feat: implement parameter cleaning for Gemini functions

fix(relay): 优化数据流处理

README.md

@@ -117,7 +117,6 @@ New API提供了丰富的功能，详细特性请参考[特性说明](https://do
 
 > [!TIP]
 > 最新版Docker镜像：`calciumion/new-api:latest`  
-> 默认账号root 密码123456
 
 ### 多机部署注意事项
 - 必须设置环境变量 `SESSION_SECRET`，否则会导致多机部署时登录状态不一致

common/constants.go

@@ -1,8 +1,8 @@
 package common
 
 import (
-	"os"
-	"strconv"
+	//"os"
+	//"strconv"
 	"sync"
 	"time"
 
@@ -63,8 +63,8 @@ var EmailDomainWhitelist = []string{
 	"foxmail.com",
 }
 
-var DebugEnabled = os.Getenv("DEBUG") == "true"
-var MemoryCacheEnabled = os.Getenv("MEMORY_CACHE_ENABLED") == "true"
+var DebugEnabled bool
+var MemoryCacheEnabled bool
 
 var LogConsumeEnabled = true
 
@@ -103,22 +103,22 @@ var RetryTimes = 0
 
 //var RootUserEmail = ""
 
-var IsMasterNode = os.Getenv("NODE_TYPE") != "slave"
+var IsMasterNode bool
 
-var requestInterval, _ = strconv.Atoi(os.Getenv("POLLING_INTERVAL"))
-var RequestInterval = time.Duration(requestInterval) * time.Second
+var requestInterval int
+var RequestInterval time.Duration
 
-var SyncFrequency = GetEnvOrDefault("SYNC_FREQUENCY", 60) // unit is second
+var SyncFrequency int // unit is second
 
 var BatchUpdateEnabled = false
-var BatchUpdateInterval = GetEnvOrDefault("BATCH_UPDATE_INTERVAL", 5)
+var BatchUpdateInterval int
 
-var RelayTimeout = GetEnvOrDefault("RELAY_TIMEOUT", 0) // unit is second
+var RelayTimeout int // unit is second
 
-var GeminiSafetySetting = GetEnvOrDefaultString("GEMINI_SAFETY_SETTING", "BLOCK_NONE")
+var GeminiSafetySetting string
 
 // https://docs.cohere.com/docs/safety-modes Type; NONE/CONTEXTUAL/STRICT
-var CohereSafetySetting = GetEnvOrDefaultString("COHERE_SAFETY_SETTING", "NONE")
+var CohereSafetySetting string
 
 const (
 	RequestIdKey = "X-Oneapi-Request-Id"
@@ -145,13 +145,13 @@ var (
 // All duration's unit is seconds
 // Shouldn't larger then RateLimitKeyExpirationDuration
 var (
-	GlobalApiRateLimitEnable   = GetEnvOrDefaultBool("GLOBAL_API_RATE_LIMIT_ENABLE", true)
-	GlobalApiRateLimitNum      = GetEnvOrDefault("GLOBAL_API_RATE_LIMIT", 180)
-	GlobalApiRateLimitDuration = int64(GetEnvOrDefault("GLOBAL_API_RATE_LIMIT_DURATION", 180))
+	GlobalApiRateLimitEnable   bool
+	GlobalApiRateLimitNum      int
+	GlobalApiRateLimitDuration int64
 
-	GlobalWebRateLimitEnable   = GetEnvOrDefaultBool("GLOBAL_WEB_RATE_LIMIT_ENABLE", true)
-	GlobalWebRateLimitNum      = GetEnvOrDefault("GLOBAL_WEB_RATE_LIMIT", 60)
-	GlobalWebRateLimitDuration = int64(GetEnvOrDefault("GLOBAL_WEB_RATE_LIMIT_DURATION", 180))
+	GlobalWebRateLimitEnable   bool
+	GlobalWebRateLimitNum      int
+	GlobalWebRateLimitDuration int64
 
 	UploadRateLimitNum            = 10
 	UploadRateLimitDuration int64 = 60
@@ -235,6 +235,7 @@ const (
 	ChannelTypeVolcEngine     = 45
 	ChannelTypeBaiduV2        = 46
 	ChannelTypeXinference     = 47
+	ChannelTypeXai            = 48
 	ChannelTypeDummy          // this one is only for count, do not add any channel after this
 
 )
@@ -288,4 +289,5 @@ var ChannelBaseURLs = []string{
 	"https://ark.cn-beijing.volces.com",         //45
 	"https://qianfan.baidubce.com",              //46
 	"",                                          //47
+	"https://api.x.ai",                          //48
 }

common/init.go

@@ -6,6 +6,8 @@ import (
 	"log"
 	"os"
 	"path/filepath"
+	"strconv"
+	"time"
 )
 
 var (
@@ -66,4 +68,31 @@ func LoadEnv() {
 			}
 		}
 	}
+
+	// Initialize variables from constants.go that were using environment variables
+	DebugEnabled = os.Getenv("DEBUG") == "true"
+	MemoryCacheEnabled = os.Getenv("MEMORY_CACHE_ENABLED") == "true"
+	IsMasterNode = os.Getenv("NODE_TYPE") != "slave"
+	
+	// Parse requestInterval and set RequestInterval
+	requestInterval, _ = strconv.Atoi(os.Getenv("POLLING_INTERVAL"))
+	RequestInterval = time.Duration(requestInterval) * time.Second
+	
+	// Initialize variables with GetEnvOrDefault
+	SyncFrequency = GetEnvOrDefault("SYNC_FREQUENCY", 60)
+	BatchUpdateInterval = GetEnvOrDefault("BATCH_UPDATE_INTERVAL", 5)
+	RelayTimeout = GetEnvOrDefault("RELAY_TIMEOUT", 0)
+	
+	// Initialize string variables with GetEnvOrDefaultString
+	GeminiSafetySetting = GetEnvOrDefaultString("GEMINI_SAFETY_SETTING", "BLOCK_NONE")
+	CohereSafetySetting = GetEnvOrDefaultString("COHERE_SAFETY_SETTING", "NONE")
+	
+	// Initialize rate limit variables
+	GlobalApiRateLimitEnable = GetEnvOrDefaultBool("GLOBAL_API_RATE_LIMIT_ENABLE", true)
+	GlobalApiRateLimitNum = GetEnvOrDefault("GLOBAL_API_RATE_LIMIT", 180)
+	GlobalApiRateLimitDuration = int64(GetEnvOrDefault("GLOBAL_API_RATE_LIMIT_DURATION", 180))
+	
+	GlobalWebRateLimitEnable = GetEnvOrDefaultBool("GLOBAL_WEB_RATE_LIMIT_ENABLE", true)
+	GlobalWebRateLimitNum = GetEnvOrDefault("GLOBAL_WEB_RATE_LIMIT", 60)
+	GlobalWebRateLimitDuration = int64(GetEnvOrDefault("GLOBAL_WEB_RATE_LIMIT_DURATION", 180))
 }

common/json.go

@@ -12,3 +12,7 @@ func DecodeJson(data []byte, v any) error {
 func DecodeJsonStr(data string, v any) error {
 	return DecodeJson(StringToByteSlice(data), v)
 }
+
+func EncodeJson(v any) ([]byte, error) {
+	return json.Marshal(v)
+}

constant/env.go

@@ -4,32 +4,39 @@ import (
 	"one-api/common"
 )
 
