merge upstream

Signed-off-by: wozulong <>

relay/channel/adapter.go

@@ -11,9 +11,11 @@ import (
 type Adaptor interface {
 	// Init IsStream bool
 	Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest)
+	InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest)
 	GetRequestURL(info *relaycommon.RelayInfo) (string, error)
 	SetupRequestHeader(c *gin.Context, req *http.Request, info *relaycommon.RelayInfo) error
 	ConvertRequest(c *gin.Context, relayMode int, request *dto.GeneralOpenAIRequest) (any, error)
+	ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error)
 	DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error)
 	DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage *dto.Usage, err *dto.OpenAIErrorWithStatusCode)
 	GetModelList() []string

relay/channel/ali/adaptor.go

@@ -15,6 +15,9 @@ import (
 type Adaptor struct {
 }
 
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 
 }
@@ -53,6 +56,10 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.Gen
 	}
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
 	return channel.DoApiRequest(a, c, info, requestBody)
 }

relay/channel/aws/adaptor.go

@@ -20,6 +20,11 @@ type Adaptor struct {
 	RequestMode int
 }
 
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+	//TODO implement me
+
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 	if strings.HasPrefix(info.UpstreamModelName, "claude-3") {
 		a.RequestMode = RequestModeMessage
@@ -53,13 +58,17 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.Gen
 	return claudeReq, err
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
 	return nil, nil
 }
 
 func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage *dto.Usage, err *dto.OpenAIErrorWithStatusCode) {
 	if info.IsStream {
-		err, usage = awsStreamHandler(c, info, a.RequestMode)
+		err, usage = awsStreamHandler(c, resp, info, a.RequestMode)
 	} else {
 		err, usage = awsHandler(c, info, a.RequestMode)
 	}

relay/channel/aws/relay-aws.go

@@ -13,6 +13,7 @@ import (
 	relaymodel "one-api/dto"
 	"one-api/relay/channel/claude"
 	relaycommon "one-api/relay/common"
+	"one-api/service"
 	"strings"
 	"time"
 
@@ -112,7 +113,7 @@ func awsHandler(c *gin.Context, info *relaycommon.RelayInfo, requestMode int) (*
 	return nil, &usage
 }
 
-func awsStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, requestMode int) (*relaymodel.OpenAIErrorWithStatusCode, *relaymodel.Usage) {
+func awsStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo, requestMode int) (*relaymodel.OpenAIErrorWithStatusCode, *relaymodel.Usage) {
 	awsCli, err := newAwsClient(c, info)
 	if err != nil {
 		return wrapErr(errors.Wrap(err, "newAwsClient")), nil
@@ -162,7 +163,6 @@ func awsStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, requestMode i
 	c.Stream(func(w io.Writer) bool {
 		event, ok := <-stream.Events()
 		if !ok {
-			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
 			return false
 		}
 
@@ -214,6 +214,17 @@ func awsStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, requestMode i
 			return false
 		}
 	})
-
+	if info.ShouldIncludeUsage {
+		response := service.GenerateFinalUsageResponse(id, createdTime, info.UpstreamModelName, usage)
+		err := service.ObjectData(c, response)
+		if err != nil {
+			common.SysError("send final response failed: " + err.Error())
+		}
+	}
+	service.Done(c)
+	err = resp.Body.Close()
+	if err != nil {
+		return service.OpenAIErrorWrapperLocal(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
 	return nil, &usage
 }

relay/channel/baidu/adaptor.go

@@ -2,6 +2,7 @@ package baidu
 
 import (
 	"errors"
+	"fmt"
 	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
@@ -15,44 +16,74 @@ import (
 type Adaptor struct {
 }
 
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+	//TODO implement me
+
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 
 }
 
 func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
-	var fullRequestURL string
-	switch info.UpstreamModelName {
-	case "ERNIE-Bot-4":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions_pro"
-	case "ERNIE-Bot-8K":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie_bot_8k"
-	case "ERNIE-Bot":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions"
-	case "ERNIE-Speed":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie_speed"
-	case "ERNIE-Bot-turbo":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/eb-instant"
-	case "BLOOMZ-7B":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/bloomz_7b1"
-	case "ERNIE-4.0-8K":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions_pro"
-	case "ERNIE-3.5-8K":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions"
-	case "ERNIE-Speed-8K":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie_speed"
-	case "ERNIE-Character-8K":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie-char-8k"
-	case "ERNIE-Functions-8K":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie-func-8k"
-	case "ERNIE-Lite-8K-0922":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/eb-instant"
-	case "Yi-34B-Chat":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/yi_34b_chat"
-	case "Embedding-V1":
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/embeddings/embedding-v1"
-	default:
-		fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/" + strings.ToLower(info.UpstreamModelName)
+	// https://cloud.baidu.com/doc/WENXINWORKSHOP/s/clntwmv7t
+	suffix := "chat/"
+	if strings.HasPrefix(info.UpstreamModelName, "Embedding") {
+		suffix = "embeddings/"
 	}
+	if strings.HasPrefix(info.UpstreamModelName, "bge-large") {
+		suffix = "embeddings/"
+	}
+	if strings.HasPrefix(info.UpstreamModelName, "tao-8k") {
+		suffix = "embeddings/"
+	}
+	switch info.UpstreamModelName {
+	case "ERNIE-4.0":
+		suffix += "completions_pro"
+	case "ERNIE-Bot-4":
+		suffix += "completions_pro"
+	case "ERNIE-Bot":
+		suffix += "completions"
+	case "ERNIE-Bot-turbo":
+		suffix += "eb-instant"
+	case "ERNIE-Speed":
+		suffix += "ernie_speed"
+	case "ERNIE-4.0-8K":
+		suffix += "completions_pro"
+	case "ERNIE-3.5-8K":
+		suffix += "completions"
+	case "ERNIE-3.5-8K-0205":
+		suffix += "ernie-3.5-8k-0205"
+	case "ERNIE-3.5-8K-1222":
+		suffix += "ernie-3.5-8k-1222"
+	case "ERNIE-Bot-8K":
+		suffix += "ernie_bot_8k"
+	case "ERNIE-3.5-4K-0205":
+		suffix += "ernie-3.5-4k-0205"
+	case "ERNIE-Speed-8K":
+		suffix += "ernie_speed"
+	case "ERNIE-Speed-128K":
+		suffix += "ernie-speed-128k"
+	case "ERNIE-Lite-8K-0922":
+		suffix += "eb-instant"
+	case "ERNIE-Lite-8K-0308":
+		suffix += "ernie-lite-8k"
+	case "ERNIE-Tiny-8K":
+		suffix += "ernie-tiny-8k"
+	case "BLOOMZ-7B":
+		suffix += "bloomz_7b1"
+	case "Embedding-V1":
+		suffix += "embedding-v1"
+	case "bge-large-zh":
+		suffix += "bge_large_zh"
+	case "bge-large-en":
+		suffix += "bge_large_en"
+	case "tao-8k":
+		suffix += "tao_8k"
+	default:
+		suffix += strings.ToLower(info.UpstreamModelName)
+	}
+	fullRequestURL := fmt.Sprintf("%s/rpc/2.0/ai_custom/v1/wenxinworkshop/%s", info.BaseUrl, suffix)
 	var accessToken string
 	var err error
 	if accessToken, err = getBaiduAccessToken(info.ApiKey); err != nil {
@@ -82,6 +113,10 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.Gen
 	}
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
 	return channel.DoApiRequest(a, c, info, requestBody)
 }

relay/channel/baidu/constants.go

@@ -1,20 +1,22 @@
 package baidu
 
 var ModelList = []string{
-	"ERNIE-3.5-8K",
 	"ERNIE-4.0-8K",
+	"ERNIE-3.5-8K",
+	"ERNIE-3.5-8K-0205",
+	"ERNIE-3.5-8K-1222",
+	"ERNIE-Bot-8K",
+	"ERNIE-3.5-4K-0205",
 	"ERNIE-Speed-8K",
 	"ERNIE-Speed-128K",
-	"ERNIE-Lite-8K",
+	"ERNIE-Lite-8K-0922",
+	"ERNIE-Lite-8K-0308",
 	"ERNIE-Tiny-8K",
-	"ERNIE-Character-8K",
-	"ERNIE-Functions-8K",
-	//"ERNIE-Bot-4",
-	//"ERNIE-Bot-8K",
-	//"ERNIE-Bot",
-	//"ERNIE-Speed",
-	//"ERNIE-Bot-turbo",
+	"BLOOMZ-7B",
 	"Embedding-V1",
+	"bge-large-zh",
+	"bge-large-en",
+	"tao-8k",
 }
 
