Merge branch 'main' into dev

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

@@ -235,6 +235,10 @@ 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{
-			Account:            dto.AccountFromService(&accounts[i]),
-			CurrentConcurrency: concurrencyCounts[accounts[i].ID],
+		acc := &accounts[i]
+		item := AccountWithConcurrency{
+			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/handler/openai_gateway_handler.go

@@ -186,8 +186,8 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 		return
 	}
 
-	// Generate session hash (from header for OpenAI)
-	sessionHash := h.gatewayService.GenerateSessionHash(c)
+	// Generate session hash (header first; fallback to prompt_cache_key)
+	sessionHash := h.gatewayService.GenerateSessionHash(c, reqBody)
 
 	const maxAccountSwitches = 3
 	switchCount := 0

backend/internal/pkg/gemini/models.go

@@ -16,14 +16,11 @@ type ModelsListResponse struct {
 func DefaultModels() []Model {
 	methods := []string{"generateContent", "streamGenerateContent"}
 	return []Model{
-		{Name: "models/gemini-3-pro-preview", SupportedGenerationMethods: methods},
-		{Name: "models/gemini-3-flash-preview", SupportedGenerationMethods: methods},
-		{Name: "models/gemini-2.5-pro", SupportedGenerationMethods: methods},
-		{Name: "models/gemini-2.5-flash", SupportedGenerationMethods: methods},
 		{Name: "models/gemini-2.0-flash", SupportedGenerationMethods: methods},
-		{Name: "models/gemini-1.5-pro", SupportedGenerationMethods: methods},
-		{Name: "models/gemini-1.5-flash", SupportedGenerationMethods: methods},
-		{Name: "models/gemini-1.5-flash-8b", SupportedGenerationMethods: methods},
+		{Name: "models/gemini-2.5-flash", SupportedGenerationMethods: methods},
+		{Name: "models/gemini-2.5-pro", SupportedGenerationMethods: methods},
+		{Name: "models/gemini-3-flash-preview", SupportedGenerationMethods: methods},
+		{Name: "models/gemini-3-pro-preview", SupportedGenerationMethods: methods},
 	}
 }
 

backend/internal/pkg/geminicli/models.go

@@ -12,10 +12,10 @@ type Model struct {
 // DefaultModels is the curated Gemini model list used by the admin UI "test account" flow.
 var DefaultModels = []Model{
 	{ID: "gemini-2.0-flash", Type: "model", DisplayName: "Gemini 2.0 Flash", CreatedAt: ""},
-	{ID: "gemini-2.5-pro", Type: "model", DisplayName: "Gemini 2.5 Pro", CreatedAt: ""},
 	{ID: "gemini-2.5-flash", Type: "model", DisplayName: "Gemini 2.5 Flash", CreatedAt: ""},
-	{ID: "gemini-3-pro-preview", Type: "model", DisplayName: "Gemini 3 Pro Preview", CreatedAt: ""},
+	{ID: "gemini-2.5-pro", Type: "model", DisplayName: "Gemini 2.5 Pro", CreatedAt: ""},
 	{ID: "gemini-3-flash-preview", Type: "model", DisplayName: "Gemini 3 Flash Preview", CreatedAt: ""},
+	{ID: "gemini-3-pro-preview", Type: "model", DisplayName: "Gemini 3 Pro Preview", CreatedAt: ""},
 }
 
 // DefaultTestModel is the default model to preselect in test flows.

backend/internal/repository/ent.go

@@ -65,5 +65,18 @@ func InitEnt(cfg *config.Config) (*ent.Client, *sql.DB, error) {
 
 	// 创建 Ent 客户端，绑定到已配置的数据库驱动。
 	client := ent.NewClient(ent.Driver(drv))
+
+	// SIMPLE 模式：启动时补齐各平台默认分组。
+	// - anthropic/openai/gemini: 确保存在 <platform>-default
+	// - antigravity: 仅要求存在 >=2 个未软删除分组（用于 claude/gemini 混合调度场景）
+	if cfg.RunMode == config.RunModeSimple {
+		seedCtx, seedCancel := context.WithTimeout(context.Background(), 30*time.Second)
+		defer seedCancel()
+		if err := ensureSimpleModeDefaultGroups(seedCtx, client); err != nil {
+			_ = client.Close()
+			return nil, nil, err
+		}
+	}
+
 	return client, drv.DB(), nil
 }

backend/internal/repository/ops_repo.go

@@ -992,7 +992,8 @@ func buildOpsErrorLogsWhere(filter *service.OpsErrorLogFilter) (string, []any) {
 	}
 
 	// View filter: errors vs excluded vs all.
-	// Excluded = upstream 429/529 and business-limited (quota/concurrency/billing) errors.
+	// Excluded = business-limited errors (quota/concurrency/billing).
+	// Upstream 429/529 are included in errors view to match SLA calculation.
 	view := ""
 	if filter != nil {
 		view = strings.ToLower(strings.TrimSpace(filter.View))
@@ -1000,15 +1001,13 @@ func buildOpsErrorLogsWhere(filter *service.OpsErrorLogFilter) (string, []any) {
 	switch view {
 	case "", "errors":
 		clauses = append(clauses, "COALESCE(is_business_limited,false) = false")
-		clauses = append(clauses, "COALESCE(upstream_status_code, status_code, 0) NOT IN (429, 529)")
 	case "excluded":
-		clauses = append(clauses, "(COALESCE(is_business_limited,false) = true OR COALESCE(upstream_status_code, status_code, 0) IN (429, 529))")
+		clauses = append(clauses, "COALESCE(is_business_limited,false) = true")
 	case "all":
 		// no-op
 	default:
 		// treat unknown as default 'errors'
 		clauses = append(clauses, "COALESCE(is_business_limited,false) = false")
-		clauses = append(clauses, "COALESCE(upstream_status_code, status_code, 0) NOT IN (429, 529)")
 	}
 	if len(filter.StatusCodes) > 0 {
 		args = append(args, pq.Array(filter.StatusCodes))

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/simple_mode_default_groups.go

