feat: support chatglm-4 (close #1045, close #952, close #952, close #943)

.github/workflows/linux-release.yml

@@ -23,7 +23,7 @@ jobs:
       - uses: actions/setup-node@v3
         with:
           node-version: 16
-      - name: Build Frontend (theme default)
+      - name: Build Frontend
         env:
           CI: ""
         run: |
@@ -38,7 +38,7 @@ jobs:
       - name: Build Backend (amd64)
         run: |
           go mod download
-          go build -ldflags "-s -w -X 'one-api/common.Version=$(git describe --tags)' -extldflags '-static'" -o one-api
+          go build -ldflags "-s -w -X 'github.com/songquanpeng/one-api/common.Version=$(git describe --tags)' -extldflags '-static'" -o one-api
 
       - name: Build Backend (arm64)
         run: |

.github/workflows/macos-release.yml

@@ -23,7 +23,7 @@ jobs:
       - uses: actions/setup-node@v3
         with:
           node-version: 16
-      - name: Build Frontend (theme default)
+      - name: Build Frontend
         env:
           CI: ""
         run: |
@@ -38,7 +38,7 @@ jobs:
       - name: Build Backend
         run: |
           go mod download
-          go build -ldflags "-X 'one-api/common.Version=$(git describe --tags)'" -o one-api-macos
+          go build -ldflags "-X 'github.com/songquanpeng/one-api/common.Version=$(git describe --tags)'" -o one-api-macos
       - name: Release
         uses: softprops/action-gh-release@v1
         if: startsWith(github.ref, 'refs/tags/')

.github/workflows/windows-release.yml

@@ -26,7 +26,7 @@ jobs:
       - uses: actions/setup-node@v3
         with:
           node-version: 16
-      - name: Build Frontend (theme default)
+      - name: Build Frontend
         env:
           CI: ""
         run: |
@@ -41,7 +41,7 @@ jobs:
       - name: Build Backend
         run: |
           go mod download
-          go build -ldflags "-s -w -X 'one-api/common.Version=$(git describe --tags)'" -o one-api.exe
+          go build -ldflags "-s -w -X 'github.com/songquanpeng/one-api/common.Version=$(git describe --tags)'" -o one-api.exe
       - name: Release
         uses: softprops/action-gh-release@v1
         if: startsWith(github.ref, 'refs/tags/')

README.md

@@ -74,8 +74,9 @@ _✨ 通过标准的 OpenAI API 格式访问所有的大模型，开箱即用
    + [x] [360 智脑](https://ai.360.cn)
    + [x] [腾讯混元大模型](https://cloud.tencent.com/document/product/1729)
    + [x] [Moonshot AI](https://platform.moonshot.cn/)
+   + [x] [百川大模型](https://platform.baichuan-ai.com)
    + [ ] [字节云雀大模型](https://www.volcengine.com/product/ark) (WIP)
-   + [ ] [MINIMAX](https://api.minimax.chat/) (WIP)
+   + [x] [MINIMAX](https://api.minimax.chat/)
 2. 支持配置镜像以及众多[第三方代理服务](https://iamazing.cn/page/openai-api-third-party-services)。
 3. 支持通过**负载均衡**的方式访问多个渠道。
 4. 支持 **stream 模式**，可以通过流式传输实现打字机效果。

common/constants.go

@@ -64,6 +64,8 @@ const (
 	ChannelTypeTencent        = 23
 	ChannelTypeGemini         = 24
 	ChannelTypeMoonshot       = 25
+	ChannelTypeBaichuan       = 26
+	ChannelTypeMinimax        = 27
 )
 
 var ChannelBaseURLs = []string{
@@ -93,6 +95,8 @@ var ChannelBaseURLs = []string{
 	"https://hunyuan.cloud.tencent.com",         // 23
 	"https://generativelanguage.googleapis.com", // 24
 	"https://api.moonshot.cn",                   // 25
+	"https://api.baichuan-ai.com",               // 26
+	"https://api.minimax.chat",                  // 27
 }
 
 const (

common/gin.go

@@ -8,12 +8,24 @@ import (
 	"strings"
 )
 
-func UnmarshalBodyReusable(c *gin.Context, v any) error {
+const KeyRequestBody = "key_request_body"
+
+func GetRequestBody(c *gin.Context) ([]byte, error) {
+	requestBody, _ := c.Get(KeyRequestBody)
+	if requestBody != nil {
+		return requestBody.([]byte), nil
+	}
 	requestBody, err := io.ReadAll(c.Request.Body)
 	if err != nil {
-		return err
+		return nil, err
 	}
-	err = c.Request.Body.Close()
+	_ = c.Request.Body.Close()
+	c.Set(KeyRequestBody, requestBody)
+	return requestBody.([]byte), nil
+}
+
+func UnmarshalBodyReusable(c *gin.Context, v any) error {
+	requestBody, err := GetRequestBody(c)
 	if err != nil {
 		return err
 	}

common/helper/helper.go

@@ -137,6 +137,7 @@ func GetUUID() string {
 }
 
 const keyChars = "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ"
+const keyNumbers = "0123456789"
 
 func init() {
 	rand.Seed(time.Now().UnixNano())
@@ -168,6 +169,15 @@ func GetRandomString(length int) string {
 	return string(key)
 }
 