 var ChannelName = "baidu"

relay/channel/baidu/dto.go

@@ -11,9 +11,16 @@ type BaiduMessage struct {
 }
 
 type BaiduChatRequest struct {
-	Messages []BaiduMessage `json:"messages"`
-	Stream   bool           `json:"stream"`
-	UserId   string         `json:"user_id,omitempty"`
+	Messages        []BaiduMessage `json:"messages"`
+	Temperature     float64        `json:"temperature,omitempty"`
+	TopP            float64        `json:"top_p,omitempty"`
+	PenaltyScore    float64        `json:"penalty_score,omitempty"`
+	Stream          bool           `json:"stream,omitempty"`
+	System          string         `json:"system,omitempty"`
+	DisableSearch   bool           `json:"disable_search,omitempty"`
+	EnableCitation  bool           `json:"enable_citation,omitempty"`
+	MaxOutputTokens *int           `json:"max_output_tokens,omitempty"`
+	UserId          string         `json:"user_id,omitempty"`
 }
 
 type Error struct {

relay/channel/baidu/relay-baidu.go

@@ -22,17 +22,33 @@ import (
 var baiduTokenStore sync.Map
 
 func requestOpenAI2Baidu(request dto.GeneralOpenAIRequest) *BaiduChatRequest {
-	messages := make([]BaiduMessage, 0, len(request.Messages))
+	baiduRequest := BaiduChatRequest{
+		Temperature:    request.Temperature,
+		TopP:           request.TopP,
+		PenaltyScore:   request.FrequencyPenalty,
+		Stream:         request.Stream,
+		DisableSearch:  false,
+		EnableCitation: false,
+		UserId:         request.User,
+	}
+	if request.MaxTokens != 0 {
+		maxTokens := int(request.MaxTokens)
+		if request.MaxTokens == 1 {
+			maxTokens = 2
+		}
+		baiduRequest.MaxOutputTokens = &maxTokens
+	}
 	for _, message := range request.Messages {
-		messages = append(messages, BaiduMessage{
-			Role:    message.Role,
-			Content: message.StringContent(),
-		})
-	}
-	return &BaiduChatRequest{
-		Messages: messages,
-		Stream:   request.Stream,
+		if message.Role == "system" {
+			baiduRequest.System = message.StringContent()
+		} else {
+			baiduRequest.Messages = append(baiduRequest.Messages, BaiduMessage{
+				Role:    message.Role,
+				Content: message.StringContent(),
+			})
+		}
 	}
+	return &baiduRequest
 }
 
 func responseBaidu2OpenAI(response *BaiduChatResponse) *dto.OpenAITextResponse {

relay/channel/claude/adaptor.go

@@ -21,6 +21,11 @@ type Adaptor struct {
 	RequestMode int
 }
 
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+	//TODO implement me
+
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 	if strings.HasPrefix(info.UpstreamModelName, "claude-3") {
 		a.RequestMode = RequestModeMessage
@@ -59,6 +64,10 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.Gen
 	}
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
 	return channel.DoApiRequest(a, c, info, requestBody)
 }

relay/channel/claude/relay-claude.go

@@ -330,22 +330,15 @@ func claudeStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.
 			response.Created = createdTime
 			response.Model = info.UpstreamModelName
 
-			jsonStr, err := json.Marshal(response)
+			err = service.ObjectData(c, response)
 			if err != nil {
-				common.SysError("error marshalling stream response: " + err.Error())
-				return true
+				common.SysError(err.Error())
 			}
-			c.Render(-1, common.CustomEvent{Data: "data: " + string(jsonStr)})
 			return true
 		case <-stopChan:
-			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
 			return false
 		}
 	})
-	err := resp.Body.Close()
-	if err != nil {
-		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
-	}
 	if requestMode == RequestModeCompletion {
 		usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, info.PromptTokens)
 	} else {
@@ -356,6 +349,18 @@ func claudeStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.
 			usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, usage.PromptTokens)
 		}
 	}
+	if info.ShouldIncludeUsage {
+		response := service.GenerateFinalUsageResponse(responseId, createdTime, info.UpstreamModelName, *usage)
+		err := service.ObjectData(c, response)
+		if err != nil {
+			common.SysError("send final response failed: " + err.Error())
+		}
+	}
+	service.Done(c)
+	err := resp.Body.Close()
+	if err != nil {
+		return service.OpenAIErrorWrapperLocal(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
 	return nil, usage
 }
 

relay/channel/cohere/adaptor.go

@@ -8,16 +8,24 @@ import (
 	"one-api/dto"
 	"one-api/relay/channel"
 	relaycommon "one-api/relay/common"
+	"one-api/relay/constant"
 )
 
 type Adaptor struct {
 }
 
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 }
 
 func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
-	return fmt.Sprintf("%s/v1/chat", info.BaseUrl), nil
+	if info.RelayMode == constant.RelayModeRerank {
+		return fmt.Sprintf("%s/v1/rerank", info.BaseUrl), nil
+	} else {
+		return fmt.Sprintf("%s/v1/chat", info.BaseUrl), nil
+	}
 }
 
 func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Request, info *relaycommon.RelayInfo) error {
@@ -34,11 +42,19 @@ func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, request
 	return channel.DoApiRequest(a, c, info, requestBody)
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return requestConvertRerank2Cohere(request), nil
+}
+
 func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage *dto.Usage, err *dto.OpenAIErrorWithStatusCode) {
-	if info.IsStream {
-		err, usage = cohereStreamHandler(c, resp, info)
+	if info.RelayMode == constant.RelayModeRerank {
+		err, usage = cohereRerankHandler(c, resp, info)
 	} else {
-		err, usage = cohereHandler(c, resp, info.UpstreamModelName, info.PromptTokens)
+		if info.IsStream {
+			err, usage = cohereStreamHandler(c, resp, info)
+		} else {
+			err, usage = cohereHandler(c, resp, info.UpstreamModelName, info.PromptTokens)
+		}
 	}
 	return
 }

relay/channel/cohere/constant.go

@@ -2,6 +2,7 @@ package cohere
 
 var ModelList = []string{
 	"command-r", "command-r-plus", "command-light", "command-light-nightly", "command", "command-nightly",
+	"rerank-english-v3.0", "rerank-multilingual-v3.0", "rerank-english-v2.0", "rerank-multilingual-v2.0",
 }
 
 var ChannelName = "cohere"

relay/channel/cohere/dto.go

@@ -1,5 +1,7 @@
 package cohere
 
+import "one-api/dto"
+
 type CohereRequest struct {
 	Model       string        `json:"model"`
 	ChatHistory []ChatHistory `json:"chat_history"`
@@ -28,6 +30,19 @@ type CohereResponseResult struct {
 	Meta         CohereMeta `json:"meta"`
 }
 
+type CohereRerankRequest struct {
+	Documents       []any  `json:"documents"`
+	Query           string `json:"query"`
+	Model           string `json:"model"`
+	TopN            int    `json:"top_n"`
+	ReturnDocuments bool   `json:"return_documents"`
+}
+
+type CohereRerankResponseResult struct {
+	Results []dto.RerankResponseDocument `json:"results"`
+	Meta    CohereMeta                   `json:"meta"`
+}
+
 type CohereMeta struct {
 	//Tokens CohereTokens `json:"tokens"`
 	BilledUnits CohereBilledUnits `json:"billed_units"`

relay/channel/cohere/relay-cohere.go

@@ -47,6 +47,20 @@ func requestOpenAI2Cohere(textRequest dto.GeneralOpenAIRequest) *CohereRequest {
 	return &cohereReq
 }
 
