Merge pull request #796 from touwaeriol/feature/apikey-quota-limit

feat: add configurable spending limit for API Key accounts

backend/internal/handler/admin/account_handler.go

@@ -1328,6 +1328,29 @@ func (h *AccountHandler) ClearRateLimit(c *gin.Context) {
 	response.Success(c, h.buildAccountResponseWithRuntime(c.Request.Context(), account))
 }
 
+// ResetQuota handles resetting account quota usage
+// POST /api/v1/admin/accounts/:id/reset-quota
+func (h *AccountHandler) ResetQuota(c *gin.Context) {
+	accountID, err := strconv.ParseInt(c.Param("id"), 10, 64)
+	if err != nil {
+		response.BadRequest(c, "Invalid account ID")
+		return
+	}
+
+	if err := h.adminService.ResetAccountQuota(c.Request.Context(), accountID); err != nil {
+		response.InternalError(c, "Failed to reset account quota: "+err.Error())
+		return
+	}
+
+	account, err := h.adminService.GetAccount(c.Request.Context(), accountID)
+	if err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	response.Success(c, h.buildAccountResponseWithRuntime(c.Request.Context(), account))
+}
+
 // GetTempUnschedulable handles getting temporary unschedulable status
 // GET /api/v1/admin/accounts/:id/temp-unschedulable
 func (h *AccountHandler) GetTempUnschedulable(c *gin.Context) {

backend/internal/handler/admin/admin_service_stub_test.go

@@ -425,5 +425,9 @@ func (s *stubAdminService) AdminUpdateAPIKeyGroupID(ctx context.Context, keyID i
 	return nil, service.ErrAPIKeyNotFound
 }
 
+func (s *stubAdminService) ResetAccountQuota(ctx context.Context, id int64) error {
+	return nil
+}
+
 // Ensure stub implements interface.
 var _ service.AdminService = (*stubAdminService)(nil)

backend/internal/handler/dto/mappers.go

@@ -248,6 +248,17 @@ func AccountFromServiceShallow(a *service.Account) *Account {
 		}
 	}
 
+	// 提取 API Key 账号配额限制（仅 apikey 类型有效）
+	if a.Type == service.AccountTypeAPIKey {
+		if limit := a.GetQuotaLimit(); limit > 0 {
+			out.QuotaLimit = &limit
+		}
+		used := a.GetQuotaUsed()
+		if out.QuotaLimit != nil {
+			out.QuotaUsed = &used
+		}
+	}
+
 	return out
 }
 

backend/internal/handler/dto/types.go

@@ -185,6 +185,10 @@ type Account struct {
 	CacheTTLOverrideEnabled *bool   `json:"cache_ttl_override_enabled,omitempty"`
 	CacheTTLOverrideTarget  *string `json:"cache_ttl_override_target,omitempty"`
 
+	// API Key 账号配额限制
+	QuotaLimit *float64 `json:"quota_limit,omitempty"`
+	QuotaUsed  *float64 `json:"quota_used,omitempty"`
+
 	Proxy         *Proxy         `json:"proxy,omitempty"`
 	AccountGroups []AccountGroup `json:"account_groups,omitempty"`
 

backend/internal/handler/sora_client_handler_test.go

@@ -2132,6 +2132,14 @@ func (r *stubAccountRepoForHandler) BulkUpdate(context.Context, []int64, service
 	return 0, nil
 }
 
+func (r *stubAccountRepoForHandler) IncrementQuotaUsed(context.Context, int64, float64) error {
+	return nil
+}
+
+func (r *stubAccountRepoForHandler) ResetQuotaUsed(context.Context, int64) error {
+	return nil
+}
+
 // ==================== Stub: SoraClient (用于 SoraGatewayService) ====================
 
 var _ service.SoraClient = (*stubSoraClientForHandler)(nil)

backend/internal/handler/sora_gateway_handler_test.go

@@ -216,6 +216,14 @@ func (r *stubAccountRepo) BulkUpdate(ctx context.Context, ids []int64, updates s
 	return 0, nil
 }
 
+func (r *stubAccountRepo) IncrementQuotaUsed(ctx context.Context, id int64, amount float64) error {
+	return nil
+}
+
+func (r *stubAccountRepo) ResetQuotaUsed(ctx context.Context, id int64) error {
+	return nil
+}
+
 func (r *stubAccountRepo) listSchedulable() []service.Account {
 	var result []service.Account
 	for _, acc := range r.accounts {

backend/internal/repository/account_repo.go

@@ -1657,3 +1657,60 @@ func (r *accountRepository) FindByExtraField(ctx context.Context, key string, va
 
