feat: add admin provider analytics dashboard

admin-frontend/src/views/AdminProviders.vue

@@ -29,6 +29,248 @@
         </div>
       </header>
 
+      <BaseCard padding="lg">
+        <div class="flex flex-col gap-5 xl:flex-row xl:items-start xl:justify-between">
+          <div class="max-w-2xl">
+            <div class="flex flex-wrap items-center gap-3">
+              <h2 class="text-xl font-bold text-gray-900">当前环境 Provider 运营摘要</h2>
+              <span class="rounded-full bg-emerald-50 px-3 py-1 text-xs font-medium text-emerald-700">
+                跨用户 / 当前环境
+              </span>
+            </div>
+            <p class="mt-2 text-sm leading-6 text-gray-500">
+              这里展示的是当前部署环境内所有生成任务留下的 Provider 调用轨迹，便于运营和排障。
+              跨环境对比仍需要后续独立汇聚层。
+            </p>
+            <div class="mt-4 flex flex-wrap gap-2">
+              <button
+                type="button"
+                class="rounded-lg border px-3 py-1.5 text-sm transition-colors"
+                :class="analyticsWindow === '7' ? 'border-gray-900 bg-gray-900 text-white' : 'border-gray-200 bg-white text-gray-600 hover:border-gray-400'"
+                @click="analyticsWindow = '7'"
+              >
+                最近 7 天
+              </button>
+              <button
+                type="button"
+                class="rounded-lg border px-3 py-1.5 text-sm transition-colors"
+                :class="analyticsWindow === '30' ? 'border-gray-900 bg-gray-900 text-white' : 'border-gray-200 bg-white text-gray-600 hover:border-gray-400'"
+                @click="analyticsWindow = '30'"
+              >
+                最近 30 天
+              </button>
+              <button
+                type="button"
+                class="rounded-lg border px-3 py-1.5 text-sm transition-colors"
+                :class="analyticsWindow === 'all' ? 'border-gray-900 bg-gray-900 text-white' : 'border-gray-200 bg-white text-gray-600 hover:border-gray-400'"
+                @click="analyticsWindow = 'all'"
+              >
+                全部
+              </button>
+            </div>
+            <div class="mt-3 flex flex-wrap gap-2">
+              <button
+                type="button"
+                class="rounded-lg border px-3 py-1.5 text-sm transition-colors"
+                :class="analyticsCapability === 'all' ? 'border-indigo-600 bg-indigo-600 text-white' : 'border-gray-200 bg-white text-gray-600 hover:border-gray-400'"
+                @click="analyticsCapability = 'all'"
+              >
+                全部能力
+              </button>
+              <button
+                type="button"
+                class="rounded-lg border px-3 py-1.5 text-sm transition-colors"
+                :class="analyticsCapability === 'text' ? 'border-indigo-600 bg-indigo-600 text-white' : 'border-gray-200 bg-white text-gray-600 hover:border-gray-400'"
+                @click="analyticsCapability = 'text'"
+              >
+                文本
+              </button>
+              <button
+                type="button"
+                class="rounded-lg border px-3 py-1.5 text-sm transition-colors"
+                :class="analyticsCapability === 'image' ? 'border-indigo-600 bg-indigo-600 text-white' : 'border-gray-200 bg-white text-gray-600 hover:border-gray-400'"
+                @click="analyticsCapability = 'image'"
+              >
+                图片
+              </button>
+              <button
+                type="button"
+                class="rounded-lg border px-3 py-1.5 text-sm transition-colors"
+                :class="analyticsCapability === 'tts' ? 'border-indigo-600 bg-indigo-600 text-white' : 'border-gray-200 bg-white text-gray-600 hover:border-gray-400'"
+                @click="analyticsCapability = 'tts'"
+              >
+                语音
+              </button>
+              <button
+                type="button"
+                class="rounded-lg border px-3 py-1.5 text-sm transition-colors"
+                :class="analyticsCapability === 'storybook' ? 'border-indigo-600 bg-indigo-600 text-white' : 'border-gray-200 bg-white text-gray-600 hover:border-gray-400'"
+                @click="analyticsCapability = 'storybook'"
+              >
+                绘本
+              </button>
+            </div>
+          </div>
+
+          <div class="grid grid-cols-2 gap-3 sm:grid-cols-4 xl:min-w-[420px]">
+            <div class="rounded-xl border border-gray-100 bg-gray-50 px-4 py-3">
+              <div class="text-xs text-gray-500">活跃用户</div>
+              <div class="mt-1 text-2xl font-semibold text-gray-900">
+                {{ analytics?.user_count ?? 0 }}
+              </div>
+            </div>
+            <div class="rounded-xl border border-gray-100 bg-gray-50 px-4 py-3">
+              <div class="text-xs text-gray-500">总调用</div>
+              <div class="mt-1 text-2xl font-semibold text-gray-900">
+                {{ analytics?.total_calls ?? 0 }}
+              </div>
+            </div>
+            <div class="rounded-xl border border-gray-100 bg-gray-50 px-4 py-3">
+              <div class="text-xs text-gray-500">成功率</div>
+              <div class="mt-1 text-2xl font-semibold text-gray-900">
+                {{ providerSuccessRate ?? '--' }}<span v-if="providerSuccessRate !== null">%</span>
+              </div>
+            </div>
+            <div class="rounded-xl border border-gray-100 bg-gray-50 px-4 py-3">
+              <div class="text-xs text-gray-500">预估成本</div>
+              <div class="mt-1 text-2xl font-semibold text-gray-900">
+                {{ formatCost(analytics?.estimated_cost_usd) }}
+              </div>
+            </div>
+          </div>
+        </div>
+
+        <div v-if="analyticsLoading" class="mt-5 rounded-xl border border-dashed border-gray-200 px-4 py-6 text-sm text-gray-500">
+          正在更新运营摘要...
+        </div>
+        <div v-else-if="analyticsError" class="mt-5 rounded-xl border border-rose-100 bg-rose-50 px-4 py-4 text-sm text-rose-600">
+          {{ analyticsError }}
+        </div>
+        <template v-else-if="analytics">
+          <div class="mt-6 grid grid-cols-2 gap-3 lg:grid-cols-4">
+            <div class="rounded-xl border border-gray-100 bg-white px-4 py-3">
+              <div class="text-xs text-gray-500">覆盖故事</div>
+              <div class="mt-1 text-lg font-semibold text-gray-900">{{ analytics.story_count }}</div>
