components/ui/dropdown-menu';
+	import { cn } from '$lib/components/ui/utils';
+	import { SearchInput } from '$lib/components/app';
+
+	interface Props {
+		open?: boolean;
+		onOpenChange?: (open: boolean) => void;
+		placeholder?: string;
+		searchValue?: string;
+		onSearchChange?: (value: string) => void;
+		onSearchKeyDown?: (event: KeyboardEvent) => void;
+		align?: 'start' | 'center' | 'end';
+		contentClass?: string;
+		emptyMessage?: string;
+		isEmpty?: boolean;
+		disabled?: boolean;
+		trigger: Snippet;
+		children: Snippet;
+		footer?: Snippet;
+	}
+
+	let {
+		open = $bindable(false),
+		onOpenChange,
+		placeholder = 'Search...',
+		searchValue = $bindable(''),
+		onSearchChange,
+		onSearchKeyDown,
+		align = 'start',
+		contentClass = 'w-72',
+		emptyMessage = 'No items found',
+		isEmpty = false,
+		disabled = false,
+		trigger,
+		children,
+		footer
+	}: Props = $props();
+
+	function handleOpenChange(newOpen: boolean) {
+		open = newOpen;
+
+		if (!newOpen) {
+			searchValue = '';
+			onSearchChange?.('');
+		}
+
+		onOpenChange?.(newOpen);
+	}
+</script>
+
+<DropdownMenu.Root bind:open onOpenChange={handleOpenChange}>
+	<DropdownMenu.Trigger
+		{disabled}
+		onclick={(e) => {
+			e.preventDefault();
+			e.stopPropagation();
+		}}
+	>
+		{@render trigger()}
+	</DropdownMenu.Trigger>
+
+	<DropdownMenu.Content {align} class={cn(contentClass, 'pt-0')}>
+		<div class="sticky top-0 z-10 mb-2 bg-popover p-1 pt-2">
+			<SearchInput
+				{placeholder}
+				bind:value={searchValue}
+				onInput={onSearchChange}
+				onKeyDown={onSearchKeyDown}
+			/>
+		</div>
+
+		<div class={cn('overflow-y-auto')}>
+			{@render children()}
+
+			{#if isEmpty}
+				<div class="px-2 py-3 text-center text-sm text-muted-foreground">{emptyMessage}</div>
+			{/if}
+		</div>
+
+		{#if footer}
+			<DropdownMenu.Separator />
+
+			{@render footer()}
+		{/if}
+	</DropdownMenu.Content>
+</DropdownMenu.Root>

tools/server/webui/src/lib/components/app/misc/MarkdownContent.svelte

@@ -486,6 +486,8 @@
 		text-decoration: underline;
 		text-underline-offset: 2px;
 		transition: color 0.2s ease;
+		overflow-wrap: anywhere;
+		word-break: break-all;
 	}
 
 	div :global(a:hover) {

tools/server/webui/src/lib/constants/settings-config.ts

@@ -7,7 +7,8 @@ export const SETTING_CONFIG_DEFAULT: Record<string, string | number | boolean> =
 	theme: 'system',
 	showThoughtInProgress: false,
 	showToolCalls: false,
-	disableReasoningFormat: false,
+	disableReasoningParsing: false,
+	showRawOutputSwitch: false,
 	keepStatsVisible: false,
 	showMessageStats: true,
 	askForTitleConfirmation: false,
@@ -92,8 +93,10 @@ export const SETTING_CONFIG_INFO: Record<string, string> = {
 	showThoughtInProgress: 'Expand thought process by default when generating messages.',
 	showToolCalls:
 		'Display tool call labels and payloads from Harmony-compatible delta.tool_calls data below assistant messages.',
-	disableReasoningFormat:
-		'Show raw LLM output without backend parsing and frontend Markdown rendering to inspect streaming across different models.',
+	disableReasoningParsing:
+		'Send reasoning_format=none to prevent server-side extraction of reasoning tokens into separate field',
+	showRawOutputSwitch:
+		'Show toggle button to display messages as plain text instead of Markdown-formatted content',
 	keepStatsVisible: 'Keep processing statistics visible after generation finishes.',
 	showMessageStats:
 		'Display generation statistics (tokens/second, token count, duration) below each assistant message.',

tools/server/webui/src/lib/services/chat.ts

@@ -90,7 +90,7 @@ export class ChatService {
 			custom,
 			timings_per_token,
 			// Config options
-			disableReasoningFormat
+			disableReasoningParsing
 		} = options;
 
 		const normalizedMessages: ApiChatMessageData[] = messages
@@ -127,7 +127,7 @@ export class ChatService {
 			requestBody.model = options.model;
 		}
 