-var StreamingTimeout = common.GetEnvOrDefault("STREAMING_TIMEOUT", 60)
-var DifyDebug = common.GetEnvOrDefaultBool("DIFY_DEBUG", true)
-
-var MaxFileDownloadMB = common.GetEnvOrDefault("MAX_FILE_DOWNLOAD_MB", 20)
-
-// ForceStreamOption 覆盖请求参数，强制返回usage信息
-var ForceStreamOption = common.GetEnvOrDefaultBool("FORCE_STREAM_OPTION", true)
-
-var GetMediaToken = common.GetEnvOrDefaultBool("GET_MEDIA_TOKEN", true)
-
-var GetMediaTokenNotStream = common.GetEnvOrDefaultBool("GET_MEDIA_TOKEN_NOT_STREAM", true)
-
-var UpdateTask = common.GetEnvOrDefaultBool("UPDATE_TASK", true)
-
-var AzureDefaultAPIVersion = common.GetEnvOrDefaultString("AZURE_DEFAULT_API_VERSION", "2024-12-01-preview")
+var StreamingTimeout int
+var DifyDebug bool
+var MaxFileDownloadMB int
+var ForceStreamOption bool
+var GetMediaToken bool
+var GetMediaTokenNotStream bool
+var UpdateTask bool
+var AzureDefaultAPIVersion string
+var GeminiVisionMaxImageNum int
+var NotifyLimitCount int
+var NotificationLimitDurationMinute int
+var GenerateDefaultToken bool
 
 //var GeminiModelMap = map[string]string{
 //	"gemini-1.0-pro": "v1",
 //}
 
-var GeminiVisionMaxImageNum = common.GetEnvOrDefault("GEMINI_VISION_MAX_IMAGE_NUM", 16)
-
-var NotifyLimitCount = common.GetEnvOrDefault("NOTIFY_LIMIT_COUNT", 2)
-var NotificationLimitDurationMinute = common.GetEnvOrDefault("NOTIFICATION_LIMIT_DURATION_MINUTE", 10)
-
 func InitEnv() {
+	StreamingTimeout = common.GetEnvOrDefault("STREAMING_TIMEOUT", 60)
+	DifyDebug = common.GetEnvOrDefaultBool("DIFY_DEBUG", true)
+	MaxFileDownloadMB = common.GetEnvOrDefault("MAX_FILE_DOWNLOAD_MB", 20)
+	// ForceStreamOption 覆盖请求参数，强制返回usage信息
+	ForceStreamOption = common.GetEnvOrDefaultBool("FORCE_STREAM_OPTION", true)
+	GetMediaToken = common.GetEnvOrDefaultBool("GET_MEDIA_TOKEN", true)
+	GetMediaTokenNotStream = common.GetEnvOrDefaultBool("GET_MEDIA_TOKEN_NOT_STREAM", true)
+	UpdateTask = common.GetEnvOrDefaultBool("UPDATE_TASK", true)
+	AzureDefaultAPIVersion = common.GetEnvOrDefaultString("AZURE_DEFAULT_API_VERSION", "2024-12-01-preview")
+	GeminiVisionMaxImageNum = common.GetEnvOrDefault("GEMINI_VISION_MAX_IMAGE_NUM", 16)
+	NotifyLimitCount = common.GetEnvOrDefault("NOTIFY_LIMIT_COUNT", 2)
+	NotificationLimitDurationMinute = common.GetEnvOrDefault("NOTIFICATION_LIMIT_DURATION_MINUTE", 10)
+	// GenerateDefaultToken 是否生成初始令牌，默认关闭。
+	GenerateDefaultToken = common.GetEnvOrDefaultBool("GENERATE_DEFAULT_TOKEN", false)
+
 	//modelVersionMapStr := strings.TrimSpace(os.Getenv("GEMINI_MODEL_MAP"))
 	//if modelVersionMapStr == "" {
 	//	return
@@ -43,6 +50,3 @@ func InitEnv() {
 	//	}
 	//}
 }
-
-// GenerateDefaultToken 是否生成初始令牌，默认关闭。
-var GenerateDefaultToken = common.GetEnvOrDefaultBool("GENERATE_DEFAULT_TOKEN", false)

controller/channel-test.go

@@ -192,6 +192,8 @@ func buildTestRequest(model string) *dto.GeneralOpenAIRequest {
 		if !strings.Contains(model, "claude") {
 			testRequest.MaxTokens = 50
 		}
+	} else if strings.Contains(model, "gemini") {
+		testRequest.MaxTokens = 300
 	} else {
 		testRequest.MaxTokens = 10
 	}

controller/channel.go

@@ -119,6 +119,9 @@ func FetchUpstreamModels(c *gin.Context) {
 		baseURL = channel.GetBaseURL()
 	}
 	url := fmt.Sprintf("%s/v1/models", baseURL)
+	if channel.Type == common.ChannelTypeGemini {
+		url = fmt.Sprintf("%s/v1beta/openai/models", baseURL)
+	}
 	body, err := GetResponseBody("GET", url, channel, GetAuthHeader(channel.Key))
 	if err != nil {
 		c.JSON(http.StatusOK, gin.H{
@@ -139,7 +142,11 @@ func FetchUpstreamModels(c *gin.Context) {
 
 	var ids []string
 	for _, model := range result.Data {
-		ids = append(ids, model.ID)
+		id := model.ID
+		if channel.Type == common.ChannelTypeGemini {
+			id = strings.TrimPrefix(id, "models/")
+		}
+		ids = append(ids, id)
 	}
 
 	c.JSON(http.StatusOK, gin.H{

controller/setup.go

@@ -6,6 +6,7 @@ import (
 	"one-api/constant"
 	"one-api/model"
 	"one-api/setting/operation_setting"
+	"time"
 )
 
 type Setup struct {
@@ -124,7 +125,7 @@ func PostSetup(c *gin.Context) {
 	operation_setting.DemoSiteEnabled = req.DemoSiteEnabled
 
 	// Save operation modes to database for persistence
-	err = model.UpdateOption("self_use_mode", boolToString(req.SelfUseModeEnabled))
+	err = model.UpdateOption("SelfUseModeEnabled", boolToString(req.SelfUseModeEnabled))
 	if err != nil {
 		c.JSON(500, gin.H{
 			"success": false,
@@ -133,7 +134,7 @@ func PostSetup(c *gin.Context) {
 		return
 	}
 
-	err = model.UpdateOption("demo_site_mode", boolToString(req.DemoSiteEnabled))
+	err = model.UpdateOption("DemoSiteEnabled", boolToString(req.DemoSiteEnabled))
 	if err != nil {
 		c.JSON(500, gin.H{
 			"success": false,
@@ -144,11 +145,16 @@ func PostSetup(c *gin.Context) {
 
 	// Update setup status
 	constant.Setup = true
-	err = model.UpdateOption("setup", "true")
+
+	setup := model.Setup{
+		Version:       common.Version,
+		InitializedAt: time.Now().Unix(),
+	}
+	err = model.DB.Create(&setup).Error
 	if err != nil {
 		c.JSON(500, gin.H{
 			"success": false,
-			"message": "设置初始化状态失败: " + err.Error(),
+			"message": "系统初始化失败: " + err.Error(),
 		})
 		return
 	}

dto/realtime.go

@@ -45,15 +45,16 @@ type RealtimeUsage struct {
 
 type InputTokenDetails struct {
 	CachedTokens         int `json:"cached_tokens"`
-	CachedCreationTokens int
+	CachedCreationTokens int `json:"-"`
 	TextTokens           int `json:"text_tokens"`
 	AudioTokens          int `json:"audio_tokens"`
 	ImageTokens          int `json:"image_tokens"`
 }
 
 type OutputTokenDetails struct {
-	TextTokens  int `json:"text_tokens"`
-	AudioTokens int `json:"audio_tokens"`
+	TextTokens      int `json:"text_tokens"`
+	AudioTokens     int `json:"audio_tokens"`
+	ReasoningTokens int `json:"reasoning_tokens"`
 }
 
 type RealtimeSession struct {

main.go

@@ -12,6 +12,7 @@ import (
 	"one-api/model"
 	"one-api/router"
 	"one-api/service"
+	"one-api/setting/operation_setting"
 	"os"
 	"strconv"
 
@@ -51,6 +52,9 @@ func main() {
 	if err != nil {
 		common.FatalLog("failed to initialize database: " + err.Error())
 	}
+
+	model.CheckSetup()
+
 	// Initialize SQL Database
 	err = model.InitLogDB()
 	if err != nil {
@@ -69,10 +73,13 @@ func main() {
 		common.FatalLog("failed to initialize Redis: " + err.Error())
 	}
 