@@ -0,0 +1,82 @@
+package repository
+
+import (
+	"context"
+	"fmt"
+
+	dbent "github.com/Wei-Shaw/sub2api/ent"
+	"github.com/Wei-Shaw/sub2api/ent/group"
+	"github.com/Wei-Shaw/sub2api/internal/service"
+)
+
+func ensureSimpleModeDefaultGroups(ctx context.Context, client *dbent.Client) error {
+	if client == nil {
+		return fmt.Errorf("nil ent client")
+	}
+
+	requiredByPlatform := map[string]int{
+		service.PlatformAnthropic:   1,
+		service.PlatformOpenAI:      1,
+		service.PlatformGemini:      1,
+		service.PlatformAntigravity: 2,
+	}
+
+	for platform, minCount := range requiredByPlatform {
+		count, err := client.Group.Query().
+			Where(group.PlatformEQ(platform), group.DeletedAtIsNil()).
+			Count(ctx)
+		if err != nil {
+			return fmt.Errorf("count groups for platform %s: %w", platform, err)
+		}
+
+		if platform == service.PlatformAntigravity {
+			if count < minCount {
+				for i := count; i < minCount; i++ {
+					name := fmt.Sprintf("%s-default-%d", platform, i+1)
+					if err := createGroupIfNotExists(ctx, client, name, platform); err != nil {
+						return err
+					}
+				}
+			}
+			continue
+		}
+
+		// Non-antigravity platforms: ensure <platform>-default exists.
+		name := platform + "-default"
+		if err := createGroupIfNotExists(ctx, client, name, platform); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+func createGroupIfNotExists(ctx context.Context, client *dbent.Client, name, platform string) error {
+	exists, err := client.Group.Query().
+		Where(group.NameEQ(name), group.DeletedAtIsNil()).
+		Exist(ctx)
+	if err != nil {
+		return fmt.Errorf("check group exists %s: %w", name, err)
+	}
+	if exists {
+		return nil
+	}
+
+	_, err = client.Group.Create().
+		SetName(name).
+		SetDescription("Auto-created default group").
+		SetPlatform(platform).
+		SetStatus(service.StatusActive).
+		SetSubscriptionType(service.SubscriptionTypeStandard).
+		SetRateMultiplier(1.0).
+		SetIsExclusive(false).
+		Save(ctx)
+	if err != nil {
+		if dbent.IsConstraintError(err) {
+			// Concurrent server startups may race on creation; treat as success.
+			return nil
+		}
+		return fmt.Errorf("create default group %s: %w", name, err)
+	}
+	return nil
+}

backend/internal/repository/simple_mode_default_groups_integration_test.go

@@ -0,0 +1,84 @@
+//go:build integration
+
+package repository
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"github.com/Wei-Shaw/sub2api/ent/group"
+	"github.com/Wei-Shaw/sub2api/internal/service"
+	"github.com/stretchr/testify/require"
+)
+
+func TestEnsureSimpleModeDefaultGroups_CreatesMissingDefaults(t *testing.T) {
+	ctx := context.Background()
+	tx := testEntTx(t)
+	client := tx.Client()
+
+	seedCtx, cancel := context.WithTimeout(ctx, 10*time.Second)
+	defer cancel()
+
+	require.NoError(t, ensureSimpleModeDefaultGroups(seedCtx, client))
+
+	assertGroupExists := func(name string) {
+		exists, err := client.Group.Query().Where(group.NameEQ(name), group.DeletedAtIsNil()).Exist(seedCtx)
+		require.NoError(t, err)
+		require.True(t, exists, "expected group %s to exist", name)
+	}
+
+	assertGroupExists(service.PlatformAnthropic + "-default")
+	assertGroupExists(service.PlatformOpenAI + "-default")
+	assertGroupExists(service.PlatformGemini + "-default")
+	assertGroupExists(service.PlatformAntigravity + "-default-1")
+	assertGroupExists(service.PlatformAntigravity + "-default-2")
+}
+
+func TestEnsureSimpleModeDefaultGroups_IgnoresSoftDeletedGroups(t *testing.T) {
+	ctx := context.Background()
+	tx := testEntTx(t)
+	client := tx.Client()
+
+	seedCtx, cancel := context.WithTimeout(ctx, 10*time.Second)
+	defer cancel()
+
+	// Create and then soft-delete an anthropic default group.
+	g, err := client.Group.Create().
+		SetName(service.PlatformAnthropic + "-default").
+		SetPlatform(service.PlatformAnthropic).
+		SetStatus(service.StatusActive).
+		SetSubscriptionType(service.SubscriptionTypeStandard).
+		SetRateMultiplier(1.0).
+		SetIsExclusive(false).
+		Save(seedCtx)
+	require.NoError(t, err)
+
+	_, err = client.Group.Delete().Where(group.IDEQ(g.ID)).Exec(seedCtx)
+	require.NoError(t, err)
+
+	require.NoError(t, ensureSimpleModeDefaultGroups(seedCtx, client))
+
+	// New active one should exist.
+	count, err := client.Group.Query().Where(group.NameEQ(service.PlatformAnthropic+"-default"), group.DeletedAtIsNil()).Count(seedCtx)
+	require.NoError(t, err)
+	require.Equal(t, 1, count)
+}
+
+func TestEnsureSimpleModeDefaultGroups_AntigravityNeedsTwoGroupsOnlyByCount(t *testing.T) {
+	ctx := context.Background()
+	tx := testEntTx(t)
+	client := tx.Client()
+
+	seedCtx, cancel := context.WithTimeout(ctx, 10*time.Second)
+	defer cancel()
+
+	mustCreateGroup(t, client, &service.Group{Name: "ag-custom-1-" + time.Now().Format(time.RFC3339Nano), Platform: service.PlatformAntigravity})
+	mustCreateGroup(t, client, &service.Group{Name: "ag-custom-2-" + time.Now().Format(time.RFC3339Nano), Platform: service.PlatformAntigravity})
+
+	require.NoError(t, ensureSimpleModeDefaultGroups(seedCtx, client))
+
+	count, err := client.Group.Query().Where(group.PlatformEQ(service.PlatformAntigravity), group.DeletedAtIsNil()).Count(seedCtx)
+	require.NoError(t, err)
+	require.GreaterOrEqual(t, count, 2)
+}

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,
@@ -62,6 +72,7 @@ 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

@@ -557,3 +557,141 @@ 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/antigravity_model_mapping_test.go

@@ -30,7 +30,7 @@ func TestIsAntigravityModelSupported(t *testing.T) {
 		{"可映射 - claude-3-haiku-20240307", "claude-3-haiku-20240307", true},
 
 		// Gemini 前缀透传
-		{"Gemini前缀 - gemini-1.5-pro", "gemini-1.5-pro", true},
+		{"Gemini前缀 - gemini-2.5-pro", "gemini-2.5-pro", true},
 		{"Gemini前缀 - gemini-unknown-model", "gemini-unknown-model", true},
 		{"Gemini前缀 - gemini-future-version", "gemini-future-version", true},
 
@@ -142,10 +142,10 @@ func TestAntigravityGatewayService_GetMappedModel(t *testing.T) {
 			expected:       "gemini-2.5-flash",
 		},
 		{
-			name:           "Gemini透传 - gemini-1.5-pro",
-			requestedModel: "gemini-1.5-pro",
+			name:           "Gemini透传 - gemini-2.5-pro",
+			requestedModel: "gemini-2.5-pro",
 			accountMapping: nil,
-			expected:       "gemini-1.5-pro",
+			expected:       "gemini-2.5-pro",
 		},
 		{
 			name:           "Gemini透传 - gemini-future-model",

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

@@ -176,6 +176,7 @@ 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
@@ -196,6 +197,7 @@ func NewGatewayService(
 	httpUpstream HTTPUpstream,
 	deferredService *DeferredService,
 	claudeTokenProvider *ClaudeTokenProvider,
+	sessionLimitCache SessionLimitCache,
 ) *GatewayService {
 	return &GatewayService{
 		accountRepo:         accountRepo,
@@ -214,6 +216,7 @@ func NewGatewayService(
 		httpUpstream:        httpUpstream,
 		deferredService:     deferredService,
 		claudeTokenProvider: claudeTokenProvider,
+		sessionLimitCache:   sessionLimitCache,
 	}
 }
 
