ui-svelte: add prompt processing histogram (#705)

Activities page shows histograms for prompt processing and token generation times. 

Fix: #691
Fix: #703
This commit is contained in:
Benson Wong
2026-04-25 16:13:07 -07:00
committed by GitHub
parent 3cd7837b1f
commit ce28485be2
6 changed files with 117 additions and 52 deletions
+51 -26
View File
@@ -11,57 +11,82 @@
const totalRequests = $metrics.length;
const totalInputTokens = $metrics.reduce((sum, m) => sum + m.input_tokens, 0);
const totalOutputTokens = $metrics.reduce((sum, m) => sum + m.output_tokens, 0);
const totalCacheTokens = $metrics.reduce((sum, m) => sum + m.cache_tokens, 0);
const tokensPerSecond = $metrics
.filter((m) => m.tokens_per_second > 0)
.map((m) => m.tokens_per_second);
const promptPerSecond = $metrics.filter((m) => m.prompt_per_second > 0).map((m) => m.prompt_per_second);
const histogramData = tokensPerSecond.length > 0
? calculateHistogramData(tokensPerSecond, { minBins: 20, maxBins: 80, binScaling: 3 })
: null;
const tokensPerSecond = $metrics.filter((m) => m.tokens_per_second > 0).map((m) => m.tokens_per_second);
const promptHistogramData =
promptPerSecond.length > 0 ? calculateHistogramData(promptPerSecond) : null;
const genHistogramData =
tokensPerSecond.length > 0 ? calculateHistogramData(tokensPerSecond) : null;
return {
totalRequests,
totalInputTokens,
totalOutputTokens,
totalCacheTokens,
inFlightRequests: $inFlightRequests,
histogramData,
promptHistogramData,
genHistogramData,
};
});
</script>
<div class="card">
<div class="card relative p-3">
<button
class="flex items-center gap-1 px-4 pt-3 text-xs font-medium text-gray-500 dark:text-gray-400 hover:text-gray-700 dark:hover:text-gray-200 transition-colors"
onclick={() => $histogramCollapsed = !$histogramCollapsed}
class="absolute top-2 right-2 w-6 h-6 flex items-center justify-center rounded-full border border-gray-300 dark:border-gray-600 text-gray-400 dark:text-gray-500 hover:text-gray-600 dark:hover:text-gray-300 hover:border-gray-400 dark:hover:border-gray-400 transition-colors"
onclick={() => ($histogramCollapsed = !$histogramCollapsed)}
title={$histogramCollapsed ? "Show histograms" : "Hide histograms"}
>
<svg
class="w-3 h-3 transition-transform"
style="transform: rotate({$histogramCollapsed ? -90 : 0}deg)"
viewBox="0 0 16 16"
fill="currentColor"
>
<path d="M4.5 6l3.5 4 3.5-4H4.5z" />
</svg>
Tokens/sec Distribution
{#if $histogramCollapsed}
<svg class="w-3.5 h-3.5" viewBox="0 0 16 16" fill="currentColor">
<path d="M4.5 6l3.5 4 3.5-4H4.5z" />
</svg>
{:else}
<svg class="w-3 h-3" viewBox="0 0 16 16" fill="currentColor">
<path d="M3.5 3.5l9 9M12.5 3.5l-9 9" stroke="currentColor" stroke-width="2" stroke-linecap="round" fill="none" />
</svg>
{/if}
</button>
{#if !$histogramCollapsed}
{#if stats.histogramData}
<TokenHistogram data={stats.histogramData} />
{:else}
<div class="px-4 py-6 text-center text-sm text-gray-500 dark:text-gray-400">
No token speed data yet
<div class="flex flex-col sm:flex-row gap-6 mb-3">
<div class="w-full sm:w-1/2 min-w-0">
<div class="text-sm font-medium text-gray-500 dark:text-gray-400 mb-1">Prompt Processing</div>
{#if stats.promptHistogramData}
<TokenHistogram
data={stats.promptHistogramData}
unit="prompt tokens/sec"
colorClass="text-amber-500 dark:text-amber-400"
/>
{:else}
<div class="py-6 text-center text-sm text-gray-500 dark:text-gray-400">No prompt speed data yet</div>
{/if}
</div>
{/if}
<div class="w-full sm:w-1/2 min-w-0">
<div class="text-sm font-medium text-gray-500 dark:text-gray-400 mb-1">Token Generation</div>
{#if stats.genHistogramData}
<TokenHistogram data={stats.genHistogramData} unit="tokens/sec" />
{:else}
<div class="py-6 text-center text-sm text-gray-500 dark:text-gray-400">No generation speed data yet</div>
{/if}
</div>
</div>
{/if}
<div class="grid grid-cols-3 gap-x-6 gap-y-1 px-4 pb-3 text-sm">
<div class="grid grid-cols-4 gap-x-6 gap-y-1 text-sm">
<div class="text-xs uppercase tracking-wider text-gray-500 dark:text-gray-400">Requests</div>
<div class="text-xs uppercase tracking-wider text-gray-500 dark:text-gray-400">Cached</div>
<div class="text-xs uppercase tracking-wider text-gray-500 dark:text-gray-400">Processed</div>
<div class="text-xs uppercase tracking-wider text-gray-500 dark:text-gray-400">Generated</div>
<div class="text-sm text-gray-700 dark:text-gray-300">
<span class="font-semibold">{nf.format(stats.totalRequests)}</span> completed,
<span class="font-semibold">{nf.format(stats.inFlightRequests)}</span> waiting
</div>
<div class="text-sm text-gray-700 dark:text-gray-300">
<span class="font-semibold">{nf.format(stats.totalCacheTokens)}</span> tokens
</div>
<div class="text-sm text-gray-700 dark:text-gray-300">
<span class="font-semibold">{nf.format(stats.totalInputTokens)}</span> tokens
</div>