+	// Initialize model settings
+	operation_setting.InitModelSettings()
 	// Initialize constants
 	constant.InitEnv()
 	// Initialize options
 	model.InitOptionMap()
+
 	if common.RedisEnabled {
 		// for compatibility with old versions
 		common.MemoryCacheEnabled = true

model/main.go

@@ -56,23 +56,30 @@ func createRootAccountIfNeed() error {
 	return nil
 }
 
-func checkSetup() {
-	if GetSetup() == nil {
+func CheckSetup() {
+	setup := GetSetup()
+	if setup == nil {
+		// No setup record exists, check if we have a root user
 		if RootUserExists() {
 			common.SysLog("system is not initialized, but root user exists")
 			// Create setup record
-			setup := Setup{
+			newSetup := Setup{
 				Version:       common.Version,
 				InitializedAt: time.Now().Unix(),
 			}
-			err := DB.Create(&setup).Error
+			err := DB.Create(&newSetup).Error
 			if err != nil {
 				common.SysLog("failed to create setup record: " + err.Error())
 			}
 			constant.Setup = true
 		} else {
+			common.SysLog("system is not initialized and no root user exists")
 			constant.Setup = false
 		}
+	} else {
+		// Setup record exists, system is initialized
+		common.SysLog("system is already initialized at: " + time.Unix(setup.InitializedAt, 0).String())
+		constant.Setup = true
 	}
 }
 
@@ -237,7 +244,6 @@ func migrateDB() error {
 	}
 	err = DB.AutoMigrate(&Setup{})
 	common.SysLog("database migrated")
-	checkSetup()
 	//err = createRootAccountIfNeed()
 	return err
 }

relay/channel/dify/relay-dify.go

@@ -1,7 +1,6 @@
 package dify
 
 import (
-	"bufio"
 	"bytes"
 	"encoding/base64"
 	"encoding/json"
@@ -213,12 +212,8 @@ func streamResponseDify2OpenAI(difyResponse DifyChunkChatCompletionResponse) *dt
 func difyStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
 	var responseText string
 	usage := &dto.Usage{}
-	scanner := bufio.NewScanner(resp.Body)
-	scanner.Split(bufio.ScanLines)
 	var nodeToken int
-
 	helper.SetEventStreamHeaders(c)
-
 	helper.StreamScannerHandler(c, resp, info, func(data string) bool {
 		var difyResponse DifyChunkChatCompletionResponse
 		err := json.Unmarshal([]byte(data), &difyResponse)
@@ -247,13 +242,10 @@ func difyStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.Re
 		}
 		return true
 	})
-	if err := scanner.Err(); err != nil {
-		common.SysError("error reading stream: " + err.Error())
-	}
 	helper.Done(c)
 	err := resp.Body.Close()
 	if err != nil {
-		//return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+		// return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
 		common.SysError("close_response_body_failed: " + err.Error())
 	}
 	if usage.TotalTokens == 0 {

relay/channel/gemini/constant.go

@@ -16,6 +16,8 @@ var ModelList = []string{
 	"gemini-2.0-pro-exp",
 	// thinking exp
 	"gemini-2.0-flash-thinking-exp",
+	"gemini-2.5-pro-exp-03-25",
+	"gemini-2.5-pro-preview-03-25",
 	// imagen models
 	"imagen-3.0-generate-002",
 	// embedding models

relay/channel/gemini/relay-gemini.go

@@ -56,6 +56,7 @@ func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) (*GeminiChatReque
 				continue
 			}
 			if tool.Function.Parameters != nil {
+
 				params, ok := tool.Function.Parameters.(map[string]interface{})
 				if ok {
 					if props, hasProps := params["properties"].(map[string]interface{}); hasProps {
@@ -65,6 +66,9 @@ func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) (*GeminiChatReque
 					}
 				}
 			}
+			// Clean the parameters before appending
+			cleanedParams := cleanFunctionParameters(tool.Function.Parameters)
+			tool.Function.Parameters = cleanedParams
 			functions = append(functions, tool.Function)
 		}
 		if codeExecution {
@@ -86,11 +90,11 @@ func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) (*GeminiChatReque
 		// json_data, _ := json.Marshal(geminiRequest.Tools)
 		// common.SysLog("tools_json: " + string(json_data))
 	} else if textRequest.Functions != nil {
-		geminiRequest.Tools = []GeminiChatTool{
-			{
-				FunctionDeclarations: textRequest.Functions,
-			},
-		}
+		//geminiRequest.Tools = []GeminiChatTool{
+		//	{
+		//		FunctionDeclarations: textRequest.Functions,
+		//	},
+		//}
 	}
 
 	if textRequest.ResponseFormat != nil && (textRequest.ResponseFormat.Type == "json_schema" || textRequest.ResponseFormat.Type == "json_object") {
@@ -229,6 +233,96 @@ func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) (*GeminiChatReque
 	return &geminiRequest, nil
 }
 
+
+// cleanFunctionParameters recursively removes unsupported fields from Gemini function parameters.
+func cleanFunctionParameters(params interface{}) interface{} {
+	if params == nil {
+		return nil
+	}
+
+	paramMap, ok := params.(map[string]interface{})
+	if !ok {
+		// Not a map, return as is (e.g., could be an array or primitive)
+		return params
+	}
+
+	// Create a copy to avoid modifying the original
+	cleanedMap := make(map[string]interface{})
+	for k, v := range paramMap {
+		cleanedMap[k] = v
+	}
+
+	// Clean properties
+	if props, ok := cleanedMap["properties"].(map[string]interface{}); ok && props != nil {
+		cleanedProps := make(map[string]interface{})
+		for propName, propValue := range props {
+			propMap, ok := propValue.(map[string]interface{})
+			if !ok {
+				cleanedProps[propName] = propValue // Keep non-map properties
+				continue
+			}
+
+			// Create a copy of the property map
+			cleanedPropMap := make(map[string]interface{})
+			for k, v := range propMap {
+				cleanedPropMap[k] = v
+			}
+
+			// Remove unsupported fields
+			delete(cleanedPropMap, "default")
+			delete(cleanedPropMap, "exclusiveMaximum")
+			delete(cleanedPropMap, "exclusiveMinimum")
+
+			// Check and clean 'format' for string types
+			if propType, typeExists := cleanedPropMap["type"].(string); typeExists && propType == "string" {
+				if formatValue, formatExists := cleanedPropMap["format"].(string); formatExists {
+					if formatValue != "enum" && formatValue != "date-time" {
+						delete(cleanedPropMap, "format")
+					}
+				}
+			}
+
+			// Recursively clean nested properties within this property if it's an object/array
+			// Check the type before recursing
+			if propType, typeExists := cleanedPropMap["type"].(string); typeExists && (propType == "object" || propType == "array") {
+				cleanedProps[propName] = cleanFunctionParameters(cleanedPropMap)
+			} else {
+				cleanedProps[propName] = cleanedPropMap // Assign the cleaned map back if not recursing
+			}
+
+		}
+		cleanedMap["properties"] = cleanedProps
+	}
+
+	// Recursively clean items in arrays if needed (e.g., type: array, items: { ... })
+	if items, ok := cleanedMap["items"].(map[string]interface{}); ok && items != nil {
+		cleanedMap["items"] = cleanFunctionParameters(items)
+	}
+	// Also handle items if it's an array of schemas
+	if itemsArray, ok := cleanedMap["items"].([]interface{}); ok {
+		cleanedItemsArray := make([]interface{}, len(itemsArray))
+		for i, item := range itemsArray {
+			cleanedItemsArray[i] = cleanFunctionParameters(item)
+		}
+		cleanedMap["items"] = cleanedItemsArray
+	}
+
+
+	// Recursively clean other schema composition keywords if necessary
+	for _, field := range []string{"allOf", "anyOf", "oneOf"} {
+		if nested, ok := cleanedMap[field].([]interface{}); ok {
+			cleanedNested := make([]interface{}, len(nested))
+			for i, item := range nested {
+				cleanedNested[i] = cleanFunctionParameters(item)
+			}
+			cleanedMap[field] = cleanedNested
+		}
+	}
+
+	return cleanedMap
+}
+
+
 func removeAdditionalPropertiesWithDepth(schema interface{}, depth int) interface{} {
 	if depth >= 5 {
 		return schema