 	return r.accountsToService(ctx, accounts)
 }
+
+// IncrementQuotaUsed 原子递增账号的 extra.quota_used 字段
+func (r *accountRepository) IncrementQuotaUsed(ctx context.Context, id int64, amount float64) error {
+	rows, err := r.sql.QueryContext(ctx,
+		`UPDATE accounts SET extra = jsonb_set(
+			COALESCE(extra, '{}'::jsonb),
+			'{quota_used}',
+			to_jsonb(COALESCE((extra->>'quota_used')::numeric, 0) + $1)
+		), updated_at = NOW()
+		WHERE id = $2 AND deleted_at IS NULL
+		RETURNING
+			COALESCE((extra->>'quota_used')::numeric, 0),
+			COALESCE((extra->>'quota_limit')::numeric, 0)`,
+		amount, id)
+	if err != nil {
+		return err
+	}
+	defer func() { _ = rows.Close() }()
+
+	var newUsed, limit float64
+	if rows.Next() {
+		if err := rows.Scan(&newUsed, &limit); err != nil {
+			return err
+		}
+	}
+	if err := rows.Err(); err != nil {
+		return err
+	}
+
+	// 配额刚超限时触发调度快照刷新，使账号及时从调度候选中移除
+	if limit > 0 && newUsed >= limit && (newUsed-amount) < limit {
+		if err := enqueueSchedulerOutbox(ctx, r.sql, service.SchedulerOutboxEventAccountChanged, &id, nil, nil); err != nil {
+			logger.LegacyPrintf("repository.account", "[SchedulerOutbox] enqueue quota exceeded failed: account=%d err=%v", id, err)
+		}
+	}
+	return nil
+}
+
+// ResetQuotaUsed 重置账号的 extra.quota_used 为 0
+func (r *accountRepository) ResetQuotaUsed(ctx context.Context, id int64) error {
+	_, err := r.sql.ExecContext(ctx,
+		`UPDATE accounts SET extra = jsonb_set(
+			COALESCE(extra, '{}'::jsonb),
+			'{quota_used}',
+			'0'::jsonb
+		), updated_at = NOW()
+		WHERE id = $1 AND deleted_at IS NULL`,
+		id)
+	if err != nil {
+		return err
+	}
+	// 重置配额后触发调度快照刷新，使账号重新参与调度
+	if err := enqueueSchedulerOutbox(ctx, r.sql, service.SchedulerOutboxEventAccountChanged, &id, nil, nil); err != nil {
+		logger.LegacyPrintf("repository.account", "[SchedulerOutbox] enqueue quota reset failed: account=%d err=%v", id, err)
+	}
+	return nil
+}

backend/internal/server/api_contract_test.go

@@ -1096,6 +1096,14 @@ func (s *stubAccountRepo) UpdateExtra(ctx context.Context, id int64, updates map
 	return errors.New("not implemented")
 }
 
+func (s *stubAccountRepo) IncrementQuotaUsed(ctx context.Context, id int64, amount float64) error {
+	return errors.New("not implemented")
+}
+
+func (s *stubAccountRepo) ResetQuotaUsed(ctx context.Context, id int64) error {
+	return errors.New("not implemented")
+}
+
 func (s *stubAccountRepo) BulkUpdate(ctx context.Context, ids []int64, updates service.AccountBulkUpdate) (int64, error) {
 	s.bulkUpdateIDs = append([]int64{}, ids...)
 	return int64(len(ids)), nil

backend/internal/server/routes/admin.go

@@ -252,6 +252,7 @@ func registerAccountRoutes(admin *gin.RouterGroup, h *handler.Handlers) {
 		accounts.GET("/:id/today-stats", h.Admin.Account.GetTodayStats)
 		accounts.POST("/today-stats/batch", h.Admin.Account.GetBatchTodayStats)
 		accounts.POST("/:id/clear-rate-limit", h.Admin.Account.ClearRateLimit)
+		accounts.POST("/:id/reset-quota", h.Admin.Account.ResetQuota)
 		accounts.GET("/:id/temp-unschedulable", h.Admin.Account.GetTempUnschedulable)
 		accounts.DELETE("/:id/temp-unschedulable", h.Admin.Account.ClearTempUnschedulable)
 		accounts.POST("/:id/schedulable", h.Admin.Account.SetSchedulable)

backend/internal/service/account.go

@@ -1117,6 +1117,38 @@ func (a *Account) GetCacheTTLOverrideTarget() string {
 	return "5m"
 }
 
