fix(gateway): 修复账号跨分组调度问题

问题：账号可能被调度到未分配的分组（如 simon 账号被调度到 claude_default）

根因：
- 强制平台模式下分组查询失败时回退到全平台查询
- listSchedulableAccounts 中分组为空时回退到无分组查询
- 粘性会话只检查平台匹配，未校验账号分组归属

修复：
- 移除强制平台模式的回退逻辑，分组内无账号时返回错误
- 移除 listSchedulableAccounts 的回退逻辑
- 新增 isAccountInGroup 方法用于分组校验
- 在三处粘性会话检查中增加分组归属验证

backend/internal/service/gateway_service.go

@@ -35,6 +35,7 @@ const (
 	stickySessionTTL        = time.Hour // 粘性会话TTL
 	defaultMaxLineSize      = 10 * 1024 * 1024
 	claudeCodeSystemPrompt  = "You are Claude Code, Anthropic's official CLI for Claude."
+	maxCacheControlBlocks   = 4 // Anthropic API 允许的最大 cache_control 块数量
 )
 
 // sseDataRe matches SSE data lines with optional whitespace after colon.
@@ -43,6 +44,16 @@ var (
 	sseDataRe            = regexp.MustCompile(`^data:\s*`)
 	sessionIDRegex       = regexp.MustCompile(`session_([a-f0-9-]{36})`)
 	claudeCliUserAgentRe = regexp.MustCompile(`^claude-cli/\d+\.\d+\.\d+`)
+
+	// claudeCodePromptPrefixes 用于检测 Claude Code 系统提示词的前缀列表
+	// 支持多种变体：标准版、Agent SDK 版、Explore Agent 版、Compact 版等
+	// 注意：前缀之间不应存在包含关系，否则会导致冗余匹配
+	claudeCodePromptPrefixes = []string{
+		"You are Claude Code, Anthropic's official CLI for Claude",             // 标准版 & Agent SDK 版（含 running within...）
+		"You are a Claude agent, built on Anthropic's Claude Agent SDK",        // Agent SDK 变体
+		"You are a file search specialist for Claude Code",                     // Explore Agent 版
+		"You are a helpful AI assistant tasked with summarizing conversations", // Compact 版
+	}
 )
 
 // allowedHeaders 白名单headers（参考CRS项目）
@@ -355,17 +366,8 @@ func (s *GatewayService) SelectAccountForModelWithExclusions(ctx context.Context
 		return s.selectAccountWithMixedScheduling(ctx, groupID, sessionHash, requestedModel, excludedIDs, platform)
 	}
 
-	// 强制平台模式：优先按分组查找，找不到再查全部该平台账户
-	if hasForcePlatform && groupID != nil {
-		account, err := s.selectAccountForModelWithPlatform(ctx, groupID, sessionHash, requestedModel, excludedIDs, platform)
-		if err == nil {
-			return account, nil
-		}
-		// 分组中找不到，回退查询全部该平台账户
-		groupID = nil
-	}
-
 	// antigravity 分组、强制平台模式或无分组使用单平台选择
+	// 注意：强制平台模式也必须遵守分组限制，不再回退到全平台查询
 	return s.selectAccountForModelWithPlatform(ctx, groupID, sessionHash, requestedModel, excludedIDs, platform)
 }
 
