feat: support zhipu's ChatGLM (close #289)

README.md

@@ -61,6 +61,10 @@ _✨ All in one 的 OpenAI 接口，整合各种 API 访问方式，开箱即用
 1. 支持多种 API 访问渠道：
    + [x] OpenAI 官方通道（支持配置镜像）
    + [x] **Azure OpenAI API**
+   + [x] [Anthropic Claude 系列模型](https://anthropic.com)
+   + [x] [Google PaLM2 系列模型](https://developers.generativeai.google)
+   + [x] [百度文心一言系列模型](https://cloud.baidu.com/doc/WENXINWORKSHOP/index.html)
+   + [x] [智谱 ChatGLM 系列模型](https://bigmodel.cn)
    + [x] [API Distribute](https://api.gptjk.top/register?aff=QGxj)
    + [x] [OpenAI-SB](https://openai-sb.com)
    + [x] [API2D](https://api2d.com/r/197971)
@@ -81,17 +85,17 @@ _✨ All in one 的 OpenAI 接口，整合各种 API 访问方式，开箱即用
 12. 支持以美元为单位显示额度。
 13. 支持发布公告，设置充值链接，设置新用户初始额度。
 14. 支持模型映射，重定向用户的请求模型。
-15. 支持绘图接口。
-16. 支持丰富的**自定义**设置，
+15. 支持失败自动重试。
+16. 支持绘图接口。
+17. 支持丰富的**自定义**设置，
     1. 支持自定义系统名称，logo 以及页脚。
     2. 支持自定义首页和关于页面，可以选择使用 HTML & Markdown 代码进行自定义，或者使用一个单独的网页通过 iframe 嵌入。
-17. 支持通过系统访问令牌访问管理 API。
-18. 支持 Cloudflare Turnstile 用户校验。
-19. 支持用户管理，支持**多种用户登录注册方式**：
+18. 支持通过系统访问令牌访问管理 API。
+19. 支持 Cloudflare Turnstile 用户校验。
+20. 支持用户管理，支持**多种用户登录注册方式**：
     + 邮箱登录注册以及通过邮箱进行密码重置。
     + [GitHub 开放授权](https://github.com/settings/applications/new)。
     + 微信公众号授权（需要额外部署 [WeChat Server](https://github.com/songquanpeng/wechat-server)）。
-20. 未来其他大模型开放 API 后，将第一时间支持，并将其封装成同样的 API 访问方式。
 
 ## 部署
 ### 基于 Docker 进行部署

common/constants.go

@@ -68,6 +68,7 @@ var AutomaticDisableChannelEnabled = false
 var QuotaRemindThreshold = 1000
 var PreConsumedQuota = 500
 var ApproximateTokenEnabled = false
+var RetryTimes = 0
 
 var RootUserEmail = ""
 
@@ -150,6 +151,9 @@ const (
 	ChannelTypePaLM      = 11
 	ChannelTypeAPI2GPT   = 12
 	ChannelTypeAIGC2D    = 13
+	ChannelTypeAnthropic = 14
+	ChannelTypeBaidu     = 15
+	ChannelTypeZhipu     = 16
 )
 
 var ChannelBaseURLs = []string{
@@ -167,4 +171,7 @@ var ChannelBaseURLs = []string{
 	"",                              // 11
 	"https://api.api2gpt.com",       // 12
 	"https://api.aigc2d.com",        // 13
+	"https://api.anthropic.com",     // 14
+	"https://aip.baidubce.com",      // 15
+	"https://open.bigmodel.cn",      // 16
 }

common/model-ratio.go

@@ -4,9 +4,11 @@ import "encoding/json"
 
 // ModelRatio
 // https://platform.openai.com/docs/models/model-endpoint-compatibility
+// https://cloud.baidu.com/doc/WENXINWORKSHOP/s/Blfmc9dlf
 // https://openai.com/pricing
 // TODO: when a new api is enabled, check the pricing here
 // 1 === $0.002 / 1K tokens
+// 1 === ￥0.014 / 1k tokens
 var ModelRatio = map[string]float64{
 	"gpt-4":                   15,
 	"gpt-4-0314":              15,
@@ -36,6 +38,14 @@ var ModelRatio = map[string]float64{
 	"text-moderation-stable":  0.1,
 	"text-moderation-latest":  0.1,
 	"dall-e":                  8,
+	"claude-instant-1":        0.75,
+	"claude-2":                30,
+	"ERNIE-Bot":               0.8572, // ￥0.012 / 1k tokens
+	"ERNIE-Bot-turbo":         0.5715, // ￥0.008 / 1k tokens
+	"PaLM-2":                  1,
+	"chatglm_pro":             0.7143, // ￥0.01 / 1k tokens
+	"chatglm_std":             0.3572, // ￥0.005 / 1k tokens
+	"chatglm_lite":            0.1429, // ￥0.002 / 1k tokens
 }
 
 func ModelRatio2JSONString() string {

controller/billing.go

@@ -7,16 +7,19 @@ import (
 )
 
 func GetSubscription(c *gin.Context) {
-	var quota int
+	var remainQuota int
+	var usedQuota int
 	var err error
 	var token *model.Token
 	if common.DisplayTokenStatEnabled {
 		tokenId := c.GetInt("token_id")
 		token, err = model.GetTokenById(tokenId)
-		quota = token.RemainQuota
+		remainQuota = token.RemainQuota
+		usedQuota = token.UsedQuota
 	} else {
 		userId := c.GetInt("id")
-		quota, err = model.GetUserQuota(userId)
+		remainQuota, err = model.GetUserQuota(userId)
+		usedQuota, err = model.GetUserUsedQuota(userId)
 	}
 	if err != nil {
 		openAIError := OpenAIError{
@@ -28,6 +31,7 @@ func GetSubscription(c *gin.Context) {
 		})
 		return
 	}
+	quota := remainQuota + usedQuota
 	amount := float64(quota)
 	if common.DisplayInCurrencyEnabled {
 		amount /= common.QuotaPerUnit

controller/channel-test.go

@@ -14,7 +14,7 @@ import (
 	"time"
 )
 
-func testChannel(channel *model.Channel, request ChatRequest) error {
+func testChannel(channel *model.Channel, request ChatRequest) (error, *OpenAIError) {
 	switch channel.Type {
 	case common.ChannelTypeAzure:
 		request.Model = "gpt-35-turbo"
@@ -33,11 +33,11 @@ func testChannel(channel *model.Channel, request ChatRequest) error {
 
 	jsonData, err := json.Marshal(request)
 	if err != nil {
-		return err
+		return err, nil
 	}
 	req, err := http.NewRequest("POST", requestURL, bytes.NewBuffer(jsonData))
 	if err != nil {
-		return err
+		return err, nil
 	}
 	if channel.Type == common.ChannelTypeAzure {
 		req.Header.Set("api-key", channel.Key)
@@ -48,18 +48,18 @@ func testChannel(channel *model.Channel, request ChatRequest) error {
 	client := &http.Client{}
 	resp, err := client.Do(req)
 	if err != nil {
-		return err
+		return err, nil
 	}
 	defer resp.Body.Close()
 	var response TextResponse
 	err = json.NewDecoder(resp.Body).Decode(&response)
 	if err != nil {
-		return err
+		return err, nil
 	}
 	if response.Usage.CompletionTokens == 0 {
-		return errors.New(fmt.Sprintf("type %s, code %v, message %s", response.Error.Type, response.Error.Code, response.Error.Message))
+		return errors.New(fmt.Sprintf("type %s, code %v, message %s", response.Error.Type, response.Error.Code, response.Error.Message)), &response.Error
 	}
-	return nil
+	return nil, nil
 }
 
 func buildTestRequest() *ChatRequest {
@@ -94,7 +94,7 @@ func TestChannel(c *gin.Context) {
 	}
 	testRequest := buildTestRequest()
 	tik := time.Now()
-	err = testChannel(channel, *testRequest)
+	err, _ = testChannel(channel, *testRequest)
 	tok := time.Now()
 	milliseconds := tok.Sub(tik).Milliseconds()
 	go channel.UpdateResponseTime(milliseconds)
@@ -158,13 +158,14 @@ func testAllChannels(notify bool) error {
 				continue
 			}
 			tik := time.Now()
-			err := testChannel(channel, *testRequest)
+			err, openaiErr := testChannel(channel, *testRequest)
 			tok := time.Now()
 			milliseconds := tok.Sub(tik).Milliseconds()
-			if err != nil || milliseconds > disableThreshold {
-				if milliseconds > disableThreshold {
-					err = errors.New(fmt.Sprintf("响应时间 %.2fs 超过阈值 %.2fs", float64(milliseconds)/1000.0, float64(disableThreshold)/1000.0))
-				}
+			if milliseconds > disableThreshold {
+				err = errors.New(fmt.Sprintf("响应时间 %.2fs 超过阈值 %.2fs", float64(milliseconds)/1000.0, float64(disableThreshold)/1000.0))
+				disableChannel(channel.Id, channel.Name, err.Error())
+			}
+			if shouldDisableChannel(openaiErr) {
 				disableChannel(channel.Id, channel.Name, err.Error())
 			}
 			channel.UpdateResponseTime(milliseconds)