+// GetQuotaLimit 获取 API Key 账号的配额限制（美元）
+// 返回 0 表示未启用
+func (a *Account) GetQuotaLimit() float64 {
+	if a.Extra == nil {
+		return 0
+	}
+	if v, ok := a.Extra["quota_limit"]; ok {
+		return parseExtraFloat64(v)
+	}
+	return 0
+}
+
+// GetQuotaUsed 获取 API Key 账号的已用配额（美元）
+func (a *Account) GetQuotaUsed() float64 {
+	if a.Extra == nil {
+		return 0
+	}
+	if v, ok := a.Extra["quota_used"]; ok {
+		return parseExtraFloat64(v)
+	}
+	return 0
+}
+
+// IsQuotaExceeded 检查 API Key 账号配额是否已超限
+func (a *Account) IsQuotaExceeded() bool {
+	limit := a.GetQuotaLimit()
+	if limit <= 0 {
+		return false
+	}
+	return a.GetQuotaUsed() >= limit
+}
+
 // 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 {

backend/internal/service/account_service.go

@@ -68,6 +68,10 @@ type AccountRepository interface {
 	UpdateSessionWindow(ctx context.Context, id int64, start, end *time.Time, status string) error
 	UpdateExtra(ctx context.Context, id int64, updates map[string]any) error
 	BulkUpdate(ctx context.Context, ids []int64, updates AccountBulkUpdate) (int64, error)
+	// IncrementQuotaUsed 原子递增 API Key 账号的配额用量
+	IncrementQuotaUsed(ctx context.Context, id int64, amount float64) error
+	// ResetQuotaUsed 重置 API Key 账号的配额用量为 0
+	ResetQuotaUsed(ctx context.Context, id int64) error
 }
 
 // AccountBulkUpdate describes the fields that can be updated in a bulk operation.

backend/internal/service/account_service_delete_test.go

@@ -199,6 +199,14 @@ func (s *accountRepoStub) BulkUpdate(ctx context.Context, ids []int64, updates A
 	panic("unexpected BulkUpdate call")
 }
 
+func (s *accountRepoStub) IncrementQuotaUsed(ctx context.Context, id int64, amount float64) error {
+	return nil
+}
+
+func (s *accountRepoStub) ResetQuotaUsed(ctx context.Context, id int64) error {
+	return nil
+}
+
 // TestAccountService_Delete_NotFound 测试删除不存在的账号时返回正确的错误。
 // 预期行为：
 //   - ExistsByID 返回 false（账号不存在）

backend/internal/service/account_test_service.go

@@ -180,7 +180,7 @@ func (s *AccountTestService) TestAccountConnection(c *gin.Context, accountID int
 	}
 
 	if account.Platform == PlatformAntigravity {
-		return s.testAntigravityAccountConnection(c, account, modelID)
+		return s.routeAntigravityTest(c, account, modelID)
 	}
 
 	if account.Platform == PlatformSora {
@@ -1177,6 +1177,18 @@ func truncateSoraErrorBody(body []byte, max int) string {
 	return soraerror.TruncateBody(body, max)
 }
 
+// routeAntigravityTest 路由 Antigravity 账号的测试请求。
+// APIKey 类型走原生协议（与 gateway_handler 路由一致），OAuth/Upstream 走 CRS 中转。
+func (s *AccountTestService) routeAntigravityTest(c *gin.Context, account *Account, modelID string) error {
+	if account.Type == AccountTypeAPIKey {
+		if strings.HasPrefix(modelID, "gemini-") {
+			return s.testGeminiAccountConnection(c, account, modelID)
+		}
+		return s.testClaudeAccountConnection(c, account, modelID)
+	}
+	return s.testAntigravityAccountConnection(c, account, modelID)
+}
+
 // testAntigravityAccountConnection tests an Antigravity account's connection
 // 支持 Claude 和 Gemini 两种协议，使用非流式请求
 func (s *AccountTestService) testAntigravityAccountConnection(c *gin.Context, account *Account, modelID string) error {

backend/internal/service/admin_service.go

@@ -84,6 +84,7 @@ type AdminService interface {
 	DeleteRedeemCode(ctx context.Context, id int64) error
 	BatchDeleteRedeemCodes(ctx context.Context, ids []int64) (int64, error)
 	ExpireRedeemCode(ctx context.Context, id int64) (*RedeemCode, error)
+	ResetAccountQuota(ctx context.Context, id int64) error
 }
 
 // CreateUserInput represents input for creating a new user via admin operations.
@@ -1458,6 +1459,10 @@ func (s *adminServiceImpl) UpdateAccount(ctx context.Context, id int64, input *U
 		account.Credentials = input.Credentials
 	}
 	if len(input.Extra) > 0 {
+		// 保留 quota_used，防止编辑账号时意外重置配额用量
+		if oldQuotaUsed, ok := account.Extra["quota_used"]; ok {
+			input.Extra["quota_used"] = oldQuotaUsed
+		}
 		account.Extra = input.Extra
 	}
 	if input.ProxyID != nil {
@@ -2439,3 +2444,7 @@ func (e *MixedChannelError) Error() string {
 	return fmt.Sprintf("mixed_channel_warning: Group '%s' contains both %s and %s accounts. Using mixed channels in the same context may cause thinking block signature validation issues, which will fallback to non-thinking mode for historical messages.",
 		e.GroupName, e.CurrentPlatform, e.OtherPlatform)
 }
+
+func (s *adminServiceImpl) ResetAccountQuota(ctx context.Context, id int64) error {
+	return s.accountRepo.ResetQuotaUsed(ctx, id)
+}

backend/internal/service/gateway_multiplatform_test.go

@@ -187,6 +187,14 @@ func (m *mockAccountRepoForPlatform) BulkUpdate(ctx context.Context, ids []int64
 	return 0, nil
 }
 