relay/channel/xai/adaptor.go

@@ -0,0 +1,105 @@
+package xai
+
+import (
+	"errors"
+	"fmt"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/dto"
+	"one-api/relay/channel"
+	relaycommon "one-api/relay/common"
+	"strings"
+)
+
+type Adaptor struct {
+}
+
+func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
+	//TODO implement me
+	//panic("implement me")
+	return nil, errors.New("not available")
+}
+
+func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
+	//not available
+	return nil, errors.New("not available")
+}
+
+func (a *Adaptor) ConvertImageRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.ImageRequest) (any, error) {
+	request.Size = ""
+	return request, nil
+}
+
+func (a *Adaptor) Init(info *relaycommon.RelayInfo) {
+}
+
+func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
+	return fmt.Sprintf("%s/v1/chat/completions", info.BaseUrl), nil
+}
+
+func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *relaycommon.RelayInfo) error {
+	channel.SetupApiRequestHeader(info, c, req)
+	req.Set("Authorization", "Bearer "+info.ApiKey)
+	return nil
+}
+
+func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
+	if request == nil {
+		return nil, errors.New("request is nil")
+	}
+	request.StreamOptions = nil
+	if strings.HasPrefix(request.Model, "grok-3-mini") {
+		if request.MaxCompletionTokens == 0 && request.MaxTokens != 0 {
+			request.MaxCompletionTokens = request.MaxTokens
+			request.MaxTokens = 0
+		}
+		if strings.HasSuffix(request.Model, "-high") {
+			request.ReasoningEffort = "high"
+			request.Model = strings.TrimSuffix(request.Model, "-high")
+		} else if strings.HasSuffix(request.Model, "-low") {
+			request.ReasoningEffort = "low"
+			request.Model = strings.TrimSuffix(request.Model, "-low")
+		} else if strings.HasSuffix(request.Model, "-medium") {
+			request.ReasoningEffort = "medium"
+			request.Model = strings.TrimSuffix(request.Model, "-medium")
+		}
+		info.ReasoningEffort = request.ReasoningEffort
+		info.UpstreamModelName = request.Model
+	}
+	return request, nil
+}
+
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, nil
+}
+
+func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.EmbeddingRequest) (any, error) {
+	//not available
+	return nil, errors.New("not available")
+}
+
+func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (any, error) {
+	return channel.DoApiRequest(a, c, info, requestBody)
+}
+
+func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage any, err *dto.OpenAIErrorWithStatusCode) {
+	if info.IsStream {
+		err, usage = xAIStreamHandler(c, resp, info)
+	} else {
+		err, usage = xAIHandler(c, resp, info)
+	}
+	//if _, ok := usage.(*dto.Usage); ok && usage != nil {
+	//	usage.(*dto.Usage).CompletionTokens = usage.(*dto.Usage).TotalTokens - usage.(*dto.Usage).PromptTokens
+	//}
+
+	return
+}
+
+func (a *Adaptor) GetModelList() []string {
+	return ModelList
+}
+
+func (a *Adaptor) GetChannelName() string {
+	return ChannelName
+}

relay/channel/xai/constants.go

@@ -0,0 +1,18 @@
+package xai
+
+var ModelList = []string{
+	// grok-3
+	"grok-3-beta", "grok-3-mini-beta",
+	// grok-3 mini
+	"grok-3-fast-beta", "grok-3-mini-fast-beta",
+	// extend grok-3-mini reasoning
+	"grok-3-mini-beta-high", "grok-3-mini-beta-low", "grok-3-mini-beta-medium",
+	"grok-3-mini-fast-beta-high", "grok-3-mini-fast-beta-low", "grok-3-mini-fast-beta-medium",
+	// image model
+	"grok-2-image",
+	// legacy models
+	"grok-2", "grok-2-vision",
+	"grok-beta", "grok-vision-beta",
+}
+
+var ChannelName = "xai"

relay/channel/xai/dto.go

@@ -0,0 +1,14 @@
+package xai
+
+import "one-api/dto"
+
+// ChatCompletionResponse represents the response from XAI chat completion API
+type ChatCompletionResponse struct {
+	Id                string `json:"id"`
+	Object            string `json:"object"`
+	Created           int64  `json:"created"`
+	Model             string `json:"model"`
+	Choices           []dto.ChatCompletionsStreamResponseChoice
+	Usage             *dto.Usage `json:"usage"`
+	SystemFingerprint string     `json:"system_fingerprint"`
+}

relay/channel/xai/text.go

@@ -0,0 +1,107 @@
+package xai
+
+import (
+	"bytes"
+	"encoding/json"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
+	"one-api/dto"
+	relaycommon "one-api/relay/common"
+	"one-api/relay/helper"
+	"one-api/service"
+)
+
+func streamResponseXAI2OpenAI(xAIResp *dto.ChatCompletionsStreamResponse, usage *dto.Usage) *dto.ChatCompletionsStreamResponse {
+	if xAIResp == nil {
+		return nil
+	}
+	if xAIResp.Usage != nil {
+		xAIResp.Usage.CompletionTokens = usage.CompletionTokens
+	}
+	openAIResp := &dto.ChatCompletionsStreamResponse{
+		Id:      xAIResp.Id,
+		Object:  xAIResp.Object,
+		Created: xAIResp.Created,
+		Model:   xAIResp.Model,
+		Choices: xAIResp.Choices,
+		Usage:   xAIResp.Usage,
+	}
+
+	return openAIResp
+}
+
+func xAIStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
+	usage := &dto.Usage{}
+
+	helper.SetEventStreamHeaders(c)
+
+	helper.StreamScannerHandler(c, resp, info, func(data string) bool {
+		var xAIResp *dto.ChatCompletionsStreamResponse
+		err := json.Unmarshal([]byte(data), &xAIResp)
+		if err != nil {
+			common.SysError("error unmarshalling stream response: " + err.Error())
+			return true
+		}
+
+		// 把 xAI 的usage转换为 OpenAI 的usage
+		if xAIResp.Usage != nil {
+			usage.PromptTokens = xAIResp.Usage.PromptTokens
+			usage.TotalTokens = xAIResp.Usage.TotalTokens
+			usage.CompletionTokens = usage.TotalTokens - usage.PromptTokens
+		}
+
+		openaiResponse := streamResponseXAI2OpenAI(xAIResp, usage)
+		err = helper.ObjectData(c, openaiResponse)
+		if err != nil {
+			common.SysError(err.Error())
+		}
+		return true
+	})
+
+	helper.Done(c)
+	err := resp.Body.Close()
+	if err != nil {
+		//return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+		common.SysError("close_response_body_failed: " + err.Error())
+	}
+	return nil, usage
+}
+
+func xAIHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
+	responseBody, err := io.ReadAll(resp.Body)
+	var response *dto.TextResponse
+	err = common.DecodeJson(responseBody, &response)
+	if err != nil {
+		common.SysError("error unmarshalling stream response: " + err.Error())
+		return nil, nil
+	}
+	response.Usage.CompletionTokens = response.Usage.TotalTokens - response.Usage.PromptTokens
+	response.Usage.CompletionTokenDetails.TextTokens = response.Usage.CompletionTokens - response.Usage.CompletionTokenDetails.ReasoningTokens
+
+	// new body
+	encodeJson, err := common.EncodeJson(response)
+	if err != nil {
+		common.SysError("error marshalling stream response: " + err.Error())
+		return nil, nil
+	}
+
+	// set new body
+	resp.Body = io.NopCloser(bytes.NewBuffer(encodeJson))
+
+	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 service.OpenAIErrorWrapper(err, "copy_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+
+	return nil, &response.Usage
+}