@@ -407,8 +410,12 @@ 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 {
@@ -527,7 +534,7 @@ 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++
@@ -554,13 +561,18 @@ 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 {
@@ -573,9 +585,15 @@ 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 {
+								// 会话数量限制检查
+								if !s.checkAndRegisterSession(ctx, stickyAccount, sessionUUID) {
+									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)
@@ -586,6 +604,7 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 										ReleaseFunc: result.ReleaseFunc,
 									}, nil
 								}
+							}
 
 							waitingCount, _ := s.concurrencyService.GetAccountWaitingCount(ctx, stickyAccountID)
 							if waitingCount < cfg.StickySessionMaxWaiting {
@@ -657,6 +676,11 @@ 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)
 						}
@@ -699,9 +723,14 @@ 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 {
+					// 会话数量限制检查
+					if !s.checkAndRegisterSession(ctx, account, sessionUUID) {
+						result.ReleaseFunc() // 释放槽位，继续到 Layer 2
+					} else {
 						_ = s.cache.RefreshSessionTTL(ctx, derefGroupID(groupID), sessionHash, stickySessionTTL)
 						return &AccountSelectionResult{
 							Account:     account,
@@ -709,6 +738,7 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 							ReleaseFunc: result.ReleaseFunc,
 						}, nil
 					}
+				}
 
 				waitingCount, _ := s.concurrencyService.GetAccountWaitingCount(ctx, accountID)
 				if waitingCount < cfg.StickySessionMaxWaiting {
@@ -748,6 +778,10 @@ 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)
 	}
 
@@ -765,7 +799,7 @@ 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 {
@@ -814,6 +848,11 @@ 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)
 					}
@@ -843,13 +882,18 @@ 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)
 			}
@@ -1081,6 +1125,107 @@ 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)

backend/internal/service/gemini_multiplatform_test.go

@@ -599,7 +599,7 @@ func TestGeminiMessagesCompatService_isModelSupportedByAccount(t *testing.T) {
 			name: "Gemini平台-有映射配置-只支持配置的模型",
 			account: &Account{
 				Platform:    PlatformGemini,
-				Credentials: map[string]any{"model_mapping": map[string]any{"gemini-1.5-pro": "x"}},
+				Credentials: map[string]any{"model_mapping": map[string]any{"gemini-2.5-pro": "x"}},
 			},
 			model:    "gemini-2.5-flash",
 			expected: false,

backend/internal/service/openai_codex_transform.go

@@ -394,19 +394,35 @@ func normalizeCodexTools(reqBody map[string]any) bool {
 	}
 
 	modified := false
-	for idx, tool := range tools {
+	validTools := make([]any, 0, len(tools))
+
+	for _, tool := range tools {
 		toolMap, ok := tool.(map[string]any)
 		if !ok {
+			// Keep unknown structure as-is to avoid breaking upstream behavior.
+			validTools = append(validTools, tool)
 			continue
 		}
 
 		toolType, _ := toolMap["type"].(string)
-		if strings.TrimSpace(toolType) != "function" {
+		toolType = strings.TrimSpace(toolType)
+		if toolType != "function" {
+			validTools = append(validTools, toolMap)
 			continue
 		}
 
-		function, ok := toolMap["function"].(map[string]any)
-		if !ok {
+		// OpenAI Responses-style tools use top-level name/parameters.
+		if name, ok := toolMap["name"].(string); ok && strings.TrimSpace(name) != "" {
+			validTools = append(validTools, toolMap)
+			continue
+		}
+
+		// ChatCompletions-style tools use {type:"function", function:{...}}.
+		functionValue, hasFunction := toolMap["function"]
+		function, ok := functionValue.(map[string]any)
+		if !hasFunction || functionValue == nil || !ok || function == nil {
+			// Drop invalid function tools.
+			modified = true
 			continue
 		}
 
@@ -435,11 +451,11 @@ func normalizeCodexTools(reqBody map[string]any) bool {
 			}
 		}
 
-		tools[idx] = toolMap
+		validTools = append(validTools, toolMap)
 	}
 
 	if modified {
-		reqBody["tools"] = tools
+		reqBody["tools"] = validTools
 	}
 
 	return modified

backend/internal/service/openai_codex_transform_test.go

@@ -129,6 +129,37 @@ func TestFilterCodexInput_RemovesItemReferenceWhenNotPreserved(t *testing.T) {
 	require.False(t, hasID)
 }
 
+func TestApplyCodexOAuthTransform_NormalizeCodexTools_PreservesResponsesFunctionTools(t *testing.T) {
+	setupCodexCache(t)
+
+	reqBody := map[string]any{
+		"model": "gpt-5.1",
+		"tools": []any{
+			map[string]any{
+				"type":        "function",
+				"name":        "bash",
+				"description": "desc",
+				"parameters":  map[string]any{"type": "object"},
+			},
+			map[string]any{
+				"type":     "function",
+				"function": nil,
+			},
+		},
+	}
+
+	applyCodexOAuthTransform(reqBody)
+
+	tools, ok := reqBody["tools"].([]any)
+	require.True(t, ok)
+	require.Len(t, tools, 1)
+
+	first, ok := tools[0].(map[string]any)
+	require.True(t, ok)
+	require.Equal(t, "function", first["type"])
+	require.Equal(t, "bash", first["name"])
+}
+
 func TestApplyCodexOAuthTransform_EmptyInput(t *testing.T) {
 	// 空 input 应保持为空且不触发异常。
 	setupCodexCache(t)

backend/internal/service/openai_gateway_service.go

@@ -133,12 +133,30 @@ func NewOpenAIGatewayService(
 	}
 }
 
-// GenerateSessionHash generates session hash from header (OpenAI uses session_id header)
-func (s *OpenAIGatewayService) GenerateSessionHash(c *gin.Context) string {
-	sessionID := c.GetHeader("session_id")
+// GenerateSessionHash generates a sticky-session hash for OpenAI requests.
+//
+// Priority:
+//  1. Header: session_id
+//  2. Header: conversation_id
+//  3. Body:   prompt_cache_key (opencode)
+func (s *OpenAIGatewayService) GenerateSessionHash(c *gin.Context, reqBody map[string]any) string {
+	if c == nil {
+		return ""
+	}
+
+	sessionID := strings.TrimSpace(c.GetHeader("session_id"))
+	if sessionID == "" {
+		sessionID = strings.TrimSpace(c.GetHeader("conversation_id"))
+	}
+	if sessionID == "" && reqBody != nil {
+		if v, ok := reqBody["prompt_cache_key"].(string); ok {
+			sessionID = strings.TrimSpace(v)
+		}
+	}
 	if sessionID == "" {
 		return ""
 	}
+
 	hash := sha256.Sum256([]byte(sessionID))
 	return hex.EncodeToString(hash[:])
 }

backend/internal/service/openai_gateway_service_test.go

@@ -49,6 +49,49 @@ func (c stubConcurrencyCache) GetAccountsLoadBatch(ctx context.Context, accounts
 	return out, nil
 }
 
