refactor: add scheduling mode support and optimize system settings UI

- Add fixed_order and cache_affinity scheduling modes to CacheAwareScheduler
- Only apply cache affinity in cache_affinity mode; use fixed order otherwise
- Simplify Dialog components with title/description props
- Remove unnecessary button shadows in SystemSettings
- Optimize import dialog UI structure
- Update ModelAliasesTab shadow styling
- Fix fallback orchestrator type hints
- Add scheduling_mode configuration in system config

frontend/src/components/ui/button.vue

@@ -34,11 +34,10 @@ const buttonClass = computed(() => {
     'inline-flex items-center justify-center rounded-xl text-sm font-semibold transition-all duration-200 ring-offset-background focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-2 disabled:pointer-events-none disabled:opacity-50 active:scale-[0.98]'
 
   const variantClasses = {
-    default:
-      'bg-primary text-white shadow-[0_20px_35px_rgba(204,120,92,0.35)] hover:bg-primary/90 hover:shadow-[0_25px_45px_rgba(204,120,92,0.45)]',
-    destructive: 'bg-destructive text-destructive-foreground hover:bg-destructive/85 shadow-sm',
+    default: 'bg-primary text-white hover:bg-primary/90',
+    destructive: 'bg-destructive text-destructive-foreground hover:bg-destructive/85',
     outline:
-      'border border-border/60 bg-card/60 text-foreground hover:border-primary/60 hover:text-primary hover:bg-primary/10 shadow-sm backdrop-blur transition-all',
+      'border border-border/60 bg-card/60 text-foreground hover:border-primary/60 hover:text-primary hover:bg-primary/10 backdrop-blur transition-all',
     secondary:
       'bg-secondary text-secondary-foreground shadow-inner hover:bg-secondary/80',
     ghost: 'hover:bg-accent hover:text-accent-foreground',

frontend/src/components/ui/dialog/Dialog.vue

@@ -2,7 +2,7 @@
   <Teleport to="body">
     <div
       v-if="isOpen"
-      class="fixed inset-0 overflow-y-auto"
+      class="fixed inset-0 overflow-y-auto pointer-events-none"
       :style="{ zIndex: containerZIndex }"
     >
       <!-- 背景遮罩 -->
@@ -16,7 +16,7 @@
       >
         <div
           v-if="isOpen"
-          class="fixed inset-0 bg-black/40 backdrop-blur-sm transition-opacity"
+          class="fixed inset-0 bg-black/40 backdrop-blur-sm transition-opacity pointer-events-auto"
           :style="{ zIndex: backdropZIndex }"
           @click="handleClose"
         />
@@ -34,7 +34,7 @@
         >
           <div
             v-if="isOpen"
-            class="relative transform rounded-lg bg-background text-left shadow-2xl transition-all sm:my-8 sm:w-full border border-border"
+            class="relative transform rounded-lg bg-background text-left shadow-2xl transition-all sm:my-8 sm:w-full border border-border pointer-events-auto"
             :style="{ zIndex: contentZIndex }"
             :class="maxWidthClass"
             @click.stop

frontend/src/components/ui/select-content.vue

@@ -45,7 +45,7 @@ const props = withDefaults(defineProps<Props>(), {
 
 const contentClass = computed(() =>
   cn(
-    'z-[100] max-h-96 min-w-[8rem] overflow-hidden rounded-2xl border border-border bg-card text-foreground shadow-2xl backdrop-blur-xl pointer-events-auto',
+    'z-[200] max-h-96 min-w-[8rem] overflow-hidden rounded-2xl border border-border bg-card text-foreground shadow-2xl backdrop-blur-xl pointer-events-auto',
     'data-[state=open]:animate-in data-[state=closed]:animate-out data-[state=closed]:fade-out-0 data-[state=open]:fade-in-0 data-[state=closed]:zoom-out-95 data-[state=open]:zoom-in-95',
     'data-[side=bottom]:slide-in-from-top-2 data-[side=left]:slide-in-from-right-2 data-[side=right]:slide-in-from-left-2 data-[side=top]:slide-in-from-bottom-2',
     props.class

