merge: resolve conflicts with main

backend/cmd/server/wire_gen.go

@@ -118,7 +118,8 @@ func initializeApplication(buildInfo handler.BuildInfo) (*Application, error) {
 	concurrencyCache := repository.ProvideConcurrencyCache(redisClient, configConfig)
 	concurrencyService := service.ProvideConcurrencyService(concurrencyCache, accountRepository, configConfig)
 	crsSyncService := service.NewCRSSyncService(accountRepository, proxyRepository, oAuthService, openAIOAuthService, geminiOAuthService, configConfig)
-	accountHandler := admin.NewAccountHandler(adminService, oAuthService, openAIOAuthService, geminiOAuthService, antigravityOAuthService, rateLimitService, accountUsageService, accountTestService, concurrencyService, crsSyncService)
+	sessionLimitCache := repository.ProvideSessionLimitCache(redisClient, configConfig)
+	accountHandler := admin.NewAccountHandler(adminService, oAuthService, openAIOAuthService, geminiOAuthService, antigravityOAuthService, rateLimitService, accountUsageService, accountTestService, concurrencyService, crsSyncService, sessionLimitCache)
 	oAuthHandler := admin.NewOAuthHandler(oAuthService)
 	openAIOAuthHandler := admin.NewOpenAIOAuthHandler(openAIOAuthService, adminService)
 	geminiOAuthHandler := admin.NewGeminiOAuthHandler(geminiOAuthService)
@@ -140,7 +141,7 @@ func initializeApplication(buildInfo handler.BuildInfo) (*Application, error) {
 	identityService := service.NewIdentityService(identityCache)
 	deferredService := service.ProvideDeferredService(accountRepository, timingWheelService)
 	claudeTokenProvider := service.NewClaudeTokenProvider(accountRepository, geminiTokenCache, oAuthService)
-	gatewayService := service.NewGatewayService(accountRepository, groupRepository, usageLogRepository, userRepository, userSubscriptionRepository, gatewayCache, configConfig, schedulerSnapshotService, concurrencyService, billingService, rateLimitService, billingCacheService, identityService, httpUpstream, deferredService, claudeTokenProvider)
+	gatewayService := service.NewGatewayService(accountRepository, groupRepository, usageLogRepository, userRepository, userSubscriptionRepository, gatewayCache, configConfig, schedulerSnapshotService, concurrencyService, billingService, rateLimitService, billingCacheService, identityService, httpUpstream, deferredService, claudeTokenProvider, sessionLimitCache)
 	openAITokenProvider := service.NewOpenAITokenProvider(accountRepository, geminiTokenCache, openAIOAuthService)
 	openAIGatewayService := service.NewOpenAIGatewayService(accountRepository, usageLogRepository, userRepository, userSubscriptionRepository, gatewayCache, configConfig, schedulerSnapshotService, concurrencyService, billingService, rateLimitService, billingCacheService, httpUpstream, deferredService, openAITokenProvider)
 	geminiMessagesCompatService := service.NewGeminiMessagesCompatService(accountRepository, groupRepository, gatewayCache, schedulerSnapshotService, geminiTokenProvider, rateLimitService, httpUpstream, antigravityGatewayService, configConfig)

backend/internal/config/config.go

@@ -234,6 +234,10 @@ type GatewayConfig struct {
 	// ConcurrencySlotTTLMinutes: 并发槽位过期时间（分钟）
 	// 应大于最长 LLM 请求时间，防止请求完成前槽位过期
 	ConcurrencySlotTTLMinutes int `mapstructure:"concurrency_slot_ttl_minutes"`
+	// SessionIdleTimeoutMinutes: 会话空闲超时时间（分钟），默认 5 分钟
+	// 用于 Anthropic OAuth/SetupToken 账号的会话数量限制功能
+	// 空闲超过此时间的会话将被自动释放
+	SessionIdleTimeoutMinutes int `mapstructure:"session_idle_timeout_minutes"`
 
 	// StreamDataIntervalTimeout: 流数据间隔超时（秒），0表示禁用
 	StreamDataIntervalTimeout int `mapstructure:"stream_data_interval_timeout"`

backend/internal/handler/admin/account_handler.go

@@ -44,6 +44,7 @@ type AccountHandler struct {
 	accountTestService      *service.AccountTestService
 	concurrencyService      *service.ConcurrencyService
 	crsSyncService          *service.CRSSyncService
+	sessionLimitCache       service.SessionLimitCache
 }
 
 // NewAccountHandler creates a new admin account handler
@@ -58,6 +59,7 @@ func NewAccountHandler(
 	accountTestService *service.AccountTestService,
 	concurrencyService *service.ConcurrencyService,
 	crsSyncService *service.CRSSyncService,
+	sessionLimitCache service.SessionLimitCache,
 ) *AccountHandler {
 	return &AccountHandler{
 		adminService:            adminService,
@@ -70,6 +72,7 @@ func NewAccountHandler(
 		accountTestService:      accountTestService,
 		concurrencyService:      concurrencyService,
 		crsSyncService:          crsSyncService,
+		sessionLimitCache:       sessionLimitCache,
 	}
 }
 
@@ -130,6 +133,9 @@ type BulkUpdateAccountsRequest struct {
 type AccountWithConcurrency struct {
 	*dto.Account
 	CurrentConcurrency int `json:"current_concurrency"`
+	// 以下字段仅对 Anthropic OAuth/SetupToken 账号有效，且仅在启用相应功能时返回
+	CurrentWindowCost *float64 `json:"current_window_cost,omitempty"` // 当前窗口费用
+	ActiveSessions    *int     `json:"active_sessions,omitempty"`     // 当前活跃会话数
 }
 
 // List handles listing all accounts with pagination
@@ -164,13 +170,89 @@ func (h *AccountHandler) List(c *gin.Context) {
 		concurrencyCounts = make(map[int64]int)
 	}
 
+	// 识别需要查询窗口费用和会话数的账号（Anthropic OAuth/SetupToken 且启用了相应功能）
+	windowCostAccountIDs := make([]int64, 0)
+	sessionLimitAccountIDs := make([]int64, 0)
+	for i := range accounts {
+		acc := &accounts[i]
+		if acc.IsAnthropicOAuthOrSetupToken() {
+			if acc.GetWindowCostLimit() > 0 {
+				windowCostAccountIDs = append(windowCostAccountIDs, acc.ID)
+			}
+			if acc.GetMaxSessions() > 0 {
+				sessionLimitAccountIDs = append(sessionLimitAccountIDs, acc.ID)
+			}
+		}
+	}
+
+	// 并行获取窗口费用和活跃会话数
+	var windowCosts map[int64]float64
+	var activeSessions map[int64]int
+
+	// 获取活跃会话数（批量查询）
+	if len(sessionLimitAccountIDs) > 0 && h.sessionLimitCache != nil {
+		activeSessions, _ = h.sessionLimitCache.GetActiveSessionCountBatch(c.Request.Context(), sessionLimitAccountIDs)
+		if activeSessions == nil {
+			activeSessions = make(map[int64]int)
+		}
+	}
+
+	// 获取窗口费用（并行查询）
+	if len(windowCostAccountIDs) > 0 {
+		windowCosts = make(map[int64]float64)
+		var mu sync.Mutex
+		g, gctx := errgroup.WithContext(c.Request.Context())
+		g.SetLimit(10) // 限制并发数
+
+		for i := range accounts {
+			acc := &accounts[i]
+			if !acc.IsAnthropicOAuthOrSetupToken() || acc.GetWindowCostLimit() <= 0 {
+				continue
+			}
+			accCopy := acc // 闭包捕获
+			g.Go(func() error {
+				var startTime time.Time
+				if accCopy.SessionWindowStart != nil {
+					startTime = *accCopy.SessionWindowStart
+				} else {
+					startTime = time.Now().Add(-5 * time.Hour)
+				}
+				stats, err := h.accountUsageService.GetAccountWindowStats(gctx, accCopy.ID, startTime)
+				if err == nil && stats != nil {
+					mu.Lock()
+					windowCosts[accCopy.ID] = stats.StandardCost // 使用标准费用
+					mu.Unlock()
+				}
+				return nil // 不返回错误，允许部分失败
+			})
+		}
+		_ = g.Wait()
+	}
+
 	// Build response with concurrency info
 	result := make([]AccountWithConcurrency, len(accounts))
 	for i := range accounts {
-		result[i] = AccountWithConcurrency{
+		acc := &accounts[i]
-			Account:            dto.AccountFromService(&accounts[i]),
+		item := AccountWithConcurrency{
-			CurrentConcurrency: concurrencyCounts[accounts[i].ID],
+			Account:            dto.AccountFromService(acc),
+			CurrentConcurrency: concurrencyCounts[acc.ID],
 		}
+
+		// 添加窗口费用（仅当启用时）
+		if windowCosts != nil {
+			if cost, ok := windowCosts[acc.ID]; ok {
+				item.CurrentWindowCost = &cost
+			}
+		}
+
+		// 添加活跃会话数（仅当启用时）
+		if activeSessions != nil {
+			if count, ok := activeSessions[acc.ID]; ok {
+				item.ActiveSessions = &count
+			}
+		}
+
+		result[i] = item
 	}
 
 	response.Paginated(c, result, total, page, pageSize)

backend/internal/handler/dto/mappers.go

@@ -116,7 +116,7 @@ func AccountFromServiceShallow(a *service.Account) *Account {
 	if a == nil {
 		return nil
 	}
-	return &Account{
+	out := &Account{
 		ID:                      a.ID,
 		Name:                    a.Name,
 		Notes:                   a.Notes,
@@ -146,6 +146,24 @@ func AccountFromServiceShallow(a *service.Account) *Account {
 		SessionWindowStatus:     a.SessionWindowStatus,
 		GroupIDs:                a.GroupIDs,
 	}
+
+	// 提取 5h 窗口费用控制和会话数量控制配置（仅 Anthropic OAuth/SetupToken 账号有效）
+	if a.IsAnthropicOAuthOrSetupToken() {
+		if limit := a.GetWindowCostLimit(); limit > 0 {
+			out.WindowCostLimit = &limit
+		}
+		if reserve := a.GetWindowCostStickyReserve(); reserve > 0 {
+			out.WindowCostStickyReserve = &reserve
+		}
+		if maxSessions := a.GetMaxSessions(); maxSessions > 0 {
+			out.MaxSessions = &maxSessions
+		}
+		if idleTimeout := a.GetSessionIdleTimeoutMinutes(); idleTimeout > 0 {
+			out.SessionIdleTimeoutMin = &idleTimeout
+		}
+	}
+
+	return out
 }
 
 func AccountFromService(a *service.Account) *Account {

backend/internal/handler/dto/types.go

@@ -102,6 +102,16 @@ type Account struct {
 	SessionWindowEnd    *time.Time `json:"session_window_end"`
 	SessionWindowStatus string     `json:"session_window_status"`
 
+	// 5h窗口费用控制（仅 Anthropic OAuth/SetupToken 账号有效）
+	// 从 extra 字段提取，方便前端显示和编辑
+	WindowCostLimit         *float64 `json:"window_cost_limit,omitempty"`
+	WindowCostStickyReserve *float64 `json:"window_cost_sticky_reserve,omitempty"`
+
+	// 会话数量控制（仅 Anthropic OAuth/SetupToken 账号有效）
+	// 从 extra 字段提取，方便前端显示和编辑
+	MaxSessions           *int `json:"max_sessions,omitempty"`
+	SessionIdleTimeoutMin *int `json:"session_idle_timeout_minutes,omitempty"`
+
 	Proxy         *Proxy         `json:"proxy,omitempty"`
 	AccountGroups []AccountGroup `json:"account_groups,omitempty"`
 

backend/internal/handler/gateway_handler.go

@@ -185,7 +185,7 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 		lastFailoverStatus := 0
 
 		for {
-			selection, err := h.gatewayService.SelectAccountWithLoadAwareness(c.Request.Context(), apiKey.GroupID, sessionKey, reqModel, failedAccountIDs)
+			selection, err := h.gatewayService.SelectAccountWithLoadAwareness(c.Request.Context(), apiKey.GroupID, sessionKey, reqModel, failedAccountIDs, "") // Gemini 不使用会话限制
 			if err != nil {
 				if len(failedAccountIDs) == 0 {
 					h.handleStreamingAwareError(c, http.StatusServiceUnavailable, "api_error", "No available accounts: "+err.Error(), streamStarted)
@@ -320,7 +320,7 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 
 	for {
 		// 选择支持该模型的账号
-		selection, err := h.gatewayService.SelectAccountWithLoadAwareness(c.Request.Context(), apiKey.GroupID, sessionKey, reqModel, failedAccountIDs)
+		selection, err := h.gatewayService.SelectAccountWithLoadAwareness(c.Request.Context(), apiKey.GroupID, sessionKey, reqModel, failedAccountIDs, parsedReq.MetadataUserID)
 		if err != nil {
 			if len(failedAccountIDs) == 0 {
 				h.handleStreamingAwareError(c, http.StatusServiceUnavailable, "api_error", "No available accounts: "+err.Error(), streamStarted)

backend/internal/handler/gemini_v1beta_handler.go

@@ -226,7 +226,7 @@ func (h *GatewayHandler) GeminiV1BetaModels(c *gin.Context) {
 	lastFailoverStatus := 0
 
 	for {
-		selection, err := h.gatewayService.SelectAccountWithLoadAwareness(c.Request.Context(), apiKey.GroupID, sessionKey, modelName, failedAccountIDs)
+		selection, err := h.gatewayService.SelectAccountWithLoadAwareness(c.Request.Context(), apiKey.GroupID, sessionKey, modelName, failedAccountIDs, "") // Gemini 不使用会话限制
 		if err != nil {
 			if len(failedAccountIDs) == 0 {
 				googleError(c, http.StatusServiceUnavailable, "No available Gemini accounts: "+err.Error())