+func GetRandomNumberString(length int) string {
+	rand.Seed(time.Now().UnixNano())
+	key := make([]byte, length)
+	for i := 0; i < length; i++ {
+		key[i] = keyNumbers[rand.Intn(len(keyNumbers))]
+	}
+	return string(key)
+}
+
 func GetTimestamp() int64 {
 	return time.Now().Unix()
 }

common/logger/logger.go

@@ -13,6 +13,7 @@ import (
 )
 
 const (
+	loggerDEBUG = "DEBUG"
 	loggerINFO  = "INFO"
 	loggerWarn  = "WARN"
 	loggerError = "ERR"
@@ -55,6 +56,10 @@ func SysError(s string) {
 	_, _ = fmt.Fprintf(gin.DefaultErrorWriter, "[SYS] %v | %s \n", t.Format("2006/01/02 - 15:04:05"), s)
 }
 
+func Debug(ctx context.Context, msg string) {
+	logHelper(ctx, loggerDEBUG, msg)
+}
+
 func Info(ctx context.Context, msg string) {
 	logHelper(ctx, loggerINFO, msg)
 }
@@ -67,6 +72,10 @@ func Error(ctx context.Context, msg string) {
 	logHelper(ctx, loggerError, msg)
 }
 
+func Debugf(ctx context.Context, format string, a ...any) {
+	Debug(ctx, fmt.Sprintf(format, a...))
+}
+
 func Infof(ctx context.Context, format string, a ...any) {
 	Info(ctx, fmt.Sprintf(format, a...))
 }

common/model-ratio.go

@@ -102,6 +102,10 @@ var ModelRatio = map[string]float64{
 	"PaLM-2":            1,
 	"gemini-pro":        1, // $0.00025 / 1k characters -> $0.001 / 1k tokens
 	"gemini-pro-vision": 1, // $0.00025 / 1k characters -> $0.001 / 1k tokens
+	// https://open.bigmodel.cn/pricing
+	"glm-4":                     0.1 * RMB,
+	"glm-4v":                    0.1 * RMB,
+	"glm-3-turbo":               0.005 * RMB,
 	"chatglm_turbo":             0.3572, // ￥0.005 / 1k tokens
 	"chatglm_pro":               0.7143, // ￥0.01 / 1k tokens
 	"chatglm_std":               0.3572, // ￥0.005 / 1k tokens
@@ -127,6 +131,23 @@ var ModelRatio = map[string]float64{
 	"moonshot-v1-8k":   0.012 * RMB,
 	"moonshot-v1-32k":  0.024 * RMB,
 	"moonshot-v1-128k": 0.06 * RMB,
+	// https://platform.baichuan-ai.com/price
+	"Baichuan2-Turbo":      0.008 * RMB,
+	"Baichuan2-Turbo-192k": 0.016 * RMB,
+	"Baichuan2-53B":        0.02 * RMB,
+	// https://api.minimax.chat/document/price
+	"abab6-chat":    0.1 * RMB,
+	"abab5.5-chat":  0.015 * RMB,
+	"abab5.5s-chat": 0.005 * RMB,
+}
+
+var DefaultModelRatio map[string]float64
+
+func init() {
+	DefaultModelRatio = make(map[string]float64)
+	for k, v := range ModelRatio {
+		DefaultModelRatio[k] = v
+	}
 }
 
 func ModelRatio2JSONString() string {
@@ -147,6 +168,9 @@ func GetModelRatio(name string) float64 {
 		name = strings.TrimSuffix(name, "-internet")
 	}
 	ratio, ok := ModelRatio[name]
+	if !ok {
+		ratio, ok = DefaultModelRatio[name]
+	}
 	if !ok {
 		logger.SysError("model ratio not found: " + name)
 		return 30