frontend/src/features/providers/components/PriorityManagementDialog.vue

@@ -312,9 +312,42 @@
 
     <template #footer>
       <div class="flex items-center justify-between w-full">
+        <div class="flex items-center gap-4">
           <div class="text-xs text-muted-foreground">
             当前模式: <span class="font-medium">{{ activeMainTab === 'provider' ? '提供商优先' : 'Key 优先' }}</span>
           </div>
+          <div class="flex items-center gap-2 pl-4 border-l border-border">
+            <span class="text-xs text-muted-foreground">调度:</span>
+            <div class="flex gap-0.5 p-0.5 bg-muted/40 rounded-md">
+              <button
+                type="button"
+                class="px-2 py-1 text-xs font-medium rounded transition-all"
+                :class="[
+                  schedulingMode === 'fixed_order'
+                    ? 'bg-primary text-primary-foreground shadow-sm'
+                    : 'text-muted-foreground hover:text-foreground hover:bg-muted/50'
+                ]"
+                title="严格按优先级顺序，不考虑缓存"
+                @click="schedulingMode = 'fixed_order'"
+              >
+                固定顺序
+              </button>
+              <button
+                type="button"
+                class="px-2 py-1 text-xs font-medium rounded transition-all"
+                :class="[
+                  schedulingMode === 'cache_affinity'
+                    ? 'bg-primary text-primary-foreground shadow-sm'
+                    : 'text-muted-foreground hover:text-foreground hover:bg-muted/50'
+                ]"
+                title="优先使用已缓存的Provider，利用Prompt Cache"
+                @click="schedulingMode = 'cache_affinity'"
+              >
+                缓存亲和
+              </button>
+            </div>
+          </div>
+        </div>
         <div class="flex gap-2">
           <Button
             size="sm"
@@ -410,6 +443,9 @@ const saving = ref(false)
 // Key 优先级编辑状态
 const editingKeyPriority = ref<Record<string, string | null>>({})  // format -> keyId
 
+// 调度模式状态
+const schedulingMode = ref<'fixed_order' | 'cache_affinity'>('cache_affinity')
+
 // 可用的 API 格式
 const availableFormats = computed(() => {
   return Object.keys(keysByFormat.value).sort()
@@ -433,11 +469,18 @@ watch(internalOpen, async (open) => {
 // 加载当前的优先级模式配置
 async function loadCurrentPriorityMode() {
   try {
-    const response = await adminApi.getSystemConfig('provider_priority_mode')
-    const currentMode = response.value || 'provider'
+    const [priorityResponse, schedulingResponse] = await Promise.all([
+      adminApi.getSystemConfig('provider_priority_mode'),
+      adminApi.getSystemConfig('scheduling_mode')
+    ])
+    const currentMode = priorityResponse.value || 'provider'
     activeMainTab.value = currentMode === 'global_key' ? 'key' : 'provider'
+
+    const currentSchedulingMode = schedulingResponse.value || 'cache_affinity'
+    schedulingMode.value = currentSchedulingMode === 'fixed_order' ? 'fixed_order' : 'cache_affinity'
   } catch {
     activeMainTab.value = 'provider'
+    schedulingMode.value = 'cache_affinity'
   }
 }
 
@@ -611,11 +654,19 @@ async function save() {
 
     const newMode = activeMainTab.value === 'key' ? 'global_key' : 'provider'
 
-    await adminApi.updateSystemConfig(
+    // 保存优先级模式和调度模式
+    await Promise.all([
+      adminApi.updateSystemConfig(
         'provider_priority_mode',
         newMode,
         'Provider/Key 优先级策略：provider(提供商优先模式) 或 global_key(全局Key优先模式)'
+      ),
+      adminApi.updateSystemConfig(
+        'scheduling_mode',
+        schedulingMode.value,
+        '调度模式：fixed_order(固定顺序模式) 或 cache_affinity(缓存亲和模式)'
       )
+    ])
 
     const providerUpdates = sortedProviders.value.map((provider, index) =>
       updateProvider(provider.id, { provider_priority: index + 1 })

frontend/src/features/providers/components/provider-tabs/ModelAliasesTab.vue

@@ -286,7 +286,9 @@
                     @click="addUpstreamModel(model.id)"
                   >
                     <div class="flex-1 min-w-0">
-                      <div class="font-mono text-xs truncate">{{ model.id }}</div>
+                      <div class="font-mono text-xs truncate">
+                        {{ model.id }}
+                      </div>
                       <div
                         v-if="model.owned_by"
                         class="text-xs text-muted-foreground truncate"

frontend/src/views/admin/SystemSettings.vue

@@ -7,6 +7,7 @@
       <template #actions>
         <Button
           :disabled="loading"
+          class="shadow-none hover:shadow-none"
           @click="saveSystemConfig"
         >
           {{ loading ? '保存中...' : '保存所有配置' }}
@@ -465,16 +466,12 @@
     </div>
 