+            </div>
+            <div class="rounded-xl border border-gray-100 bg-white px-4 py-3">
+              <div class="text-xs text-gray-500">覆盖任务</div>
+              <div class="mt-1 text-lg font-semibold text-gray-900">{{ analytics.job_count }}</div>
+            </div>
+            <div class="rounded-xl border border-gray-100 bg-white px-4 py-3">
+              <div class="text-xs text-gray-500">平均耗时</div>
+              <div class="mt-1 text-lg font-semibold text-gray-900">{{ formatLatency(analytics.avg_latency_ms) }}</div>
+            </div>
+            <div class="rounded-xl border border-gray-100 bg-white px-4 py-3">
+              <div class="text-xs text-gray-500">配置中 Provider</div>
+              <div class="mt-1 text-lg font-semibold text-gray-900">{{ enabledProviderCount }}/{{ providers.length }}</div>
+            </div>
+          </div>
+
+          <div class="mt-6 grid grid-cols-1 gap-6 xl:grid-cols-[minmax(0,2fr)_minmax(0,1fr)]">
+            <div class="rounded-2xl border border-gray-100 bg-white">
+              <div class="flex items-center justify-between border-b border-gray-100 px-5 py-4">
+                <div>
+                  <h3 class="font-semibold text-gray-900">Provider 调用明细</h3>
+                  <p class="mt-1 text-xs text-gray-500">按能力和 adapter 聚合的当前环境视图</p>
+                </div>
+                <span class="text-xs text-gray-400">{{ analyticsProviderRows.length }} 个组合</span>
+              </div>
+              <div class="divide-y divide-gray-100">
+                <div
+                  v-for="row in analyticsProviderRows"
+                  :key="`${row.capability}:${row.adapter}`"
+                  class="grid grid-cols-1 gap-3 px-5 py-4 md:grid-cols-[minmax(0,1.2fr)_repeat(4,minmax(0,0.8fr))]"
+                >
+                  <div>
+                    <div class="flex items-center gap-2">
+                      <span class="rounded-full bg-gray-100 px-2 py-0.5 text-[11px] font-medium text-gray-600">
+                        {{ formatCapability(row.capability) }}
+                      </span>
+                      <span class="font-medium text-gray-900">{{ row.adapter }}</span>
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-[11px] text-gray-400">调用</div>
+                    <div class="mt-1 text-sm font-medium text-gray-900">{{ row.call_count }}</div>
+                  </div>
+                  <div>
+                    <div class="text-[11px] text-gray-400">成功率</div>
+                    <div class="mt-1 text-sm font-medium text-gray-900">
+                      {{ getSuccessRate(row.success_count, row.call_count) ?? '--' }}<span v-if="getSuccessRate(row.success_count, row.call_count) !== null">%</span>
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-[11px] text-gray-400">耗时</div>
+                    <div class="mt-1 text-sm font-medium text-gray-900">{{ formatLatency(row.avg_latency_ms) }}</div>
+                  </div>
+                  <div>
+                    <div class="text-[11px] text-gray-400">成本</div>
+                    <div class="mt-1 text-sm font-medium text-gray-900">{{ formatCost(row.estimated_cost_usd) }}</div>
+                  </div>
+                </div>
+                <div v-if="analyticsProviderRows.length === 0" class="px-5 py-8 text-sm text-gray-500">
+                  当前筛选条件下还没有 Provider 调用样本。
+                </div>
+              </div>
+            </div>
+
+            <div class="space-y-6">
+              <div class="rounded-2xl border border-gray-100 bg-white">
+                <div class="border-b border-gray-100 px-5 py-4">
+                  <h3 class="font-semibold text-gray-900">跨用户分布</h3>
+                  <p class="mt-1 text-xs text-gray-500">按调用量排序，帮助快速识别主要使用者</p>
+                </div>
+                <div class="divide-y divide-gray-100">
+                  <div
+                    v-for="row in analyticsUserRows.slice(0, 5)"
+                    :key="row.user_id"
+                    class="px-5 py-4"
+                  >
+                    <div class="flex items-start justify-between gap-3">
+                      <div>
+                        <div class="text-sm font-medium text-gray-900">{{ row.user_id }}</div>
+                        <div class="mt-1 text-xs text-gray-500">
+                          {{ row.story_count }} 个故事 · {{ row.job_count }} 个任务
+                        </div>
+                      </div>
+                      <div class="text-right">
+                        <div class="text-sm font-semibold text-gray-900">{{ row.call_count }} 次</div>
+                        <div class="mt-1 text-xs text-gray-500">{{ formatCost(row.estimated_cost_usd) }}</div>
+                      </div>
+                    </div>
+                  </div>
+                  <div v-if="analyticsUserRows.length === 0" class="px-5 py-8 text-sm text-gray-500">
+                    当前还没有跨用户样本。
+                  </div>
+                </div>
+              </div>
+
+              <div class="rounded-2xl border border-gray-100 bg-white">
+                <div class="border-b border-gray-100 px-5 py-4">
+                  <h3 class="font-semibold text-gray-900">最近失败原因</h3>
+                  <p class="mt-1 text-xs text-gray-500">先看最常见原因，再决定是否调整路由或优先级</p>
+                </div>
+                <div class="divide-y divide-gray-100">
+                  <div
+                    v-for="reason in analytics.failure_reasons.slice(0, 5)"
+                    :key="reason.reason"
+                    class="flex items-center justify-between gap-3 px-5 py-4"
+                  >
+                    <span class="text-sm text-gray-700">{{ reason.reason }}</span>
+                    <span class="rounded-full bg-rose-50 px-2.5 py-1 text-xs font-medium text-rose-600">
+                      {{ reason.count }} 次
+                    </span>
+                  </div>
+                  <div v-if="analytics.failure_reasons.length === 0" class="px-5 py-8 text-sm text-gray-500">
+                    当前窗口内还没有失败调用。
+                  </div>
+                </div>
+              </div>
+            </div>
+          </div>
+        </template>
+      </BaseCard>
+
       <!-- Main Content Grid -->
       <div class="grid grid-cols-1 lg:grid-cols-4 gap-8">
         