backend/internal/repository/session_limit_cache.go

@@ -0,0 +1,321 @@
+package repository
+
+import (
+	"context"
+	"fmt"
+	"strconv"
+	"time"
+
+	"github.com/Wei-Shaw/sub2api/internal/service"
+	"github.com/redis/go-redis/v9"
+)
+
+// 会话限制缓存常量定义
+//
+// 设计说明：
+// 使用 Redis 有序集合（Sorted Set）跟踪每个账号的活跃会话：
+// - Key: session_limit:account:{accountID}
+// - Member: sessionUUID（从 metadata.user_id 中提取）
+// - Score: Unix 时间戳（会话最后活跃时间）
+//
+// 通过 ZREMRANGEBYSCORE 自动清理过期会话，无需手动管理 TTL
+const (
+	// 会话限制键前缀
+	// 格式: session_limit:account:{accountID}
+	sessionLimitKeyPrefix = "session_limit:account:"
+
+	// 窗口费用缓存键前缀
+	// 格式: window_cost:account:{accountID}
+	windowCostKeyPrefix = "window_cost:account:"
+
+	// 窗口费用缓存 TTL（30秒）
+	windowCostCacheTTL = 30 * time.Second
+)
+
+var (
+	// registerSessionScript 注册会话活动
+	// 使用 Redis TIME 命令获取服务器时间，避免多实例时钟不同步
+	// KEYS[1] = session_limit:account:{accountID}
+	// ARGV[1] = maxSessions
+	// ARGV[2] = idleTimeout（秒）
+	// ARGV[3] = sessionUUID
+	// 返回: 1 = 允许, 0 = 拒绝
+	registerSessionScript = redis.NewScript(`
+		local key = KEYS[1]
+		local maxSessions = tonumber(ARGV[1])
+		local idleTimeout = tonumber(ARGV[2])
+		local sessionUUID = ARGV[3]
+
+		-- 使用 Redis 服务器时间，确保多实例时钟一致
+		local timeResult = redis.call('TIME')
+		local now = tonumber(timeResult[1])
+		local expireBefore = now - idleTimeout
+
+		-- 清理过期会话
+		redis.call('ZREMRANGEBYSCORE', key, '-inf', expireBefore)
+
+		-- 检查会话是否已存在（支持刷新时间戳）
+		local exists = redis.call('ZSCORE', key, sessionUUID)
+		if exists ~= false then
+			-- 会话已存在，刷新时间戳
+			redis.call('ZADD', key, now, sessionUUID)
+			redis.call('EXPIRE', key, idleTimeout + 60)
+			return 1
+		end
+
+		-- 检查是否达到会话数量上限
+		local count = redis.call('ZCARD', key)
+		if count < maxSessions then
+			-- 未达上限，添加新会话
+			redis.call('ZADD', key, now, sessionUUID)
+			redis.call('EXPIRE', key, idleTimeout + 60)
+			return 1
+		end
+
+		-- 达到上限，拒绝新会话
+		return 0
+	`)
+
+	// refreshSessionScript 刷新会话时间戳
+	// KEYS[1] = session_limit:account:{accountID}
+	// ARGV[1] = idleTimeout（秒）
+	// ARGV[2] = sessionUUID
+	refreshSessionScript = redis.NewScript(`
+		local key = KEYS[1]
+		local idleTimeout = tonumber(ARGV[1])
+		local sessionUUID = ARGV[2]
+
+		local timeResult = redis.call('TIME')
+		local now = tonumber(timeResult[1])
+
+		-- 检查会话是否存在
+		local exists = redis.call('ZSCORE', key, sessionUUID)
+		if exists ~= false then
+			redis.call('ZADD', key, now, sessionUUID)
+			redis.call('EXPIRE', key, idleTimeout + 60)
+		end
+		return 1
+	`)
+
+	// getActiveSessionCountScript 获取活跃会话数
+	// KEYS[1] = session_limit:account:{accountID}
+	// ARGV[1] = idleTimeout（秒）
+	getActiveSessionCountScript = redis.NewScript(`
+		local key = KEYS[1]
+		local idleTimeout = tonumber(ARGV[1])
+
+		local timeResult = redis.call('TIME')
+		local now = tonumber(timeResult[1])
+		local expireBefore = now - idleTimeout
+
+		-- 清理过期会话
+		redis.call('ZREMRANGEBYSCORE', key, '-inf', expireBefore)
+
+		return redis.call('ZCARD', key)
+	`)
+
+	// isSessionActiveScript 检查会话是否活跃
+	// KEYS[1] = session_limit:account:{accountID}
+	// ARGV[1] = idleTimeout（秒）
+	// ARGV[2] = sessionUUID
+	isSessionActiveScript = redis.NewScript(`
+		local key = KEYS[1]
+		local idleTimeout = tonumber(ARGV[1])
+		local sessionUUID = ARGV[2]
+
+		local timeResult = redis.call('TIME')
+		local now = tonumber(timeResult[1])
+		local expireBefore = now - idleTimeout
+
+		-- 获取会话的时间戳
+		local score = redis.call('ZSCORE', key, sessionUUID)
+		if score == false then
+			return 0
+		end
+
+		-- 检查是否过期
+		if tonumber(score) <= expireBefore then
+			return 0
+		end
+
+		return 1
+	`)
+)
+
+type sessionLimitCache struct {
+	rdb                *redis.Client
+	defaultIdleTimeout time.Duration // 默认空闲超时（用于 GetActiveSessionCount）
+}
+
+// NewSessionLimitCache 创建会话限制缓存
+// defaultIdleTimeoutMinutes: 默认空闲超时时间（分钟），用于无参数查询
+func NewSessionLimitCache(rdb *redis.Client, defaultIdleTimeoutMinutes int) service.SessionLimitCache {
+	if defaultIdleTimeoutMinutes <= 0 {
+		defaultIdleTimeoutMinutes = 5 // 默认 5 分钟
+	}
+	return &sessionLimitCache{
+		rdb:                rdb,
+		defaultIdleTimeout: time.Duration(defaultIdleTimeoutMinutes) * time.Minute,
+	}
+}
+
+// sessionLimitKey 生成会话限制的 Redis 键
+func sessionLimitKey(accountID int64) string {
+	return fmt.Sprintf("%s%d", sessionLimitKeyPrefix, accountID)
+}
+
+// windowCostKey 生成窗口费用缓存的 Redis 键
+func windowCostKey(accountID int64) string {
+	return fmt.Sprintf("%s%d", windowCostKeyPrefix, accountID)
+}
+
+// RegisterSession 注册会话活动
+func (c *sessionLimitCache) RegisterSession(ctx context.Context, accountID int64, sessionUUID string, maxSessions int, idleTimeout time.Duration) (bool, error) {
+	if sessionUUID == "" || maxSessions <= 0 {
+		return true, nil // 无效参数，默认允许
+	}
+
+	key := sessionLimitKey(accountID)
+	idleTimeoutSeconds := int(idleTimeout.Seconds())
+	if idleTimeoutSeconds <= 0 {
+		idleTimeoutSeconds = int(c.defaultIdleTimeout.Seconds())
+	}
+
+	result, err := registerSessionScript.Run(ctx, c.rdb, []string{key}, maxSessions, idleTimeoutSeconds, sessionUUID).Int()
+	if err != nil {
+		return true, err // 失败开放：缓存错误时允许请求通过
+	}
+	return result == 1, nil
+}
+
+// RefreshSession 刷新会话时间戳
+func (c *sessionLimitCache) RefreshSession(ctx context.Context, accountID int64, sessionUUID string, idleTimeout time.Duration) error {
+	if sessionUUID == "" {
+		return nil
+	}
+
+	key := sessionLimitKey(accountID)
+	idleTimeoutSeconds := int(idleTimeout.Seconds())
+	if idleTimeoutSeconds <= 0 {
+		idleTimeoutSeconds = int(c.defaultIdleTimeout.Seconds())
+	}
+
+	_, err := refreshSessionScript.Run(ctx, c.rdb, []string{key}, idleTimeoutSeconds, sessionUUID).Result()
+	return err
+}
+
+// GetActiveSessionCount 获取活跃会话数
+func (c *sessionLimitCache) GetActiveSessionCount(ctx context.Context, accountID int64) (int, error) {
+	key := sessionLimitKey(accountID)
+	idleTimeoutSeconds := int(c.defaultIdleTimeout.Seconds())
+
+	result, err := getActiveSessionCountScript.Run(ctx, c.rdb, []string{key}, idleTimeoutSeconds).Int()
+	if err != nil {
+		return 0, err
+	}
+	return result, nil
+}
+
+// GetActiveSessionCountBatch 批量获取多个账号的活跃会话数
+func (c *sessionLimitCache) GetActiveSessionCountBatch(ctx context.Context, accountIDs []int64) (map[int64]int, error) {
+	if len(accountIDs) == 0 {
+		return make(map[int64]int), nil
+	}
+
+	results := make(map[int64]int, len(accountIDs))
+
+	// 使用 pipeline 批量执行
+	pipe := c.rdb.Pipeline()
+	idleTimeoutSeconds := int(c.defaultIdleTimeout.Seconds())
+
+	cmds := make(map[int64]*redis.Cmd, len(accountIDs))
+	for _, accountID := range accountIDs {
+		key := sessionLimitKey(accountID)
+		cmds[accountID] = getActiveSessionCountScript.Run(ctx, pipe, []string{key}, idleTimeoutSeconds)
+	}
+
+	// 执行 pipeline，即使部分失败也尝试获取成功的结果
+	_, _ = pipe.Exec(ctx)
+
+	for accountID, cmd := range cmds {
+		if result, err := cmd.Int(); err == nil {
+			results[accountID] = result
+		}
+	}
+
+	return results, nil
+}
+
+// IsSessionActive 检查会话是否活跃
+func (c *sessionLimitCache) IsSessionActive(ctx context.Context, accountID int64, sessionUUID string) (bool, error) {
+	if sessionUUID == "" {
+		return false, nil
+	}
+
+	key := sessionLimitKey(accountID)
+	idleTimeoutSeconds := int(c.defaultIdleTimeout.Seconds())
+
+	result, err := isSessionActiveScript.Run(ctx, c.rdb, []string{key}, idleTimeoutSeconds, sessionUUID).Int()
+	if err != nil {
+		return false, err
+	}
+	return result == 1, nil
+}
+
+// ========== 5h窗口费用缓存实现 ==========
+
+// GetWindowCost 获取缓存的窗口费用
+func (c *sessionLimitCache) GetWindowCost(ctx context.Context, accountID int64) (float64, bool, error) {
+	key := windowCostKey(accountID)
+	val, err := c.rdb.Get(ctx, key).Float64()
+	if err == redis.Nil {
+		return 0, false, nil // 缓存未命中
+	}
+	if err != nil {
+		return 0, false, err
+	}
+	return val, true, nil
+}
+
+// SetWindowCost 设置窗口费用缓存
+func (c *sessionLimitCache) SetWindowCost(ctx context.Context, accountID int64, cost float64) error {
+	key := windowCostKey(accountID)
+	return c.rdb.Set(ctx, key, cost, windowCostCacheTTL).Err()
+}
+
+// GetWindowCostBatch 批量获取窗口费用缓存
+func (c *sessionLimitCache) GetWindowCostBatch(ctx context.Context, accountIDs []int64) (map[int64]float64, error) {
+	if len(accountIDs) == 0 {
+		return make(map[int64]float64), nil
+	}
+
+	// 构建批量查询的 keys
+	keys := make([]string, len(accountIDs))
+	for i, accountID := range accountIDs {
+		keys[i] = windowCostKey(accountID)
+	}
+
+	// 使用 MGET 批量获取
+	vals, err := c.rdb.MGet(ctx, keys...).Result()
+	if err != nil {
+		return nil, err
+	}
+
+	results := make(map[int64]float64, len(accountIDs))
+	for i, val := range vals {
+		if val == nil {
+			continue // 缓存未命中
+		}
+		// 尝试解析为 float64
+		switch v := val.(type) {
+		case string:
+			if cost, err := strconv.ParseFloat(v, 64); err == nil {
+				results[accountIDs[i]] = cost
+			}
+		case float64:
+			results[accountIDs[i]] = v
+		}
+	}
+
+	return results, nil
+}

backend/internal/repository/wire.go

@@ -37,6 +37,16 @@ func ProvidePricingRemoteClient(cfg *config.Config) service.PricingRemoteClient
 	return NewPricingRemoteClient(cfg.Update.ProxyURL)
 }
 
+// ProvideSessionLimitCache 创建会话限制缓存
+// 用于 Anthropic OAuth/SetupToken 账号的并发会话数量控制
+func ProvideSessionLimitCache(rdb *redis.Client, cfg *config.Config) service.SessionLimitCache {
+	defaultIdleTimeoutMinutes := 5 // 默认 5 分钟空闲超时
+	if cfg != nil && cfg.Gateway.SessionIdleTimeoutMinutes > 0 {
+		defaultIdleTimeoutMinutes = cfg.Gateway.SessionIdleTimeoutMinutes
+	}
+	return NewSessionLimitCache(rdb, defaultIdleTimeoutMinutes)
+}
+
 // ProviderSet is the Wire provider set for all repositories
 var ProviderSet = wire.NewSet(
 	NewUserRepository,
@@ -61,6 +71,7 @@ var ProviderSet = wire.NewSet(
 	NewTempUnschedCache,
 	NewTimeoutCounterCache,
 	ProvideConcurrencyCache,
+	ProvideSessionLimitCache,
 	NewDashboardCache,
 	NewEmailCache,
 	NewIdentityCache,

backend/internal/server/api_contract_test.go

@@ -441,7 +441,7 @@ func newContractDeps(t *testing.T) *contractDeps {
 	apiKeyHandler := handler.NewAPIKeyHandler(apiKeyService)
 	usageHandler := handler.NewUsageHandler(usageService, apiKeyService)
 	adminSettingHandler := adminhandler.NewSettingHandler(settingService, nil, nil, nil)
-	adminAccountHandler := adminhandler.NewAccountHandler(adminService, nil, nil, nil, nil, nil, nil, nil, nil, nil)
+	adminAccountHandler := adminhandler.NewAccountHandler(adminService, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil)
 
 	jwtAuth := func(c *gin.Context) {
 		c.Set(string(middleware.ContextKeyUser), middleware.AuthSubject{

backend/internal/service/account.go

@@ -573,3 +573,141 @@ func (a *Account) IsMixedSchedulingEnabled() bool {
 	}
 	return false
 }
+
+// WindowCostSchedulability 窗口费用调度状态
+type WindowCostSchedulability int
+
+const (
+	// WindowCostSchedulable 可正常调度
+	WindowCostSchedulable WindowCostSchedulability = iota
+	// WindowCostStickyOnly 仅允许粘性会话
+	WindowCostStickyOnly
+	// WindowCostNotSchedulable 完全不可调度
+	WindowCostNotSchedulable
+)
+
+// IsAnthropicOAuthOrSetupToken 判断是否为 Anthropic OAuth 或 SetupToken 类型账号
+// 仅这两类账号支持 5h 窗口额度控制和会话数量控制
+func (a *Account) IsAnthropicOAuthOrSetupToken() bool {
+	return a.Platform == PlatformAnthropic && (a.Type == AccountTypeOAuth || a.Type == AccountTypeSetupToken)
+}
+
+// GetWindowCostLimit 获取 5h 窗口费用阈值（美元）
+// 返回 0 表示未启用
+func (a *Account) GetWindowCostLimit() float64 {
+	if a.Extra == nil {
+		return 0
+	}
+	if v, ok := a.Extra["window_cost_limit"]; ok {
+		return parseExtraFloat64(v)
+	}
+	return 0
+}
+
+// GetWindowCostStickyReserve 获取粘性会话预留额度（美元）
+// 默认值为 10
+func (a *Account) GetWindowCostStickyReserve() float64 {
+	if a.Extra == nil {
+		return 10.0
+	}
+	if v, ok := a.Extra["window_cost_sticky_reserve"]; ok {
+		val := parseExtraFloat64(v)
+		if val > 0 {
+			return val
+		}
+	}
+	return 10.0
+}
+
+// GetMaxSessions 获取最大并发会话数
+// 返回 0 表示未启用
+func (a *Account) GetMaxSessions() int {
+	if a.Extra == nil {
+		return 0
+	}
+	if v, ok := a.Extra["max_sessions"]; ok {
+		return parseExtraInt(v)
+	}
+	return 0
+}
+
+// GetSessionIdleTimeoutMinutes 获取会话空闲超时分钟数
+// 默认值为 5 分钟
+func (a *Account) GetSessionIdleTimeoutMinutes() int {
+	if a.Extra == nil {
+		return 5
+	}
+	if v, ok := a.Extra["session_idle_timeout_minutes"]; ok {
+		val := parseExtraInt(v)
+		if val > 0 {
+			return val
+		}
+	}
+	return 5
+}
+
+// CheckWindowCostSchedulability 根据当前窗口费用检查调度状态
+// - 费用 < 阈值: WindowCostSchedulable（可正常调度）
+// - 费用 >= 阈值 且 < 阈值+预留: WindowCostStickyOnly（仅粘性会话）
+// - 费用 >= 阈值+预留: WindowCostNotSchedulable（不可调度）
+func (a *Account) CheckWindowCostSchedulability(currentWindowCost float64) WindowCostSchedulability {
+	limit := a.GetWindowCostLimit()
+	if limit <= 0 {
+		return WindowCostSchedulable
+	}
+
+	if currentWindowCost < limit {
+		return WindowCostSchedulable
+	}
+
+	stickyReserve := a.GetWindowCostStickyReserve()
+	if currentWindowCost < limit+stickyReserve {
+		return WindowCostStickyOnly
+	}
+
+	return WindowCostNotSchedulable
+}
+
+// parseExtraFloat64 从 extra 字段解析 float64 值
+func parseExtraFloat64(value any) float64 {
+	switch v := value.(type) {
+	case float64:
+		return v
+	case float32:
+		return float64(v)
+	case int:
+		return float64(v)
+	case int64:
+		return float64(v)
+	case json.Number:
+		if f, err := v.Float64(); err == nil {
+			return f
+		}
+	case string:
+		if f, err := strconv.ParseFloat(strings.TrimSpace(v), 64); err == nil {
+			return f
+		}
+	}
+	return 0
+}
+
+// parseExtraInt 从 extra 字段解析 int 值
+func parseExtraInt(value any) int {
+	switch v := value.(type) {
+	case int:
+		return v
+	case int64:
+		return int(v)
+	case float64:
+		return int(v)
+	case json.Number:
+		if i, err := v.Int64(); err == nil {
+			return int(i)
+		}
+	case string:
+		if i, err := strconv.Atoi(strings.TrimSpace(v)); err == nil {
+			return i
+		}
+	}
+	return 0
+}

backend/internal/service/account_usage_service.go

@@ -575,3 +575,9 @@ func buildGeminiUsageProgress(used, limit int64, resetAt time.Time, tokens int64
 		},
 	}
 }
+
+// GetAccountWindowStats 获取账号在指定时间窗口内的使用统计
+// 用于账号列表页面显示当前窗口费用
+func (s *AccountUsageService) GetAccountWindowStats(ctx context.Context, accountID int64, startTime time.Time) (*usagestats.AccountStats, error) {
+	return s.usageLogRepo.GetAccountWindowStats(ctx, accountID, startTime)
+}

backend/internal/service/gateway_multiplatform_test.go

@@ -1052,7 +1052,7 @@ func TestGatewayService_SelectAccountWithLoadAwareness(t *testing.T) {
 			concurrencyService: nil, // No concurrency service
 		}
 
-		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "", "claude-3-5-sonnet-20241022", nil)
+		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "", "claude-3-5-sonnet-20241022", nil, "")
 		require.NoError(t, err)
 		require.NotNil(t, result)
 		require.NotNil(t, result.Account)
@@ -1105,7 +1105,7 @@ func TestGatewayService_SelectAccountWithLoadAwareness(t *testing.T) {
 			concurrencyService: nil, // legacy path
 		}
 
-		result, err := svc.SelectAccountWithLoadAwareness(ctx, &groupID, sessionHash, "claude-b", nil)
+		result, err := svc.SelectAccountWithLoadAwareness(ctx, &groupID, sessionHash, "claude-b", nil, "")
 		require.NoError(t, err)
 		require.NotNil(t, result)
 		require.NotNil(t, result.Account)
@@ -1137,7 +1137,7 @@ func TestGatewayService_SelectAccountWithLoadAwareness(t *testing.T) {
 			concurrencyService: nil,
 		}
 
-		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "", "claude-3-5-sonnet-20241022", nil)
+		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "", "claude-3-5-sonnet-20241022", nil, "")
 		require.NoError(t, err)
 		require.NotNil(t, result)
 		require.NotNil(t, result.Account)
@@ -1169,7 +1169,7 @@ func TestGatewayService_SelectAccountWithLoadAwareness(t *testing.T) {
 		}
 
 		excludedIDs := map[int64]struct{}{1: {}}
-		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "", "claude-3-5-sonnet-20241022", excludedIDs)
+		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "", "claude-3-5-sonnet-20241022", excludedIDs, "")
 		require.NoError(t, err)
 		require.NotNil(t, result)
 		require.NotNil(t, result.Account)
@@ -1203,7 +1203,7 @@ func TestGatewayService_SelectAccountWithLoadAwareness(t *testing.T) {
 			concurrencyService: NewConcurrencyService(concurrencyCache),
 		}
 
-		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "sticky", "claude-3-5-sonnet-20241022", nil)
+		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "sticky", "claude-3-5-sonnet-20241022", nil, "")
 		require.NoError(t, err)
 		require.NotNil(t, result)
 		require.NotNil(t, result.Account)
@@ -1239,7 +1239,7 @@ func TestGatewayService_SelectAccountWithLoadAwareness(t *testing.T) {
 			concurrencyService: NewConcurrencyService(concurrencyCache),
 		}
 
-		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "sticky", "claude-3-5-sonnet-20241022", nil)
+		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "sticky", "claude-3-5-sonnet-20241022", nil, "")
 		require.NoError(t, err)
 		require.NotNil(t, result)
 		require.NotNil(t, result.Account)
@@ -1266,7 +1266,7 @@ func TestGatewayService_SelectAccountWithLoadAwareness(t *testing.T) {
 			concurrencyService: nil,
 		}
 
-		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "", "claude-3-5-sonnet-20241022", nil)
+		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "", "claude-3-5-sonnet-20241022", nil, "")
 		require.Error(t, err)
 		require.Nil(t, result)
 		require.Contains(t, err.Error(), "no available accounts")
@@ -1298,7 +1298,7 @@ func TestGatewayService_SelectAccountWithLoadAwareness(t *testing.T) {
 			concurrencyService: nil,
 		}
 
-		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "", "claude-3-5-sonnet-20241022", nil)
+		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "", "claude-3-5-sonnet-20241022", nil, "")
 		require.NoError(t, err)
 		require.NotNil(t, result)
 		require.NotNil(t, result.Account)
@@ -1331,7 +1331,7 @@ func TestGatewayService_SelectAccountWithLoadAwareness(t *testing.T) {
 			concurrencyService: nil,
 		}
 
-		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "", "claude-3-5-sonnet-20241022", nil)
+		result, err := svc.SelectAccountWithLoadAwareness(ctx, nil, "", "claude-3-5-sonnet-20241022", nil, "")
 		require.NoError(t, err)
 		require.NotNil(t, result)
 		require.NotNil(t, result.Account)

backend/internal/service/gateway_service.go

@@ -213,6 +213,7 @@ type GatewayService struct {
 	deferredService     *DeferredService
 	concurrencyService  *ConcurrencyService
 	claudeTokenProvider *ClaudeTokenProvider
+	sessionLimitCache   SessionLimitCache // 会话数量限制缓存（仅 Anthropic OAuth/SetupToken）
 }
 
 // NewGatewayService creates a new GatewayService
@@ -233,6 +234,7 @@ func NewGatewayService(
 	httpUpstream HTTPUpstream,
 	deferredService *DeferredService,
 	claudeTokenProvider *ClaudeTokenProvider,
+	sessionLimitCache SessionLimitCache,
 ) *GatewayService {
 	return &GatewayService{
 		accountRepo:         accountRepo,
@@ -251,6 +253,7 @@ func NewGatewayService(
 		httpUpstream:        httpUpstream,
 		deferredService:     deferredService,
 		claudeTokenProvider: claudeTokenProvider,
+		sessionLimitCache:   sessionLimitCache,
 	}
 }
 