+func TestOpenAIGatewayService_GenerateSessionHash_Priority(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	c.Request = httptest.NewRequest(http.MethodPost, "/openai/v1/responses", nil)
+
+	svc := &OpenAIGatewayService{}
+
+	// 1) session_id header wins
+	c.Request.Header.Set("session_id", "sess-123")
+	c.Request.Header.Set("conversation_id", "conv-456")
+	h1 := svc.GenerateSessionHash(c, map[string]any{"prompt_cache_key": "ses_aaa"})
+	if h1 == "" {
+		t.Fatalf("expected non-empty hash")
+	}
+
+	// 2) conversation_id used when session_id absent
+	c.Request.Header.Del("session_id")
+	h2 := svc.GenerateSessionHash(c, map[string]any{"prompt_cache_key": "ses_aaa"})
+	if h2 == "" {
+		t.Fatalf("expected non-empty hash")
+	}
+	if h1 == h2 {
+		t.Fatalf("expected different hashes for different keys")
+	}
+
+	// 3) prompt_cache_key used when both headers absent
+	c.Request.Header.Del("conversation_id")
+	h3 := svc.GenerateSessionHash(c, map[string]any{"prompt_cache_key": "ses_aaa"})
+	if h3 == "" {
+		t.Fatalf("expected non-empty hash")
+	}
+	if h2 == h3 {
+		t.Fatalf("expected different hashes for different keys")
+	}
+
+	// 4) empty when no signals
+	h4 := svc.GenerateSessionHash(c, map[string]any{})
+	if h4 != "" {
+		t.Fatalf("expected empty hash when no signals")
+	}
+}
+
 func TestOpenAISelectAccountWithLoadAwareness_FiltersUnschedulable(t *testing.T) {
 	now := time.Now()
 	resetAt := now.Add(10 * time.Minute)

backend/internal/service/openai_tool_corrector.go

@@ -27,6 +27,11 @@ var codexToolNameMapping = map[string]string{
 	"executeBash":  "bash",
 	"exec_bash":    "bash",
 	"execBash":     "bash",
+
+	// Some clients output generic fetch names.
+	"fetch":     "webfetch",
+	"web_fetch": "webfetch",
+	"webFetch":  "webfetch",
 }
 
 // ToolCorrectionStats 记录工具修正的统计信息（导出用于 JSON 序列化）
@@ -208,27 +213,67 @@ func (c *CodexToolCorrector) correctToolParameters(toolName string, functionCall
 	// 根据工具名称应用特定的参数修正规则
 	switch toolName {
 	case "bash":
-		// 移除 workdir 参数（OpenCode 不支持）
-		if _, exists := argsMap["workdir"]; exists {
-			delete(argsMap, "workdir")
+		// OpenCode bash 支持 workdir；有些来源会输出 work_dir。
+		if _, hasWorkdir := argsMap["workdir"]; !hasWorkdir {
+			if workDir, exists := argsMap["work_dir"]; exists {
+				argsMap["workdir"] = workDir
+				delete(argsMap, "work_dir")
 				corrected = true
-			log.Printf("[CodexToolCorrector] Removed 'workdir' parameter from bash tool")
+				log.Printf("[CodexToolCorrector] Renamed 'work_dir' to 'workdir' in bash tool")
 			}
+		} else {
 			if _, exists := argsMap["work_dir"]; exists {
 				delete(argsMap, "work_dir")
 				corrected = true
-			log.Printf("[CodexToolCorrector] Removed 'work_dir' parameter from bash tool")
+				log.Printf("[CodexToolCorrector] Removed duplicate 'work_dir' parameter from bash tool")
+			}
 		}
 
 	case "edit":
-		// OpenCode edit 使用 old_string/new_string，Codex 可能使用其他名称
-		// 这里可以添加参数名称的映射逻辑
-		if _, exists := argsMap["file_path"]; !exists {
-			if path, exists := argsMap["path"]; exists {
-				argsMap["file_path"] = path
+		// OpenCode edit 参数为 filePath/oldString/newString（camelCase）。
+		if _, exists := argsMap["filePath"]; !exists {
+			if filePath, exists := argsMap["file_path"]; exists {
+				argsMap["filePath"] = filePath
+				delete(argsMap, "file_path")
+				corrected = true
+				log.Printf("[CodexToolCorrector] Renamed 'file_path' to 'filePath' in edit tool")
+			} else if filePath, exists := argsMap["path"]; exists {
+				argsMap["filePath"] = filePath
 				delete(argsMap, "path")
 				corrected = true
-				log.Printf("[CodexToolCorrector] Renamed 'path' to 'file_path' in edit tool")
+				log.Printf("[CodexToolCorrector] Renamed 'path' to 'filePath' in edit tool")
+			} else if filePath, exists := argsMap["file"]; exists {
+				argsMap["filePath"] = filePath
+				delete(argsMap, "file")
+				corrected = true
+				log.Printf("[CodexToolCorrector] Renamed 'file' to 'filePath' in edit tool")
+			}
+		}
+
+		if _, exists := argsMap["oldString"]; !exists {
+			if oldString, exists := argsMap["old_string"]; exists {
+				argsMap["oldString"] = oldString
+				delete(argsMap, "old_string")
+				corrected = true
+				log.Printf("[CodexToolCorrector] Renamed 'old_string' to 'oldString' in edit tool")
+			}
+		}
+
+		if _, exists := argsMap["newString"]; !exists {
+			if newString, exists := argsMap["new_string"]; exists {
+				argsMap["newString"] = newString
+				delete(argsMap, "new_string")
+				corrected = true
+				log.Printf("[CodexToolCorrector] Renamed 'new_string' to 'newString' in edit tool")
+			}
+		}
+
+		if _, exists := argsMap["replaceAll"]; !exists {
+			if replaceAll, exists := argsMap["replace_all"]; exists {
+				argsMap["replaceAll"] = replaceAll
+				delete(argsMap, "replace_all")
+				corrected = true
+				log.Printf("[CodexToolCorrector] Renamed 'replace_all' to 'replaceAll' in edit tool")
 			}
 		}
 	}

backend/internal/service/openai_tool_corrector_test.go

@@ -416,22 +416,23 @@ func TestCorrectToolParameters(t *testing.T) {
 		expected map[string]bool // key: 期待存在的参数, value: true表示应该存在
 	}{
 		{
-			name: "remove workdir from bash tool",
+			name: "rename work_dir to workdir in bash tool",
 			input: `{
 				"tool_calls": [{
 					"function": {
 						"name": "bash",
-						"arguments": "{\"command\":\"ls\",\"workdir\":\"/tmp\"}"
+						"arguments": "{\"command\":\"ls\",\"work_dir\":\"/tmp\"}"
 					}
 				}]
 			}`,
 			expected: map[string]bool{
 				"command":  true,
-				"workdir": false,
+				"workdir":  true,
+				"work_dir": false,
 			},
 		},
 		{
-			name: "rename path to file_path in edit tool",
+			name: "rename snake_case edit params to camelCase",
 			input: `{
 				"tool_calls": [{
 					"function": {
@@ -441,10 +442,12 @@ func TestCorrectToolParameters(t *testing.T) {
 				}]
 			}`,
 			expected: map[string]bool{
-				"file_path":  true,
+				"filePath":   true,
 				"path":       false,
-				"old_string": true,
-				"new_string": true,
+				"oldString":  true,
+				"old_string": false,
+				"newString":  true,
+				"new_string": false,
 			},
 		},
 	}

