Merge pull request #3400 from seefs001/fix/openai-usage

refactor: optimize billing flow for OpenAI-to-Anthropic convert

dto/openai_response.go

@@ -220,10 +220,12 @@ type CompletionsStreamResponse struct {
 }
 
 type Usage struct {
-	PromptTokens         int `json:"prompt_tokens"`
+	PromptTokens         int    `json:"prompt_tokens"`
-	CompletionTokens     int `json:"completion_tokens"`
+	CompletionTokens     int    `json:"completion_tokens"`
-	TotalTokens          int `json:"total_tokens"`
+	TotalTokens          int    `json:"total_tokens"`
-	PromptCacheHitTokens int `json:"prompt_cache_hit_tokens,omitempty"`
+	PromptCacheHitTokens int    `json:"prompt_cache_hit_tokens,omitempty"`
+	UsageSemantic        string `json:"usage_semantic,omitempty"`
+	UsageSource          string `json:"usage_source,omitempty"`
 
 	PromptTokensDetails    InputTokenDetails  `json:"prompt_tokens_details"`
 	CompletionTokenDetails OutputTokenDetails `json:"completion_tokens_details"`
@@ -251,7 +253,7 @@ type OpenAIVideoResponse struct {
 
 type InputTokenDetails struct {
 	CachedTokens         int `json:"cached_tokens"`
-	CachedCreationTokens int `json:"-"`
+	CachedCreationTokens int `json:"cached_creation_tokens,omitempty"`
 	TextTokens           int `json:"text_tokens"`
 	AudioTokens          int `json:"audio_tokens"`
 	ImageTokens          int `json:"image_tokens"`

relay/audio_handler.go

@@ -70,7 +70,7 @@ func AudioHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *type
 	if usage.(*dto.Usage).CompletionTokenDetails.AudioTokens > 0 || usage.(*dto.Usage).PromptTokensDetails.AudioTokens > 0 {
 		service.PostAudioConsumeQuota(c, info, usage.(*dto.Usage), "")
 	} else {
-		postConsumeQuota(c, info, usage.(*dto.Usage))
+		service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
 	}
 
 	return nil

relay/channel/claude/relay-claude.go

@@ -555,6 +555,35 @@ type ClaudeResponseInfo struct {
 	Done         bool
 }
 
+func cacheCreationTokensForOpenAIUsage(usage *dto.Usage) int {
+	if usage == nil {
+		return 0
+	}
+	splitCacheCreationTokens := usage.ClaudeCacheCreation5mTokens + usage.ClaudeCacheCreation1hTokens
+	if splitCacheCreationTokens == 0 {
+		return usage.PromptTokensDetails.CachedCreationTokens
+	}
+	if usage.PromptTokensDetails.CachedCreationTokens > splitCacheCreationTokens {
+		return usage.PromptTokensDetails.CachedCreationTokens
+	}
+	return splitCacheCreationTokens
+}
+
+func buildOpenAIStyleUsageFromClaudeUsage(usage *dto.Usage) dto.Usage {
+	if usage == nil {
+		return dto.Usage{}
+	}
+	clone := *usage
+	cacheCreationTokens := cacheCreationTokensForOpenAIUsage(usage)
+	totalInputTokens := usage.PromptTokens + usage.PromptTokensDetails.CachedTokens + cacheCreationTokens
+	clone.PromptTokens = totalInputTokens
+	clone.InputTokens = totalInputTokens
+	clone.TotalTokens = totalInputTokens + usage.CompletionTokens
+	clone.UsageSemantic = "openai"
+	clone.UsageSource = "anthropic"
+	return clone
+}
+
 func buildMessageDeltaPatchUsage(claudeResponse *dto.ClaudeResponse, claudeInfo *ClaudeResponseInfo) *dto.ClaudeUsage {
 	usage := &dto.ClaudeUsage{}
 	if claudeResponse != nil && claudeResponse.Usage != nil {
@@ -643,6 +672,7 @@ func FormatClaudeResponseInfo(claudeResponse *dto.ClaudeResponse, oaiResponse *d
 		// message_start, 获取usage
 		if claudeResponse.Message != nil && claudeResponse.Message.Usage != nil {
 			claudeInfo.Usage.PromptTokens = claudeResponse.Message.Usage.InputTokens
+			claudeInfo.Usage.UsageSemantic = "anthropic"
 			claudeInfo.Usage.PromptTokensDetails.CachedTokens = claudeResponse.Message.Usage.CacheReadInputTokens
 			claudeInfo.Usage.PromptTokensDetails.CachedCreationTokens = claudeResponse.Message.Usage.CacheCreationInputTokens
 			claudeInfo.Usage.ClaudeCacheCreation5mTokens = claudeResponse.Message.Usage.GetCacheCreation5mTokens()
@@ -661,6 +691,7 @@ func FormatClaudeResponseInfo(claudeResponse *dto.ClaudeResponse, oaiResponse *d
 	} else if claudeResponse.Type == "message_delta" {
 		// 最终的usage获取
 		if claudeResponse.Usage != nil {
+			claudeInfo.Usage.UsageSemantic = "anthropic"
 			if claudeResponse.Usage.InputTokens > 0 {
 				// 不叠加，只取最新的
 				claudeInfo.Usage.PromptTokens = claudeResponse.Usage.InputTokens
@@ -754,12 +785,16 @@ func HandleStreamFinalResponse(c *gin.Context, info *relaycommon.RelayInfo, clau
 		}
 		claudeInfo.Usage = service.ResponseText2Usage(c, claudeInfo.ResponseText.String(), info.UpstreamModelName, claudeInfo.Usage.PromptTokens)
 	}
+	if claudeInfo.Usage != nil {
+		claudeInfo.Usage.UsageSemantic = "anthropic"
+	}
 
 	if info.RelayFormat == types.RelayFormatClaude {
 		//
 	} else if info.RelayFormat == types.RelayFormatOpenAI {
 		if info.ShouldIncludeUsage {
-			response := helper.GenerateFinalUsageResponse(claudeInfo.ResponseId, claudeInfo.Created, info.UpstreamModelName, *claudeInfo.Usage)
+			openAIUsage := buildOpenAIStyleUsageFromClaudeUsage(claudeInfo.Usage)
+			response := helper.GenerateFinalUsageResponse(claudeInfo.ResponseId, claudeInfo.Created, info.UpstreamModelName, openAIUsage)
 			err := helper.ObjectData(c, response)
 			if err != nil {
 				common.SysLog("send final response failed: " + err.Error())
@@ -810,6 +845,7 @@ func HandleClaudeResponseData(c *gin.Context, info *relaycommon.RelayInfo, claud
 		claudeInfo.Usage.PromptTokens = claudeResponse.Usage.InputTokens
 		claudeInfo.Usage.CompletionTokens = claudeResponse.Usage.OutputTokens
 		claudeInfo.Usage.TotalTokens = claudeResponse.Usage.InputTokens + claudeResponse.Usage.OutputTokens
+		claudeInfo.Usage.UsageSemantic = "anthropic"
 		claudeInfo.Usage.PromptTokensDetails.CachedTokens = claudeResponse.Usage.CacheReadInputTokens
 		claudeInfo.Usage.PromptTokensDetails.CachedCreationTokens = claudeResponse.Usage.CacheCreationInputTokens
 		claudeInfo.Usage.ClaudeCacheCreation5mTokens = claudeResponse.Usage.GetCacheCreation5mTokens()
@@ -819,7 +855,7 @@ func HandleClaudeResponseData(c *gin.Context, info *relaycommon.RelayInfo, claud
 	switch info.RelayFormat {
 	case types.RelayFormatOpenAI:
 		openaiResponse := ResponseClaude2OpenAI(&claudeResponse)
-		openaiResponse.Usage = *claudeInfo.Usage
+		openaiResponse.Usage = buildOpenAIStyleUsageFromClaudeUsage(claudeInfo.Usage)
 		responseData, err = json.Marshal(openaiResponse)
 		if err != nil {
 			return types.NewError(err, types.ErrorCodeBadResponseBody)