@@ -816,8 +819,12 @@ func (s *GatewayService) SelectAccountForModelWithExclusions(ctx context.Context
 }
 
 // SelectAccountWithLoadAwareness selects account with load-awareness and wait plan.
-func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, groupID *int64, sessionHash string, requestedModel string, excludedIDs map[int64]struct{}) (*AccountSelectionResult, error) {
+// metadataUserID: 原始 metadata.user_id 字段（用于提取会话 UUID 进行会话数量限制）
+func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, groupID *int64, sessionHash string, requestedModel string, excludedIDs map[int64]struct{}, metadataUserID string) (*AccountSelectionResult, error) {
 	cfg := s.schedulingConfig()
+	// 提取会话 UUID（用于会话数量限制）
+	sessionUUID := extractSessionUUID(metadataUserID)
+
 	var stickyAccountID int64
 	if sessionHash != "" && s.cache != nil {
 		if accountID, err := s.cache.GetSessionAccountID(ctx, derefGroupID(groupID), sessionHash); err == nil {
@@ -936,7 +943,7 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 	if len(routingAccountIDs) > 0 && s.concurrencyService != nil {
 		// 1. 过滤出路由列表中可调度的账号
 		var routingCandidates []*Account
-		var filteredExcluded, filteredMissing, filteredUnsched, filteredPlatform, filteredModelScope, filteredModelMapping int
+		var filteredExcluded, filteredMissing, filteredUnsched, filteredPlatform, filteredModelScope, filteredModelMapping, filteredWindowCost int
 		for _, routingAccountID := range routingAccountIDs {
 			if isExcluded(routingAccountID) {
 				filteredExcluded++
@@ -963,13 +970,18 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 				filteredModelMapping++
 				continue
 			}
+			// 窗口费用检查（非粘性会话路径）
+			if !s.isAccountSchedulableForWindowCost(ctx, account, false) {
+				filteredWindowCost++
+				continue
+			}
 			routingCandidates = append(routingCandidates, account)
 		}
 
 		if s.debugModelRoutingEnabled() {
-			log.Printf("[ModelRoutingDebug] routed candidates: group_id=%v model=%s routed=%d candidates=%d filtered(excluded=%d missing=%d unsched=%d platform=%d model_scope=%d model_mapping=%d)",
+			log.Printf("[ModelRoutingDebug] routed candidates: group_id=%v model=%s routed=%d candidates=%d filtered(excluded=%d missing=%d unsched=%d platform=%d model_scope=%d model_mapping=%d window_cost=%d)",
 				derefGroupID(groupID), requestedModel, len(routingAccountIDs), len(routingCandidates),
-				filteredExcluded, filteredMissing, filteredUnsched, filteredPlatform, filteredModelScope, filteredModelMapping)
+				filteredExcluded, filteredMissing, filteredUnsched, filteredPlatform, filteredModelScope, filteredModelMapping, filteredWindowCost)
 		}
 
 		if len(routingCandidates) > 0 {
@@ -982,18 +994,25 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 						if stickyAccount.IsSchedulable() &&
 							s.isAccountAllowedForPlatform(stickyAccount, platform, useMixed) &&
 							stickyAccount.IsSchedulableForModel(requestedModel) &&
-							(requestedModel == "" || s.isModelSupportedByAccount(stickyAccount, requestedModel)) {
+							(requestedModel == "" || s.isModelSupportedByAccount(stickyAccount, requestedModel)) &&
+							s.isAccountSchedulableForWindowCost(ctx, stickyAccount, true) { // 粘性会话窗口费用检查
 							result, err := s.tryAcquireAccountSlot(ctx, stickyAccountID, stickyAccount.Concurrency)
 							if err == nil && result.Acquired {
-								_ = s.cache.RefreshSessionTTL(ctx, derefGroupID(groupID), sessionHash, stickySessionTTL)
+								// 会话数量限制检查
-								if s.debugModelRoutingEnabled() {
+								if !s.checkAndRegisterSession(ctx, stickyAccount, sessionUUID) {
-									log.Printf("[ModelRoutingDebug] routed sticky hit: group_id=%v model=%s session=%s account=%d", derefGroupID(groupID), requestedModel, shortSessionHash(sessionHash), stickyAccountID)
+									result.ReleaseFunc() // 释放槽位
+									// 继续到负载感知选择
+								} else {
+									_ = s.cache.RefreshSessionTTL(ctx, derefGroupID(groupID), sessionHash, stickySessionTTL)
+									if s.debugModelRoutingEnabled() {
+										log.Printf("[ModelRoutingDebug] routed sticky hit: group_id=%v model=%s session=%s account=%d", derefGroupID(groupID), requestedModel, shortSessionHash(sessionHash), stickyAccountID)
+									}
+									return &AccountSelectionResult{
+										Account:     stickyAccount,
+										Acquired:    true,
+										ReleaseFunc: result.ReleaseFunc,
+									}, nil
 								}
-								return &AccountSelectionResult{
-									Account:     stickyAccount,
-									Acquired:    true,
-									ReleaseFunc: result.ReleaseFunc,
-								}, nil
 							}
 
 							waitingCount, _ := s.concurrencyService.GetAccountWaitingCount(ctx, stickyAccountID)
@@ -1066,6 +1085,11 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 				for _, item := range routingAvailable {
 					result, err := s.tryAcquireAccountSlot(ctx, item.account.ID, item.account.Concurrency)
 					if err == nil && result.Acquired {
+						// 会话数量限制检查
+						if !s.checkAndRegisterSession(ctx, item.account, sessionUUID) {
+							result.ReleaseFunc() // 释放槽位，继续尝试下一个账号
+							continue
+						}
 						if sessionHash != "" && s.cache != nil {
 							_ = s.cache.SetSessionAccountID(ctx, derefGroupID(groupID), sessionHash, item.account.ID, stickySessionTTL)
 						}
@@ -1108,15 +1132,21 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 			if ok && s.isAccountInGroup(account, groupID) &&
 				s.isAccountAllowedForPlatform(account, platform, useMixed) &&
 				account.IsSchedulableForModel(requestedModel) &&
-				(requestedModel == "" || s.isModelSupportedByAccount(account, requestedModel)) {
+				(requestedModel == "" || s.isModelSupportedByAccount(account, requestedModel)) &&
+				s.isAccountSchedulableForWindowCost(ctx, account, true) { // 粘性会话窗口费用检查
 				result, err := s.tryAcquireAccountSlot(ctx, accountID, account.Concurrency)
 				if err == nil && result.Acquired {
-					_ = s.cache.RefreshSessionTTL(ctx, derefGroupID(groupID), sessionHash, stickySessionTTL)
+					// 会话数量限制检查
-					return &AccountSelectionResult{
+					if !s.checkAndRegisterSession(ctx, account, sessionUUID) {
-						Account:     account,
+						result.ReleaseFunc() // 释放槽位，继续到 Layer 2
-						Acquired:    true,
+					} else {
-						ReleaseFunc: result.ReleaseFunc,
+						_ = s.cache.RefreshSessionTTL(ctx, derefGroupID(groupID), sessionHash, stickySessionTTL)
-					}, nil
+						return &AccountSelectionResult{
+							Account:     account,
+							Acquired:    true,
+							ReleaseFunc: result.ReleaseFunc,
+						}, nil
+					}
 				}
 
 				waitingCount, _ := s.concurrencyService.GetAccountWaitingCount(ctx, accountID)
@@ -1157,6 +1187,10 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 		if requestedModel != "" && !s.isModelSupportedByAccount(acc, requestedModel) {
 			continue
 		}
+		// 窗口费用检查（非粘性会话路径）
+		if !s.isAccountSchedulableForWindowCost(ctx, acc, false) {
+			continue
+		}
 		candidates = append(candidates, acc)
 	}
 
@@ -1174,7 +1208,7 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 
 	loadMap, err := s.concurrencyService.GetAccountsLoadBatch(ctx, accountLoads)
 	if err != nil {
-		if result, ok := s.tryAcquireByLegacyOrder(ctx, candidates, groupID, sessionHash, preferOAuth); ok {
+		if result, ok := s.tryAcquireByLegacyOrder(ctx, candidates, groupID, sessionHash, preferOAuth, sessionUUID); ok {
 			return result, nil
 		}
 	} else {
@@ -1223,6 +1257,11 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 			for _, item := range available {
 				result, err := s.tryAcquireAccountSlot(ctx, item.account.ID, item.account.Concurrency)
 				if err == nil && result.Acquired {
+					// 会话数量限制检查
+					if !s.checkAndRegisterSession(ctx, item.account, sessionUUID) {
+						result.ReleaseFunc() // 释放槽位，继续尝试下一个账号
+						continue
+					}
 					if sessionHash != "" && s.cache != nil {
 						_ = s.cache.SetSessionAccountID(ctx, derefGroupID(groupID), sessionHash, item.account.ID, stickySessionTTL)
 					}
@@ -1252,13 +1291,18 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 	return nil, errors.New("no available accounts")
 }
 
-func (s *GatewayService) tryAcquireByLegacyOrder(ctx context.Context, candidates []*Account, groupID *int64, sessionHash string, preferOAuth bool) (*AccountSelectionResult, bool) {
+func (s *GatewayService) tryAcquireByLegacyOrder(ctx context.Context, candidates []*Account, groupID *int64, sessionHash string, preferOAuth bool, sessionUUID string) (*AccountSelectionResult, bool) {
 	ordered := append([]*Account(nil), candidates...)
 	sortAccountsByPriorityAndLastUsed(ordered, preferOAuth)
 
 	for _, acc := range ordered {
 		result, err := s.tryAcquireAccountSlot(ctx, acc.ID, acc.Concurrency)
 		if err == nil && result.Acquired {
+			// 会话数量限制检查
+			if !s.checkAndRegisterSession(ctx, acc, sessionUUID) {
+				result.ReleaseFunc() // 释放槽位，继续尝试下一个账号
+				continue
+			}
 			if sessionHash != "" && s.cache != nil {
 				_ = s.cache.SetSessionAccountID(ctx, derefGroupID(groupID), sessionHash, acc.ID, stickySessionTTL)
 			}
@@ -1490,6 +1534,107 @@ func (s *GatewayService) tryAcquireAccountSlot(ctx context.Context, accountID in
 	return s.concurrencyService.AcquireAccountSlot(ctx, accountID, maxConcurrency)
 }
 
+// isAccountSchedulableForWindowCost 检查账号是否可根据窗口费用进行调度
+// 仅适用于 Anthropic OAuth/SetupToken 账号
+// 返回 true 表示可调度，false 表示不可调度
+func (s *GatewayService) isAccountSchedulableForWindowCost(ctx context.Context, account *Account, isSticky bool) bool {
+	// 只检查 Anthropic OAuth/SetupToken 账号
+	if !account.IsAnthropicOAuthOrSetupToken() {
+		return true
+	}
+
+	limit := account.GetWindowCostLimit()
+	if limit <= 0 {
+		return true // 未启用窗口费用限制
+	}
+
+	// 尝试从缓存获取窗口费用
+	var currentCost float64
+	if s.sessionLimitCache != nil {
+		if cost, hit, err := s.sessionLimitCache.GetWindowCost(ctx, account.ID); err == nil && hit {
+			currentCost = cost
+			goto checkSchedulability
+		}
+	}
+
+	// 缓存未命中，从数据库查询
+	{
+		var startTime time.Time
+		if account.SessionWindowStart != nil {
+			startTime = *account.SessionWindowStart
+		} else {
+			startTime = time.Now().Add(-5 * time.Hour)
+		}
+
+		stats, err := s.usageLogRepo.GetAccountWindowStats(ctx, account.ID, startTime)
+		if err != nil {
+			// 失败开放：查询失败时允许调度
+			return true
+		}
+
+		// 使用标准费用（不含账号倍率）
+		currentCost = stats.StandardCost
+
+		// 设置缓存（忽略错误）
+		if s.sessionLimitCache != nil {
+			_ = s.sessionLimitCache.SetWindowCost(ctx, account.ID, currentCost)
+		}
+	}
+
+checkSchedulability:
+	schedulability := account.CheckWindowCostSchedulability(currentCost)
+
+	switch schedulability {
+	case WindowCostSchedulable:
+		return true
+	case WindowCostStickyOnly:
+		return isSticky
+	case WindowCostNotSchedulable:
+		return false
+	}
+	return true
+}
+
+// checkAndRegisterSession 检查并注册会话，用于会话数量限制
+// 仅适用于 Anthropic OAuth/SetupToken 账号
+// 返回 true 表示允许（在限制内或会话已存在），false 表示拒绝（超出限制且是新会话）
+func (s *GatewayService) checkAndRegisterSession(ctx context.Context, account *Account, sessionUUID string) bool {
+	// 只检查 Anthropic OAuth/SetupToken 账号
+	if !account.IsAnthropicOAuthOrSetupToken() {
+		return true
+	}
+
+	maxSessions := account.GetMaxSessions()
+	if maxSessions <= 0 || sessionUUID == "" {
+		return true // 未启用会话限制或无会话ID
+	}
+
+	if s.sessionLimitCache == nil {
+		return true // 缓存不可用时允许通过
+	}
+
+	idleTimeout := time.Duration(account.GetSessionIdleTimeoutMinutes()) * time.Minute
+
+	allowed, err := s.sessionLimitCache.RegisterSession(ctx, account.ID, sessionUUID, maxSessions, idleTimeout)
+	if err != nil {
+		// 失败开放：缓存错误时允许通过
+		return true
+	}
+	return allowed
+}
+
+// extractSessionUUID 从 metadata.user_id 中提取会话 UUID
+// 格式: user_{64位hex}_account__session_{uuid}
+func extractSessionUUID(metadataUserID string) string {
+	if metadataUserID == "" {
+		return ""
+	}
+	if match := sessionIDRegex.FindStringSubmatch(metadataUserID); len(match) > 1 {
+		return match[1]
+	}
+	return ""
+}
+
 func (s *GatewayService) getSchedulableAccount(ctx context.Context, accountID int64) (*Account, error) {
 	if s.schedulerSnapshot != nil {
 		return s.schedulerSnapshot.GetAccount(ctx, accountID)
@@ -2384,9 +2529,9 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 	retryStart := time.Now()
 	for attempt := 1; attempt <= maxRetryAttempts; attempt++ {
 		// 构建上游请求（每次重试需要重新构建，因为请求体需要重新读取）
-		upstreamReq, err := s.buildUpstreamRequest(ctx, c, account, body, token, tokenType, reqModel, reqStream, shouldMimicClaudeCode)
 		// Capture upstream request body for ops retry of this attempt.
 		c.Set(OpsUpstreamRequestBodyKey, string(body))
+		upstreamReq, err := s.buildUpstreamRequest(ctx, c, account, body, token, tokenType, reqModel, reqStream, shouldMimicClaudeCode)
 		if err != nil {
 			return nil, err
 		}

backend/internal/service/openai_gateway_service.go

@@ -1067,15 +1067,29 @@ func (s *OpenAIGatewayService) handleStreamingResponse(ctx context.Context, resp
 	// 记录上次收到上游数据的时间，用于控制 keepalive 发送频率
 	lastDataAt := time.Now()
 
-	// 仅发送一次错误事件，避免多次写入导致协议混乱（写失败时尽力通知客户端）
+	// 仅发送一次错误事件，避免多次写入导致协议混乱。
+	// 注意：OpenAI `/v1/responses` streaming 事件必须符合 OpenAI Responses schema；
+	// 否则下游 SDK（例如 OpenCode）会因为类型校验失败而报错。
 	errorEventSent := false
+	clientDisconnected := false // 客户端断开后继续 drain 上游以收集 usage
 	sendErrorEvent := func(reason string) {
-		if errorEventSent {
+		if errorEventSent || clientDisconnected {
 			return
 		}
 		errorEventSent = true
-		_, _ = fmt.Fprintf(w, "event: error\ndata: {\"error\":\"%s\"}\n\n", reason)
+		payload := map[string]any{
-		flusher.Flush()
+			"type":            "error",
+			"sequence_number": 0,
+			"error": map[string]any{
+				"type":    "upstream_error",
+				"message": reason,
+				"code":    reason,
+			},
+		}
+		if b, err := json.Marshal(payload); err == nil {
+			_, _ = fmt.Fprintf(w, "data: %s\n\n", b)
+			flusher.Flush()
+		}
 	}
 
 	needModelReplace := originalModel != mappedModel
@@ -1087,6 +1101,17 @@ func (s *OpenAIGatewayService) handleStreamingResponse(ctx context.Context, resp
 				return &openaiStreamingResult{usage: usage, firstTokenMs: firstTokenMs}, nil
 			}
 			if ev.err != nil {
+				// 客户端断开/取消请求时，上游读取往往会返回 context canceled。
+				// /v1/responses 的 SSE 事件必须符合 OpenAI 协议；这里不注入自定义 error event，避免下游 SDK 解析失败。
+				if errors.Is(ev.err, context.Canceled) || errors.Is(ev.err, context.DeadlineExceeded) {
+					log.Printf("Context canceled during streaming, returning collected usage")
+					return &openaiStreamingResult{usage: usage, firstTokenMs: firstTokenMs}, nil
+				}
+				// 客户端已断开时，上游出错仅影响体验，不影响计费；返回已收集 usage
+				if clientDisconnected {
+					log.Printf("Upstream read error after client disconnect: %v, returning collected usage", ev.err)
+					return &openaiStreamingResult{usage: usage, firstTokenMs: firstTokenMs}, nil
+				}
 				if errors.Is(ev.err, bufio.ErrTooLong) {
 					log.Printf("SSE line too long: account=%d max_size=%d error=%v", account.ID, maxLineSize, ev.err)
 					sendErrorEvent("response_too_large")
@@ -1110,15 +1135,19 @@ func (s *OpenAIGatewayService) handleStreamingResponse(ctx context.Context, resp
 
 				// Correct Codex tool calls if needed (apply_patch -> edit, etc.)
 				if correctedData, corrected := s.toolCorrector.CorrectToolCallsInSSEData(data); corrected {
+					data = correctedData
 					line = "data: " + correctedData
 				}
 
-				// Forward line
+				// 写入客户端（客户端断开后继续 drain 上游）
-				if _, err := fmt.Fprintf(w, "%s\n", line); err != nil {
+				if !clientDisconnected {
-					sendErrorEvent("write_failed")
+					if _, err := fmt.Fprintf(w, "%s\n", line); err != nil {
-					return &openaiStreamingResult{usage: usage, firstTokenMs: firstTokenMs}, err
+						clientDisconnected = true
+						log.Printf("Client disconnected during streaming, continuing to drain upstream for billing")
+					} else {
+						flusher.Flush()
+					}
 				}
-				flusher.Flush()
 