+func (m *mockAccountRepoForPlatform) IncrementQuotaUsed(ctx context.Context, id int64, amount float64) error {
+	return nil
+}
+
+func (m *mockAccountRepoForPlatform) ResetQuotaUsed(ctx context.Context, id int64) error {
+	return nil
+}
+
 // Verify interface implementation
 var _ AccountRepository = (*mockAccountRepoForPlatform)(nil)
 

backend/internal/service/gateway_service.go

@@ -1228,6 +1228,10 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 				modelScopeSkippedIDs = append(modelScopeSkippedIDs, account.ID)
 				continue
 			}
+			// 配额检查
+			if !s.isAccountSchedulableForQuota(account) {
+				continue
+			}
 			// 窗口费用检查（非粘性会话路径）
 			if !s.isAccountSchedulableForWindowCost(ctx, account, false) {
 				filteredWindowCost++
@@ -1260,6 +1264,7 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 							s.isAccountAllowedForPlatform(stickyAccount, platform, useMixed) &&
 							(requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, stickyAccount, requestedModel)) &&
 							s.isAccountSchedulableForModelSelection(ctx, stickyAccount, requestedModel) &&
+							s.isAccountSchedulableForQuota(stickyAccount) &&
 							s.isAccountSchedulableForWindowCost(ctx, stickyAccount, true) &&
 
 							s.isAccountSchedulableForRPM(ctx, stickyAccount, true) { // 粘性会话窗口费用+RPM 检查
@@ -1416,6 +1421,7 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 					s.isAccountAllowedForPlatform(account, platform, useMixed) &&
 					(requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) &&
 					s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) &&
+					s.isAccountSchedulableForQuota(account) &&
 					s.isAccountSchedulableForWindowCost(ctx, account, true) &&
 
 					s.isAccountSchedulableForRPM(ctx, account, true) { // 粘性会话窗口费用+RPM 检查
@@ -1480,6 +1486,10 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 		if !s.isAccountSchedulableForModelSelection(ctx, acc, requestedModel) {
 			continue
 		}
+		// 配额检查
+		if !s.isAccountSchedulableForQuota(acc) {
+			continue
+		}
 		// 窗口费用检查（非粘性会话路径）
 		if !s.isAccountSchedulableForWindowCost(ctx, acc, false) {
 			continue
@@ -2113,6 +2123,15 @@ func (s *GatewayService) withWindowCostPrefetch(ctx context.Context, accounts []
 	return context.WithValue(ctx, windowCostPrefetchContextKey, costs)
 }
 