relay/channel/claude/relay_claude_test.go

@@ -173,3 +173,85 @@ func TestFormatClaudeResponseInfo_ContentBlockDelta(t *testing.T) {
 		t.Errorf("ResponseText = %q, want %q", claudeInfo.ResponseText.String(), "hello")
 	}
 }
+
+func TestBuildOpenAIStyleUsageFromClaudeUsage(t *testing.T) {
+	usage := &dto.Usage{
+		PromptTokens:     100,
+		CompletionTokens: 20,
+		PromptTokensDetails: dto.InputTokenDetails{
+			CachedTokens:         30,
+			CachedCreationTokens: 50,
+		},
+		ClaudeCacheCreation5mTokens: 10,
+		ClaudeCacheCreation1hTokens: 20,
+		UsageSemantic:               "anthropic",
+	}
+
+	openAIUsage := buildOpenAIStyleUsageFromClaudeUsage(usage)
+
+	if openAIUsage.PromptTokens != 180 {
+		t.Fatalf("PromptTokens = %d, want 180", openAIUsage.PromptTokens)
+	}
+	if openAIUsage.InputTokens != 180 {
+		t.Fatalf("InputTokens = %d, want 180", openAIUsage.InputTokens)
+	}
+	if openAIUsage.TotalTokens != 200 {
+		t.Fatalf("TotalTokens = %d, want 200", openAIUsage.TotalTokens)
+	}
+	if openAIUsage.UsageSemantic != "openai" {
+		t.Fatalf("UsageSemantic = %s, want openai", openAIUsage.UsageSemantic)
+	}
+	if openAIUsage.UsageSource != "anthropic" {
+		t.Fatalf("UsageSource = %s, want anthropic", openAIUsage.UsageSource)
+	}
+}
+
+func TestBuildOpenAIStyleUsageFromClaudeUsagePreservesCacheCreationRemainder(t *testing.T) {
+	tests := []struct {
+		name                    string
+		cachedCreationTokens    int
+		cacheCreationTokens5m   int
+		cacheCreationTokens1h   int
+		expectedTotalInputToken int
+	}{
+		{
+			name:                    "prefers aggregate when it includes remainder",
+			cachedCreationTokens:    50,
+			cacheCreationTokens5m:   10,
+			cacheCreationTokens1h:   20,
+			expectedTotalInputToken: 180,
+		},
+		{
+			name:                    "falls back to split tokens when aggregate missing",
+			cachedCreationTokens:    0,
+			cacheCreationTokens5m:   10,
+			cacheCreationTokens1h:   20,
+			expectedTotalInputToken: 160,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			usage := &dto.Usage{
+				PromptTokens:     100,
+				CompletionTokens: 20,
+				PromptTokensDetails: dto.InputTokenDetails{
+					CachedTokens:         30,
+					CachedCreationTokens: tt.cachedCreationTokens,
+				},
+				ClaudeCacheCreation5mTokens: tt.cacheCreationTokens5m,
+				ClaudeCacheCreation1hTokens: tt.cacheCreationTokens1h,
+				UsageSemantic:               "anthropic",
+			}
+
+			openAIUsage := buildOpenAIStyleUsageFromClaudeUsage(usage)
+
+			if openAIUsage.PromptTokens != tt.expectedTotalInputToken {
+				t.Fatalf("PromptTokens = %d, want %d", openAIUsage.PromptTokens, tt.expectedTotalInputToken)
+			}
+			if openAIUsage.InputTokens != tt.expectedTotalInputToken {
+				t.Fatalf("InputTokens = %d, want %d", openAIUsage.InputTokens, tt.expectedTotalInputToken)
+			}
+		})
+	}
+}

relay/claude_handler.go

@@ -122,7 +122,7 @@ func ClaudeHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ
 			return newApiErr
 		}
 
-		service.PostClaudeConsumeQuota(c, info, usage)
+		service.PostTextConsumeQuota(c, info, usage, nil)
 		return nil
 	}
 
@@ -190,6 +190,6 @@ func ClaudeHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ
 		return newAPIError
 	}
 
-	service.PostClaudeConsumeQuota(c, info, usage.(*dto.Usage))
+	service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
 	return nil
 }

relay/compatible_handler.go

@@ -6,25 +6,20 @@ import (
 	"io"
 	"net/http"
 	"strings"
-	"time"
 
 	"github.com/QuantumNous/new-api/common"
 	"github.com/QuantumNous/new-api/constant"
 	"github.com/QuantumNous/new-api/dto"
 	"github.com/QuantumNous/new-api/logger"
-	"github.com/QuantumNous/new-api/model"
 	relaycommon "github.com/QuantumNous/new-api/relay/common"
 	relayconstant "github.com/QuantumNous/new-api/relay/constant"
 	"github.com/QuantumNous/new-api/relay/helper"
 	"github.com/QuantumNous/new-api/service"
 	"github.com/QuantumNous/new-api/setting/model_setting"
-	"github.com/QuantumNous/new-api/setting/operation_setting"
 	"github.com/QuantumNous/new-api/setting/ratio_setting"
 	"github.com/QuantumNous/new-api/types"
 	"github.com/samber/lo"
 
-	"github.com/shopspring/decimal"
-
 	"github.com/gin-gonic/gin"
 )
 