@@ -275,7 +517,7 @@
 </template>
 
 <script setup lang="ts">
-import { ref, computed, onMounted } from 'vue'
+import { computed, onMounted, ref, watch } from 'vue'
 import BaseButton from '../components/ui/BaseButton.vue'
 import BaseCard from '../components/ui/BaseCard.vue'
 import BaseInput from '../components/ui/BaseInput.vue'
@@ -299,7 +541,46 @@ type Provider = {
   config_json: Record<string, any>
 }
 
-// State
+type ProviderAnalyticsBucket = {
+  capability: string
+  adapter: string
+  call_count: number
+  success_count: number
+  failure_count: number
+  avg_latency_ms: number | null
+  estimated_cost_usd: number
+}
+
+type ProviderAnalyticsUserBucket = {
+  user_id: string
+  call_count: number
+  success_count: number
+  failure_count: number
+  job_count: number
+  story_count: number
+  estimated_cost_usd: number
+}
+
+type ProviderAnalyticsResponse = {
+  scope: string
+  window_days: number | null
+  capability: string | null
+  total_calls: number
+  successful_calls: number
+  failed_calls: number
+  avg_latency_ms: number | null
+  estimated_cost_usd: number
+  user_count: number
+  job_count: number
+  story_count: number
+  by_provider: ProviderAnalyticsBucket[]
+  by_user: ProviderAnalyticsUserBucket[]
+  failure_reasons: Array<{
+    reason: string
+    count: number
+  }>
+}
+
 // State
 const loginForm = ref({ username: '', password: '' })
 const loginError = ref('')
@@ -308,6 +589,11 @@ const activeTab = ref('text')
 const providers = ref<Provider[]>([])
 const defaults = ref<Record<string, string[]>>({})
 const availableAdapters = ref<string[]>([])
+const analytics = ref<ProviderAnalyticsResponse | null>(null)
+const analyticsLoading = ref(false)
+const analyticsError = ref('')
+const analyticsWindow = ref<'7' | '30' | 'all'>('30')
+const analyticsCapability = ref<'all' | 'text' | 'image' | 'tts' | 'storybook'>('all')
 const editing = ref(false)
 const form = ref<Partial<Provider> & { api_key?: string; config_json: Record<string, any> }>({
   type: 'text',
@@ -322,6 +608,46 @@ function getAuthHeader(): string {
   return sessionStorage.getItem('admin_auth') || ''
 }
 
+function buildAnalyticsPath() {
+  const params = new URLSearchParams()
+  if (analyticsWindow.value !== 'all') {
+    params.set('days', analyticsWindow.value)
+  }
+  if (analyticsCapability.value !== 'all') {
+    params.set('capability', analyticsCapability.value)
+  }
+  const query = params.toString()
+  return `${apiBase}/admin/providers/analytics${query ? `?${query}` : ''}`
+}
+
+function formatLatency(value?: number | null) {
+  return typeof value === 'number' ? `${Math.round(value)}ms` : '暂无'
+}
+
+function formatCost(value?: number | null) {
+  return typeof value === 'number' ? `$${value.toFixed(4)}` : '$0.0000'
+}
+
+function formatCapability(value: string) {
+  switch (value) {
+    case 'text':
+      return '文本'
+    case 'image':
+      return '图片'
+    case 'tts':
+      return '语音'
+    case 'storybook':
+      return '绘本'
+    default:
+      return value
+  }
+}
+
+function getSuccessRate(successCount: number, totalCount: number) {
+  if (!totalCount) return null
+  return Math.round((successCount / totalCount) * 100)
+}
+
 // Actions
 async function login() {
   loginError.value = ''
@@ -349,24 +675,58 @@ function logout() {
   sessionStorage.removeItem('admin_auth')
   isLoggedIn.value = false
   providers.value = []
+  analytics.value = null
   window.location.reload()
 }
 
+async function loadAnalytics() {
+  if (!isLoggedIn.value) return
+
+  analyticsLoading.value = true
+  analyticsError.value = ''
+  try {
+    const response = await fetch(buildAnalyticsPath(), {
+      headers: { Authorization: getAuthHeader() },
+    })
+    if (!response.ok) {
+      throw new Error('运营数据加载失败')
+    }
+    analytics.value = await response.json()
+  } catch (error) {
+    analyticsError.value = error instanceof Error ? error.message : '运营数据加载失败'
+    analytics.value = null
+  } finally {
+    analyticsLoading.value = false
+  }
+}
+
 async function loadData() {
   if (!isLoggedIn.value) return
 
   const headers = { Authorization: getAuthHeader() }
+  analyticsLoading.value = true
 
-  // Parallel fetch
-  const [pRes, dRes, aRes] = await Promise.all([
-    fetch(`${apiBase}/admin/providers`, { headers }),
-    fetch(`${apiBase}/admin/providers/defaults`, { headers }),
-    fetch(`${apiBase}/admin/providers/adapters`, { headers })
-  ])
+  try {
+    const [pRes, dRes, aRes, analyticsRes] = await Promise.all([
+      fetch(`${apiBase}/admin/providers`, { headers }),
+      fetch(`${apiBase}/admin/providers/defaults`, { headers }),
+      fetch(`${apiBase}/admin/providers/adapters`, { headers }),
+      fetch(buildAnalyticsPath(), { headers }),
+    ])
 