@@ -443,7 +445,8 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 		accountID, err := s.cache.GetSessionAccountID(ctx, sessionHash)
 		if err == nil && accountID > 0 && !isExcluded(accountID) {
 			account, err := s.accountRepo.GetByID(ctx, accountID)
-			if err == nil && s.isAccountAllowedForPlatform(account, platform, useMixed) &&
+			if err == nil && s.isAccountInGroup(account, groupID) &&
+				s.isAccountAllowedForPlatform(account, platform, useMixed) &&
 				account.IsSchedulable() &&
 				(requestedModel == "" || s.isModelSupportedByAccount(account, requestedModel)) {
 				result, err := s.tryAcquireAccountSlot(ctx, accountID, account.Concurrency)
@@ -660,9 +663,7 @@ func (s *GatewayService) listSchedulableAccounts(ctx context.Context, groupID *i
 		accounts, err = s.accountRepo.ListSchedulableByPlatform(ctx, platform)
 	} else if groupID != nil {
 		accounts, err = s.accountRepo.ListSchedulableByGroupIDAndPlatform(ctx, *groupID, platform)
-		if err == nil && len(accounts) == 0 && hasForcePlatform {
+		// 分组内无账号则返回空列表，由上层处理错误，不再回退到全平台查询
-			accounts, err = s.accountRepo.ListSchedulableByPlatform(ctx, platform)
-		}
 	} else {
 		accounts, err = s.accountRepo.ListSchedulableByPlatform(ctx, platform)
 	}
@@ -685,6 +686,23 @@ func (s *GatewayService) isAccountAllowedForPlatform(account *Account, platform
 	return account.Platform == platform
 }
 
+// isAccountInGroup checks if the account belongs to the specified group.
+// Returns true if groupID is nil (no group restriction) or account belongs to the group.
+func (s *GatewayService) isAccountInGroup(account *Account, groupID *int64) bool {
+	if groupID == nil {
+		return true // 无分组限制
+	}
+	if account == nil {
+		return false
+	}
+	for _, ag := range account.AccountGroups {
+		if ag.GroupID == *groupID {
+			return true
+		}
+	}
+	return false
+}
+
 func (s *GatewayService) tryAcquireAccountSlot(ctx context.Context, accountID int64, maxConcurrency int) (*AcquireResult, error) {
 	if s.concurrencyService == nil {
 		return &AcquireResult{Acquired: true, ReleaseFunc: func() {}}, nil
@@ -723,8 +741,8 @@ func (s *GatewayService) selectAccountForModelWithPlatform(ctx context.Context,
 		if err == nil && accountID > 0 {
 			if _, excluded := excludedIDs[accountID]; !excluded {
 				account, err := s.accountRepo.GetByID(ctx, accountID)
-				// 检查账号平台是否匹配（确保粘性会话不会跨平台）
+				// 检查账号分组归属和平台匹配（确保粘性会话不会跨分组或跨平台）
-				if err == nil && account.Platform == platform && account.IsSchedulable() && (requestedModel == "" || s.isModelSupportedByAccount(account, requestedModel)) {
+				if err == nil && s.isAccountInGroup(account, groupID) && account.Platform == platform && account.IsSchedulable() && (requestedModel == "" || s.isModelSupportedByAccount(account, requestedModel)) {
 					if err := s.cache.RefreshSessionTTL(ctx, sessionHash, stickySessionTTL); err != nil {
 						log.Printf("refresh session ttl failed: session=%s err=%v", sessionHash, err)
 					}
@@ -812,8 +830,8 @@ func (s *GatewayService) selectAccountWithMixedScheduling(ctx context.Context, g
 		if err == nil && accountID > 0 {
 			if _, excluded := excludedIDs[accountID]; !excluded {
 				account, err := s.accountRepo.GetByID(ctx, accountID)
-				// 检查账号是否有效：原生平台直接匹配，antigravity 需要启用混合调度
+				// 检查账号分组归属和有效性：原生平台直接匹配，antigravity 需要启用混合调度
-				if err == nil && account.IsSchedulable() && (requestedModel == "" || s.isModelSupportedByAccount(account, requestedModel)) {
+				if err == nil && s.isAccountInGroup(account, groupID) && account.IsSchedulable() && (requestedModel == "" || s.isModelSupportedByAccount(account, requestedModel)) {
 					if account.Platform == nativePlatform || (account.Platform == PlatformAntigravity && account.IsMixedSchedulingEnabled()) {
 						if err := s.cache.RefreshSessionTTL(ctx, sessionHash, stickySessionTTL); err != nil {
 							log.Printf("refresh session ttl failed: session=%s err=%v", sessionHash, err)
@@ -1013,15 +1031,15 @@ func isClaudeCodeClient(userAgent string, metadataUserID string) bool {
 }
 