@@ -93,7 +88,7 @@ func TextHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *types
 		if containAudioTokens && containsAudioRatios {
 			service.PostAudioConsumeQuota(c, info, usage, "")
 		} else {
-			postConsumeQuota(c, info, usage)
+			service.PostTextConsumeQuota(c, info, usage, nil)
 		}
 		return nil
 	}
@@ -216,293 +211,7 @@ func TextHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *types
 	if containAudioTokens && containsAudioRatios {
 		service.PostAudioConsumeQuota(c, info, usage.(*dto.Usage), "")
 	} else {
-		postConsumeQuota(c, info, usage.(*dto.Usage))
+		service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
 	}
 	return nil
 }
-
-func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usage *dto.Usage, extraContent ...string) {
-	originUsage := usage
-	if usage == nil {
-		usage = &dto.Usage{
-			PromptTokens:     relayInfo.GetEstimatePromptTokens(),
-			CompletionTokens: 0,
-			TotalTokens:      relayInfo.GetEstimatePromptTokens(),
-		}
-		extraContent = append(extraContent, "上游无计费信息")
-	}
-
-	if originUsage != nil {
-		service.ObserveChannelAffinityUsageCacheByRelayFormat(ctx, usage, relayInfo.GetFinalRequestRelayFormat())
-	}
-
-	adminRejectReason := common.GetContextKeyString(ctx, constant.ContextKeyAdminRejectReason)
-
-	useTimeSeconds := time.Now().Unix() - relayInfo.StartTime.Unix()
-	promptTokens := usage.PromptTokens
-	cacheTokens := usage.PromptTokensDetails.CachedTokens
-	imageTokens := usage.PromptTokensDetails.ImageTokens
-	audioTokens := usage.PromptTokensDetails.AudioTokens
-	completionTokens := usage.CompletionTokens
-	cachedCreationTokens := usage.PromptTokensDetails.CachedCreationTokens
-
-	modelName := relayInfo.OriginModelName
-
-	tokenName := ctx.GetString("token_name")
-	completionRatio := relayInfo.PriceData.CompletionRatio
-	cacheRatio := relayInfo.PriceData.CacheRatio
-	imageRatio := relayInfo.PriceData.ImageRatio
-	modelRatio := relayInfo.PriceData.ModelRatio
-	groupRatio := relayInfo.PriceData.GroupRatioInfo.GroupRatio
-	modelPrice := relayInfo.PriceData.ModelPrice
-	cachedCreationRatio := relayInfo.PriceData.CacheCreationRatio
-
-	// Convert values to decimal for precise calculation
-	dPromptTokens := decimal.NewFromInt(int64(promptTokens))
-	dCacheTokens := decimal.NewFromInt(int64(cacheTokens))
-	dImageTokens := decimal.NewFromInt(int64(imageTokens))
-	dAudioTokens := decimal.NewFromInt(int64(audioTokens))
-	dCompletionTokens := decimal.NewFromInt(int64(completionTokens))
-	dCachedCreationTokens := decimal.NewFromInt(int64(cachedCreationTokens))
-	dCompletionRatio := decimal.NewFromFloat(completionRatio)
-	dCacheRatio := decimal.NewFromFloat(cacheRatio)
-	dImageRatio := decimal.NewFromFloat(imageRatio)
-	dModelRatio := decimal.NewFromFloat(modelRatio)
-	dGroupRatio := decimal.NewFromFloat(groupRatio)
-	dModelPrice := decimal.NewFromFloat(modelPrice)
-	dCachedCreationRatio := decimal.NewFromFloat(cachedCreationRatio)
-	dQuotaPerUnit := decimal.NewFromFloat(common.QuotaPerUnit)
-
-	ratio := dModelRatio.Mul(dGroupRatio)
-
-	// openai web search 工具计费
-	var dWebSearchQuota decimal.Decimal
-	var webSearchPrice float64
-	// response api 格式工具计费
-	if relayInfo.ResponsesUsageInfo != nil {
-		if webSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolWebSearchPreview]; exists && webSearchTool.CallCount > 0 {
-			// 计算 web search 调用的配额 (配额 = 价格 * 调用次数 / 1000 * 分组倍率)
-			webSearchPrice = operation_setting.GetWebSearchPricePerThousand(modelName, webSearchTool.SearchContextSize)
-			dWebSearchQuota = decimal.NewFromFloat(webSearchPrice).
-				Mul(decimal.NewFromInt(int64(webSearchTool.CallCount))).
-				Div(decimal.NewFromInt(1000)).Mul(dGroupRatio).Mul(dQuotaPerUnit)
-			extraContent = append(extraContent, fmt.Sprintf("Web Search 调用 %d 次，上下文大小 %s，调用花费 %s",
-				webSearchTool.CallCount, webSearchTool.SearchContextSize, dWebSearchQuota.String()))
-		}
-	} else if strings.HasSuffix(modelName, "search-preview") {
-		// search-preview 模型不支持 response api
-		searchContextSize := ctx.GetString("chat_completion_web_search_context_size")
-		if searchContextSize == "" {
-			searchContextSize = "medium"
-		}
-		webSearchPrice = operation_setting.GetWebSearchPricePerThousand(modelName, searchContextSize)
-		dWebSearchQuota = decimal.NewFromFloat(webSearchPrice).
-			Div(decimal.NewFromInt(1000)).Mul(dGroupRatio).Mul(dQuotaPerUnit)
-		extraContent = append(extraContent, fmt.Sprintf("Web Search 调用 1 次，上下文大小 %s，调用花费 %s",
-			searchContextSize, dWebSearchQuota.String()))
-	}
-	// claude web search tool 计费
-	var dClaudeWebSearchQuota decimal.Decimal
-	var claudeWebSearchPrice float64
-	claudeWebSearchCallCount := ctx.GetInt("claude_web_search_requests")
-	if claudeWebSearchCallCount > 0 {
-		claudeWebSearchPrice = operation_setting.GetClaudeWebSearchPricePerThousand()
-		dClaudeWebSearchQuota = decimal.NewFromFloat(claudeWebSearchPrice).
-			Div(decimal.NewFromInt(1000)).Mul(dGroupRatio).Mul(dQuotaPerUnit).Mul(decimal.NewFromInt(int64(claudeWebSearchCallCount)))
-		extraContent = append(extraContent, fmt.Sprintf("Claude Web Search 调用 %d 次，调用花费 %s",
-			claudeWebSearchCallCount, dClaudeWebSearchQuota.String()))
-	}
-	// file search tool 计费
-	var dFileSearchQuota decimal.Decimal
-	var fileSearchPrice float64
-	if relayInfo.ResponsesUsageInfo != nil {
-		if fileSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolFileSearch]; exists && fileSearchTool.CallCount > 0 {
-			fileSearchPrice = operation_setting.GetFileSearchPricePerThousand()
-			dFileSearchQuota = decimal.NewFromFloat(fileSearchPrice).
-				Mul(decimal.NewFromInt(int64(fileSearchTool.CallCount))).
-				Div(decimal.NewFromInt(1000)).Mul(dGroupRatio).Mul(dQuotaPerUnit)
-			extraContent = append(extraContent, fmt.Sprintf("File Search 调用 %d 次，调用花费 %s",
-				fileSearchTool.CallCount, dFileSearchQuota.String()))
-		}
-	}
-	var dImageGenerationCallQuota decimal.Decimal
-	var imageGenerationCallPrice float64
-	if ctx.GetBool("image_generation_call") {
-		imageGenerationCallPrice = operation_setting.GetGPTImage1PriceOnceCall(ctx.GetString("image_generation_call_quality"), ctx.GetString("image_generation_call_size"))
-		dImageGenerationCallQuota = decimal.NewFromFloat(imageGenerationCallPrice).Mul(dGroupRatio).Mul(dQuotaPerUnit)
-		extraContent = append(extraContent, fmt.Sprintf("Image Generation Call 花费 %s", dImageGenerationCallQuota.String()))
-	}
-
-	var quotaCalculateDecimal decimal.Decimal
-
-	var audioInputQuota decimal.Decimal
-	var audioInputPrice float64
-	isClaudeUsageSemantic := relayInfo.GetFinalRequestRelayFormat() == types.RelayFormatClaude
-	if !relayInfo.PriceData.UsePrice {
-		baseTokens := dPromptTokens
-		// 减去 cached tokens
-		// Anthropic API 的 input_tokens 已经不包含缓存 tokens，不需要减去
-		// OpenAI/OpenRouter 等 API 的 prompt_tokens 包含缓存 tokens，需要减去
-		var cachedTokensWithRatio decimal.Decimal
-		if !dCacheTokens.IsZero() {
-			if !isClaudeUsageSemantic {
-				baseTokens = baseTokens.Sub(dCacheTokens)
-			}
-			cachedTokensWithRatio = dCacheTokens.Mul(dCacheRatio)
-		}
-		var dCachedCreationTokensWithRatio decimal.Decimal
-		if !dCachedCreationTokens.IsZero() {
-			if !isClaudeUsageSemantic {
-				baseTokens = baseTokens.Sub(dCachedCreationTokens)
-			}
-			dCachedCreationTokensWithRatio = dCachedCreationTokens.Mul(dCachedCreationRatio)
-		}
-
-		// 减去 image tokens
-		var imageTokensWithRatio decimal.Decimal
-		if !dImageTokens.IsZero() {
-			baseTokens = baseTokens.Sub(dImageTokens)
-			imageTokensWithRatio = dImageTokens.Mul(dImageRatio)
-		}
-
-		// 减去 Gemini audio tokens
-		if !dAudioTokens.IsZero() {
-			audioInputPrice = operation_setting.GetGeminiInputAudioPricePerMillionTokens(modelName)
-			if audioInputPrice > 0 {
-				// 重新计算 base tokens
-				baseTokens = baseTokens.Sub(dAudioTokens)
-				audioInputQuota = decimal.NewFromFloat(audioInputPrice).Div(decimal.NewFromInt(1000000)).Mul(dAudioTokens).Mul(dGroupRatio).Mul(dQuotaPerUnit)
-				extraContent = append(extraContent, fmt.Sprintf("Audio Input 花费 %s", audioInputQuota.String()))
-			}
-		}
-		promptQuota := baseTokens.Add(cachedTokensWithRatio).
-			Add(imageTokensWithRatio).
-			Add(dCachedCreationTokensWithRatio)
-
-		completionQuota := dCompletionTokens.Mul(dCompletionRatio)
-
-		quotaCalculateDecimal = promptQuota.Add(completionQuota).Mul(ratio)
-
-		if !ratio.IsZero() && quotaCalculateDecimal.LessThanOrEqual(decimal.Zero) {
-			quotaCalculateDecimal = decimal.NewFromInt(1)
-		}
-	} else {
-		quotaCalculateDecimal = dModelPrice.Mul(dQuotaPerUnit).Mul(dGroupRatio)
-	}
-	// 添加 responses tools call 调用的配额
-	quotaCalculateDecimal = quotaCalculateDecimal.Add(dWebSearchQuota)
-	quotaCalculateDecimal = quotaCalculateDecimal.Add(dFileSearchQuota)
-	// 添加 audio input 独立计费
-	quotaCalculateDecimal = quotaCalculateDecimal.Add(audioInputQuota)
-	// 添加 image generation call 计费
-	quotaCalculateDecimal = quotaCalculateDecimal.Add(dImageGenerationCallQuota)
-
-	if len(relayInfo.PriceData.OtherRatios) > 0 {
-		for key, otherRatio := range relayInfo.PriceData.OtherRatios {
-			dOtherRatio := decimal.NewFromFloat(otherRatio)
-			quotaCalculateDecimal = quotaCalculateDecimal.Mul(dOtherRatio)
-			extraContent = append(extraContent, fmt.Sprintf("其他倍率 %s: %f", key, otherRatio))
-		}
-	}
-
-	quota := int(quotaCalculateDecimal.Round(0).IntPart())
-	totalTokens := promptTokens + completionTokens
-
-	//var logContent string
-
-	// record all the consume log even if quota is 0
-	if totalTokens == 0 {
-		// in this case, must be some error happened
-		// we cannot just return, because we may have to return the pre-consumed quota
-		quota = 0
-		extraContent = append(extraContent, "上游没有返回计费信息，无法扣费（可能是上游超时）")
-		logger.LogError(ctx, fmt.Sprintf("total tokens is 0, cannot consume quota, userId %d, channelId %d, "+
-			"tokenId %d, model %s， pre-consumed quota %d", relayInfo.UserId, relayInfo.ChannelId, relayInfo.TokenId, modelName, relayInfo.FinalPreConsumedQuota))
-	} else {
-		if !ratio.IsZero() && quota == 0 {
-			quota = 1
-		}
-		model.UpdateUserUsedQuotaAndRequestCount(relayInfo.UserId, quota)
-		model.UpdateChannelUsedQuota(relayInfo.ChannelId, quota)
-	}
-
-	if err := service.SettleBilling(ctx, relayInfo, quota); err != nil {
-		logger.LogError(ctx, "error settling billing: "+err.Error())
-	}
-
-	logModel := modelName
-	if strings.HasPrefix(logModel, "gpt-4-gizmo") {
-		logModel = "gpt-4-gizmo-*"
-		extraContent = append(extraContent, fmt.Sprintf("模型 %s", modelName))
-	}
-	if strings.HasPrefix(logModel, "gpt-4o-gizmo") {
-		logModel = "gpt-4o-gizmo-*"
-		extraContent = append(extraContent, fmt.Sprintf("模型 %s", modelName))
-	}
-	logContent := strings.Join(extraContent, ", ")
-	other := service.GenerateTextOtherInfo(ctx, relayInfo, modelRatio, groupRatio, completionRatio, cacheTokens, cacheRatio, modelPrice, relayInfo.PriceData.GroupRatioInfo.GroupSpecialRatio)
-	if adminRejectReason != "" {
-		other["reject_reason"] = adminRejectReason
-	}
-	// For chat-based calls to the Claude model, tagging is required. Using Claude's rendering logs, the two approaches handle input rendering differently.
-	if isClaudeUsageSemantic {
-		other["claude"] = true
-		other["usage_semantic"] = "anthropic"
-	}
-	if imageTokens != 0 {
-		other["image"] = true
-		other["image_ratio"] = imageRatio
-		other["image_output"] = imageTokens
-	}
-	if cachedCreationTokens != 0 {
-		other["cache_creation_tokens"] = cachedCreationTokens
-		other["cache_creation_ratio"] = cachedCreationRatio
-	}
-	if !dWebSearchQuota.IsZero() {
-		if relayInfo.ResponsesUsageInfo != nil {
-			if webSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolWebSearchPreview]; exists {
-				other["web_search"] = true
-				other["web_search_call_count"] = webSearchTool.CallCount
-				other["web_search_price"] = webSearchPrice
-			}
-		} else if strings.HasSuffix(modelName, "search-preview") {
-			other["web_search"] = true
-			other["web_search_call_count"] = 1
-			other["web_search_price"] = webSearchPrice
-		}
-	} else if !dClaudeWebSearchQuota.IsZero() {
-		other["web_search"] = true
-		other["web_search_call_count"] = claudeWebSearchCallCount
-		other["web_search_price"] = claudeWebSearchPrice
-	}
-	if !dFileSearchQuota.IsZero() && relayInfo.ResponsesUsageInfo != nil {
-		if fileSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolFileSearch]; exists {
-			other["file_search"] = true
-			other["file_search_call_count"] = fileSearchTool.CallCount
-			other["file_search_price"] = fileSearchPrice
-		}
-	}
-	if !audioInputQuota.IsZero() {
-		other["audio_input_seperate_price"] = true
-		other["audio_input_token_count"] = audioTokens
-		other["audio_input_price"] = audioInputPrice
-	}
-	if !dImageGenerationCallQuota.IsZero() {
-		other["image_generation_call"] = true
-		other["image_generation_call_price"] = imageGenerationCallPrice
-	}
-	model.RecordConsumeLog(ctx, relayInfo.UserId, model.RecordConsumeLogParams{
-		ChannelId:        relayInfo.ChannelId,
-		PromptTokens:     promptTokens,
-		CompletionTokens: completionTokens,
-		ModelName:        logModel,
-		TokenName:        tokenName,
-		Quota:            quota,
-		Content:          logContent,
-		TokenId:          relayInfo.TokenId,
-		UseTimeSeconds:   int(useTimeSeconds),
-		IsStream:         relayInfo.IsStream,
-		Group:            relayInfo.UsingGroup,
-		Other:            other,
-	})
-}

