feat(billing): 网关计费迁移到 CalculateCostUnified + 模型限制错误统一

- GatewayService/OpenAIGatewayService 注入 ModelPricingResolver
- RecordUsage 从旧路径迁移到 CalculateCostUnified（支持 per_request/image 模式）
- 无渠道时自动回退旧路径，保持原有行为
- 长上下文双倍计费仅在无渠道定价时生效
- CostBreakdown 新增 BillingMode 字段，使用日志记录实际计费模式
- 模型限制错误改为与"无可用账号"相同的 503 响应

backend/cmd/server/VERSION

@@ -1 +1 @@
-0.1.105.14
+0.1.105.15

backend/cmd/server/wire_gen.go

@@ -177,10 +177,9 @@ func initializeApplication(buildInfo handler.BuildInfo) (*Application, error) {
 	digestSessionStore := service.NewDigestSessionStore()
 	channelService := service.NewChannelService(channelRepository, apiKeyAuthCacheInvalidator)
 	modelPricingResolver := service.NewModelPricingResolver(channelService, billingService)
-	_ = modelPricingResolver // Phase 4: 已注册，后续 Gateway 迁移时使用
-	gatewayService := service.NewGatewayService(accountRepository, groupRepository, usageLogRepository, usageBillingRepository, userRepository, userSubscriptionRepository, userGroupRateRepository, gatewayCache, configConfig, schedulerSnapshotService, concurrencyService, billingService, rateLimitService, billingCacheService, identityService, httpUpstream, deferredService, claudeTokenProvider, sessionLimitCache, rpmCache, digestSessionStore, settingService, tlsFingerprintProfileService, channelService)
+	gatewayService := service.NewGatewayService(accountRepository, groupRepository, usageLogRepository, usageBillingRepository, userRepository, userSubscriptionRepository, userGroupRateRepository, gatewayCache, configConfig, schedulerSnapshotService, concurrencyService, billingService, rateLimitService, billingCacheService, identityService, httpUpstream, deferredService, claudeTokenProvider, sessionLimitCache, rpmCache, digestSessionStore, settingService, tlsFingerprintProfileService, channelService, modelPricingResolver)
 	openAITokenProvider := service.ProvideOpenAITokenProvider(accountRepository, geminiTokenCache, openAIOAuthService, oauthRefreshAPI)
-	openAIGatewayService := service.NewOpenAIGatewayService(accountRepository, usageLogRepository, usageBillingRepository, userRepository, userSubscriptionRepository, userGroupRateRepository, gatewayCache, configConfig, schedulerSnapshotService, concurrencyService, billingService, rateLimitService, billingCacheService, httpUpstream, deferredService, openAITokenProvider)
+	openAIGatewayService := service.NewOpenAIGatewayService(accountRepository, usageLogRepository, usageBillingRepository, userRepository, userSubscriptionRepository, userGroupRateRepository, gatewayCache, configConfig, schedulerSnapshotService, concurrencyService, billingService, rateLimitService, billingCacheService, httpUpstream, deferredService, openAITokenProvider, modelPricingResolver)
 	geminiMessagesCompatService := service.NewGeminiMessagesCompatService(accountRepository, groupRepository, gatewayCache, schedulerSnapshotService, geminiTokenProvider, rateLimitService, httpUpstream, antigravityGatewayService, configConfig)
 	opsSystemLogSink := service.ProvideOpsSystemLogSink(opsRepository)
 	opsService := service.NewOpsService(opsRepository, settingRepository, configConfig, accountRepository, userRepository, concurrencyService, gatewayService, openAIGatewayService, geminiMessagesCompatService, antigravityGatewayService, opsSystemLogSink)

backend/internal/handler/gateway_handler.go

@@ -164,14 +164,10 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 		channelMapping = h.gatewayService.ResolveChannelMapping(c.Request.Context(), *apiKey.GroupID, reqModel)
 	}
 