controller/model.go

@@ -252,6 +252,78 @@ func init() {
 			Root:       "code-davinci-edit-001",
 			Parent:     nil,
 		},
+		{
+			Id:         "claude-instant-1",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "anturopic",
+			Permission: permission,
+			Root:       "claude-instant-1",
+			Parent:     nil,
+		},
+		{
+			Id:         "claude-2",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "anturopic",
+			Permission: permission,
+			Root:       "claude-2",
+			Parent:     nil,
+		},
+		{
+			Id:         "ERNIE-Bot",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "baidu",
+			Permission: permission,
+			Root:       "ERNIE-Bot",
+			Parent:     nil,
+		},
+		{
+			Id:         "ERNIE-Bot-turbo",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "baidu",
+			Permission: permission,
+			Root:       "ERNIE-Bot-turbo",
+			Parent:     nil,
+		},
+		{
+			Id:         "PaLM-2",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "google",
+			Permission: permission,
+			Root:       "PaLM-2",
+			Parent:     nil,
+		},
+		{
+			Id:         "chatglm_pro",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "zhipu",
+			Permission: permission,
+			Root:       "chatglm_pro",
+			Parent:     nil,
+		},
+		{
+			Id:         "chatglm_std",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "zhipu",
+			Permission: permission,
+			Root:       "chatglm_std",
+			Parent:     nil,
+		},
+		{
+			Id:         "chatglm_lite",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "zhipu",
+			Permission: permission,
+			Root:       "chatglm_lite",
+			Parent:     nil,
+		},
 	}
 	openAIModelsMap = make(map[string]OpenAIModels)
 	for _, model := range openAIModels {

controller/relay-baidu.go

@@ -0,0 +1,203 @@
+package controller
+
+import (
+	"bufio"
+	"encoding/json"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
+	"strings"
+)
+
+// https://cloud.baidu.com/doc/WENXINWORKSHOP/s/flfmc9do2
+
+type BaiduTokenResponse struct {
+	RefreshToken  string `json:"refresh_token"`
+	ExpiresIn     int    `json:"expires_in"`
+	SessionKey    string `json:"session_key"`
+	AccessToken   string `json:"access_token"`
+	Scope         string `json:"scope"`
+	SessionSecret string `json:"session_secret"`
+}
+
+type BaiduMessage struct {
+	Role    string `json:"role"`
+	Content string `json:"content"`
+}
+
+type BaiduChatRequest struct {
+	Messages []BaiduMessage `json:"messages"`
+	Stream   bool           `json:"stream"`
+	UserId   string         `json:"user_id,omitempty"`
+}
+
+type BaiduError struct {
+	ErrorCode int    `json:"error_code"`
+	ErrorMsg  string `json:"error_msg"`
+}
+
+type BaiduChatResponse struct {
+	Id               string `json:"id"`
+	Object           string `json:"object"`
+	Created          int64  `json:"created"`
+	Result           string `json:"result"`
+	IsTruncated      bool   `json:"is_truncated"`
+	NeedClearHistory bool   `json:"need_clear_history"`
+	Usage            Usage  `json:"usage"`
+	BaiduError
+}
+
+type BaiduChatStreamResponse struct {
+	BaiduChatResponse
+	SentenceId int  `json:"sentence_id"`
+	IsEnd      bool `json:"is_end"`
+}
+
+func requestOpenAI2Baidu(request GeneralOpenAIRequest) *BaiduChatRequest {
+	messages := make([]BaiduMessage, 0, len(request.Messages))
+	for _, message := range request.Messages {
+		messages = append(messages, BaiduMessage{
+			Role:    message.Role,
+			Content: message.Content,
+		})
+	}
+	return &BaiduChatRequest{
+		Messages: messages,
+		Stream:   request.Stream,
+	}
+}
+
+func responseBaidu2OpenAI(response *BaiduChatResponse) *OpenAITextResponse {
+	choice := OpenAITextResponseChoice{
+		Index: 0,
+		Message: Message{
+			Role:    "assistant",
+			Content: response.Result,
+		},
+		FinishReason: "stop",
+	}
+	fullTextResponse := OpenAITextResponse{
+		Id:      response.Id,
+		Object:  "chat.completion",
+		Created: response.Created,
+		Choices: []OpenAITextResponseChoice{choice},
+		Usage:   response.Usage,
+	}
+	return &fullTextResponse
+}
+
+func streamResponseBaidu2OpenAI(baiduResponse *BaiduChatStreamResponse) *ChatCompletionsStreamResponse {
+	var choice ChatCompletionsStreamResponseChoice
+	choice.Delta.Content = baiduResponse.Result
+	choice.FinishReason = "stop"
+	response := ChatCompletionsStreamResponse{
+		Id:      baiduResponse.Id,
+		Object:  "chat.completion.chunk",
+		Created: baiduResponse.Created,
+		Model:   "ernie-bot",
+		Choices: []ChatCompletionsStreamResponseChoice{choice},
+	}
+	return &response
+}
+
+func baiduStreamHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, *Usage) {
+	var usage Usage
+	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) < 6 { // ignore blank line or wrong format
+				continue
+			}
+			data = data[6:]
+			dataChan <- data
+		}
+		stopChan <- true
+	}()
+	c.Writer.Header().Set("Content-Type", "text/event-stream")
+	c.Writer.Header().Set("Cache-Control", "no-cache")
+	c.Writer.Header().Set("Connection", "keep-alive")
+	c.Writer.Header().Set("Transfer-Encoding", "chunked")
+	c.Writer.Header().Set("X-Accel-Buffering", "no")
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case data := <-dataChan:
+			var baiduResponse BaiduChatStreamResponse
+			err := json.Unmarshal([]byte(data), &baiduResponse)
+			if err != nil {
+				common.SysError("error unmarshalling stream response: " + err.Error())
+				return true
+			}
+			usage.PromptTokens += baiduResponse.Usage.PromptTokens
+			usage.CompletionTokens += baiduResponse.Usage.CompletionTokens
+			usage.TotalTokens += baiduResponse.Usage.TotalTokens
+			response := streamResponseBaidu2OpenAI(&baiduResponse)
+			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
+		}
+	})
+	err := resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	return nil, &usage
+}
+
+func baiduHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, *Usage) {
+	var baiduResponse BaiduChatResponse
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = json.Unmarshal(responseBody, &baiduResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if baiduResponse.ErrorMsg != "" {
+		return &OpenAIErrorWithStatusCode{
+			OpenAIError: OpenAIError{
+				Message: baiduResponse.ErrorMsg,
+				Type:    "baidu_error",
+				Param:   "",
+				Code:    baiduResponse.ErrorCode,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+	fullTextResponse := responseBaidu2OpenAI(&baiduResponse)
+	jsonResponse, err := json.Marshal(fullTextResponse)
+	if err != nil {
+		return errorWrapper(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, &fullTextResponse.Usage
+}

controller/relay-claude.go

@@ -0,0 +1,221 @@
+package controller
+
+import (
+	"bufio"
+	"encoding/json"
+	"fmt"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
+	"strings"
+)
+
+type ClaudeMetadata struct {
+	UserId string `json:"user_id"`
+}
+
+type ClaudeRequest struct {
+	Model             string   `json:"model"`
+	Prompt            string   `json:"prompt"`
+	MaxTokensToSample int      `json:"max_tokens_to_sample"`
+	StopSequences     []string `json:"stop_sequences,omitempty"`
+	Temperature       float64  `json:"temperature,omitempty"`
+	TopP              float64  `json:"top_p,omitempty"`
+	TopK              int      `json:"top_k,omitempty"`
+	//ClaudeMetadata    `json:"metadata,omitempty"`
+	Stream bool `json:"stream,omitempty"`
+}
+
+type ClaudeError struct {
+	Type    string `json:"type"`
+	Message string `json:"message"`
+}
+
+type ClaudeResponse struct {
+	Completion string      `json:"completion"`
+	StopReason string      `json:"stop_reason"`
+	Model      string      `json:"model"`
+	Error      ClaudeError `json:"error"`
+}
+
+func stopReasonClaude2OpenAI(reason string) string {
+	switch reason {
+	case "stop_sequence":
+		return "stop"
+	case "max_tokens":
+		return "length"
+	default:
+		return reason
+	}
+}
+
+func requestOpenAI2Claude(textRequest GeneralOpenAIRequest) *ClaudeRequest {
+	claudeRequest := ClaudeRequest{
+		Model:             textRequest.Model,
+		Prompt:            "",
+		MaxTokensToSample: textRequest.MaxTokens,
+		StopSequences:     nil,
+		Temperature:       textRequest.Temperature,
+		TopP:              textRequest.TopP,
+		Stream:            textRequest.Stream,
+	}
+	if claudeRequest.MaxTokensToSample == 0 {
+		claudeRequest.MaxTokensToSample = 1000000
+	}
+	prompt := ""
+	for _, message := range textRequest.Messages {
+		if message.Role == "user" {
+			prompt += fmt.Sprintf("\n\nHuman: %s", message.Content)
+		} else if message.Role == "assistant" {
+			prompt += fmt.Sprintf("\n\nAssistant: %s", message.Content)
+		} else {
+			// ignore other roles
+		}
+		prompt += "\n\nAssistant:"
+	}
+	claudeRequest.Prompt = prompt
+	return &claudeRequest
+}
+
+func streamResponseClaude2OpenAI(claudeResponse *ClaudeResponse) *ChatCompletionsStreamResponse {
+	var choice ChatCompletionsStreamResponseChoice
+	choice.Delta.Content = claudeResponse.Completion
+	choice.FinishReason = stopReasonClaude2OpenAI(claudeResponse.StopReason)
+	var response ChatCompletionsStreamResponse
+	response.Object = "chat.completion.chunk"
+	response.Model = claudeResponse.Model
+	response.Choices = []ChatCompletionsStreamResponseChoice{choice}
+	return &response
+}
+
+func responseClaude2OpenAI(claudeResponse *ClaudeResponse) *OpenAITextResponse {
+	choice := OpenAITextResponseChoice{
+		Index: 0,
+		Message: Message{
+			Role:    "assistant",
+			Content: strings.TrimPrefix(claudeResponse.Completion, " "),
+			Name:    nil,
+		},
+		FinishReason: stopReasonClaude2OpenAI(claudeResponse.StopReason),
+	}
+	fullTextResponse := OpenAITextResponse{
+		Id:      fmt.Sprintf("chatcmpl-%s", common.GetUUID()),
+		Object:  "chat.completion",
+		Created: common.GetTimestamp(),
+		Choices: []OpenAITextResponseChoice{choice},
+	}
+	return &fullTextResponse
+}
+
+func claudeStreamHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, string) {
+	responseText := ""
+	responseId := fmt.Sprintf("chatcmpl-%s", common.GetUUID())
+	createdTime := common.GetTimestamp()
+	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), "\r\n\r\n"); i >= 0 {
+			return i + 4, 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 !strings.HasPrefix(data, "event: completion") {
+				continue
+			}
+			data = strings.TrimPrefix(data, "event: completion\r\ndata: ")
+			dataChan <- data
+		}
+		stopChan <- true
+	}()
+	c.Writer.Header().Set("Content-Type", "text/event-stream")
+	c.Writer.Header().Set("Cache-Control", "no-cache")
+	c.Writer.Header().Set("Connection", "keep-alive")
+	c.Writer.Header().Set("Transfer-Encoding", "chunked")
+	c.Writer.Header().Set("X-Accel-Buffering", "no")
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case data := <-dataChan:
+			// some implementations may add \r at the end of data
+			data = strings.TrimSuffix(data, "\r")
+			var claudeResponse ClaudeResponse
+			err := json.Unmarshal([]byte(data), &claudeResponse)
+			if err != nil {
+				common.SysError("error unmarshalling stream response: " + err.Error())
+				return true
+			}
+			responseText += claudeResponse.Completion
+			response := streamResponseClaude2OpenAI(&claudeResponse)
+			response.Id = responseId
+			response.Created = createdTime
+			jsonStr, 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(jsonStr)})
+			return true
+		case <-stopChan:
+			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
+			return false
+		}
+	})
+	err := resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), ""
+	}
+	return nil, responseText
+}
+
+func claudeHandler(c *gin.Context, resp *http.Response, promptTokens int, model string) (*OpenAIErrorWithStatusCode, *Usage) {
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	var claudeResponse ClaudeResponse
+	err = json.Unmarshal(responseBody, &claudeResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if claudeResponse.Error.Type != "" {
+		return &OpenAIErrorWithStatusCode{
+			OpenAIError: OpenAIError{
+				Message: claudeResponse.Error.Message,
+				Type:    claudeResponse.Error.Type,
+				Param:   "",
+				Code:    claudeResponse.Error.Type,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+	fullTextResponse := responseClaude2OpenAI(&claudeResponse)
+	completionTokens := countTokenText(claudeResponse.Completion, model)
+	usage := Usage{
+		PromptTokens:     promptTokens,
+		CompletionTokens: completionTokens,
+		TotalTokens:      promptTokens + completionTokens,
+	}
+	fullTextResponse.Usage = usage
+	jsonResponse, err := json.Marshal(fullTextResponse)
+	if err != nil {
+		return errorWrapper(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
+}