relay/channel/zhipu_4v/adaptor.go

@@ -10,6 +10,7 @@ import (
 	"one-api/relay/channel"
 	"one-api/relay/channel/openai"
 	relaycommon "one-api/relay/common"
+	relayconstant "one-api/relay/constant"
 )
 
 type Adaptor struct {
@@ -35,7 +36,13 @@ func (a *Adaptor) Init(info *relaycommon.RelayInfo) {
 }
 
 func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
-	return fmt.Sprintf("%s/api/paas/v4/chat/completions", info.BaseUrl), nil
+	baseUrl := fmt.Sprintf("%s/api/paas/v4", info.BaseUrl)
+	switch info.RelayMode {
+	case relayconstant.RelayModeEmbeddings:
+		return fmt.Sprintf("%s/embeddings", baseUrl), nil
+	default:
+		return fmt.Sprintf("%s/chat/completions", baseUrl), nil
+	}
 }
 
 func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *relaycommon.RelayInfo) error {
@@ -60,8 +67,7 @@ func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dt
 }
 
 func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.EmbeddingRequest) (any, error) {
-	//TODO implement me
-	return nil, errors.New("not implemented")
+	return request, nil
 }
 
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (any, error) {

relay/channel/zhipu_4v/relay-zhipu_v4.go

@@ -1,17 +1,9 @@
 package zhipu_4v
 
 import (
-	"bufio"
-	"bytes"
-	"encoding/json"
-	"github.com/gin-gonic/gin"
 	"github.com/golang-jwt/jwt"
-	"io"
-	"net/http"
 	"one-api/common"
 	"one-api/dto"
-	"one-api/relay/helper"
-	"one-api/service"
 	"strings"
 	"sync"
 	"time"
@@ -119,163 +111,3 @@ func requestOpenAI2Zhipu(request dto.GeneralOpenAIRequest) *dto.GeneralOpenAIReq
 		ToolChoice:  request.ToolChoice,
 	}
 }
-
-//func responseZhipu2OpenAI(response *dto.OpenAITextResponse) *dto.OpenAITextResponse {
-//	fullTextResponse := dto.OpenAITextResponse{
-//		Id:      response.Id,
-//		Object:  "chat.completion",
-//		Created: common.GetTimestamp(),
-//		Choices: make([]dto.OpenAITextResponseChoice, 0, len(response.TextResponseChoices)),
-//		Usage:   response.Usage,
-//	}
-//	for i, choice := range response.TextResponseChoices {
-//		content, _ := json.Marshal(strings.Trim(choice.Content, "\""))
-//		openaiChoice := dto.OpenAITextResponseChoice{
-//			Index: i,
-//			Message: dto.Message{
-//				Role:    choice.Role,
-//				Content: content,
-//			},
-//			FinishReason: "",
-//		}
-//		if i == len(response.TextResponseChoices)-1 {
-//			openaiChoice.FinishReason = "stop"
-//		}
-//		fullTextResponse.Choices = append(fullTextResponse.Choices, openaiChoice)
-//	}
-//	return &fullTextResponse
-//}
-
-func streamResponseZhipu2OpenAI(zhipuResponse *ZhipuV4StreamResponse) *dto.ChatCompletionsStreamResponse {
-	var choice dto.ChatCompletionsStreamResponseChoice
-	choice.Delta.Content = zhipuResponse.Choices[0].Delta.Content
-	choice.Delta.Role = zhipuResponse.Choices[0].Delta.Role
-	choice.Delta.ToolCalls = zhipuResponse.Choices[0].Delta.ToolCalls
-	choice.Index = zhipuResponse.Choices[0].Index
-	choice.FinishReason = zhipuResponse.Choices[0].FinishReason
-	response := dto.ChatCompletionsStreamResponse{
-		Id:      zhipuResponse.Id,
-		Object:  "chat.completion.chunk",
-		Created: zhipuResponse.Created,
-		Model:   "glm-4v",
-		Choices: []dto.ChatCompletionsStreamResponseChoice{choice},
-	}
-	return &response
-}
-
-func lastStreamResponseZhipuV42OpenAI(zhipuResponse *ZhipuV4StreamResponse) (*dto.ChatCompletionsStreamResponse, *dto.Usage) {
-	response := streamResponseZhipu2OpenAI(zhipuResponse)
-	return response, &zhipuResponse.Usage
-}
-
-func zhipuStreamHandler(c *gin.Context, resp *http.Response) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
-	var usage *dto.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
-			}
-			if data[:6] != "data: " && data[:6] != "[DONE]" {
-				continue
-			}
-			dataChan <- data
-		}
-		stopChan <- true
-	}()
-	helper.SetEventStreamHeaders(c)
-	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")
-
-			var streamResponse ZhipuV4StreamResponse
-			err := json.Unmarshal([]byte(data), &streamResponse)
-			if err != nil {
-				common.SysError("error unmarshalling stream response: " + err.Error())
-			}
-			var response *dto.ChatCompletionsStreamResponse
-			if strings.Contains(data, "prompt_tokens") {
-				response, usage = lastStreamResponseZhipuV42OpenAI(&streamResponse)
-			} else {
-				response = streamResponseZhipu2OpenAI(&streamResponse)
-			}
-			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:
-			return false
-		}
-	})
-	err := resp.Body.Close()
-	if err != nil {
-		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
-	}
-	return nil, usage
-}
-
-func zhipuHandler(c *gin.Context, resp *http.Response) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
-	var textResponse ZhipuV4Response
-	responseBody, err := io.ReadAll(resp.Body)
-	if err != nil {
-		return service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
-	}
-	err = resp.Body.Close()
-	if err != nil {
-		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
-	}
-	err = json.Unmarshal(responseBody, &textResponse)
-	if err != nil {
-		return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
-	}
-	if textResponse.Error.Type != "" {
-		return &dto.OpenAIErrorWithStatusCode{
-			Error:      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 HTTPClient 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 service.OpenAIErrorWrapper(err, "copy_response_body_failed", http.StatusInternalServerError), nil
-	}
-	err = resp.Body.Close()
-	if err != nil {
-		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
-	}
-
-	return nil, &textResponse.Usage
-}

relay/constant/api_type.go

@@ -32,6 +32,7 @@ const (
 	APITypeBaiduV2
 	APITypeOpenRouter
 	APITypeXinference
+	APITypeXai
 	APITypeDummy // this one is only for count, do not add any channel after this
 )
 
@@ -92,6 +93,8 @@ func ChannelType2APIType(channelType int) (int, bool) {
 		apiType = APITypeOpenRouter
 	case common.ChannelTypeXinference:
 		apiType = APITypeXinference
+	case common.ChannelTypeXai:
+		apiType = APITypeXai
 	}
 	if apiType == -1 {
 		return APITypeOpenAI, false

relay/helper/common.go

@@ -56,6 +56,9 @@ func StringData(c *gin.Context, str string) error {
 }
 
 func ObjectData(c *gin.Context, object interface{}) error {
+	if object == nil {
+		return errors.New("object is nil")
+	}
 	jsonData, err := json.Marshal(object)
 	if err != nil {
 		return fmt.Errorf("error marshalling object: %w", err)

relay/helper/stream_scanner.go

@@ -14,6 +14,11 @@ import (
 	"github.com/gin-gonic/gin"
 )
 