+// isAccountSchedulableForQuota 检查 API Key 账号是否在配额限制内
+// 仅适用于配置了 quota_limit 的 apikey 类型账号
+func (s *GatewayService) isAccountSchedulableForQuota(account *Account) bool {
+	if account.Type != AccountTypeAPIKey {
+		return true
+	}
+	return !account.IsQuotaExceeded()
+}
+
 // 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
@@ -2590,7 +2609,7 @@ func (s *GatewayService) selectAccountForModelWithPlatform(ctx context.Context,
 						if clearSticky {
 							_ = s.cache.DeleteSessionAccountID(ctx, derefGroupID(groupID), sessionHash)
 						}
-						if !clearSticky && s.isAccountInGroup(account, groupID) && account.Platform == platform && (requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) && s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) && s.isAccountSchedulableForWindowCost(ctx, account, true) && s.isAccountSchedulableForRPM(ctx, account, true) {
+						if !clearSticky && s.isAccountInGroup(account, groupID) && account.Platform == platform && (requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) && s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) && s.isAccountSchedulableForQuota(account) && s.isAccountSchedulableForWindowCost(ctx, account, true) && s.isAccountSchedulableForRPM(ctx, account, true) {
 							if s.debugModelRoutingEnabled() {
 								logger.LegacyPrintf("service.gateway", "[ModelRoutingDebug] legacy routed sticky hit: group_id=%v model=%s session=%s account=%d", derefGroupID(groupID), requestedModel, shortSessionHash(sessionHash), accountID)
 							}
@@ -2644,6 +2663,9 @@ func (s *GatewayService) selectAccountForModelWithPlatform(ctx context.Context,
 			if !s.isAccountSchedulableForModelSelection(ctx, acc, requestedModel) {
 				continue
 			}
+			if !s.isAccountSchedulableForQuota(acc) {
+				continue
+			}
 			if !s.isAccountSchedulableForWindowCost(ctx, acc, false) {
 				continue
 			}
@@ -2700,7 +2722,7 @@ func (s *GatewayService) selectAccountForModelWithPlatform(ctx context.Context,
 					if clearSticky {
 						_ = s.cache.DeleteSessionAccountID(ctx, derefGroupID(groupID), sessionHash)
 					}
-					if !clearSticky && s.isAccountInGroup(account, groupID) && account.Platform == platform && (requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) && s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) && s.isAccountSchedulableForWindowCost(ctx, account, true) && s.isAccountSchedulableForRPM(ctx, account, true) {
+					if !clearSticky && s.isAccountInGroup(account, groupID) && account.Platform == platform && (requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) && s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) && s.isAccountSchedulableForQuota(account) && s.isAccountSchedulableForWindowCost(ctx, account, true) && s.isAccountSchedulableForRPM(ctx, account, true) {
 						return account, nil
 					}
 				}
@@ -2743,6 +2765,9 @@ func (s *GatewayService) selectAccountForModelWithPlatform(ctx context.Context,
 		if !s.isAccountSchedulableForModelSelection(ctx, acc, requestedModel) {
 			continue
 		}
+		if !s.isAccountSchedulableForQuota(acc) {
+			continue
+		}
 		if !s.isAccountSchedulableForWindowCost(ctx, acc, false) {
 			continue
 		}
@@ -2818,7 +2843,7 @@ func (s *GatewayService) selectAccountWithMixedScheduling(ctx context.Context, g
 						if clearSticky {
 							_ = s.cache.DeleteSessionAccountID(ctx, derefGroupID(groupID), sessionHash)
 						}
-						if !clearSticky && s.isAccountInGroup(account, groupID) && (requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) && s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) && s.isAccountSchedulableForWindowCost(ctx, account, true) && s.isAccountSchedulableForRPM(ctx, account, true) {
+						if !clearSticky && s.isAccountInGroup(account, groupID) && (requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) && s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) && s.isAccountSchedulableForQuota(account) && s.isAccountSchedulableForWindowCost(ctx, account, true) && s.isAccountSchedulableForRPM(ctx, account, true) {
 							if account.Platform == nativePlatform || (account.Platform == PlatformAntigravity && account.IsMixedSchedulingEnabled()) {
 								if s.debugModelRoutingEnabled() {
 									logger.LegacyPrintf("service.gateway", "[ModelRoutingDebug] legacy mixed routed sticky hit: group_id=%v model=%s session=%s account=%d", derefGroupID(groupID), requestedModel, shortSessionHash(sessionHash), accountID)
@@ -2874,6 +2899,9 @@ func (s *GatewayService) selectAccountWithMixedScheduling(ctx context.Context, g
 			if !s.isAccountSchedulableForModelSelection(ctx, acc, requestedModel) {
 				continue
 			}
+			if !s.isAccountSchedulableForQuota(acc) {
+				continue
+			}
 			if !s.isAccountSchedulableForWindowCost(ctx, acc, false) {
 				continue
 			}
@@ -2930,7 +2958,7 @@ func (s *GatewayService) selectAccountWithMixedScheduling(ctx context.Context, g
 					if clearSticky {
 						_ = s.cache.DeleteSessionAccountID(ctx, derefGroupID(groupID), sessionHash)
 					}
-					if !clearSticky && s.isAccountInGroup(account, groupID) && (requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) && s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) && s.isAccountSchedulableForWindowCost(ctx, account, true) && s.isAccountSchedulableForRPM(ctx, account, true) {
+					if !clearSticky && s.isAccountInGroup(account, groupID) && (requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) && s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) && s.isAccountSchedulableForQuota(account) && s.isAccountSchedulableForWindowCost(ctx, account, true) && s.isAccountSchedulableForRPM(ctx, account, true) {
 						if account.Platform == nativePlatform || (account.Platform == PlatformAntigravity && account.IsMixedSchedulingEnabled()) {
 							return account, nil
 						}
@@ -2975,6 +3003,9 @@ func (s *GatewayService) selectAccountWithMixedScheduling(ctx context.Context, g
 		if !s.isAccountSchedulableForModelSelection(ctx, acc, requestedModel) {
 			continue
 		}
+		if !s.isAccountSchedulableForQuota(acc) {
+			continue
+		}
 		if !s.isAccountSchedulableForWindowCost(ctx, acc, false) {
 			continue
 		}
@@ -6379,6 +6410,89 @@ type APIKeyQuotaUpdater interface {
 	UpdateRateLimitUsage(ctx context.Context, apiKeyID int64, cost float64) error
 }
 