controller/relay-openai.go

@@ -0,0 +1,133 @@
+package controller
+
+import (
+	"bufio"
+	"bytes"
+	"encoding/json"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
+	"strings"
+)
+
+func openaiStreamHandler(c *gin.Context, resp *http.Response, relayMode int) (*OpenAIErrorWithStatusCode, string) {
+	responseText := ""
+	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) < 6 { // ignore blank line or wrong format
+				continue
+			}
+			dataChan <- data
+			data = data[6:]
+			if !strings.HasPrefix(data, "[DONE]") {
+				switch relayMode {
+				case RelayModeChatCompletions:
+					var streamResponse ChatCompletionsStreamResponse
+					err := json.Unmarshal([]byte(data), &streamResponse)
+					if err != nil {
+						common.SysError("error unmarshalling stream response: " + err.Error())
+						return
+					}
+					for _, choice := range streamResponse.Choices {
+						responseText += choice.Delta.Content
+					}
+				case RelayModeCompletions:
+					var streamResponse CompletionsStreamResponse
+					err := json.Unmarshal([]byte(data), &streamResponse)
+					if err != nil {
+						common.SysError("error unmarshalling stream response: " + err.Error())
+						return
+					}
+					for _, choice := range streamResponse.Choices {
+						responseText += choice.Text
+					}
+				}
+			}
+		}
+		stopChan <- true
+	}()
+	c.Writer.Header().Set("Content-Type", "text/event-stream")
+	c.Writer.Header().Set("Cache-Control", "no-cache")
+	c.Writer.Header().Set("Connection", "keep-alive")
+	c.Writer.Header().Set("Transfer-Encoding", "chunked")
+	c.Writer.Header().Set("X-Accel-Buffering", "no")
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case data := <-dataChan:
+			if strings.HasPrefix(data, "data: [DONE]") {
+				data = data[:12]
+			}
+			// some implementations may add \r at the end of data
+			data = strings.TrimSuffix(data, "\r")
+			c.Render(-1, common.CustomEvent{Data: data})
+			return true
+		case <-stopChan:
+			return false
+		}
+	})
+	err := resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), ""
+	}
+	return nil, responseText
+}
+
+func openaiHandler(c *gin.Context, resp *http.Response, consumeQuota bool) (*OpenAIErrorWithStatusCode, *Usage) {
+	var textResponse TextResponse
+	if consumeQuota {
+		responseBody, err := io.ReadAll(resp.Body)
+		if err != nil {
+			return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+		}
+		err = resp.Body.Close()
+		if err != nil {
+			return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+		}
+		err = json.Unmarshal(responseBody, &textResponse)
+		if err != nil {
+			return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+		}
+		if textResponse.Error.Type != "" {
+			return &OpenAIErrorWithStatusCode{
+				OpenAIError: textResponse.Error,
+				StatusCode:  resp.StatusCode,
+			}, nil
+		}
+		// Reset response body
+		resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
+	}
+	// We shouldn't set the header before we parse the response body, because the parse part may fail.
+	// And then we will have to send an error response, but in this case, the header has already been set.
+	// So the client will be confused by the response.
+	// For example, Postman will report error, and we cannot check the response at all.
+	for k, v := range resp.Header {
+		c.Writer.Header().Set(k, v[0])
+	}
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err := io.Copy(c.Writer, resp.Body)
+	if err != nil {
+		return errorWrapper(err, "copy_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	return nil, &textResponse.Usage
+}

controller/relay-palm.go

@@ -1,10 +1,17 @@
 package controller
 
 import (
+	"encoding/json"
 	"fmt"
 	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
 )
 
+// https://developers.generativeai.google/api/rest/generativelanguage/models/generateMessage#request-body
+// https://developers.generativeai.google/api/rest/generativelanguage/models/generateMessage#response-body
+
 type PaLMChatMessage struct {
 	Author  string `json:"author"`
 	Content string `json:"content"`
@@ -15,45 +22,188 @@ type PaLMFilter struct {
 	Message string `json:"message"`
 }
 
-// https://developers.generativeai.google/api/rest/generativelanguage/models/generateMessage#request-body
+type PaLMPrompt struct {
+	Messages []PaLMChatMessage `json:"messages"`
+}
+
 type PaLMChatRequest struct {
-	Prompt         []Message `json:"prompt"`
-	Temperature    float64   `json:"temperature"`
-	CandidateCount int       `json:"candidateCount"`
-	TopP           float64   `json:"topP"`
-	TopK           int       `json:"topK"`
+	Prompt         PaLMPrompt `json:"prompt"`
+	Temperature    float64    `json:"temperature,omitempty"`
+	CandidateCount int        `json:"candidateCount,omitempty"`
+	TopP           float64    `json:"topP,omitempty"`
+	TopK           int        `json:"topK,omitempty"`
+}
+
+type PaLMError struct {
+	Code    int    `json:"code"`
+	Message string `json:"message"`
+	Status  string `json:"status"`
 }
 
-// https://developers.generativeai.google/api/rest/generativelanguage/models/generateMessage#response-body
 type PaLMChatResponse struct {
-	Candidates []Message    `json:"candidates"`
-	Messages   []Message    `json:"messages"`
-	Filters    []PaLMFilter `json:"filters"`
+	Candidates []PaLMChatMessage `json:"candidates"`
+	Messages   []Message         `json:"messages"`
+	Filters    []PaLMFilter      `json:"filters"`
+	Error      PaLMError         `json:"error"`
 }
 
