fix: restore pre-3400 OpenRouter billing semantics

service/convert.go

@@ -616,10 +616,7 @@ func ResponseOpenAI2Claude(openAIResponse *dto.OpenAITextResponse, info *relayco
 	}
 	claudeResponse.Content = contents
 	claudeResponse.StopReason = stopReason
-	claudeResponse.Usage = &dto.ClaudeUsage{
-		InputTokens:  openAIResponse.PromptTokens,
-		OutputTokens: openAIResponse.CompletionTokens,
-	}
+	claudeResponse.Usage = buildClaudeUsageFromOpenAIUsage(&openAIResponse.Usage)
 
 	return claudeResponse
 }

service/text_quota.go

@@ -113,8 +113,11 @@ func calculateTextQuotaSummary(ctx *gin.Context, relayInfo *relaycommon.RelayInf
 	summary.ImageTokens = usage.PromptTokensDetails.ImageTokens
 	summary.AudioTokens = usage.PromptTokensDetails.AudioTokens
 	legacyClaudeDerived := isLegacyClaudeDerivedOpenAIUsage(relayInfo, usage)
+	isOpenRouterClaudeBilling := relayInfo.ChannelMeta != nil &&
+		relayInfo.ChannelType == constant.ChannelTypeOpenRouter &&
+		summary.IsClaudeUsageSemantic
 
-	if relayInfo.ChannelMeta != nil && relayInfo.ChannelType == constant.ChannelTypeOpenRouter {
+	if isOpenRouterClaudeBilling {
 		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 {

service/text_quota_test.go

@@ -5,6 +5,7 @@ import (
 	"testing"
 	"time"
 
+	"github.com/QuantumNous/new-api/constant"
 	"github.com/QuantumNous/new-api/dto"
 	relaycommon "github.com/QuantumNous/new-api/relay/common"
 	"github.com/QuantumNous/new-api/types"
@@ -204,3 +205,114 @@ func TestCalculateTextQuotaSummaryHandlesLegacyClaudeDerivedOpenAIUsage(t *testi
 	// 62 + 3544*0.1 + 586*1.25 + 95*5 = 1624.9 => 1624
 	require.Equal(t, 1624, summary.Quota)
 }
+
+func TestCalculateTextQuotaSummarySeparatesOpenRouterCacheReadFromPromptBilling(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	ctx, _ := gin.CreateTestContext(w)
+
+	relayInfo := &relaycommon.RelayInfo{
+		OriginModelName: "openai/gpt-4.1",
+		ChannelMeta: &relaycommon.ChannelMeta{
+			ChannelType: constant.ChannelTypeOpenRouter,
+		},
+		PriceData: types.PriceData{
+			ModelRatio:         1,
+			CompletionRatio:    1,
+			CacheRatio:         0.1,
+			CacheCreationRatio: 1.25,
+			GroupRatioInfo:     types.GroupRatioInfo{GroupRatio: 1},
+		},
+		StartTime: time.Now(),
+	}
+
+	usage := &dto.Usage{
+		PromptTokens:     2604,
+		CompletionTokens: 383,
+		PromptTokensDetails: dto.InputTokenDetails{
+			CachedTokens: 2432,
+		},
+	}
+
+	summary := calculateTextQuotaSummary(ctx, relayInfo, usage)
+
+	// OpenRouter OpenAI-format display keeps prompt_tokens as total input,
+	// but billing still separates normal input from cache read tokens.
+	// quota = (2604 - 2432) + 2432*0.1 + 383 = 798.2 => 798
+	require.Equal(t, 2604, summary.PromptTokens)
+	require.Equal(t, 798, summary.Quota)
+}
+
+func TestCalculateTextQuotaSummarySeparatesOpenRouterCacheCreationFromPromptBilling(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	ctx, _ := gin.CreateTestContext(w)
+
+	relayInfo := &relaycommon.RelayInfo{
+		OriginModelName: "openai/gpt-4.1",
+		ChannelMeta: &relaycommon.ChannelMeta{
+			ChannelType: constant.ChannelTypeOpenRouter,
+		},
+		PriceData: types.PriceData{
+			ModelRatio:         1,
+			CompletionRatio:    1,
+			CacheCreationRatio: 1.25,
+			GroupRatioInfo:     types.GroupRatioInfo{GroupRatio: 1},
+		},
+		StartTime: time.Now(),
+	}
+
+	usage := &dto.Usage{
+		PromptTokens:     2604,
+		CompletionTokens: 383,
+		PromptTokensDetails: dto.InputTokenDetails{
+			CachedCreationTokens: 100,
+		},
+	}
+
+	summary := calculateTextQuotaSummary(ctx, relayInfo, usage)
+
+	// prompt_tokens is still logged as total input, but cache creation is billed separately.
+	// quota = (2604 - 100) + 100*1.25 + 383 = 3012
+	require.Equal(t, 2604, summary.PromptTokens)
+	require.Equal(t, 3012, summary.Quota)
+}
+
+func TestCalculateTextQuotaSummaryKeepsPrePRClaudeOpenRouterBilling(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	ctx, _ := gin.CreateTestContext(w)
+
+	relayInfo := &relaycommon.RelayInfo{
+		FinalRequestRelayFormat: types.RelayFormatClaude,
+		OriginModelName:         "anthropic/claude-3.7-sonnet",
+		ChannelMeta: &relaycommon.ChannelMeta{
+			ChannelType: constant.ChannelTypeOpenRouter,
+		},
+		PriceData: types.PriceData{
+			ModelRatio:         1,
+			CompletionRatio:    1,
+			CacheRatio:         0.1,
+			CacheCreationRatio: 1.25,
+			GroupRatioInfo:     types.GroupRatioInfo{GroupRatio: 1},
+		},
+		StartTime: time.Now(),
+	}
+
+	usage := &dto.Usage{
+		PromptTokens:     2604,
+		CompletionTokens: 383,
+		PromptTokensDetails: dto.InputTokenDetails{
+			CachedTokens: 2432,
+		},
+	}
+
+	summary := calculateTextQuotaSummary(ctx, relayInfo, usage)
+
+	// Pre-PR PostClaudeConsumeQuota behavior for OpenRouter:
+	// prompt = 2604 - 2432 = 172
+	// quota = 172 + 2432*0.1 + 383 = 798.2 => 798
+	require.True(t, summary.IsClaudeUsageSemantic)
+	require.Equal(t, 172, summary.PromptTokens)
+	require.Equal(t, 798, summary.Quota)
+}