     <!-- 导入配置对话框 -->
-    <Dialog v-model:open="importDialogOpen">
-      <DialogContent class="max-w-lg">
-        <DialogHeader>
-          <DialogTitle>导入配置</DialogTitle>
-          <DialogDescription>
-            选择冲突处理模式并确认导入
-          </DialogDescription>
-        </DialogHeader>
-
-        <div class="space-y-4 py-4">
+    <Dialog
+      v-model:open="importDialogOpen"
+      title="导入配置"
+      description="选择冲突处理模式并确认导入"
+    >
+      <div class="space-y-4">
         <div
           v-if="importPreview"
           class="p-3 bg-muted rounded-lg text-sm"
@@ -496,7 +493,10 @@
 
         <div>
           <Label class="block text-sm font-medium mb-2">冲突处理模式</Label>
-            <Select v-model="mergeMode">
+          <Select
+            v-model="mergeMode"
+            v-model:open="mergeModeSelectOpen"
+          >
             <SelectTrigger>
               <SelectValue />
             </SelectTrigger>
@@ -525,17 +525,15 @@
           </p>
         </div>
 
-          <div class="p-3 bg-yellow-500/10 border border-yellow-500/20 rounded-lg">
-            <p class="text-sm text-yellow-600 dark:text-yellow-400">
+        <p class="text-xs text-muted-foreground">
           注意：相同的 API Keys 会自动跳过，不会创建重复记录。
         </p>
       </div>
-        </div>
 
-        <DialogFooter>
+      <template #footer>
         <Button
           variant="outline"
-            @click="importDialogOpen = false"
+          @click="importDialogOpen = false; mergeModeSelectOpen = false"
         >
           取消
         </Button>
@@ -545,20 +543,17 @@
         >
           {{ importLoading ? '导入中...' : '确认导入' }}
         </Button>
-        </DialogFooter>
-      </DialogContent>
+      </template>
     </Dialog>
 
     <!-- 导入结果对话框 -->
-    <Dialog v-model:open="importResultDialogOpen">
-      <DialogContent class="max-w-lg">
-        <DialogHeader>
-          <DialogTitle>导入完成</DialogTitle>
-        </DialogHeader>
-
+    <Dialog
+      v-model:open="importResultDialogOpen"
+      title="导入完成"
+    >
       <div
         v-if="importResult"
-          class="space-y-4 py-4"
+        class="space-y-4"
       >
         <div class="grid grid-cols-2 gap-4 text-sm">
           <div class="p-3 bg-muted rounded-lg">
@@ -614,12 +609,12 @@
 
         <div
           v-if="importResult.stats.errors.length > 0"
-            class="p-3 bg-red-500/10 border border-red-500/20 rounded-lg"
+          class="p-3 bg-destructive/10 rounded-lg"
         >
-            <p class="font-medium text-red-600 dark:text-red-400 mb-2">
+          <p class="font-medium text-destructive mb-2">
             警告信息
           </p>
-            <ul class="text-sm text-red-600 dark:text-red-400 space-y-1">
+          <ul class="text-sm text-destructive space-y-1">
             <li
               v-for="(err, index) in importResult.stats.errors"
               :key="index"
@@ -630,25 +625,20 @@
         </div>
       </div>
 
-        <DialogFooter>
+      <template #footer>
         <Button @click="importResultDialogOpen = false">
           确定
         </Button>
-        </DialogFooter>
-      </DialogContent>
+      </template>
     </Dialog>
 