controller/model.go

@@ -4,6 +4,8 @@ import (
 	"fmt"
 	"github.com/gin-gonic/gin"
 	"github.com/songquanpeng/one-api/relay/channel/ai360"
+	"github.com/songquanpeng/one-api/relay/channel/baichuan"
+	"github.com/songquanpeng/one-api/relay/channel/minimax"
 	"github.com/songquanpeng/one-api/relay/channel/moonshot"
 	"github.com/songquanpeng/one-api/relay/constant"
 	"github.com/songquanpeng/one-api/relay/helper"
@@ -98,6 +100,28 @@ func init() {
 			Parent:     nil,
 		})
 	}
+	for _, modelName := range baichuan.ModelList {
+		openAIModels = append(openAIModels, OpenAIModels{
+			Id:         modelName,
+			Object:     "model",
+			Created:    1626777600,
+			OwnedBy:    "baichuan",
+			Permission: permission,
+			Root:       modelName,
+			Parent:     nil,
+		})
+	}
+	for _, modelName := range minimax.ModelList {
+		openAIModels = append(openAIModels, OpenAIModels{
+			Id:         modelName,
+			Object:     "model",
+			Created:    1626777600,
+			OwnedBy:    "minimax",
+			Permission: permission,
+			Root:       modelName,
+			Parent:     nil,
+		})
+	}
 	openAIModelsMap = make(map[string]OpenAIModels)
 	for _, model := range openAIModels {
 		openAIModelsMap[model.Id] = model

controller/relay.go

@@ -1,23 +1,27 @@
 package controller
 
 import (
+	"bytes"
+	"context"
 	"fmt"
 	"github.com/gin-gonic/gin"
+	"github.com/songquanpeng/one-api/common"
 	"github.com/songquanpeng/one-api/common/config"
 	"github.com/songquanpeng/one-api/common/helper"
 	"github.com/songquanpeng/one-api/common/logger"
+	"github.com/songquanpeng/one-api/middleware"
+	dbmodel "github.com/songquanpeng/one-api/model"
 	"github.com/songquanpeng/one-api/relay/constant"
 	"github.com/songquanpeng/one-api/relay/controller"
 	"github.com/songquanpeng/one-api/relay/model"
 	"github.com/songquanpeng/one-api/relay/util"
+	"io"
 	"net/http"
-	"strconv"
 )
 
 // https://platform.openai.com/docs/api-reference/chat
 
-func Relay(c *gin.Context) {
-	relayMode := constant.Path2RelayMode(c.Request.URL.Path)
+func relay(c *gin.Context, relayMode int) *model.ErrorWithStatusCode {
 	var err *model.ErrorWithStatusCode
 	switch relayMode {
 	case constant.RelayModeImagesGenerations:
@@ -31,33 +35,90 @@ func Relay(c *gin.Context) {
 	default:
 		err = controller.RelayTextHelper(c)
 	}
-	if err != nil {
+	return err
+}
+
+func Relay(c *gin.Context) {
+	ctx := c.Request.Context()
+	relayMode := constant.Path2RelayMode(c.Request.URL.Path)
+	if config.DebugEnabled {
+		requestBody, _ := common.GetRequestBody(c)
+		logger.Debugf(ctx, "request body: %s", string(requestBody))
+	}
+	bizErr := relay(c, relayMode)
+	if bizErr == nil {
+		return
+	}
+	channelId := c.GetInt("channel_id")
+	lastFailedChannelId := channelId
+	channelName := c.GetString("channel_name")
+	group := c.GetString("group")
+	originalModel := c.GetString("original_model")
+	go processChannelRelayError(ctx, channelId, channelName, bizErr)
 	requestId := c.GetString(logger.RequestIdKey)
-		retryTimesStr := c.Query("retry")
-		retryTimes, _ := strconv.Atoi(retryTimesStr)
-		if retryTimesStr == "" {
-			retryTimes = config.RetryTimes
+	retryTimes := config.RetryTimes
+	if !shouldRetry(c, bizErr.StatusCode) {
+		logger.Errorf(ctx, "relay error happen, status code is %d, won't retry in this case", bizErr.StatusCode)
+		retryTimes = 0
 	}
-		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.Error.Message = "当前分组上游负载已饱和，请稍后再试"
+	for i := retryTimes; i > 0; i-- {
+		channel, err := dbmodel.CacheGetRandomSatisfiedChannel(group, originalModel)
+		if err != nil {
+			logger.Errorf(ctx, "CacheGetRandomSatisfiedChannel failed: %w", err)
+			break
 		}
-			err.Error.Message = helper.MessageWithRequestId(err.Error.Message, requestId)
-			c.JSON(err.StatusCode, gin.H{
-				"error": err.Error,
+		logger.Infof(ctx, "using channel #%d to retry (remain times %d)", channel.Id, i)
+		if channel.Id == lastFailedChannelId {
+			continue
+		}
+		middleware.SetupContextForSelectedChannel(c, channel, originalModel)
+		requestBody, err := common.GetRequestBody(c)
+		c.Request.Body = io.NopCloser(bytes.NewBuffer(requestBody))
+		bizErr = relay(c, relayMode)
+		if bizErr == nil {
+			return
+		}
+		channelId := c.GetInt("channel_id")
+		lastFailedChannelId = channelId
+		channelName := c.GetString("channel_name")
+		go processChannelRelayError(ctx, channelId, channelName, bizErr)
+	}
+	if bizErr != nil {
+		if bizErr.StatusCode == http.StatusTooManyRequests {
+			bizErr.Error.Message = "当前分组上游负载已饱和，请稍后再试"
+		}
+		bizErr.Error.Message = helper.MessageWithRequestId(bizErr.Error.Message, requestId)
+		c.JSON(bizErr.StatusCode, gin.H{
+			"error": bizErr.Error,
 		})
 	}
-		channelId := c.GetInt("channel_id")
-		logger.Error(c.Request.Context(), fmt.Sprintf("relay error (channel #%d): %s", channelId, err.Message))
+}
+
+func shouldRetry(c *gin.Context, statusCode int) bool {
+	if _, ok := c.Get("specific_channel_id"); ok {
+		return false
+	}
+	if statusCode == http.StatusTooManyRequests {
+		return true
+	}
+	if statusCode/100 == 5 {
+		return true
+	}
+	if statusCode == http.StatusBadRequest {
+		return false
+	}
+	if statusCode/100 == 2 {
+		return false
+	}
+	return true
+}
+
+func processChannelRelayError(ctx context.Context, channelId int, channelName string, err *model.ErrorWithStatusCode) {
+	logger.Errorf(ctx, "relay error (channel #%d): %s", channelId, err.Message)
 	// https://platform.openai.com/docs/guides/error-codes/api-errors
 	if util.ShouldDisableChannel(&err.Error, err.StatusCode) {
-			channelId := c.GetInt("channel_id")
-			channelName := c.GetString("channel_name")
 		disableChannel(channelId, channelName, err.Message)
 	}
-	}
 }
 