+func requestConvertRerank2Cohere(rerankRequest dto.RerankRequest) *CohereRerankRequest {
+	if rerankRequest.TopN == 0 {
+		rerankRequest.TopN = 1
+	}
+	cohereReq := CohereRerankRequest{
+		Query:           rerankRequest.Query,
+		Documents:       rerankRequest.Documents,
+		Model:           rerankRequest.Model,
+		TopN:            rerankRequest.TopN,
+		ReturnDocuments: true,
+	}
+	return &cohereReq
+}
+
 func stopReasonCohere2OpenAI(reason string) string {
 	switch reason {
 	case "COMPLETE":
@@ -194,3 +208,42 @@ func cohereHandler(c *gin.Context, resp *http.Response, modelName string, prompt
 	_, err = c.Writer.Write(jsonResponse)
 	return nil, &usage
 }
+
+func cohereRerankHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	var cohereResp CohereRerankResponseResult
+	err = json.Unmarshal(responseBody, &cohereResp)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	usage := dto.Usage{}
+	if cohereResp.Meta.BilledUnits.InputTokens == 0 {
+		usage.PromptTokens = info.PromptTokens
+		usage.CompletionTokens = 0
+		usage.TotalTokens = info.PromptTokens
+	} else {
+		usage.PromptTokens = cohereResp.Meta.BilledUnits.InputTokens
+		usage.CompletionTokens = cohereResp.Meta.BilledUnits.OutputTokens
+		usage.TotalTokens = cohereResp.Meta.BilledUnits.InputTokens + cohereResp.Meta.BilledUnits.OutputTokens
+	}
+
+	var rerankResp dto.RerankResponse
+	rerankResp.Results = cohereResp.Results
+	rerankResp.Usage = usage
+
+	jsonResponse, err := json.Marshal(rerankResp)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	c.Writer.Header().Set("Content-Type", "application/json")
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err = c.Writer.Write(jsonResponse)
+	return nil, &usage
+}

relay/channel/dify/adaptor.go

@@ -0,0 +1,65 @@
+package dify
+
+import (
+	"errors"
+	"fmt"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/dto"
+	"one-api/relay/channel"
+	relaycommon "one-api/relay/common"
+)
+
+type Adaptor struct {
+}
+
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+	//TODO implement me
+
+}
+
+func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
+}
+
+func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
+	return fmt.Sprintf("%s/v1/chat-messages", info.BaseUrl), nil
+}
+
+func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Request, info *relaycommon.RelayInfo) error {
+	channel.SetupApiRequestHeader(info, c, req)
+	req.Header.Set("Authorization", "Bearer "+info.ApiKey)
+	return nil
+}
+
+func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.GeneralOpenAIRequest) (any, error) {
+	if request == nil {
+		return nil, errors.New("request is nil")
+	}
+	return requestOpenAI2Dify(*request), nil
+}
+
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
+func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
+	return channel.DoApiRequest(a, c, info, requestBody)
+}
+
+func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage *dto.Usage, err *dto.OpenAIErrorWithStatusCode) {
+	if info.IsStream {
+		err, usage = difyStreamHandler(c, resp, info)
+	} else {
+		err, usage = difyHandler(c, resp, info)
+	}
+	return
+}
+
+func (a *Adaptor) GetModelList() []string {
+	return ModelList
+}
+
+func (a *Adaptor) GetChannelName() string {
+	return ChannelName
+}

relay/channel/dify/constants.go

@@ -0,0 +1,5 @@
+package dify
+
+var ModelList []string
+
+var ChannelName = "dify"

relay/channel/dify/dto.go

@@ -0,0 +1,35 @@
+package dify
+
+import "one-api/dto"
+
+type DifyChatRequest struct {
+	Inputs           map[string]interface{} `json:"inputs"`
+	Query            string                 `json:"query"`
+	ResponseMode     string                 `json:"response_mode"`
+	User             string                 `json:"user"`
+	AutoGenerateName bool                   `json:"auto_generate_name"`
+}
+
+type DifyMetaData struct {
+	Usage dto.Usage `json:"usage"`
+}
+
+type DifyData struct {
+	WorkflowId string `json:"workflow_id"`
+	NodeId     string `json:"node_id"`
+}
+
+type DifyChatCompletionResponse struct {
+	ConversationId string       `json:"conversation_id"`
+	Answer         string       `json:"answer"`
+	CreateAt       int64        `json:"create_at"`
+	MetaData       DifyMetaData `json:"metadata"`
+}
+
+type DifyChunkChatCompletionResponse struct {
+	Event          string       `json:"event"`
+	ConversationId string       `json:"conversation_id"`
+	Answer         string       `json:"answer"`
+	Data           DifyData     `json:"data"`
+	MetaData       DifyMetaData `json:"metadata"`
+}

relay/channel/dify/relay-dify.go

@@ -0,0 +1,156 @@
+package dify
+
+import (
+	"bufio"
+	"encoding/json"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
+	"one-api/constant"
+	"one-api/dto"
+	relaycommon "one-api/relay/common"
+	"one-api/service"
+	"strings"
+)
+
+func requestOpenAI2Dify(request dto.GeneralOpenAIRequest) *DifyChatRequest {
+	content := ""
+	for _, message := range request.Messages {
+		if message.Role == "system" {
+			content += "SYSTEM: \n" + message.StringContent() + "\n"
+		} else if message.Role == "assistant" {
+			content += "ASSISTANT: \n" + message.StringContent() + "\n"
+		} else {
+			content += "USER: \n" + message.StringContent() + "\n"
+		}
+	}
+	mode := "blocking"
+	if request.Stream {
+		mode = "streaming"
+	}
+	user := request.User
+	if user == "" {
+		user = "api-user"
+	}
+	return &DifyChatRequest{
+		Inputs:           make(map[string]interface{}),
+		Query:            content,
+		ResponseMode:     mode,
+		User:             user,
+		AutoGenerateName: false,
+	}
+}
+
+func streamResponseDify2OpenAI(difyResponse DifyChunkChatCompletionResponse) *dto.ChatCompletionsStreamResponse {
+	response := dto.ChatCompletionsStreamResponse{
+		Object:  "chat.completion.chunk",
+		Created: common.GetTimestamp(),
+		Model:   "dify",
+	}
+	var choice dto.ChatCompletionsStreamResponseChoice
+	if constant.DifyDebug && difyResponse.Event == "workflow_started" {
+		choice.Delta.SetContentString("Workflow: " + difyResponse.Data.WorkflowId + "\n")
+	} else if constant.DifyDebug && difyResponse.Event == "node_started" {
+		choice.Delta.SetContentString("Node: " + difyResponse.Data.NodeId + "\n")
+	} else if difyResponse.Event == "message" {
+		choice.Delta.SetContentString(difyResponse.Answer)
+	}
+	response.Choices = append(response.Choices, choice)
+	return &response
+}
+
+func difyStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
+	var responseText string
+	usage := &dto.Usage{}
+	scanner := bufio.NewScanner(resp.Body)
+	scanner.Split(bufio.ScanLines)
+
+	service.SetEventStreamHeaders(c)
+
+	for scanner.Scan() {
+		data := scanner.Text()
+		if len(data) < 5 || !strings.HasPrefix(data, "data:") {
+			continue
+		}
+		data = strings.TrimPrefix(data, "data:")
+		var difyResponse DifyChunkChatCompletionResponse
+		err := json.Unmarshal([]byte(data), &difyResponse)
+		if err != nil {
+			common.SysError("error unmarshalling stream response: " + err.Error())
+			continue
+		}
+		var openaiResponse dto.ChatCompletionsStreamResponse
+		if difyResponse.Event == "message_end" {
+			usage = &difyResponse.MetaData.Usage
+			break
+		} else if difyResponse.Event == "error" {
+			break
+		} else {
+			openaiResponse = *streamResponseDify2OpenAI(difyResponse)
+			if len(openaiResponse.Choices) != 0 {
+				responseText += openaiResponse.Choices[0].Delta.GetContentString()
+			}
+		}
+		err = service.ObjectData(c, openaiResponse)
+		if err != nil {
+			common.SysError(err.Error())
+		}
+	}
+	if err := scanner.Err(); err != nil {
+		common.SysError("error reading stream: " + err.Error())
+	}
+	service.Done(c)
+	err := resp.Body.Close()
+	if err != nil {
+		//return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+		common.SysError("close_response_body_failed: " + err.Error())
+	}
+	if usage.TotalTokens == 0 {
+		usage.PromptTokens = info.PromptTokens
+		usage.CompletionTokens, _ = service.CountTokenText("gpt-3.5-turbo", responseText)
+		usage.TotalTokens = usage.PromptTokens + usage.CompletionTokens
+	}
+	return nil, usage
+}
+
+func difyHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
+	var difyResponse DifyChatCompletionResponse
+	responseBody, err := io.ReadAll(resp.Body)
+
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = json.Unmarshal(responseBody, &difyResponse)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	fullTextResponse := dto.OpenAITextResponse{
+		Id:      difyResponse.ConversationId,
+		Object:  "chat.completion",
+		Created: common.GetTimestamp(),
+		Usage:   difyResponse.MetaData.Usage,
+	}
+	content, _ := json.Marshal(difyResponse.Answer)
+	choice := dto.OpenAITextResponseChoice{
+		Index: 0,
+		Message: dto.Message{
+			Role:    "assistant",
+			Content: content,
+		},
+		FinishReason: "stop",
+	}
+	fullTextResponse.Choices = append(fullTextResponse.Choices, choice)
+	jsonResponse, err := json.Marshal(fullTextResponse)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	c.Writer.Header().Set("Content-Type", "application/json")
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err = c.Writer.Write(jsonResponse)
+	return nil, &difyResponse.MetaData.Usage
+}