     <!-- 用户数据导入对话框 -->
-    <Dialog v-model:open="importUsersDialogOpen">
-      <DialogContent class="max-w-lg">
-        <DialogHeader>
-          <DialogTitle>导入用户数据</DialogTitle>
-          <DialogDescription>
-            选择冲突处理模式并确认导入
-          </DialogDescription>
-        </DialogHeader>
-
-        <div class="space-y-4 py-4">
+    <Dialog
+      v-model:open="importUsersDialogOpen"
+      title="导入用户数据"
+      description="选择冲突处理模式并确认导入"
+    >
+      <div class="space-y-4">
         <div
           v-if="importUsersPreview"
           class="p-3 bg-muted rounded-lg text-sm"
@@ -666,7 +656,10 @@
 
         <div>
           <Label class="block text-sm font-medium mb-2">冲突处理模式</Label>
-            <Select v-model="usersMergeMode">
+          <Select
+            v-model="usersMergeMode"
+            v-model:open="usersMergeModeSelectOpen"
+          >
             <SelectTrigger>
               <SelectValue />
             </SelectTrigger>
@@ -695,17 +688,15 @@
           </p>
         </div>
 
-          <div class="p-3 bg-yellow-500/10 border border-yellow-500/20 rounded-lg">
-            <p class="text-sm text-yellow-600 dark:text-yellow-400">
+        <p class="text-xs text-muted-foreground">
           注意：用户 API Keys 需要目标系统使用相同的 ENCRYPTION_KEY 环境变量才能正常工作。
         </p>
       </div>
-        </div>
 
-        <DialogFooter>
+      <template #footer>
         <Button
           variant="outline"
-            @click="importUsersDialogOpen = false"
+          @click="importUsersDialogOpen = false; usersMergeModeSelectOpen = false"
         >
           取消
         </Button>
@@ -715,20 +706,17 @@
         >
           {{ importUsersLoading ? '导入中...' : '确认导入' }}
         </Button>
-        </DialogFooter>
-      </DialogContent>
+      </template>
     </Dialog>
 
     <!-- 用户数据导入结果对话框 -->
-    <Dialog v-model:open="importUsersResultDialogOpen">
-      <DialogContent class="max-w-lg">
-        <DialogHeader>
-          <DialogTitle>用户数据导入完成</DialogTitle>
-        </DialogHeader>
-
+    <Dialog
+      v-model:open="importUsersResultDialogOpen"
+      title="用户数据导入完成"
+    >
       <div
         v-if="importUsersResult"
-          class="space-y-4 py-4"
+        class="space-y-4"
       >
         <div class="grid grid-cols-2 gap-4 text-sm">
           <div class="p-3 bg-muted rounded-lg">
@@ -754,12 +742,12 @@
 
         <div
           v-if="importUsersResult.stats.errors.length > 0"
-            class="p-3 bg-red-500/10 border border-red-500/20 rounded-lg"
+          class="p-3 bg-destructive/10 rounded-lg"
         >
-            <p class="font-medium text-red-600 dark:text-red-400 mb-2">
+          <p class="font-medium text-destructive mb-2">
             警告信息
           </p>
-            <ul class="text-sm text-red-600 dark:text-red-400 space-y-1">
+          <ul class="text-sm text-destructive space-y-1">
             <li
               v-for="(err, index) in importUsersResult.stats.errors"
               :key="index"
@@ -770,12 +758,11 @@
         </div>
       </div>
 
-        <DialogFooter>
+      <template #footer>
         <Button @click="importUsersResultDialogOpen = false">
           确定
         </Button>
-        </DialogFooter>
-      </DialogContent>
+      </template>
     </Dialog>
   </PageContainer>
 </template>
@@ -794,11 +781,6 @@ import SelectContent from '@/components/ui/select-content.vue'
 import SelectItem from '@/components/ui/select-item.vue'
 import {
   Dialog,
-  DialogContent,
-  DialogHeader,
-  DialogTitle,
-  DialogDescription,
-  DialogFooter
 } from '@/components/ui'
 import { PageHeader, PageContainer, CardSection } from '@/components/layout'
 import { useToast } from '@/composables/useToast'
@@ -843,6 +825,7 @@ const configFileInput = ref<HTMLInputElement | null>(null)
 const importPreview = ref<ConfigExportData | null>(null)
 const importResult = ref<ConfigImportResponse | null>(null)
 const mergeMode = ref<'skip' | 'overwrite' | 'error'>('skip')
+const mergeModeSelectOpen = ref(false)
 