 // systemIncludesClaudeCodePrompt 检查 system 中是否已包含 Claude Code 提示词
-// 支持 string 和 []any 两种格式
+// 使用前缀匹配支持多种变体（标准版、Agent SDK 版等）
 func systemIncludesClaudeCodePrompt(system any) bool {
 	switch v := system.(type) {
 	case string:
-		return v == claudeCodeSystemPrompt
+		return hasClaudeCodePrefix(v)
 	case []any:
 		for _, item := range v {
 			if m, ok := item.(map[string]any); ok {
-				if text, ok := m["text"].(string); ok && text == claudeCodeSystemPrompt {
+				if text, ok := m["text"].(string); ok && hasClaudeCodePrefix(text) {
 					return true
 				}
 			}
@@ -1030,6 +1048,16 @@ func systemIncludesClaudeCodePrompt(system any) bool {
 	return false
 }
 
+// hasClaudeCodePrefix 检查文本是否以 Claude Code 提示词的特征前缀开头
+func hasClaudeCodePrefix(text string) bool {
+	for _, prefix := range claudeCodePromptPrefixes {
+		if strings.HasPrefix(text, prefix) {
+			return true
+		}
+	}
+	return false
+}
+
 // injectClaudeCodePrompt 在 system 开头注入 Claude Code 提示词
 // 处理 null、字符串、数组三种格式
 func injectClaudeCodePrompt(body []byte, system any) []byte {
@@ -1073,6 +1101,124 @@ func injectClaudeCodePrompt(body []byte, system any) []byte {
 	return result
 }
 
+// enforceCacheControlLimit 强制执行 cache_control 块数量限制（最多 4 个）
+// 超限时优先从 messages 中移除 cache_control，保护 system 中的缓存控制
+func enforceCacheControlLimit(body []byte) []byte {
+	var data map[string]any
+	if err := json.Unmarshal(body, &data); err != nil {
+		return body
+	}
+
+	// 计算当前 cache_control 块数量
+	count := countCacheControlBlocks(data)
+	if count <= maxCacheControlBlocks {
+		return body
+	}
+
+	// 超限：优先从 messages 中移除，再从 system 中移除
+	for count > maxCacheControlBlocks {
+		if removeCacheControlFromMessages(data) {
+			count--
+			continue
+		}
+		if removeCacheControlFromSystem(data) {
+			count--
+			continue
+		}
+		break
+	}
+
+	result, err := json.Marshal(data)
+	if err != nil {
+		return body
+	}
+	return result
+}
+
+// countCacheControlBlocks 统计 system 和 messages 中的 cache_control 块数量
+func countCacheControlBlocks(data map[string]any) int {
+	count := 0
+
+	// 统计 system 中的块
+	if system, ok := data["system"].([]any); ok {
+		for _, item := range system {
+			if m, ok := item.(map[string]any); ok {
+				if _, has := m["cache_control"]; has {
+					count++
+				}
+			}
+		}
+	}
+
+	// 统计 messages 中的块
+	if messages, ok := data["messages"].([]any); ok {
+		for _, msg := range messages {
+			if msgMap, ok := msg.(map[string]any); ok {
+				if content, ok := msgMap["content"].([]any); ok {
+					for _, item := range content {
+						if m, ok := item.(map[string]any); ok {
+							if _, has := m["cache_control"]; has {
+								count++
+							}
+						}
+					}
+				}
+			}
+		}
+	}
+
+	return count
+}
+
+// removeCacheControlFromMessages 从 messages 中移除一个 cache_control（从头开始）
+// 返回 true 表示成功移除，false 表示没有可移除的
+func removeCacheControlFromMessages(data map[string]any) bool {
+	messages, ok := data["messages"].([]any)
+	if !ok {
+		return false
+	}
+
+	for _, msg := range messages {
+		msgMap, ok := msg.(map[string]any)
+		if !ok {
+			continue
+		}
+		content, ok := msgMap["content"].([]any)
+		if !ok {
+			continue
+		}
+		for _, item := range content {
+			if m, ok := item.(map[string]any); ok {
+				if _, has := m["cache_control"]; has {
+					delete(m, "cache_control")
+					return true
+				}
+			}
+		}
+	}
+	return false
+}
+
+// removeCacheControlFromSystem 从 system 中移除一个 cache_control（从尾部开始，保护注入的 prompt）
+// 返回 true 表示成功移除，false 表示没有可移除的
+func removeCacheControlFromSystem(data map[string]any) bool {
+	system, ok := data["system"].([]any)
+	if !ok {
+		return false
+	}
+
+	// 从尾部开始移除，保护开头注入的 Claude Code prompt
+	for i := len(system) - 1; i >= 0; i-- {
+		if m, ok := system[i].(map[string]any); ok {
+			if _, has := m["cache_control"]; has {
+				delete(m, "cache_control")
+				return true
+			}
+		}
+	}
+	return false
+}
+
 // Forward 转发请求到Claude API
 func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *Account, parsed *ParsedRequest) (*ForwardResult, error) {
 	startTime := time.Now()
@@ -1093,6 +1239,9 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 		body = injectClaudeCodePrompt(body, parsed.System)
 	}
 