relay/channel/gemini/adaptor.go

@@ -9,12 +9,14 @@ import (
 	"one-api/dto"
 	"one-api/relay/channel"
 	relaycommon "one-api/relay/common"
-	"one-api/service"
 )
 
 type Adaptor struct {
 }
 
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 }
 
@@ -56,15 +58,17 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.Gen
 	return CovertGemini2OpenAI(*request), nil
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
 	return channel.DoApiRequest(a, c, info, requestBody)
 }
 
 func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage *dto.Usage, err *dto.OpenAIErrorWithStatusCode) {
 	if info.IsStream {
-		var responseText string
-		err, responseText = geminiChatStreamHandler(c, resp, info)
-		usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, info.PromptTokens)
+		err, usage = geminiChatStreamHandler(c, resp, info)
 	} else {
 		err, usage = geminiChatHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
 	}

relay/channel/gemini/dto.go

@@ -59,4 +59,11 @@ type GeminiChatPromptFeedback struct {
 type GeminiChatResponse struct {
 	Candidates     []GeminiChatCandidate    `json:"candidates"`
 	PromptFeedback GeminiChatPromptFeedback `json:"promptFeedback"`
+	UsageMetadata  GeminiUsageMetadata      `json:"usageMetadata"`
+}
+
+type GeminiUsageMetadata struct {
+	PromptTokenCount     int `json:"promptTokenCount"`
+	CandidatesTokenCount int `json:"candidatesTokenCount"`
+	TotalTokenCount      int `json:"totalTokenCount"`
 }

relay/channel/gemini/relay-gemini.go

@@ -5,6 +5,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"io"
+	"log"
 	"net/http"
 	"one-api/common"
 	"one-api/constant"
@@ -162,8 +163,12 @@ func streamResponseGeminiChat2OpenAI(geminiResponse *GeminiChatResponse) *dto.Ch
 	return &response
 }
 
-func geminiChatStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, string) {
+func geminiChatStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
 	responseText := ""
+	responseJson := ""
+	id := fmt.Sprintf("chatcmpl-%s", common.GetUUID())
+	createAt := common.GetTimestamp()
+	var usage = &dto.Usage{}
 	dataChan := make(chan string, 5)
 	stopChan := make(chan bool, 2)
 	scanner := bufio.NewScanner(resp.Body)
@@ -182,6 +187,7 @@ func geminiChatStreamHandler(c *gin.Context, resp *http.Response, info *relaycom
 	go func() {
 		for scanner.Scan() {
 			data := scanner.Text()
+			responseJson += data
 			data = strings.TrimSpace(data)
 			if !strings.HasPrefix(data, "\"text\": \"") {
 				continue
@@ -216,10 +222,10 @@ func geminiChatStreamHandler(c *gin.Context, resp *http.Response, info *relaycom
 			var choice dto.ChatCompletionsStreamResponseChoice
 			choice.Delta.SetContentString(dummy.Content)
 			response := dto.ChatCompletionsStreamResponse{
-				Id:      fmt.Sprintf("chatcmpl-%s", common.GetUUID()),
+				Id:      id,
 				Object:  "chat.completion.chunk",
-				Created: common.GetTimestamp(),
-				Model:   "gemini-pro",
+				Created: createAt,
+				Model:   info.UpstreamModelName,
 				Choices: []dto.ChatCompletionsStreamResponseChoice{choice},
 			}
 			jsonResponse, err := json.Marshal(response)
@@ -230,15 +236,34 @@ func geminiChatStreamHandler(c *gin.Context, resp *http.Response, info *relaycom
 			c.Render(-1, common.CustomEvent{Data: "data: " + string(jsonResponse)})
 			return true
 		case <-stopChan:
-			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
 			return false
 		}
 	})
-	err := resp.Body.Close()
+	var geminiChatResponses []GeminiChatResponse
+	err := json.Unmarshal([]byte(responseJson), &geminiChatResponses)
 	if err != nil {
-		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), ""
+		log.Printf("cannot get gemini usage: %s", err.Error())
+		usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, info.PromptTokens)
+	} else {
+		for _, response := range geminiChatResponses {
+			usage.PromptTokens = response.UsageMetadata.PromptTokenCount
+			usage.CompletionTokens = response.UsageMetadata.CandidatesTokenCount
+		}
+		usage.TotalTokens = usage.PromptTokens + usage.CompletionTokens
 	}
-	return nil, responseText
+	if info.ShouldIncludeUsage {
+		response := service.GenerateFinalUsageResponse(id, createAt, info.UpstreamModelName, *usage)
+		err := service.ObjectData(c, response)
+		if err != nil {
+			common.SysError("send final response failed: " + err.Error())
+		}
+	}
+	service.Done(c)
+	err = resp.Body.Close()
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), usage
+	}
+	return nil, usage
 }
 
 func geminiChatHandler(c *gin.Context, resp *http.Response, promptTokens int, model string) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
@@ -267,11 +292,10 @@ func geminiChatHandler(c *gin.Context, resp *http.Response, promptTokens int, mo
 		}, nil
 	}
 	fullTextResponse := responseGeminiChat2OpenAI(&geminiResponse)
-	completionTokens, _ := service.CountTokenText(geminiResponse.GetResponseText(), model)
 	usage := dto.Usage{
-		PromptTokens:     promptTokens,
-		CompletionTokens: completionTokens,
-		TotalTokens:      promptTokens + completionTokens,
+		PromptTokens:     geminiResponse.UsageMetadata.PromptTokenCount,
+		CompletionTokens: geminiResponse.UsageMetadata.CandidatesTokenCount,
+		TotalTokens:      geminiResponse.UsageMetadata.TotalTokenCount,
 	}
 	fullTextResponse.Usage = usage
 	jsonResponse, err := json.Marshal(fullTextResponse)

relay/channel/jina/adaptor.go

@@ -0,0 +1,64 @@
+package jina
+
+import (
+	"errors"
+	"fmt"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/dto"
+	"one-api/relay/channel"
+	relaycommon "one-api/relay/common"
+	"one-api/relay/constant"
+)
+
+type Adaptor struct {
+}
+
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+}
+
+func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
+}
+
+func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
+	if info.RelayMode == constant.RelayModeRerank {
+		return fmt.Sprintf("%s/v1/rerank", info.BaseUrl), nil
+	} else if info.RelayMode == constant.RelayModeEmbeddings {
+		return fmt.Sprintf("%s/v1/embeddings ", info.BaseUrl), nil
+	}
+	return "", errors.New("invalid relay mode")
+}
+
+func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Request, info *relaycommon.RelayInfo) error {
+	channel.SetupApiRequestHeader(info, c, req)
+	req.Header.Set("Authorization", fmt.Sprintf("Bearer %s", info.ApiKey))
+	return nil
+}
+
+func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.GeneralOpenAIRequest) (any, error) {
+	return request, nil
+}
+
+func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
+	return channel.DoApiRequest(a, c, info, requestBody)
+}
+
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return request, nil
+}
+
+func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage *dto.Usage, err *dto.OpenAIErrorWithStatusCode) {
+	if info.RelayMode == constant.RelayModeRerank {
+		err, usage = jinaRerankHandler(c, resp)
+	}
+	return
+}
+
+func (a *Adaptor) GetModelList() []string {
+	return ModelList
+}
+
+func (a *Adaptor) GetChannelName() string {
+	return ChannelName
+}