 // 用户数据导出/导入相关
 const exportUsersLoading = ref(false)
@@ -853,6 +836,7 @@ const usersFileInput = ref<HTMLInputElement | null>(null)
 const importUsersPreview = ref<UsersExportData | null>(null)
 const importUsersResult = ref<UsersImportResponse | null>(null)
 const usersMergeMode = ref<'skip' | 'overwrite' | 'error'>('skip')
+const usersMergeModeSelectOpen = ref(false)
 
 const systemConfig = ref<SystemConfig>({
   // 基础配置
@@ -1136,6 +1120,7 @@ async function confirmImport() {
     })
     importResult.value = result
     importDialogOpen.value = false
+    mergeModeSelectOpen.value = false
     importResultDialogOpen.value = true
     success('配置导入成功')
   } catch (err: any) {
@@ -1224,6 +1209,7 @@ async function confirmImportUsers() {
     })
     importUsersResult.value = result
     importUsersDialogOpen.value = false
+    usersMergeModeSelectOpen.value = false
     importUsersResultDialogOpen.value = true
     success('用户数据导入成功')
   } catch (err: any) {

frontend/src/views/shared/Dashboard.vue

@@ -103,7 +103,7 @@
           </div>
           <div class="grid grid-cols-2 gap-2 sm:gap-3 xl:grid-cols-4">
             <Card class="relative p-3 sm:p-4 border-book-cloth/30">
-              <Clock class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-book-cloth" />
+              <Clock class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-muted-foreground" />
               <div class="pr-6">
                 <p class="text-[9px] sm:text-[10px] font-semibold uppercase tracking-[0.2em] sm:tracking-[0.3em] text-muted-foreground">
                   平均响应
@@ -114,7 +114,7 @@
               </div>
             </Card>
             <Card class="relative p-3 sm:p-4 border-kraft/30">
-              <AlertTriangle class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-kraft" />
+              <AlertTriangle class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-muted-foreground" />
               <div class="pr-6">
                 <p class="text-[9px] sm:text-[10px] font-semibold uppercase tracking-[0.2em] sm:tracking-[0.3em] text-muted-foreground">
                   错误率
@@ -128,7 +128,7 @@
               </div>
             </Card>
             <Card class="relative p-3 sm:p-4 border-book-cloth/25">
-              <Shuffle class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-kraft" />
+              <Shuffle class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-muted-foreground" />
               <div class="pr-6">
                 <p class="text-[9px] sm:text-[10px] font-semibold uppercase tracking-[0.2em] sm:tracking-[0.3em] text-muted-foreground">
                   转移次数
@@ -142,7 +142,7 @@
               v-if="costStats"
               class="relative p-3 sm:p-4 border-manilla/40"
             >
-              <DollarSign class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-book-cloth" />
+              <DollarSign class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-muted-foreground" />
               <div class="pr-6">
                 <p class="text-[9px] sm:text-[10px] font-semibold uppercase tracking-[0.2em] sm:tracking-[0.3em] text-muted-foreground">
                   实际成本
@@ -180,7 +180,7 @@
           </div>
           <div class="grid grid-cols-2 gap-2 sm:gap-3 xl:grid-cols-4">
             <Card class="relative p-3 sm:p-4 border-book-cloth/30">
-              <Database class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-book-cloth" />
+              <Database class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-muted-foreground" />
               <div class="pr-6">
                 <p class="text-[9px] sm:text-[10px] font-semibold uppercase tracking-[0.2em] sm:tracking-[0.3em] text-muted-foreground">
                   缓存命中率
@@ -191,7 +191,7 @@
               </div>
             </Card>
             <Card class="relative p-3 sm:p-4 border-kraft/30">
-              <Hash class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-kraft" />
+              <Hash class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-muted-foreground" />
               <div class="pr-6">
                 <p class="text-[9px] sm:text-[10px] font-semibold uppercase tracking-[0.2em] sm:tracking-[0.3em] text-muted-foreground">
                   缓存读取
@@ -202,7 +202,7 @@
               </div>
             </Card>
             <Card class="relative p-3 sm:p-4 border-book-cloth/25">
-              <Database class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-kraft" />
+              <Database class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-muted-foreground" />
               <div class="pr-6">
                 <p class="text-[9px] sm:text-[10px] font-semibold uppercase tracking-[0.2em] sm:tracking-[0.3em] text-muted-foreground">
                   缓存创建
@@ -216,7 +216,7 @@
               v-if="tokenBreakdown"
               class="relative p-3 sm:p-4 border-manilla/40"
             >
-              <Hash class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-book-cloth" />
+              <Hash class="absolute top-3 right-3 h-3.5 w-3.5 sm:h-4 sm:w-4 text-muted-foreground" />
               <div class="pr-6">
                 <p class="text-[9px] sm:text-[10px] font-semibold uppercase tracking-[0.2em] sm:tracking-[0.3em] text-muted-foreground">
                   总Token
@@ -254,16 +254,16 @@
         <Card class="overflow-hidden p-4 flex flex-col flex-1 min-h-0 h-full max-h-[280px] sm:max-h-none">
           <div
             v-if="loadingAnnouncements"
-            class="py-8 text-center"
+            class="flex-1 flex items-center justify-center"
           >
-            <Loader2 class="h-5 w-5 animate-spin mx-auto text-muted-foreground" />
+            <Loader2 class="h-5 w-5 animate-spin text-muted-foreground" />
           </div>
 