-func relayPaLM(openAIRequest GeneralOpenAIRequest, c *gin.Context) *OpenAIErrorWithStatusCode {
-	// https://developers.generativeai.google/api/rest/generativelanguage/models/generateMessage
-	messages := make([]PaLMChatMessage, 0, len(openAIRequest.Messages))
-	for _, message := range openAIRequest.Messages {
-		var author string
-		if message.Role == "user" {
-			author = "0"
-		} else {
-			author = "1"
-		}
-		messages = append(messages, PaLMChatMessage{
-			Author:  author,
+func requestOpenAI2PaLM(textRequest GeneralOpenAIRequest) *PaLMChatRequest {
+	palmRequest := PaLMChatRequest{
+		Prompt: PaLMPrompt{
+			Messages: make([]PaLMChatMessage, 0, len(textRequest.Messages)),
+		},
+		Temperature:    textRequest.Temperature,
+		CandidateCount: textRequest.N,
+		TopP:           textRequest.TopP,
+		TopK:           textRequest.MaxTokens,
+	}
+	for _, message := range textRequest.Messages {
+		palmMessage := PaLMChatMessage{
 			Content: message.Content,
-		})
+		}
+		if message.Role == "user" {
+			palmMessage.Author = "0"
+		} else {
+			palmMessage.Author = "1"
+		}
+		palmRequest.Prompt.Messages = append(palmRequest.Prompt.Messages, palmMessage)
 	}
-	request := PaLMChatRequest{
-		Prompt:         nil,
-		Temperature:    openAIRequest.Temperature,
-		CandidateCount: openAIRequest.N,
-		TopP:           openAIRequest.TopP,
-		TopK:           openAIRequest.MaxTokens,
-	}
-	// TODO: forward request to PaLM & convert response
-	fmt.Print(request)
-	return nil
+	return &palmRequest
+}
+
+func responsePaLM2OpenAI(response *PaLMChatResponse) *OpenAITextResponse {
+	fullTextResponse := OpenAITextResponse{
+		Choices: make([]OpenAITextResponseChoice, 0, len(response.Candidates)),
+	}
+	for i, candidate := range response.Candidates {
+		choice := OpenAITextResponseChoice{
+			Index: i,
+			Message: Message{
+				Role:    "assistant",
+				Content: candidate.Content,
+			},
+			FinishReason: "stop",
+		}
+		fullTextResponse.Choices = append(fullTextResponse.Choices, choice)
+	}
+	return &fullTextResponse
+}
+
+func streamResponsePaLM2OpenAI(palmResponse *PaLMChatResponse) *ChatCompletionsStreamResponse {
+	var choice ChatCompletionsStreamResponseChoice
+	if len(palmResponse.Candidates) > 0 {
+		choice.Delta.Content = palmResponse.Candidates[0].Content
+	}
+	choice.FinishReason = "stop"
+	var response ChatCompletionsStreamResponse
+	response.Object = "chat.completion.chunk"
+	response.Model = "palm2"
+	response.Choices = []ChatCompletionsStreamResponseChoice{choice}
+	return &response
+}
+
+func palmStreamHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, string) {
+	responseText := ""
+	responseId := fmt.Sprintf("chatcmpl-%s", common.GetUUID())
+	createdTime := common.GetTimestamp()
+	dataChan := make(chan string)
+	stopChan := make(chan bool)
+	go func() {
+		responseBody, err := io.ReadAll(resp.Body)
+		if err != nil {
+			common.SysError("error reading stream response: " + err.Error())
+			stopChan <- true
+			return
+		}
+		err = resp.Body.Close()
+		if err != nil {
+			common.SysError("error closing stream response: " + err.Error())
+			stopChan <- true
+			return
+		}
+		var palmResponse PaLMChatResponse
+		err = json.Unmarshal(responseBody, &palmResponse)
+		if err != nil {
+			common.SysError("error unmarshalling stream response: " + err.Error())
+			stopChan <- true
+			return
+		}
+		fullTextResponse := streamResponsePaLM2OpenAI(&palmResponse)
+		fullTextResponse.Id = responseId
+		fullTextResponse.Created = createdTime
+		if len(palmResponse.Candidates) > 0 {
+			responseText = palmResponse.Candidates[0].Content
+		}
+		jsonResponse, err := json.Marshal(fullTextResponse)
+		if err != nil {
+			common.SysError("error marshalling stream response: " + err.Error())
+			stopChan <- true
+			return
+		}
+		dataChan <- string(jsonResponse)
+		stopChan <- true
+	}()
+	c.Writer.Header().Set("Content-Type", "text/event-stream")
+	c.Writer.Header().Set("Cache-Control", "no-cache")
+	c.Writer.Header().Set("Connection", "keep-alive")
+	c.Writer.Header().Set("Transfer-Encoding", "chunked")
+	c.Writer.Header().Set("X-Accel-Buffering", "no")
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case data := <-dataChan:
+			c.Render(-1, common.CustomEvent{Data: "data: " + data})
+			return true
+		case <-stopChan:
+			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
+			return false
+		}
+	})
+	err := resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), ""
+	}
+	return nil, responseText
+}
+
+func palmHandler(c *gin.Context, resp *http.Response, promptTokens int, model string) (*OpenAIErrorWithStatusCode, *Usage) {
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	var palmResponse PaLMChatResponse
+	err = json.Unmarshal(responseBody, &palmResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if palmResponse.Error.Code != 0 || len(palmResponse.Candidates) == 0 {
+		return &OpenAIErrorWithStatusCode{
+			OpenAIError: OpenAIError{
+				Message: palmResponse.Error.Message,
+				Type:    palmResponse.Error.Status,
+				Param:   "",
+				Code:    palmResponse.Error.Code,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+	fullTextResponse := responsePaLM2OpenAI(&palmResponse)
+	completionTokens := countTokenText(palmResponse.Candidates[0].Content, model)
+	usage := Usage{
+		PromptTokens:     promptTokens,
+		CompletionTokens: completionTokens,
+		TotalTokens:      promptTokens + completionTokens,
+	}
+	fullTextResponse.Usage = usage
+	jsonResponse, err := json.Marshal(fullTextResponse)
+	if err != nil {
+		return errorWrapper(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
 }

controller/relay-text.go

@@ -1,7 +1,6 @@
 package controller
 
 import (
-	"bufio"
 	"bytes"
 	"encoding/json"
 	"errors"
@@ -15,6 +14,14 @@ import (
 	"github.com/gin-gonic/gin"
 )
 