relay/channel/jina/constant.go

@@ -0,0 +1,8 @@
+package jina
+
+var ModelList = []string{
+	"jina-clip-v1",
+	"jina-reranker-v2-base-multilingual",
+}
+
+var ChannelName = "jina"

relay/channel/jina/relay-jina.go

@@ -0,0 +1,35 @@
+package jina
+
+import (
+	"encoding/json"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/dto"
+	"one-api/service"
+)
+
+func jinaRerankHandler(c *gin.Context, resp *http.Response) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	var jinaResp dto.RerankResponse
+	err = json.Unmarshal(responseBody, &jinaResp)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+
+	jsonResponse, err := json.Marshal(jinaResp)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	c.Writer.Header().Set("Content-Type", "application/json")
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err = c.Writer.Write(jsonResponse)
+	return nil, &jinaResp.Usage
+}

relay/channel/ollama/adaptor.go

@@ -16,6 +16,9 @@ import (
 type Adaptor struct {
 }
 
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 }
 
@@ -45,6 +48,10 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.Gen
 	}
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
 	return channel.DoApiRequest(a, c, info, requestBody)
 }
@@ -52,8 +59,10 @@ func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, request
 func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage *dto.Usage, err *dto.OpenAIErrorWithStatusCode) {
 	if info.IsStream {
 		var responseText string
-		err, responseText, _ = openai.OpenaiStreamHandler(c, resp, info)
-		usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, info.PromptTokens)
+		err, usage, responseText, _ = openai.OpenaiStreamHandler(c, resp, info)
+		if usage == nil || usage.TotalTokens == 0 || (usage.PromptTokens+usage.CompletionTokens) == 0 {
+			usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, info.PromptTokens)
+		}
 	} else {
 		if info.RelayMode == relayconstant.RelayModeEmbeddings {
 			err, usage = ollamaEmbeddingHandler(c, resp, info.PromptTokens, info.UpstreamModelName, info.RelayMode)

relay/channel/openai/adaptor.go

@@ -22,6 +22,13 @@ type Adaptor struct {
 	ChannelType int
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 	a.ChannelType = info.ChannelType
 }
@@ -82,9 +89,11 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
 	if info.IsStream {
 		var responseText string
 		var toolCount int
-		err, responseText, toolCount = OpenaiStreamHandler(c, resp, info)
-		usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, info.PromptTokens)
-		usage.CompletionTokens += toolCount * 7
+		err, usage, responseText, toolCount = OpenaiStreamHandler(c, resp, info)
+		if usage == nil || usage.TotalTokens == 0 || (usage.PromptTokens+usage.CompletionTokens) == 0 {
+			usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, info.PromptTokens)
+			usage.CompletionTokens += toolCount * 7
+		}
 	} else {
 		err, usage = OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
 	}

relay/channel/openai/relay-openai.go

@@ -18,9 +18,10 @@ import (
 	"time"
 )
 
-func OpenaiStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, string, int) {
+func OpenaiStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage, string, int) {
 	//checkSensitive := constant.ShouldCheckCompletionSensitive()
 	var responseTextBuilder strings.Builder
+	var usage dto.Usage
 	toolCount := 0
 	scanner := bufio.NewScanner(resp.Body)
 	scanner.Split(func(data []byte, atEOF bool) (advance int, token []byte, err error) {
@@ -62,17 +63,24 @@ func OpenaiStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.
 				streamItems = append(streamItems, data)
 			}
 		}
+		// 计算token
 		streamResp := "[" + strings.Join(streamItems, ",") + "]"
 		switch info.RelayMode {
 		case relayconstant.RelayModeChatCompletions:
 			var streamResponses []dto.ChatCompletionsStreamResponseSimple
 			err := json.Unmarshal(common.StringToByteSlice(streamResp), &streamResponses)
 			if err != nil {
+				// 一次性解析失败，逐个解析
 				common.SysError("error unmarshalling stream response: " + err.Error())
 				for _, item := range streamItems {
 					var streamResponse dto.ChatCompletionsStreamResponseSimple
 					err := json.Unmarshal(common.StringToByteSlice(item), &streamResponse)
 					if err == nil {
+						if streamResponse.Usage != nil {
+							if streamResponse.Usage.TotalTokens != 0 {
+								usage = *streamResponse.Usage
+							}
+						}
 						for _, choice := range streamResponse.Choices {
 							responseTextBuilder.WriteString(choice.Delta.GetContentString())
 							if choice.Delta.ToolCalls != nil {
@@ -89,6 +97,11 @@ func OpenaiStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.
 				}
 			} else {
 				for _, streamResponse := range streamResponses {
+					if streamResponse.Usage != nil {
+						if streamResponse.Usage.TotalTokens != 0 {
+							usage = *streamResponse.Usage
+						}
+					}
 					for _, choice := range streamResponse.Choices {
 						responseTextBuilder.WriteString(choice.Delta.GetContentString())
 						if choice.Delta.ToolCalls != nil {
@@ -107,6 +120,7 @@ func OpenaiStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.
 			var streamResponses []dto.CompletionsStreamResponse
 			err := json.Unmarshal(common.StringToByteSlice(streamResp), &streamResponses)
 			if err != nil {
+				// 一次性解析失败，逐个解析
 				common.SysError("error unmarshalling stream response: " + err.Error())
 				for _, item := range streamItems {
 					var streamResponse dto.CompletionsStreamResponse
@@ -133,13 +147,19 @@ func OpenaiStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.
 	}()
 	service.SetEventStreamHeaders(c)
 	isFirst := true
+	ticker := time.NewTicker(time.Duration(constant.StreamingTimeout) * time.Second)
+	defer ticker.Stop()
 	c.Stream(func(w io.Writer) bool {
 		select {
+		case <-ticker.C:
+			common.LogError(c, "reading data from upstream timeout")
+			return false
 		case data := <-dataChan:
 			if isFirst {
 				isFirst = false
 				info.FirstResponseTime = time.Now()
 			}
+			ticker.Reset(time.Duration(constant.StreamingTimeout) * time.Second)
 			if strings.HasPrefix(data, "data: [DONE]") {
 				data = data[:12]
 			}
@@ -153,10 +173,10 @@ func OpenaiStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.
 	})
 	err := resp.Body.Close()
 	if err != nil {
-		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), "", toolCount
+		return service.OpenAIErrorWrapperLocal(err, "close_response_body_failed", http.StatusInternalServerError), nil, "", toolCount
 	}
 	wg.Wait()
-	return nil, responseTextBuilder.String(), toolCount
+	return nil, &usage, responseTextBuilder.String(), toolCount
 }
 
 func OpenaiHandler(c *gin.Context, resp *http.Response, promptTokens int, model string) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {

relay/channel/palm/adaptor.go

@@ -15,6 +15,11 @@ import (
 type Adaptor struct {
 }
 
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+	//TODO implement me
+
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 }
 
@@ -35,6 +40,10 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.Gen
 	return request, nil
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
 	return channel.DoApiRequest(a, c, info, requestBody)
 }

relay/channel/perplexity/adaptor.go

@@ -16,6 +16,11 @@ import (
 type Adaptor struct {
 }
 
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+	//TODO implement me
+
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 }
 