-	// 渠道模型限制检查
+	// 渠道模型限制检查：使用原始请求模型名，因为定价列表中注册的是用户请求的模型名
 	if apiKey.GroupID != nil {
-		checkModel := reqModel
-		if channelMapping.Mapped {
-			checkModel = channelMapping.MappedModel
-		}
-		if h.gatewayService.IsModelRestricted(c.Request.Context(), *apiKey.GroupID, checkModel) {
-			h.errorResponse(c, http.StatusForbidden, "invalid_request_error", "Model not available in current channel: "+reqModel)
+		if h.gatewayService.IsModelRestricted(c.Request.Context(), *apiKey.GroupID, reqModel) {
+			h.errorResponse(c, http.StatusServiceUnavailable, "api_error", "No available accounts")
 			return
 		}
 	}

backend/internal/handler/gateway_handler_warmup_intercept_unit_test.go

@@ -161,6 +161,8 @@ func newTestGatewayHandler(t *testing.T, group *service.Group, accounts []*servi
 		nil, // digestStore
 		nil, // settingService
 		nil, // tlsFPProfileService
+		nil, // channelService
+		nil, // resolver
 	)
 
 	// RunModeSimple：跳过计费检查，避免引入 repo/cache 依赖。

backend/internal/handler/sora_client_handler_test.go

@@ -2225,6 +2225,7 @@ func newMinimalGatewayService(accountRepo service.AccountRepository) *service.Ga
 	return service.NewGatewayService(
 		accountRepo, nil, nil, nil, nil, nil, nil, nil, nil,
 		nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil,
+		nil, nil,
 	)
 }
 

backend/internal/handler/sora_gateway_handler_test.go

@@ -465,6 +465,8 @@ func TestSoraGatewayHandler_ChatCompletions(t *testing.T) {
 		nil, // digestStore
 		nil, // settingService
 		nil, // tlsFPProfileService
+		nil, // channelService
+		nil, // resolver
 	)
 
 	soraClient := &stubSoraClient{imageURLs: []string{"https://example.com/a.png"}}

backend/internal/service/billing_service.go

@@ -104,6 +104,7 @@ type CostBreakdown struct {
 	CacheReadCost     float64
 	TotalCost         float64
 	ActualCost        float64 // 应用倍率后的实际费用
+	BillingMode       string  // 计费模式（"token"/"per_request"/"image"），由 CalculateCostUnified 填充
 }
 
 // BillingService 计费服务
@@ -439,12 +440,21 @@ func (s *BillingService) CalculateCostUnified(input CostInput) (*CostBreakdown,
 		input.RateMultiplier = 1.0
 	}
 
+	var breakdown *CostBreakdown
+	var err error
 	switch resolved.Mode {
 	case BillingModePerRequest, BillingModeImage:
-		return s.calculatePerRequestCost(resolved, input)
+		breakdown, err = s.calculatePerRequestCost(resolved, input)
 	default: // BillingModeToken
-		return s.calculateTokenCost(resolved, input)
+		breakdown, err = s.calculateTokenCost(resolved, input)
 	}
+	if err == nil && breakdown != nil {
+		breakdown.BillingMode = string(resolved.Mode)
+		if breakdown.BillingMode == "" {
+			breakdown.BillingMode = string(BillingModeToken)
+		}
+	}
+	return breakdown, err
 }
 
 // calculateTokenCost 按 token 区间计费

backend/internal/service/gateway_record_usage_test.go

@@ -42,6 +42,7 @@ func newGatewayRecordUsageServiceForTest(usageRepo UsageLogRepository, userRepo
 		nil,
 		nil,
 		nil,
+		nil,
 	)
 }
 