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/pricing_service.go

@@ -531,8 +531,8 @@ func (s *PricingService) buildModelLookupCandidates(modelLower string) []string
 func normalizeModelNameForPricing(model string) string {
 	// Common Gemini/VertexAI forms:
 	// - models/gemini-2.0-flash-exp
-	// - publishers/google/models/gemini-1.5-pro
-	// - projects/.../locations/.../publishers/google/models/gemini-1.5-pro
+	// - publishers/google/models/gemini-2.5-pro
+	// - projects/.../locations/.../publishers/google/models/gemini-2.5-pro
 	model = strings.TrimSpace(model)
 	model = strings.TrimLeft(model, "/")
 	model = strings.TrimPrefix(model, "models/")

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)
+}

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/AccountTestModal.vue

@@ -292,8 +292,11 @@ const loadAvailableModels = async () => {
     if (availableModels.value.length > 0) {
       if (props.account.platform === 'gemini') {
         const preferred =
+          availableModels.value.find((m) => m.id === 'gemini-2.0-flash') ||
+          availableModels.value.find((m) => m.id === 'gemini-2.5-flash') ||
           availableModels.value.find((m) => m.id === 'gemini-2.5-pro') ||
-          availableModels.value.find((m) => m.id === 'gemini-3-pro')
+          availableModels.value.find((m) => m.id === 'gemini-3-flash-preview') ||
+          availableModels.value.find((m) => m.id === 'gemini-3-pro-preview')
         selectedModelId.value = preferred?.id || availableModels.value[0].id
       } else {
         // Try to select Sonnet as default, otherwise use first model

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/admin/account/AccountTableActions.vue

@@ -1,5 +1,6 @@
 <template>
   <div class="flex flex-wrap items-center gap-3">
+    <slot name="before"></slot>
     <button @click="$emit('refresh')" :disabled="loading" class="btn btn-secondary">
       <Icon name="refresh" size="md" :class="[loading ? 'animate-spin' : '']" />
     </button>

frontend/src/components/admin/account/AccountTestModal.vue

@@ -232,8 +232,11 @@ const loadAvailableModels = async () => {
     if (availableModels.value.length > 0) {
       if (props.account.platform === 'gemini') {
         const preferred =
+          availableModels.value.find((m) => m.id === 'gemini-2.0-flash') ||
+          availableModels.value.find((m) => m.id === 'gemini-2.5-flash') ||
           availableModels.value.find((m) => m.id === 'gemini-2.5-pro') ||
-          availableModels.value.find((m) => m.id === 'gemini-3-pro')
+          availableModels.value.find((m) => m.id === 'gemini-3-flash-preview') ||
+          availableModels.value.find((m) => m.id === 'gemini-3-pro-preview')
         selectedModelId.value = preferred?.id || availableModels.value[0].id
       } else {
         // Try to select Sonnet as default, otherwise use first model

frontend/src/components/keys/UseKeyModal.vue

@@ -443,7 +443,7 @@ $env:ANTHROPIC_AUTH_TOKEN="${apiKey}"`
 }
 
 function generateGeminiCliContent(baseUrl: string, apiKey: string): FileConfig {
-  const model = 'gemini-2.5-pro'
+  const model = 'gemini-2.0-flash'
   const modelComment = t('keys.useKeyModal.gemini.modelComment')
   let path: string
   let content: string
@@ -548,14 +548,22 @@ function generateOpenCodeConfig(platform: string, baseUrl: string, apiKey: strin
     }
   }
   const geminiModels = {
-    'gemini-3-pro-high': { name: 'Gemini 3 Pro High' },
-    'gemini-3-pro-low': { name: 'Gemini 3 Pro Low' },
-    'gemini-3-pro-preview': { name: 'Gemini 3 Pro Preview' },
-    'gemini-3-pro-image': { name: 'Gemini 3 Pro Image' },
-    'gemini-3-flash': { name: 'Gemini 3 Flash' },
-    'gemini-2.5-flash-thinking': { name: 'Gemini 2.5 Flash Thinking' },
+    'gemini-2.0-flash': { name: 'Gemini 2.0 Flash' },
     'gemini-2.5-flash': { name: 'Gemini 2.5 Flash' },
-    'gemini-2.5-flash-lite': { name: 'Gemini 2.5 Flash Lite' }
+    'gemini-2.5-pro': { name: 'Gemini 2.5 Pro' },
+    'gemini-3-flash-preview': { name: 'Gemini 3 Flash Preview' },
+    'gemini-3-pro-preview': { name: 'Gemini 3 Pro Preview' }
+  }
+
+  const antigravityGeminiModels = {
+    'gemini-2.5-flash': { name: 'Gemini 2.5 Flash' },
+    'gemini-2.5-flash-lite': { name: 'Gemini 2.5 Flash Lite' },
+    'gemini-2.5-flash-thinking': { name: 'Gemini 2.5 Flash Thinking' },
+    'gemini-3-flash': { name: 'Gemini 3 Flash' },
+    'gemini-3-pro-low': { name: 'Gemini 3 Pro Low' },
+    'gemini-3-pro-high': { name: 'Gemini 3 Pro High' },
+    'gemini-3-pro-preview': { name: 'Gemini 3 Pro Preview' },
+    'gemini-3-pro-image': { name: 'Gemini 3 Pro Image' }
   }
   const claudeModels = {
     'claude-opus-4-5-thinking': { name: 'Claude Opus 4.5 Thinking' },
@@ -575,7 +583,7 @@ function generateOpenCodeConfig(platform: string, baseUrl: string, apiKey: strin
   } else if (platform === 'antigravity-gemini') {
     provider[platform].npm = '@ai-sdk/google'
     provider[platform].name = 'Antigravity (Gemini)'
-    provider[platform].models = geminiModels
+    provider[platform].models = antigravityGeminiModels
   } else if (platform === 'openai') {
     provider[platform].models = openaiModels
   }

frontend/src/composables/useModelWhitelist.ts

@@ -43,13 +43,13 @@ export const claudeModels = [
 
 // Google Gemini
 const geminiModels = [
-  'gemini-2.0-flash', 'gemini-2.0-flash-lite-preview', 'gemini-2.0-flash-exp',
-  'gemini-2.0-pro-exp', 'gemini-2.0-flash-thinking-exp',
-  'gemini-2.5-pro-exp-03-25', 'gemini-2.5-pro-preview-03-25',
-  'gemini-3-pro-preview',
-  'gemini-1.5-pro', 'gemini-1.5-pro-latest',
-  'gemini-1.5-flash', 'gemini-1.5-flash-latest', 'gemini-1.5-flash-8b',
-  'gemini-exp-1206'
+  // Keep in sync with backend curated Gemini lists.
+  // This list is intentionally conservative (models commonly available across OAuth/API key).
+  'gemini-2.0-flash',
+  'gemini-2.5-flash',
+  'gemini-2.5-pro',
+  'gemini-3-flash-preview',
+  'gemini-3-pro-preview'
 ]
 