+	// 强制执行 cache_control 块数量限制（最多 4 个）
+	body = enforceCacheControlLimit(body)
+
 	// 应用模型映射（仅对apikey类型账号）
 	originalModel := reqModel
 	if account.Type == AccountTypeAPIKey {

frontend/src/components/admin/usage/UsageFilters.vue

@@ -85,9 +85,40 @@
         </div>
 
         <!-- Account Filter -->
-        <div class="w-full sm:w-auto sm:min-w-[220px]">
+        <div ref="accountSearchRef" class="usage-filter-dropdown relative w-full sm:w-auto sm:min-w-[220px]">
           <label class="input-label">{{ t('admin.usage.account') }}</label>
-          <Select v-model="filters.account_id" :options="accountOptions" searchable @change="emitChange" />
+          <input
+            v-model="accountKeyword"
+            type="text"
+            class="input pr-8"
+            :placeholder="t('admin.usage.searchAccountPlaceholder')"
+            @input="debounceAccountSearch"
+            @focus="showAccountDropdown = true"
+          />
+          <button
+            v-if="filters.account_id"
+            type="button"
+            @click="clearAccount"
+            class="absolute right-2 top-9 text-gray-400"
+            aria-label="Clear account filter"
+          >
+            ✕
+          </button>
+          <div
+            v-if="showAccountDropdown && (accountResults.length > 0 || accountKeyword)"
+            class="absolute z-50 mt-1 max-h-60 w-full overflow-auto rounded-lg border bg-white shadow-lg dark:bg-gray-800"
+          >
+            <button
+              v-for="a in accountResults"
+              :key="a.id"
+              type="button"
+              @click="selectAccount(a)"
+              class="w-full px-4 py-2 text-left hover:bg-gray-100 dark:hover:bg-gray-700"
+            >
+              <span class="truncate">{{ a.name }}</span>
+              <span class="ml-2 text-xs text-gray-400">#{{ a.id }}</span>
+            </button>
+          </div>
         </div>
 
         <!-- Stream Type Filter -->
@@ -166,6 +197,7 @@ const filters = toRef(props, 'modelValue')
 
 const userSearchRef = ref<HTMLElement | null>(null)
 const apiKeySearchRef = ref<HTMLElement | null>(null)
+const accountSearchRef = ref<HTMLElement | null>(null)
 
 const userKeyword = ref('')
 const userResults = ref<SimpleUser[]>([])
@@ -177,9 +209,17 @@ const apiKeyResults = ref<SimpleApiKey[]>([])
 const showApiKeyDropdown = ref(false)
 let apiKeySearchTimeout: ReturnType<typeof setTimeout> | null = null
 