 func RelayNotImplemented(c *gin.Context) {

i18n/en.json

@@ -456,6 +456,7 @@
   "已绑定的邮箱账户": "Email Account Bound",
   "用户信息更新成功！": "User information updated successfully!",
   "模型倍率 %.2f，分组倍率 %.2f": "model rate %.2f, group rate %.2f",
+  "模型倍率 %.2f，分组倍率 %.2f，补全倍率 %.2f": "model rate %.2f, group rate %.2f, completion rate %.2f",
   "使用明细（总消耗额度：{renderQuota(stat.quota)}）": "Usage Details (Total Consumption Quota: {renderQuota(stat.quota)})",
   "用户名称": "User Name",
   "令牌名称": "Token Name",

middleware/auth.go

@@ -108,7 +108,7 @@ func TokenAuth() func(c *gin.Context) {
 		c.Set("token_name", token.Name)
 		if len(parts) > 1 {
 			if model.IsAdmin(token.UserId) {
-				c.Set("channelId", parts[1])
+				c.Set("specific_channel_id", parts[1])
 			} else {
 				abortWithMessage(c, http.StatusForbidden, "普通用户不支持指定渠道")
 				return

middleware/distributor.go

@@ -21,8 +21,9 @@ func Distribute() func(c *gin.Context) {
 		userId := c.GetInt("id")
 		userGroup, _ := model.CacheGetUserGroup(userId)
 		c.Set("group", userGroup)
+		var requestModel string
 		var channel *model.Channel
-		channelId, ok := c.Get("channelId")
+		channelId, ok := c.Get("specific_channel_id")
 		if ok {
 			id, err := strconv.Atoi(channelId.(string))
 			if err != nil {
@@ -66,6 +67,7 @@ func Distribute() func(c *gin.Context) {
 					modelRequest.Model = "whisper-1"
 				}
 			}
+			requestModel = modelRequest.Model
 			channel, err = model.CacheGetRandomSatisfiedChannel(userGroup, modelRequest.Model)
 			if err != nil {
 				message := fmt.Sprintf("当前分组 %s 下对于模型 %s 无可用渠道", userGroup, modelRequest.Model)
@@ -77,10 +79,17 @@ func Distribute() func(c *gin.Context) {
 				return
 			}
 		}
+		SetupContextForSelectedChannel(c, channel, requestModel)
+		c.Next()
+	}
+}
+
+func SetupContextForSelectedChannel(c *gin.Context, channel *model.Channel, modelName string) {
 	c.Set("channel", channel.Type)
 	c.Set("channel_id", channel.Id)
 	c.Set("channel_name", channel.Name)
 	c.Set("model_mapping", channel.GetModelMapping())
+	c.Set("original_model", modelName) // for retry
 	c.Request.Header.Set("Authorization", fmt.Sprintf("Bearer %s", channel.Key))
 	c.Set("base_url", channel.GetBaseURL())
 	// this is for backward compatibility
@@ -100,6 +109,4 @@ func Distribute() func(c *gin.Context) {
 	for k, v := range cfg {
 		c.Set(common.ConfigKeyPrefix+k, v)
 	}
-		c.Next()
-	}
 }

