webui: display prompt processing stats (#18146)

* webui: display prompt processing stats

* feat: Improve UI of Chat Message Statistics

* chore: update webui build output

* refactor: Post-review improvements

* chore: update webui build output

---------

Co-authored-by: Aleksander Grygier <aleksander.grygier@gmail.com>

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

@@ -244,7 +244,7 @@
 
 	<div class="info my-6 grid gap-4">
 		{#if displayedModel()}
-			<span class="inline-flex flex-wrap items-center gap-2 text-xs text-muted-foreground">
+			<div class="inline-flex flex-wrap items-start gap-2 text-xs text-muted-foreground">
 				{#if isRouter}
 					<ModelsSelector
 						currentModel={displayedModel()}
@@ -258,11 +258,13 @@
 
 				{#if currentConfig.showMessageStats && message.timings && message.timings.predicted_n && message.timings.predicted_ms}
 					<ChatMessageStatistics
+						promptTokens={message.timings.prompt_n}
+						promptMs={message.timings.prompt_ms}
 						predictedTokens={message.timings.predicted_n}
 						predictedMs={message.timings.predicted_ms}
 					/>
 				{/if}
-			</span>
+			</div>
 		{/if}
 
 		{#if config().showToolCalls}

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

@@ -1,20 +1,122 @@
 <script lang="ts">
-	import { Clock, Gauge, WholeWord } from '@lucide/svelte';
+	import { Clock, Gauge, WholeWord, BookOpenText, Sparkles } from '@lucide/svelte';
 	import { BadgeChatStatistic } from '$lib/components/app';
+	import * as Tooltip from '$lib/components/ui/tooltip';
+	import { ChatMessageStatsView } from '$lib/enums';
 
 	interface Props {
 		predictedTokens: number;
 		predictedMs: number;
+		promptTokens?: number;
+		promptMs?: number;
 	}
 
-	let { predictedTokens, predictedMs }: Props = $props();
+	let { predictedTokens, predictedMs, promptTokens, promptMs }: Props = $props();
+
+	let activeView: ChatMessageStatsView = $state(ChatMessageStatsView.GENERATION);
 
 	let tokensPerSecond = $derived((predictedTokens / predictedMs) * 1000);
 	let timeInSeconds = $derived((predictedMs / 1000).toFixed(2));
+
+	let promptTokensPerSecond = $derived(
+		promptTokens !== undefined && promptMs !== undefined
+			? (promptTokens / promptMs) * 1000
+			: undefined
+	);
+
+	let promptTimeInSeconds = $derived(
+		promptMs !== undefined ? (promptMs / 1000).toFixed(2) : undefined
+	);
+
+	let hasPromptStats = $derived(
+		promptTokens !== undefined &&
+			promptMs !== undefined &&
+			promptTokensPerSecond !== undefined &&
+			promptTimeInSeconds !== undefined
+	);
 </script>
 
-<BadgeChatStatistic icon={WholeWord} value="{predictedTokens} tokens" />
+<div class="inline-flex items-center text-xs text-muted-foreground">
+	<div class="inline-flex items-center rounded-sm bg-muted-foreground/15 p-0.5">
+		{#if hasPromptStats}
+			<Tooltip.Root>
+				<Tooltip.Trigger>
+					<button
+						type="button"
+						class="inline-flex h-5 w-5 items-center justify-center rounded-sm transition-colors {activeView ===
+						ChatMessageStatsView.READING
+							? 'bg-background text-foreground shadow-sm'
+							: 'hover:text-foreground'}"
+						onclick={() => (activeView = ChatMessageStatsView.READING)}
+					>
+						<BookOpenText class="h-3 w-3" />
+						<span class="sr-only">Reading</span>
+					</button>
+				</Tooltip.Trigger>
+				<Tooltip.Content>
+					<p>Reading (prompt processing)</p>
+				</Tooltip.Content>
+			</Tooltip.Root>
+		{/if}
+		<Tooltip.Root>
+			<Tooltip.Trigger>
+				<button
+					type="button"
+					class="inline-flex h-5 w-5 items-center justify-center rounded-sm transition-colors {activeView ===
+					ChatMessageStatsView.GENERATION
+						? 'bg-background text-foreground shadow-sm'
+						: 'hover:text-foreground'}"
+					onclick={() => (activeView = ChatMessageStatsView.GENERATION)}
+				>
+					<Sparkles class="h-3 w-3" />
+					<span class="sr-only">Generation</span>
+				</button>
+			</Tooltip.Trigger>
+			<Tooltip.Content>
+				<p>Generation (token output)</p>
+			</Tooltip.Content>
+		</Tooltip.Root>
+	</div>
 
-<BadgeChatStatistic icon={Clock} value="{timeInSeconds}s" />
-
-<BadgeChatStatistic icon={Gauge} value="{tokensPerSecond.toFixed(2)} tokens/s" />
+	<div class="flex items-center gap-1 px-2">
+		{#if activeView === ChatMessageStatsView.GENERATION}
+			<BadgeChatStatistic
+				class="bg-transparent"
+				icon={WholeWord}
+				value="{predictedTokens} tokens"
+				tooltipLabel="Generated tokens"
+			/>
+			<BadgeChatStatistic
+				class="bg-transparent"
+				icon={Clock}
+				value="{timeInSeconds}s"
+				tooltipLabel="Generation time"
+			/>
+			<BadgeChatStatistic
+				class="bg-transparent"
+				icon={Gauge}
+				value="{tokensPerSecond.toFixed(2)} tokens/s"
+				tooltipLabel="Generation speed"
+			/>
+		{:else if hasPromptStats}
+			<BadgeChatStatistic
+				class="bg-transparent"
+				icon={WholeWord}
+				value="{promptTokens} tokens"
+				tooltipLabel="Prompt tokens"
+			/>
+			<BadgeChatStatistic
+				class="bg-transparent"
+				icon={Clock}
+				value="{promptTimeInSeconds}s"
+				tooltipLabel="Prompt processing time"
+			/>
+			<BadgeChatStatistic
+				class="bg-transparent"
+				icon={Gauge}
+				value="{promptTokensPerSecond!.toFixed(2)} tokens/s"
+				tooltipLabel="Prompt processing speed"
+			/>
+		{/if}
+	</div>
+</div>