+interface SimpleAccount {
+  id: number
+  name: string
+}
+const accountKeyword = ref('')
+const accountResults = ref<SimpleAccount[]>([])
+const showAccountDropdown = ref(false)
+let accountSearchTimeout: ReturnType<typeof setTimeout> | null = null
+
 const modelOptions = ref<SelectOption[]>([{ value: null, label: t('admin.usage.allModels') }])
 const groupOptions = ref<SelectOption[]>([{ value: null, label: t('admin.usage.allGroups') }])
-const accountOptions = ref<SelectOption[]>([{ value: null, label: t('admin.usage.allAccounts') }])
 
 const streamTypeOptions = ref<SelectOption[]>([
   { value: null, label: t('admin.usage.allTypes') },
@@ -278,6 +318,37 @@ const onClearApiKey = () => {
   emitChange()
 }
 
+const debounceAccountSearch = () => {
+  if (accountSearchTimeout) clearTimeout(accountSearchTimeout)
+  accountSearchTimeout = setTimeout(async () => {
+    if (!accountKeyword.value) {
+      accountResults.value = []
+      return
+    }
+    try {
+      const res = await adminAPI.accounts.list(1, 20, { search: accountKeyword.value })
+      accountResults.value = res.items.map((a) => ({ id: a.id, name: a.name }))
+    } catch {
+      accountResults.value = []
+    }
+  }, 300)
+}
+
+const selectAccount = (a: SimpleAccount) => {
+  accountKeyword.value = a.name
+  showAccountDropdown.value = false
+  filters.value.account_id = a.id
+  emitChange()
+}
+
+const clearAccount = () => {
+  accountKeyword.value = ''
+  accountResults.value = []
+  showAccountDropdown.value = false
+  filters.value.account_id = undefined
+  emitChange()
+}
+
 const onApiKeyFocus = () => {
   showApiKeyDropdown.value = true
   // Trigger search if no results yet
@@ -292,9 +363,11 @@ const onDocumentClick = (e: MouseEvent) => {
 
   const clickedInsideUser = userSearchRef.value?.contains(target) ?? false
   const clickedInsideApiKey = apiKeySearchRef.value?.contains(target) ?? false
+  const clickedInsideAccount = accountSearchRef.value?.contains(target) ?? false
 
   if (!clickedInsideUser) showUserDropdown.value = false
   if (!clickedInsideApiKey) showApiKeyDropdown.value = false
+  if (!clickedInsideAccount) showAccountDropdown.value = false
 }
 
 watch(
@@ -333,20 +406,27 @@ watch(
   }
 )
 
+watch(
+  () => filters.value.account_id,
+  (accountId) => {
+    if (!accountId) {
+      accountKeyword.value = ''
+      accountResults.value = []
+    }
+  }
+)
+
 onMounted(async () => {
   document.addEventListener('click', onDocumentClick)
 
   try {
-    const [gs, ms, as] = await Promise.all([
+    const [gs, ms] = await Promise.all([
       adminAPI.groups.list(1, 1000),
-      adminAPI.dashboard.getModelStats({ start_date: props.startDate, end_date: props.endDate }),
+      adminAPI.dashboard.getModelStats({ start_date: props.startDate, end_date: props.endDate })
-      adminAPI.accounts.list(1, 1000)
     ])
 
     groupOptions.value.push(...gs.items.map((g: any) => ({ value: g.id, label: g.name })))
 