middleware/request-id.go

@@ -9,7 +9,7 @@ import (
 
 func RequestId() func(c *gin.Context) {
 	return func(c *gin.Context) {
-		id := helper.GetTimeString() + helper.GetRandomString(8)
+		id := helper.GetTimeString() + helper.GetRandomNumberString(8)
 		c.Set(logger.RequestIdKey, id)
 		ctx := context.WithValue(c.Request.Context(), logger.RequestIdKey, id)
 		c.Request = c.Request.WithContext(ctx)

model/cache.go

@@ -94,7 +94,7 @@ func CacheUpdateUserQuota(id int) error {
 	if !common.RedisEnabled {
 		return nil
 	}
-	quota, err := GetUserQuota(id)
+	quota, err := CacheGetUserQuota(id)
 	if err != nil {
 		return err
 	}

relay/channel/aiproxy/main.go

@@ -53,7 +53,7 @@ func responseAIProxyLibrary2OpenAI(response *LibraryResponse) *openai.TextRespon
 		FinishReason: "stop",
 	}
 	fullTextResponse := openai.TextResponse{
-		Id:      helper.GetUUID(),
+		Id:      fmt.Sprintf("chatcmpl-%s", helper.GetUUID()),
 		Object:  "chat.completion",
 		Created: helper.GetTimestamp(),
 		Choices: []openai.TextResponseChoice{choice},
@@ -66,7 +66,7 @@ func documentsAIProxyLibrary(documents []LibraryDocument) *openai.ChatCompletion
 	choice.Delta.Content = aiProxyDocuments2Markdown(documents)
 	choice.FinishReason = &constant.StopFinishReason
 	return &openai.ChatCompletionsStreamResponse{
-		Id:      helper.GetUUID(),
+		Id:      fmt.Sprintf("chatcmpl-%s", helper.GetUUID()),
 		Object:  "chat.completion.chunk",
 		Created: helper.GetTimestamp(),
 		Model:   "",
@@ -78,7 +78,7 @@ func streamResponseAIProxyLibrary2OpenAI(response *LibraryStreamResponse) *opena
 	var choice openai.ChatCompletionsStreamResponseChoice
 	choice.Delta.Content = response.Content
 	return &openai.ChatCompletionsStreamResponse{
-		Id:      helper.GetUUID(),
+		Id:      fmt.Sprintf("chatcmpl-%s", helper.GetUUID()),
 		Object:  "chat.completion.chunk",
 		Created: helper.GetTimestamp(),
 		Model:   response.Model,

relay/channel/baichuan/constants.go

@@ -0,0 +1,7 @@
+package baichuan
+
+var ModelList = []string{
+	"Baichuan2-Turbo",
+	"Baichuan2-Turbo-192k",
+	"Baichuan-Text-Embedding",
+}

relay/channel/minimax/constants.go

@@ -0,0 +1,7 @@
+package minimax
+
+var ModelList = []string{
+	"abab5.5s-chat",
+	"abab5.5-chat",
+	"abab6-chat",
+}

relay/channel/minimax/main.go

@@ -0,0 +1,14 @@
+package minimax
+
+import (
+	"fmt"
+	"github.com/songquanpeng/one-api/relay/constant"
+	"github.com/songquanpeng/one-api/relay/util"
+)
+
+func GetRequestURL(meta *util.RelayMeta) (string, error) {
+	if meta.Mode == constant.RelayModeChatCompletions {
+		return fmt.Sprintf("%s/v1/text/chatcompletion_v2", meta.BaseURL), nil
+	}
+	return "", fmt.Errorf("unsupported relay mode %d for minimax", meta.Mode)
+}

relay/channel/openai/adaptor.go

@@ -7,6 +7,8 @@ import (
 	"github.com/songquanpeng/one-api/common"
 	"github.com/songquanpeng/one-api/relay/channel"
 	"github.com/songquanpeng/one-api/relay/channel/ai360"
+	"github.com/songquanpeng/one-api/relay/channel/baichuan"
+	"github.com/songquanpeng/one-api/relay/channel/minimax"
 	"github.com/songquanpeng/one-api/relay/channel/moonshot"
 	"github.com/songquanpeng/one-api/relay/model"
 	"github.com/songquanpeng/one-api/relay/util"
@@ -24,7 +26,8 @@ func (a *Adaptor) Init(meta *util.RelayMeta) {
 }
 