relay/embedding_handler.go

@@ -82,6 +82,6 @@ func EmbeddingHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *
 		service.ResetStatusCode(newAPIError, statusCodeMappingStr)
 		return newAPIError
 	}
-	postConsumeQuota(c, info, usage.(*dto.Usage))
+	service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
 	return nil
 }

relay/gemini_handler.go

@@ -194,7 +194,7 @@ func GeminiHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ
 		return openaiErr
 	}
 
-	postConsumeQuota(c, info, usage.(*dto.Usage))
+	service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
 	return nil
 }
 
@@ -288,6 +288,6 @@ func GeminiEmbeddingHandler(c *gin.Context, info *relaycommon.RelayInfo) (newAPI
 		return openaiErr
 	}
 
-	postConsumeQuota(c, info, usage.(*dto.Usage))
+	service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
 	return nil
 }

relay/image_handler.go

@@ -141,6 +141,6 @@ func ImageHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *type
 		logContent = append(logContent, fmt.Sprintf("生成数量 %d", imageN))
 	}
 
-	postConsumeQuota(c, info, usage.(*dto.Usage), logContent...)
+	service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), logContent)
 	return nil
 }

relay/rerank_handler.go

@@ -96,6 +96,6 @@ func RerankHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ
 		service.ResetStatusCode(newAPIError, statusCodeMappingStr)
 		return newAPIError
 	}
-	postConsumeQuota(c, info, usage.(*dto.Usage))
+	service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
 	return nil
 }

relay/responses_handler.go

@@ -145,7 +145,7 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *
 			info.PriceData = originPriceData
 			return types.NewError(err, types.ErrorCodeModelPriceError, types.ErrOptionWithSkipRetry())
 		}
-		postConsumeQuota(c, info, usageDto)
+		service.PostTextConsumeQuota(c, info, usageDto, nil)
 
 		info.OriginModelName = originModelName
 		info.PriceData = originPriceData
@@ -155,7 +155,7 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *
 	if strings.HasPrefix(info.OriginModelName, "gpt-4o-audio") {
 		service.PostAudioConsumeQuota(c, info, usageDto, "")
 	} else {
-		postConsumeQuota(c, info, usageDto)
+		service.PostTextConsumeQuota(c, info, usageDto, nil)
 	}
 	return nil
 }

service/convert.go

@@ -223,6 +223,25 @@ func generateStopBlock(index int) *dto.ClaudeResponse {
 	}
 }
 
+func buildClaudeUsageFromOpenAIUsage(oaiUsage *dto.Usage) *dto.ClaudeUsage {
+	if oaiUsage == nil {
+		return nil
+	}
+	usage := &dto.ClaudeUsage{
+		InputTokens:              oaiUsage.PromptTokens,
+		OutputTokens:             oaiUsage.CompletionTokens,
+		CacheCreationInputTokens: oaiUsage.PromptTokensDetails.CachedCreationTokens,
+		CacheReadInputTokens:     oaiUsage.PromptTokensDetails.CachedTokens,
+	}
+	if oaiUsage.ClaudeCacheCreation5mTokens > 0 || oaiUsage.ClaudeCacheCreation1hTokens > 0 {
+		usage.CacheCreation = &dto.ClaudeCacheCreationUsage{
+			Ephemeral5mInputTokens: oaiUsage.ClaudeCacheCreation5mTokens,
+			Ephemeral1hInputTokens: oaiUsage.ClaudeCacheCreation1hTokens,
+		}
+	}
+	return usage
+}
+
 func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamResponse, info *relaycommon.RelayInfo) []*dto.ClaudeResponse {
 	if info.ClaudeConvertInfo.Done {
 		return nil
@@ -391,13 +410,8 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon
 			}
 			if oaiUsage != nil {
 				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
-					Type: "message_delta",
+					Type:  "message_delta",
-					Usage: &dto.ClaudeUsage{
+					Usage: buildClaudeUsageFromOpenAIUsage(oaiUsage),
-						InputTokens:              oaiUsage.PromptTokens,
-						OutputTokens:             oaiUsage.CompletionTokens,
-						CacheCreationInputTokens: oaiUsage.PromptTokensDetails.CachedCreationTokens,
-						CacheReadInputTokens:     oaiUsage.PromptTokensDetails.CachedTokens,
-					},
 					Delta: &dto.ClaudeMediaMessage{
 						StopReason: common.GetPointer[string](stopReasonOpenAI2Claude(info.FinishReason)),
 					},
@@ -419,13 +433,8 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon
 			oaiUsage := info.ClaudeConvertInfo.Usage
 			if oaiUsage != nil {
 				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
-					Type: "message_delta",
+					Type:  "message_delta",
-					Usage: &dto.ClaudeUsage{
+					Usage: buildClaudeUsageFromOpenAIUsage(oaiUsage),
-						InputTokens:              oaiUsage.PromptTokens,
-						OutputTokens:             oaiUsage.CompletionTokens,
-						CacheCreationInputTokens: oaiUsage.PromptTokensDetails.CachedCreationTokens,
-						CacheReadInputTokens:     oaiUsage.PromptTokensDetails.CachedTokens,
-					},
 					Delta: &dto.ClaudeMediaMessage{
 						StopReason: common.GetPointer[string](stopReasonOpenAI2Claude(info.FinishReason)),
 					},
@@ -555,13 +564,8 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon
 			}
 			if oaiUsage != nil {
 				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
-					Type: "message_delta",
+					Type:  "message_delta",
-					Usage: &dto.ClaudeUsage{
+					Usage: buildClaudeUsageFromOpenAIUsage(oaiUsage),
-						InputTokens:              oaiUsage.PromptTokens,
-						OutputTokens:             oaiUsage.CompletionTokens,
-						CacheCreationInputTokens: oaiUsage.PromptTokensDetails.CachedCreationTokens,
-						CacheReadInputTokens:     oaiUsage.PromptTokensDetails.CachedTokens,
-					},
 					Delta: &dto.ClaudeMediaMessage{
 						StopReason: common.GetPointer[string](stopReasonOpenAI2Claude(info.FinishReason)),
 					},

service/log_info_generate.go

@@ -73,6 +73,7 @@ func GenerateTextOtherInfo(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, m
 	other["admin_info"] = adminInfo
 	appendRequestPath(ctx, relayInfo, other)
 	appendRequestConversionChain(relayInfo, other)
+	appendFinalRequestFormat(relayInfo, other)
 	appendBillingInfo(relayInfo, other)
 	appendParamOverrideInfo(relayInfo, other)
 	return other
@@ -167,6 +168,17 @@ func appendRequestConversionChain(relayInfo *relaycommon.RelayInfo, other map[st
 	other["request_conversion"] = chain
 }
 
+func appendFinalRequestFormat(relayInfo *relaycommon.RelayInfo, other map[string]interface{}) {
+	if relayInfo == nil || other == nil {
+		return
+	}
+	if relayInfo.GetFinalRequestRelayFormat() == types.RelayFormatClaude {
+		// claude indicates the final upstream request format is Claude Messages.
+		// Frontend log rendering uses this to keep the original Claude input display.
+		other["claude"] = true
+	}
+}
+
 func GenerateWssOtherInfo(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usage *dto.RealtimeUsage, modelRatio, groupRatio, completionRatio, audioRatio, audioCompletionRatio, modelPrice, userGroupRatio float64) map[string]interface{} {
 	info := GenerateTextOtherInfo(ctx, relayInfo, modelRatio, groupRatio, completionRatio, 0, 0.0, modelPrice, userGroupRatio)
 	info["ws"] = true

service/quota.go

@@ -235,108 +235,6 @@ func PostWssConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, mod
 	})
 }
 
