merge upstream

Signed-off-by: wozulong <>
2025-11-12 17:13:43 +08:00 · 2024-05-15 12:36:29 +08:00
parent e0f80cdb8f c222bc8752
commit 5ffb520363
49 changed files with 2313 additions and 845 deletions
--- a/relay/channel/ai360/constants.go
+++ b/relay/channel/ai360/constants.go
@@ -6,3 +6,5 @@ var ModelList = []string{
 	"embedding_s1_v1",
 	"semantic_similarity_s1_v1",
 }
+
+var ChannelName = "ai360"
--- a/relay/channel/ollama/constants.go
+++ b/relay/channel/ollama/constants.go
@@ -1,5 +1,7 @@
 package ollama

-var ModelList []string
+var ModelList = []string{
+	"llama3-7b",
+}

 var ChannelName = "ollama"
--- a/relay/channel/openai/constant.go
+++ b/relay/channel/openai/constant.go
@@ -1,16 +1,14 @@
 package openai

 var ModelList = []string{
-	"gpt-4o", "gpt-4o-2024-05-13",
 	"gpt-3.5-turbo", "gpt-3.5-turbo-0301", "gpt-3.5-turbo-0613", "gpt-3.5-turbo-1106", "gpt-3.5-turbo-0125",
 	"gpt-3.5-turbo-16k", "gpt-3.5-turbo-16k-0613",
 	"gpt-3.5-turbo-instruct",
 	"gpt-4", "gpt-4-0314", "gpt-4-0613", "gpt-4-1106-preview", "gpt-4-0125-preview",
 	"gpt-4-32k", "gpt-4-32k-0314", "gpt-4-32k-0613",
-	"gpt-4-turbo",
-	"gpt-4-turbo-2024-04-09",
-	"gpt-4-turbo-preview",
+	"gpt-4-turbo-preview", "gpt-4-turbo", "gpt-4-turbo-2024-04-09",
 	"gpt-4-vision-preview",
+	"gpt-4o", "gpt-4o-2024-05-13",
 	"text-embedding-ada-002", "text-embedding-3-small", "text-embedding-3-large",
 	"text-curie-001", "text-babbage-001", "text-ada-001", "text-davinci-002", "text-davinci-003",
 	"text-moderation-latest", "text-moderation-stable",
--- a/relay/common/relay_info.go
+++ b/relay/common/relay_info.go
@@ -38,7 +38,7 @@ func GenRelayInfo(c *gin.Context) *RelayInfo {
 	tokenUnlimited := c.GetBool("token_unlimited_quota")
 	startTime := time.Now()

-	apiType := constant.ChannelType2APIType(channelType)
+	apiType, _ := constant.ChannelType2APIType(channelType)

 	info := &RelayInfo{
 		RelayMode:      constant.Path2RelayMode(c.Request.URL.Path),
--- a/relay/constant/api_type.go
+++ b/relay/constant/api_type.go
@@ -24,9 +24,11 @@ const (
 	APITypeDummy // this one is only for count, do not add any channel after this
 )

-func ChannelType2APIType(channelType int) int {
-	apiType := APITypeOpenAI
+func ChannelType2APIType(channelType int) (int, bool) {
+	apiType := -1
 	switch channelType {
+	case common.ChannelTypeOpenAI:
+		apiType = APITypeOpenAI
 	case common.ChannelTypeAnthropic:
 		apiType = APITypeAnthropic
 	case common.ChannelTypeBaidu:
@@ -56,5 +58,8 @@ func ChannelType2APIType(channelType int) int {
 	case common.ChannelTypeCohere:
 		apiType = APITypeCohere
 	}
-	return apiType
+	if apiType == -1 {
+		return APITypeOpenAI, false
+	}
+	return apiType, true
 }
--- a/relay/relay-audio.go
+++ b/relay/relay-audio.go
@@ -196,7 +196,10 @@ func AudioHelper(c *gin.Context, relayMode int) *dto.OpenAIErrorWithStatusCode {
 			if quota != 0 {
 				tokenName := c.GetString("token_name")
 				logContent := fmt.Sprintf("模型倍率 %.2f，分组倍率 %.2f", modelRatio, groupRatio)
-				model.RecordConsumeLog(ctx, userId, channelId, promptTokens, 0, audioRequest.Model, tokenName, quota, logContent, tokenId, userQuota, int(useTimeSeconds), false)
+				other := make(map[string]interface{})
+				other["model_ratio"] = modelRatio
+				other["group_ratio"] = groupRatio
+				model.RecordConsumeLog(ctx, userId, channelId, promptTokens, 0, audioRequest.Model, tokenName, quota, logContent, tokenId, userQuota, int(useTimeSeconds), false, other)
 				model.UpdateUserUsedQuotaAndRequestCount(userId, quota)
 				channelId := c.GetInt("channel_id")
 				model.UpdateChannelUsedQuota(channelId, quota)
--- a/relay/relay-image.go
+++ b/relay/relay-image.go
@@ -106,21 +106,26 @@ func RelayImageHelper(c *gin.Context, relayMode int) *dto.OpenAIErrorWithStatusC
 		requestBody = c.Request.Body
 	}