 				// Record first token time
 				if firstTokenMs == nil && data != "" && data != "[DONE]" {
@@ -1128,11 +1157,14 @@ func (s *OpenAIGatewayService) handleStreamingResponse(ctx context.Context, resp
 				s.parseSSEUsage(data, usage)
 			} else {
 				// Forward non-data lines as-is
-				if _, err := fmt.Fprintf(w, "%s\n", line); err != nil {
+				if !clientDisconnected {
-					sendErrorEvent("write_failed")
+					if _, err := fmt.Fprintf(w, "%s\n", line); err != nil {
-					return &openaiStreamingResult{usage: usage, firstTokenMs: firstTokenMs}, err
+						clientDisconnected = true
+						log.Printf("Client disconnected during streaming, continuing to drain upstream for billing")
+					} else {
+						flusher.Flush()
+					}
 				}
-				flusher.Flush()
 			}
 
 		case <-intervalCh:
@@ -1140,6 +1172,10 @@ func (s *OpenAIGatewayService) handleStreamingResponse(ctx context.Context, resp
 			if time.Since(lastRead) < streamInterval {
 				continue
 			}
+			if clientDisconnected {
+				log.Printf("Upstream timeout after client disconnect, returning collected usage")
+				return &openaiStreamingResult{usage: usage, firstTokenMs: firstTokenMs}, nil
+			}
 			log.Printf("Stream data interval timeout: account=%d model=%s interval=%s", account.ID, originalModel, streamInterval)
 			// 处理流超时，可能标记账户为临时不可调度或错误状态
 			if s.rateLimitService != nil {
@@ -1149,11 +1185,16 @@ func (s *OpenAIGatewayService) handleStreamingResponse(ctx context.Context, resp
 			return &openaiStreamingResult{usage: usage, firstTokenMs: firstTokenMs}, fmt.Errorf("stream data interval timeout")
 
 		case <-keepaliveCh:
+			if clientDisconnected {
+				continue
+			}
 			if time.Since(lastDataAt) < keepaliveInterval {
 				continue
 			}
 			if _, err := fmt.Fprint(w, ":\n\n"); err != nil {
-				return &openaiStreamingResult{usage: usage, firstTokenMs: firstTokenMs}, err
+				clientDisconnected = true
+				log.Printf("Client disconnected during streaming, continuing to drain upstream for billing")
+				continue
 			}
 			flusher.Flush()
 		}

backend/internal/service/openai_gateway_service_test.go

@@ -33,6 +33,25 @@ type stubConcurrencyCache struct {
 	ConcurrencyCache
 }
 
+type cancelReadCloser struct{}
+
+func (c cancelReadCloser) Read(p []byte) (int, error) { return 0, context.Canceled }
+func (c cancelReadCloser) Close() error               { return nil }
+
+type failingGinWriter struct {
+	gin.ResponseWriter
+	failAfter int
+	writes    int
+}
+
+func (w *failingGinWriter) Write(p []byte) (int, error) {
+	if w.writes >= w.failAfter {
+		return 0, errors.New("write failed")
+	}
+	w.writes++
+	return w.ResponseWriter.Write(p)
+}
+
 func (c stubConcurrencyCache) AcquireAccountSlot(ctx context.Context, accountID int64, maxConcurrency int, requestID string) (bool, error) {
 	return true, nil
 }
@@ -169,8 +188,85 @@ func TestOpenAIStreamingTimeout(t *testing.T) {
 	if err == nil || !strings.Contains(err.Error(), "stream data interval timeout") {
 		t.Fatalf("expected stream timeout error, got %v", err)
 	}
-	if !strings.Contains(rec.Body.String(), "stream_timeout") {
+	if !strings.Contains(rec.Body.String(), "\"type\":\"error\"") || !strings.Contains(rec.Body.String(), "stream_timeout") {
-		t.Fatalf("expected stream_timeout SSE error, got %q", rec.Body.String())
+		t.Fatalf("expected OpenAI-compatible error SSE event, got %q", rec.Body.String())
+	}
+}
+
+func TestOpenAIStreamingContextCanceledDoesNotInjectErrorEvent(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	cfg := &config.Config{
+		Gateway: config.GatewayConfig{
+			StreamDataIntervalTimeout: 0,
+			StreamKeepaliveInterval:   0,
+			MaxLineSize:               defaultMaxLineSize,
+		},
+	}
+	svc := &OpenAIGatewayService{cfg: cfg}
+
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+	c.Request = httptest.NewRequest(http.MethodPost, "/", nil).WithContext(ctx)
+
+	resp := &http.Response{
+		StatusCode: http.StatusOK,
+		Body:       cancelReadCloser{},
+		Header:     http.Header{},
+	}
+
+	_, err := svc.handleStreamingResponse(c.Request.Context(), resp, c, &Account{ID: 1}, time.Now(), "model", "model")
+	if err != nil {
+		t.Fatalf("expected nil error, got %v", err)
+	}
+	if strings.Contains(rec.Body.String(), "event: error") || strings.Contains(rec.Body.String(), "stream_read_error") {
+		t.Fatalf("expected no injected SSE error event, got %q", rec.Body.String())
+	}
+}
+
+func TestOpenAIStreamingClientDisconnectDrainsUpstreamUsage(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	cfg := &config.Config{
+		Gateway: config.GatewayConfig{
+			StreamDataIntervalTimeout: 0,
+			StreamKeepaliveInterval:   0,
+			MaxLineSize:               defaultMaxLineSize,
+		},
+	}
+	svc := &OpenAIGatewayService{cfg: cfg}
+
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	c.Request = httptest.NewRequest(http.MethodPost, "/", nil)
+	c.Writer = &failingGinWriter{ResponseWriter: c.Writer, failAfter: 0}
+
+	pr, pw := io.Pipe()
+	resp := &http.Response{
+		StatusCode: http.StatusOK,
+		Body:       pr,
+		Header:     http.Header{},
+	}
+
+	go func() {
+		defer func() { _ = pw.Close() }()
+		_, _ = pw.Write([]byte("data: {\"type\":\"response.in_progress\",\"response\":{}}\n\n"))
+		_, _ = pw.Write([]byte("data: {\"type\":\"response.completed\",\"response\":{\"usage\":{\"input_tokens\":3,\"output_tokens\":5,\"input_tokens_details\":{\"cached_tokens\":1}}}}\n\n"))
+	}()
+
+	result, err := svc.handleStreamingResponse(c.Request.Context(), resp, c, &Account{ID: 1}, time.Now(), "model", "model")
+	_ = pr.Close()
+	if err != nil {
+		t.Fatalf("expected nil error, got %v", err)
+	}
+	if result == nil || result.usage == nil {
+		t.Fatalf("expected usage result")
+	}
+	if result.usage.InputTokens != 3 || result.usage.OutputTokens != 5 || result.usage.CacheReadInputTokens != 1 {
+		t.Fatalf("unexpected usage: %+v", *result.usage)
+	}
+	if strings.Contains(rec.Body.String(), "event: error") || strings.Contains(rec.Body.String(), "write_failed") {
+		t.Fatalf("expected no injected SSE error event, got %q", rec.Body.String())
 	}
 }
 
@@ -209,8 +305,8 @@ func TestOpenAIStreamingTooLong(t *testing.T) {
 	if !errors.Is(err, bufio.ErrTooLong) {
 		t.Fatalf("expected ErrTooLong, got %v", err)
 	}
-	if !strings.Contains(rec.Body.String(), "response_too_large") {
+	if !strings.Contains(rec.Body.String(), "\"type\":\"error\"") || !strings.Contains(rec.Body.String(), "response_too_large") {
-		t.Fatalf("expected response_too_large SSE error, got %q", rec.Body.String())
+		t.Fatalf("expected OpenAI-compatible error SSE event, got %q", rec.Body.String())
 	}
 }
 

backend/internal/service/ops_retry.go

@@ -514,7 +514,7 @@ func (s *OpsService) selectAccountForRetry(ctx context.Context, reqType opsRetry
 		if s.gatewayService == nil {
 			return nil, fmt.Errorf("gateway service not available")
 		}
-		return s.gatewayService.SelectAccountWithLoadAwareness(ctx, groupID, "", model, excludedIDs)
+		return s.gatewayService.SelectAccountWithLoadAwareness(ctx, groupID, "", model, excludedIDs, "") // 重试不使用会话限制
 	default:
 		return nil, fmt.Errorf("unsupported retry type: %s", reqType)
 	}

backend/internal/service/session_limit_cache.go

@@ -0,0 +1,63 @@
+package service
+
+import (
+	"context"
+	"time"
+)
+
+// SessionLimitCache 管理账号级别的活跃会话跟踪
+// 用于 Anthropic OAuth/SetupToken 账号的会话数量限制
+//
+// Key 格式: session_limit:account:{accountID}
+// 数据结构: Sorted Set (member=sessionUUID, score=timestamp)
+//
+// 会话在空闲超时后自动过期，无需手动清理
+type SessionLimitCache interface {
+	// RegisterSession 注册会话活动
+	// - 如果会话已存在，刷新其时间戳并返回 true
+	// - 如果会话不存在且活跃会话数 < maxSessions，添加新会话并返回 true
+	// - 如果会话不存在且活跃会话数 >= maxSessions，返回 false（拒绝）
+	//
+	// 参数:
+	//   accountID: 账号 ID
+	//   sessionUUID: 从 metadata.user_id 中提取的会话 UUID
+	//   maxSessions: 最大并发会话数限制
+	//   idleTimeout: 会话空闲超时时间
+	//
+	// 返回:
+	//   allowed: true 表示允许（在限制内或会话已存在），false 表示拒绝（超出限制且是新会话）
+	//   error: 操作错误
+	RegisterSession(ctx context.Context, accountID int64, sessionUUID string, maxSessions int, idleTimeout time.Duration) (allowed bool, err error)
+
+	// RefreshSession 刷新现有会话的时间戳
+	// 用于活跃会话保持活动状态
+	RefreshSession(ctx context.Context, accountID int64, sessionUUID string, idleTimeout time.Duration) error
+
+	// GetActiveSessionCount 获取当前活跃会话数
+	// 返回未过期的会话数量
+	GetActiveSessionCount(ctx context.Context, accountID int64) (int, error)
+
+	// GetActiveSessionCountBatch 批量获取多个账号的活跃会话数
+	// 返回 map[accountID]count，查询失败的账号不在 map 中
+	GetActiveSessionCountBatch(ctx context.Context, accountIDs []int64) (map[int64]int, error)
+
+	// IsSessionActive 检查特定会话是否活跃（未过期）
+	IsSessionActive(ctx context.Context, accountID int64, sessionUUID string) (bool, error)
+
+	// ========== 5h窗口费用缓存 ==========
+	// Key 格式: window_cost:account:{accountID}
+	// 用于缓存账号在当前5h窗口内的标准费用，减少数据库聚合查询压力
+
+	// GetWindowCost 获取缓存的窗口费用
+	// 返回 (cost, true, nil) 如果缓存命中
+	// 返回 (0, false, nil) 如果缓存未命中
+	// 返回 (0, false, err) 如果发生错误
+	GetWindowCost(ctx context.Context, accountID int64) (cost float64, hit bool, err error)
+
+	// SetWindowCost 设置窗口费用缓存
+	SetWindowCost(ctx context.Context, accountID int64, cost float64) error
+
+	// GetWindowCostBatch 批量获取窗口费用缓存
+	// 返回 map[accountID]cost，缓存未命中的账号不在 map 中
+	GetWindowCostBatch(ctx context.Context, accountIDs []int64) (map[int64]float64, error)
+}

deploy/Caddyfile

@@ -33,6 +33,22 @@
 
 # 修改为你的域名
 example.com {
+	# =========================================================================
+	# 静态资源长期缓存（高优先级，放在最前面）
+	# 带 hash 的文件可以永久缓存，浏览器和 CDN 都会缓存
+	# =========================================================================
+	@static {
+		path /assets/*
+		path /logo.png
+		path /favicon.ico
+	}
+	header @static {
+		Cache-Control "public, max-age=31536000, immutable"
+		# 移除可能干扰缓存的头
+		-Pragma
+		-Expires
+	}
+
 	# =========================================================================
 	# TLS 安全配置
 	# =========================================================================

frontend/package.json

@@ -9,7 +9,10 @@
     "preview": "vite preview",
     "lint": "eslint . --ext .vue,.js,.jsx,.cjs,.mjs,.ts,.tsx,.cts,.mts --fix",
     "lint:check": "eslint . --ext .vue,.js,.jsx,.cjs,.mjs,.ts,.tsx,.cts,.mts",
-    "typecheck": "vue-tsc --noEmit"
+    "typecheck": "vue-tsc --noEmit",
+    "test": "vitest",
+    "test:run": "vitest run",
+    "test:coverage": "vitest run --coverage"
   },
   "dependencies": {
     "@lobehub/icons": "^4.0.2",
@@ -29,17 +32,21 @@
     "@types/file-saver": "^2.0.7",
     "@types/mdx": "^2.0.13",
     "@types/node": "^20.10.5",
-    "@vitejs/plugin-vue": "^5.2.3",
-    "autoprefixer": "^10.4.16",
     "@typescript-eslint/eslint-plugin": "^7.18.0",
     "@typescript-eslint/parser": "^7.18.0",
+    "@vitejs/plugin-vue": "^5.2.3",
+    "@vitest/coverage-v8": "^2.1.9",
+    "@vue/test-utils": "^2.4.6",
+    "autoprefixer": "^10.4.16",
     "eslint": "^8.57.0",
     "eslint-plugin-vue": "^9.25.0",
+    "jsdom": "^24.1.3",
     "postcss": "^8.4.32",
     "tailwindcss": "^3.4.0",
     "typescript": "~5.6.0",
     "vite": "^5.0.10",
     "vite-plugin-checker": "^0.9.1",
+    "vitest": "^2.1.9",
     "vue-tsc": "^2.2.0"
   }
 }

frontend/src/App.vue

@@ -2,6 +2,7 @@
 import { RouterView, useRouter, useRoute } from 'vue-router'
 import { onMounted, watch } from 'vue'
 import Toast from '@/components/common/Toast.vue'
+import NavigationProgress from '@/components/common/NavigationProgress.vue'
 import { useAppStore, useAuthStore, useSubscriptionStore } from '@/stores'
 import { getSetupStatus } from '@/api/setup'
 