-  if (pRes.ok) providers.value = await pRes.json()
-  if (dRes.ok) defaults.value = await dRes.json()
-  if (aRes.ok) availableAdapters.value = await aRes.json()
+    if (pRes.ok) providers.value = await pRes.json()
+    if (dRes.ok) defaults.value = await dRes.json()
+    if (aRes.ok) availableAdapters.value = await aRes.json()
+    if (analyticsRes.ok) {
+      analytics.value = await analyticsRes.json()
+      analyticsError.value = ''
+    } else {
+      analytics.value = null
+      analyticsError.value = '运营数据加载失败'
+    }
+  } finally {
+    analyticsLoading.value = false
+  }
 }
 
 // Computed
@@ -385,6 +745,37 @@ const adapterOptions = computed(() => {
     })
 })
 
+const enabledProviderCount = computed(() => providers.value.filter((provider) => provider.enabled).length)
+
+const providerSuccessRate = computed(() => {
+  if (!analytics.value?.total_calls) return null
+  return getSuccessRate(analytics.value.successful_calls, analytics.value.total_calls)
+})
+
+const analyticsProviderRows = computed(() => {
+  return [...(analytics.value?.by_provider || [])].sort((left, right) => {
+    if (right.call_count !== left.call_count) {
+      return right.call_count - left.call_count
+    }
+    if (right.estimated_cost_usd !== left.estimated_cost_usd) {
+      return right.estimated_cost_usd - left.estimated_cost_usd
+    }
+    return left.adapter.localeCompare(right.adapter)
+  })
+})
+
+const analyticsUserRows = computed(() => {
+  return [...(analytics.value?.by_user || [])].sort((left, right) => {
+    if (right.call_count !== left.call_count) {
+      return right.call_count - left.call_count
+    }
+    if (right.estimated_cost_usd !== left.estimated_cost_usd) {
+      return right.estimated_cost_usd - left.estimated_cost_usd
+    }
+    return left.user_id.localeCompare(right.user_id)
+  })
+})
+
 // UI Actions
 function cloneDefault(type: string, name: string) {
   // 根据默认名称推断配置
@@ -439,9 +830,9 @@ async function submit() {
     method,
     headers: {
       'Content-Type': 'application/json',
-      Authorization: getAuthHeader()
+      Authorization: getAuthHeader(),
     },
-    body: JSON.stringify(form.value)
+    body: JSON.stringify(form.value),
   })
 
   await loadData()
@@ -449,10 +840,10 @@ async function submit() {
 }
 
 async function remove(p: Provider) {
-  if(!confirm(`确认删除 ${p.name}?`)) return
+  if (!confirm(`确认删除 ${p.name}?`)) return
   await fetch(`${apiBase}/admin/providers/${p.id}`, {
     method: 'DELETE',
-    headers: { Authorization: getAuthHeader() }
+    headers: { Authorization: getAuthHeader() },
   })
   await loadData()
 }
@@ -462,9 +853,9 @@ async function toggleEnabled(p: Provider) {
     method: 'PUT',
     headers: {
       'Content-Type': 'application/json',
-      Authorization: getAuthHeader()
+      Authorization: getAuthHeader(),
     },
-    body: JSON.stringify({ enabled: !p.enabled })
+    body: JSON.stringify({ enabled: !p.enabled }),
   })
   await loadData()
 }
@@ -472,4 +863,10 @@ async function toggleEnabled(p: Provider) {
 onMounted(() => {
   if (isLoggedIn.value) loadData()
 })
+
+watch([analyticsWindow, analyticsCapability], () => {
+  if (isLoggedIn.value) {
+    void loadAnalytics()
+  }
+})
 </script>

backend/app/api/admin_providers.py

@@ -1,4 +1,4 @@
-from fastapi import APIRouter, Depends, HTTPException
+from fastapi import APIRouter, Depends, HTTPException, Query
 from pydantic import BaseModel, ConfigDict, Field
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
@@ -8,6 +8,7 @@ from app.db.admin_models import Provider
 from app.db.database import get_db
 from app.services.adapters.registry import AdapterRegistry
 from app.services.cost_tracker import cost_tracker
+from app.services.generation_jobs import get_admin_provider_analytics
 from app.services.provider_policy import DEFAULT_PROVIDERS, list_capability_policies
 from app.services.secret_service import SecretService
 
@@ -56,6 +57,48 @@ class ProviderResponse(BaseModel):
 
     model_config = ConfigDict(from_attributes=True)
 
+
+class ProviderAnalyticsBucket(BaseModel):
+    capability: str
+    adapter: str
+    call_count: int
+    success_count: int
+    failure_count: int
+    avg_latency_ms: float | None = None
+    estimated_cost_usd: float
+
+
+class ProviderAnalyticsUserBucket(BaseModel):
+    user_id: str
+    call_count: int
+    success_count: int
+    failure_count: int
+    job_count: int
+    story_count: int
+    estimated_cost_usd: float
+
+
+class ProviderAnalyticsFailureReason(BaseModel):
+    reason: str
+    count: int
+
+
+class ProviderAnalyticsResponse(BaseModel):
+    scope: str
+    window_days: int | None = None
+    capability: str | None = None
+    total_calls: int
+    successful_calls: int
+    failed_calls: int
+    avg_latency_ms: float | None = None
+    estimated_cost_usd: float
+    user_count: int
+    job_count: int
+    story_count: int
+    by_provider: list[ProviderAnalyticsBucket]
+    by_user: list[ProviderAnalyticsUserBucket]
+    failure_reasons: list[ProviderAnalyticsFailureReason]
+
 @router.get("/providers/adapters")
 async def list_available_adapters():
     """获取所有可用的适配器类型 (定义的类)。"""
@@ -74,6 +117,20 @@ async def list_provider_capabilities():
     return list_capability_policies()
 