@@ -39,6 +44,10 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.Gen
 	return requestOpenAI2Perplexity(*request), nil
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
 	return channel.DoApiRequest(a, c, info, requestBody)
 }
@@ -46,8 +55,10 @@ func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, request
 func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage *dto.Usage, err *dto.OpenAIErrorWithStatusCode) {
 	if info.IsStream {
 		var responseText string
-		err, responseText, _ = openai.OpenaiStreamHandler(c, resp, info)
-		usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, info.PromptTokens)
+		err, usage, responseText, _ = openai.OpenaiStreamHandler(c, resp, info)
+		if usage == nil || usage.TotalTokens == 0 || (usage.PromptTokens+usage.CompletionTokens) == 0 {
+			usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, info.PromptTokens)
+		}
 	} else {
 		err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
 	}

relay/channel/tencent/adaptor.go

@@ -6,28 +6,44 @@ import (
 	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
+	"one-api/common"
 	"one-api/dto"
 	"one-api/relay/channel"
 	relaycommon "one-api/relay/common"
 	"one-api/service"
+	"strconv"
 	"strings"
 )
 
 type Adaptor struct {
-	Sign string
+	Sign      string
+	AppID     int64
+	Action    string
+	Version   string
+	Timestamp int64
+}
+
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+	//TODO implement me
+
 }
 
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
+	a.Action = "ChatCompletions"
+	a.Version = "2023-09-01"
+	a.Timestamp = common.GetTimestamp()
 }
 
 func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
-	return fmt.Sprintf("%s/hyllm/v1/chat/completions", info.BaseUrl), nil
+	return fmt.Sprintf("%s/", info.BaseUrl), nil
 }
 
 func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Request, info *relaycommon.RelayInfo) error {
 	channel.SetupApiRequestHeader(info, c, req)
 	req.Header.Set("Authorization", a.Sign)
-	req.Header.Set("X-TC-Action", info.UpstreamModelName)
+	req.Header.Set("X-TC-Action", a.Action)
+	req.Header.Set("X-TC-Version", a.Version)
+	req.Header.Set("X-TC-Timestamp", strconv.FormatInt(a.Timestamp, 10))
 	return nil
 }
 
@@ -38,17 +54,20 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.Gen
 	apiKey := c.Request.Header.Get("Authorization")
 	apiKey = strings.TrimPrefix(apiKey, "Bearer ")
 	appId, secretId, secretKey, err := parseTencentConfig(apiKey)
+	a.AppID = appId
 	if err != nil {
 		return nil, err
 	}
-	tencentRequest := requestOpenAI2Tencent(*request)
-	tencentRequest.AppId = appId
-	tencentRequest.SecretId = secretId
+	tencentRequest := requestOpenAI2Tencent(a, *request)
 	// we have to calculate the sign here
-	a.Sign = getTencentSign(*tencentRequest, secretKey)
+	a.Sign = getTencentSign(*tencentRequest, a, secretId, secretKey)
 	return tencentRequest, nil
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
 	return channel.DoApiRequest(a, c, info, requestBody)
 }

relay/channel/tencent/constants.go

@@ -1,9 +1,10 @@
 package tencent
 
 var ModelList = []string{
-	"ChatPro",
-	"ChatStd",
-	"hunyuan",
+	"hunyuan-lite",
+	"hunyuan-standard",
+	"hunyuan-standard-256K",
+	"hunyuan-pro",
 }
 
 var ChannelName = "tencent"

relay/channel/tencent/dto.go

@@ -1,62 +1,75 @@
 package tencent
 
-import "one-api/dto"
-
 type TencentMessage struct {
-	Role    string `json:"role"`
-	Content string `json:"content"`
+	Role    string `json:"Role"`
+	Content string `json:"Content"`
 }
 
 type TencentChatRequest struct {
-	AppId    int64  `json:"app_id"`    // 腾讯云账号的 APPID
-	SecretId string `json:"secret_id"` // 官网 SecretId
-	// Timestamp当前 UNIX 时间戳，单位为秒，可记录发起 API 请求的时间。
-	// 例如1529223702，如果与当前时间相差过大，会引起签名过期错误
-	Timestamp int64 `json:"timestamp"`
-	// Expired 签名的有效期，是一个符合 UNIX Epoch 时间戳规范的数值，
-	// 单位为秒；Expired 必须大于 Timestamp 且 Expired-Timestamp 小于90天
-	Expired int64  `json:"expired"`
-	QueryID string `json:"query_id"` //请求 Id，用于问题排查
-	// Temperature 较高的数值会使输出更加随机，而较低的数值会使其更加集中和确定
-	// 默认 1.0，取值区间为[0.0,2.0]，非必要不建议使用,不合理的取值会影响效果
-	// 建议该参数和 top_p 只设置1个，不要同时更改 top_p
-	Temperature float64 `json:"temperature"`
-	// TopP 影响输出文本的多样性，取值越大，生成文本的多样性越强
-	// 默认1.0，取值区间为[0.0, 1.0]，非必要不建议使用, 不合理的取值会影响效果
-	// 建议该参数和 temperature 只设置1个，不要同时更改
-	TopP float64 `json:"top_p"`
-	// Stream 0：同步，1：流式 （默认，协议：SSE)
-	// 同步请求超时：60s，如果内容较长建议使用流式
-	Stream int `json:"stream"`
-	// Messages 会话内容, 长度最多为40, 按对话时间从旧到新在数组中排列
-	// 输入 content 总数最大支持 3000 token。
-	Messages []TencentMessage `json:"messages"`
-	Model    string           `json:"model"` // 模型名称
+	// 模型名称，可选值包括 hunyuan-lite、hunyuan-standard、hunyuan-standard-256K、hunyuan-pro。
+	// 各模型介绍请阅读 [产品概述](https://cloud.tencent.com/document/product/1729/104753) 中的说明。
+	//
+	// 注意：
+	// 不同的模型计费不同，请根据 [购买指南](https://cloud.tencent.com/document/product/1729/97731) 按需调用。
+	Model *string `json:"Model"`
+	// 聊天上下文信息。
+	// 说明：
+	// 1. 长度最多为 40，按对话时间从旧到新在数组中排列。
+	// 2. Message.Role 可选值：system、user、assistant。
+	// 其中，system 角色可选，如存在则必须位于列表的最开始。user 和 assistant 需交替出现（一问一答），以 user 提问开始和结束，且 Content 不能为空。Role 的顺序示例：[system（可选） user assistant user assistant user ...]。
+	// 3. Messages 中 Content 总长度不能超过模型输入长度上限（可参考 [产品概述](https://cloud.tencent.com/document/product/1729/104753) 文档），超过则会截断最前面的内容，只保留尾部内容。
+	Messages []*TencentMessage `json:"Messages"`
+	// 流式调用开关。
+	// 说明：
+	// 1. 未传值时默认为非流式调用（false）。
+	// 2. 流式调用时以 SSE 协议增量返回结果（返回值取 Choices[n].Delta 中的值，需要拼接增量数据才能获得完整结果）。
+	// 3. 非流式调用时：
+	// 调用方式与普通 HTTP 请求无异。
+	// 接口响应耗时较长，**如需更低时延建议设置为 true**。
+	// 只返回一次最终结果（返回值取 Choices[n].Message 中的值）。
+	//
+	// 注意：
+	// 通过 SDK 调用时，流式和非流式调用需用**不同的方式**获取返回值，具体参考 SDK 中的注释或示例（在各语言 SDK 代码仓库的 examples/hunyuan/v20230901/ 目录中）。
+	Stream *bool `json:"Stream,omitempty"`
+	// 说明：
+	// 1. 影响输出文本的多样性，取值越大，生成文本的多样性越强。
+	// 2. 取值区间为 [0.0, 1.0]，未传值时使用各模型推荐值。
+	// 3. 非必要不建议使用，不合理的取值会影响效果。
+	TopP *float64 `json:"TopP,omitempty"`
+	// 说明：
+	// 1. 较高的数值会使输出更加随机，而较低的数值会使其更加集中和确定。
+	// 2. 取值区间为 [0.0, 2.0]，未传值时使用各模型推荐值。
+	// 3. 非必要不建议使用，不合理的取值会影响效果。
+	Temperature *float64 `json:"Temperature,omitempty"`
 }
 
 type TencentError struct {
-	Code    int    `json:"code"`
-	Message string `json:"message"`
+	Code    int    `json:"Code"`
+	Message string `json:"Message"`
 }
 