 // 智谱 GLM
@@ -229,9 +229,8 @@ const openaiPresetMappings = [
 
 const geminiPresetMappings = [
   { label: 'Flash 2.0', from: 'gemini-2.0-flash', to: 'gemini-2.0-flash', color: 'bg-blue-100 text-blue-700 hover:bg-blue-200 dark:bg-blue-900/30 dark:text-blue-400' },
-  { label: 'Flash Lite', from: 'gemini-2.0-flash-lite-preview', to: 'gemini-2.0-flash-lite-preview', color: 'bg-indigo-100 text-indigo-700 hover:bg-indigo-200 dark:bg-indigo-900/30 dark:text-indigo-400' },
-  { label: '1.5 Pro', from: 'gemini-1.5-pro', to: 'gemini-1.5-pro', color: 'bg-purple-100 text-purple-700 hover:bg-purple-200 dark:bg-purple-900/30 dark:text-purple-400' },
-  { label: '1.5 Flash', from: 'gemini-1.5-flash', to: 'gemini-1.5-flash', color: 'bg-emerald-100 text-emerald-700 hover:bg-emerald-200 dark:bg-emerald-900/30 dark:text-emerald-400' }
+  { label: '2.5 Flash', from: 'gemini-2.5-flash', to: 'gemini-2.5-flash', color: 'bg-indigo-100 text-indigo-700 hover:bg-indigo-200 dark:bg-indigo-900/30 dark:text-indigo-400' },
+  { label: '2.5 Pro', from: 'gemini-2.5-pro', to: 'gemini-2.5-pro', color: 'bg-purple-100 text-purple-700 hover:bg-purple-200 dark:bg-purple-900/30 dark:text-purple-400' }
 ]
 
 // =====================

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',
@@ -675,6 +676,7 @@ export default {
       updating: 'Updating...',
       columns: {
         user: 'User',
+        email: 'Email',
         username: 'Username',
         notes: 'Notes',
         role: 'Role',
@@ -1085,7 +1087,7 @@ export default {
         platformType: 'Platform/Type',
         platform: 'Platform',
         type: 'Type',
-        concurrencyStatus: 'Concurrency',
+        capacity: 'Capacity',
         notes: 'Notes',
         priority: 'Priority',
         billingRateMultiplier: 'Billing Rate',
@@ -1095,10 +1097,23 @@ export default {
         todayStats: 'Today Stats',
         groups: 'Groups',
         usageWindows: 'Usage Windows',
+        proxy: 'Proxy',
         lastUsed: 'Last Used',
         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',
@@ -1250,6 +1265,31 @@ 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',
@@ -1134,7 +1135,7 @@ 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',
@@ -1144,10 +1145,23 @@ export default {
         todayStats: '今日统计',
         groups: '分组',
         usageWindows: '用量窗口',
+        proxy: '代理',
         lastUsed: '最近使用',
         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',
@@ -1383,6 +1397,31 @@ 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/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

@@ -15,7 +15,40 @@
             @refresh="load"
             @sync="showSync = true"
             @create="showCreate = true"
-          />
+          >
+            <template #before>
+              <!-- Column Settings Dropdown -->
+              <div class="relative" ref="columnDropdownRef">
+                <button
+                  @click="showColumnDropdown = !showColumnDropdown"
+                  class="btn btn-secondary px-2 md:px-3"
+                  :title="t('admin.users.columnSettings')"
+                >
+                  <svg class="h-4 w-4 md:mr-1.5" fill="none" stroke="currentColor" viewBox="0 0 24 24" stroke-width="1.5">
+                    <path stroke-linecap="round" stroke-linejoin="round" d="M9 4.5v15m6-15v15m-10.875 0h15.75c.621 0 1.125-.504 1.125-1.125V5.625c0-.621-.504-1.125-1.125-1.125H4.125C3.504 4.5 3 5.004 3 5.625v12.75c0 .621.504 1.125 1.125 1.125z" />
+                  </svg>
+                  <span class="hidden md:inline">{{ t('admin.users.columnSettings') }}</span>
+                </button>
+                <!-- Dropdown menu -->
+                <div
+                  v-if="showColumnDropdown"
+                  class="absolute right-0 z-50 mt-2 w-48 origin-top-right rounded-lg border border-gray-200 bg-white shadow-lg dark:border-gray-700 dark:bg-gray-800"
+                >
+                  <div class="max-h-80 overflow-y-auto p-2">
+                    <button
+                      v-for="col in toggleableColumns"
+                      :key="col.key"
+                      @click="toggleColumn(col.key)"
+                      class="flex w-full items-center justify-between rounded-md px-3 py-2 text-sm text-gray-700 hover:bg-gray-100 dark:text-gray-200 dark:hover:bg-gray-700"
+                    >
+                      <span>{{ col.label }}</span>
+                      <Icon v-if="isColumnVisible(col.key)" name="check" size="sm" class="text-primary-500" />
+                    </button>
+                  </div>
+                </div>
+              </div>
+            </template>
+          </AccountTableActions>
         </div>
       </template>
       <template #table>
@@ -34,15 +67,8 @@
           <template #cell-platform_type="{ row }">
             <PlatformTypeBadge :platform="row.platform" :type="row.type" />
           </template>
-          <template #cell-concurrency="{ row }">
-            <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', (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 #cell-capacity="{ row }">
+            <AccountCapacityCell :account="row" />
           </template>
           <template #cell-status="{ row }">
             <AccountStatusIndicator :account="row" @show-temp-unsched="handleShowTempUnsched" />
@@ -61,6 +87,15 @@
           <template #cell-usage="{ row }">
             <AccountUsageCell :account="row" />
           </template>
+          <template #cell-proxy="{ row }">
+            <div v-if="row.proxy" class="flex items-center gap-2">
+              <span class="text-sm text-gray-700 dark:text-gray-300">{{ row.proxy.name }}</span>
+              <span v-if="row.proxy.country_code" class="text-xs text-gray-500 dark:text-gray-400">
+                ({{ row.proxy.country_code }})
+              </span>
+            </div>
+            <span v-else class="text-sm text-gray-400 dark:text-dark-500">-</span>
+          </template>
           <template #cell-rate_multiplier="{ row }">
             <span class="text-sm font-mono text-gray-700 dark:text-gray-300">
               {{ (row.rate_multiplier ?? 1).toFixed(2) }}x
@@ -148,7 +183,9 @@ 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 Icon from '@/components/icons/Icon.vue'
 import { formatDateTime, formatRelativeTime } from '@/utils/format'
 import type { Account, Proxy, Group } from '@/types'
 