+const (
+	InitialScannerBufferSize = 1 << 20  // 1MB (1*1024*1024)
+	MaxScannerBufferSize     = 10 << 20 // 10MB (10*1024*1024)
+)
+
 func StreamScannerHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo, dataHandler func(data string) bool) {
 
 	if resp == nil {
@@ -38,7 +43,7 @@ func StreamScannerHandler(c *gin.Context, resp *http.Response, info *relaycommon
 		ticker.Stop()
 		close(stopChan)
 	}()
-
+	scanner.Buffer(make([]byte, InitialScannerBufferSize), MaxScannerBufferSize)
 	scanner.Split(bufio.ScanLines)
 	SetEventStreamHeaders(c)
 

relay/relay_adaptor.go

@@ -25,6 +25,7 @@ import (
 	"one-api/relay/channel/tencent"
 	"one-api/relay/channel/vertex"
 	"one-api/relay/channel/volcengine"
+	"one-api/relay/channel/xai"
 	"one-api/relay/channel/xunfei"
 	"one-api/relay/channel/zhipu"
 	"one-api/relay/channel/zhipu_4v"
@@ -85,6 +86,8 @@ func GetAdaptor(apiType int) channel.Adaptor {
 		return &openai.Adaptor{}
 	case constant.APITypeXinference:
 		return &openai.Adaptor{}
+	case constant.APITypeXai:
+		return &xai.Adaptor{}
 	}
 	return nil
 }

setting/group_ratio.go

@@ -4,6 +4,7 @@ import (
 	"encoding/json"
 	"errors"
 	"one-api/common"
+	"sync"
 )
 
 var groupRatio = map[string]float64{
@@ -11,8 +12,12 @@ var groupRatio = map[string]float64{
 	"vip":     1,
 	"svip":    1,
 }
+var groupRatioMutex sync.RWMutex
 
 func GetGroupRatioCopy() map[string]float64 {
+	groupRatioMutex.RLock()
+	defer groupRatioMutex.RUnlock()
+	
 	groupRatioCopy := make(map[string]float64)
 	for k, v := range groupRatio {
 		groupRatioCopy[k] = v
@@ -21,11 +26,17 @@ func GetGroupRatioCopy() map[string]float64 {
 }
 
 func ContainsGroupRatio(name string) bool {
+	groupRatioMutex.RLock()
+	defer groupRatioMutex.RUnlock()
+	
 	_, ok := groupRatio[name]
 	return ok
 }
 
 func GroupRatio2JSONString() string {
+	groupRatioMutex.RLock()
+	defer groupRatioMutex.RUnlock()
+	
 	jsonBytes, err := json.Marshal(groupRatio)
 	if err != nil {
 		common.SysError("error marshalling model ratio: " + err.Error())
@@ -34,11 +45,17 @@ func GroupRatio2JSONString() string {
 }
 
 func UpdateGroupRatioByJSONString(jsonStr string) error {
+	groupRatioMutex.Lock()
+	defer groupRatioMutex.Unlock()
+	
 	groupRatio = make(map[string]float64)
 	return json.Unmarshal([]byte(jsonStr), &groupRatio)
 }
 
 func GetGroupRatio(name string) float64 {
+	groupRatioMutex.RLock()
+	defer groupRatioMutex.RUnlock()
+	
 	ratio, ok := groupRatio[name]
 	if !ok {
 		common.SysError("group ratio not found: " + name)

setting/operation_setting/cache_ratio.go

@@ -56,17 +56,15 @@ var cacheRatioMapMutex sync.RWMutex
 
 // GetCacheRatioMap returns the cache ratio map
 func GetCacheRatioMap() map[string]float64 {
-	cacheRatioMapMutex.Lock()
-	defer cacheRatioMapMutex.Unlock()
-	if cacheRatioMap == nil {
-		cacheRatioMap = defaultCacheRatio
-	}
+	cacheRatioMapMutex.RLock()
+	defer cacheRatioMapMutex.RUnlock()
 	return cacheRatioMap
 }
 
 // CacheRatio2JSONString converts the cache ratio map to a JSON string
 func CacheRatio2JSONString() string {
-	GetCacheRatioMap()
+	cacheRatioMapMutex.RLock()
+	defer cacheRatioMapMutex.RUnlock()
 	jsonBytes, err := json.Marshal(cacheRatioMap)
 	if err != nil {
 		common.SysError("error marshalling cache ratio: " + err.Error())
@@ -84,10 +82,11 @@ func UpdateCacheRatioByJSONString(jsonStr string) error {
 
 // GetCacheRatio returns the cache ratio for a model
 func GetCacheRatio(name string) (float64, bool) {
-	GetCacheRatioMap()
+	cacheRatioMapMutex.RLock()
+	defer cacheRatioMapMutex.RUnlock()
 	ratio, ok := cacheRatioMap[name]
 	if !ok {
-		return 1, false // Default to 0.5 if not found
+		return 1, false // Default to 1 if not found
 	}
 	return ratio, true
 }

setting/operation_setting/model-ratio.go

@@ -131,17 +131,12 @@ var defaultModelRatio = map[string]float64{
 	"bge-large-en":                        0.002 * RMB,
 	"tao-8k":                              0.002 * RMB,
 	"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
-	"gemini-1.0-pro-vision-001":           1,
-	"gemini-1.0-pro-001":                  1,
-	"gemini-1.5-pro-latest":               1.75, // $3.5 / 1M tokens
-	"gemini-1.5-pro-exp-0827":             1.75, // $3.5 / 1M tokens
-	"gemini-1.5-flash-latest":             1,
-	"gemini-1.5-flash-exp-0827":           1,
-	"gemini-1.0-pro-latest":               1,
-	"gemini-1.0-pro-vision-latest":        1,
-	"gemini-ultra":                        1,
+	"gemini-1.5-pro-latest":               1.25, // $3.5 / 1M tokens
+	"gemini-1.5-flash-latest":             0.075,
+	"gemini-2.0-flash":                    0.05,
+	"gemini-2.5-pro-exp-03-25":            0.625,
+	"gemini-2.5-pro-preview-03-25":        0.625,
+	"text-embedding-004":                  0.001,
 	"chatglm_turbo":                       0.3572,     // ￥0.005 / 1k tokens
 	"chatglm_pro":                         0.7143,     // ￥0.01 / 1k tokens
 	"chatglm_std":                         0.3572,     // ￥0.005 / 1k tokens
@@ -204,29 +199,39 @@ var defaultModelRatio = map[string]float64{
 	"llama-3-sonar-small-32k-online": 0.2 / 1000 * USD,
 	"llama-3-sonar-large-32k-chat":   1 / 1000 * USD,
 	"llama-3-sonar-large-32k-online": 1 / 1000 * USD,
+	// grok
+	"grok-3-beta":           1.5,
+	"grok-3-mini-beta":      0.15,
+	"grok-2":                1,
+	"grok-2-vision":         1,
+	"grok-beta":             2.5,
+	"grok-vision-beta":      2.5,
+	"grok-3-fast-beta":      2.5,
+	"grok-3-mini-fast-beta": 0.3,
 }
 
 var defaultModelPrice = map[string]float64{
-	"suno_music":        0.1,
-	"suno_lyrics":       0.01,
-	"dall-e-3":          0.04,
-	"gpt-4-gizmo-*":     0.1,
-	"mj_imagine":        0.1,
-	"mj_variation":      0.1,
-	"mj_reroll":         0.1,
-	"mj_blend":          0.1,
-	"mj_modal":          0.1,
-	"mj_zoom":           0.1,
-	"mj_shorten":        0.1,
-	"mj_high_variation": 0.1,
-	"mj_low_variation":  0.1,
-	"mj_pan":            0.1,
-	"mj_inpaint":        0,
-	"mj_custom_zoom":    0,
-	"mj_describe":       0.05,
-	"mj_upscale":        0.05,
-	"swap_face":         0.05,
-	"mj_upload":         0.05,
+	"suno_music":              0.1,
+	"suno_lyrics":             0.01,
+	"dall-e-3":                0.04,
+	"imagen-3.0-generate-002": 0.03,
+	"gpt-4-gizmo-*":           0.1,
+	"mj_imagine":              0.1,
+	"mj_variation":            0.1,
+	"mj_reroll":               0.1,
+	"mj_blend":                0.1,
+	"mj_modal":                0.1,
+	"mj_zoom":                 0.1,
+	"mj_shorten":              0.1,
+	"mj_high_variation":       0.1,
+	"mj_low_variation":        0.1,
+	"mj_pan":                  0.1,
+	"mj_inpaint":              0,
+	"mj_custom_zoom":          0,
+	"mj_describe":             0.05,
+	"mj_upscale":              0.05,
+	"swap_face":               0.05,
+	"mj_upload":               0.05,
 }
 
 var (
@@ -249,17 +254,41 @@ var defaultCompletionRatio = map[string]float64{
 	"gpt-4-all":      2,
 }
 