           <div
             v-else-if="announcements.length === 0"
-            class="py-8 text-center"
+            class="flex-1 flex flex-col items-center justify-center"
           >
-            <Bell class="h-8 w-8 mx-auto text-muted-foreground/40" />
+            <Bell class="h-8 w-8 text-muted-foreground/40" />
             <p class="mt-2 text-xs text-muted-foreground">
               暂无公告
             </p>
@@ -793,9 +793,8 @@ const statCardGlows = [
   'bg-kraft/30'
 ]
 
-const getStatIconColor = (index: number): string => {
-  const colors = ['text-book-cloth', 'text-kraft', 'text-book-cloth', 'text-kraft']
-  return colors[index % colors.length]
+const getStatIconColor = (_index: number): string => {
+  return 'text-muted-foreground'
 }
 
 // 统计数据

src/api/admin/monitoring/cache.py

@@ -21,7 +21,8 @@ from src.core.logger import logger
 from src.database import get_db
 from src.models.database import ApiKey, User
 from src.services.cache.affinity_manager import get_affinity_manager
-from src.services.cache.aware_scheduler import get_cache_aware_scheduler
+from src.services.cache.aware_scheduler import CacheAwareScheduler, get_cache_aware_scheduler
+from src.services.system.config import SystemConfigService
 
 router = APIRouter(prefix="/api/admin/monitoring/cache", tags=["Admin - Monitoring: Cache"])
 pipeline = ApiRequestPipeline()
@@ -250,7 +251,22 @@ class AdminCacheStatsAdapter(AdminApiAdapter):
     async def handle(self, context: ApiRequestContext) -> Dict[str, Any]:  # type: ignore[override]
         try:
             redis_client = get_redis_client_sync()
-            scheduler = await get_cache_aware_scheduler(redis_client)
+            # 读取系统配置，确保监控接口与编排器使用一致的模式
+            priority_mode = SystemConfigService.get_config(
+                context.db,
+                "provider_priority_mode",
+                CacheAwareScheduler.PRIORITY_MODE_PROVIDER,
+            )
+            scheduling_mode = SystemConfigService.get_config(
+                context.db,
+                "scheduling_mode",
+                CacheAwareScheduler.SCHEDULING_MODE_CACHE_AFFINITY,
+            )
+            scheduler = await get_cache_aware_scheduler(
+                redis_client,
+                priority_mode=priority_mode,
+                scheduling_mode=scheduling_mode,
+            )
             stats = await scheduler.get_stats()
             logger.info("缓存统计信息查询成功")
             context.add_audit_metadata(
@@ -270,7 +286,22 @@ class AdminCacheMetricsAdapter(AdminApiAdapter):
     async def handle(self, context: ApiRequestContext) -> PlainTextResponse:
         try:
             redis_client = get_redis_client_sync()
-            scheduler = await get_cache_aware_scheduler(redis_client)
+            # 读取系统配置，确保监控接口与编排器使用一致的模式
+            priority_mode = SystemConfigService.get_config(
+                context.db,
+                "provider_priority_mode",
+                CacheAwareScheduler.PRIORITY_MODE_PROVIDER,
+            )
+            scheduling_mode = SystemConfigService.get_config(
+                context.db,
+                "scheduling_mode",
+                CacheAwareScheduler.SCHEDULING_MODE_CACHE_AFFINITY,
+            )
+            scheduler = await get_cache_aware_scheduler(
+                redis_client,
+                priority_mode=priority_mode,
+                scheduling_mode=scheduling_mode,
+            )
             stats = await scheduler.get_stats()
             payload = self._format_prometheus(stats)
             context.add_audit_metadata(