@@ -177,17 +214,59 @@ const statsAcc = ref<Account | null>(null)
 const togglingSchedulable = ref<number | null>(null)
 const menu = reactive<{show:boolean, acc:Account|null, pos:{top:number, left:number}|null}>({ show: false, acc: null, pos: null })
 
+// Column settings
+const showColumnDropdown = ref(false)
+const columnDropdownRef = ref<HTMLElement | null>(null)
+const hiddenColumns = reactive<Set<string>>(new Set())
+const DEFAULT_HIDDEN_COLUMNS = ['proxy', 'notes', 'priority', 'rate_multiplier']
+const HIDDEN_COLUMNS_KEY = 'account-hidden-columns'
+
+const loadSavedColumns = () => {
+  try {
+    const saved = localStorage.getItem(HIDDEN_COLUMNS_KEY)
+    if (saved) {
+      const parsed = JSON.parse(saved) as string[]
+      parsed.forEach(key => hiddenColumns.add(key))
+    } else {
+      DEFAULT_HIDDEN_COLUMNS.forEach(key => hiddenColumns.add(key))
+    }
+  } catch (e) {
+    console.error('Failed to load saved columns:', e)
+    DEFAULT_HIDDEN_COLUMNS.forEach(key => hiddenColumns.add(key))
+  }
+}
+
+const saveColumnsToStorage = () => {
+  try {
+    localStorage.setItem(HIDDEN_COLUMNS_KEY, JSON.stringify([...hiddenColumns]))
+  } catch (e) {
+    console.error('Failed to save columns:', e)
+  }
+}
+
+const toggleColumn = (key: string) => {
+  if (hiddenColumns.has(key)) {
+    hiddenColumns.delete(key)
+  } else {
+    hiddenColumns.add(key)
+  }
+  saveColumnsToStorage()
+}
+
+const isColumnVisible = (key: string) => !hiddenColumns.has(key)
+
 const { items: accounts, loading, params, pagination, load, reload, debouncedReload, handlePageChange, handlePageSizeChange } = useTableLoader<Account, any>({
   fetchFn: adminAPI.accounts.list,
   initialParams: { platform: '', type: '', status: '', search: '' }
 })
 
-const cols = computed(() => {
+// All available columns
+const allColumns = computed(() => {
   const c = [
     { 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 }
@@ -197,6 +276,7 @@ const cols = computed(() => {
   }
   c.push(
     { key: 'usage', label: t('admin.accounts.columns.usageWindows'), sortable: false },
+    { key: 'proxy', label: t('admin.accounts.columns.proxy'), sortable: false },
     { key: 'priority', label: t('admin.accounts.columns.priority'), sortable: true },
     { key: 'rate_multiplier', label: t('admin.accounts.columns.billingRateMultiplier'), sortable: true },
     { key: 'last_used_at', label: t('admin.accounts.columns.lastUsed'), sortable: true },
@@ -207,6 +287,18 @@ const cols = computed(() => {
   return c
 })
 
+// Columns that can be toggled (exclude select, name, and actions)
+const toggleableColumns = computed(() =>
+  allColumns.value.filter(col => col.key !== 'select' && col.key !== 'name' && col.key !== 'actions')
+)
+
+// Filtered columns based on visibility
+const cols = computed(() =>
+  allColumns.value.filter(col =>
+    col.key === 'select' || col.key === 'name' || col.key === 'actions' || !hiddenColumns.has(col.key)
+  )
+)
+
 const handleEdit = (a: Account) => { edAcc.value = a; showEdit.value = true }
 const openMenu = (a: Account, e: MouseEvent) => {
   menu.acc = a
@@ -409,12 +501,21 @@ const isExpired = (value: number | null) => {
   return value * 1000 <= Date.now()
 }
 
-// 滚动时关闭菜单
+// 滚动时关闭操作菜单（不关闭列设置下拉菜单）
 const handleScroll = () => {
   menu.show = false
 }
 
+// 点击外部关闭列设置下拉菜单
+const handleClickOutside = (event: MouseEvent) => {
+  const target = event.target as HTMLElement
+  if (columnDropdownRef.value && !columnDropdownRef.value.contains(target)) {
+    showColumnDropdown.value = false
+  }
+}
+
 onMounted(async () => {
+  loadSavedColumns()
   load()
   try {
     const [p, g] = await Promise.all([adminAPI.proxies.getAll(), adminAPI.groups.getAll()])
@@ -424,9 +525,11 @@ onMounted(async () => {
     console.error('Failed to load proxies/groups:', error)
   }
   window.addEventListener('scroll', handleScroll, true)
+  document.addEventListener('click', handleClickOutside)
 })
 
 onUnmounted(() => {
   window.removeEventListener('scroll', handleScroll, true)
+  document.removeEventListener('click', handleClickOutside)
 })
 </script>

frontend/src/views/admin/SubscriptionsView.vue

@@ -85,6 +85,57 @@
 
           <!-- Right: Actions -->
           <div class="ml-auto flex flex-wrap items-center justify-end gap-3">
+            <!-- Column Settings Dropdown -->
+            <div class="relative" ref="columnDropdownRef">
+              <button
+                @click="showColumnDropdown = !showColumnDropdown"
+                class="btn btn-secondary px-2 md:px-3"
+                :title="t('admin.users.columnSettings')"
+              >
+                <svg class="h-4 w-4 md:mr-1.5" fill="none" stroke="currentColor" viewBox="0 0 24 24" stroke-width="1.5">
+                  <path stroke-linecap="round" stroke-linejoin="round" d="M9 4.5v15m6-15v15m-10.875 0h15.75c.621 0 1.125-.504 1.125-1.125V5.625c0-.621-.504-1.125-1.125-1.125H4.125C3.504 4.5 3 5.004 3 5.625v12.75c0 .621.504 1.125 1.125 1.125z" />
+                </svg>
+                <span class="hidden md:inline">{{ t('admin.users.columnSettings') }}</span>
+              </button>
+              <!-- Dropdown menu -->
+              <div
+                v-if="showColumnDropdown"
+                class="absolute right-0 z-50 mt-2 w-48 origin-top-right rounded-lg border border-gray-200 bg-white shadow-lg dark:border-gray-700 dark:bg-gray-800"
+              >
+                <div class="p-2">
+                  <!-- User column mode selection -->
+                  <div class="mb-2 border-b border-gray-200 pb-2 dark:border-gray-700">
+                    <div class="px-3 py-1 text-xs font-medium text-gray-500 dark:text-gray-400">
+                      {{ t('admin.subscriptions.columns.user') }}
+                    </div>
+                    <button
+                      @click="setUserColumnMode('email')"
+                      class="flex w-full items-center justify-between rounded-md px-3 py-2 text-sm text-gray-700 hover:bg-gray-100 dark:text-gray-200 dark:hover:bg-gray-700"
+                    >
+                      <span>{{ t('admin.users.columns.email') }}</span>
+                      <Icon v-if="userColumnMode === 'email'" name="check" size="sm" class="text-primary-500" />
+                    </button>
+                    <button
+                      @click="setUserColumnMode('username')"
+                      class="flex w-full items-center justify-between rounded-md px-3 py-2 text-sm text-gray-700 hover:bg-gray-100 dark:text-gray-200 dark:hover:bg-gray-700"
+                    >
+                      <span>{{ t('admin.users.columns.username') }}</span>
+                      <Icon v-if="userColumnMode === 'username'" name="check" size="sm" class="text-primary-500" />
+                    </button>
+                  </div>
+                  <!-- Other columns toggle -->
+                  <button
+                    v-for="col in toggleableColumns"
+                    :key="col.key"
+                    @click="toggleColumn(col.key)"
+                    class="flex w-full items-center justify-between rounded-md px-3 py-2 text-sm text-gray-700 hover:bg-gray-100 dark:text-gray-200 dark:hover:bg-gray-700"
+                  >
+                    <span>{{ col.label }}</span>
+                    <Icon v-if="isColumnVisible(col.key)" name="check" size="sm" class="text-primary-500" />
+                  </button>
+                </div>
+              </div>
+            </div>
             <button
               @click="loadSubscriptions"
               :disabled="loading"
@@ -110,12 +161,18 @@
                 class="flex h-8 w-8 items-center justify-center rounded-full bg-primary-100 dark:bg-primary-900/30"
               >
                 <span class="text-sm font-medium text-primary-700 dark:text-primary-300">
-                  {{ row.user?.email?.charAt(0).toUpperCase() || '?' }}
+                  {{ userColumnMode === 'email'
+                    ? (row.user?.email?.charAt(0).toUpperCase() || '?')
+                    : (row.user?.username?.charAt(0).toUpperCase() || '?')
+                  }}
                 </span>
               </div>
-              <span class="font-medium text-gray-900 dark:text-white">{{
-                row.user?.email || t('admin.redeem.userPrefix', { id: row.user_id })
-              }}</span>
+              <span class="font-medium text-gray-900 dark:text-white">
+                {{ userColumnMode === 'email'
+                  ? (row.user?.email || t('admin.redeem.userPrefix', { id: row.user_id }))
+                  : (row.user?.username || '-')
+                }}
+              </span>
             </div>
           </template>
 