 
+@router.get("/providers/analytics", response_model=ProviderAnalyticsResponse)
+async def get_provider_analytics(
+    days: int | None = Query(default=None, ge=1, le=365),
+    capability: str | None = Query(default=None),
+    db: AsyncSession = Depends(get_db),
+):
+    """获取当前环境跨用户的 Provider 运营摘要。"""
+    return await get_admin_provider_analytics(
+        db,
+        days=days,
+        capability=capability,
+    )
+
+
 @router.get("/providers", response_model=list[ProviderResponse])
 async def list_providers(db: AsyncSession = Depends(get_db)):
     result = await db.execute(select(Provider))

backend/app/services/generation_jobs.py

@@ -606,23 +606,37 @@ def _aggregate_provider_events(
     }
 
 
+def _event_matches_capability(
+    event: GenerationJobEvent,
+    capability: str | None = None,
+) -> bool:
+    event_capability = str((event.event_metadata or {}).get("capability") or "unknown")
+    return capability is None or event_capability == capability
+
+
 def _provider_events_query(
     *,
-    user_id: str,
+    user_id: str | None = None,
     story_id: int | None = None,
     days: int | None = None,
 ):
     query = (
-        select(GenerationJobEvent)
+        select(
+            GenerationJobEvent,
+            GenerationJob.user_id,
+            GenerationJob.story_id,
+        )
         .join(GenerationJob, GenerationJobEvent.job_id == GenerationJob.id)
         .where(
-            GenerationJob.user_id == user_id,
             GenerationJobEvent.event_type.in_(
                 ["provider_call_succeeded", "provider_call_failed"]
             ),
         )
     )
 
+    if user_id is not None:
+        query = query.where(GenerationJob.user_id == user_id)
+
     if story_id is not None:
         query = query.where(GenerationJob.story_id == story_id)
 
@@ -681,17 +695,12 @@ async def get_user_provider_analytics(
     filtered_event_job_ids = {
         event.job_id
         for event in events
-        if capability is None
-        or str((event.event_metadata or {}).get("capability") or "unknown") == capability
+        if _event_matches_capability(event, capability)
     }
     filtered_story_ids = {
         event.story_id
         for event in events
-        if event.story_id is not None
-        and (
-            capability is None
-            or str((event.event_metadata or {}).get("capability") or "unknown") == capability
-        )
+        if event.story_id is not None and _event_matches_capability(event, capability)
     }
 
     return {
@@ -703,6 +712,87 @@ async def get_user_provider_analytics(
     }
 
 
+async def get_admin_provider_analytics(
+    db: AsyncSession,
+    *,
+    days: int | None = None,
+    capability: str | None = None,
+) -> dict[str, Any]:
+    """Aggregate provider telemetry across every user in the current environment."""
+
+    rows = (await db.execute(_provider_events_query(days=days))).all()
+    events = [event for event, _, _ in rows]
+    filtered_rows = [
+        (event, user_id, story_id)
+        for event, user_id, story_id in rows
+        if _event_matches_capability(event, capability)
+    ]
+
+    by_user: dict[str, dict[str, Any]] = {}
+    filtered_job_ids = {event.job_id for event, _, _ in filtered_rows}
+    filtered_story_ids = {
+        story_id for _, _, story_id in filtered_rows if story_id is not None
+    }
+    filtered_user_ids = {user_id for _, user_id, _ in filtered_rows}
+
+    for event, user_id, story_id in filtered_rows:
+        bucket = by_user.setdefault(
+            user_id,
+            {
+                "user_id": user_id,
+                "call_count": 0,
+                "success_count": 0,
+                "failure_count": 0,
+                "estimated_cost_usd": 0.0,
+                "job_ids": set(),
+                "story_ids": set(),
+            },
+        )
+        bucket["call_count"] += 1
+        bucket["job_ids"].add(event.job_id)
+        if story_id is not None:
+            bucket["story_ids"].add(story_id)
+
+        if event.event_type == "provider_call_succeeded":
+            bucket["success_count"] += 1
+            bucket["estimated_cost_usd"] += (
+                _as_float((event.event_metadata or {}).get("estimated_cost_usd")) or 0.0
+            )
+        else:
+            bucket["failure_count"] += 1
+
+    serialized_users = [
+        {
+            "user_id": user_id,
+            "call_count": bucket["call_count"],
+            "success_count": bucket["success_count"],
+            "failure_count": bucket["failure_count"],
+            "job_count": len(bucket["job_ids"]),
+            "story_count": len(bucket["story_ids"]),
+            "estimated_cost_usd": round(bucket["estimated_cost_usd"], 6),
+        }
+        for user_id, bucket in by_user.items()
+    ]
+    serialized_users.sort(
+        key=lambda item: (
+            -int(item["call_count"]),
+            -float(item["estimated_cost_usd"]),
+            str(item["user_id"]),
+        )
+    )
+
+    return {
+        "scope": "current_environment",
+        "window_days": days,
+        "capability": capability,
+        **_aggregate_provider_events(events, capability=capability),
+        "user_count": len(filtered_user_ids),
+        "job_count": len(filtered_job_ids),
+        "story_count": len(filtered_story_ids),
+        "by_user": serialized_users,
+    }
+
+
 async def get_user_generation_ops_summary(
     db: AsyncSession,
     *,