-    accountOptions.value.push(...as.items.map((a: any) => ({ value: a.id, label: a.name })))
-
     const uniqueModels = new Set<string>()
     ms.models?.forEach((s: any) => s.model && uniqueModels.add(s.model))
     modelOptions.value.push(

frontend/src/components/admin/usage/UsageTable.vue

@@ -143,8 +143,8 @@
     >
       <div class="whitespace-nowrap rounded-lg border border-gray-700 bg-gray-900 px-3 py-2.5 text-xs text-white shadow-xl dark:border-gray-600 dark:bg-gray-800">
         <div class="space-y-1.5">
-          <div class="mb-2 border-b border-gray-700 pb-1.5">
+          <div>
-            <div class="text-xs font-semibold text-gray-300 mb-1">Token {{ t('usage.details') }}</div>
+            <div class="text-xs font-semibold text-gray-300 mb-1">{{ t('usage.tokenDetails') }}</div>
             <div v-if="tokenTooltipData && tokenTooltipData.input_tokens > 0" class="flex items-center justify-between gap-4">
               <span class="text-gray-400">{{ t('admin.usage.inputTokens') }}</span>
               <span class="font-medium text-white">{{ tokenTooltipData.input_tokens.toLocaleString() }}</span>
@@ -184,6 +184,27 @@
     >
       <div class="whitespace-nowrap rounded-lg border border-gray-700 bg-gray-900 px-3 py-2.5 text-xs text-white shadow-xl dark:border-gray-600 dark:bg-gray-800">
         <div class="space-y-1.5">
+          <!-- Cost Breakdown -->
+          <div class="mb-2 border-b border-gray-700 pb-1.5">
+            <div class="text-xs font-semibold text-gray-300 mb-1">{{ t('usage.costDetails') }}</div>
+            <div v-if="tooltipData && tooltipData.input_cost > 0" class="flex items-center justify-between gap-4">
+              <span class="text-gray-400">{{ t('admin.usage.inputCost') }}</span>
+              <span class="font-medium text-white">${{ tooltipData.input_cost.toFixed(6) }}</span>
+            </div>
+            <div v-if="tooltipData && tooltipData.output_cost > 0" class="flex items-center justify-between gap-4">
+              <span class="text-gray-400">{{ t('admin.usage.outputCost') }}</span>
+              <span class="font-medium text-white">${{ tooltipData.output_cost.toFixed(6) }}</span>
+            </div>
+            <div v-if="tooltipData && tooltipData.cache_creation_cost > 0" class="flex items-center justify-between gap-4">
+              <span class="text-gray-400">{{ t('admin.usage.cacheCreationCost') }}</span>
+              <span class="font-medium text-white">${{ tooltipData.cache_creation_cost.toFixed(6) }}</span>
+            </div>
+            <div v-if="tooltipData && tooltipData.cache_read_cost > 0" class="flex items-center justify-between gap-4">
+              <span class="text-gray-400">{{ t('admin.usage.cacheReadCost') }}</span>
+              <span class="font-medium text-white">${{ tooltipData.cache_read_cost.toFixed(6) }}</span>
+            </div>
+          </div>
+          <!-- Rate and Summary -->
           <div class="flex items-center justify-between gap-6">
             <span class="text-gray-400">{{ t('usage.rate') }}</span>
             <span class="font-semibold text-blue-400">{{ (tooltipData?.rate_multiplier || 1).toFixed(2) }}x</span>

frontend/src/components/keys/UseKeyModal.vue

@@ -105,10 +105,7 @@
                 </button>
               </div>
               <!-- Code Content -->
-              <pre class="p-4 text-sm font-mono text-gray-100 overflow-x-auto">
+              <pre class="p-4 text-sm font-mono text-gray-100 overflow-x-auto"><code v-if="file.highlighted" v-html="file.highlighted"></code><code v-else v-text="file.content"></code></pre>
-                <code v-if="file.highlighted" v-html="file.highlighted"></code>
-                <code v-else v-text="file.content"></code>
-              </pre>
             </div>
           </div>
         </div>