+const (
+	APITypeOpenAI = iota
+	APITypeClaude
+	APITypePaLM
+	APITypeBaidu
+	APITypeZhipu
+)
+
 func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 	channelType := c.GetInt("channel")
 	tokenId := c.GetInt("token_id")
@@ -71,33 +78,71 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			isModelMapped = true
 		}
 	}
+	apiType := APITypeOpenAI
+	if strings.HasPrefix(textRequest.Model, "claude") {
+		apiType = APITypeClaude
+	} else if strings.HasPrefix(textRequest.Model, "ERNIE") {
+		apiType = APITypeBaidu
+	} else if strings.HasPrefix(textRequest.Model, "PaLM") {
+		apiType = APITypePaLM
+	} else if strings.HasPrefix(textRequest.Model, "chatglm_") {
+		apiType = APITypeZhipu
+	}
 	baseURL := common.ChannelBaseURLs[channelType]
 	requestURL := c.Request.URL.String()
 	if c.GetString("base_url") != "" {
 		baseURL = c.GetString("base_url")
 	}
 	fullRequestURL := fmt.Sprintf("%s%s", baseURL, requestURL)
-	if channelType == common.ChannelTypeAzure {
-		// https://learn.microsoft.com/en-us/azure/cognitive-services/openai/chatgpt-quickstart?pivots=rest-api&tabs=command-line#rest-api
-		query := c.Request.URL.Query()
-		apiVersion := query.Get("api-version")
-		if apiVersion == "" {
-			apiVersion = c.GetString("api_version")
+	switch apiType {
+	case APITypeOpenAI:
+		if channelType == common.ChannelTypeAzure {
+			// https://learn.microsoft.com/en-us/azure/cognitive-services/openai/chatgpt-quickstart?pivots=rest-api&tabs=command-line#rest-api
+			query := c.Request.URL.Query()
+			apiVersion := query.Get("api-version")
+			if apiVersion == "" {
+				apiVersion = c.GetString("api_version")
+			}
+			requestURL := strings.Split(requestURL, "?")[0]
+			requestURL = fmt.Sprintf("%s?api-version=%s", requestURL, apiVersion)
+			baseURL = c.GetString("base_url")
+			task := strings.TrimPrefix(requestURL, "/v1/")
+			model_ := textRequest.Model
+			model_ = strings.Replace(model_, ".", "", -1)
+			// https://github.com/songquanpeng/one-api/issues/67
+			model_ = strings.TrimSuffix(model_, "-0301")
+			model_ = strings.TrimSuffix(model_, "-0314")
+			model_ = strings.TrimSuffix(model_, "-0613")
+			fullRequestURL = fmt.Sprintf("%s/openai/deployments/%s/%s", baseURL, model_, task)
 		}
-		requestURL := strings.Split(requestURL, "?")[0]
-		requestURL = fmt.Sprintf("%s?api-version=%s", requestURL, apiVersion)
-		baseURL = c.GetString("base_url")
-		task := strings.TrimPrefix(requestURL, "/v1/")
-		model_ := textRequest.Model
-		model_ = strings.Replace(model_, ".", "", -1)
-		// https://github.com/songquanpeng/one-api/issues/67
-		model_ = strings.TrimSuffix(model_, "-0301")
-		model_ = strings.TrimSuffix(model_, "-0314")
-		model_ = strings.TrimSuffix(model_, "-0613")
-		fullRequestURL = fmt.Sprintf("%s/openai/deployments/%s/%s", baseURL, model_, task)
-	} else if channelType == common.ChannelTypePaLM {
-		err := relayPaLM(textRequest, c)
-		return err
+	case APITypeClaude:
+		fullRequestURL = "https://api.anthropic.com/v1/complete"
+		if baseURL != "" {
+			fullRequestURL = fmt.Sprintf("%s/v1/complete", baseURL)
+		}
+	case APITypeBaidu:
+		switch textRequest.Model {
+		case "ERNIE-Bot":
+			fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions"
+		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"
+		}
+		apiKey := c.Request.Header.Get("Authorization")
+		apiKey = strings.TrimPrefix(apiKey, "Bearer ")
+		fullRequestURL += "?access_token=" + apiKey // TODO: access token expire in 30 days
+	case APITypePaLM:
+		fullRequestURL = "https://generativelanguage.googleapis.com/v1beta2/models/chat-bison-001:generateMessage"
+		apiKey := c.Request.Header.Get("Authorization")
+		apiKey = strings.TrimPrefix(apiKey, "Bearer ")
+		fullRequestURL += "?key=" + apiKey
+	case APITypeZhipu:
+		method := "invoke"
+		if textRequest.Stream {
+			method = "sse-invoke"
+		}
+		fullRequestURL = fmt.Sprintf("https://open.bigmodel.cn/api/paas/v3/model-api/%s/%s", textRequest.Model, method)
 	}
 	var promptTokens int
 	var completionTokens int
@@ -142,16 +187,59 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 	} else {
 		requestBody = c.Request.Body
 	}
+	switch apiType {
+	case APITypeClaude:
+		claudeRequest := requestOpenAI2Claude(textRequest)
+		jsonStr, err := json.Marshal(claudeRequest)
+		if err != nil {
+			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
+		}
+		requestBody = bytes.NewBuffer(jsonStr)
+	case APITypeBaidu:
+		baiduRequest := requestOpenAI2Baidu(textRequest)
+		jsonStr, err := json.Marshal(baiduRequest)
+		if err != nil {
+			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
+		}
+		requestBody = bytes.NewBuffer(jsonStr)
+	case APITypePaLM:
+		palmRequest := requestOpenAI2PaLM(textRequest)
+		jsonStr, err := json.Marshal(palmRequest)
+		if err != nil {
+			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
+		}
+		requestBody = bytes.NewBuffer(jsonStr)
+	case APITypeZhipu:
+		zhipuRequest := requestOpenAI2Zhipu(textRequest)
+		jsonStr, err := json.Marshal(zhipuRequest)
+		if err != nil {
+			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
+		}
+		requestBody = bytes.NewBuffer(jsonStr)
+	}
 	req, err := http.NewRequest(c.Request.Method, fullRequestURL, requestBody)
 	if err != nil {
 		return errorWrapper(err, "new_request_failed", http.StatusInternalServerError)
 	}
-	if channelType == common.ChannelTypeAzure {
-		key := c.Request.Header.Get("Authorization")
-		key = strings.TrimPrefix(key, "Bearer ")
-		req.Header.Set("api-key", key)
-	} else {
-		req.Header.Set("Authorization", c.Request.Header.Get("Authorization"))
+	apiKey := c.Request.Header.Get("Authorization")
+	apiKey = strings.TrimPrefix(apiKey, "Bearer ")
+	switch apiType {
+	case APITypeOpenAI:
+		if channelType == common.ChannelTypeAzure {
+			req.Header.Set("api-key", apiKey)
+		} else {
+			req.Header.Set("Authorization", c.Request.Header.Get("Authorization"))
+		}
+	case APITypeClaude:
+		req.Header.Set("x-api-key", apiKey)
+		anthropicVersion := c.Request.Header.Get("anthropic-version")
+		if anthropicVersion == "" {
+			anthropicVersion = "2023-06-01"
+		}
+		req.Header.Set("anthropic-version", anthropicVersion)
+	case APITypeZhipu:
+		token := getZhipuToken(apiKey)
+		req.Header.Set("Authorization", token)
 	}
 	req.Header.Set("Content-Type", c.Request.Header.Get("Content-Type"))
 	req.Header.Set("Accept", c.Request.Header.Get("Accept"))
@@ -183,11 +271,15 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			if strings.HasPrefix(textRequest.Model, "gpt-4") {
 				completionRatio = 2
 			}
-			if isStream {
+			if isStream && apiType != APITypeBaidu && apiType != APITypeZhipu {
 				completionTokens = countTokenText(streamResponseText, textRequest.Model)
 			} else {
 				promptTokens = textResponse.Usage.PromptTokens
 				completionTokens = textResponse.Usage.CompletionTokens
+				if apiType == APITypeZhipu {
+					// zhipu's API does not return prompt tokens & completion tokens
+					promptTokens = textResponse.Usage.TotalTokens
+				}
 			}
 			quota = promptTokens + int(float64(completionTokens)*completionRatio)
 			quota = int(float64(quota) * ratio)
@@ -219,123 +311,102 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			}
 		}
 	}()
-
-	if isStream {
-		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\n"); i >= 0 {
-				return i + 2, 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) < 6 { // must be something wrong!
-					common.SysError("invalid stream response: " + data)
-					continue
-				}
-				dataChan <- data
-				data = data[6:]
-				if !strings.HasPrefix(data, "[DONE]") {
-					switch relayMode {
-					case RelayModeChatCompletions:
-						var streamResponse ChatCompletionsStreamResponse
-						err = json.Unmarshal([]byte(data), &streamResponse)
-						if err != nil {
-							common.SysError("error unmarshalling stream response: " + err.Error())
-							return
-						}
-						for _, choice := range streamResponse.Choices {
-							streamResponseText += choice.Delta.Content
-						}
-					case RelayModeCompletions:
-						var streamResponse CompletionsStreamResponse
-						err = json.Unmarshal([]byte(data), &streamResponse)
-						if err != nil {
-							common.SysError("error unmarshalling stream response: " + err.Error())
-							return
-						}
-						for _, choice := range streamResponse.Choices {
-							streamResponseText += choice.Text
-						}
-					}
-				}
-			}
-			stopChan <- true
-		}()
-		c.Writer.Header().Set("Content-Type", "text/event-stream")
-		c.Writer.Header().Set("Cache-Control", "no-cache")
-		c.Writer.Header().Set("Connection", "keep-alive")
-		c.Writer.Header().Set("Transfer-Encoding", "chunked")
-		c.Writer.Header().Set("X-Accel-Buffering", "no")
-		c.Stream(func(w io.Writer) bool {
-			select {
-			case data := <-dataChan:
-				if strings.HasPrefix(data, "data: [DONE]") {
-					data = data[:12]
-				}
-				c.Render(-1, common.CustomEvent{Data: data})
-				return true
-			case <-stopChan:
-				return false
-			}
-		})
-		err = resp.Body.Close()
-		if err != nil {
-			return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError)
-		}
-		return nil
-	} else {
-		if consumeQuota {
-			responseBody, err := io.ReadAll(resp.Body)
+	switch apiType {
+	case APITypeOpenAI:
+		if isStream {
+			err, responseText := openaiStreamHandler(c, resp, relayMode)
 			if err != nil {
-				return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError)
+				return err
 			}
-			err = resp.Body.Close()
+			streamResponseText = responseText
+			return nil
+		} else {
+			err, usage := openaiHandler(c, resp, consumeQuota)
 			if err != nil {
-				return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError)
+				return err
 			}
-			err = json.Unmarshal(responseBody, &textResponse)
+			if usage != nil {
+				textResponse.Usage = *usage
+			}
+			return nil
+		}
+	case APITypeClaude:
+		if isStream {
+			err, responseText := claudeStreamHandler(c, resp)
 			if err != nil {
-				return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError)
+				return err
 			}
-			if textResponse.Error.Type != "" {
-				return &OpenAIErrorWithStatusCode{
-					OpenAIError: textResponse.Error,
-					StatusCode:  resp.StatusCode,
-				}
+			streamResponseText = responseText
+			return nil
+		} else {
+			err, usage := claudeHandler(c, resp, promptTokens, textRequest.Model)
+			if err != nil {
+				return err
 			}
-			// Reset response body
-			resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
+			if usage != nil {
+				textResponse.Usage = *usage
+			}
+			return nil
 		}
-		// We shouldn't set the header before we parse the response body, because the parse part may fail.
-		// And then we will have to send an error response, but in this case, the header has already been set.
-		// So the client will be confused by the response.
-		// For example, Postman will report error, and we cannot check the response at all.
-		for k, v := range resp.Header {
-			c.Writer.Header().Set(k, v[0])
+	case APITypeBaidu:
+		if isStream {
+			err, usage := baiduStreamHandler(c, resp)
+			if err != nil {
+				return err
+			}
+			if usage != nil {
+				textResponse.Usage = *usage
+			}
+			return nil
+		} else {
+			err, usage := baiduHandler(c, resp)
+			if err != nil {
+				return err
+			}
+			if usage != nil {
+				textResponse.Usage = *usage
+			}
+			return nil
 		}
-		c.Writer.WriteHeader(resp.StatusCode)
-		_, err = io.Copy(c.Writer, resp.Body)
-		if err != nil {
-			return errorWrapper(err, "copy_response_body_failed", http.StatusInternalServerError)
+	case APITypePaLM:
+		if textRequest.Stream { // PaLM2 API does not support stream
+			err, responseText := palmStreamHandler(c, resp)
+			if err != nil {
+				return err
+			}
+			streamResponseText = responseText
+			return nil
+		} else {
+			err, usage := palmHandler(c, resp, promptTokens, textRequest.Model)
+			if err != nil {
+				return err
+			}
+			if usage != nil {
+				textResponse.Usage = *usage
+			}
+			return nil
 		}
-		err = resp.Body.Close()
-		if err != nil {
-			return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError)
+	case APITypeZhipu:
+		if isStream {
+			err, usage := zhipuStreamHandler(c, resp)
+			if err != nil {
+				return err
+			}
+			if usage != nil {
+				textResponse.Usage = *usage
+			}
+			return nil
+		} else {
+			err, usage := zhipuHandler(c, resp)
+			if err != nil {
+				return err
+			}
+			if usage != nil {
+				textResponse.Usage = *usage
+			}
+			return nil
 		}
-		return nil
+	default:
+		return errorWrapper(errors.New("unknown api type"), "unknown_api_type", http.StatusInternalServerError)
 	}
 }

controller/relay-utils.go

@@ -91,3 +91,16 @@ func errorWrapper(err error, code string, statusCode int) *OpenAIErrorWithStatus
 		StatusCode:  statusCode,
 	}
 }
+
+func shouldDisableChannel(err *OpenAIError) bool {
+	if !common.AutomaticDisableChannelEnabled {
+		return false
+	}
+	if err == nil {
+		return false
+	}
+	if err.Type == "insufficient_quota" || err.Code == "invalid_api_key" || err.Code == "account_deactivated" {
+		return true
+	}
+	return false
+}