-func PostClaudeConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usage *dto.Usage) {
-	if usage != nil {
-		ObserveChannelAffinityUsageCacheByRelayFormat(ctx, usage, relayInfo.GetFinalRequestRelayFormat())
-	}
-
-	useTimeSeconds := time.Now().Unix() - relayInfo.StartTime.Unix()
-	promptTokens := usage.PromptTokens
-	completionTokens := usage.CompletionTokens
-	modelName := relayInfo.OriginModelName
-
-	tokenName := ctx.GetString("token_name")
-	completionRatio := relayInfo.PriceData.CompletionRatio
-	modelRatio := relayInfo.PriceData.ModelRatio
-	groupRatio := relayInfo.PriceData.GroupRatioInfo.GroupRatio
-	modelPrice := relayInfo.PriceData.ModelPrice
-	cacheRatio := relayInfo.PriceData.CacheRatio
-	cacheTokens := usage.PromptTokensDetails.CachedTokens
-
-	cacheCreationRatio := relayInfo.PriceData.CacheCreationRatio
-	cacheCreationRatio5m := relayInfo.PriceData.CacheCreation5mRatio
-	cacheCreationRatio1h := relayInfo.PriceData.CacheCreation1hRatio
-	cacheCreationTokens := usage.PromptTokensDetails.CachedCreationTokens
-	cacheCreationTokens5m := usage.ClaudeCacheCreation5mTokens
-	cacheCreationTokens1h := usage.ClaudeCacheCreation1hTokens
-
-	if relayInfo.ChannelType == constant.ChannelTypeOpenRouter {
-		promptTokens -= cacheTokens
-		isUsingCustomSettings := relayInfo.PriceData.UsePrice || hasCustomModelRatio(modelName, relayInfo.PriceData.ModelRatio)
-		if cacheCreationTokens == 0 && relayInfo.PriceData.CacheCreationRatio != 1 && usage.Cost != 0 && !isUsingCustomSettings {
-			maybeCacheCreationTokens := CalcOpenRouterCacheCreateTokens(*usage, relayInfo.PriceData)
-			if maybeCacheCreationTokens >= 0 && promptTokens >= maybeCacheCreationTokens {
-				cacheCreationTokens = maybeCacheCreationTokens
-			}
-		}
-		promptTokens -= cacheCreationTokens
-	}
-
-	calculateQuota := 0.0
-	if !relayInfo.PriceData.UsePrice {
-		calculateQuota = float64(promptTokens)
-		calculateQuota += float64(cacheTokens) * cacheRatio
-		calculateQuota += float64(cacheCreationTokens5m) * cacheCreationRatio5m
-		calculateQuota += float64(cacheCreationTokens1h) * cacheCreationRatio1h
-		remainingCacheCreationTokens := cacheCreationTokens - cacheCreationTokens5m - cacheCreationTokens1h
-		if remainingCacheCreationTokens > 0 {
-			calculateQuota += float64(remainingCacheCreationTokens) * cacheCreationRatio
-		}
-		calculateQuota += float64(completionTokens) * completionRatio
-		calculateQuota = calculateQuota * groupRatio * modelRatio
-	} else {
-		calculateQuota = modelPrice * common.QuotaPerUnit * groupRatio
-	}
-
-	if modelRatio != 0 && calculateQuota <= 0 {
-		calculateQuota = 1
-	}
-
-	quota := int(calculateQuota)
-
-	totalTokens := promptTokens + completionTokens
-
-	var logContent string
-	// record all the consume log even if quota is 0
-	if totalTokens == 0 {
-		// in this case, must be some error happened
-		// we cannot just return, because we may have to return the pre-consumed quota
-		quota = 0
-		logContent += fmt.Sprintf("（可能是上游出错）")
-		logger.LogError(ctx, fmt.Sprintf("total tokens is 0, cannot consume quota, userId %d, channelId %d, "+
-			"tokenId %d, model %s， pre-consumed quota %d", relayInfo.UserId, relayInfo.ChannelId, relayInfo.TokenId, modelName, relayInfo.FinalPreConsumedQuota))
-	} else {
-		model.UpdateUserUsedQuotaAndRequestCount(relayInfo.UserId, quota)
-		model.UpdateChannelUsedQuota(relayInfo.ChannelId, quota)
-	}
-
-	if err := SettleBilling(ctx, relayInfo, quota); err != nil {
-		logger.LogError(ctx, "error settling billing: "+err.Error())
-	}
-
-	other := GenerateClaudeOtherInfo(ctx, relayInfo, modelRatio, groupRatio, completionRatio,
-		cacheTokens, cacheRatio,
-		cacheCreationTokens, cacheCreationRatio,
-		cacheCreationTokens5m, cacheCreationRatio5m,
-		cacheCreationTokens1h, cacheCreationRatio1h,
-		modelPrice, relayInfo.PriceData.GroupRatioInfo.GroupSpecialRatio)
-	model.RecordConsumeLog(ctx, relayInfo.UserId, model.RecordConsumeLogParams{
-		ChannelId:        relayInfo.ChannelId,
-		PromptTokens:     promptTokens,
-		CompletionTokens: completionTokens,
-		ModelName:        modelName,
-		TokenName:        tokenName,
-		Quota:            quota,
-		Content:          logContent,
-		TokenId:          relayInfo.TokenId,
-		UseTimeSeconds:   int(useTimeSeconds),
-		IsStream:         relayInfo.IsStream,
-		Group:            relayInfo.UsingGroup,
-		Other:            other,
-	})
-
-}
-
 func CalcOpenRouterCacheCreateTokens(usage dto.Usage, priceData types.PriceData) int {
 	if priceData.CacheCreationRatio == 1 {
 		return 0