@@ -84,6 +85,7 @@ onMounted(async () => {
 </script>
 
 <template>
+  <NavigationProgress />
   <RouterView />
   <Toast />
 </template>

frontend/src/__tests__/integration/navigation.spec.ts

@@ -0,0 +1,478 @@
+/**
+ * 导航集成测试
+ * 测试完整的页面导航流程、预加载和错误恢复机制
+ */
+import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest'
+import { createRouter, createWebHistory, type Router } from 'vue-router'
+import { createPinia, setActivePinia } from 'pinia'
+import { mount, flushPromises } from '@vue/test-utils'
+import { defineComponent, h, nextTick } from 'vue'
+import { useNavigationLoadingState, _resetNavigationLoadingInstance } from '@/composables/useNavigationLoading'
+import { useRoutePrefetch } from '@/composables/useRoutePrefetch'
+
+// Mock 视图组件
+const MockDashboard = defineComponent({
+  name: 'MockDashboard',
+  render() {
+    return h('div', { class: 'dashboard' }, 'Dashboard')
+  }
+})
+
+const MockKeys = defineComponent({
+  name: 'MockKeys',
+  render() {
+    return h('div', { class: 'keys' }, 'Keys')
+  }
+})
+
+const MockUsage = defineComponent({
+  name: 'MockUsage',
+  render() {
+    return h('div', { class: 'usage' }, 'Usage')
+  }
+})
+
+// Mock stores
+vi.mock('@/stores/auth', () => ({
+  useAuthStore: () => ({
+    isAuthenticated: true,
+    isAdmin: false,
+    isSimpleMode: false,
+    checkAuth: vi.fn()
+  })
+}))
+
+vi.mock('@/stores/app', () => ({
+  useAppStore: () => ({
+    siteName: 'Test Site'
+  })
+}))
+
+// 创建测试路由
+function createTestRouter(): Router {
+  return createRouter({
+    history: createWebHistory(),
+    routes: [
+      {
+        path: '/',
+        redirect: '/dashboard'
+      },
+      {
+        path: '/dashboard',
+        name: 'Dashboard',
+        component: MockDashboard,
+        meta: { requiresAuth: true, title: 'Dashboard' }
+      },
+      {
+        path: '/keys',
+        name: 'Keys',
+        component: MockKeys,
+        meta: { requiresAuth: true, title: 'Keys' }
+      },
+      {
+        path: '/usage',
+        name: 'Usage',
+        component: MockUsage,
+        meta: { requiresAuth: true, title: 'Usage' }
+      }
+    ]
+  })
+}
+
+// 测试用 App 组件
+const TestApp = defineComponent({
+  name: 'TestApp',
+  setup() {
+    return () => h('div', { id: 'app' }, [h('router-view')])
+  }
+})
+
+describe('Navigation Integration Tests', () => {
+  let router: Router
+  let originalRequestIdleCallback: typeof window.requestIdleCallback
+  let originalCancelIdleCallback: typeof window.cancelIdleCallback
+
+  beforeEach(() => {
+    // 设置 Pinia
+    setActivePinia(createPinia())
+
+    // 重置导航加载状态
+    _resetNavigationLoadingInstance()
+
+    // 创建新的路由实例
+    router = createTestRouter()
+
+    // Mock requestIdleCallback
+    originalRequestIdleCallback = window.requestIdleCallback
+    originalCancelIdleCallback = window.cancelIdleCallback
+
+    vi.stubGlobal('requestIdleCallback', (cb: IdleRequestCallback) => {
+      const id = setTimeout(() => cb({ didTimeout: false, timeRemaining: () => 50 }), 0)
+      return id
+    })
+    vi.stubGlobal('cancelIdleCallback', (id: number) => clearTimeout(id))
+  })
+
+  afterEach(() => {
+    vi.restoreAllMocks()
+    window.requestIdleCallback = originalRequestIdleCallback
+    window.cancelIdleCallback = originalCancelIdleCallback
+  })
+
+  describe('完整页面导航流程', () => {
+    it('导航时应该触发加载状态变化', async () => {
+      const navigationLoading = useNavigationLoadingState()
+
+      // 初始状态
+      expect(navigationLoading.isLoading.value).toBe(false)
+
+      // 挂载应用
+      const wrapper = mount(TestApp, {
+        global: {
+          plugins: [router]
+        }
+      })
+
+      // 等待路由初始化
+      await router.isReady()
+      await flushPromises()
+
+      // 导航到 /dashboard
+      await router.push('/dashboard')
+      await flushPromises()
+      await nextTick()
+
+      // 导航结束后状态应该重置
+      expect(navigationLoading.isLoading.value).toBe(false)
+
+      wrapper.unmount()
+    })
+
+    it('导航到新页面应该正确渲染组件', async () => {
+      const wrapper = mount(TestApp, {
+        global: {
+          plugins: [router]
+        }
+      })
+
+      await router.isReady()
+      await router.push('/dashboard')
+      await flushPromises()
+      await nextTick()
+
+      // 检查当前路由
+      expect(router.currentRoute.value.path).toBe('/dashboard')
+
+      wrapper.unmount()
+    })
+
+    it('连续快速导航应该正确处理路由状态', async () => {
+      const wrapper = mount(TestApp, {
+        global: {
+          plugins: [router]
+        }
+      })
+
+      await router.isReady()
+      await router.push('/dashboard')
+
+      // 快速连续导航
+      router.push('/keys')
+      router.push('/usage')
+      router.push('/dashboard')
+
+      await flushPromises()
+      await nextTick()
+
+      // 应该最终停在 /dashboard
+      expect(router.currentRoute.value.path).toBe('/dashboard')
+
+      wrapper.unmount()
+    })
+  })
+
+  describe('路由预加载', () => {
+    it('导航后应该触发相关路由预加载', async () => {
+      const routePrefetch = useRoutePrefetch()
+      const triggerSpy = vi.spyOn(routePrefetch, 'triggerPrefetch')
+
+      // 设置 afterEach 守卫
+      router.afterEach((to) => {
+        routePrefetch.triggerPrefetch(to)
+      })
+
+      const wrapper = mount(TestApp, {
+        global: {
+          plugins: [router]
+        }
+      })
+
+      await router.isReady()
+      await router.push('/dashboard')
+      await flushPromises()
+
+      // 应该触发预加载
+      expect(triggerSpy).toHaveBeenCalled()
+
+      wrapper.unmount()
+    })
+
+    it('已预加载的路由不应重复预加载', async () => {
+      const routePrefetch = useRoutePrefetch()
+
+      const wrapper = mount(TestApp, {
+        global: {
+          plugins: [router]
+        }
+      })
+
+      await router.isReady()
+      await router.push('/dashboard')
+      await flushPromises()
+
+      // 手动触发预加载
+      routePrefetch.triggerPrefetch(router.currentRoute.value)
+      await new Promise((resolve) => setTimeout(resolve, 100))
+
+      const prefetchedCount = routePrefetch.prefetchedRoutes.value.size
+
+      // 再次触发相同路由预加载
+      routePrefetch.triggerPrefetch(router.currentRoute.value)
+      await new Promise((resolve) => setTimeout(resolve, 100))
+
+      // 预加载数量不应增加
+      expect(routePrefetch.prefetchedRoutes.value.size).toBe(prefetchedCount)
+
+      wrapper.unmount()
+    })
+
+    it('路由变化时应取消之前的预加载任务', async () => {
+      const routePrefetch = useRoutePrefetch()
+
+      const wrapper = mount(TestApp, {
+        global: {
+          plugins: [router]
+        }
+      })
+
+      await router.isReady()
+
+      // 触发预加载
+      routePrefetch.triggerPrefetch(router.currentRoute.value)
+
+      // 立即导航到新路由（这会在内部调用 cancelPendingPrefetch）
+      routePrefetch.triggerPrefetch({ path: '/keys' } as any)
+
+      // 由于 triggerPrefetch 内部调用 cancelPendingPrefetch，检查是否有预加载被正确管理
+      expect(routePrefetch.prefetchedRoutes.value.size).toBeLessThanOrEqual(2)
+
+      wrapper.unmount()
+    })
+  })
+
+  describe('Chunk 加载错误恢复', () => {
+    it('chunk 加载失败应该被正确捕获', async () => {
+      const errorHandler = vi.fn()
+
+      // 创建带错误处理的路由
+      const errorRouter = createRouter({
+        history: createWebHistory(),
+        routes: [
+          {
+            path: '/dashboard',
+            name: 'Dashboard',
+            component: MockDashboard
+          },
+          {
+            path: '/error-page',
+            name: 'ErrorPage',
+            // 模拟加载失败的组件
+            component: () => Promise.reject(new Error('Failed to fetch dynamically imported module'))
+          }
+        ]
+      })
+
+      errorRouter.onError(errorHandler)
+
+      const wrapper = mount(TestApp, {
+        global: {
+          plugins: [errorRouter]
+        }
+      })
+
+      await errorRouter.isReady()
+      await errorRouter.push('/dashboard')
+      await flushPromises()
+
+      // 尝试导航到会失败的页面
+      try {
+        await errorRouter.push('/error-page')
+      } catch {
+        // 预期会失败
+      }
+
+      await flushPromises()
+
+      // 错误处理器应该被调用
+      expect(errorHandler).toHaveBeenCalled()
+
+      wrapper.unmount()
+    })
+
+    it('chunk 加载错误应该包含正确的错误信息', async () => {
+      let capturedError: Error | null = null
+
+      const errorRouter = createRouter({
+        history: createWebHistory(),
+        routes: [
+          {
+            path: '/dashboard',
+            name: 'Dashboard',
+            component: MockDashboard
+          },
+          {
+            path: '/chunk-error',
+            name: 'ChunkError',
+            component: () => {
+              const error = new Error('Loading chunk failed')
+              error.name = 'ChunkLoadError'
+              return Promise.reject(error)
+            }
+          }
+        ]
+      })
+
+      errorRouter.onError((error) => {
+        capturedError = error
+      })
+
+      const wrapper = mount(TestApp, {
+        global: {
+          plugins: [errorRouter]
+        }
+      })
+
+      await errorRouter.isReady()
+
+      try {
+        await errorRouter.push('/chunk-error')
+      } catch {
+        // 预期会失败
+      }
+
+      await flushPromises()
+
+      expect(capturedError).not.toBeNull()
+      expect(capturedError!.name).toBe('ChunkLoadError')
+
+      wrapper.unmount()
+    })
+  })
+
+  describe('导航状态管理', () => {
+    it('导航开始时 isLoading 应该变为 true', async () => {
+      const navigationLoading = useNavigationLoadingState()
+
+      // 创建一个延迟加载的组件来模拟真实场景
+      const DelayedComponent = defineComponent({
+        name: 'DelayedComponent',
+        async setup() {
+          await new Promise((resolve) => setTimeout(resolve, 50))
+          return () => h('div', 'Delayed')
+        }
+      })
+
+      const delayRouter = createRouter({
+        history: createWebHistory(),
+        routes: [
+          {
+            path: '/dashboard',
+            name: 'Dashboard',
+            component: MockDashboard
+          },
+          {
+            path: '/delayed',
+            name: 'Delayed',
+            component: DelayedComponent
+          }
+        ]
+      })
+
+      // 设置导航守卫
+      delayRouter.beforeEach(() => {
+        navigationLoading.startNavigation()
+      })
+
+      delayRouter.afterEach(() => {
+        navigationLoading.endNavigation()
+      })
+
+      const wrapper = mount(TestApp, {
+        global: {
+          plugins: [delayRouter]
+        }
+      })
+
+      await delayRouter.isReady()
+      await delayRouter.push('/dashboard')
+      await flushPromises()
+
+      // 导航结束后 isLoading 应该为 false
+      expect(navigationLoading.isLoading.value).toBe(false)
+
+      wrapper.unmount()
+    })
+
+    it('导航取消时应该正确重置状态', async () => {
+      const navigationLoading = useNavigationLoadingState()
+
+      const testRouter = createRouter({
+        history: createWebHistory(),
+        routes: [
+          {
+            path: '/dashboard',
+            name: 'Dashboard',
+            component: MockDashboard
+          },
+          {
+            path: '/keys',
+            name: 'Keys',
+            component: MockKeys,
+            beforeEnter: (_to, _from, next) => {
+              // 模拟导航取消
+              next(false)
+            }
+          }
+        ]
+      })
+
+      testRouter.beforeEach(() => {
+        navigationLoading.startNavigation()
+      })
+
+      testRouter.afterEach(() => {
+        navigationLoading.endNavigation()
+      })
+
+      const wrapper = mount(TestApp, {
+        global: {
+          plugins: [testRouter]
+        }
+      })
+
+      await testRouter.isReady()
+      await testRouter.push('/dashboard')
+      await flushPromises()
+
+      // 尝试导航到被取消的路由
+      await testRouter.push('/keys').catch(() => {})
+      await flushPromises()
+
+      // 导航被取消后，状态应该被重置
+      // 注意：由于 afterEach 仍然会被调用，isLoading 应该为 false
+      expect(navigationLoading.isLoading.value).toBe(false)
+
+      wrapper.unmount()
+    })
+  })
+})

frontend/src/__tests__/setup.ts

@@ -0,0 +1,45 @@
+/**
+ * Vitest 测试环境设置
+ * 提供全局 mock 和测试工具
+ */
+import { config } from '@vue/test-utils'
+import { vi } from 'vitest'
+
+// Mock requestIdleCallback (Safari < 15 不支持)
+if (typeof globalThis.requestIdleCallback === 'undefined') {
+  globalThis.requestIdleCallback = ((callback: IdleRequestCallback) => {
+    return window.setTimeout(() => callback({ didTimeout: false, timeRemaining: () => 50 }), 1)
+  }) as unknown as typeof requestIdleCallback
+}
+
+if (typeof globalThis.cancelIdleCallback === 'undefined') {
+  globalThis.cancelIdleCallback = ((id: number) => {
+    window.clearTimeout(id)
+  }) as unknown as typeof cancelIdleCallback
+}
+
+// Mock IntersectionObserver
+class MockIntersectionObserver {
+  observe = vi.fn()
+  disconnect = vi.fn()
+  unobserve = vi.fn()
+}
+
+globalThis.IntersectionObserver = MockIntersectionObserver as unknown as typeof IntersectionObserver
+
+// Mock ResizeObserver
+class MockResizeObserver {
+  observe = vi.fn()
+  disconnect = vi.fn()
+  unobserve = vi.fn()
+}
+
+globalThis.ResizeObserver = MockResizeObserver as unknown as typeof ResizeObserver
+
+// Vue Test Utils 全局配置
+config.global.stubs = {
+  // 可以在这里添加全局 stub
+}
+
+// 设置全局测试超时
+vi.setConfig({ testTimeout: 10000 })

frontend/src/components/account/AccountCapacityCell.vue

@@ -0,0 +1,199 @@
+<template>
+  <div class="flex flex-col gap-1.5">
+    <!-- 并发槽位 -->
+    <div class="flex items-center gap-1.5">
+      <span
+        :class="[
+          'inline-flex items-center gap-1 rounded-md px-2 py-0.5 text-xs font-medium',
+          concurrencyClass
+        ]"
+      >
+        <svg class="h-3 w-3" fill="none" viewBox="0 0 24 24" stroke="currentColor" stroke-width="2">
+          <path stroke-linecap="round" stroke-linejoin="round" d="M3.75 6A2.25 2.25 0 016 3.75h2.25A2.25 2.25 0 0110.5 6v2.25a2.25 2.25 0 01-2.25 2.25H6a2.25 2.25 0 01-2.25-2.25V6zM3.75 15.75A2.25 2.25 0 016 13.5h2.25a2.25 2.25 0 012.25 2.25V18a2.25 2.25 0 01-2.25 2.25H6A2.25 2.25 0 013.75 18v-2.25zM13.5 6a2.25 2.25 0 012.25-2.25H18A2.25 2.25 0 0120.25 6v2.25A2.25 2.25 0 0118 10.5h-2.25a2.25 2.25 0 01-2.25-2.25V6zM13.5 15.75a2.25 2.25 0 012.25-2.25H18a2.25 2.25 0 012.25 2.25V18A2.25 2.25 0 0118 20.25h-2.25A2.25 2.25 0 0113.5 18v-2.25z" />
+        </svg>
+        <span class="font-mono">{{ currentConcurrency }}</span>
+        <span class="text-gray-400 dark:text-gray-500">/</span>
+        <span class="font-mono">{{ account.concurrency }}</span>
+      </span>
+    </div>
+
+    <!-- 5h窗口费用限制（仅 Anthropic OAuth/SetupToken 且启用时显示） -->
+    <div v-if="showWindowCost" class="flex items-center gap-1">
+      <span
+        :class="[
+          'inline-flex items-center gap-1 rounded-md px-1.5 py-0.5 text-[10px] font-medium',
+          windowCostClass
+        ]"
+        :title="windowCostTooltip"
+      >
+        <svg class="h-2.5 w-2.5" fill="none" viewBox="0 0 24 24" stroke="currentColor" stroke-width="2">
+          <path stroke-linecap="round" stroke-linejoin="round" d="M12 6v12m-3-2.818l.879.659c1.171.879 3.07.879 4.242 0 1.172-.879 1.172-2.303 0-3.182C13.536 12.219 12.768 12 12 12c-.725 0-1.45-.22-2.003-.659-1.106-.879-1.106-2.303 0-3.182s2.9-.879 4.006 0l.415.33M21 12a9 9 0 11-18 0 9 9 0 0118 0z" />
+        </svg>
+        <span class="font-mono">${{ formatCost(currentWindowCost) }}</span>
+        <span class="text-gray-400 dark:text-gray-500">/</span>
+        <span class="font-mono">${{ formatCost(account.window_cost_limit) }}</span>
+      </span>
+    </div>
+
+    <!-- 会话数量限制（仅 Anthropic OAuth/SetupToken 且启用时显示） -->
+    <div v-if="showSessionLimit" class="flex items-center gap-1">
+      <span
+        :class="[
+          'inline-flex items-center gap-1 rounded-md px-1.5 py-0.5 text-[10px] font-medium',
+          sessionLimitClass
+        ]"
+        :title="sessionLimitTooltip"
+      >
+        <svg class="h-2.5 w-2.5" fill="none" viewBox="0 0 24 24" stroke="currentColor" stroke-width="2">
+          <path stroke-linecap="round" stroke-linejoin="round" d="M15 19.128a9.38 9.38 0 002.625.372 9.337 9.337 0 004.121-.952 4.125 4.125 0 00-7.533-2.493M15 19.128v-.003c0-1.113-.285-2.16-.786-3.07M15 19.128v.106A12.318 12.318 0 018.624 21c-2.331 0-4.512-.645-6.374-1.766l-.001-.109a6.375 6.375 0 0111.964-3.07M12 6.375a3.375 3.375 0 11-6.75 0 3.375 3.375 0 016.75 0zm8.25 2.25a2.625 2.625 0 11-5.25 0 2.625 2.625 0 015.25 0z" />
+        </svg>
+        <span class="font-mono">{{ activeSessions }}</span>
+        <span class="text-gray-400 dark:text-gray-500">/</span>
+        <span class="font-mono">{{ account.max_sessions }}</span>
+      </span>
+    </div>
+  </div>
+</template>
+
+<script setup lang="ts">
+import { computed } from 'vue'
+import { useI18n } from 'vue-i18n'
+import type { Account } from '@/types'
+
+const props = defineProps<{
+  account: Account
+}>()
+
+const { t } = useI18n()
+
+// 当前并发数
+const currentConcurrency = computed(() => props.account.current_concurrency || 0)
+
+// 是否为 Anthropic OAuth/SetupToken 账号
+const isAnthropicOAuthOrSetupToken = computed(() => {
+  return (
+    props.account.platform === 'anthropic' &&
+    (props.account.type === 'oauth' || props.account.type === 'setup-token')
+  )
+})
+
+// 是否显示窗口费用限制
+const showWindowCost = computed(() => {
+  return (
+    isAnthropicOAuthOrSetupToken.value &&
+    props.account.window_cost_limit !== undefined &&
+    props.account.window_cost_limit !== null &&
+    props.account.window_cost_limit > 0
+  )
+})
+
+// 当前窗口费用
+const currentWindowCost = computed(() => props.account.current_window_cost ?? 0)
+
+// 是否显示会话限制
+const showSessionLimit = computed(() => {
+  return (
+    isAnthropicOAuthOrSetupToken.value &&
+    props.account.max_sessions !== undefined &&
+    props.account.max_sessions !== null &&
+    props.account.max_sessions > 0
+  )
+})
+
+// 当前活跃会话数
+const activeSessions = computed(() => props.account.active_sessions ?? 0)
+
+// 并发状态样式
+const concurrencyClass = computed(() => {
+  const current = currentConcurrency.value
+  const max = props.account.concurrency
+
+  if (current >= max) {
+    return 'bg-red-100 text-red-700 dark:bg-red-900/30 dark:text-red-400'
+  }
+  if (current > 0) {
+    return 'bg-yellow-100 text-yellow-700 dark:bg-yellow-900/30 dark:text-yellow-400'
+  }
+  return 'bg-gray-100 text-gray-600 dark:bg-gray-800 dark:text-gray-400'
+})
+
+// 窗口费用状态样式
+const windowCostClass = computed(() => {
+  if (!showWindowCost.value) return ''
+
+  const current = currentWindowCost.value
+  const limit = props.account.window_cost_limit || 0
+  const reserve = props.account.window_cost_sticky_reserve || 10
+
+  // >= 阈值+预留: 完全不可调度 (红色)
+  if (current >= limit + reserve) {
+    return 'bg-red-100 text-red-700 dark:bg-red-900/30 dark:text-red-400'
+  }
+  // >= 阈值: 仅粘性会话 (橙色)
+  if (current >= limit) {
+    return 'bg-orange-100 text-orange-700 dark:bg-orange-900/30 dark:text-orange-400'
+  }
+  // >= 80% 阈值: 警告 (黄色)
+  if (current >= limit * 0.8) {
+    return 'bg-yellow-100 text-yellow-700 dark:bg-yellow-900/30 dark:text-yellow-400'
+  }
+  // 正常 (绿色)
+  return 'bg-emerald-100 text-emerald-700 dark:bg-emerald-900/30 dark:text-emerald-400'
+})
+
+// 窗口费用提示文字
+const windowCostTooltip = computed(() => {
+  if (!showWindowCost.value) return ''
+
+  const current = currentWindowCost.value
+  const limit = props.account.window_cost_limit || 0
+  const reserve = props.account.window_cost_sticky_reserve || 10
+
+  if (current >= limit + reserve) {
+    return t('admin.accounts.capacity.windowCost.blocked')
+  }
+  if (current >= limit) {
+    return t('admin.accounts.capacity.windowCost.stickyOnly')
+  }
+  return t('admin.accounts.capacity.windowCost.normal')
+})
+
+// 会话限制状态样式
+const sessionLimitClass = computed(() => {
+  if (!showSessionLimit.value) return ''
+
+  const current = activeSessions.value
+  const max = props.account.max_sessions || 0
+
+  // >= 最大: 完全占满 (红色)
+  if (current >= max) {
+    return 'bg-red-100 text-red-700 dark:bg-red-900/30 dark:text-red-400'
+  }
+  // >= 80%: 警告 (黄色)
+  if (current >= max * 0.8) {
+    return 'bg-yellow-100 text-yellow-700 dark:bg-yellow-900/30 dark:text-yellow-400'
+  }
+  // 正常 (绿色)
+  return 'bg-emerald-100 text-emerald-700 dark:bg-emerald-900/30 dark:text-emerald-400'
+})
+
+// 会话限制提示文字
+const sessionLimitTooltip = computed(() => {
+  if (!showSessionLimit.value) return ''
+
+  const current = activeSessions.value
+  const max = props.account.max_sessions || 0
+  const idle = props.account.session_idle_timeout_minutes || 5
+
+  if (current >= max) {
+    return t('admin.accounts.capacity.sessions.full', { idle })
+  }
+  return t('admin.accounts.capacity.sessions.normal', { idle })
+})
+
+// 格式化费用显示
+const formatCost = (value: number | null | undefined) => {
+  if (value === null || value === undefined) return '0'
+  return value.toFixed(2)
+}
+</script>