src/api/admin/system.py

@@ -852,7 +852,7 @@ class AdminImportConfigAdapter(AdminApiAdapter):
             from src.services.cache.invalidation import get_cache_invalidation_service
 
             cache_service = get_cache_invalidation_service()
-            cache_service.invalidate_all()
+            cache_service.clear_all_caches()
 
             return {
                 "message": "配置导入成功",

src/services/cache/aware_scheduler.py

@@ -121,8 +121,17 @@ class CacheAwareScheduler:
         PRIORITY_MODE_PROVIDER,
         PRIORITY_MODE_GLOBAL_KEY,
     }
+    # 调度模式常量
+    SCHEDULING_MODE_FIXED_ORDER = "fixed_order"  # 固定顺序模式
+    SCHEDULING_MODE_CACHE_AFFINITY = "cache_affinity"  # 缓存亲和模式
+    ALLOWED_SCHEDULING_MODES = {
+        SCHEDULING_MODE_FIXED_ORDER,
+        SCHEDULING_MODE_CACHE_AFFINITY,
+    }
 
-    def __init__(self, redis_client=None, priority_mode: Optional[str] = None):
+    def __init__(
+        self, redis_client=None, priority_mode: Optional[str] = None, scheduling_mode: Optional[str] = None
+    ):
         """
         初始化调度器
 
@@ -132,12 +141,16 @@ class CacheAwareScheduler:
         Args:
             redis_client: Redis客户端（可选）
             priority_mode: 候选排序策略（provider | global_key）
+            scheduling_mode: 调度模式（fixed_order | cache_affinity）
         """
         self.redis = redis_client
         self.priority_mode = self._normalize_priority_mode(
             priority_mode or self.PRIORITY_MODE_PROVIDER
         )
-        logger.debug(f"[CacheAwareScheduler] 初始化优先级模式: {self.priority_mode}")
+        self.scheduling_mode = self._normalize_scheduling_mode(
+            scheduling_mode or self.SCHEDULING_MODE_CACHE_AFFINITY
+        )
+        logger.debug(f"[CacheAwareScheduler] 初始化优先级模式: {self.priority_mode}, 调度模式: {self.scheduling_mode}")
 
         # 初始化子组件（将在第一次使用时异步初始化）
         self._affinity_manager: Optional[CacheAffinityManager] = None
@@ -673,7 +686,8 @@ class CacheAwareScheduler:
             f"(api_format={target_format.value}, model={model_name})"
         )
 
-        # 4. 应用缓存亲和性排序（使用 global_model_id 作为模型标识）
+        # 4. 应用缓存亲和性排序（仅在缓存亲和模式下启用）
+        if self.scheduling_mode == self.SCHEDULING_MODE_CACHE_AFFINITY:
             if affinity_key and candidates:
                 candidates = await self._apply_cache_affinity(
                     candidates=candidates,
@@ -681,6 +695,10 @@ class CacheAwareScheduler:
                     api_format=target_format,
                     global_model_id=global_model_id,
                 )
+        else:
+            # 固定顺序模式：标记所有候选为非缓存
+            for candidate in candidates:
+                candidate.is_cached = False
 
         return candidates, global_model_id
 
@@ -1060,6 +1078,22 @@ class CacheAwareScheduler:
         self.priority_mode = normalized
         logger.debug(f"[CacheAwareScheduler] 切换优先级模式为: {self.priority_mode}")
 