-func GetModelPriceMap() map[string]float64 {
+// InitModelSettings initializes all model related settings maps
+func InitModelSettings() {
+	// Initialize modelPriceMap
 	modelPriceMapMutex.Lock()
-	defer modelPriceMapMutex.Unlock()
-	if modelPriceMap == nil {
-		modelPriceMap = defaultModelPrice
-	}
+	modelPriceMap = defaultModelPrice
+	modelPriceMapMutex.Unlock()
+
+	// Initialize modelRatioMap
+	modelRatioMapMutex.Lock()
+	modelRatioMap = defaultModelRatio
+	modelRatioMapMutex.Unlock()
+
+	// Initialize CompletionRatio
+	CompletionRatioMutex.Lock()
+	CompletionRatio = defaultCompletionRatio
+	CompletionRatioMutex.Unlock()
+
+	// Initialize cacheRatioMap
+	cacheRatioMapMutex.Lock()
+	cacheRatioMap = defaultCacheRatio
+	cacheRatioMapMutex.Unlock()
+
+	common.SysLog("model settings initialized")
+}
+
+func GetModelPriceMap() map[string]float64 {
+	modelPriceMapMutex.RLock()
+	defer modelPriceMapMutex.RUnlock()
 	return modelPriceMap
 }
 
 func ModelPrice2JSONString() string {
-	GetModelPriceMap()
+	modelPriceMapMutex.RLock()
+	defer modelPriceMapMutex.RUnlock()
+
 	jsonBytes, err := json.Marshal(modelPriceMap)
 	if err != nil {
 		common.SysError("error marshalling model price: " + err.Error())
@@ -276,7 +305,9 @@ func UpdateModelPriceByJSONString(jsonStr string) error {
 
 // GetModelPrice 返回模型的价格，如果模型不存在则返回-1，false
 func GetModelPrice(name string, printErr bool) (float64, bool) {
-	GetModelPriceMap()
+	modelPriceMapMutex.RLock()
+	defer modelPriceMapMutex.RUnlock()
+
 	if strings.HasPrefix(name, "gpt-4-gizmo") {
 		name = "gpt-4-gizmo-*"
 	}
@@ -293,24 +324,6 @@ func GetModelPrice(name string, printErr bool) (float64, bool) {
 	return price, true
 }
 
-func GetModelRatioMap() map[string]float64 {
-	modelRatioMapMutex.Lock()
-	defer modelRatioMapMutex.Unlock()
-	if modelRatioMap == nil {
-		modelRatioMap = defaultModelRatio
-	}
-	return modelRatioMap
-}
-
-func ModelRatio2JSONString() string {
-	GetModelRatioMap()
-	jsonBytes, err := json.Marshal(modelRatioMap)
-	if err != nil {
-		common.SysError("error marshalling model ratio: " + err.Error())
-	}
-	return string(jsonBytes)
-}
-
 func UpdateModelRatioByJSONString(jsonStr string) error {
 	modelRatioMapMutex.Lock()
 	defer modelRatioMapMutex.Unlock()
@@ -319,7 +332,9 @@ func UpdateModelRatioByJSONString(jsonStr string) error {
 }
 
 func GetModelRatio(name string) (float64, bool) {
-	GetModelRatioMap()
+	modelRatioMapMutex.RLock()
+	defer modelRatioMapMutex.RUnlock()
+
 	if strings.HasPrefix(name, "gpt-4-gizmo") {
 		name = "gpt-4-gizmo-*"
 	}
@@ -343,16 +358,15 @@ func GetDefaultModelRatioMap() map[string]float64 {
 }
 
 func GetCompletionRatioMap() map[string]float64 {
-	CompletionRatioMutex.Lock()
-	defer CompletionRatioMutex.Unlock()
-	if CompletionRatio == nil {
-		CompletionRatio = defaultCompletionRatio
-	}
+	CompletionRatioMutex.RLock()
+	defer CompletionRatioMutex.RUnlock()
 	return CompletionRatio
 }
 
 func CompletionRatio2JSONString() string {
-	GetCompletionRatioMap()
+	CompletionRatioMutex.RLock()
+	defer CompletionRatioMutex.RUnlock()
+
 	jsonBytes, err := json.Marshal(CompletionRatio)
 	if err != nil {
 		common.SysError("error marshalling completion ratio: " + err.Error())
@@ -368,7 +382,8 @@ func UpdateCompletionRatioByJSONString(jsonStr string) error {
 }
 
 func GetCompletionRatio(name string) float64 {
-	GetCompletionRatioMap()
+	CompletionRatioMutex.RLock()
+	defer CompletionRatioMutex.RUnlock()
 
 	if strings.Contains(name, "/") {
 		if ratio, ok := CompletionRatio[name]; ok {
@@ -438,7 +453,14 @@ func getHardcodedCompletionModelRatio(name string) (float64, bool) {
 		return 3, true
 	}
 	if strings.HasPrefix(name, "gemini-") {
-		return 4, true
+		if strings.HasPrefix(name, "gemini-1.5") {
+			return 4, true
+		} else if strings.HasPrefix(name, "gemini-2.0") {
+			return 4, true
+		} else if strings.HasPrefix(name, "gemini-2.5-pro-preview") {
+			return 8, true
+		}
+		return 4, false
 	}
 	if strings.HasPrefix(name, "command") {
 		switch name {
@@ -451,7 +473,7 @@ func getHardcodedCompletionModelRatio(name string) (float64, bool) {
 		case "command-r-plus-08-2024":
 			return 4, true
 		default:
-			return 4, true
+			return 4, false
 		}
 	}
 	// hint 只给官方上4倍率，由于开源模型供应商自行定价，不对其进行补全倍率进行强制对齐
@@ -508,3 +530,14 @@ func GetAudioCompletionRatio(name string) float64 {
 	}
 	return 2
 }
+
+func ModelRatio2JSONString() string {
+	modelRatioMapMutex.RLock()
+	defer modelRatioMapMutex.RUnlock()
+
+	jsonBytes, err := json.Marshal(modelRatioMap)
+	if err != nil {
+		common.SysError("error marshalling model ratio: " + err.Error())
+	}
+	return string(jsonBytes)
+}

web/package.json

@@ -23,7 +23,7 @@
     "react-turnstile": "^1.0.5",
     "semantic-ui-offline": "^2.5.0",
     "semantic-ui-react": "^2.1.3",
-    "sse": "github:mpetazzoni/sse.js",
+    "sse": "https://github.com/mpetazzoni/sse.js",
     "i18next": "^23.16.8",
     "react-i18next": "^13.0.0",
     "i18next-browser-languagedetector": "^7.2.0"

web/src/App.js

@@ -26,6 +26,7 @@ import Playground from './pages/Playground/Playground.js';
 import OAuth2Callback from "./components/OAuth2Callback.js";
 import PersonalSetting from './components/PersonalSetting.js';
 import Setup from './pages/Setup/index.js';
+import SetupCheck from './components/SetupCheck';
 
 const Home = lazy(() => import('./pages/Home'));
 const Detail = lazy(() => import('./pages/Detail'));
@@ -35,7 +36,7 @@ function App() {
   const location = useLocation();
   
   return (
-    <>
+    <SetupCheck>
       <Routes>
         <Route
           path='/'
@@ -286,7 +287,7 @@ function App() {
           />
           <Route path='*' element={<NotFound />} />
         </Routes>
-      </>
+      </SetupCheck>
   );
 }
 