frontend/src/components/account/EditAccountModal.vue

@@ -604,6 +604,136 @@
         </div>
       </div>
 
+      <!-- Quota Control Section (Anthropic OAuth/SetupToken only) -->
+      <div
+        v-if="account?.platform === 'anthropic' && (account?.type === 'oauth' || account?.type === 'setup-token')"
+        class="border-t border-gray-200 pt-4 dark:border-dark-600 space-y-4"
+      >
+        <div class="mb-3">
+          <h3 class="input-label mb-0 text-base font-semibold">{{ t('admin.accounts.quotaControl.title') }}</h3>
+          <p class="mt-1 text-xs text-gray-500 dark:text-gray-400">
+            {{ t('admin.accounts.quotaControl.hint') }}
+          </p>
+        </div>
+
+        <!-- Window Cost Limit -->
+        <div class="rounded-lg border border-gray-200 p-4 dark:border-dark-600">
+          <div class="mb-3 flex items-center justify-between">
+            <div>
+              <label class="input-label mb-0">{{ t('admin.accounts.quotaControl.windowCost.label') }}</label>
+              <p class="mt-1 text-xs text-gray-500 dark:text-gray-400">
+                {{ t('admin.accounts.quotaControl.windowCost.hint') }}
+              </p>
+            </div>
+            <button
+              type="button"
+              @click="windowCostEnabled = !windowCostEnabled"
+              :class="[
+                'relative inline-flex h-6 w-11 flex-shrink-0 cursor-pointer rounded-full border-2 border-transparent transition-colors duration-200 ease-in-out focus:outline-none focus:ring-2 focus:ring-primary-500 focus:ring-offset-2',
+                windowCostEnabled ? 'bg-primary-600' : 'bg-gray-200 dark:bg-dark-600'
+              ]"
+            >
+              <span
+                :class="[
+                  'pointer-events-none inline-block h-5 w-5 transform rounded-full bg-white shadow ring-0 transition duration-200 ease-in-out',
+                  windowCostEnabled ? 'translate-x-5' : 'translate-x-0'
+                ]"
+              />
+            </button>
+          </div>
+
+          <div v-if="windowCostEnabled" class="grid grid-cols-2 gap-4">
+            <div>
+              <label class="input-label">{{ t('admin.accounts.quotaControl.windowCost.limit') }}</label>
+              <div class="relative">
+                <span class="absolute left-3 top-1/2 -translate-y-1/2 text-gray-500 dark:text-gray-400">$</span>
+                <input
+                  v-model.number="windowCostLimit"
+                  type="number"
+                  min="0"
+                  step="1"
+                  class="input pl-7"
+                  :placeholder="t('admin.accounts.quotaControl.windowCost.limitPlaceholder')"
+                />
+              </div>
+              <p class="input-hint">{{ t('admin.accounts.quotaControl.windowCost.limitHint') }}</p>
+            </div>
+            <div>
+              <label class="input-label">{{ t('admin.accounts.quotaControl.windowCost.stickyReserve') }}</label>
+              <div class="relative">
+                <span class="absolute left-3 top-1/2 -translate-y-1/2 text-gray-500 dark:text-gray-400">$</span>
+                <input
+                  v-model.number="windowCostStickyReserve"
+                  type="number"
+                  min="0"
+                  step="1"
+                  class="input pl-7"
+                  :placeholder="t('admin.accounts.quotaControl.windowCost.stickyReservePlaceholder')"
+                />
+              </div>
+              <p class="input-hint">{{ t('admin.accounts.quotaControl.windowCost.stickyReserveHint') }}</p>
+            </div>
+          </div>
+        </div>
+
+        <!-- Session Limit -->
+        <div class="rounded-lg border border-gray-200 p-4 dark:border-dark-600">
+          <div class="mb-3 flex items-center justify-between">
+            <div>
+              <label class="input-label mb-0">{{ t('admin.accounts.quotaControl.sessionLimit.label') }}</label>
+              <p class="mt-1 text-xs text-gray-500 dark:text-gray-400">
+                {{ t('admin.accounts.quotaControl.sessionLimit.hint') }}
+              </p>
+            </div>
+            <button
+              type="button"
+              @click="sessionLimitEnabled = !sessionLimitEnabled"
+              :class="[
+                'relative inline-flex h-6 w-11 flex-shrink-0 cursor-pointer rounded-full border-2 border-transparent transition-colors duration-200 ease-in-out focus:outline-none focus:ring-2 focus:ring-primary-500 focus:ring-offset-2',
+                sessionLimitEnabled ? 'bg-primary-600' : 'bg-gray-200 dark:bg-dark-600'
+              ]"
+            >
+              <span
+                :class="[
+                  'pointer-events-none inline-block h-5 w-5 transform rounded-full bg-white shadow ring-0 transition duration-200 ease-in-out',
+                  sessionLimitEnabled ? 'translate-x-5' : 'translate-x-0'
+                ]"
+              />
+            </button>
+          </div>
+
+          <div v-if="sessionLimitEnabled" class="grid grid-cols-2 gap-4">
+            <div>
+              <label class="input-label">{{ t('admin.accounts.quotaControl.sessionLimit.maxSessions') }}</label>
+              <input
+                v-model.number="maxSessions"
+                type="number"
+                min="1"
+                step="1"
+                class="input"
+                :placeholder="t('admin.accounts.quotaControl.sessionLimit.maxSessionsPlaceholder')"
+              />
+              <p class="input-hint">{{ t('admin.accounts.quotaControl.sessionLimit.maxSessionsHint') }}</p>
+            </div>
+            <div>
+              <label class="input-label">{{ t('admin.accounts.quotaControl.sessionLimit.idleTimeout') }}</label>
+              <div class="relative">
+                <input
+                  v-model.number="sessionIdleTimeout"
+                  type="number"
+                  min="1"
+                  step="1"
+                  class="input pr-12"
+                  :placeholder="t('admin.accounts.quotaControl.sessionLimit.idleTimeoutPlaceholder')"
+                />
+                <span class="absolute right-3 top-1/2 -translate-y-1/2 text-gray-500 dark:text-gray-400">{{ t('common.minutes') }}</span>
+              </div>
+              <p class="input-hint">{{ t('admin.accounts.quotaControl.sessionLimit.idleTimeoutHint') }}</p>
+            </div>
+          </div>
+        </div>
+      </div>
+
       <div class="border-t border-gray-200 pt-4 dark:border-dark-600">
         <div>
           <label class="input-label">{{ t('common.status') }}</label>
@@ -767,6 +897,14 @@ const mixedScheduling = ref(false) // For antigravity accounts: enable mixed sch
 const tempUnschedEnabled = ref(false)
 const tempUnschedRules = ref<TempUnschedRuleForm[]>([])
 
+// Quota control state (Anthropic OAuth/SetupToken only)
+const windowCostEnabled = ref(false)
+const windowCostLimit = ref<number | null>(null)
+const windowCostStickyReserve = ref<number | null>(null)
+const sessionLimitEnabled = ref(false)
+const maxSessions = ref<number | null>(null)
+const sessionIdleTimeout = ref<number | null>(null)
+
 // Computed: current preset mappings based on platform
 const presetMappings = computed(() => getPresetMappingsByPlatform(props.account?.platform || 'anthropic'))
 const tempUnschedPresets = computed(() => [
@@ -854,6 +992,9 @@ watch(
       const extra = newAccount.extra as Record<string, unknown> | undefined
       mixedScheduling.value = extra?.mixed_scheduling === true
 
+      // Load quota control settings (Anthropic OAuth/SetupToken only)
+      loadQuotaControlSettings(newAccount)
+
       loadTempUnschedRules(credentials)
 
       // Initialize API Key fields for apikey type
@@ -1087,6 +1228,35 @@ function loadTempUnschedRules(credentials?: Record<string, unknown>) {
   })
 }
 
+// Load quota control settings from account (Anthropic OAuth/SetupToken only)
+function loadQuotaControlSettings(account: Account) {
+  // Reset all quota control state first
+  windowCostEnabled.value = false
+  windowCostLimit.value = null
+  windowCostStickyReserve.value = null
+  sessionLimitEnabled.value = false
+  maxSessions.value = null
+  sessionIdleTimeout.value = null
+
+  // Only applies to Anthropic OAuth/SetupToken accounts
+  if (account.platform !== 'anthropic' || (account.type !== 'oauth' && account.type !== 'setup-token')) {
+    return
+  }
+
+  // Load from extra field (via backend DTO fields)
+  if (account.window_cost_limit != null && account.window_cost_limit > 0) {
+    windowCostEnabled.value = true
+    windowCostLimit.value = account.window_cost_limit
+    windowCostStickyReserve.value = account.window_cost_sticky_reserve ?? 10
+  }
+
+  if (account.max_sessions != null && account.max_sessions > 0) {
+    sessionLimitEnabled.value = true
+    maxSessions.value = account.max_sessions
+    sessionIdleTimeout.value = account.session_idle_timeout_minutes ?? 5
+  }
+}
+
 function formatTempUnschedKeywords(value: unknown) {
   if (Array.isArray(value)) {
     return value
@@ -1214,6 +1384,32 @@ const handleSubmit = async () => {
       updatePayload.extra = newExtra
     }
 
+    // For Anthropic OAuth/SetupToken accounts, handle quota control settings in extra
+    if (props.account.platform === 'anthropic' && (props.account.type === 'oauth' || props.account.type === 'setup-token')) {
+      const currentExtra = (props.account.extra as Record<string, unknown>) || {}
+      const newExtra: Record<string, unknown> = { ...currentExtra }
+
+      // Window cost limit settings
+      if (windowCostEnabled.value && windowCostLimit.value != null && windowCostLimit.value > 0) {
+        newExtra.window_cost_limit = windowCostLimit.value
+        newExtra.window_cost_sticky_reserve = windowCostStickyReserve.value ?? 10
+      } else {
+        delete newExtra.window_cost_limit
+        delete newExtra.window_cost_sticky_reserve
+      }
+
+      // Session limit settings
+      if (sessionLimitEnabled.value && maxSessions.value != null && maxSessions.value > 0) {
+        newExtra.max_sessions = maxSessions.value
+        newExtra.session_idle_timeout_minutes = sessionIdleTimeout.value ?? 5
+      } else {
+        delete newExtra.max_sessions
+        delete newExtra.session_idle_timeout_minutes
+      }
+
+      updatePayload.extra = newExtra
+    }
+
     await adminAPI.accounts.update(props.account.id, updatePayload)
     appStore.showSuccess(t('admin.accounts.accountUpdated'))
     emit('updated')

frontend/src/components/common/NavigationProgress.vue

@@ -0,0 +1,109 @@
+<script setup lang="ts">
+/**
+ * 导航进度条组件
+ * 在页面顶部显示加载进度，提供导航反馈
+ */
+import { computed } from 'vue'
+import { useNavigationLoadingState } from '@/composables/useNavigationLoading'
+
+const { isLoading } = useNavigationLoadingState()
+
+// 进度条可见性
+const isVisible = computed(() => isLoading.value)
+</script>
+
+<template>
+  <Transition name="progress-fade">
+    <div
+      v-show="isVisible"
+      class="navigation-progress"
+      role="progressbar"
+      aria-label="Loading"
+      aria-valuenow="0"
+      aria-valuemin="0"
+      aria-valuemax="100"
+    >
+      <div class="navigation-progress-bar" />
+    </div>
+  </Transition>
+</template>
+
+<style scoped>
+.navigation-progress {
+  position: fixed;
+  top: 0;
+  left: 0;
+  right: 0;
+  height: 3px;
+  z-index: 9999;
+  overflow: hidden;
+  background: transparent;
+}
+
+.navigation-progress-bar {
+  height: 100%;
+  width: 100%;
+  background: linear-gradient(
+    90deg,
+    transparent 0%,
+    theme('colors.primary.400') 20%,
+    theme('colors.primary.500') 50%,
+    theme('colors.primary.400') 80%,
+    transparent 100%
+  );
+  animation: progress-slide 1.5s ease-in-out infinite;
+}
+
+/* 暗色模式下的进度条颜色 */
+:root.dark .navigation-progress-bar {
+  background: linear-gradient(
+    90deg,
+    transparent 0%,
+    theme('colors.primary.500') 20%,
+    theme('colors.primary.400') 50%,
+    theme('colors.primary.500') 80%,
+    transparent 100%
+  );
+}
+
+/* 进度条滑动动画 */
+@keyframes progress-slide {
+  0% {
+    transform: translateX(-100%);
+  }
+  100% {
+    transform: translateX(100%);
+  }
+}
+
+/* 淡入淡出过渡 */
+.progress-fade-enter-active {
+  transition: opacity 0.15s ease-out;
+}
+
+.progress-fade-leave-active {
+  transition: opacity 0.3s ease-out;
+}
+
+.progress-fade-enter-from,
+.progress-fade-leave-to {
+  opacity: 0;
+}
+
+/* 减少动画模式 */
+@media (prefers-reduced-motion: reduce) {
+  .navigation-progress-bar {
+    animation: progress-pulse 2s ease-in-out infinite;
+  }
+
+  @keyframes progress-pulse {
+    0%,
+    100% {
+      opacity: 0.4;
+    }
+    50% {
+      opacity: 1;
+    }
+  }
+}
+</style>

frontend/src/components/common/__tests__/NavigationProgress.spec.ts

@@ -0,0 +1,83 @@
+/**
+ * NavigationProgress 组件单元测试
+ */
+import { describe, it, expect, vi, beforeEach } from 'vitest'
+import { mount } from '@vue/test-utils'
+import { ref } from 'vue'
+import NavigationProgress from '../../common/NavigationProgress.vue'
+
+// Mock useNavigationLoadingState
+const mockIsLoading = ref(false)
+
+vi.mock('@/composables/useNavigationLoading', () => ({
+  useNavigationLoadingState: () => ({
+    isLoading: mockIsLoading
+  })
+}))
+
+describe('NavigationProgress', () => {
+  beforeEach(() => {
+    mockIsLoading.value = false
+  })
+
+  it('isLoading=false 时进度条应该隐藏', () => {
+    mockIsLoading.value = false
+    const wrapper = mount(NavigationProgress)
+
+    const progressBar = wrapper.find('.navigation-progress')
+    // v-show 会设置 display: none
+    expect(progressBar.isVisible()).toBe(false)
+  })
+
+  it('isLoading=true 时进度条应该可见', async () => {
+    mockIsLoading.value = true
+    const wrapper = mount(NavigationProgress)
+
+    await wrapper.vm.$nextTick()
+
+    const progressBar = wrapper.find('.navigation-progress')
+    expect(progressBar.exists()).toBe(true)
+    expect(progressBar.isVisible()).toBe(true)
+  })
+
+  it('应该有正确的 ARIA 属性', () => {
+    mockIsLoading.value = true
+    const wrapper = mount(NavigationProgress)
+
+    const progressBar = wrapper.find('.navigation-progress')
+    expect(progressBar.attributes('role')).toBe('progressbar')
+    expect(progressBar.attributes('aria-label')).toBe('Loading')
+    expect(progressBar.attributes('aria-valuemin')).toBe('0')
+    expect(progressBar.attributes('aria-valuemax')).toBe('100')
+  })
+
+  it('进度条应该有动画 class', () => {
+    mockIsLoading.value = true
+    const wrapper = mount(NavigationProgress)
+
+    const bar = wrapper.find('.navigation-progress-bar')
+    expect(bar.exists()).toBe(true)
+  })
+
+  it('应该正确响应 isLoading 状态变化', async () => {
+    // 测试初始状态为 false
+    mockIsLoading.value = false
+    const wrapper = mount(NavigationProgress)
+    await wrapper.vm.$nextTick()
+
+    // 初始状态隐藏
+    expect(wrapper.find('.navigation-progress').isVisible()).toBe(false)
+
+    // 卸载后重新挂载以测试 true 状态
+    wrapper.unmount()
+
+    // 改变为 true 后重新挂载
+    mockIsLoading.value = true
+    const wrapper2 = mount(NavigationProgress)
+    await wrapper2.vm.$nextTick()
+    expect(wrapper2.find('.navigation-progress').isVisible()).toBe(true)
+
+    // 清理
+    wrapper2.unmount()
+  })
+})

frontend/src/composables/__tests__/useNavigationLoading.spec.ts

@@ -0,0 +1,176 @@
+/**
+ * useNavigationLoading 组合式函数单元测试
+ */
+import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest'
+import {
+  useNavigationLoading,
+  _resetNavigationLoadingInstance
+} from '../useNavigationLoading'
+
+describe('useNavigationLoading', () => {
+  beforeEach(() => {
+    vi.useFakeTimers()
+    _resetNavigationLoadingInstance()
+  })
+
+  afterEach(() => {
+    vi.useRealTimers()
+  })
+
+  describe('startNavigation', () => {
+    it('导航开始时 isNavigating 应变为 true', () => {
+      const { isNavigating, startNavigation } = useNavigationLoading()
+
+      expect(isNavigating.value).toBe(false)
+
+      startNavigation()
+
+      expect(isNavigating.value).toBe(true)
+    })
+
+    it('导航开始后延迟显示加载指示器（防闪烁）', () => {
+      const { isLoading, startNavigation, ANTI_FLICKER_DELAY } = useNavigationLoading()
+
+      startNavigation()
+
+      // 立即检查，不应该显示
+      expect(isLoading.value).toBe(false)
+
+      // 经过防闪烁延迟后应该显示
+      vi.advanceTimersByTime(ANTI_FLICKER_DELAY)
+      expect(isLoading.value).toBe(true)
+    })
+  })
+
+  describe('endNavigation', () => {
+    it('导航结束时 isLoading 应变为 false', () => {
+      const { isLoading, startNavigation, endNavigation, ANTI_FLICKER_DELAY } = useNavigationLoading()
+
+      startNavigation()
+      vi.advanceTimersByTime(ANTI_FLICKER_DELAY)
+      expect(isLoading.value).toBe(true)
+
+      endNavigation()
+      expect(isLoading.value).toBe(false)
+    })
+
+    it('导航结束时 isNavigating 应变为 false', () => {
+      const { isNavigating, startNavigation, endNavigation } = useNavigationLoading()
+
+      startNavigation()
+      expect(isNavigating.value).toBe(true)
+
+      endNavigation()
+      expect(isNavigating.value).toBe(false)
+    })
+  })
+
+  describe('快速导航（< 100ms）防闪烁', () => {
+    it('快速导航不应触发显示加载指示器', () => {
+      const { isLoading, startNavigation, endNavigation, ANTI_FLICKER_DELAY } = useNavigationLoading()
+
+      startNavigation()
+
+      // 在防闪烁延迟之前结束导航
+      vi.advanceTimersByTime(ANTI_FLICKER_DELAY - 50)
+      endNavigation()
+
+      // 不应该显示加载指示器
+      expect(isLoading.value).toBe(false)
+
+      // 即使继续等待也不应该显示
+      vi.advanceTimersByTime(100)
+      expect(isLoading.value).toBe(false)
+    })
+  })
+
+  describe('cancelNavigation', () => {
+    it('导航取消时应正确重置状态', () => {
+      const { isLoading, startNavigation, cancelNavigation, ANTI_FLICKER_DELAY } = useNavigationLoading()
+
+      startNavigation()
+      vi.advanceTimersByTime(ANTI_FLICKER_DELAY / 2)
+
+      cancelNavigation()
+
+      // 取消后不应该触发显示
+      vi.advanceTimersByTime(ANTI_FLICKER_DELAY)
+      expect(isLoading.value).toBe(false)
+    })
+  })
+
+  describe('getNavigationDuration', () => {
+    it('应该返回正确的导航持续时间', () => {
+      const { startNavigation, getNavigationDuration } = useNavigationLoading()
+
+      expect(getNavigationDuration()).toBeNull()
+
+      startNavigation()
+      vi.advanceTimersByTime(500)
+
+      const duration = getNavigationDuration()
+      expect(duration).toBe(500)
+    })
+
+    it('导航结束后应返回 null', () => {
+      const { startNavigation, endNavigation, getNavigationDuration } = useNavigationLoading()
+
+      startNavigation()
+      vi.advanceTimersByTime(500)
+      endNavigation()
+
+      expect(getNavigationDuration()).toBeNull()
+    })
+  })
+
+  describe('resetState', () => {
+    it('应该重置所有状态', () => {
+      const { isLoading, isNavigating, startNavigation, resetState, ANTI_FLICKER_DELAY } = useNavigationLoading()
+
+      startNavigation()
+      vi.advanceTimersByTime(ANTI_FLICKER_DELAY)
+
+      expect(isLoading.value).toBe(true)
+      expect(isNavigating.value).toBe(true)
+
+      resetState()
+
+      expect(isLoading.value).toBe(false)
+      expect(isNavigating.value).toBe(false)
+    })
+  })
+
+  describe('连续导航场景', () => {
+    it('连续快速导航应正确处理状态', () => {
+      const { isLoading, startNavigation, cancelNavigation, endNavigation, ANTI_FLICKER_DELAY } = useNavigationLoading()
+
+      // 第一次导航
+      startNavigation()
+      vi.advanceTimersByTime(30)
+
+      // 第二次导航（取消第一次）
+      cancelNavigation()
+      startNavigation()
+      vi.advanceTimersByTime(30)
+
+      // 第三次导航（取消第二次）
+      cancelNavigation()
+      startNavigation()
+
+      // 这次等待足够长时间
+      vi.advanceTimersByTime(ANTI_FLICKER_DELAY)
+      expect(isLoading.value).toBe(true)
+
+      // 结束导航
+      endNavigation()
+      expect(isLoading.value).toBe(false)
+    })
+  })
+
+  describe('ANTI_FLICKER_DELAY 常量', () => {
+    it('应该为 100ms', () => {
+      const { ANTI_FLICKER_DELAY } = useNavigationLoading()
+      expect(ANTI_FLICKER_DELAY).toBe(100)
+    })
+  })
+})