backend/tests/test_admin_providers.py

@@ -0,0 +1,285 @@
+from datetime import datetime, timedelta, timezone
+
+from fastapi import FastAPI
+from httpx import ASGITransport, AsyncClient
+
+from app.api import admin_providers
+from app.core.admin_auth import admin_guard
+from app.db.database import get_db
+from app.db.models import Story, User
+from app.services.generation_jobs import create_generation_job, record_generation_event
+
+
+def _build_admin_test_app(db_session) -> FastAPI:
+    app = FastAPI()
+    app.include_router(admin_providers.router, prefix="/admin")
+
+    async def override_get_db():
+        yield db_session
+
+    async def override_admin_guard():
+        return True
+
+    app.dependency_overrides[get_db] = override_get_db
+    app.dependency_overrides[admin_guard] = override_admin_guard
+    return app
+
+
+async def _create_story(
+    db_session,
+    *,
+    user_id: str,
+    title: str,
+    mode: str = "generated",
+) -> Story:
+    story = Story(
+        user_id=user_id,
+        title=title,
+        story_text="测试内容",
+        cover_prompt="A gentle moonlit forest",
+        mode=mode,
+        generation_status="partial_ready",
+        text_status="ready",
+        image_status="not_requested",
+        audio_status="not_requested",
+    )
+    db_session.add(story)
+    await db_session.commit()
+    await db_session.refresh(story)
+    return story
+
+
+async def test_admin_provider_analytics_aggregate_across_users(db_session, test_user):
+    second_user = User(
+        id="github:67890",
+        name="Another User",
+        avatar_url="https://example.com/avatar-2.png",
+        provider="github",
+    )
+    db_session.add(second_user)
+    await db_session.commit()
+
+    first_story = await _create_story(db_session, user_id=test_user.id, title="第一位用户的故事")
+    second_story = await _create_story(
+        db_session,
+        user_id=second_user.id,
+        title="第二位用户的故事",
+    )
+
+    image_job = await create_generation_job(
+        db_session,
+        user_id=test_user.id,
+        output_mode="asset_retry",
+        input_type="image",
+        request_payload={"assets": ["image"]},
+        story_id=first_story.id,
+    )
+    await record_generation_event(
+        db_session,
+        job=image_job,
+        story_id=first_story.id,
+        event_type="provider_call_succeeded",
+        status="succeeded",
+        metadata={
+            "capability": "image",
+            "adapter": "demo",
+            "strategy": "priority",
+            "latency_ms": 42,
+            "estimated_cost_usd": 0.01,
+        },
+    )
+    await record_generation_event(
+        db_session,
+        job=image_job,
+        story_id=first_story.id,
+        event_type="provider_call_failed",
+        status="failed",
+        metadata={
+            "capability": "image",
+            "adapter": "cqtai",
+            "strategy": "priority",
+            "latency_ms": 120,
+            "error": "timeout",
+        },
+    )
+
+    audio_job = await create_generation_job(
+        db_session,
+        user_id=second_user.id,
+        output_mode="asset_retry",
+        input_type="audio",
+        request_payload={"assets": ["audio"]},
+        story_id=second_story.id,
+    )
+    await record_generation_event(
+        db_session,
+        job=audio_job,
+        story_id=second_story.id,
+        event_type="provider_call_succeeded",
+        status="succeeded",
+        metadata={
+            "capability": "tts",
+            "adapter": "edge_tts",
+            "strategy": "priority",
+            "latency_ms": 18,
+            "estimated_cost_usd": 0.003,
+        },
+    )
+
+    admin_app = _build_admin_test_app(db_session)
+    transport = ASGITransport(app=admin_app)
+
+    async with AsyncClient(transport=transport, base_url="http://test") as client:
+        response = await client.get("/admin/providers/analytics")
+
+    assert response.status_code == 200
+    data = response.json()
+    assert data["scope"] == "current_environment"
+    assert data["user_count"] == 2
+    assert data["job_count"] == 2
+    assert data["story_count"] == 2
+    assert data["total_calls"] == 3
+    assert data["successful_calls"] == 2
+    assert data["failed_calls"] == 1
+    assert data["avg_latency_ms"] == 60.0
+    assert data["estimated_cost_usd"] == 0.013
+    assert data["failure_reasons"] == [{"reason": "timeout", "count": 1}]
+    assert data["by_provider"] == [
+        {
+            "capability": "image",
+            "adapter": "cqtai",
+            "call_count": 1,
+            "success_count": 0,
+            "failure_count": 1,
+            "avg_latency_ms": 120.0,
+            "estimated_cost_usd": 0.0,
+        },
+        {
+            "capability": "image",
+            "adapter": "demo",
+            "call_count": 1,
+            "success_count": 1,
+            "failure_count": 0,
+            "avg_latency_ms": 42.0,
+            "estimated_cost_usd": 0.01,
+        },
+        {
+            "capability": "tts",
+            "adapter": "edge_tts",
+            "call_count": 1,
+            "success_count": 1,
+            "failure_count": 0,
+            "avg_latency_ms": 18.0,
+            "estimated_cost_usd": 0.003,
+        },
+    ]
+    assert data["by_user"] == [
+        {
+            "user_id": test_user.id,
+            "call_count": 2,
+            "success_count": 1,
+            "failure_count": 1,
+            "job_count": 1,
+            "story_count": 1,
+            "estimated_cost_usd": 0.01,
+        },
+        {
+            "user_id": second_user.id,
+            "call_count": 1,
+            "success_count": 1,
+            "failure_count": 0,
+            "job_count": 1,
+            "story_count": 1,
+            "estimated_cost_usd": 0.003,
+        },
+    ]
+
+
+async def test_admin_provider_analytics_support_days_and_capability_filters(
+    db_session,
+    test_user,
+):
+    second_user = User(
+        id="google:22222",
+        name="Filter User",
+        avatar_url="https://example.com/avatar-3.png",
+        provider="google",
+    )
+    db_session.add(second_user)
+    await db_session.commit()
+
+    first_story = await _create_story(db_session, user_id=test_user.id, title="旧事件故事")
+    second_story = await _create_story(db_session, user_id=second_user.id, title="最近事件故事")
+
+    image_job = await create_generation_job(
+        db_session,
+        user_id=test_user.id,
+        output_mode="asset_retry",
+        input_type="image",
+        request_payload={"assets": ["image"]},
+        story_id=first_story.id,
+    )
+    old_event = await record_generation_event(
+        db_session,
+        job=image_job,
+        story_id=first_story.id,
+        event_type="provider_call_failed",
+        status="failed",
+        metadata={
+            "capability": "image",
+            "adapter": "cqtai",
+            "strategy": "priority",
+            "latency_ms": 120,
+            "error": "timeout",
+        },
+    )
+    old_event.created_at = datetime.now(timezone.utc) - timedelta(days=10)
+    await db_session.commit()
+
+    audio_job = await create_generation_job(
+        db_session,
+        user_id=second_user.id,
+        output_mode="asset_retry",
+        input_type="audio",
+        request_payload={"assets": ["audio"]},
+        story_id=second_story.id,
+    )
+    await record_generation_event(
+        db_session,
+        job=audio_job,
+        story_id=second_story.id,
+        event_type="provider_call_succeeded",
+        status="succeeded",
+        metadata={
+            "capability": "tts",
+            "adapter": "edge_tts",
+            "strategy": "priority",
+            "latency_ms": 18,
+            "estimated_cost_usd": 0.003,
+        },
+    )
+
+    admin_app = _build_admin_test_app(db_session)
+    transport = ASGITransport(app=admin_app)
+
+    async with AsyncClient(transport=transport, base_url="http://test") as client:
+        response = await client.get("/admin/providers/analytics?days=7")
+        assert response.status_code == 200
+        data = response.json()
+        assert data["window_days"] == 7
+        assert data["total_calls"] == 1
+        assert data["user_count"] == 1
+        assert data["job_count"] == 1
+        assert data["story_count"] == 1
+        assert data["failure_reasons"] == []
+
+        response = await client.get("/admin/providers/analytics?capability=image")
+        assert response.status_code == 200
+        data = response.json()
+        assert data["capability"] == "image"
+        assert data["total_calls"] == 1
+        assert data["failed_calls"] == 1
+        assert data["user_count"] == 1
+        assert data["job_count"] == 1
+        assert data["story_count"] == 1
+        assert data["failure_reasons"] == [{"reason": "timeout", "count": 1}]