 func (a *Adaptor) GetRequestURL(meta *util.RelayMeta) (string, error) {
-	if meta.ChannelType == common.ChannelTypeAzure {
+	switch meta.ChannelType {
+	case common.ChannelTypeAzure:
 		// https://learn.microsoft.com/en-us/azure/cognitive-services/openai/chatgpt-quickstart?pivots=rest-api&tabs=command-line#rest-api
 		requestURL := strings.Split(meta.RequestURLPath, "?")[0]
 		requestURL = fmt.Sprintf("%s?api-version=%s", requestURL, meta.APIVersion)
@@ -38,8 +41,11 @@ func (a *Adaptor) GetRequestURL(meta *util.RelayMeta) (string, error) {
 
 		requestURL = fmt.Sprintf("/openai/deployments/%s/%s", model_, task)
 		return util.GetFullRequestURL(meta.BaseURL, requestURL, meta.ChannelType), nil
-	}
+	case common.ChannelTypeMinimax:
+		return minimax.GetRequestURL(meta)
+	default:
 		return util.GetFullRequestURL(meta.BaseURL, meta.RequestURLPath, meta.ChannelType), nil
+	}
 }
 
 func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Request, meta *util.RelayMeta) error {
@@ -70,7 +76,7 @@ func (a *Adaptor) DoRequest(c *gin.Context, meta *util.RelayMeta, requestBody io
 func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, meta *util.RelayMeta) (usage *model.Usage, err *model.ErrorWithStatusCode) {
 	if meta.IsStream {
 		var responseText string
-		err, responseText = StreamHandler(c, resp, meta.Mode)
+		err, responseText, _ = StreamHandler(c, resp, meta.Mode)
 		usage = ResponseText2Usage(responseText, meta.ActualModelName, meta.PromptTokens)
 	} else {
 		err, usage = Handler(c, resp, meta.PromptTokens, meta.ActualModelName)
@@ -84,6 +90,10 @@ func (a *Adaptor) GetModelList() []string {
 		return ai360.ModelList
 	case common.ChannelTypeMoonshot:
 		return moonshot.ModelList
+	case common.ChannelTypeBaichuan:
+		return baichuan.ModelList
+	case common.ChannelTypeMinimax:
+		return minimax.ModelList
 	default:
 		return ModelList
 	}
@@ -97,6 +107,10 @@ func (a *Adaptor) GetChannelName() string {
 		return "360"
 	case common.ChannelTypeMoonshot:
 		return "moonshot"
+	case common.ChannelTypeBaichuan:
+		return "baichuan"
+	case common.ChannelTypeMinimax:
+		return "minimax"
 	default:
 		return "openai"
 	}

relay/channel/openai/main.go

@@ -14,7 +14,7 @@ import (
 	"strings"
 )
 
-func StreamHandler(c *gin.Context, resp *http.Response, relayMode int) (*model.ErrorWithStatusCode, string) {
+func StreamHandler(c *gin.Context, resp *http.Response, relayMode int) (*model.ErrorWithStatusCode, string, *model.Usage) {
 	responseText := ""
 	scanner := bufio.NewScanner(resp.Body)
 	scanner.Split(func(data []byte, atEOF bool) (advance int, token []byte, err error) {
@@ -31,6 +31,7 @@ func StreamHandler(c *gin.Context, resp *http.Response, relayMode int) (*model.E
 	})
 	dataChan := make(chan string)
 	stopChan := make(chan bool)
+	var usage *model.Usage
 	go func() {
 		for scanner.Scan() {
 			data := scanner.Text()
@@ -54,6 +55,9 @@ func StreamHandler(c *gin.Context, resp *http.Response, relayMode int) (*model.E
 					for _, choice := range streamResponse.Choices {
 						responseText += choice.Delta.Content
 					}
+					if streamResponse.Usage != nil {
+						usage = streamResponse.Usage
+					}
 				case constant.RelayModeCompletions:
 					var streamResponse CompletionsStreamResponse
 					err := json.Unmarshal([]byte(data), &streamResponse)
@@ -86,9 +90,9 @@ func StreamHandler(c *gin.Context, resp *http.Response, relayMode int) (*model.E
 	})
 	err := resp.Body.Close()
 	if err != nil {
-		return ErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), ""
+		return ErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), "", nil
 	}
-	return nil, responseText
+	return nil, responseText, usage
 }
 
 func Handler(c *gin.Context, resp *http.Response, promptTokens int, modelName string) (*model.ErrorWithStatusCode, *model.Usage) {

relay/channel/openai/model.go

@@ -118,8 +118,10 @@ type ImageResponse struct {
 }
 
 type ChatCompletionsStreamResponseChoice struct {
+	Index int `json:"index"`
 	Delta struct {
 		Content string `json:"content"`
+		Role    string `json:"role,omitempty"`
 	} `json:"delta"`
 	FinishReason *string `json:"finish_reason,omitempty"`
 }
@@ -130,6 +132,7 @@ type ChatCompletionsStreamResponse struct {
 	Created int64                                 `json:"created"`
 	Model   string                                `json:"model"`
 	Choices []ChatCompletionsStreamResponseChoice `json:"choices"`
+	Usage   *model.Usage                          `json:"usage"`
 }
 