-	modelRatio := common.GetModelRatio(imageRequest.Model)
+	modelPrice, success := common.GetModelPrice(imageRequest.Model, true)
+	if !success {
+		modelRatio := common.GetModelRatio(imageRequest.Model)
+		// modelRatio 16 = modelPrice $0.04
+		// per 1 modelRatio = $0.04 / 16
+		modelPrice = 0.0025 * modelRatio
+	}
 	groupRatio := common.GetGroupRatio(group)
-	ratio := modelRatio * groupRatio
 	userQuota, err := model.CacheGetUserQuota(userId)

 	sizeRatio := 1.0
 	// Size
 	if imageRequest.Size == "256x256" {
-		sizeRatio = 1
+		sizeRatio = 0.4
 	} else if imageRequest.Size == "512x512" {
-		sizeRatio = 1.125
+		sizeRatio = 0.45
 	} else if imageRequest.Size == "1024x1024" {
-		sizeRatio = 1.25
+		sizeRatio = 1
 	} else if imageRequest.Size == "1024x1792" || imageRequest.Size == "1792x1024" {
-		sizeRatio = 2.5
+		sizeRatio = 2
 	}

 	qualityRatio := 1.0
@@ -131,7 +136,7 @@ func RelayImageHelper(c *gin.Context, relayMode int) *dto.OpenAIErrorWithStatusC
 		}
 	}

-	quota := int(ratio*sizeRatio*qualityRatio*1000) * imageRequest.N
+	quota := int(modelPrice*groupRatio*common.QuotaPerUnit*sizeRatio*qualityRatio) * imageRequest.N

 	if userQuota-quota < 0 {
 		return service.OpenAIErrorWrapper(errors.New("user quota is not enough"), "insufficient_user_quota", http.StatusForbidden)
@@ -190,8 +195,11 @@ func RelayImageHelper(c *gin.Context, relayMode int) *dto.OpenAIErrorWithStatusC
 			if imageRequest.Quality == "hd" {
 				quality = "hd"
 			}
-			logContent := fmt.Sprintf("模型倍率 %.2f，分组倍率 %.2f, 大小 %s, 品质 %s", modelRatio, groupRatio, imageRequest.Size, quality)
-			model.RecordConsumeLog(ctx, userId, channelId, 0, 0, imageRequest.Model, tokenName, quota, logContent, tokenId, userQuota, int(useTimeSeconds), false)
+			logContent := fmt.Sprintf("模型价格 %.2f，分组倍率 %.2f, 大小 %s, 品质 %s", modelPrice, groupRatio, imageRequest.Size, quality)
+			other := make(map[string]interface{})
+			other["model_price"] = modelPrice
+			other["group_ratio"] = groupRatio
+			model.RecordConsumeLog(ctx, userId, channelId, 0, 0, imageRequest.Model, tokenName, quota, logContent, tokenId, userQuota, int(useTimeSeconds), false, other)
 			model.UpdateUserUsedQuotaAndRequestCount(userId, quota)
 			channelId := c.GetInt("channel_id")
 			model.UpdateChannelUsedQuota(channelId, quota)
--- a/relay/relay-mj.go
+++ b/relay/relay-mj.go
@@ -155,9 +155,9 @@ func RelaySwapFace(c *gin.Context) *dto.MidjourneyResponse {
 		return service.MidjourneyErrorWrapper(constant.MjRequestError, "sour_base64_and_target_base64_is_required")
 	}
 	modelName := service.CoverActionToModelName(constant.MjActionSwapFace)
-	modelPrice := common.GetModelPrice(modelName, true)
+	modelPrice, success := common.GetModelPrice(modelName, true)
 	// 如果没有配置价格，则使用默认价格
-	if modelPrice == -1 {
+	if !success {
 		defaultPrice, ok := common.DefaultModelPrice[modelName]
 		if !ok {
 			modelPrice = 0.1
@@ -202,7 +202,10 @@ func RelaySwapFace(c *gin.Context) *dto.MidjourneyResponse {
 			if quota != 0 {
 				tokenName := c.GetString("token_name")
 				logContent := fmt.Sprintf("模型固定价格 %.2f，分组倍率 %.2f，操作 %s", modelPrice, groupRatio, constant.MjActionSwapFace)
-				model.RecordConsumeLog(ctx, userId, channelId, 0, 0, modelName, tokenName, quota, logContent, tokenId, userQuota, 0, false)
+				other := make(map[string]interface{})
+				other["model_price"] = modelPrice
+				other["group_ratio"] = groupRatio
+				model.RecordConsumeLog(ctx, userId, channelId, 0, 0, modelName, tokenName, quota, logContent, tokenId, userQuota, 0, false, other)
 				model.UpdateUserUsedQuotaAndRequestCount(userId, quota)
 				channelId := c.GetInt("channel_id")
 				model.UpdateChannelUsedQuota(channelId, quota)
@@ -451,9 +454,9 @@ func RelayMidjourneySubmit(c *gin.Context, relayMode int) *dto.MidjourneyRespons
 	fullRequestURL := fmt.Sprintf("%s%s", baseURL, requestURL)