 type TencentUsage struct {
-	InputTokens  int `json:"input_tokens"`
-	OutputTokens int `json:"output_tokens"`
-	TotalTokens  int `json:"total_tokens"`
+	PromptTokens     int `json:"PromptTokens"`
+	CompletionTokens int `json:"CompletionTokens"`
+	TotalTokens      int `json:"TotalTokens"`
 }
 
 type TencentResponseChoices struct {
-	FinishReason string         `json:"finish_reason,omitempty"` // 流式结束标志位，为 stop 则表示尾包
-	Messages     TencentMessage `json:"messages,omitempty"`      // 内容，同步模式返回内容，流模式为 null 输出 content 内容总数最多支持 1024token。
-	Delta        TencentMessage `json:"delta,omitempty"`         // 内容，流模式返回内容，同步模式为 null 输出 content 内容总数最多支持 1024token。
+	FinishReason string         `json:"FinishReason,omitempty"` // 流式结束标志位，为 stop 则表示尾包
+	Messages     TencentMessage `json:"Message,omitempty"`      // 内容，同步模式返回内容，流模式为 null 输出 content 内容总数最多支持 1024token。
+	Delta        TencentMessage `json:"Delta,omitempty"`        // 内容，流模式返回内容，同步模式为 null 输出 content 内容总数最多支持 1024token。
 }
 
 type TencentChatResponse struct {
-	Choices []TencentResponseChoices `json:"choices,omitempty"` // 结果
-	Created string                   `json:"created,omitempty"` // unix 时间戳的字符串
-	Id      string                   `json:"id,omitempty"`      // 会话 id
-	Usage   dto.Usage                `json:"usage,omitempty"`   // token 数量
-	Error   TencentError             `json:"error,omitempty"`   // 错误信息 注意：此字段可能返回 null，表示取不到有效值
-	Note    string                   `json:"note,omitempty"`    // 注释
-	ReqID   string                   `json:"req_id,omitempty"`  // 唯一请求 Id，每次请求都会返回。用于反馈接口入参
+	Choices []TencentResponseChoices `json:"Choices,omitempty"` // 结果
+	Created int64                    `json:"Created,omitempty"` // unix 时间戳的字符串
+	Id      string                   `json:"Id,omitempty"`      // 会话 id
+	Usage   TencentUsage             `json:"Usage,omitempty"`   // token 数量
+	Error   TencentError             `json:"Error,omitempty"`   // 错误信息 注意：此字段可能返回 null，表示取不到有效值
+	Note    string                   `json:"Note,omitempty"`    // 注释
+	ReqID   string                   `json:"Req_id,omitempty"`  // 唯一请求 Id，每次请求都会返回。用于反馈接口入参
+}
+
+type TencentChatResponseSB struct {
+	Response TencentChatResponse `json:"Response,omitempty"`
 }

relay/channel/tencent/relay-tencent.go

@@ -3,8 +3,8 @@ package tencent
 import (
 	"bufio"
 	"crypto/hmac"
-	"crypto/sha1"
-	"encoding/base64"
+	"crypto/sha256"
+	"encoding/hex"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -15,54 +15,46 @@ import (
 	"one-api/dto"
 	relaycommon "one-api/relay/common"
 	"one-api/service"
-	"sort"
 	"strconv"
 	"strings"
+	"time"
 )
 
 // https://cloud.tencent.com/document/product/1729/97732
 
-func requestOpenAI2Tencent(request dto.GeneralOpenAIRequest) *TencentChatRequest {
-	messages := make([]TencentMessage, 0, len(request.Messages))
+func requestOpenAI2Tencent(a *Adaptor, request dto.GeneralOpenAIRequest) *TencentChatRequest {
+	messages := make([]*TencentMessage, 0, len(request.Messages))
 	for i := 0; i < len(request.Messages); i++ {
 		message := request.Messages[i]
-		if message.Role == "system" {
-			messages = append(messages, TencentMessage{
-				Role:    "user",
-				Content: message.StringContent(),
-			})
-			messages = append(messages, TencentMessage{
-				Role:    "assistant",
-				Content: "Okay",
-			})
-			continue
-		}
-		messages = append(messages, TencentMessage{
+		messages = append(messages, &TencentMessage{
 			Content: message.StringContent(),
 			Role:    message.Role,
 		})
 	}
-	stream := 0
-	if request.Stream {
-		stream = 1
+	var req = TencentChatRequest{
+		Stream:   &request.Stream,
+		Messages: messages,
+		Model:    &request.Model,
 	}
-	return &TencentChatRequest{
-		Timestamp:   common.GetTimestamp(),
-		Expired:     common.GetTimestamp() + 24*60*60,
-		QueryID:     common.GetUUID(),
-		Temperature: request.Temperature,
-		TopP:        request.TopP,
-		Stream:      stream,
-		Messages:    messages,
-		Model:       request.Model,
+	if request.TopP != 0 {
+		req.TopP = &request.TopP
 	}
+	if request.Temperature != 0 {
+		req.Temperature = &request.Temperature
+	}
+	return &req
 }
 
 func responseTencent2OpenAI(response *TencentChatResponse) *dto.OpenAITextResponse {
 	fullTextResponse := dto.OpenAITextResponse{
+		Id:      response.Id,
 		Object:  "chat.completion",
 		Created: common.GetTimestamp(),
-		Usage:   response.Usage,
+		Usage: dto.Usage{
+			PromptTokens:     response.Usage.PromptTokens,
+			CompletionTokens: response.Usage.CompletionTokens,
+			TotalTokens:      response.Usage.TotalTokens,
+		},
 	}
 	if len(response.Choices) > 0 {
 		content, _ := json.Marshal(response.Choices[0].Messages.Content)
@@ -99,69 +91,51 @@ func streamResponseTencent2OpenAI(TencentResponse *TencentChatResponse) *dto.Cha
 func tencentStreamHandler(c *gin.Context, resp *http.Response) (*dto.OpenAIErrorWithStatusCode, string) {
 	var responseText string
 	scanner := bufio.NewScanner(resp.Body)
-	scanner.Split(func(data []byte, atEOF bool) (advance int, token []byte, err error) {
-		if atEOF && len(data) == 0 {
-			return 0, nil, nil
-		}
-		if i := strings.Index(string(data), "\n"); i >= 0 {
-			return i + 1, data[0:i], nil
-		}
-		if atEOF {
-			return len(data), data, nil
-		}
-		return 0, nil, nil
-	})
-	dataChan := make(chan string)
-	stopChan := make(chan bool)
-	go func() {
-		for scanner.Scan() {
-			data := scanner.Text()
-			if len(data) < 5 { // ignore blank line or wrong format
-				continue
-			}
-			if data[:5] != "data:" {
-				continue
-			}
-			data = data[5:]
-			dataChan <- data
-		}
-		stopChan <- true
-	}()
+	scanner.Split(bufio.ScanLines)
+
 	service.SetEventStreamHeaders(c)
-	c.Stream(func(w io.Writer) bool {
-		select {
-		case data := <-dataChan:
-			var TencentResponse TencentChatResponse
-			err := json.Unmarshal([]byte(data), &TencentResponse)
-			if err != nil {
-				common.SysError("error unmarshalling stream response: " + err.Error())
-				return true
-			}
-			response := streamResponseTencent2OpenAI(&TencentResponse)
-			if len(response.Choices) != 0 {
-				responseText += response.Choices[0].Delta.GetContentString()
-			}
-			jsonResponse, err := json.Marshal(response)
-			if err != nil {
-				common.SysError("error marshalling stream response: " + err.Error())
-				return true
-			}
-			c.Render(-1, common.CustomEvent{Data: "data: " + string(jsonResponse)})
-			return true
-		case <-stopChan:
-			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
-			return false
+
+	for scanner.Scan() {
+		data := scanner.Text()
+		if len(data) < 5 || !strings.HasPrefix(data, "data:") {
+			continue
 		}
-	})
+		data = strings.TrimPrefix(data, "data:")
+
+		var tencentResponse TencentChatResponse
+		err := json.Unmarshal([]byte(data), &tencentResponse)
+		if err != nil {
+			common.SysError("error unmarshalling stream response: " + err.Error())
+			continue
+		}
+
+		response := streamResponseTencent2OpenAI(&tencentResponse)
+		if len(response.Choices) != 0 {
+			responseText += response.Choices[0].Delta.GetContentString()
+		}
+
+		err = service.ObjectData(c, response)
+		if err != nil {
+			common.SysError(err.Error())
+		}
+	}
+
+	if err := scanner.Err(); err != nil {
+		common.SysError("error reading stream: " + err.Error())
+	}
+
+	service.Done(c)
+
 	err := resp.Body.Close()
 	if err != nil {
 		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), ""
 	}
+
 	return nil, responseText
 }
 