service/text_quota.go

@@ -0,0 +1,427 @@
+package service
+
+import (
+	"fmt"
+	"strings"
+	"time"
+
+	"github.com/QuantumNous/new-api/common"
+	"github.com/QuantumNous/new-api/constant"
+	"github.com/QuantumNous/new-api/dto"
+	"github.com/QuantumNous/new-api/logger"
+	"github.com/QuantumNous/new-api/model"
+	relaycommon "github.com/QuantumNous/new-api/relay/common"
+	"github.com/QuantumNous/new-api/setting/operation_setting"
+	"github.com/QuantumNous/new-api/types"
+
+	"github.com/gin-gonic/gin"
+	"github.com/shopspring/decimal"
+)
+
+type textQuotaSummary struct {
+	PromptTokens             int
+	CompletionTokens         int
+	TotalTokens              int
+	CacheTokens              int
+	CacheCreationTokens      int
+	CacheCreationTokens5m    int
+	CacheCreationTokens1h    int
+	ImageTokens              int
+	AudioTokens              int
+	ModelName                string
+	TokenName                string
+	UseTimeSeconds           int64
+	CompletionRatio          float64
+	CacheRatio               float64
+	ImageRatio               float64
+	ModelRatio               float64
+	GroupRatio               float64
+	ModelPrice               float64
+	CacheCreationRatio       float64
+	CacheCreationRatio5m     float64
+	CacheCreationRatio1h     float64
+	Quota                    int
+	IsClaudeUsageSemantic    bool
+	UsageSemantic            string
+	WebSearchPrice           float64
+	WebSearchCallCount       int
+	ClaudeWebSearchPrice     float64
+	ClaudeWebSearchCallCount int
+	FileSearchPrice          float64
+	FileSearchCallCount      int
+	AudioInputPrice          float64
+	ImageGenerationCallPrice float64
+}
+
+func cacheWriteTokensTotal(summary textQuotaSummary) int {
+	if summary.CacheCreationTokens5m > 0 || summary.CacheCreationTokens1h > 0 {
+		splitCacheWriteTokens := summary.CacheCreationTokens5m + summary.CacheCreationTokens1h
+		if summary.CacheCreationTokens > splitCacheWriteTokens {
+			return summary.CacheCreationTokens
+		}
+		return splitCacheWriteTokens
+	}
+	return summary.CacheCreationTokens
+}
+
+func isLegacyClaudeDerivedOpenAIUsage(relayInfo *relaycommon.RelayInfo, usage *dto.Usage) bool {
+	if relayInfo == nil || usage == nil {
+		return false
+	}
+	if relayInfo.GetFinalRequestRelayFormat() == types.RelayFormatClaude {
+		return false
+	}
+	if usage.UsageSource != "" || usage.UsageSemantic != "" {
+		return false
+	}
+	return usage.ClaudeCacheCreation5mTokens > 0 || usage.ClaudeCacheCreation1hTokens > 0
+}
+
+func calculateTextQuotaSummary(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usage *dto.Usage) textQuotaSummary {
+	summary := textQuotaSummary{
+		ModelName:            relayInfo.OriginModelName,
+		TokenName:            ctx.GetString("token_name"),
+		UseTimeSeconds:       time.Now().Unix() - relayInfo.StartTime.Unix(),
+		CompletionRatio:      relayInfo.PriceData.CompletionRatio,
+		CacheRatio:           relayInfo.PriceData.CacheRatio,
+		ImageRatio:           relayInfo.PriceData.ImageRatio,
+		ModelRatio:           relayInfo.PriceData.ModelRatio,
+		GroupRatio:           relayInfo.PriceData.GroupRatioInfo.GroupRatio,
+		ModelPrice:           relayInfo.PriceData.ModelPrice,
+		CacheCreationRatio:   relayInfo.PriceData.CacheCreationRatio,
+		CacheCreationRatio5m: relayInfo.PriceData.CacheCreation5mRatio,
+		CacheCreationRatio1h: relayInfo.PriceData.CacheCreation1hRatio,
+		UsageSemantic:        usageSemanticFromUsage(relayInfo, usage),
+	}
+	summary.IsClaudeUsageSemantic = summary.UsageSemantic == "anthropic"
+
+	if usage == nil {
+		usage = &dto.Usage{
+			PromptTokens:     relayInfo.GetEstimatePromptTokens(),
+			CompletionTokens: 0,
+			TotalTokens:      relayInfo.GetEstimatePromptTokens(),
+		}
+	}
+
+	summary.PromptTokens = usage.PromptTokens
+	summary.CompletionTokens = usage.CompletionTokens
+	summary.TotalTokens = usage.PromptTokens + usage.CompletionTokens
+	summary.CacheTokens = usage.PromptTokensDetails.CachedTokens
+	summary.CacheCreationTokens = usage.PromptTokensDetails.CachedCreationTokens
+	summary.CacheCreationTokens5m = usage.ClaudeCacheCreation5mTokens
+	summary.CacheCreationTokens1h = usage.ClaudeCacheCreation1hTokens
+	summary.ImageTokens = usage.PromptTokensDetails.ImageTokens
+	summary.AudioTokens = usage.PromptTokensDetails.AudioTokens
+	legacyClaudeDerived := isLegacyClaudeDerivedOpenAIUsage(relayInfo, usage)
+
+	if relayInfo.ChannelMeta != nil && relayInfo.ChannelType == constant.ChannelTypeOpenRouter {
+		summary.PromptTokens -= summary.CacheTokens
+		isUsingCustomSettings := relayInfo.PriceData.UsePrice || hasCustomModelRatio(summary.ModelName, relayInfo.PriceData.ModelRatio)
+		if summary.CacheCreationTokens == 0 && relayInfo.PriceData.CacheCreationRatio != 1 && usage.Cost != 0 && !isUsingCustomSettings {
+			maybeCacheCreationTokens := CalcOpenRouterCacheCreateTokens(*usage, relayInfo.PriceData)
+			if maybeCacheCreationTokens >= 0 && summary.PromptTokens >= maybeCacheCreationTokens {
+				summary.CacheCreationTokens = maybeCacheCreationTokens
+			}
+		}
+		summary.PromptTokens -= summary.CacheCreationTokens
+	}
+
+	dPromptTokens := decimal.NewFromInt(int64(summary.PromptTokens))
+	dCacheTokens := decimal.NewFromInt(int64(summary.CacheTokens))
+	dImageTokens := decimal.NewFromInt(int64(summary.ImageTokens))
+	dAudioTokens := decimal.NewFromInt(int64(summary.AudioTokens))
+	dCompletionTokens := decimal.NewFromInt(int64(summary.CompletionTokens))
+	dCachedCreationTokens := decimal.NewFromInt(int64(summary.CacheCreationTokens))
+	dCompletionRatio := decimal.NewFromFloat(summary.CompletionRatio)
+	dCacheRatio := decimal.NewFromFloat(summary.CacheRatio)
+	dImageRatio := decimal.NewFromFloat(summary.ImageRatio)
+	dModelRatio := decimal.NewFromFloat(summary.ModelRatio)
+	dGroupRatio := decimal.NewFromFloat(summary.GroupRatio)
+	dModelPrice := decimal.NewFromFloat(summary.ModelPrice)
+	dCacheCreationRatio := decimal.NewFromFloat(summary.CacheCreationRatio)
+	dCacheCreationRatio5m := decimal.NewFromFloat(summary.CacheCreationRatio5m)
+	dCacheCreationRatio1h := decimal.NewFromFloat(summary.CacheCreationRatio1h)
+	dQuotaPerUnit := decimal.NewFromFloat(common.QuotaPerUnit)
+
+	ratio := dModelRatio.Mul(dGroupRatio)
+
+	var dWebSearchQuota decimal.Decimal
+	if relayInfo.ResponsesUsageInfo != nil {
+		if webSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolWebSearchPreview]; exists && webSearchTool.CallCount > 0 {
+			summary.WebSearchCallCount = webSearchTool.CallCount
+			summary.WebSearchPrice = operation_setting.GetWebSearchPricePerThousand(summary.ModelName, webSearchTool.SearchContextSize)
+			dWebSearchQuota = decimal.NewFromFloat(summary.WebSearchPrice).
+				Mul(decimal.NewFromInt(int64(webSearchTool.CallCount))).
+				Div(decimal.NewFromInt(1000)).Mul(dGroupRatio).Mul(dQuotaPerUnit)
+		}
+	} else if strings.HasSuffix(summary.ModelName, "search-preview") {
+		searchContextSize := ctx.GetString("chat_completion_web_search_context_size")
+		if searchContextSize == "" {
+			searchContextSize = "medium"
+		}
+		summary.WebSearchCallCount = 1
+		summary.WebSearchPrice = operation_setting.GetWebSearchPricePerThousand(summary.ModelName, searchContextSize)
+		dWebSearchQuota = decimal.NewFromFloat(summary.WebSearchPrice).
+			Div(decimal.NewFromInt(1000)).Mul(dGroupRatio).Mul(dQuotaPerUnit)
+	}
+
+	var dClaudeWebSearchQuota decimal.Decimal
+	summary.ClaudeWebSearchCallCount = ctx.GetInt("claude_web_search_requests")
+	if summary.ClaudeWebSearchCallCount > 0 {
+		summary.ClaudeWebSearchPrice = operation_setting.GetClaudeWebSearchPricePerThousand()
+		dClaudeWebSearchQuota = decimal.NewFromFloat(summary.ClaudeWebSearchPrice).
+			Div(decimal.NewFromInt(1000)).Mul(dGroupRatio).Mul(dQuotaPerUnit).
+			Mul(decimal.NewFromInt(int64(summary.ClaudeWebSearchCallCount)))
+	}
+
+	var dFileSearchQuota decimal.Decimal
+	if relayInfo.ResponsesUsageInfo != nil {
+		if fileSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolFileSearch]; exists && fileSearchTool.CallCount > 0 {
+			summary.FileSearchCallCount = fileSearchTool.CallCount
+			summary.FileSearchPrice = operation_setting.GetFileSearchPricePerThousand()
+			dFileSearchQuota = decimal.NewFromFloat(summary.FileSearchPrice).
+				Mul(decimal.NewFromInt(int64(fileSearchTool.CallCount))).
+				Div(decimal.NewFromInt(1000)).Mul(dGroupRatio).Mul(dQuotaPerUnit)
+		}
+	}
+
+	var dImageGenerationCallQuota decimal.Decimal
+	if ctx.GetBool("image_generation_call") {
+		summary.ImageGenerationCallPrice = operation_setting.GetGPTImage1PriceOnceCall(ctx.GetString("image_generation_call_quality"), ctx.GetString("image_generation_call_size"))
+		dImageGenerationCallQuota = decimal.NewFromFloat(summary.ImageGenerationCallPrice).Mul(dGroupRatio).Mul(dQuotaPerUnit)
+	}
+
+	var audioInputQuota decimal.Decimal
+	if !relayInfo.PriceData.UsePrice {
+		baseTokens := dPromptTokens
+
+		var cachedTokensWithRatio decimal.Decimal
+		if !dCacheTokens.IsZero() {
+			if !summary.IsClaudeUsageSemantic && !legacyClaudeDerived {
+				baseTokens = baseTokens.Sub(dCacheTokens)
+			}
+			cachedTokensWithRatio = dCacheTokens.Mul(dCacheRatio)
+		}
+
+		var cachedCreationTokensWithRatio decimal.Decimal
+		hasSplitCacheCreationTokens := summary.CacheCreationTokens5m > 0 || summary.CacheCreationTokens1h > 0
+		if !dCachedCreationTokens.IsZero() || hasSplitCacheCreationTokens {
+			if !summary.IsClaudeUsageSemantic && !legacyClaudeDerived {
+				baseTokens = baseTokens.Sub(dCachedCreationTokens)
+				cachedCreationTokensWithRatio = dCachedCreationTokens.Mul(dCacheCreationRatio)
+			} else {
+				remaining := summary.CacheCreationTokens - summary.CacheCreationTokens5m - summary.CacheCreationTokens1h
+				if remaining < 0 {
+					remaining = 0
+				}
+				cachedCreationTokensWithRatio = decimal.NewFromInt(int64(remaining)).Mul(dCacheCreationRatio)
+				cachedCreationTokensWithRatio = cachedCreationTokensWithRatio.Add(decimal.NewFromInt(int64(summary.CacheCreationTokens5m)).Mul(dCacheCreationRatio5m))
+				cachedCreationTokensWithRatio = cachedCreationTokensWithRatio.Add(decimal.NewFromInt(int64(summary.CacheCreationTokens1h)).Mul(dCacheCreationRatio1h))
+			}
+		}
+
+		var imageTokensWithRatio decimal.Decimal