+    def _normalize_scheduling_mode(self, mode: Optional[str]) -> str:
+        normalized = (mode or "").strip().lower()
+        if normalized not in self.ALLOWED_SCHEDULING_MODES:
+            if normalized:
+                logger.warning(f"[CacheAwareScheduler] 无效的调度模式 '{mode}'，回退为 cache_affinity")
+            return self.SCHEDULING_MODE_CACHE_AFFINITY
+        return normalized
+
+    def set_scheduling_mode(self, mode: Optional[str]) -> None:
+        """运行时更新调度模式"""
+        normalized = self._normalize_scheduling_mode(mode)
+        if normalized == self.scheduling_mode:
+            return
+        self.scheduling_mode = normalized
+        logger.debug(f"[CacheAwareScheduler] 切换调度模式为: {self.scheduling_mode}")
+
     def _apply_priority_mode_sort(
         self, candidates: List[ProviderCandidate], affinity_key: Optional[str] = None
     ) -> List[ProviderCandidate]:
@@ -1307,6 +1341,7 @@ _scheduler: Optional[CacheAwareScheduler] = None
 async def get_cache_aware_scheduler(
     redis_client=None,
     priority_mode: Optional[str] = None,
+    scheduling_mode: Optional[str] = None,
 ) -> CacheAwareScheduler:
     """
     获取全局CacheAwareScheduler实例
@@ -1317,6 +1352,7 @@ async def get_cache_aware_scheduler(
     Args:
         redis_client: Redis客户端（可选）
         priority_mode: 外部覆盖的优先级模式（provider | global_key）
+        scheduling_mode: 外部覆盖的调度模式（fixed_order | cache_affinity）
 
     Returns:
         CacheAwareScheduler实例
@@ -1324,8 +1360,13 @@ async def get_cache_aware_scheduler(
     global _scheduler
 
     if _scheduler is None:
-        _scheduler = CacheAwareScheduler(redis_client, priority_mode=priority_mode)
-    elif priority_mode:
+        _scheduler = CacheAwareScheduler(
+            redis_client, priority_mode=priority_mode, scheduling_mode=scheduling_mode
+        )
+    else:
+        if priority_mode:
             _scheduler.set_priority_mode(priority_mode)
+        if scheduling_mode:
+            _scheduler.set_scheduling_mode(scheduling_mode)
 
     return _scheduler

src/services/orchestration/fallback_orchestrator.py

@@ -102,9 +102,15 @@ class FallbackOrchestrator:
                 "provider_priority_mode",
                 CacheAwareScheduler.PRIORITY_MODE_PROVIDER,
             )
+            scheduling_mode = SystemConfigService.get_config(
+                self.db,
+                "scheduling_mode",
+                CacheAwareScheduler.SCHEDULING_MODE_CACHE_AFFINITY,
+            )
             self.cache_scheduler = await get_cache_aware_scheduler(
                 self.redis,
                 priority_mode=priority_mode,
+                scheduling_mode=scheduling_mode,
             )
         else:
             # 确保运行时配置变更能生效
@@ -113,7 +119,13 @@ class FallbackOrchestrator:
                 "provider_priority_mode",
                 CacheAwareScheduler.PRIORITY_MODE_PROVIDER,
             )
+            scheduling_mode = SystemConfigService.get_config(
+                self.db,
+                "scheduling_mode",
+                CacheAwareScheduler.SCHEDULING_MODE_CACHE_AFFINITY,
+            )
             self.cache_scheduler.set_priority_mode(priority_mode)
+            self.cache_scheduler.set_scheduling_mode(scheduling_mode)
 
         # 确保 cache_scheduler 内部组件也已初始化
         await self.cache_scheduler._ensure_initialized()

src/services/system/config.py

@@ -71,6 +71,10 @@ class SystemConfigService:
             "value": "provider",
             "description": "优先级策略：provider(提供商优先模式) 或 global_key(全局Key优先模式)",
         },
+        "scheduling_mode": {
+            "value": "cache_affinity",
+            "description": "调度模式：fixed_order(固定顺序模式，严格按优先级顺序) 或 cache_affinity(缓存亲和模式，优先使用已缓存的Provider)",
+        },
         "auto_delete_expired_keys": {
             "value": False,
             "description": "是否自动删除过期的API Key（True=物理删除，False=仅禁用），仅管理员可配置",