controller/relay-zhipu.go

@@ -0,0 +1,290 @@
+package controller
+
+import (
+	"bufio"
+	"encoding/json"
+	"github.com/gin-gonic/gin"
+	"github.com/golang-jwt/jwt"
+	"io"
+	"net/http"
+	"one-api/common"
+	"strings"
+	"sync"
+	"time"
+)
+
+// https://open.bigmodel.cn/doc/api#chatglm_std
+// chatglm_std, chatglm_lite
+// https://open.bigmodel.cn/api/paas/v3/model-api/chatglm_std/invoke
+// https://open.bigmodel.cn/api/paas/v3/model-api/chatglm_std/sse-invoke
+
+type ZhipuMessage struct {
+	Role    string `json:"role"`
+	Content string `json:"content"`
+}
+
+type ZhipuRequest struct {
+	Prompt      []ZhipuMessage `json:"prompt"`
+	Temperature float64        `json:"temperature,omitempty"`
+	TopP        float64        `json:"top_p,omitempty"`
+	RequestId   string         `json:"request_id,omitempty"`
+	Incremental bool           `json:"incremental,omitempty"`
+}
+
+type ZhipuResponseData struct {
+	TaskId     string         `json:"task_id"`
+	RequestId  string         `json:"request_id"`
+	TaskStatus string         `json:"task_status"`
+	Choices    []ZhipuMessage `json:"choices"`
+	Usage      `json:"usage"`
+}
+
+type ZhipuResponse struct {
+	Code    int               `json:"code"`
+	Msg     string            `json:"msg"`
+	Success bool              `json:"success"`
+	Data    ZhipuResponseData `json:"data"`
+}
+
+type ZhipuStreamMetaResponse struct {
+	RequestId  string `json:"request_id"`
+	TaskId     string `json:"task_id"`
+	TaskStatus string `json:"task_status"`
+	Usage      `json:"usage"`
+}
+
+type zhipuTokenData struct {
+	Token      string
+	ExpiryTime time.Time
+}
+
+var zhipuTokens sync.Map
+var expSeconds int64 = 24 * 3600
+
+func getZhipuToken(apikey string) string {
+	data, ok := zhipuTokens.Load(apikey)
+	if ok {
+		tokenData := data.(zhipuTokenData)
+		if time.Now().Before(tokenData.ExpiryTime) {
+			return tokenData.Token
+		}
+	}
+
+	split := strings.Split(apikey, ".")
+	if len(split) != 2 {
+		common.SysError("invalid zhipu key: " + apikey)
+		return ""
+	}
+
+	id := split[0]
+	secret := split[1]
+
+	expMillis := time.Now().Add(time.Duration(expSeconds)*time.Second).UnixNano() / 1e6
+	expiryTime := time.Now().Add(time.Duration(expSeconds) * time.Second)
+
+	timestamp := time.Now().UnixNano() / 1e6
+
+	payload := jwt.MapClaims{
+		"api_key":   id,
+		"exp":       expMillis,
+		"timestamp": timestamp,
+	}
+
+	token := jwt.NewWithClaims(jwt.SigningMethodHS256, payload)
+
+	token.Header["alg"] = "HS256"
+	token.Header["sign_type"] = "SIGN"
+
+	tokenString, err := token.SignedString([]byte(secret))
+	if err != nil {
+		return ""
+	}
+
+	zhipuTokens.Store(apikey, zhipuTokenData{
+		Token:      tokenString,
+		ExpiryTime: expiryTime,
+	})
+
+	return tokenString
+}
+
+func requestOpenAI2Zhipu(request GeneralOpenAIRequest) *ZhipuRequest {
+	messages := make([]ZhipuMessage, 0, len(request.Messages))
+	for _, message := range request.Messages {
+		messages = append(messages, ZhipuMessage{
+			Role:    message.Role,
+			Content: message.Content,
+		})
+	}
+	return &ZhipuRequest{
+		Prompt:      messages,
+		Temperature: request.Temperature,
+		TopP:        request.TopP,
+		Incremental: false,
+	}
+}
+
+func responseZhipu2OpenAI(response *ZhipuResponse) *OpenAITextResponse {
+	fullTextResponse := OpenAITextResponse{
+		Id:      response.Data.TaskId,
+		Object:  "chat.completion",
+		Created: common.GetTimestamp(),
+		Choices: make([]OpenAITextResponseChoice, 0, len(response.Data.Choices)),
+		Usage:   response.Data.Usage,
+	}
+	for i, choice := range response.Data.Choices {
+		openaiChoice := OpenAITextResponseChoice{
+			Index: i,
+			Message: Message{
+				Role:    choice.Role,
+				Content: strings.Trim(choice.Content, "\""),
+			},
+			FinishReason: "",
+		}
+		if i == len(response.Data.Choices)-1 {
+			openaiChoice.FinishReason = "stop"
+		}
+		fullTextResponse.Choices = append(fullTextResponse.Choices, openaiChoice)
+	}
+	return &fullTextResponse
+}
+
+func streamResponseZhipu2OpenAI(zhipuResponse string) *ChatCompletionsStreamResponse {
+	var choice ChatCompletionsStreamResponseChoice
+	choice.Delta.Content = zhipuResponse
+	choice.FinishReason = ""
+	response := ChatCompletionsStreamResponse{
+		Object:  "chat.completion.chunk",
+		Created: common.GetTimestamp(),
+		Model:   "chatglm",
+		Choices: []ChatCompletionsStreamResponseChoice{choice},
+	}
+	return &response
+}
+
+func streamMetaResponseZhipu2OpenAI(zhipuResponse *ZhipuStreamMetaResponse) (*ChatCompletionsStreamResponse, *Usage) {
+	var choice ChatCompletionsStreamResponseChoice
+	choice.Delta.Content = ""
+	choice.FinishReason = "stop"
+	response := ChatCompletionsStreamResponse{
+		Id:      zhipuResponse.RequestId,
+		Object:  "chat.completion.chunk",
+		Created: common.GetTimestamp(),
+		Model:   "chatglm",
+		Choices: []ChatCompletionsStreamResponseChoice{choice},
+	}
+	return &response, &zhipuResponse.Usage
+}
+
+func zhipuStreamHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, *Usage) {
+	var usage *Usage
+	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)
+	metaChan := make(chan string)
+	stopChan := make(chan bool)
+	go func() {
+		for scanner.Scan() {
+			data := scanner.Text()
+			data = strings.Trim(data, "\"")
+			if len(data) < 5 { // ignore blank line or wrong format
+				continue
+			}
+			if data[:5] == "data:" {
+				dataChan <- data[5:]
+			} else if data[:5] == "meta:" {
+				metaChan <- data[5:]
+			}
+		}
+		stopChan <- true
+	}()
+	c.Writer.Header().Set("Content-Type", "text/event-stream")
+	c.Writer.Header().Set("Cache-Control", "no-cache")
+	c.Writer.Header().Set("Connection", "keep-alive")
+	c.Writer.Header().Set("Transfer-Encoding", "chunked")
+	c.Writer.Header().Set("X-Accel-Buffering", "no")
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case data := <-dataChan:
+			response := streamResponseZhipu2OpenAI(data)
+			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 data := <-metaChan:
+			var zhipuResponse ZhipuStreamMetaResponse
+			err := json.Unmarshal([]byte(data), &zhipuResponse)
+			if err != nil {
+				common.SysError("error unmarshalling stream response: " + err.Error())
+				return true
+			}
+			response, zhipuUsage := streamMetaResponseZhipu2OpenAI(&zhipuResponse)
+			jsonResponse, err := json.Marshal(response)
+			if err != nil {
+				common.SysError("error marshalling stream response: " + err.Error())
+				return true
+			}
+			usage = zhipuUsage
+			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()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	return nil, usage
+}
+
+func zhipuHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, *Usage) {
+	var zhipuResponse ZhipuResponse
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = json.Unmarshal(responseBody, &zhipuResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if !zhipuResponse.Success {
+		return &OpenAIErrorWithStatusCode{
+			OpenAIError: OpenAIError{
+				Message: zhipuResponse.Msg,
+				Type:    "zhipu_error",
+				Param:   "",
+				Code:    zhipuResponse.Code,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+	fullTextResponse := responseZhipu2OpenAI(&zhipuResponse)
+	jsonResponse, err := json.Marshal(fullTextResponse)
+	if err != nil {
+		return errorWrapper(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, &fullTextResponse.Usage
+}

controller/relay.go

@@ -4,6 +4,7 @@ import (
 	"fmt"
 	"net/http"
 	"one-api/common"
+	"strconv"
 	"strings"
 
 	"github.com/gin-gonic/gin"
@@ -84,6 +85,20 @@ type TextResponse struct {
 	Error OpenAIError `json:"error"`
 }
 
+type OpenAITextResponseChoice struct {
+	Index        int `json:"index"`
+	Message      `json:"message"`
+	FinishReason string `json:"finish_reason"`
+}
+
+type OpenAITextResponse struct {
+	Id      string                     `json:"id"`
+	Object  string                     `json:"object"`
+	Created int64                      `json:"created"`
+	Choices []OpenAITextResponseChoice `json:"choices"`
+	Usage   `json:"usage"`
+}
+
 type ImageResponse struct {
 	Created int `json:"created"`
 	Data    []struct {
@@ -91,13 +106,19 @@ type ImageResponse struct {
 	}
 }
 