+		if !dImageTokens.IsZero() {
+			baseTokens = baseTokens.Sub(dImageTokens)
+			imageTokensWithRatio = dImageTokens.Mul(dImageRatio)
+		}
+
+		if !dAudioTokens.IsZero() {
+			summary.AudioInputPrice = operation_setting.GetGeminiInputAudioPricePerMillionTokens(summary.ModelName)
+			if summary.AudioInputPrice > 0 {
+				baseTokens = baseTokens.Sub(dAudioTokens)
+				audioInputQuota = decimal.NewFromFloat(summary.AudioInputPrice).
+					Div(decimal.NewFromInt(1000000)).Mul(dAudioTokens).Mul(dGroupRatio).Mul(dQuotaPerUnit)
+			}
+		}
+
+		promptQuota := baseTokens.Add(cachedTokensWithRatio).Add(imageTokensWithRatio).Add(cachedCreationTokensWithRatio)
+		completionQuota := dCompletionTokens.Mul(dCompletionRatio)
+		quotaCalculateDecimal := promptQuota.Add(completionQuota).Mul(ratio)
+		quotaCalculateDecimal = quotaCalculateDecimal.Add(dWebSearchQuota)
+		quotaCalculateDecimal = quotaCalculateDecimal.Add(dClaudeWebSearchQuota)
+		quotaCalculateDecimal = quotaCalculateDecimal.Add(dFileSearchQuota)
+		quotaCalculateDecimal = quotaCalculateDecimal.Add(audioInputQuota)
+		quotaCalculateDecimal = quotaCalculateDecimal.Add(dImageGenerationCallQuota)
+
+		if len(relayInfo.PriceData.OtherRatios) > 0 {
+			for _, otherRatio := range relayInfo.PriceData.OtherRatios {
+				quotaCalculateDecimal = quotaCalculateDecimal.Mul(decimal.NewFromFloat(otherRatio))
+			}
+		}
+
+		if !ratio.IsZero() && quotaCalculateDecimal.LessThanOrEqual(decimal.Zero) {
+			quotaCalculateDecimal = decimal.NewFromInt(1)
+		}
+		summary.Quota = int(quotaCalculateDecimal.Round(0).IntPart())
+	} else {
+		quotaCalculateDecimal := dModelPrice.Mul(dQuotaPerUnit).Mul(dGroupRatio)
+		quotaCalculateDecimal = quotaCalculateDecimal.Add(dWebSearchQuota)
+		quotaCalculateDecimal = quotaCalculateDecimal.Add(dClaudeWebSearchQuota)
+		quotaCalculateDecimal = quotaCalculateDecimal.Add(dFileSearchQuota)
+		quotaCalculateDecimal = quotaCalculateDecimal.Add(audioInputQuota)
+		quotaCalculateDecimal = quotaCalculateDecimal.Add(dImageGenerationCallQuota)
+		if len(relayInfo.PriceData.OtherRatios) > 0 {
+			for _, otherRatio := range relayInfo.PriceData.OtherRatios {
+				quotaCalculateDecimal = quotaCalculateDecimal.Mul(decimal.NewFromFloat(otherRatio))
+			}
+		}
+		summary.Quota = int(quotaCalculateDecimal.Round(0).IntPart())
+	}
+
+	if summary.TotalTokens == 0 {
+		summary.Quota = 0
+	} else if !ratio.IsZero() && summary.Quota == 0 {
+		summary.Quota = 1
+	}
+
+	return summary
+}
+
+func usageSemanticFromUsage(relayInfo *relaycommon.RelayInfo, usage *dto.Usage) string {
+	if usage != nil && usage.UsageSemantic != "" {
+		return usage.UsageSemantic
+	}
+	if relayInfo != nil && relayInfo.GetFinalRequestRelayFormat() == types.RelayFormatClaude {
+		return "anthropic"
+	}
+	return "openai"
+}
+
+func PostTextConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usage *dto.Usage, extraContent []string) {
+	originUsage := usage
+	if usage == nil {
+		extraContent = append(extraContent, "上游无计费信息")
+	}
+	if originUsage != nil {
+		ObserveChannelAffinityUsageCacheByRelayFormat(ctx, usage, relayInfo.GetFinalRequestRelayFormat())
+	}
+
+	adminRejectReason := common.GetContextKeyString(ctx, constant.ContextKeyAdminRejectReason)
+	summary := calculateTextQuotaSummary(ctx, relayInfo, usage)
+
+	if summary.WebSearchCallCount > 0 {
+		extraContent = append(extraContent, fmt.Sprintf("Web Search 调用 %d 次，调用花费 %s", summary.WebSearchCallCount, decimal.NewFromFloat(summary.WebSearchPrice).Mul(decimal.NewFromInt(int64(summary.WebSearchCallCount))).Div(decimal.NewFromInt(1000)).Mul(decimal.NewFromFloat(summary.GroupRatio)).Mul(decimal.NewFromFloat(common.QuotaPerUnit)).String()))
+	}
+	if summary.ClaudeWebSearchCallCount > 0 {
+		extraContent = append(extraContent, fmt.Sprintf("Claude Web Search 调用 %d 次，调用花费 %s", summary.ClaudeWebSearchCallCount, decimal.NewFromFloat(summary.ClaudeWebSearchPrice).Div(decimal.NewFromInt(1000)).Mul(decimal.NewFromFloat(summary.GroupRatio)).Mul(decimal.NewFromFloat(common.QuotaPerUnit)).Mul(decimal.NewFromInt(int64(summary.ClaudeWebSearchCallCount))).String()))
+	}
+	if summary.FileSearchCallCount > 0 {
+		extraContent = append(extraContent, fmt.Sprintf("File Search 调用 %d 次，调用花费 %s", summary.FileSearchCallCount, decimal.NewFromFloat(summary.FileSearchPrice).Mul(decimal.NewFromInt(int64(summary.FileSearchCallCount))).Div(decimal.NewFromInt(1000)).Mul(decimal.NewFromFloat(summary.GroupRatio)).Mul(decimal.NewFromFloat(common.QuotaPerUnit)).String()))
+	}
+	if summary.AudioInputPrice > 0 && summary.AudioTokens > 0 {
+		extraContent = append(extraContent, fmt.Sprintf("Audio Input 花费 %s", decimal.NewFromFloat(summary.AudioInputPrice).Div(decimal.NewFromInt(1000000)).Mul(decimal.NewFromInt(int64(summary.AudioTokens))).Mul(decimal.NewFromFloat(summary.GroupRatio)).Mul(decimal.NewFromFloat(common.QuotaPerUnit)).String()))
+	}
+	if summary.ImageGenerationCallPrice > 0 {
+		extraContent = append(extraContent, fmt.Sprintf("Image Generation Call 花费 %s", decimal.NewFromFloat(summary.ImageGenerationCallPrice).Mul(decimal.NewFromFloat(summary.GroupRatio)).Mul(decimal.NewFromFloat(common.QuotaPerUnit)).String()))
+	}
+
+	if summary.TotalTokens == 0 {
+		extraContent = append(extraContent, "上游没有返回计费信息，无法扣费（可能是上游超时）")
+		logger.LogError(ctx, fmt.Sprintf("total tokens is 0, cannot consume quota, userId %d, channelId %d, tokenId %d, model %s， pre-consumed quota %d", relayInfo.UserId, relayInfo.ChannelId, relayInfo.TokenId, summary.ModelName, relayInfo.FinalPreConsumedQuota))
+	} else {
+		model.UpdateUserUsedQuotaAndRequestCount(relayInfo.UserId, summary.Quota)
+		model.UpdateChannelUsedQuota(relayInfo.ChannelId, summary.Quota)
+	}
+
+	if err := SettleBilling(ctx, relayInfo, summary.Quota); err != nil {
+		logger.LogError(ctx, "error settling billing: "+err.Error())
+	}
+
+	logModel := summary.ModelName
+	if strings.HasPrefix(logModel, "gpt-4-gizmo") {
+		logModel = "gpt-4-gizmo-*"
+		extraContent = append(extraContent, fmt.Sprintf("模型 %s", summary.ModelName))
+	}
+	if strings.HasPrefix(logModel, "gpt-4o-gizmo") {
+		logModel = "gpt-4o-gizmo-*"
+		extraContent = append(extraContent, fmt.Sprintf("模型 %s", summary.ModelName))
+	}
+
+	logContent := strings.Join(extraContent, ", ")
+	var other map[string]interface{}
+	if summary.IsClaudeUsageSemantic {
+		other = GenerateClaudeOtherInfo(ctx, relayInfo,
+			summary.ModelRatio, summary.GroupRatio, summary.CompletionRatio,
+			summary.CacheTokens, summary.CacheRatio,
+			summary.CacheCreationTokens, summary.CacheCreationRatio,
+			summary.CacheCreationTokens5m, summary.CacheCreationRatio5m,
+			summary.CacheCreationTokens1h, summary.CacheCreationRatio1h,
+			summary.ModelPrice, relayInfo.PriceData.GroupRatioInfo.GroupSpecialRatio)
+		other["usage_semantic"] = "anthropic"
+	} else {
+		other = GenerateTextOtherInfo(ctx, relayInfo, summary.ModelRatio, summary.GroupRatio, summary.CompletionRatio, summary.CacheTokens, summary.CacheRatio, summary.ModelPrice, relayInfo.PriceData.GroupRatioInfo.GroupSpecialRatio)
+	}
+	if adminRejectReason != "" {
+		other["reject_reason"] = adminRejectReason
+	}
+	if summary.ImageTokens != 0 {
+		other["image"] = true
+		other["image_ratio"] = summary.ImageRatio
+		other["image_output"] = summary.ImageTokens
+	}
+	if summary.WebSearchCallCount > 0 {
+		other["web_search"] = true
+		other["web_search_call_count"] = summary.WebSearchCallCount
+		other["web_search_price"] = summary.WebSearchPrice
+	} else if summary.ClaudeWebSearchCallCount > 0 {
+		other["web_search"] = true
+		other["web_search_call_count"] = summary.ClaudeWebSearchCallCount
+		other["web_search_price"] = summary.ClaudeWebSearchPrice
+	}
+	if summary.FileSearchCallCount > 0 {
+		other["file_search"] = true
+		other["file_search_call_count"] = summary.FileSearchCallCount
+		other["file_search_price"] = summary.FileSearchPrice
+	}
+	if summary.AudioInputPrice > 0 && summary.AudioTokens > 0 {
+		other["audio_input_seperate_price"] = true
+		other["audio_input_token_count"] = summary.AudioTokens
+		other["audio_input_price"] = summary.AudioInputPrice
+	}
+	if summary.ImageGenerationCallPrice > 0 {
+		other["image_generation_call"] = true
+		other["image_generation_call_price"] = summary.ImageGenerationCallPrice
+	}
+	if summary.CacheCreationTokens > 0 {
+		other["cache_creation_tokens"] = summary.CacheCreationTokens
+		other["cache_creation_ratio"] = summary.CacheCreationRatio
+	}
+	if summary.CacheCreationTokens5m > 0 {
+		other["cache_creation_tokens_5m"] = summary.CacheCreationTokens5m
+		other["cache_creation_ratio_5m"] = summary.CacheCreationRatio5m
+	}
+	if summary.CacheCreationTokens1h > 0 {
+		other["cache_creation_tokens_1h"] = summary.CacheCreationTokens1h
+		other["cache_creation_ratio_1h"] = summary.CacheCreationRatio1h
+	}
+	cacheWriteTokens := cacheWriteTokensTotal(summary)
+	if cacheWriteTokens > 0 {
+		// cache_write_tokens: normalized cache creation total for UI display.
+		// If split 5m/1h values are present, this is their sum; otherwise it falls back
+		// to cache_creation_tokens.
+		other["cache_write_tokens"] = cacheWriteTokens
+	}
+	if relayInfo.GetFinalRequestRelayFormat() != types.RelayFormatClaude && usage != nil && usage.UsageSource != "" && usage.InputTokens > 0 {
+		// input_tokens_total: explicit normalized total input used by the usage log UI.
+		// Only write this field when upstream/current conversion has already provided a
+		// reliable total input value and tagged the usage source. Do not infer it from
+		// prompt/cache fields here, otherwise old upstream payloads may be double-counted.
+		other["input_tokens_total"] = usage.InputTokens
+	}
+
+	model.RecordConsumeLog(ctx, relayInfo.UserId, model.RecordConsumeLogParams{
+		ChannelId:        relayInfo.ChannelId,
+		PromptTokens:     summary.PromptTokens,
+		CompletionTokens: summary.CompletionTokens,
+		ModelName:        logModel,
+		TokenName:        summary.TokenName,
+		Quota:            summary.Quota,
+		Content:          logContent,
+		TokenId:          relayInfo.TokenId,
+		UseTimeSeconds:   int(summary.UseTimeSeconds),
+		IsStream:         relayInfo.IsStream,
+		Group:            relayInfo.UsingGroup,
+		Other:            other,
+	})
+}