 type CompletionsStreamResponse struct {

relay/channel/tencent/main.go

@@ -81,6 +81,7 @@ func responseTencent2OpenAI(response *ChatResponse) *openai.TextResponse {
 
 func streamResponseTencent2OpenAI(TencentResponse *ChatResponse) *openai.ChatCompletionsStreamResponse {
 	response := openai.ChatCompletionsStreamResponse{
+		Id:      fmt.Sprintf("chatcmpl-%s", helper.GetUUID()),
 		Object:  "chat.completion.chunk",
 		Created: helper.GetTimestamp(),
 		Model:   "tencent-hunyuan",

relay/channel/xunfei/main.go

@@ -70,6 +70,7 @@ func responseXunfei2OpenAI(response *ChatResponse) *openai.TextResponse {
 		FinishReason: constant.StopFinishReason,
 	}
 	fullTextResponse := openai.TextResponse{
+		Id:      fmt.Sprintf("chatcmpl-%s", helper.GetUUID()),
 		Object:  "chat.completion",
 		Created: helper.GetTimestamp(),
 		Choices: []openai.TextResponseChoice{choice},
@@ -92,6 +93,7 @@ func streamResponseXunfei2OpenAI(xunfeiResponse *ChatResponse) *openai.ChatCompl
 		choice.FinishReason = &constant.StopFinishReason
 	}
 	response := openai.ChatCompletionsStreamResponse{
+		Id:      fmt.Sprintf("chatcmpl-%s", helper.GetUUID()),
 		Object:  "chat.completion.chunk",
 		Created: helper.GetTimestamp(),
 		Model:   "SparkDesk",

relay/channel/zhipu/adaptor.go

@@ -5,20 +5,35 @@ import (
 	"fmt"
 	"github.com/gin-gonic/gin"
 	"github.com/songquanpeng/one-api/relay/channel"
+	"github.com/songquanpeng/one-api/relay/channel/openai"
 	"github.com/songquanpeng/one-api/relay/model"
 	"github.com/songquanpeng/one-api/relay/util"
 	"io"
 	"net/http"
+	"strings"
 )
 
 type Adaptor struct {
+	APIVersion string
 }
 
 func (a *Adaptor) Init(meta *util.RelayMeta) {
 
 }
 
+func (a *Adaptor) SetVersionByModeName(modelName string) {
+	if strings.HasPrefix(modelName, "glm-") {
+		a.APIVersion = "v4"
+	} else {
+		a.APIVersion = "v3"
+	}
+}
+
 func (a *Adaptor) GetRequestURL(meta *util.RelayMeta) (string, error) {
+	a.SetVersionByModeName(meta.ActualModelName)
+	if a.APIVersion == "v4" {
+		return fmt.Sprintf("%s/api/paas/v4/chat/completions", meta.BaseURL), nil
+	}
 	method := "invoke"
 	if meta.IsStream {
 		method = "sse-invoke"
@@ -37,6 +52,13 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *model.G
 	if request == nil {
 		return nil, errors.New("request is nil")
 	}
+	if request.TopP >= 1 {
+		request.TopP = 0.99
+	}
+	a.SetVersionByModeName(request.Model)
+	if a.APIVersion == "v4" {
+		return request, nil
+	}
 	return ConvertRequest(*request), nil
 }
 
@@ -44,7 +66,19 @@ func (a *Adaptor) DoRequest(c *gin.Context, meta *util.RelayMeta, requestBody io
 	return channel.DoRequestHelper(a, c, meta, requestBody)
 }
 
+func (a *Adaptor) DoResponseV4(c *gin.Context, resp *http.Response, meta *util.RelayMeta) (usage *model.Usage, err *model.ErrorWithStatusCode) {
+	if meta.IsStream {
+		err, _, usage = openai.StreamHandler(c, resp, meta.Mode)
+	} else {
+		err, usage = openai.Handler(c, resp, meta.PromptTokens, meta.ActualModelName)
+	}
+	return
+}
+
 func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, meta *util.RelayMeta) (usage *model.Usage, err *model.ErrorWithStatusCode) {
+	if a.APIVersion == "v4" {
+		return a.DoResponseV4(c, resp, meta)
+	}
 	if meta.IsStream {
 		err, usage = StreamHandler(c, resp)
 	} else {

relay/channel/zhipu/constants.go

@@ -2,4 +2,5 @@ package zhipu
 
 var ModelList = []string{
 	"chatglm_turbo", "chatglm_pro", "chatglm_std", "chatglm_lite",
+	"glm-4", "glm-4v", "glm-3-turbo",
 }