docs/planning/demo-checklist.md

@@ -126,7 +126,7 @@ DreamWeaver 是面向 3-8 岁亲子场景的个性化 AI 绘本与陪伴式讲
 
 ### 2:20 - 3:00 取舍与下一步
 
-求职版优先稳定闭环和可解释性，不做支付、多租户和复杂监控。现在 job/event 已能查询 workflow、资产补全、provider 调用轨迹和聚合指标，用户端和管理端也能展示生成轨迹与跨故事 Provider 运营摘要；统一生成也已经迁移到后台 worker，下一步是补取消/重试队列。
+求职版优先稳定闭环和可解释性，不做支付、多租户和复杂监控。现在 job/event 已能查询 workflow、资产补全、provider 调用轨迹和聚合指标，统一生成已迁移到后台 worker，取消/重试队列也已打通；用户端可看跨故事运营摘要，管理端可看当前环境跨用户 Provider dashboard。下一步应补跨环境汇聚、断点续跑和更完整监控。
 
 ---
 

docs/planning/demo-package.md

@@ -51,7 +51,7 @@ SMOKE_AUDIO=1 ./scripts/demo_smoke.sh
 - **AI 不确定性处理**：主内容和资产拆开，图片/音频失败不阻塞阅读。
 - **Provider 产品化**：用户看到稳定能力，系统内部用 Capability / Provider / Adapter / Routing Policy 管供应链。
 - **可观测性**：generation job/event 让生成过程、失败恢复和 Provider 成本可解释。
-- **可继续生产化**：统一生成已经迁移到 worker，前端轮询和任务事件模型也已打通，下一步是补取消/重试队列和更完整监控。
+- **可继续生产化**：统一生成已迁移到 worker，前端轮询、任务事件模型、取消/重试队列和管理台当前环境 dashboard 也已打通，下一步是补跨环境汇聚、断点续跑和更完整监控。
 
 ---
 
@@ -63,4 +63,4 @@ SMOKE_AUDIO=1 ./scripts/demo_smoke.sh
 | 图片生成失败 | 展示 `degraded_completed` 与资源重试 |
 | Docker 冷启动慢 | 演示前先跑 smoke 并保持容器运行 |
 | Provider 追问过深 | 回到 Capability / Provider / Adapter / Routing Policy 四层解释 |
-| 生产化追问 | 说明下一步是取消/重试队列、监控告警、密钥治理和 Provider analytics 扩展 |
+| 生产化追问 | 说明下一步是跨环境 Provider 汇聚、断点续跑、监控告警和密钥治理 |

docs/planning/interview-pitch.md

@@ -83,4 +83,4 @@ AI 生成产品最大的问题不是“能不能调模型”，而是结果不
 
 ### 这个项目下一步怎么上线？
 
-我已经把当前轻量 job/event 模型迁移到后台 worker，并打通了前端进度轮询；下一步会补取消/重试队列，再继续扩展跨时间窗口和跨用户维度的 provider 运营分析。生产上线前还需要补真实用户鉴权配置、密钥管理、监控告警和部署策略。
+我已经把当前轻量 job/event 模型迁移到后台 worker，并打通了前端进度轮询、取消/重试队列和管理台当前环境运营视图；下一步会补跨环境 Provider 汇聚、断点续跑和更完整监控。生产上线前还需要补真实用户鉴权配置、密钥管理和部署策略。

docs/planning/week-2-to-4-execution-backlog.md

@@ -72,6 +72,7 @@ Week 2 已完成演示闭环、统一生成工作流、generation job/event、
 | W4-09 | Workflow | 卡住任务自动收敛 | `GENERATION_JOB_STALE_MINUTES` + Celery beat stale job maintenance | P1 | Done |
 | W4-10 | Workflow | 防止重复资产任务 | 运行中故事拒绝重复封面/音频/资产重试请求 | P1 | Done |
 | W4-11 | Workflow | 生成任务取消与重新排队 | 取消已提交任务，失败/取消任务可重新排队 | P1 | Done |
+| W4-12 | Ops | 管理台当前环境跨用户 Provider dashboard | `GET /admin/providers/analytics` + admin console 运营摘要面板 | P1 | Done |
 
 ---
 

docs/planning/week-4-sprint-review.md

@@ -40,6 +40,7 @@ DreamWeaver 已经具备求职演示所需的完整闭环：
 - Provider failover 和聚合指标
 - 跨故事 Provider analytics
 - 任务运行概览、最近失败摘要与卡住任务收敛
+- 当前环境跨用户 Provider dashboard
 - 前端生成轨迹和自动轮询形态
 