service/text_quota_test.go

@@ -0,0 +1,206 @@
+package service
+
+import (
+	"net/http/httptest"
+	"testing"
+	"time"
+
+	"github.com/QuantumNous/new-api/dto"
+	relaycommon "github.com/QuantumNous/new-api/relay/common"
+	"github.com/QuantumNous/new-api/types"
+
+	"github.com/gin-gonic/gin"
+	"github.com/stretchr/testify/require"
+)
+
+func TestCalculateTextQuotaSummaryUnifiedForClaudeSemantic(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	ctx, _ := gin.CreateTestContext(w)
+
+	usage := &dto.Usage{
+		PromptTokens:     1000,
+		CompletionTokens: 200,
+		PromptTokensDetails: dto.InputTokenDetails{
+			CachedTokens:         100,
+			CachedCreationTokens: 50,
+		},
+		ClaudeCacheCreation5mTokens: 10,
+		ClaudeCacheCreation1hTokens: 20,
+	}
+
+	priceData := types.PriceData{
+		ModelRatio:           1,
+		CompletionRatio:      2,
+		CacheRatio:           0.1,
+		CacheCreationRatio:   1.25,
+		CacheCreation5mRatio: 1.25,
+		CacheCreation1hRatio: 2,
+		GroupRatioInfo: types.GroupRatioInfo{
+			GroupRatio: 1,
+		},
+	}
+
+	chatRelayInfo := &relaycommon.RelayInfo{
+		RelayFormat:             types.RelayFormatOpenAI,
+		FinalRequestRelayFormat: types.RelayFormatClaude,
+		OriginModelName:         "claude-3-7-sonnet",
+		PriceData:               priceData,
+		StartTime:               time.Now(),
+	}
+	messageRelayInfo := &relaycommon.RelayInfo{
+		RelayFormat:             types.RelayFormatClaude,
+		FinalRequestRelayFormat: types.RelayFormatClaude,
+		OriginModelName:         "claude-3-7-sonnet",
+		PriceData:               priceData,
+		StartTime:               time.Now(),
+	}
+
+	chatSummary := calculateTextQuotaSummary(ctx, chatRelayInfo, usage)
+	messageSummary := calculateTextQuotaSummary(ctx, messageRelayInfo, usage)
+
+	require.Equal(t, messageSummary.Quota, chatSummary.Quota)
+	require.Equal(t, messageSummary.CacheCreationTokens5m, chatSummary.CacheCreationTokens5m)
+	require.Equal(t, messageSummary.CacheCreationTokens1h, chatSummary.CacheCreationTokens1h)
+	require.True(t, chatSummary.IsClaudeUsageSemantic)
+	require.Equal(t, 1488, chatSummary.Quota)
+}
+
+func TestCalculateTextQuotaSummaryUsesSplitClaudeCacheCreationRatios(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	ctx, _ := gin.CreateTestContext(w)
+
+	relayInfo := &relaycommon.RelayInfo{
+		RelayFormat:             types.RelayFormatOpenAI,
+		FinalRequestRelayFormat: types.RelayFormatClaude,
+		OriginModelName:         "claude-3-7-sonnet",
+		PriceData: types.PriceData{
+			ModelRatio:           1,
+			CompletionRatio:      1,
+			CacheRatio:           0,
+			CacheCreationRatio:   1,
+			CacheCreation5mRatio: 2,
+			CacheCreation1hRatio: 3,
+			GroupRatioInfo: types.GroupRatioInfo{
+				GroupRatio: 1,
+			},
+		},
+		StartTime: time.Now(),
+	}
+
+	usage := &dto.Usage{
+		PromptTokens:     100,
+		CompletionTokens: 0,
+		PromptTokensDetails: dto.InputTokenDetails{
+			CachedCreationTokens: 10,
+		},
+		ClaudeCacheCreation5mTokens: 2,
+		ClaudeCacheCreation1hTokens: 3,
+	}
+
+	summary := calculateTextQuotaSummary(ctx, relayInfo, usage)
+
+	// 100 + remaining(5)*1 + 2*2 + 3*3 = 118
+	require.Equal(t, 118, summary.Quota)
+}
+
+func TestCalculateTextQuotaSummaryUsesAnthropicUsageSemanticFromUpstreamUsage(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	ctx, _ := gin.CreateTestContext(w)
+
+	relayInfo := &relaycommon.RelayInfo{
+		RelayFormat:     types.RelayFormatOpenAI,
+		OriginModelName: "claude-3-7-sonnet",
+		PriceData: types.PriceData{
+			ModelRatio:           1,
+			CompletionRatio:      2,
+			CacheRatio:           0.1,
+			CacheCreationRatio:   1.25,
+			CacheCreation5mRatio: 1.25,
+			CacheCreation1hRatio: 2,
+			GroupRatioInfo: types.GroupRatioInfo{
+				GroupRatio: 1,
+			},
+		},
+		StartTime: time.Now(),
+	}
+
+	usage := &dto.Usage{
+		PromptTokens:     1000,
+		CompletionTokens: 200,
+		UsageSemantic:    "anthropic",
+		PromptTokensDetails: dto.InputTokenDetails{
+			CachedTokens:         100,
+			CachedCreationTokens: 50,
+		},
+		ClaudeCacheCreation5mTokens: 10,
+		ClaudeCacheCreation1hTokens: 20,
+	}
+
+	summary := calculateTextQuotaSummary(ctx, relayInfo, usage)
+
+	require.True(t, summary.IsClaudeUsageSemantic)
+	require.Equal(t, "anthropic", summary.UsageSemantic)
+	require.Equal(t, 1488, summary.Quota)
+}
+
+func TestCacheWriteTokensTotal(t *testing.T) {
+	t.Run("split cache creation", func(t *testing.T) {
+		summary := textQuotaSummary{
+			CacheCreationTokens:   50,
+			CacheCreationTokens5m: 10,
+			CacheCreationTokens1h: 20,
+		}
+		require.Equal(t, 50, cacheWriteTokensTotal(summary))
+	})
+
+	t.Run("legacy cache creation", func(t *testing.T) {
+		summary := textQuotaSummary{CacheCreationTokens: 50}
+		require.Equal(t, 50, cacheWriteTokensTotal(summary))
+	})
+
+	t.Run("split cache creation without aggregate remainder", func(t *testing.T) {
+		summary := textQuotaSummary{
+			CacheCreationTokens5m: 10,
+			CacheCreationTokens1h: 20,
+		}
+		require.Equal(t, 30, cacheWriteTokensTotal(summary))
+	})
+}
+
+func TestCalculateTextQuotaSummaryHandlesLegacyClaudeDerivedOpenAIUsage(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	ctx, _ := gin.CreateTestContext(w)
+
+	relayInfo := &relaycommon.RelayInfo{
+		RelayFormat:     types.RelayFormatOpenAI,
+		OriginModelName: "claude-3-7-sonnet",
+		PriceData: types.PriceData{
+			ModelRatio:           1,
+			CompletionRatio:      5,
+			CacheRatio:           0.1,
+			CacheCreationRatio:   1.25,
+			CacheCreation5mRatio: 1.25,
+			CacheCreation1hRatio: 2,
+			GroupRatioInfo:       types.GroupRatioInfo{GroupRatio: 1},
+		},
+		StartTime: time.Now(),
+	}
+
+	usage := &dto.Usage{
+		PromptTokens:     62,
+		CompletionTokens: 95,
+		PromptTokensDetails: dto.InputTokenDetails{
+			CachedTokens: 3544,
+		},
+		ClaudeCacheCreation5mTokens: 586,
+	}
+
+	summary := calculateTextQuotaSummary(ctx, relayInfo, usage)
+
+	// 62 + 3544*0.1 + 586*1.25 + 95*5 = 1624.9 => 1624
+	require.Equal(t, 1624, summary.Quota)
+}