relay/controller/text.go

@@ -55,7 +55,8 @@ func RelayTextHelper(c *gin.Context) *model.ErrorWithStatusCode {
 	var requestBody io.Reader
 	if meta.APIType == constant.APITypeOpenAI {
 		// no need to convert request for openai
-		if isModelMapped {
+		shouldResetRequestBody := isModelMapped || meta.ChannelType == common.ChannelTypeBaichuan // frequency_penalty 0 is not acceptable for baichuan
+		if shouldResetRequestBody {
 			jsonStr, err := json.Marshal(textRequest)
 			if err != nil {
 				return openai.ErrorWrapper(err, "json_marshal_failed", http.StatusInternalServerError)

web/berry/src/constants/ChannelConstants.js

@@ -71,6 +71,18 @@ export const CHANNEL_OPTIONS = {
     value: 23,
     color: 'default'
   },
+  26: {
+    key: 26,
+    text: '百川大模型',
+    value: 26,
+    color: 'default'
+  },
+  27: {
+    key: 27,
+    text: 'MiniMax',
+    value: 27,
+    color: 'default'
+  },
   8: {
     key: 8,
     text: '自定义渠道',

web/berry/src/views/Channel/type/Config.js

@@ -67,7 +67,7 @@ const typeConfig = {
   },
   16: {
     input: {
-      models: ["chatglm_turbo", "chatglm_pro", "chatglm_std", "chatglm_lite"],
+      models: ["glm-4", "glm-4v", "glm-3-turbo", "chatglm_turbo", "chatglm_pro", "chatglm_std", "chatglm_lite"],
     },
     modelGroup: "zhipu",
   },
@@ -145,6 +145,24 @@ const typeConfig = {
     },
     modelGroup: "google gemini",
   },
+  25: {
+    input: {
+      models: ['moonshot-v1-8k', 'moonshot-v1-32k', 'moonshot-v1-128k'],
+    },
+    modelGroup: "moonshot",
+  },
+  26: {
+    input: {
+      models: ['Baichuan2-Turbo', 'Baichuan2-Turbo-192k', 'Baichuan-Text-Embedding'],
+    },
+    modelGroup: "baichuan",
+  },
+  27: {
+    input: {
+      models: ['abab5.5s-chat', 'abab5.5-chat', 'abab6-chat'],
+    },
+    modelGroup: "minimax",
+  },
 };
 
 export { defaultConfig, typeConfig };

web/build.sh

@@ -1,13 +1,13 @@
 #!/bin/sh
 
 version=$(cat VERSION)
-themes=$(cat THEMES)
-IFS=$'\n'
+pwd
 
-for theme in $themes; do
+while IFS= read -r theme; do
     echo "Building theme: $theme"
-    cd $theme
+    rm -r build/$theme
+    cd "$theme"
     npm install
     DISABLE_ESLINT_PLUGIN='true' REACT_APP_VERSION=$version npm run build
     cd ..
-done
+done < THEMES

web/default/src/constants/channel.constants.js

@@ -11,6 +11,8 @@ export const CHANNEL_OPTIONS = [
   { key: 19, text: '360 智脑', value: 19, color: 'blue' },
   { key: 25, text: 'Moonshot AI', value: 25, color: 'black' },
   { key: 23, text: '腾讯混元', value: 23, color: 'teal' },
+  { key: 26, text: '百川大模型', value: 26, color: 'orange' },
+  { key: 27, text: 'MiniMax', value: 27, color: 'red' },
   { key: 8, text: '自定义渠道', value: 8, color: 'pink' },
   { key: 22, text: '知识库：FastGPT', value: 22, color: 'blue' },
   { key: 21, text: '知识库：AI Proxy', value: 21, color: 'purple' },

web/default/src/pages/Channel/EditChannel.js

@@ -79,7 +79,7 @@ const EditChannel = () => {
           localModels = [...localModels, ...withInternetVersion];
           break;
         case 16:
-          localModels = ['chatglm_turbo', 'chatglm_pro', 'chatglm_std', 'chatglm_lite'];
+          localModels = ["glm-4", "glm-4v", "glm-3-turbo",'chatglm_turbo', 'chatglm_pro', 'chatglm_std', 'chatglm_lite'];
           break;
         case 18:
           localModels = [
@@ -102,6 +102,12 @@ const EditChannel = () => {
         case 25:
           localModels = ['moonshot-v1-8k', 'moonshot-v1-32k', 'moonshot-v1-128k'];
           break;
+        case 26:
+          localModels = ['Baichuan2-Turbo', 'Baichuan2-Turbo-192k', 'Baichuan-Text-Embedding'];
+          break;
+        case 27:
+          localModels = ['abab5.5s-chat', 'abab5.5-chat', 'abab6-chat'];
+          break;
       }
       setInputs((inputs) => ({ ...inputs, models: localModels }));
     }