+// postUsageBillingParams 统一扣费所需的参数
+type postUsageBillingParams struct {
+	Cost                  *CostBreakdown
+	User                  *User
+	APIKey                *APIKey
+	Account               *Account
+	Subscription          *UserSubscription
+	IsSubscriptionBill    bool
+	AccountRateMultiplier float64
+	APIKeyService         APIKeyQuotaUpdater
+}
+
+// postUsageBilling 统一处理使用量记录后的扣费逻辑：
+//   - 订阅/余额扣费
+//   - API Key 配额更新
+//   - API Key 限速用量更新
+//   - 账号配额用量更新（账号口径：TotalCost × 账号计费倍率）
+func postUsageBilling(ctx context.Context, p *postUsageBillingParams, deps *billingDeps) {
+	cost := p.Cost
+
+	// 1. 订阅 / 余额扣费
+	if p.IsSubscriptionBill {
+		if cost.TotalCost > 0 {
+			if err := deps.userSubRepo.IncrementUsage(ctx, p.Subscription.ID, cost.TotalCost); err != nil {
+				slog.Error("increment subscription usage failed", "subscription_id", p.Subscription.ID, "error", err)
+			}
+			deps.billingCacheService.QueueUpdateSubscriptionUsage(p.User.ID, *p.APIKey.GroupID, cost.TotalCost)
+		}
+	} else {
+		if cost.ActualCost > 0 {
+			if err := deps.userRepo.DeductBalance(ctx, p.User.ID, cost.ActualCost); err != nil {
+				slog.Error("deduct balance failed", "user_id", p.User.ID, "error", err)
+			}
+			deps.billingCacheService.QueueDeductBalance(p.User.ID, cost.ActualCost)
+		}
+	}
+
+	// 2. API Key 配额
+	if cost.ActualCost > 0 && p.APIKey.Quota > 0 && p.APIKeyService != nil {
+		if err := p.APIKeyService.UpdateQuotaUsed(ctx, p.APIKey.ID, cost.ActualCost); err != nil {
+			slog.Error("update api key quota failed", "api_key_id", p.APIKey.ID, "error", err)
+		}
+	}
+
+	// 3. API Key 限速用量
+	if cost.ActualCost > 0 && p.APIKey.HasRateLimits() && p.APIKeyService != nil {
+		if err := p.APIKeyService.UpdateRateLimitUsage(ctx, p.APIKey.ID, cost.ActualCost); err != nil {
+			slog.Error("update api key rate limit usage failed", "api_key_id", p.APIKey.ID, "error", err)
+		}
+		deps.billingCacheService.QueueUpdateAPIKeyRateLimitUsage(p.APIKey.ID, cost.ActualCost)
+	}
+
+	// 4. 账号配额用量（账号口径：TotalCost × 账号计费倍率）
+	if cost.TotalCost > 0 && p.Account.Type == AccountTypeAPIKey && p.Account.GetQuotaLimit() > 0 {
+		accountCost := cost.TotalCost * p.AccountRateMultiplier
+		if err := deps.accountRepo.IncrementQuotaUsed(ctx, p.Account.ID, accountCost); err != nil {
+			slog.Error("increment account quota used failed", "account_id", p.Account.ID, "cost", accountCost, "error", err)
+		}
+	}
+
+	// 5. 更新账号最近使用时间
+	deps.deferredService.ScheduleLastUsedUpdate(p.Account.ID)
+}
+
+// billingDeps 扣费逻辑依赖的服务（由各 gateway service 提供）
+type billingDeps struct {
+	accountRepo         AccountRepository
+	userRepo            UserRepository
+	userSubRepo         UserSubscriptionRepository
+	billingCacheService *BillingCacheService
+	deferredService     *DeferredService
+}
+
+func (s *GatewayService) billingDeps() *billingDeps {
+	return &billingDeps{
+		accountRepo:         s.accountRepo,
+		userRepo:            s.userRepo,
+		userSubRepo:         s.userSubRepo,
+		billingCacheService: s.billingCacheService,
+		deferredService:     s.deferredService,
+	}
+}
+
 // RecordUsage 记录使用量并扣费（或更新订阅用量）
 func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInput) error {
 	result := input.Result
@@ -6542,44 +6656,20 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
 