frontend/src/composables/__tests__/useRoutePrefetch.spec.ts

@@ -0,0 +1,244 @@
+/**
+ * useRoutePrefetch 组合式函数单元测试
+ */
+import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest'
+import type { RouteLocationNormalized, Router, RouteRecordNormalized } from 'vue-router'
+
+import { useRoutePrefetch, _adminPrefetchMap, _userPrefetchMap } from '../useRoutePrefetch'
+
+// Mock 路由对象
+const createMockRoute = (path: string): RouteLocationNormalized => ({
+  path,
+  name: undefined,
+  params: {},
+  query: {},
+  hash: '',
+  fullPath: path,
+  matched: [],
+  meta: {},
+  redirectedFrom: undefined
+})
+
+// Mock Router
+const createMockRouter = (): Router => {
+  const mockImportFn = vi.fn().mockResolvedValue({ default: {} })
+
+  const routes: Partial<RouteRecordNormalized>[] = [
+    { path: '/admin/dashboard', components: { default: mockImportFn } },
+    { path: '/admin/accounts', components: { default: mockImportFn } },
+    { path: '/admin/users', components: { default: mockImportFn } },
+    { path: '/admin/groups', components: { default: mockImportFn } },
+    { path: '/admin/subscriptions', components: { default: mockImportFn } },
+    { path: '/admin/redeem', components: { default: mockImportFn } },
+    { path: '/dashboard', components: { default: mockImportFn } },
+    { path: '/keys', components: { default: mockImportFn } },
+    { path: '/usage', components: { default: mockImportFn } },
+    { path: '/redeem', components: { default: mockImportFn } },
+    { path: '/profile', components: { default: mockImportFn } }
+  ]
+
+  return {
+    getRoutes: () => routes as RouteRecordNormalized[]
+  } as Router
+}
+
+describe('useRoutePrefetch', () => {
+  let originalRequestIdleCallback: typeof window.requestIdleCallback
+  let originalCancelIdleCallback: typeof window.cancelIdleCallback
+  let mockRouter: Router
+
+  beforeEach(() => {
+    mockRouter = createMockRouter()
+
+    // 保存原始函数
+    originalRequestIdleCallback = window.requestIdleCallback
+    originalCancelIdleCallback = window.cancelIdleCallback
+
+    // Mock requestIdleCallback 立即执行
+    vi.stubGlobal('requestIdleCallback', (cb: IdleRequestCallback) => {
+      const id = setTimeout(() => cb({ didTimeout: false, timeRemaining: () => 50 }), 0)
+      return id
+    })
+    vi.stubGlobal('cancelIdleCallback', (id: number) => clearTimeout(id))
+  })
+
+  afterEach(() => {
+    vi.restoreAllMocks()
+    // 恢复原始函数
+    window.requestIdleCallback = originalRequestIdleCallback
+    window.cancelIdleCallback = originalCancelIdleCallback
+  })
+
+  describe('_isAdminRoute', () => {
+    it('应该正确识别管理员路由', () => {
+      const { _isAdminRoute } = useRoutePrefetch(mockRouter)
+      expect(_isAdminRoute('/admin/dashboard')).toBe(true)
+      expect(_isAdminRoute('/admin/users')).toBe(true)
+      expect(_isAdminRoute('/admin/accounts')).toBe(true)
+    })
+
+    it('应该正确识别非管理员路由', () => {
+      const { _isAdminRoute } = useRoutePrefetch(mockRouter)
+      expect(_isAdminRoute('/dashboard')).toBe(false)
+      expect(_isAdminRoute('/keys')).toBe(false)
+      expect(_isAdminRoute('/usage')).toBe(false)
+    })
+  })
+
+  describe('_getPrefetchConfig', () => {
+    it('管理员 dashboard 应该返回正确的预加载配置', () => {
+      const { _getPrefetchConfig } = useRoutePrefetch(mockRouter)
+      const route = createMockRoute('/admin/dashboard')
+      const config = _getPrefetchConfig(route)
+
+      expect(config).toHaveLength(2)
+    })
+
+    it('普通用户 dashboard 应该返回正确的预加载配置', () => {
+      const { _getPrefetchConfig } = useRoutePrefetch(mockRouter)
+      const route = createMockRoute('/dashboard')
+      const config = _getPrefetchConfig(route)
+
+      expect(config).toHaveLength(2)
+    })
+
+    it('未定义的路由应该返回空数组', () => {
+      const { _getPrefetchConfig } = useRoutePrefetch(mockRouter)
+      const route = createMockRoute('/unknown-route')
+      const config = _getPrefetchConfig(route)
+
+      expect(config).toHaveLength(0)
+    })
+  })
+
+  describe('triggerPrefetch', () => {
+    it('应该在浏览器空闲时触发预加载', async () => {
+      const { triggerPrefetch, prefetchedRoutes } = useRoutePrefetch(mockRouter)
+      const route = createMockRoute('/admin/dashboard')
+
+      triggerPrefetch(route)
+
+      // 等待 requestIdleCallback 执行
+      await new Promise((resolve) => setTimeout(resolve, 100))
+
+      expect(prefetchedRoutes.value.has('/admin/dashboard')).toBe(true)
+    })
+
+    it('应该避免重复预加载同一路由', async () => {
+      const { triggerPrefetch, prefetchedRoutes } = useRoutePrefetch(mockRouter)
+      const route = createMockRoute('/admin/dashboard')
+
+      triggerPrefetch(route)
+      await new Promise((resolve) => setTimeout(resolve, 100))
+
+      // 第二次触发
+      triggerPrefetch(route)
+      await new Promise((resolve) => setTimeout(resolve, 100))
+
+      // 只应该预加载一次
+      expect(prefetchedRoutes.value.size).toBe(1)
+    })
+  })
+
+  describe('cancelPendingPrefetch', () => {
+    it('应该取消挂起的预加载任务', () => {
+      const { triggerPrefetch, cancelPendingPrefetch, prefetchedRoutes } = useRoutePrefetch(mockRouter)
+      const route = createMockRoute('/admin/dashboard')
+
+      triggerPrefetch(route)
+      cancelPendingPrefetch()
+
+      // 不应该有预加载完成
+      expect(prefetchedRoutes.value.size).toBe(0)
+    })
+  })
+
+  describe('路由变化时取消之前的预加载', () => {
+    it('应该在路由变化时取消之前的预加载任务', async () => {
+      const { triggerPrefetch, prefetchedRoutes } = useRoutePrefetch(mockRouter)
+
+      // 触发第一个路由的预加载
+      triggerPrefetch(createMockRoute('/admin/dashboard'))
+
+      // 立即切换到另一个路由
+      triggerPrefetch(createMockRoute('/admin/users'))
+
+      // 等待执行
+      await new Promise((resolve) => setTimeout(resolve, 100))
+
+      // 只有最后一个路由应该被预加载
+      expect(prefetchedRoutes.value.has('/admin/users')).toBe(true)
+    })
+  })
+
+  describe('resetPrefetchState', () => {
+    it('应该重置所有预加载状态', async () => {
+      const { triggerPrefetch, resetPrefetchState, prefetchedRoutes } = useRoutePrefetch(mockRouter)
+      const route = createMockRoute('/admin/dashboard')
+
+      triggerPrefetch(route)
+      await new Promise((resolve) => setTimeout(resolve, 100))
+
+      expect(prefetchedRoutes.value.size).toBeGreaterThan(0)
+
+      resetPrefetchState()
+
+      expect(prefetchedRoutes.value.size).toBe(0)
+    })
+  })
+
+  describe('预加载映射表', () => {
+    it('管理员预加载映射表应该包含正确的路由', () => {
+      expect(_adminPrefetchMap).toHaveProperty('/admin/dashboard')
+      expect(_adminPrefetchMap['/admin/dashboard']).toHaveLength(2)
+    })
+
+    it('用户预加载映射表应该包含正确的路由', () => {
+      expect(_userPrefetchMap).toHaveProperty('/dashboard')
+      expect(_userPrefetchMap['/dashboard']).toHaveLength(2)
+    })
+  })
+
+  describe('requestIdleCallback 超时处理', () => {
+    it('超时后仍能正常执行预加载', async () => {
+      // 模拟超时情况
+      vi.stubGlobal('requestIdleCallback', (cb: IdleRequestCallback, options?: IdleRequestOptions) => {
+        const timeout = options?.timeout || 2000
+        return setTimeout(() => cb({ didTimeout: true, timeRemaining: () => 0 }), timeout)
+      })
+
+      const { triggerPrefetch, prefetchedRoutes } = useRoutePrefetch(mockRouter)
+      const route = createMockRoute('/dashboard')
+
+      triggerPrefetch(route)
+
+      // 等待超时执行
+      await new Promise((resolve) => setTimeout(resolve, 2100))
+
+      expect(prefetchedRoutes.value.has('/dashboard')).toBe(true)
+    })
+  })
+
+  describe('预加载失败处理', () => {
+    it('预加载失败时应该静默处理不影响页面功能', async () => {
+      const { triggerPrefetch } = useRoutePrefetch(mockRouter)
+      const route = createMockRoute('/admin/dashboard')
+
+      // 不应该抛出异常
+      expect(() => triggerPrefetch(route)).not.toThrow()
+    })
+  })
+
+  describe('无 router 时的行为', () => {
+    it('没有传入 router 时应该正常工作但不执行预加载', async () => {
+      const { triggerPrefetch, prefetchedRoutes } = useRoutePrefetch()
+      const route = createMockRoute('/admin/dashboard')
+
+      triggerPrefetch(route)
+      await new Promise((resolve) => setTimeout(resolve, 100))
+
+      // 没有 router，无法获取组件，所以不会预加载
+      expect(prefetchedRoutes.value.size).toBe(0)
+    })
+  })
+})

frontend/src/composables/useNavigationLoading.ts

@@ -0,0 +1,132 @@
+/**
+ * 导航加载状态组合式函数
+ * 管理路由切换时的加载状态，支持防闪烁逻辑
+ */
+import { ref, readonly, computed } from 'vue'
+
+/**
+ * 导航加载状态管理
+ *
+ * 功能：
+ * 1. 在路由切换时显示加载状态
+ * 2. 快速导航（< 100ms）不显示加载指示器（防闪烁）
+ * 3. 导航取消时正确重置状态
+ */
+export function useNavigationLoading() {
+  // 内部加载状态
+  const _isLoading = ref(false)
+
+  // 导航开始时间（用于防闪烁计算）
+  let navigationStartTime: number | null = null
+
+  // 防闪烁延迟计时器
+  let showLoadingTimer: ReturnType<typeof setTimeout> | null = null
+
+  // 是否应该显示加载指示器（考虑防闪烁逻辑）
+  const shouldShowLoading = ref(false)
+
+  // 防闪烁延迟时间（毫秒）
+  const ANTI_FLICKER_DELAY = 100
+
+  /**
+   * 清理计时器
+   */
+  const clearTimer = (): void => {
+    if (showLoadingTimer !== null) {
+      clearTimeout(showLoadingTimer)
+      showLoadingTimer = null
+    }
+  }
+
+  /**
+   * 导航开始时调用
+   */
+  const startNavigation = (): void => {
+    navigationStartTime = Date.now()
+    _isLoading.value = true
+
+    // 延迟显示加载指示器，实现防闪烁
+    clearTimer()
+    showLoadingTimer = setTimeout(() => {
+      if (_isLoading.value) {
+        shouldShowLoading.value = true
+      }
+    }, ANTI_FLICKER_DELAY)
+  }
+
+  /**
+   * 导航结束时调用
+   */
+  const endNavigation = (): void => {
+    clearTimer()
+    _isLoading.value = false
+    shouldShowLoading.value = false
+    navigationStartTime = null
+  }
+
+  /**
+   * 导航取消时调用（比如快速连续点击不同链接）
+   */
+  const cancelNavigation = (): void => {
+    clearTimer()
+    // 保持加载状态，因为新的导航会立即开始
+    // 但重置导航开始时间
+    navigationStartTime = null
+  }
+
+  /**
+   * 重置所有状态（用于测试）
+   */
+  const resetState = (): void => {
+    clearTimer()
+    _isLoading.value = false
+    shouldShowLoading.value = false
+    navigationStartTime = null
+  }
+
+  /**
+   * 获取导航持续时间（毫秒）
+   */
+  const getNavigationDuration = (): number | null => {
+    if (navigationStartTime === null) {
+      return null
+    }
+    return Date.now() - navigationStartTime
+  }
+
+  // 公开的加载状态（只读）
+  const isLoading = computed(() => shouldShowLoading.value)
+
+  // 内部加载状态（用于测试，不考虑防闪烁）
+  const isNavigating = readonly(_isLoading)
+
+  return {
+    isLoading,
+    isNavigating,
+    startNavigation,
+    endNavigation,
+    cancelNavigation,
+    resetState,
+    getNavigationDuration,
+    // 导出常量用于测试
+    ANTI_FLICKER_DELAY
+  }
+}
+
+// 创建单例实例，供全局使用
+let navigationLoadingInstance: ReturnType<typeof useNavigationLoading> | null = null
+
+export function useNavigationLoadingState() {
+  if (!navigationLoadingInstance) {
+    navigationLoadingInstance = useNavigationLoading()
+  }
+  return navigationLoadingInstance
+}
+
+// 导出重置函数（用于测试）
+export function _resetNavigationLoadingInstance(): void {
+  if (navigationLoadingInstance) {
+    navigationLoadingInstance.resetState()
+  }
+  navigationLoadingInstance = null
+}