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

@@ -1,5 +1,6 @@
 <script lang="ts">
 	import { BadgeInfo } from '$lib/components/app';
+	import * as Tooltip from '$lib/components/ui/tooltip';
 	import { copyToClipboard } from '$lib/utils';
 	import type { Component } from 'svelte';
 
@@ -7,19 +8,37 @@
 		class?: string;
 		icon: Component;
 		value: string | number;
+		tooltipLabel?: string;
 	}
 
-	let { class: className = '', icon: Icon, value }: Props = $props();
+	let { class: className = '', icon: Icon, value, tooltipLabel }: Props = $props();
 
 	function handleClick() {
 		void copyToClipboard(String(value));
 	}
 </script>
 
-<BadgeInfo class={className} onclick={handleClick}>
-	{#snippet icon()}
-		<Icon class="h-3 w-3" />
-	{/snippet}
+{#if tooltipLabel}
+	<Tooltip.Root>
+		<Tooltip.Trigger>
+			<BadgeInfo class={className} onclick={handleClick}>
+				{#snippet icon()}
+					<Icon class="h-3 w-3" />
+				{/snippet}
 
-	{value}
-</BadgeInfo>
+				{value}
+			</BadgeInfo>
+		</Tooltip.Trigger>
+		<Tooltip.Content>
+			<p>{tooltipLabel}</p>
+		</Tooltip.Content>
+	</Tooltip.Root>
+{:else}
+	<BadgeInfo class={className} onclick={handleClick}>
+		{#snippet icon()}
+			<Icon class="h-3 w-3" />
+		{/snippet}
+
+		{value}
+	</BadgeInfo>
+{/if}

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

@@ -0,0 +1,4 @@
+export enum ChatMessageStatsView {
+	GENERATION = 'generation',
+	READING = 'reading'
+}

tools/server/webui/src/lib/enums/index.ts

@@ -1,5 +1,7 @@
 export { AttachmentType } from './attachment';
 
+export { ChatMessageStatsView } from './chat';
+
 export {
 	FileTypeCategory,
 	FileTypeImage,

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

@@ -171,6 +171,7 @@ class ChatStore {
 	updateProcessingStateFromTimings(
 		timingData: {
 			prompt_n: number;
+			prompt_ms?: number;
 			predicted_n: number;
 			predicted_per_second: number;
 			cache_n: number;
@@ -212,6 +213,7 @@ class ChatStore {
 			if (message.role === 'assistant' && message.timings) {
 				const restoredState = this.parseTimingData({
 					prompt_n: message.timings.prompt_n || 0,
+					prompt_ms: message.timings.prompt_ms,
 					predicted_n: message.timings.predicted_n || 0,
 					predicted_per_second:
 						message.timings.predicted_n && message.timings.predicted_ms
@@ -282,6 +284,7 @@ class ChatStore {
 
 	private parseTimingData(timingData: Record<string, unknown>): ApiProcessingState | null {
 		const promptTokens = (timingData.prompt_n as number) || 0;
+		const promptMs = (timingData.prompt_ms as number) || undefined;
 		const predictedTokens = (timingData.predicted_n as number) || 0;
 		const tokensPerSecond = (timingData.predicted_per_second as number) || 0;
 		const cacheTokens = (timingData.cache_n as number) || 0;
@@ -320,6 +323,7 @@ class ChatStore {
 			speculative: false,
 			progressPercent,
 			promptTokens,
+			promptMs,
 			cacheTokens
 		};
 	}
@@ -536,6 +540,7 @@ class ChatStore {
 					this.updateProcessingStateFromTimings(
 						{
 							prompt_n: timings?.prompt_n || 0,
+							prompt_ms: timings?.prompt_ms,
 							predicted_n: timings?.predicted_n || 0,
 							predicted_per_second: tokensPerSecond,
 							cache_n: timings?.cache_n || 0,
@@ -768,10 +773,11 @@ class ChatStore {
 					content: streamingState.response
 				};
 				if (lastMessage.thinking?.trim()) updateData.thinking = lastMessage.thinking;
-				const lastKnownState = this.getCurrentProcessingStateSync();
+				const lastKnownState = this.getProcessingState(conversationId);
 				if (lastKnownState) {
 					updateData.timings = {
 						prompt_n: lastKnownState.promptTokens || 0,
+						prompt_ms: lastKnownState.promptMs,
 						predicted_n: lastKnownState.tokensDecoded || 0,
 						cache_n: lastKnownState.cacheTokens || 0,
 						predicted_ms:
@@ -1253,6 +1259,7 @@ class ChatStore {
 						this.updateProcessingStateFromTimings(
 							{
 								prompt_n: timings?.prompt_n || 0,
+								prompt_ms: timings?.prompt_ms,
 								predicted_n: timings?.predicted_n || 0,
 								predicted_per_second: tokensPerSecond,
 								cache_n: timings?.cache_n || 0,

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

@@ -342,6 +342,7 @@ export interface ApiProcessingState {
 	// Progress information from prompt_progress
 	progressPercent?: number;
 	promptTokens?: number;
+	promptMs?: number;
 	cacheTokens?: number;
 }