fix(frontend): correct Cache Hit Rate denominator to include all prompt tokens

Cache Hit Rate was calculated as cache_read / (cache_read + cache_creation),
which always yields 100% for OpenAI models since cache_creation is never
reported by the OpenAI API. The denominator should include all prompt tokens
(input_tokens + cache_read_tokens + cache_creation_tokens) so the rate
reflects the actual percentage of input tokens served from cache.

Fixes #2291

frontend/src/components/charts/TokenUsageTrend.vue

@@ -109,8 +109,8 @@ const chartData = computed(() => {
       {
         label: 'Cache Hit Rate',
         data: props.trendData.map((d) => {
-          const total = d.cache_read_tokens + d.cache_creation_tokens
-          return total > 0 ? (d.cache_read_tokens / total) * 100 : 0
+          const totalPromptTokens = d.input_tokens + d.cache_read_tokens + d.cache_creation_tokens
+          return totalPromptTokens > 0 ? (d.cache_read_tokens / totalPromptTokens) * 100 : 0
         }),
         borderColor: chartColors.value.cacheHitRate,
         backgroundColor: `${chartColors.value.cacheHitRate}20`,

frontend/src/components/charts/__tests__/TokenUsageTrend.spec.ts

@@ -0,0 +1,120 @@
+import { describe, expect, it, vi } from 'vitest'
+import { mount } from '@vue/test-utils'
+
+import TokenUsageTrend from '../TokenUsageTrend.vue'
+
+const messages: Record<string, string> = {
+  'admin.dashboard.tokenUsageTrend': 'Token Usage Trend',
+  'admin.dashboard.noDataAvailable': 'No data available',
+}
+
+vi.mock('vue-i18n', async () => {
+  const actual = await vi.importActual<typeof import('vue-i18n')>('vue-i18n')
+  return {
+    ...actual,
+    useI18n: () => ({
+      t: (key: string) => messages[key] ?? key,
+    }),
+  }
+})
+
+vi.mock('vue-chartjs', () => ({
+  Line: {
+    props: ['data', 'options'],
+    template: '<div class="chart-data">{{ JSON.stringify(data) }}</div>',
+  },
+}))
+
+describe('TokenUsageTrend', () => {
+  it('calculates cache hit rate against all prompt tokens', () => {
+    const wrapper = mount(TokenUsageTrend, {
+      props: {
+        trendData: [
+          {
+            date: '2026-05-08',
+            requests: 1,
+            input_tokens: 500,
+            output_tokens: 100,
+            cache_creation_tokens: 0,
+            cache_read_tokens: 1500,
+            cost: 0.01,
+            actual_cost: 0.005,
+          },
+        ],
+      },
+      global: {
+        stubs: {
+          LoadingSpinner: true,
+        },
+      },
+    })
+
+    const chartData = JSON.parse(wrapper.find('.chart-data').text())
+    const hitRateDataset = chartData.datasets.find(
+      (ds: any) => ds.label === 'Cache Hit Rate'
+    )
+    // Hit rate = 1500 / (500 + 1500 + 0) * 100 = 75%
+    expect(hitRateDataset.data[0]).toBe(75)
+  })
+
+  it('returns 0 hit rate when all prompt tokens are zero', () => {
+    const wrapper = mount(TokenUsageTrend, {
+      props: {
+        trendData: [
+          {
+            date: '2026-05-08',
+            requests: 0,
+            input_tokens: 0,
+            output_tokens: 0,
+            cache_creation_tokens: 0,
+            cache_read_tokens: 0,
+            cost: 0,
+            actual_cost: 0,
+          },
+        ],
+      },
+      global: {
+        stubs: {
+          LoadingSpinner: true,
+        },
+      },
+    })
+
+    const chartData = JSON.parse(wrapper.find('.chart-data').text())
+    const hitRateDataset = chartData.datasets.find(
+      (ds: any) => ds.label === 'Cache Hit Rate'
+    )
+    expect(hitRateDataset.data[0]).toBe(0)
+  })
+
+  it('includes cache_creation_tokens in denominator for Anthropic models', () => {
+    const wrapper = mount(TokenUsageTrend, {
+      props: {
+        trendData: [
+          {
+            date: '2026-05-08',
+            requests: 1,
+            input_tokens: 200,
+            output_tokens: 50,
+            cache_creation_tokens: 300,
+            cache_read_tokens: 500,
+            cost: 0.02,
+            actual_cost: 0.01,
+          },
+        ],
+      },
+      global: {
+        stubs: {
+          LoadingSpinner: true,
+        },
+      },
+    })
+
+    const chartData = JSON.parse(wrapper.find('.chart-data').text())
+    const hitRateDataset = chartData.datasets.find(
+      (ds: any) => ds.label === 'Cache Hit Rate'
+    )
+    // Hit rate = 500 / (200 + 500 + 300) * 100 = 50%
+    expect(hitRateDataset.data[0]).toBe(50)
+  })
+})