frontend/src/composables/useRoutePrefetch.ts

@@ -0,0 +1,202 @@
+/**
+ * 路由预加载组合式函数
+ * 在浏览器空闲时预加载可能访问的下一个页面，提升导航体验
+ *
+ * 优化说明：
+ * - 不使用静态 import() 映射表，避免增加入口文件大小
+ * - 通过路由配置动态获取组件的 import 函数
+ * - 只在实际需要预加载时才执行
+ */
+import { ref, readonly } from 'vue'
+import type { RouteLocationNormalized, Router } from 'vue-router'
+
+/**
+ * 组件导入函数类型
+ */
+type ComponentImportFn = () => Promise<unknown>
+
+/**
+ * 预加载邻接表：定义每个路由应该预加载哪些相邻路由
+ * 只存储路由路径，不存储 import 函数，避免打包问题
+ */
+const PREFETCH_ADJACENCY: Record<string, string[]> = {
+  // Admin routes - 预加载最常访问的相邻页面
+  '/admin/dashboard': ['/admin/accounts', '/admin/users'],
+  '/admin/accounts': ['/admin/dashboard', '/admin/users'],
+  '/admin/users': ['/admin/groups', '/admin/dashboard'],
+  '/admin/groups': ['/admin/subscriptions', '/admin/users'],
+  '/admin/subscriptions': ['/admin/groups', '/admin/redeem'],
+  // User routes
+  '/dashboard': ['/keys', '/usage'],
+  '/keys': ['/dashboard', '/usage'],
+  '/usage': ['/keys', '/redeem'],
+  '/redeem': ['/usage', '/profile'],
+  '/profile': ['/dashboard', '/keys']
+}
+
+/**
+ * requestIdleCallback 的返回类型
+ */
+type IdleCallbackHandle = number | ReturnType<typeof setTimeout>
+
+/**
+ * requestIdleCallback polyfill (Safari < 15)
+ */
+const scheduleIdleCallback = (
+  callback: IdleRequestCallback,
+  options?: IdleRequestOptions
+): IdleCallbackHandle => {
+  if (typeof window.requestIdleCallback === 'function') {
+    return window.requestIdleCallback(callback, options)
+  }
+  return setTimeout(() => {
+    callback({ didTimeout: false, timeRemaining: () => 50 })
+  }, 1000)
+}
+
+const cancelScheduledCallback = (handle: IdleCallbackHandle): void => {
+  if (typeof window.cancelIdleCallback === 'function' && typeof handle === 'number') {
+    window.cancelIdleCallback(handle)
+  } else {
+    clearTimeout(handle)
+  }
+}
+
+/**
+ * 路由预加载组合式函数
+ *
+ * @param router - Vue Router 实例，用于获取路由组件
+ */
+export function useRoutePrefetch(router?: Router) {
+  // 当前挂起的预加载任务句柄
+  const pendingPrefetchHandle = ref<IdleCallbackHandle | null>(null)
+
+  // 已预加载的路由集合
+  const prefetchedRoutes = ref<Set<string>>(new Set())
+
+  /**
+   * 从路由配置中获取组件的 import 函数
+   */
+  const getComponentImporter = (path: string): ComponentImportFn | null => {
+    if (!router) return null
+
+    const routes = router.getRoutes()
+    const route = routes.find((r) => r.path === path)
+
+    if (route && route.components?.default) {
+      const component = route.components.default
+      // 检查是否是懒加载组件（函数形式）
+      if (typeof component === 'function') {
+        return component as ComponentImportFn
+      }
+    }
+    return null
+  }
+
+  /**
+   * 获取当前路由应该预加载的路由路径列表
+   */
+  const getPrefetchPaths = (route: RouteLocationNormalized): string[] => {
+    return PREFETCH_ADJACENCY[route.path] || []
+  }
+
+  /**
+   * 执行单个组件的预加载
+   */
+  const prefetchComponent = async (importFn: ComponentImportFn): Promise<void> => {
+    try {
+      await importFn()
+    } catch (error) {
+      // 静默处理预加载错误
+      if (import.meta.env.DEV) {
+        console.debug('[Prefetch] Failed to prefetch component:', error)
+      }
+    }
+  }
+
+  /**
+   * 取消挂起的预加载任务
+   */
+  const cancelPendingPrefetch = (): void => {
+    if (pendingPrefetchHandle.value !== null) {
+      cancelScheduledCallback(pendingPrefetchHandle.value)
+      pendingPrefetchHandle.value = null
+    }
+  }
+
+  /**
+   * 触发路由预加载
+   */
+  const triggerPrefetch = (route: RouteLocationNormalized): void => {
+    cancelPendingPrefetch()
+
+    const prefetchPaths = getPrefetchPaths(route)
+    if (prefetchPaths.length === 0) return
+
+    pendingPrefetchHandle.value = scheduleIdleCallback(
+      () => {
+        pendingPrefetchHandle.value = null
+
+        const routePath = route.path
+        if (prefetchedRoutes.value.has(routePath)) return
+
+        // 获取需要预加载的组件 import 函数
+        const importFns: ComponentImportFn[] = []
+        for (const path of prefetchPaths) {
+          const importFn = getComponentImporter(path)
+          if (importFn) {
+            importFns.push(importFn)
+          }
+        }
+
+        if (importFns.length > 0) {
+          Promise.all(importFns.map(prefetchComponent)).then(() => {
+            prefetchedRoutes.value.add(routePath)
+          })
+        }
+      },
+      { timeout: 2000 }
+    )
+  }
+
+  /**
+   * 重置预加载状态
+   */
+  const resetPrefetchState = (): void => {
+    cancelPendingPrefetch()
+    prefetchedRoutes.value.clear()
+  }
+
+  /**
+   * 判断是否为管理员路由
+   */
+  const isAdminRoute = (path: string): boolean => {
+    return path.startsWith('/admin')
+  }
+
+  /**
+   * 获取预加载配置（兼容旧 API）
+   */
+  const getPrefetchConfig = (route: RouteLocationNormalized): ComponentImportFn[] => {
+    const paths = getPrefetchPaths(route)
+    const importFns: ComponentImportFn[] = []
+    for (const path of paths) {
+      const importFn = getComponentImporter(path)
+      if (importFn) importFns.push(importFn)
+    }
+    return importFns
+  }
+
+  return {
+    prefetchedRoutes: readonly(prefetchedRoutes),
+    triggerPrefetch,
+    cancelPendingPrefetch,
+    resetPrefetchState,
+    _getPrefetchConfig: getPrefetchConfig,
+    _isAdminRoute: isAdminRoute
+  }
+}
+
+// 兼容旧测试的导出
+export const _adminPrefetchMap = PREFETCH_ADJACENCY
+export const _userPrefetchMap = PREFETCH_ADJACENCY

frontend/src/i18n/locales/en.ts

@@ -163,6 +163,7 @@ export default {
     notAvailable: 'N/A',
     now: 'Now',
     unknown: 'Unknown',
+    minutes: 'min',
     time: {
       never: 'Never',
       justNow: 'Just now',
@@ -1082,7 +1083,7 @@ export default {
         platformType: 'Platform/Type',
         platform: 'Platform',
         type: 'Type',
-        concurrencyStatus: 'Concurrency',
+        capacity: 'Capacity',
         notes: 'Notes',
         priority: 'Priority',
         billingRateMultiplier: 'Billing Rate',
@@ -1096,6 +1097,18 @@ export default {
         expiresAt: 'Expires At',
         actions: 'Actions'
       },
+      // Capacity status tooltips
+      capacity: {
+        windowCost: {
+          blocked: '5h window cost exceeded, account scheduling paused',
+          stickyOnly: '5h window cost at threshold, only sticky sessions allowed',
+          normal: '5h window cost normal'
+        },
+        sessions: {
+          full: 'Active sessions full, new sessions must wait (idle timeout: {idle} min)',
+          normal: 'Active sessions normal (idle timeout: {idle} min)'
+        }
+      },
       tempUnschedulable: {
         title: 'Temp Unschedulable',
         statusTitle: 'Temp Unschedulable Status',
@@ -1247,6 +1260,31 @@ export default {
         'When enabled, warmup requests like title generation will return mock responses without consuming upstream tokens',
       autoPauseOnExpired: 'Auto Pause On Expired',
       autoPauseOnExpiredDesc: 'When enabled, the account will auto pause scheduling after it expires',
+      // Quota control (Anthropic OAuth/SetupToken only)
+      quotaControl: {
+        title: 'Quota Control',
+        hint: 'Only applies to Anthropic OAuth/Setup Token accounts',
+        windowCost: {
+          label: '5h Window Cost Limit',
+          hint: 'Limit account cost usage within the 5-hour window',
+          limit: 'Cost Threshold',
+          limitPlaceholder: '50',
+          limitHint: 'Account will not participate in new scheduling after reaching threshold',
+          stickyReserve: 'Sticky Reserve',
+          stickyReservePlaceholder: '10',
+          stickyReserveHint: 'Additional reserve for sticky sessions'
+        },
+        sessionLimit: {
+          label: 'Session Count Limit',
+          hint: 'Limit the number of active concurrent sessions',
+          maxSessions: 'Max Sessions',
+          maxSessionsPlaceholder: '3',
+          maxSessionsHint: 'Maximum number of active concurrent sessions',
+          idleTimeout: 'Idle Timeout',
+          idleTimeoutPlaceholder: '5',
+          idleTimeoutHint: 'Sessions will be released after idle timeout'
+        }
+      },
       expired: 'Expired',
       proxy: 'Proxy',
       noProxy: 'No Proxy',

frontend/src/i18n/locales/zh.ts

@@ -160,6 +160,7 @@ export default {
     notAvailable: '不可用',
     now: '现在',
     unknown: '未知',
+    minutes: '分钟',
     time: {
       never: '从未',
       justNow: '刚刚',
@@ -1131,7 +1132,7 @@ export default {
         platformType: '平台/类型',
         platform: '平台',
         type: '类型',
-        concurrencyStatus: '并发',
+        capacity: '容量',
         notes: '备注',
         priority: '优先级',
         billingRateMultiplier: '账号倍率',
@@ -1145,6 +1146,18 @@ export default {
         expiresAt: '过期时间',
         actions: '操作'
       },
+      // 容量状态提示
+      capacity: {
+        windowCost: {
+          blocked: '5h窗口费用超限，账号暂停调度',
+          stickyOnly: '5h窗口费用达阈值，仅允许粘性会话',
+          normal: '5h窗口费用正常'
+        },
+        sessions: {
+          full: '活跃会话已满，新会话需等待（空闲超时：{idle}分钟）',
+          normal: '活跃会话正常（空闲超时：{idle}分钟）'
+        }
+      },
       clearRateLimit: '清除速率限制',
       testConnection: '测试连接',
       reAuthorize: '重新授权',
@@ -1380,6 +1393,31 @@ export default {
       interceptWarmupRequestsDesc: '启用后，标题生成等预热请求将返回 mock 响应，不消耗上游 token',
       autoPauseOnExpired: '过期自动暂停调度',
       autoPauseOnExpiredDesc: '启用后，账号过期将自动暂停调度',
+      // Quota control (Anthropic OAuth/SetupToken only)
+      quotaControl: {
+        title: '配额控制',
+        hint: '仅适用于 Anthropic OAuth/Setup Token 账号',
+        windowCost: {
+          label: '5h窗口费用控制',
+          hint: '限制账号在5小时窗口内的费用使用',
+          limit: '费用阈值',
+          limitPlaceholder: '50',
+          limitHint: '达到阈值后不参与新请求调度',
+          stickyReserve: '粘性预留额度',
+          stickyReservePlaceholder: '10',
+          stickyReserveHint: '为粘性会话预留的额外额度'
+        },
+        sessionLimit: {
+          label: '会话数量控制',
+          hint: '限制同时活跃的会话数量',
+          maxSessions: '最大会话数',
+          maxSessionsPlaceholder: '3',
+          maxSessionsHint: '同时活跃的最大会话数量',
+          idleTimeout: '空闲超时',
+          idleTimeoutPlaceholder: '5',
+          idleTimeoutHint: '会话空闲超时后自动释放'
+        }
+      },
       expired: '已过期',
       proxy: '代理',
       noProxy: '无代理',

frontend/src/router/index.ts

@@ -6,6 +6,8 @@
 import { createRouter, createWebHistory, type RouteRecordRaw } from 'vue-router'
 import { useAuthStore } from '@/stores/auth'
 import { useAppStore } from '@/stores/app'
+import { useNavigationLoadingState } from '@/composables/useNavigationLoading'
+import { useRoutePrefetch } from '@/composables/useRoutePrefetch'
 
 /**
  * Route definitions with lazy loading
@@ -326,7 +328,15 @@ const router = createRouter({
  */
 let authInitialized = false
 
+// 初始化导航加载状态和预加载
+const navigationLoading = useNavigationLoadingState()
+// 延迟初始化预加载，传入 router 实例
+let routePrefetch: ReturnType<typeof useRoutePrefetch> | null = null
+
 router.beforeEach((to, _from, next) => {
+  // 开始导航加载状态
+  navigationLoading.startNavigation()
+
   const authStore = useAuthStore()
 
   // Restore auth state from localStorage on first navigation (page refresh)
@@ -398,6 +408,21 @@ router.beforeEach((to, _from, next) => {
   next()
 })
 
+/**
+ * Navigation guard: End loading and trigger prefetch
+ */
+router.afterEach((to) => {
+  // 结束导航加载状态
+  navigationLoading.endNavigation()
+
+  // 懒初始化预加载（首次导航时创建，传入 router 实例）
+  if (!routePrefetch) {
+    routePrefetch = useRoutePrefetch(router)
+  }
+  // 触发路由预加载（在浏览器空闲时执行）
+  routePrefetch.triggerPrefetch(to)
+})
+
 /**
  * Navigation guard: Error handling
  * Handles dynamic import failures caused by deployment updates

frontend/src/style.css

@@ -1,5 +1,3 @@
-@import url('https://fonts.googleapis.com/css2?family=Inter:wght@300;400;500;600;700&display=swap');
-
 @tailwind base;
 @tailwind components;
 @tailwind utilities;
@@ -11,7 +9,6 @@
 
   html {
     @apply scroll-smooth antialiased;
-    font-feature-settings: 'cv02', 'cv03', 'cv04', 'cv11';
   }
 
   body {

frontend/src/types/index.ts

@@ -471,6 +471,18 @@ export interface Account {
   session_window_start: string | null
   session_window_end: string | null
   session_window_status: 'allowed' | 'allowed_warning' | 'rejected' | null
+
+  // 5h窗口费用控制（仅 Anthropic OAuth/SetupToken 账号有效）
+  window_cost_limit?: number | null
+  window_cost_sticky_reserve?: number | null
+
+  // 会话数量控制（仅 Anthropic OAuth/SetupToken 账号有效）
+  max_sessions?: number | null
+  session_idle_timeout_minutes?: number | null
+
+  // 运行时状态（仅当启用对应限制时返回）
+  current_window_cost?: number | null // 当前窗口费用
+  active_sessions?: number | null // 当前活跃会话数
 }
 
 // Account Usage types

frontend/src/views/admin/AccountsView.vue

@@ -34,15 +34,8 @@
           <template #cell-platform_type="{ row }">
             <PlatformTypeBadge :platform="row.platform" :type="row.type" />
           </template>
-          <template #cell-concurrency="{ row }">
+          <template #cell-capacity="{ row }">
-            <div class="flex items-center gap-1.5">
+            <AccountCapacityCell :account="row" />
-              <span :class="['inline-flex items-center gap-1 rounded-md px-2 py-0.5 text-xs font-medium', (row.current_concurrency || 0) >= row.concurrency ? 'bg-red-100 text-red-700 dark:bg-red-900/30 dark:text-red-400' : (row.current_concurrency || 0) > 0 ? 'bg-yellow-100 text-yellow-700 dark:bg-yellow-900/30 dark:text-yellow-400' : 'bg-gray-100 text-gray-600 dark:bg-gray-800 dark:text-gray-400']">
-                <svg class="h-3 w-3" fill="none" viewBox="0 0 24 24" stroke="currentColor" stroke-width="2"><path stroke-linecap="round" stroke-linejoin="round" d="M3.75 6A2.25 2.25 0 016 3.75h2.25A2.25 2.25 0 0110.5 6v2.25a2.25 2.25 0 01-2.25 2.25H6a2.25 2.25 0 01-2.25-2.25V6zM3.75 15.75A2.25 2.25 0 016 13.5h2.25a2.25 2.25 0 012.25 2.25V18a2.25 2.25 0 01-2.25 2.25H6A2.25 2.25 0 013.75 18v-2.25zM13.5 6a2.25 2.25 0 012.25-2.25H18A2.25 2.25 0 0120.25 6v2.25A2.25 2.25 0 0118 10.5h-2.25a2.25 2.25 0 01-2.25-2.25V6zM13.5 15.75a2.25 2.25 0 012.25-2.25H18a2.25 2.25 0 012.25 2.25V18A2.25 2.25 0 0118 20.25h-2.25A2.25 2.25 0 0113.5 18v-2.25z" /></svg>
-                <span class="font-mono">{{ row.current_concurrency || 0 }}</span>
-                <span class="text-gray-400 dark:text-gray-500">/</span>
-                <span class="font-mono">{{ row.concurrency }}</span>
-              </span>
-            </div>
           </template>
           <template #cell-status="{ row }">
             <AccountStatusIndicator :account="row" @show-temp-unsched="handleShowTempUnsched" />
@@ -148,6 +141,7 @@ import AccountStatusIndicator from '@/components/account/AccountStatusIndicator.
 import AccountUsageCell from '@/components/account/AccountUsageCell.vue'
 import AccountTodayStatsCell from '@/components/account/AccountTodayStatsCell.vue'
 import AccountGroupsCell from '@/components/account/AccountGroupsCell.vue'
+import AccountCapacityCell from '@/components/account/AccountCapacityCell.vue'
 import PlatformTypeBadge from '@/components/common/PlatformTypeBadge.vue'
 import { formatDateTime, formatRelativeTime } from '@/utils/format'
 import type { Account, Proxy, Group } from '@/types'
@@ -187,7 +181,7 @@ const cols = computed(() => {
     { key: 'select', label: '', sortable: false },
     { key: 'name', label: t('admin.accounts.columns.name'), sortable: true },
     { key: 'platform_type', label: t('admin.accounts.columns.platformType'), sortable: false },
-    { key: 'concurrency', label: t('admin.accounts.columns.concurrencyStatus'), sortable: false },
+    { key: 'capacity', label: t('admin.accounts.columns.capacity'), sortable: false },
     { key: 'status', label: t('admin.accounts.columns.status'), sortable: true },
     { key: 'schedulable', label: t('admin.accounts.columns.schedulable'), sortable: true },
     { key: 'today_stats', label: t('admin.accounts.columns.todayStats'), sortable: false }

frontend/tailwind.config.js

@@ -50,16 +50,19 @@ export default {
       },
       fontFamily: {
         sans: [
-          'Inter',
+          'system-ui',
           '-apple-system',
           'BlinkMacSystemFont',
           'Segoe UI',
           'Roboto',
           'Helvetica Neue',
           'Arial',
+          'PingFang SC',
+          'Hiragino Sans GB',
+          'Microsoft YaHei',
           'sans-serif'
         ],
-        mono: ['JetBrains Mono', 'Fira Code', 'Monaco', 'Consolas', 'monospace']
+        mono: ['ui-monospace', 'SFMono-Regular', 'Menlo', 'Monaco', 'Consolas', 'monospace']
       },
       boxShadow: {
         glass: '0 8px 32px rgba(0, 0, 0, 0.08)',

frontend/vite.config.ts

@@ -58,7 +58,49 @@ export default defineConfig({
   },
   build: {
     outDir: '../backend/internal/web/dist',
-    emptyOutDir: true
+    emptyOutDir: true,
+    rollupOptions: {
+      output: {
+        /**
+         * 手动分包配置
+         * 分离第三方库并按功能合并应用代码，避免循环依赖
+         */
+        manualChunks(id: string) {
+          if (id.includes('node_modules')) {
+            // Vue 核心库
+            if (
+              id.includes('/vue/') ||
+              id.includes('/vue-router/') ||
+              id.includes('/pinia/') ||
+              id.includes('/@vue/')
+            ) {
+              return 'vendor-vue'
+            }
+
+            // UI 工具库（较大，单独分离）
+            if (id.includes('/@vueuse/') || id.includes('/xlsx/')) {
+              return 'vendor-ui'
+            }
+
+            // 图表库
+            if (id.includes('/chart.js/') || id.includes('/vue-chartjs/')) {
+              return 'vendor-chart'
+            }
+
+            // 国际化
+            if (id.includes('/vue-i18n/') || id.includes('/@intlify/')) {
+              return 'vendor-i18n'
+            }
+
+            // 其他小型第三方库合并
+            return 'vendor-misc'
+          }
+
+          // 应用代码：按入口点自动分包，不手动干预
+          // 这样可以避免循环依赖，同时保持合理的 chunk 数量
+        }
+      }
+    }
   },
   server: {
     host: '0.0.0.0',

frontend/vitest.config.ts

@@ -0,0 +1,35 @@
+import { defineConfig, mergeConfig } from 'vitest/config'
+import viteConfig from './vite.config'
+
+export default mergeConfig(
+  viteConfig,
+  defineConfig({
+    test: {
+      globals: true,
+      environment: 'jsdom',
+      include: ['src/**/*.{test,spec}.{js,ts,jsx,tsx}'],
+      exclude: ['node_modules', 'dist'],
+      coverage: {
+        provider: 'v8',
+        reporter: ['text', 'json', 'html'],
+        include: ['src/**/*.{js,ts,vue}'],
+        exclude: [
+          'node_modules',
+          'src/**/*.d.ts',
+          'src/**/*.spec.ts',
+          'src/**/*.test.ts',
+          'src/main.ts'
+        ],
+        thresholds: {
+          global: {
+            statements: 80,
+            branches: 80,
+            functions: 80,
+            lines: 80
+          }
+        }
+      },
+      setupFiles: ['./src/__tests__/setup.ts']
+    }
+  })
+)