 	shouldBill := inserted || err != nil
 
-	// 根据计费类型执行扣费
-	if isSubscriptionBilling {
-		// 订阅模式：更新订阅用量（使用 TotalCost 原始费用，不考虑倍率）
-		if shouldBill && cost.TotalCost > 0 {
-			if err := s.userSubRepo.IncrementUsage(ctx, subscription.ID, cost.TotalCost); err != nil {
-				logger.LegacyPrintf("service.gateway", "Increment subscription usage failed: %v", err)
-			}
-			// 异步更新订阅缓存
-			s.billingCacheService.QueueUpdateSubscriptionUsage(user.ID, *apiKey.GroupID, cost.TotalCost)
-		}
+	if shouldBill {
+		postUsageBilling(ctx, &postUsageBillingParams{
+			Cost:                  cost,
+			User:                  user,
+			APIKey:                apiKey,
+			Account:               account,
+			Subscription:          subscription,
+			IsSubscriptionBill:    isSubscriptionBilling,
+			AccountRateMultiplier: accountRateMultiplier,
+			APIKeyService:         input.APIKeyService,
+		}, s.billingDeps())
 	} else {
-		// 余额模式：扣除用户余额（使用 ActualCost 考虑倍率后的费用）
-		if shouldBill && cost.ActualCost > 0 {
-			if err := s.userRepo.DeductBalance(ctx, user.ID, cost.ActualCost); err != nil {
-				logger.LegacyPrintf("service.gateway", "Deduct balance failed: %v", err)
-			}
-			// 异步更新余额缓存
-			s.billingCacheService.QueueDeductBalance(user.ID, cost.ActualCost)
-		}
-	}
-
-	// 更新 API Key 配额（如果设置了配额限制）
-	if shouldBill && cost.ActualCost > 0 && apiKey.Quota > 0 && input.APIKeyService != nil {
-		if err := input.APIKeyService.UpdateQuotaUsed(ctx, apiKey.ID, cost.ActualCost); err != nil {
-			logger.LegacyPrintf("service.gateway", "Update API key quota failed: %v", err)
-		}
-	}
-
-	// Update API Key rate limit usage
-	if shouldBill && cost.ActualCost > 0 && apiKey.HasRateLimits() && input.APIKeyService != nil {
-		if err := input.APIKeyService.UpdateRateLimitUsage(ctx, apiKey.ID, cost.ActualCost); err != nil {
-			logger.LegacyPrintf("service.gateway", "Update API key rate limit usage failed: %v", err)
-		}
-		s.billingCacheService.QueueUpdateAPIKeyRateLimitUsage(apiKey.ID, cost.ActualCost)
-	}
-
-	// Schedule batch update for account last_used_at
 		s.deferredService.ScheduleLastUsedUpdate(account.ID)
+	}
 
 	return nil
 }
@@ -6740,43 +6830,20 @@ func (s *GatewayService) RecordUsageWithLongContext(ctx context.Context, input *
 
 	shouldBill := inserted || err != nil
 
-	// 根据计费类型执行扣费
-	if isSubscriptionBilling {
-		// 订阅模式：更新订阅用量（使用 TotalCost 原始费用，不考虑倍率）
-		if shouldBill && cost.TotalCost > 0 {
-			if err := s.userSubRepo.IncrementUsage(ctx, subscription.ID, cost.TotalCost); err != nil {
-				logger.LegacyPrintf("service.gateway", "Increment subscription usage failed: %v", err)
-			}
-			// 异步更新订阅缓存
-			s.billingCacheService.QueueUpdateSubscriptionUsage(user.ID, *apiKey.GroupID, cost.TotalCost)
-		}
+	if shouldBill {
+		postUsageBilling(ctx, &postUsageBillingParams{
+			Cost:                  cost,
+			User:                  user,
+			APIKey:                apiKey,
+			Account:               account,
+			Subscription:          subscription,
+			IsSubscriptionBill:    isSubscriptionBilling,
+			AccountRateMultiplier: accountRateMultiplier,
+			APIKeyService:         input.APIKeyService,
+		}, s.billingDeps())
 	} else {
-		// 余额模式：扣除用户余额（使用 ActualCost 考虑倍率后的费用）
-		if shouldBill && cost.ActualCost > 0 {
-			if err := s.userRepo.DeductBalance(ctx, user.ID, cost.ActualCost); err != nil {
-				logger.LegacyPrintf("service.gateway", "Deduct balance failed: %v", err)
-			}
-			// 异步更新余额缓存
-			s.billingCacheService.QueueDeductBalance(user.ID, cost.ActualCost)
-			// API Key 独立配额扣费
-			if input.APIKeyService != nil && apiKey.Quota > 0 {
-				if err := input.APIKeyService.UpdateQuotaUsed(ctx, apiKey.ID, cost.ActualCost); err != nil {
-					logger.LegacyPrintf("service.gateway", "Add API key quota used failed: %v", err)
-				}
-			}
-		}
-	}
-
-	// Update API Key rate limit usage
-	if shouldBill && cost.ActualCost > 0 && apiKey.HasRateLimits() && input.APIKeyService != nil {
-		if err := input.APIKeyService.UpdateRateLimitUsage(ctx, apiKey.ID, cost.ActualCost); err != nil {
-			logger.LegacyPrintf("service.gateway", "Update API key rate limit usage failed: %v", err)
-		}
-		s.billingCacheService.QueueUpdateAPIKeyRateLimitUsage(apiKey.ID, cost.ActualCost)
-	}
-
-	// Schedule batch update for account last_used_at
 		s.deferredService.ScheduleLastUsedUpdate(account.ID)
+	}
 
 	return nil
 }

backend/internal/service/gemini_multiplatform_test.go

@@ -176,6 +176,14 @@ func (m *mockAccountRepoForGemini) BulkUpdate(ctx context.Context, ids []int64,
 	return 0, nil
 }
 