backend/internal/service/gateway_service.go

@@ -566,6 +566,7 @@ type GatewayService struct {
 	debugModelRouting     atomic.Bool
 	debugClaudeMimic      atomic.Bool
 	channelService        *ChannelService
+	resolver              *ModelPricingResolver
 	debugGatewayBodyFile  atomic.Pointer[os.File] // non-nil when SUB2API_DEBUG_GATEWAY_BODY is set
 	tlsFPProfileService   *TLSFingerprintProfileService
 }
@@ -596,6 +597,7 @@ func NewGatewayService(
 	settingService *SettingService,
 	tlsFPProfileService *TLSFingerprintProfileService,
 	channelService *ChannelService,
+	resolver *ModelPricingResolver,
 ) *GatewayService {
 	userGroupRateTTL := resolveUserGroupRateCacheTTL(cfg)
 	modelsListTTL := resolveModelsListCacheTTL(cfg)
@@ -629,6 +631,7 @@ func NewGatewayService(
 		responseHeaderFilter: compileResponseHeaderFilter(cfg),
 		tlsFPProfileService:  tlsFPProfileService,
 		channelService:       channelService,
+		resolver:             resolver,
 	}
 	svc.userGroupRateResolver = newUserGroupRateResolver(
 		userGroupRateRepo,
@@ -7703,13 +7706,21 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
 			CacheCreation1hTokens: result.Usage.CacheCreation1hTokens,
 		}
 		var err error
-		// 渠道定价覆盖
-		var chPricing *ChannelModelPricing
-		if s.channelService != nil && apiKey.Group != nil {
-			chPricing = s.channelService.GetChannelModelPricing(ctx, apiKey.Group.ID, billingModel)
-		}
-		if chPricing != nil {
-			cost, err = s.billingService.CalculateCostWithChannel(billingModel, tokens, multiplier, chPricing)
+		if s.resolver != nil && apiKey.Group != nil {
+			var groupID *int64
+			if apiKey.Group != nil {
+				gid := apiKey.Group.ID
+				groupID = &gid
+			}
+			cost, err = s.billingService.CalculateCostUnified(CostInput{
+				Ctx:            ctx,
+				Model:          billingModel,
+				GroupID:        groupID,
+				Tokens:         tokens,
+				RequestCount:   1,
+				RateMultiplier: multiplier,
+				Resolver:       s.resolver,
+			})
 		} else {
 			cost, err = s.billingService.CalculateCost(billingModel, tokens, multiplier)
 		}
@@ -7781,6 +7792,9 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
 		if result.ImageCount > 0 {
 			billingMode := "image"
 			usageLog.BillingMode = &billingMode
+		} else if cost != nil && cost.BillingMode != "" {
+			billingMode := cost.BillingMode
+			usageLog.BillingMode = &billingMode
 		} else {
 			billingMode := "token"
 			usageLog.BillingMode = &billingMode
@@ -7929,14 +7943,30 @@ func (s *GatewayService) RecordUsageWithLongContext(ctx context.Context, input *
 			CacheCreation1hTokens: result.Usage.CacheCreation1hTokens,
 		}
 		var err error
-		// 渠道定价覆盖
-		var chPricing2 *ChannelModelPricing
-		if s.channelService != nil && apiKey.Group != nil {
-			chPricing2 = s.channelService.GetChannelModelPricing(ctx, apiKey.Group.ID, billingModel)
+		// 优先尝试 Resolver + CalculateCostUnified（仅在有渠道定价时使用）
+		useUnified := false
+		if s.resolver != nil && apiKey.Group != nil {
+			gid := apiKey.Group.ID
+			resolved := s.resolver.Resolve(ctx, PricingInput{
+				Model:   billingModel,
+				GroupID: &gid,
+			})
+			if resolved.Source == "channel" {
+				// 有渠道定价，渠道区间已包含上下文分层
+				cost, err = s.billingService.CalculateCostUnified(CostInput{
+					Ctx:            ctx,
+					Model:          billingModel,
+					GroupID:        &gid,
+					Tokens:         tokens,
+					RequestCount:   1,
+					RateMultiplier: multiplier,
+					Resolver:       s.resolver,
+				})
+				useUnified = true
+			}
 		}
-		if chPricing2 != nil {
-			cost, err = s.billingService.CalculateCostWithChannel(billingModel, tokens, multiplier, chPricing2)
-		} else {
+		if !useUnified {
+			// 无渠道定价，保持原有长上下文双倍计费逻辑（如 Gemini 200K 阈值）
 			cost, err = s.billingService.CalculateCostWithLongContext(billingModel, tokens, multiplier, input.LongContextThreshold, input.LongContextMultiplier)
 		}
 		if err != nil {
@@ -8001,6 +8031,9 @@ func (s *GatewayService) RecordUsageWithLongContext(ctx context.Context, input *
 	if result.ImageCount > 0 {
 		billingMode := "image"
 		usageLog.BillingMode = &billingMode
+	} else if cost != nil && cost.BillingMode != "" {
+		billingMode := cost.BillingMode
+		usageLog.BillingMode = &billingMode
 	} else {
 		billingMode := "token"
 		usageLog.BillingMode = &billingMode