 ---
@@ -61,7 +62,7 @@ DreamWeaver 已经具备求职演示所需的完整闭环：
 
 | Priority | Task | Why |
 | --- | --- | --- |
-| P0 | 跨用户 / 跨环境 Provider dashboard | 当前已支持单用户摘要，后续要支持运营视角 |
+| P0 | 跨环境 Provider dashboard | 当前环境跨用户摘要已落地，后续需要多部署汇聚视角 |
 | P1 | 监控告警与结构化 dashboard | 目前已有故事库级概览，后续要接入更完整观测体系 |
 | P1 | 断点续跑与更细粒度任务控制 | 让取消、重试和 worker 恢复更稳 |
 | P2 | 更细粒度叙事风格与音色策略 | 扩展体验，但不影响当前求职版主线 |

docs/product/unified-generation-workflow-prd.md

@@ -77,16 +77,20 @@ DreamWeaver 当前同时支持普通故事生成、完整故事生成和绘本
   - `POST /api/generations/jobs/{job_id}/cancel`
   - `POST /api/generations/jobs/{job_id}/retry`
   - 创建弹窗与生成轨迹都可触发取消或重新排队
+- 管理台已补当前环境的跨用户 Provider dashboard：
+  - `GET /admin/providers/analytics`
+  - 支持 `days` / `capability` 筛选
+  - 可查看跨用户调用量、成功率、平均耗时、预估成本、失败原因和 Top 用户分布
 
 ### Remaining Production Work
 
 - 普通故事、完整生成、绘本生成已有统一外部入口，内部 workflow 仍可继续减少兼容层分支
 - 统一资产重试入口已覆盖普通故事封面、绘本缺失插图和故事音频，后续可继续扩展更细的资产级审计
-- 断点续跑、跨用户/跨环境 Provider 分析，以及更细粒度的任务控制策略仍属于后续生产化增强
+- 断点续跑、跨环境 Provider 汇聚视图，以及更细粒度的任务控制策略仍属于后续生产化增强
 
 ### What This Means
 
-这份 PRD 仍然保留目标态设计，但主干能力已经可在当前代码中演示。当前最适合的继续方式，是在已落地的 worker 化与任务控制基础上，把当前首版运营摘要扩展为可筛选、可对比的分析视角，并逐步补断点续跑和更完整监控，而不是继续扩大功能范围。
+这份 PRD 仍然保留目标态设计，但主干能力已经可在当前代码中演示。当前最适合的继续方式，是在已落地的 worker 化、任务控制和当前环境运营 dashboard 基础上，继续补断点续跑、跨环境汇聚和更完整监控，而不是继续扩大功能范围。
 
 ---
 
@@ -97,7 +101,7 @@ DreamWeaver 当前同时支持普通故事生成、完整故事生成和绘本
 DreamWeaver 当前存在以下工作流层面问题：
 
 1. **生成入口已建立，内部路径正在收束**
-   当前前端已切到 `/api/generations`，旧的 `/api/stories/generate`、`/api/stories/generate/full`、`/api/storybook/generate` 仍作为兼容入口保留。service 内部已抽取上下文准备、主记录保存、封面补全、绘本插图补全和音频补全 helper，并用 `AssetCompletionResult` 表达资产补全结果。generation job/event 已落库并可查询，Provider 调用轨迹、单故事聚合指标和跨故事运营摘要也已进入用户端与管理端展示；统一生成请求现在已经交给后台 worker 执行。下一步重点是把取消/重试队列也接到这套事件模型上。
+   当前前端已切到 `/api/generations`，旧的 `/api/stories/generate`、`/api/stories/generate/full`、`/api/storybook/generate` 仍作为兼容入口保留。service 内部已抽取上下文准备、主记录保存、封面补全、绘本插图补全和音频补全 helper，并用 `AssetCompletionResult` 表达资产补全结果。generation job/event 已落库并可查询，Provider 调用轨迹、单故事聚合指标、跨故事运营摘要和管理台跨用户 dashboard 都已进入前端展示；统一生成请求现在已经交给后台 worker 执行。下一步重点应转向断点续跑和跨环境汇聚，而不是再补一套新的入口。
 
 2. **保存与资产补全过程正在统一**
    文本故事和绘本已拥有更清晰的主记录保存 helper；普通故事封面、绘本缺失插图、故事音频生成/缓存已共用各自的 asset completion helper。服务层已经能表达资产任务结果，并会把统一入口、资产重试、绘本逐页插图和音频生成的关键节点写入 job event。

docs/technical/architecture.md

@@ -100,7 +100,7 @@ flowchart LR
 
 当前仍是求职版 MVP，不引入复杂工作流引擎。下一步生产化优先级：
 
-1. 补齐生成任务取消与重新排队能力，减少误触和重复消耗。
-2. 基于现有 job 查询和前端轮询继续扩展真实异步进度与任务控制。
-3. 扩展 Provider analytics 的时间窗口、失败原因和跨用户维度。
-4. 继续补充部署、监控告警和密钥治理策略。
+1. 补跨环境 Provider 汇聚视图，避免每个部署环境各自成孤岛。
+2. 基于现有 job 查询和前端轮询继续扩展断点续跑与更细粒度任务控制。
+3. 在当前环境 dashboard 基础上继续扩展失败原因、监控告警和结构化观测能力。
+4. 继续补充部署与密钥治理策略。

docs/technical/generation-job-state.md

@@ -61,7 +61,7 @@ job 响应会返回 `progress_percent`、`progress_label` 和 `is_terminal`，
 当前已有两层记录，未来可以继续扩展字段和事件颗粒度：
 
 - 继续复用现有 job 查询和前端轮询进度条，为取消请求、重新排队和长任务通知提供统一入口。
-- 将当前跨故事 provider 指标扩展为跨用户、跨环境和更细颗粒度的失败原因维度分析。
+- 当前环境的跨用户 provider dashboard 已在 admin 端落地，下一步应补跨环境汇聚和更细颗粒度的失败原因维度分析。
 
 ## 面试表达