-		requestBody.reasoning_format = disableReasoningFormat ? 'none' : 'auto';
+		requestBody.reasoning_format = disableReasoningParsing ? 'none' : 'auto';
 
 		if (temperature !== undefined) requestBody.temperature = temperature;
 		if (max_tokens !== undefined) {

tools/server/webui/src/lib/services/parameter-sync.ts

@@ -70,12 +70,6 @@ export const SYNCABLE_PARAMETERS: SyncableParameter[] = [
 		canSync: true
 	},
 	{ key: 'showToolCalls', serverKey: 'showToolCalls', type: 'boolean', canSync: true },
-	{
-		key: 'disableReasoningFormat',
-		serverKey: 'disableReasoningFormat',
-		type: 'boolean',
-		canSync: true
-	},
 	{ key: 'keepStatsVisible', serverKey: 'keepStatsVisible', type: 'boolean', canSync: true },
 	{ key: 'showMessageStats', serverKey: 'showMessageStats', type: 'boolean', canSync: true },
 	{

tools/server/webui/src/lib/stores/chat.svelte.ts

@@ -118,6 +118,16 @@ class ChatStore {
 		this.isLoading = this.isChatLoading(convId);
 		const streamingState = this.getChatStreaming(convId);
 		this.currentResponse = streamingState?.response || '';
+		this.isStreamingActive = streamingState !== undefined;
+		this.setActiveProcessingConversation(convId);
+
+		// Sync streaming content to activeMessages so UI displays current content
+		if (streamingState?.response && streamingState?.messageId) {
+			const idx = conversationsStore.findMessageIndex(streamingState.messageId);
+			if (idx !== -1) {
+				conversationsStore.updateMessageAtIndex(idx, { content: streamingState.response });
+			}
+		}
 	}
 
 	/**
@@ -1639,7 +1649,7 @@ class ChatStore {
 
 		// Config options needed by ChatService
 		if (currentConfig.systemMessage) apiOptions.systemMessage = currentConfig.systemMessage;
-		if (currentConfig.disableReasoningFormat) apiOptions.disableReasoningFormat = true;
+		if (currentConfig.disableReasoningParsing) apiOptions.disableReasoningParsing = true;
 
 		if (hasValue(currentConfig.temperature))
 			apiOptions.temperature = Number(currentConfig.temperature);

tools/server/webui/src/lib/types/settings.d.ts

@@ -18,8 +18,8 @@ export interface SettingsChatServiceOptions {
 	model?: string;
 	// System message to inject
 	systemMessage?: string;
-	// Disable reasoning format (use 'none' instead of 'auto')
-	disableReasoningFormat?: boolean;
+	// Disable reasoning parsing (use 'none' instead of 'auto')
+	disableReasoningParsing?: boolean;
 	// Generation parameters
 	temperature?: number;
 	max_tokens?: number;

tools/server/webui/src/lib/utils/formatters.ts

@@ -51,3 +51,75 @@ export function formatNumber(num: number | unknown): string {
 
 	return num.toLocaleString();
 }
+
+/**
+ * Format JSON string with pretty printing (2-space indentation)
+ * Returns original string if parsing fails
+ *
+ * @param jsonString - JSON string to format
+ * @returns Pretty-printed JSON string or original if invalid
+ */
+export function formatJsonPretty(jsonString: string): string {
+	try {
+		const parsed = JSON.parse(jsonString);
+		return JSON.stringify(parsed, null, 2);
+	} catch {
+		return jsonString;
+	}
+}
+
+/**
+ * Format time as HH:MM:SS in 24-hour format
+ *
+ * @param date - Date object to format
+ * @returns Formatted time string (HH:MM:SS)
+ */
+export function formatTime(date: Date): string {
+	return date.toLocaleTimeString('en-US', {
+		hour12: false,
+		hour: '2-digit',
+		minute: '2-digit',
+		second: '2-digit'
+	});
+}
+
+/**
+ * Formats milliseconds to a human-readable time string for performance metrics.
+ * Examples: "4h 12min 54s", "12min 34s", "45s", "0.5s"
+ *
+ * @param ms - Time in milliseconds
+ * @returns Formatted time string
+ */
+export function formatPerformanceTime(ms: number): string {
+	if (ms < 0) return '0s';
+
+	const totalSeconds = ms / 1000;
+
+	if (totalSeconds < 1) {
+		return `${totalSeconds.toFixed(1)}s`;
+	}
+
+	if (totalSeconds < 10) {
+		return `${totalSeconds.toFixed(1)}s`;
+	}
+
+	const hours = Math.floor(totalSeconds / 3600);
+	const minutes = Math.floor((totalSeconds % 3600) / 60);
+	const seconds = Math.floor(totalSeconds % 60);
+
+	const parts: string[] = [];
+
+	if (hours > 0) {
+		parts.push(`${hours}h`);
+	}
+
+	if (minutes > 0) {
+		parts.push(`${minutes}min`);
+	}
+
+	if (seconds > 0 || parts.length === 0) {
+		parts.push(`${seconds}s`);
+	}
+
+	return parts.join(' ');
+}