 	modelName := service.CoverActionToModelName(midjRequest.Action)
-	modelPrice := common.GetModelPrice(modelName, true)
+	modelPrice, success := common.GetModelPrice(modelName, true)
 	// 如果没有配置价格，则使用默认价格
-	if modelPrice == -1 {
+	if !success {
 		defaultPrice, ok := common.DefaultModelPrice[modelName]
 		if !ok {
 			modelPrice = 0.1
@@ -498,7 +501,10 @@ func RelayMidjourneySubmit(c *gin.Context, relayMode int) *dto.MidjourneyRespons
 			if quota != 0 {
 				tokenName := c.GetString("token_name")
 				logContent := fmt.Sprintf("模型固定价格 %.2f，分组倍率 %.2f，操作 %s", modelPrice, groupRatio, midjRequest.Action)
-				model.RecordConsumeLog(ctx, userId, channelId, 0, 0, modelName, tokenName, quota, logContent, tokenId, userQuota, 0, false)
+				other := make(map[string]interface{})
+				other["model_price"] = modelPrice
+				other["group_ratio"] = groupRatio
+				model.RecordConsumeLog(ctx, userId, channelId, 0, 0, modelName, tokenName, quota, logContent, tokenId, userQuota, 0, false, other)
 				model.UpdateUserUsedQuotaAndRequestCount(userId, quota)
 				channelId := c.GetInt("channel_id")
 				model.UpdateChannelUsedQuota(channelId, quota)
--- a/relay/relay-text.go
+++ b/relay/relay-text.go
@@ -91,7 +91,7 @@ func TextHelper(c *gin.Context) *dto.OpenAIErrorWithStatusCode {
 		}
 	}
 	relayInfo.UpstreamModelName = textRequest.Model
-	modelPrice := common.GetModelPrice(textRequest.Model, false)
+	modelPrice, success := common.GetModelPrice(textRequest.Model, false)
 	groupRatio := common.GetGroupRatio(relayInfo.Group)

 	var preConsumedQuota int
@@ -108,7 +108,7 @@ func TextHelper(c *gin.Context) *dto.OpenAIErrorWithStatusCode {
 		return service.OpenAIErrorWrapper(err, "count_token_messages_failed", http.StatusInternalServerError)
 	}

-	if modelPrice == -1 {
+	if !success {
 		preConsumedTokens := common.PreConsumedQuota
 		if textRequest.MaxTokens != 0 {
 			preConsumedTokens = promptTokens + int(textRequest.MaxTokens)
@@ -178,7 +178,7 @@ func TextHelper(c *gin.Context) *dto.OpenAIErrorWithStatusCode {
 		service.ResetStatusCode(openaiErr, statusCodeMappingStr)
 		return openaiErr
 	}
-	postConsumeQuota(c, relayInfo, *textRequest, usage, ratio, preConsumedQuota, userQuota, modelRatio, groupRatio, modelPrice)
+	postConsumeQuota(c, relayInfo, *textRequest, usage, ratio, preConsumedQuota, userQuota, modelRatio, groupRatio, modelPrice, success)
 	return nil
 }

@@ -257,7 +257,7 @@ func returnPreConsumedQuota(c *gin.Context, tokenId int, userQuota int, preConsu

 func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, textRequest dto.GeneralOpenAIRequest,
 	usage *dto.Usage, ratio float64, preConsumedQuota int, userQuota int, modelRatio float64, groupRatio float64,
-	modelPrice float64) {
+	modelPrice float64, usePrice bool) {

 	useTimeSeconds := time.Now().Unix() - relayInfo.StartTime.Unix()
 	promptTokens := usage.PromptTokens
@@ -267,9 +267,9 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, textRe
 	completionRatio := common.GetCompletionRatio(textRequest.Model)

 	quota := 0
-	if modelPrice == -1 {
-		quota = promptTokens + int(float64(completionTokens)*completionRatio)
-		quota = int(float64(quota) * ratio)
+	if !usePrice {
+		quota = promptTokens + int(math.Round(float64(completionTokens)*completionRatio))
+		quota = int(math.Round(float64(quota) * ratio))
 		if ratio != 0 && quota <= 0 {
 			quota = 1
 		}
@@ -318,7 +318,12 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, textRe
 		logModel = "g-*"
 		logContent += fmt.Sprintf("，模型 %s", textRequest.Model)
 	}
-	model.RecordConsumeLog(ctx, relayInfo.UserId, relayInfo.ChannelId, promptTokens, completionTokens, logModel, tokenName, quota, logContent, relayInfo.TokenId, userQuota, int(useTimeSeconds), relayInfo.IsStream)
+	other := make(map[string]interface{})
+	other["model_ratio"] = modelRatio
+	other["group_ratio"] = groupRatio
+	other["completion_ratio"] = completionRatio
+	other["model_price"] = modelPrice
+	model.RecordConsumeLog(ctx, relayInfo.UserId, relayInfo.ChannelId, promptTokens, completionTokens, logModel, tokenName, quota, logContent, relayInfo.TokenId, userQuota, int(useTimeSeconds), relayInfo.IsStream, other)

 	//if quota != 0 {
 	//