frontend/src/i18n/locales/en.ts

@@ -376,6 +376,8 @@ export default {
   usage: {
     title: 'Usage Records',
     description: 'View and analyze your API usage history',
+    costDetails: 'Cost Breakdown',
+    tokenDetails: 'Token Breakdown',
     totalRequests: 'Total Requests',
     totalTokens: 'Total Tokens',
     totalCost: 'Total Cost',
@@ -1691,6 +1693,7 @@ export default {
       userFilter: 'User',
       searchUserPlaceholder: 'Search user by email...',
       searchApiKeyPlaceholder: 'Search API key by name...',
+      searchAccountPlaceholder: 'Search account by name...',
       selectedUser: 'Selected',
       user: 'User',
       account: 'Account',

frontend/src/i18n/locales/zh.ts

@@ -373,6 +373,8 @@ export default {
   usage: {
     title: '使用记录',
     description: '查看和分析您的 API 使用历史',
+    costDetails: '成本明细',
+    tokenDetails: 'Token 明细',
     totalRequests: '总请求数',
     totalTokens: '总 Token',
     totalCost: '总消费',
@@ -1836,6 +1838,7 @@ export default {
       userFilter: '用户',
       searchUserPlaceholder: '按邮箱搜索用户...',
       searchApiKeyPlaceholder: '按名称搜索 API 密钥...',
+      searchAccountPlaceholder: '按名称搜索账号...',
       selectedUser: '已选择',
       user: '用户',
       account: '账户',

frontend/src/views/admin/UsageView.vue

@@ -85,11 +85,48 @@ const exportToExcel = async () => {
       if (all.length >= total || res.items.length < 100) break; p++
     }
     if(!c.signal.aborted) {
-      // 动态加载 xlsx，降低首屏包体并减少高危依赖的常驻暴露面。
       const XLSX = await import('xlsx')
-      const ws = XLSX.utils.json_to_sheet(all); const wb = XLSX.utils.book_new(); XLSX.utils.book_append_sheet(wb, ws, 'Usage')
+      const headers = [
-      saveAs(new Blob([XLSX.write(wb, { bookType: 'xlsx', type: 'array' })], { type: 'application/vnd.openxmlformats-officedocument.spreadsheetml.sheet' }), `usage_${Date.now()}.xlsx`)
+        t('usage.time'), t('admin.usage.user'), t('usage.apiKeyFilter'),
-      appStore.showSuccess('Export Success')
+        t('admin.usage.account'), t('usage.model'), t('admin.usage.group'),
+        t('usage.type'),
+        t('admin.usage.inputTokens'), t('admin.usage.outputTokens'),
+        t('admin.usage.cacheReadTokens'), t('admin.usage.cacheCreationTokens'),
+        t('admin.usage.inputCost'), t('admin.usage.outputCost'),
+        t('admin.usage.cacheReadCost'), t('admin.usage.cacheCreationCost'),
+        t('usage.rate'), t('usage.original'), t('usage.billed'),
+        t('usage.billingType'), t('usage.firstToken'), t('usage.duration'),
+        t('admin.usage.requestId')
+      ]
+      const rows = all.map(log => [
+        log.created_at,
+        log.user?.email || '',
+        log.api_key?.name || '',
+        log.account?.name || '',
+        log.model,
+        log.group?.name || '',
+        log.stream ? t('usage.stream') : t('usage.sync'),
+        log.input_tokens,
+        log.output_tokens,
+        log.cache_read_tokens,
+        log.cache_creation_tokens,
+        log.input_cost?.toFixed(6) || '0.000000',
+        log.output_cost?.toFixed(6) || '0.000000',
+        log.cache_read_cost?.toFixed(6) || '0.000000',
+        log.cache_creation_cost?.toFixed(6) || '0.000000',
+        log.rate_multiplier?.toFixed(2) || '1.00',
+        log.total_cost?.toFixed(6) || '0.000000',
+        log.actual_cost?.toFixed(6) || '0.000000',
+        log.billing_type === 1 ? t('usage.subscription') : t('usage.balance'),
+        log.first_token_ms ?? '',
+        log.duration_ms,
+        log.request_id || ''
+      ])
+      const ws = XLSX.utils.aoa_to_sheet([headers, ...rows])
+      const wb = XLSX.utils.book_new()
+      XLSX.utils.book_append_sheet(wb, ws, 'Usage')
+      saveAs(new Blob([XLSX.write(wb, { bookType: 'xlsx', type: 'array' })], { type: 'application/vnd.openxmlformats-officedocument.spreadsheetml.sheet' }), `usage_${filters.value.start_date}_to_${filters.value.end_date}.xlsx`)
+      appStore.showSuccess(t('usage.exportSuccess'))
     }
   } catch (error) { console.error('Failed to export:', error); appStore.showError('Export Failed') }
   finally { if(exportAbortController === c) { exportAbortController = null; exporting.value = false; exportProgress.show = false } }