+func (m *mockAccountRepoForGemini) IncrementQuotaUsed(ctx context.Context, id int64, amount float64) error {
+	return nil
+}
+
+func (m *mockAccountRepoForGemini) ResetQuotaUsed(ctx context.Context, id int64) error {
+	return nil
+}
+
 // Verify interface implementation
 var _ AccountRepository = (*mockAccountRepoForGemini)(nil)
 

backend/internal/service/openai_gateway_service.go

@@ -319,6 +319,16 @@ func NewOpenAIGatewayService(
 	return svc
 }
 
+func (s *OpenAIGatewayService) billingDeps() *billingDeps {
+	return &billingDeps{
+		accountRepo:         s.accountRepo,
+		userRepo:            s.userRepo,
+		userSubRepo:         s.userSubRepo,
+		billingCacheService: s.billingCacheService,
+		deferredService:     s.deferredService,
+	}
+}
+
 // CloseOpenAIWSPool 关闭 OpenAI WebSocket 连接池的后台 worker 和空闲连接。
 // 应在应用优雅关闭时调用。
 func (s *OpenAIGatewayService) CloseOpenAIWSPool() {
@@ -3474,36 +3484,20 @@ func (s *OpenAIGatewayService) RecordUsage(ctx context.Context, input *OpenAIRec
 
 	shouldBill := inserted || err != nil
 
-	// Deduct based on billing type
-	if isSubscriptionBilling {
-		if shouldBill && cost.TotalCost > 0 {
-			_ = s.userSubRepo.IncrementUsage(ctx, subscription.ID, cost.TotalCost)
-			s.billingCacheService.QueueUpdateSubscriptionUsage(user.ID, *apiKey.GroupID, cost.TotalCost)
-		}
+	if shouldBill {
+		postUsageBilling(ctx, &postUsageBillingParams{
+			Cost:                  cost,
+			User:                  user,
+			APIKey:                apiKey,
+			Account:               account,
+			Subscription:          subscription,
+			IsSubscriptionBill:    isSubscriptionBilling,
+			AccountRateMultiplier: accountRateMultiplier,
+			APIKeyService:         input.APIKeyService,
+		}, s.billingDeps())
 	} else {
-		if shouldBill && cost.ActualCost > 0 {
-			_ = s.userRepo.DeductBalance(ctx, user.ID, cost.ActualCost)
-			s.billingCacheService.QueueDeductBalance(user.ID, cost.ActualCost)
-		}
-	}
-
-	// Update API key quota if applicable (only for balance mode with quota set)
-	if shouldBill && cost.ActualCost > 0 && apiKey.Quota > 0 && input.APIKeyService != nil {
-		if err := input.APIKeyService.UpdateQuotaUsed(ctx, apiKey.ID, cost.ActualCost); err != nil {
-			logger.LegacyPrintf("service.openai_gateway", "Update API key quota failed: %v", err)
-		}
-	}
-
-	// Update API Key rate limit usage
-	if shouldBill && cost.ActualCost > 0 && apiKey.HasRateLimits() && input.APIKeyService != nil {
-		if err := input.APIKeyService.UpdateRateLimitUsage(ctx, apiKey.ID, cost.ActualCost); err != nil {
-			logger.LegacyPrintf("service.openai_gateway", "Update API key rate limit usage failed: %v", err)
-		}
-		s.billingCacheService.QueueUpdateAPIKeyRateLimitUsage(apiKey.ID, cost.ActualCost)
-	}
-
-	// Schedule batch update for account last_used_at
 		s.deferredService.ScheduleLastUsedUpdate(account.ID)
+	}
 
 	return nil
 }

backend/internal/service/openai_ws_forwarder.go

@@ -864,7 +864,8 @@ func isOpenAIWSClientDisconnectError(err error) bool {
 		strings.Contains(message, "unexpected eof") ||
 		strings.Contains(message, "use of closed network connection") ||
 		strings.Contains(message, "connection reset by peer") ||
-		strings.Contains(message, "broken pipe")
+		strings.Contains(message, "broken pipe") ||
+		strings.Contains(message, "an established connection was aborted")
 }
 
 func classifyOpenAIWSReadFallbackReason(err error) string {

backend/internal/service/subscription_calculate_progress_test.go

@@ -34,7 +34,7 @@ func TestCalculateProgress_BasicFields(t *testing.T) {
 	assert.Equal(t, int64(100), progress.ID)
 	assert.Equal(t, "Premium", progress.GroupName)
 	assert.Equal(t, sub.ExpiresAt, progress.ExpiresAt)
-	assert.Equal(t, 29, progress.ExpiresInDays) // 约 30 天
+	assert.True(t, progress.ExpiresInDays == 29 || progress.ExpiresInDays == 30, "ExpiresInDays should be 29 or 30, got %d", progress.ExpiresInDays)
 	assert.Nil(t, progress.Daily, "无日限额时 Daily 应为 nil")
 	assert.Nil(t, progress.Weekly, "无周限额时 Weekly 应为 nil")
 	assert.Nil(t, progress.Monthly, "无月限额时 Monthly 应为 nil")