backend/internal/service/openai_gateway_record_usage_test.go

@@ -145,6 +145,7 @@ func newOpenAIRecordUsageServiceForTest(usageRepo UsageLogRepository, userRepo U
 		nil,
 		&DeferredService{},
 		nil,
+		nil,
 	)
 	svc.userGroupRateResolver = newUserGroupRateResolver(
 		rateRepo,

backend/internal/service/openai_gateway_service.go

@@ -322,6 +322,7 @@ type OpenAIGatewayService struct {
 	openAITokenProvider   *OpenAITokenProvider
 	toolCorrector         *CodexToolCorrector
 	openaiWSResolver      OpenAIWSProtocolResolver
+	resolver              *ModelPricingResolver
 
 	openaiWSPoolOnce              sync.Once
 	openaiWSStateStoreOnce        sync.Once
@@ -357,6 +358,7 @@ func NewOpenAIGatewayService(
 	httpUpstream HTTPUpstream,
 	deferredService *DeferredService,
 	openAITokenProvider *OpenAITokenProvider,
+	resolver *ModelPricingResolver,
 ) *OpenAIGatewayService {
 	svc := &OpenAIGatewayService{
 		accountRepo:         accountRepo,
@@ -384,6 +386,7 @@ func NewOpenAIGatewayService(
 		openAITokenProvider:   openAITokenProvider,
 		toolCorrector:         NewCodexToolCorrector(),
 		openaiWSResolver:      NewOpenAIWSProtocolResolver(cfg),
+		resolver:              resolver,
 		responseHeaderFilter:  compileResponseHeaderFilter(cfg),
 		codexSnapshotThrottle: newAccountWriteThrottle(openAICodexSnapshotPersistMinInterval),
 	}
@@ -4152,6 +4155,8 @@ func (s *OpenAIGatewayService) RecordUsage(ctx context.Context, input *OpenAIRec
 		multiplier = resolver.Resolve(ctx, user.ID, *apiKey.GroupID, apiKey.Group.RateMultiplier)
 	}
 
+	var cost *CostBreakdown
+	var err error
 	billingModel := forwardResultBillingModel(result.Model, result.UpstreamModel)
 	if result.BillingModel != "" {
 		billingModel = strings.TrimSpace(result.BillingModel)
@@ -4160,7 +4165,21 @@ func (s *OpenAIGatewayService) RecordUsage(ctx context.Context, input *OpenAIRec
 	if result.ServiceTier != nil {
 		serviceTier = strings.TrimSpace(*result.ServiceTier)
 	}
-	cost, err := s.billingService.CalculateCostWithServiceTier(billingModel, tokens, multiplier, serviceTier)
+	if s.resolver != nil && apiKey.Group != nil {
+		gid := apiKey.Group.ID
+		cost, err = s.billingService.CalculateCostUnified(CostInput{
+			Ctx:            ctx,
+			Model:          billingModel,
+			GroupID:        &gid,
+			Tokens:         tokens,
+			RequestCount:   1,
+			RateMultiplier: multiplier,
+			ServiceTier:    serviceTier,
+			Resolver:       s.resolver,
+		})
+	} else {
+		cost, err = s.billingService.CalculateCostWithServiceTier(billingModel, tokens, multiplier, serviceTier)
+	}
 	if err != nil {
 		cost = &CostBreakdown{ActualCost: 0}
 	}
@@ -4207,8 +4226,11 @@ func (s *OpenAIGatewayService) RecordUsage(ctx context.Context, input *OpenAIRec
 		FirstTokenMs:          result.FirstTokenMs,
 		CreatedAt:             time.Now(),
 	}
-	// 设置计费模式（OpenAI 网关都是 token 计费）
-	{
+	// 设置计费模式
+	if cost != nil && cost.BillingMode != "" {
+		billingMode := cost.BillingMode
+		usageLog.BillingMode = &billingMode
+	} else {
 		billingMode := "token"
 		usageLog.BillingMode = &billingMode
 	}

backend/internal/service/openai_ws_protocol_forward_test.go

@@ -615,6 +615,7 @@ func TestNewOpenAIGatewayService_InitializesOpenAIWSResolver(t *testing.T) {
 		nil,
 		nil,
 		nil,
+		nil,
 	)
 
 	decision := svc.getOpenAIWSProtocolResolver().Resolve(nil)