frontend/src/views/user/UsageView.vue

@@ -342,8 +342,8 @@
       >
         <div class="space-y-1.5">
           <!-- Token Breakdown -->
-          <div class="mb-2 border-b border-gray-700 pb-1.5">
+          <div>
-            <div class="text-xs font-semibold text-gray-300 mb-1">Token 明细</div>
+            <div class="text-xs font-semibold text-gray-300 mb-1">{{ t('usage.tokenDetails') }}</div>
             <div v-if="tokenTooltipData && tokenTooltipData.input_tokens > 0" class="flex items-center justify-between gap-4">
               <span class="text-gray-400">{{ t('admin.usage.inputTokens') }}</span>
               <span class="font-medium text-white">{{ tokenTooltipData.input_tokens.toLocaleString() }}</span>
@@ -389,6 +389,27 @@
         class="whitespace-nowrap rounded-lg border border-gray-700 bg-gray-900 px-3 py-2.5 text-xs text-white shadow-xl dark:border-gray-600 dark:bg-gray-800"
       >
         <div class="space-y-1.5">
+          <!-- Cost Breakdown -->
+          <div class="mb-2 border-b border-gray-700 pb-1.5">
+            <div class="text-xs font-semibold text-gray-300 mb-1">{{ t('usage.costDetails') }}</div>
+            <div v-if="tooltipData && tooltipData.input_cost > 0" class="flex items-center justify-between gap-4">
+              <span class="text-gray-400">{{ t('admin.usage.inputCost') }}</span>
+              <span class="font-medium text-white">${{ tooltipData.input_cost.toFixed(6) }}</span>
+            </div>
+            <div v-if="tooltipData && tooltipData.output_cost > 0" class="flex items-center justify-between gap-4">
+              <span class="text-gray-400">{{ t('admin.usage.outputCost') }}</span>
+              <span class="font-medium text-white">${{ tooltipData.output_cost.toFixed(6) }}</span>
+            </div>
+            <div v-if="tooltipData && tooltipData.cache_creation_cost > 0" class="flex items-center justify-between gap-4">
+              <span class="text-gray-400">{{ t('admin.usage.cacheCreationCost') }}</span>
+              <span class="font-medium text-white">${{ tooltipData.cache_creation_cost.toFixed(6) }}</span>
+            </div>
+            <div v-if="tooltipData && tooltipData.cache_read_cost > 0" class="flex items-center justify-between gap-4">
+              <span class="text-gray-400">{{ t('admin.usage.cacheReadCost') }}</span>
+              <span class="font-medium text-white">${{ tooltipData.cache_read_cost.toFixed(6) }}</span>
+            </div>
+          </div>
+          <!-- Rate and Summary -->
           <div class="flex items-center justify-between gap-6">
             <span class="text-gray-400">{{ t('usage.rate') }}</span>
             <span class="font-semibold text-blue-400"