 func tencentHandler(c *gin.Context, resp *http.Response) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
-	var TencentResponse TencentChatResponse
+	var tencentSb TencentChatResponseSB
 	responseBody, err := io.ReadAll(resp.Body)
 	if err != nil {
 		return service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
@@ -170,20 +144,20 @@ func tencentHandler(c *gin.Context, resp *http.Response) (*dto.OpenAIErrorWithSt
 	if err != nil {
 		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
 	}
-	err = json.Unmarshal(responseBody, &TencentResponse)
+	err = json.Unmarshal(responseBody, &tencentSb)
 	if err != nil {
 		return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
 	}
-	if TencentResponse.Error.Code != 0 {
+	if tencentSb.Response.Error.Code != 0 {
 		return &dto.OpenAIErrorWithStatusCode{
 			Error: dto.OpenAIError{
-				Message: TencentResponse.Error.Message,
-				Code:    TencentResponse.Error.Code,
+				Message: tencentSb.Response.Error.Message,
+				Code:    tencentSb.Response.Error.Code,
 			},
 			StatusCode: resp.StatusCode,
 		}, nil
 	}
-	fullTextResponse := responseTencent2OpenAI(&TencentResponse)
+	fullTextResponse := responseTencent2OpenAI(&tencentSb.Response)
 	jsonResponse, err := json.Marshal(fullTextResponse)
 	if err != nil {
 		return service.OpenAIErrorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
@@ -206,29 +180,62 @@ func parseTencentConfig(config string) (appId int64, secretId string, secretKey
 	return
 }
 
-func getTencentSign(req TencentChatRequest, secretKey string) string {
-	params := make([]string, 0)
-	params = append(params, "app_id="+strconv.FormatInt(req.AppId, 10))
-	params = append(params, "secret_id="+req.SecretId)
-	params = append(params, "timestamp="+strconv.FormatInt(req.Timestamp, 10))
-	params = append(params, "query_id="+req.QueryID)
-	params = append(params, "temperature="+strconv.FormatFloat(req.Temperature, 'f', -1, 64))
-	params = append(params, "top_p="+strconv.FormatFloat(req.TopP, 'f', -1, 64))
-	params = append(params, "stream="+strconv.Itoa(req.Stream))
-	params = append(params, "expired="+strconv.FormatInt(req.Expired, 10))
-
-	var messageStr string
-	for _, msg := range req.Messages {
-		messageStr += fmt.Sprintf(`{"role":"%s","content":"%s"},`, msg.Role, msg.Content)
-	}
-	messageStr = strings.TrimSuffix(messageStr, ",")
-	params = append(params, "messages=["+messageStr+"]")
-
-	sort.Sort(sort.StringSlice(params))
-	url := "hunyuan.cloud.tencent.com/hyllm/v1/chat/completions?" + strings.Join(params, "&")
-	mac := hmac.New(sha1.New, []byte(secretKey))
-	signURL := url
-	mac.Write([]byte(signURL))
-	sign := mac.Sum([]byte(nil))
-	return base64.StdEncoding.EncodeToString(sign)
+func sha256hex(s string) string {
+	b := sha256.Sum256([]byte(s))
+	return hex.EncodeToString(b[:])
+}
+
+func hmacSha256(s, key string) string {
+	hashed := hmac.New(sha256.New, []byte(key))
+	hashed.Write([]byte(s))
+	return string(hashed.Sum(nil))
+}
+
+func getTencentSign(req TencentChatRequest, adaptor *Adaptor, secId, secKey string) string {
+	// build canonical request string
+	host := "hunyuan.tencentcloudapi.com"
+	httpRequestMethod := "POST"
+	canonicalURI := "/"
+	canonicalQueryString := ""
+	canonicalHeaders := fmt.Sprintf("content-type:%s\nhost:%s\nx-tc-action:%s\n",
+		"application/json", host, strings.ToLower(adaptor.Action))
+	signedHeaders := "content-type;host;x-tc-action"
+	payload, _ := json.Marshal(req)
+	hashedRequestPayload := sha256hex(string(payload))
+	canonicalRequest := fmt.Sprintf("%s\n%s\n%s\n%s\n%s\n%s",
+		httpRequestMethod,
+		canonicalURI,
+		canonicalQueryString,
+		canonicalHeaders,
+		signedHeaders,
+		hashedRequestPayload)
+	// build string to sign
+	algorithm := "TC3-HMAC-SHA256"
+	requestTimestamp := strconv.FormatInt(adaptor.Timestamp, 10)
+	timestamp, _ := strconv.ParseInt(requestTimestamp, 10, 64)
+	t := time.Unix(timestamp, 0).UTC()
+	// must be the format 2006-01-02, ref to package time for more info
+	date := t.Format("2006-01-02")
+	credentialScope := fmt.Sprintf("%s/%s/tc3_request", date, "hunyuan")
+	hashedCanonicalRequest := sha256hex(canonicalRequest)
+	string2sign := fmt.Sprintf("%s\n%s\n%s\n%s",
+		algorithm,
+		requestTimestamp,
+		credentialScope,
+		hashedCanonicalRequest)
+
+	// sign string
+	secretDate := hmacSha256(date, "TC3"+secKey)
+	secretService := hmacSha256("hunyuan", secretDate)
+	secretKey := hmacSha256("tc3_request", secretService)
+	signature := hex.EncodeToString([]byte(hmacSha256(string2sign, secretKey)))
+
+	// build authorization
+	authorization := fmt.Sprintf("%s Credential=%s/%s, SignedHeaders=%s, Signature=%s",
+		algorithm,
+		secId,
+		credentialScope,
+		signedHeaders,
+		signature)
+	return authorization
 }

relay/channel/xunfei/adaptor.go

@@ -16,6 +16,11 @@ type Adaptor struct {
 	request *dto.GeneralOpenAIRequest
 }
 
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+	//TODO implement me
+
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 }
 
@@ -36,6 +41,10 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.Gen
 	return request, nil
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
 	// xunfei's request is not http request, so we don't need to do anything here
 	dummyResp := &http.Response{}

relay/channel/zhipu/adaptor.go

@@ -14,6 +14,11 @@ import (
 type Adaptor struct {
 }
 
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+	//TODO implement me
+
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 }
 
@@ -42,6 +47,10 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.Gen
 	return requestOpenAI2Zhipu(*request), nil
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
 	return channel.DoApiRequest(a, c, info, requestBody)
 }

relay/channel/zhipu_4v/adaptor.go

@@ -16,6 +16,11 @@ import (
 type Adaptor struct {
 }
 
+func (a *Adaptor) InitRerank(info *relaycommon.RelayInfo, request dto.RerankRequest) {
+	//TODO implement me
+
+}
+
 func (a *Adaptor) Init(info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) {
 }
 
@@ -40,6 +45,10 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *dto.Gen
 	return requestOpenAI2Zhipu(*request), nil
 }
 
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
 	return channel.DoApiRequest(a, c, info, requestBody)
 }
@@ -48,9 +57,11 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
 	if info.IsStream {
 		var responseText string
 		var toolCount int
-		err, responseText, toolCount = openai.OpenaiStreamHandler(c, resp, info)
-		usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, info.PromptTokens)
-		usage.CompletionTokens += toolCount * 7
+		err, usage, responseText, toolCount = openai.OpenaiStreamHandler(c, resp, info)
+		if usage == nil || usage.TotalTokens == 0 || (usage.PromptTokens+usage.CompletionTokens) == 0 {
+			usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, info.PromptTokens)
+			usage.CompletionTokens += toolCount * 7
+		}
 	} else {
 		err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
 	}