@@ -545,8 +602,43 @@ import Icon from '@/components/icons/Icon.vue'
 const { t } = useI18n()
 const appStore = useAppStore()
 
-const columns = computed<Column[]>(() => [
-  { key: 'user', label: t('admin.subscriptions.columns.user'), sortable: true },
+// User column display mode: 'email' or 'username'
+const userColumnMode = ref<'email' | 'username'>('email')
+const USER_COLUMN_MODE_KEY = 'subscription-user-column-mode'
+
+const loadUserColumnMode = () => {
+  try {
+    const saved = localStorage.getItem(USER_COLUMN_MODE_KEY)
+    if (saved === 'email' || saved === 'username') {
+      userColumnMode.value = saved
+    }
+  } catch (e) {
+    console.error('Failed to load user column mode:', e)
+  }
+}
+
+const saveUserColumnMode = () => {
+  try {
+    localStorage.setItem(USER_COLUMN_MODE_KEY, userColumnMode.value)
+  } catch (e) {
+    console.error('Failed to save user column mode:', e)
+  }
+}
+
+const setUserColumnMode = (mode: 'email' | 'username') => {
+  userColumnMode.value = mode
+  saveUserColumnMode()
+}
+
+// All available columns
+const allColumns = computed<Column[]>(() => [
+  {
+    key: 'user',
+    label: userColumnMode.value === 'email'
+      ? t('admin.subscriptions.columns.user')
+      : t('admin.users.columns.username'),
+    sortable: true
+  },
   { key: 'group', label: t('admin.subscriptions.columns.group'), sortable: true },
   { key: 'usage', label: t('admin.subscriptions.columns.usage'), sortable: false },
   { key: 'expires_at', label: t('admin.subscriptions.columns.expires'), sortable: true },
@@ -554,6 +646,69 @@ const columns = computed<Column[]>(() => [
   { key: 'actions', label: t('admin.subscriptions.columns.actions'), sortable: false }
 ])
 
+// Columns that can be toggled (exclude user and actions which are always visible)
+const toggleableColumns = computed(() =>
+  allColumns.value.filter(col => col.key !== 'user' && col.key !== 'actions')
+)
+
+// Hidden columns set
+const hiddenColumns = reactive<Set<string>>(new Set())
+
+// Default hidden columns
+const DEFAULT_HIDDEN_COLUMNS: string[] = []
+
+// localStorage key
+const HIDDEN_COLUMNS_KEY = 'subscription-hidden-columns'
+
+// Load saved column settings
+const loadSavedColumns = () => {
+  try {
+    const saved = localStorage.getItem(HIDDEN_COLUMNS_KEY)
+    if (saved) {
+      const parsed = JSON.parse(saved) as string[]
+      parsed.forEach(key => hiddenColumns.add(key))
+    } else {
+      DEFAULT_HIDDEN_COLUMNS.forEach(key => hiddenColumns.add(key))
+    }
+  } catch (e) {
+    console.error('Failed to load saved columns:', e)
+    DEFAULT_HIDDEN_COLUMNS.forEach(key => hiddenColumns.add(key))
+  }
+}
+
+// Save column settings to localStorage
+const saveColumnsToStorage = () => {
+  try {
+    localStorage.setItem(HIDDEN_COLUMNS_KEY, JSON.stringify([...hiddenColumns]))
+  } catch (e) {
+    console.error('Failed to save columns:', e)
+  }
+}
+
+// Toggle column visibility
+const toggleColumn = (key: string) => {
+  if (hiddenColumns.has(key)) {
+    hiddenColumns.delete(key)
+  } else {
+    hiddenColumns.add(key)
+  }
+  saveColumnsToStorage()
+}
+
+// Check if column is visible
+const isColumnVisible = (key: string) => !hiddenColumns.has(key)
+
+// Filtered columns for display
+const columns = computed<Column[]>(() =>
+  allColumns.value.filter(col =>
+    col.key === 'user' || col.key === 'actions' || !hiddenColumns.has(col.key)
+  )
+)
+
+// Column dropdown state
+const showColumnDropdown = ref(false)
+const columnDropdownRef = ref<HTMLElement | null>(null)
+
 // Filter options
 const statusOptions = computed(() => [
   { value: '', label: t('admin.subscriptions.allStatus') },
@@ -949,14 +1104,19 @@ const formatResetTime = (windowStart: string, period: 'daily' | 'weekly' | 'mont
   }
 }
 
-// Handle click outside to close user dropdown
+// Handle click outside to close dropdowns
 const handleClickOutside = (event: MouseEvent) => {
   const target = event.target as HTMLElement
   if (!target.closest('[data-assign-user-search]')) showUserDropdown.value = false
   if (!target.closest('[data-filter-user-search]')) showFilterUserDropdown.value = false
+  if (columnDropdownRef.value && !columnDropdownRef.value.contains(target)) {
+    showColumnDropdown.value = false
+  }
 }
 
 onMounted(() => {
+  loadUserColumnMode()
+  loadSavedColumns()
   loadSubscriptions()
   loadGroups()
   document.addEventListener('click', handleClickOutside)