+type ChatCompletionsStreamResponseChoice struct {
+	Delta struct {
+		Content string `json:"content"`
+	} `json:"delta"`
+	FinishReason string `json:"finish_reason,omitempty"`
+}
+
 type ChatCompletionsStreamResponse struct {
-	Choices []struct {
-		Delta struct {
-			Content string `json:"content"`
-		} `json:"delta"`
-		FinishReason string `json:"finish_reason"`
-	} `json:"choices"`
+	Id      string                                `json:"id"`
+	Object  string                                `json:"object"`
+	Created int64                                 `json:"created"`
+	Model   string                                `json:"model"`
+	Choices []ChatCompletionsStreamResponseChoice `json:"choices"`
 }
 
 type CompletionsStreamResponse struct {
@@ -132,16 +153,25 @@ func Relay(c *gin.Context) {
 		err = relayTextHelper(c, relayMode)
 	}
 	if err != nil {
-		if err.StatusCode == http.StatusTooManyRequests {
-			err.OpenAIError.Message = "当前分组负载已饱和，请稍后再试，或升级账户以提升服务质量。"
+		retryTimesStr := c.Query("retry")
+		retryTimes, _ := strconv.Atoi(retryTimesStr)
+		if retryTimesStr == "" {
+			retryTimes = common.RetryTimes
+		}
+		if retryTimes > 0 {
+			c.Redirect(http.StatusTemporaryRedirect, fmt.Sprintf("%s?retry=%d", c.Request.URL.Path, retryTimes-1))
+		} else {
+			if err.StatusCode == http.StatusTooManyRequests {
+				err.OpenAIError.Message = "当前分组负载已饱和，请稍后再试，或升级账户以提升服务质量。"
+			}
+			c.JSON(err.StatusCode, gin.H{
+				"error": err.OpenAIError,
+			})
 		}
-		c.JSON(err.StatusCode, gin.H{
-			"error": err.OpenAIError,
-		})
 		channelId := c.GetInt("channel_id")
 		common.SysError(fmt.Sprintf("relay error (channel #%d): %s", channelId, err.Message))
 		// https://platform.openai.com/docs/guides/error-codes/api-errors
-		if common.AutomaticDisableChannelEnabled && (err.Type == "insufficient_quota" || err.Code == "invalid_api_key" || err.Code == "account_deactivated") {
+		if shouldDisableChannel(&err.OpenAIError) {
 			channelId := c.GetInt("channel_id")
 			channelName := c.GetString("channel_name")
 			disableChannel(channelId, channelName, err.Message)

go.mod

@@ -11,6 +11,7 @@ require (
 	github.com/gin-gonic/gin v1.9.1
 	github.com/go-playground/validator/v10 v10.14.0
 	github.com/go-redis/redis/v8 v8.11.5
+	github.com/golang-jwt/jwt v3.2.2+incompatible
 	github.com/google/uuid v1.3.0
 	github.com/pkoukk/tiktoken-go v0.1.1
 	golang.org/x/crypto v0.9.0
@@ -20,7 +21,6 @@ require (
 )
 
 require (
-	github.com/boj/redistore v0.0.0-20180917114910-cd5dcc76aeff // indirect
 	github.com/bytedance/sonic v1.9.1 // indirect
 	github.com/cespare/xxhash/v2 v2.1.2 // indirect
 	github.com/chenzhuoyu/base64x v0.0.0-20221115062448-fe3a3abad311 // indirect
@@ -32,7 +32,6 @@ require (
 	github.com/go-playground/universal-translator v0.18.1 // indirect
 	github.com/go-sql-driver/mysql v1.6.0 // indirect
 	github.com/goccy/go-json v0.10.2 // indirect
-	github.com/gomodule/redigo v2.0.0+incompatible // indirect
 	github.com/gorilla/context v1.1.1 // indirect
 	github.com/gorilla/securecookie v1.1.1 // indirect
 	github.com/gorilla/sessions v1.2.1 // indirect

go.sum

@@ -1,5 +1,3 @@
-github.com/boj/redistore v0.0.0-20180917114910-cd5dcc76aeff h1:RmdPFa+slIr4SCBg4st/l/vZWVe9QJKMXGO60Bxbe04=
-github.com/boj/redistore v0.0.0-20180917114910-cd5dcc76aeff/go.mod h1:+RTT1BOk5P97fT2CiHkbFQwkK3mjsFAP6zCYV2aXtjw=
 github.com/bytedance/sonic v1.5.0/go.mod h1:ED5hyg4y6t3/9Ku1R6dU/4KyJ48DZ4jPhfY1O2AihPM=
 github.com/bytedance/sonic v1.9.1 h1:6iJ6NqdoxCDr6mbY8h18oSO+cShGSMRGCEo7F2h0x8s=
 github.com/bytedance/sonic v1.9.1/go.mod h1:i736AoUSYt75HyZLoJW9ERYxcy6eaN6h4BZXU064P/U=
@@ -54,10 +52,10 @@ github.com/go-sql-driver/mysql v1.6.0/go.mod h1:DCzpHaOWr8IXmIStZouvnhqoel9Qv2LB
 github.com/goccy/go-json v0.9.7/go.mod h1:6MelG93GURQebXPDq3khkgXZkazVtN9CRI+MGFi0w8I=
 github.com/goccy/go-json v0.10.2 h1:CrxCmQqYDkv1z7lO7Wbh2HN93uovUHgrECaO5ZrCXAU=
 github.com/goccy/go-json v0.10.2/go.mod h1:6MelG93GURQebXPDq3khkgXZkazVtN9CRI+MGFi0w8I=
+github.com/golang-jwt/jwt v3.2.2+incompatible h1:IfV12K8xAKAnZqdXVzCZ+TOjboZ2keLg81eXfW3O+oY=
+github.com/golang-jwt/jwt v3.2.2+incompatible/go.mod h1:8pz2t5EyA70fFQQSrl6XZXzqecmYZeUEB8OUGHkxJ+I=
 github.com/golang/protobuf v1.3.3/go.mod h1:vzj43D7+SQXF/4pzW/hwtAqwc6iTitCiVSaWz5lYuqw=
 github.com/golang/protobuf v1.5.0/go.mod h1:FsONVRAS9T7sI+LIUmWTfcYkHO4aIWwzhcaSAoJOfIk=
-github.com/gomodule/redigo v2.0.0+incompatible h1:K/R+8tc58AaqLkqG2Ol3Qk+DR/TlNuhuh457pBFPtt0=
-github.com/gomodule/redigo v2.0.0+incompatible/go.mod h1:B4C85qUVwatsJoIUNIfCRsp7qO0iAmpGFZ4EELWSbC4=
 github.com/google/go-cmp v0.5.5 h1:Khx7svrCpmxxtHBq5j2mp/xVjsi8hQMfNLvJFAlrGgU=
 github.com/google/go-cmp v0.5.5/go.mod h1:v8dTdLbMG2kIc/vJvl+f65V22dbkXbowE6jgT/gNBxE=
 github.com/google/gofuzz v1.0.0/go.mod h1:dBl0BpW6vV/+mYPU4Po3pmUjxk6FQPldtuIdl/M65Eg=
@@ -67,7 +65,6 @@ github.com/gorilla/context v1.1.1 h1:AWwleXJkX/nhcU9bZSnZoi3h/qGYqQAGhq6zZe/aQW8
 github.com/gorilla/context v1.1.1/go.mod h1:kBGZzfjB9CEq2AlWe17Uuf7NDRt0dE0s8S51q0aT7Yg=
 github.com/gorilla/securecookie v1.1.1 h1:miw7JPhV+b/lAHSXz4qd/nN9jRiAFV5FwjeKyCS8BvQ=
 github.com/gorilla/securecookie v1.1.1/go.mod h1:ra0sb63/xPlUeL+yeDciTfxMRAA+MP+HVt/4epWDjd4=
-github.com/gorilla/sessions v1.1.1/go.mod h1:8KCfur6+4Mqcc6S0FEfKuN15Vl5MgXW92AE8ovaJD0w=
 github.com/gorilla/sessions v1.2.1 h1:DHd3rPN5lE3Ts3D8rKkQ8x/0kqfeNmBAaiSi+o7FsgI=
 github.com/gorilla/sessions v1.2.1/go.mod h1:dk2InVEVJ0sfLlnXv9EAgkf6ecYs/i80K/zI+bUmuGM=
 github.com/jinzhu/inflection v1.0.0 h1:K317FqzuhWc8YvSVlFMCCUb36O/S9MCKRDI7QkRKD/E=

middleware/distributor.go

@@ -86,7 +86,7 @@ func Distribute() func(c *gin.Context) {
 			}
 			channel, err = model.CacheGetRandomSatisfiedChannel(userGroup, modelRequest.Model)
 			if err != nil {
-				message := "无可用渠道"
+				message := fmt.Sprintf("当前分组 %s 下对于模型 %s 无可用渠道", userGroup, modelRequest.Model)
 				if channel != nil {
 					common.SysError(fmt.Sprintf("渠道不存在：%d", channel.Id))
 					message = "数据库一致性已被破坏，请联系管理员"

model/option.go

@@ -68,6 +68,7 @@ func InitOptionMap() {
 	common.OptionMap["TopUpLink"] = common.TopUpLink
 	common.OptionMap["ChatLink"] = common.ChatLink
 	common.OptionMap["QuotaPerUnit"] = strconv.FormatFloat(common.QuotaPerUnit, 'f', -1, 64)
+	common.OptionMap["RetryTimes"] = strconv.Itoa(common.RetryTimes)
 	common.OptionMapRWMutex.Unlock()
 	loadOptionsFromDatabase()
 }
@@ -196,6 +197,8 @@ func updateOptionMap(key string, value string) (err error) {
 		common.QuotaRemindThreshold, _ = strconv.Atoi(value)
 	case "PreConsumedQuota":
 		common.PreConsumedQuota, _ = strconv.Atoi(value)
+	case "RetryTimes":
+		common.RetryTimes, _ = strconv.Atoi(value)
 	case "ModelRatio":
 		err = common.UpdateModelRatioByJSONString(value)
 	case "GroupRatio":

router/api-router.go

@@ -28,7 +28,7 @@ func SetApiRouter(router *gin.Engine) {
 		userRoute := apiRouter.Group("/user")
 		{
 			userRoute.POST("/register", middleware.CriticalRateLimit(), middleware.TurnstileCheck(), controller.Register)
-			userRoute.POST("/login", middleware.CriticalRateLimit(), middleware.TurnstileCheck(), controller.Login)
+			userRoute.POST("/login", middleware.CriticalRateLimit(), controller.Login)
 			userRoute.GET("/logout", controller.Logout)
 
 			selfRoute := userRoute.Group("/")

web/src/components/LoginForm.js

@@ -13,7 +13,6 @@ import {
 import { Link, useNavigate, useSearchParams } from 'react-router-dom';
 import { UserContext } from '../context/User';
 import { API, getLogo, showError, showSuccess, showInfo } from '../helpers';
-import Turnstile from 'react-turnstile';
 
 const LoginForm = () => {
   const [inputs, setInputs] = useState({
@@ -25,9 +24,6 @@ const LoginForm = () => {
   const [submitted, setSubmitted] = useState(false);
   const { username, password } = inputs;
   const [userState, userDispatch] = useContext(UserContext);
-  const [turnstileEnabled, setTurnstileEnabled] = useState(false);
-  const [turnstileSiteKey, setTurnstileSiteKey] = useState('');
-  const [turnstileToken, setTurnstileToken] = useState('');
   let navigate = useNavigate();
 
   const [status, setStatus] = useState({});
@@ -41,11 +37,6 @@ const LoginForm = () => {
     if (status) {
       status = JSON.parse(status);
       setStatus(status);
-
-      if (status.turnstile_check) {
-        setTurnstileEnabled(true);
-        setTurnstileSiteKey(status.turnstile_site_key);
-      }
     }
   }, []);
 
