Merge pull request #1120 from neotf/feat-04

feat: enhance token usage details for upstream OpenRouter

dto/openai_request.go

@@ -57,6 +57,7 @@ type GeneralOpenAIRequest struct {
 	ExtraBody           json.RawMessage   `json:"extra_body,omitempty"`
 	WebSearchOptions    *WebSearchOptions `json:"web_search_options,omitempty"`
 	// OpenRouter Params
+	Usage     json.RawMessage `json:"usage,omitempty"`
 	Reasoning json.RawMessage `json:"reasoning,omitempty"`
 	// Ali Qwen Params
 	VlHighResolutionImages json.RawMessage `json:"vl_high_resolution_images,omitempty"`

dto/openai_response.go

@@ -178,6 +178,8 @@ type Usage struct {
 	InputTokens            int                `json:"input_tokens"`
 	OutputTokens           int                `json:"output_tokens"`
 	InputTokensDetails     *InputTokenDetails `json:"input_tokens_details"`
+	// OpenRouter Params
+	Cost float64 `json:"cost,omitempty"`
 }
 
 type InputTokenDetails struct {

relay/channel/openai/adaptor.go

@@ -159,6 +159,11 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
 	if info.ChannelType != common.ChannelTypeOpenAI && info.ChannelType != common.ChannelTypeAzure {
 		request.StreamOptions = nil
 	}
+	if info.ChannelType == common.ChannelTypeOpenRouter {
+		if len(request.Usage) == 0 {
+			request.Usage = json.RawMessage(`{"include":true}`)
+		}
+	}
 	if strings.HasPrefix(request.Model, "o") {
 		if request.MaxCompletionTokens == 0 && request.MaxTokens != 0 {
 			request.MaxCompletionTokens = request.MaxTokens

service/convert.go

@@ -276,12 +276,15 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon
 		}
 		if info.Done {
 			claudeResponses = append(claudeResponses, generateStopBlock(info.ClaudeConvertInfo.Index))
-			if info.ClaudeConvertInfo.Usage != nil {
+			oaiUsage := info.ClaudeConvertInfo.Usage
+			if oaiUsage != nil {
 				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
 					Type: "message_delta",
 					Usage: &dto.ClaudeUsage{
-						InputTokens:  info.ClaudeConvertInfo.Usage.PromptTokens,
-						OutputTokens: info.ClaudeConvertInfo.Usage.CompletionTokens,
+						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/quota.go

@@ -4,6 +4,7 @@ import (
 	"errors"
 	"fmt"
 	"log"
+	"math"
 	"one-api/common"
 	constant2 "one-api/constant"
 	"one-api/dto"
@@ -231,6 +232,17 @@ func PostClaudeConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
 	cacheCreationRatio := priceData.CacheCreationRatio
 	cacheCreationTokens := usage.PromptTokensDetails.CachedCreationTokens
 
+	if relayInfo.ChannelType == common.ChannelTypeOpenRouter {
+		promptTokens -= cacheTokens
+		if cacheCreationTokens == 0 && priceData.CacheCreationRatio != 1 && usage.Cost != 0 {
+			maybeCacheCreationTokens := CalcOpenRouterCacheCreateTokens(*usage, priceData)
+			if promptTokens >= maybeCacheCreationTokens {
+				cacheCreationTokens = maybeCacheCreationTokens
+			}
+		}
+		promptTokens -= cacheCreationTokens
+	}
+
 	calculateQuota := 0.0
 	if !priceData.UsePrice {
 		calculateQuota = float64(promptTokens)
@@ -278,6 +290,27 @@ func PostClaudeConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
 		tokenName, quota, logContent, relayInfo.TokenId, userQuota, int(useTimeSeconds), relayInfo.IsStream, relayInfo.Group, other)
 }
 
+func CalcOpenRouterCacheCreateTokens(usage dto.Usage, priceData helper.PriceData) int {
+	if priceData.CacheCreationRatio == 1 {
+		return 0
+	}
+	quotaPrice := priceData.ModelRatio / common.QuotaPerUnit
+	promptCacheCreatePrice := quotaPrice * priceData.CacheCreationRatio
+	promptCacheReadPrice := quotaPrice * priceData.CacheRatio
+	completionPrice := quotaPrice * priceData.CompletionRatio
+
+	cost := usage.Cost
+	totalPromptTokens := float64(usage.PromptTokens)
+	completionTokens := float64(usage.CompletionTokens)
+	promptCacheReadTokens := float64(usage.PromptTokensDetails.CachedTokens)
+
+	return int(math.Round((cost -
+		totalPromptTokens*quotaPrice +
+		promptCacheReadTokens*(quotaPrice-promptCacheReadPrice) -
+		completionTokens*completionPrice) /
+		(promptCacheCreatePrice - quotaPrice)))
+}
+
 func PostAudioConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
 	usage *dto.Usage, preConsumedQuota int, userQuota int, priceData helper.PriceData, extraContent string) {