@@ -85,12 +76,7 @@ const LoginForm = () => {
   async function handleSubmit(e) {
     setSubmitted(true);
     if (username && password) {
-      if (turnstileEnabled && turnstileToken === '') {
-        showInfo('请稍后几秒重试，Turnstile 正在检查用户环境！');
-        return;
-      }
-
-      const res = await API.post(`/api/user/login?turnstile=${turnstileToken}`, {
+      const res = await API.post(`/api/user/login`, {
         username,
         password,
       });
@@ -133,16 +119,6 @@ const LoginForm = () => {
               value={password}
               onChange={handleChange}
             />
-            {turnstileEnabled ? (
-              <Turnstile
-                sitekey={turnstileSiteKey}
-                onVerify={(token) => {
-                  setTurnstileToken(token);
-                }}
-              />
-            ) : (
-              <></>
-            )}
             <Button color="" fluid size="large" onClick={handleSubmit}>
               登录
             </Button>

web/src/components/OperationSetting.js

@@ -20,6 +20,7 @@ const OperationSetting = () => {
     DisplayInCurrencyEnabled: '',
     DisplayTokenStatEnabled: '',
     ApproximateTokenEnabled: '',
+    RetryTimes: 0,
   });
   const [originInputs, setOriginInputs] = useState({});
   let [loading, setLoading] = useState(false);
@@ -122,6 +123,9 @@ const OperationSetting = () => {
         if (originInputs['QuotaPerUnit'] !== inputs.QuotaPerUnit) {
           await updateOption('QuotaPerUnit', inputs.QuotaPerUnit);
         }
+        if (originInputs['RetryTimes'] !== inputs.RetryTimes) {
+          await updateOption('RetryTimes', inputs.RetryTimes);
+        }
         break;
     }
   };
@@ -133,7 +137,7 @@ const OperationSetting = () => {
           <Header as='h3'>
             通用设置
           </Header>
-          <Form.Group widths={3}>
+          <Form.Group widths={4}>
             <Form.Input
               label='充值链接'
               name='TopUpLink'
@@ -162,6 +166,17 @@ const OperationSetting = () => {
               step='0.01'
               placeholder='一单位货币能兑换的额度'
             />
+            <Form.Input
+              label='失败重试次数'
+              name='RetryTimes'
+              type={'number'}
+              step='1'
+              min='0'
+              onChange={handleInputChange}
+              autoComplete='new-password'
+              value={inputs.RetryTimes}
+              placeholder='失败重试次数'
+            />
           </Form.Group>
           <Form.Group inline>
             <Form.Checkbox

web/src/constants/channel.constants.js

@@ -1,14 +1,18 @@
 export const CHANNEL_OPTIONS = [
   { key: 1, text: 'OpenAI', value: 1, color: 'green' },
-  { key: 8, text: '自定义', value: 8, color: 'pink' },
-  { key: 3, text: 'Azure', value: 3, color: 'olive' },
-  { key: 2, text: 'API2D', value: 2, color: 'blue' },
-  { key: 4, text: 'CloseAI', value: 4, color: 'teal' },
-  { key: 5, text: 'OpenAI-SB', value: 5, color: 'brown' },
-  { key: 6, text: 'OpenAI Max', value: 6, color: 'violet' },
-  { key: 7, text: 'OhMyGPT', value: 7, color: 'purple' },
-  { key: 9, text: 'AI.LS', value: 9, color: 'yellow' },
-  { key: 10, text: 'AI Proxy', value: 10, color: 'purple' },
-  { key: 12, text: 'API2GPT', value: 12, color: 'blue' },
-  { key: 13, text: 'AIGC2D', value: 13, color: 'purple' }
+  { key: 14, text: 'Anthropic Claude', value: 14, color: 'black' },
+  { key: 3, text: 'Azure OpenAI', value: 3, color: 'olive' },
+  { key: 11, text: 'Google PaLM2', value: 11, color: 'orange' },
+  { key: 15, text: '百度文心千帆', value: 15, color: 'blue' },
+  { key: 16, text: '智谱 ChatGLM', value: 16, color: 'violet' },
+  { key: 8, text: '自定义渠道', value: 8, color: 'pink' },
+  { key: 2, text: '代理：API2D', value: 2, color: 'blue' },
+  { key: 5, text: '代理：OpenAI-SB', value: 5, color: 'brown' },
+  { key: 7, text: '代理：OhMyGPT', value: 7, color: 'purple' },
+  { key: 10, text: '代理：AI Proxy', value: 10, color: 'purple' },
+  { key: 4, text: '代理：CloseAI', value: 4, color: 'teal' },
+  { key: 6, text: '代理：OpenAI Max', value: 6, color: 'violet' },
+  { key: 9, text: '代理：AI.LS', value: 9, color: 'yellow' },
+  { key: 12, text: '代理：API2GPT', value: 12, color: 'blue' },
+  { key: 13, text: '代理：AIGC2D', value: 13, color: 'purple' }
 ];

web/src/pages/Channel/EditChannel.js

@@ -27,6 +27,7 @@ const EditChannel = () => {
   };
   const [batch, setBatch] = useState(false);
   const [inputs, setInputs] = useState(originInputs);
+  const [originModelOptions, setOriginModelOptions] = useState([]);
   const [modelOptions, setModelOptions] = useState([]);
   const [groupOptions, setGroupOptions] = useState([]);
   const [basicModels, setBasicModels] = useState([]);
@@ -44,19 +45,6 @@ const EditChannel = () => {
         data.models = [];
       } else {
         data.models = data.models.split(',');
-        setTimeout(() => {
-          let localModelOptions = [...modelOptions];
-          data.models.forEach((model) => {
-            if (!localModelOptions.find((option) => option.key === model)) {
-              localModelOptions.push({
-                key: model,
-                text: model,
-                value: model
-              });
-            }
-          });
-          setModelOptions(localModelOptions);
-        }, 1000);
       }
       if (data.group === '') {
         data.groups = [];
@@ -76,13 +64,16 @@ const EditChannel = () => {
   const fetchModels = async () => {
     try {
       let res = await API.get(`/api/channel/models`);
-      setModelOptions(res.data.data.map((model) => ({
+      let localModelOptions = res.data.data.map((model) => ({
         key: model.id,
         text: model.id,
         value: model.id
-      })));
+      }));
+      setOriginModelOptions(localModelOptions);
       setFullModels(res.data.data.map((model) => model.id));
-      setBasicModels(res.data.data.filter((model) => !model.id.startsWith('gpt-4')).map((model) => model.id));
+      setBasicModels(res.data.data.filter((model) => {
+        return model.id.startsWith('gpt-3') || model.id.startsWith('text-');
+      }).map((model) => model.id));
     } catch (error) {
       showError(error.message);
     }
@@ -101,6 +92,20 @@ const EditChannel = () => {
     }
   };
 
+  useEffect(() => {
+    let localModelOptions = [...originModelOptions];
+    inputs.models.forEach((model) => {
+      if (!localModelOptions.find((option) => option.key === model)) {
+        localModelOptions.push({
+          key: model,
+          text: model,
+          value: model
+        });
+      }
+    });
+    setModelOptions(localModelOptions);
+  }, [originModelOptions, inputs.models]);
+
   useEffect(() => {
     if (isEdit) {
       loadChannel().then();
@@ -280,15 +285,20 @@ const EditChannel = () => {
             <Input
               action={
                 <Button type={'button'} onClick={()=>{
+                  if (customModel.trim() === "") return;
+                  if (inputs.models.includes(customModel)) return;
                   let localModels = [...inputs.models];
                   localModels.push(customModel);
-                  let localModelOptions = [...modelOptions];
+                  let localModelOptions = [];
                   localModelOptions.push({
                     key: customModel,
                     text: customModel,
                     value: customModel,
                   });
-                  setModelOptions(localModelOptions);
+                  setModelOptions(modelOptions=>{
+                    return [...modelOptions, ...localModelOptions];
+                  });
+                  setCustomModel('');
                   handleInputChange(null, { name: 'models', value: localModels });
                 }}>填入</Button>
               }
@@ -327,7 +337,7 @@ const EditChannel = () => {
                 label='密钥'
                 name='key'
                 required
-                placeholder={'请输入密钥'}
+                placeholder={inputs.type === 15 ? "请输入 access token，当前版本暂不支持自动刷新，请每 30 天更新一次" : '请输入密钥'}
                 onChange={handleInputChange}
                 value={inputs.key}
                 autoComplete='new-password'