web/src/components/SetupCheck.js

@@ -0,0 +1,18 @@
+import React, { useContext, useEffect } from 'react';
+import { Navigate, useLocation } from 'react-router-dom';
+import { StatusContext } from '../context/Status';
+
+const SetupCheck = ({ children }) => {
+  const [statusState] = useContext(StatusContext);
+  const location = useLocation();
+
+  useEffect(() => {
+    if (statusState?.status?.setup === false && location.pathname !== '/setup') {
+      window.location.href = '/setup';
+    }
+  }, [statusState?.status?.setup, location.pathname]);
+
+  return children;
+};
+
+export default SetupCheck; 

web/src/components/SystemSetting.js

@@ -619,7 +619,7 @@ const SystemSetting = () => {
                       允许通过 Telegram 进行登录
                     </Form.Checkbox>
                     <Form.Checkbox
-                      field='oidc.enabled'
+                      field="['oidc.enabled']"
                       noLabel
                       onChange={(e) => handleCheckboxChange('oidc.enabled', e)}
                     >
@@ -721,14 +721,14 @@ const SystemSetting = () => {
                 <Row gutter={{ xs: 8, sm: 16, md: 24, lg: 24, xl: 24, xxl: 24 }}>
                   <Col xs={24} sm={24} md={12} lg={12} xl={12}>
                     <Form.Input
-                      field='oidc.well_known'
+                      field="['oidc.well_known']"
                       label='Well-Known URL'
                       placeholder='请输入 OIDC 的 Well-Known URL'
                     />
                   </Col>
                   <Col xs={24} sm={24} md={12} lg={12} xl={12}>
                     <Form.Input
-                      field='oidc.client_id'
+                      field="['oidc.client_id']"
                       label='Client ID'
                       placeholder='输入 OIDC 的 Client ID'
                     />
@@ -737,7 +737,7 @@ const SystemSetting = () => {
                 <Row gutter={{ xs: 8, sm: 16, md: 24, lg: 24, xl: 24, xxl: 24 }}>
                   <Col xs={24} sm={24} md={12} lg={12} xl={12}>
                     <Form.Input
-                      field='oidc.client_secret'
+                      field="['oidc.client_secret']"
                       label='Client Secret'
                       type='password'
                       placeholder='敏感信息不会发送到前端显示'
@@ -745,7 +745,7 @@ const SystemSetting = () => {
                   </Col>
                   <Col xs={24} sm={24} md={12} lg={12} xl={12}>
                     <Form.Input
-                      field='oidc.authorization_endpoint'
+                      field="['oidc.authorization_endpoint']"
                       label='Authorization Endpoint'
                       placeholder='输入 OIDC 的 Authorization Endpoint'
                     />
@@ -754,14 +754,14 @@ const SystemSetting = () => {
                 <Row gutter={{ xs: 8, sm: 16, md: 24, lg: 24, xl: 24, xxl: 24 }}>
                   <Col xs={24} sm={24} md={12} lg={12} xl={12}>
                     <Form.Input
-                      field='oidc.token_endpoint'
+                      field="['oidc.token_endpoint']"
                       label='Token Endpoint'
                       placeholder='输入 OIDC 的 Token Endpoint'
                     />
                   </Col>
                   <Col xs={24} sm={24} md={12} lg={12} xl={12}>
                     <Form.Input
-                      field='oidc.user_info_endpoint'
+                      field="['oidc.user_info_endpoint']"
                       label='User Info Endpoint'
                       placeholder='输入 OIDC 的 Userinfo Endpoint'
                     />

web/src/constants/channel.constants.js

@@ -115,4 +115,9 @@ export const CHANNEL_OPTIONS = [
     color: 'blue',
     label: '字节火山方舟、豆包、DeepSeek通用'
   },
+  {
+    value: 48,
+    color: 'blue',
+    label: 'xAI'
+  }
 ];

web/src/i18n/locales/en.json

@@ -514,7 +514,7 @@
   "，图片演示。": "related image demo.",
   "令牌创建成功，请在列表页面点击复制获取令牌！": "Token created successfully, please click copy on the list page to get the token!",
   "代理": "Proxy",
-  "此项可选，用于通过代理站来进行 API 调用，请输入代理站地址，格式为：https://domain.com": "This is optional, used to make API calls through the proxy site, please enter the proxy site address, the format is: https://domain.com",
+  "此项可选，用于通过自定义API地址来进行 API 调用，请输入API地址，格式为：https://domain.com": "This is optional, used to make API calls through the proxy site, please enter the proxy site address, the format is: https://domain.com",
   "取消密码登录将导致所有未绑定其他登录方式的用户（包括管理员）无法通过密码登录，确认取消？": "Canceling password login will cause all users (including administrators) who have not bound other login methods to be unable to log in via password, confirm cancel?",
   "按照如下格式输入：": "Enter in the following format:",
   "模型版本": "Model version",
@@ -1111,7 +1111,7 @@
   "如果你对接的是上游One API或者New API等转发项目，请使用OpenAI类型，不要使用此类型，除非你知道你在做什么。": "If you are connecting to upstream One API or New API forwarding projects, please use OpenAI type. Do not use this type unless you know what you are doing.",
   "完整的 Base URL，支持变量{model}": "Complete Base URL, supports variable {model}",
   "请输入完整的URL，例如：https://api.openai.com/v1/chat/completions": "Please enter complete URL, e.g.: https://api.openai.com/v1/chat/completions",
-  "此项可选，用于通过代理站来进行 API 调用，末尾不要带/v1和/": "Optional for API calls through proxy sites, do not end with /v1 and /",
+  "此项可选，用于通过自定义API地址来进行 API 调用，末尾不要带/v1和/": "Optional for API calls through custom API address, do not add /v1 and / at the end",
   "私有部署地址": "Private Deployment Address",
   "请输入私有部署地址，格式为：https://fastgpt.run/api/openapi": "Please enter private deployment address, format: https://fastgpt.run/api/openapi",
   "注意非Chat API，请务必填写正确的API地址，否则可能导致无法使用": "Note: For non-Chat API, please make sure to enter the correct API address, otherwise it may not work",
@@ -1272,7 +1272,7 @@
   "通知邮箱": "Notification email",
   "设置用于接收额度预警的邮箱地址，不填则使用账号绑定的邮箱": "Set the email address for receiving quota warning notifications, if not set, the email address bound to the account will be used",
   "留空则使用账号绑定的邮箱": "If left blank, the email address bound to the account will be used",
-  "代理站地址": "Base URL",
+  "API地址": "Base URL",
   "对于官方渠道，new-api已经内置地址，除非是第三方代理站点或者Azure的特殊接入地址，否则不需要填写": "For official channels, the new-api has a built-in address. Unless it is a third-party proxy site or a special Azure access address, there is no need to fill it in",
   "渠道额外设置": "Channel extra settings",
   "参数覆盖": "Parameters override",
@@ -1346,5 +1346,26 @@
   "提示缓存倍率": "Prompt cache ratio",
   "缓存：${{price}} * {{ratio}} = ${{total}} / 1M tokens (缓存倍率: {{cacheRatio}})": "Cache: ${{price}} * {{ratio}} = ${{total}} / 1M tokens (cache ratio: {{cacheRatio}})",
   "提示 {{nonCacheInput}} tokens + 缓存 {{cacheInput}} tokens * {{cacheRatio}} / 1M tokens * ${{price}} + 补全 {{completion}} tokens / 1M tokens * ${{compPrice}} * 分组 {{ratio}} = ${{total}}": "Prompt {{nonCacheInput}} tokens + cache {{cacheInput}} tokens * {{cacheRatio}} / 1M tokens * ${{price}} + completion {{completion}} tokens / 1M tokens * ${{compPrice}} * group {{ratio}} = ${{total}}",
-  "缓存 Tokens": "Cache Tokens"
+  "缓存 Tokens": "Cache Tokens",
+  "系统初始化": "System initialization",
+  "管理员账号已经初始化过，请继续设置系统参数": "The admin account has already been initialized, please continue to set the system parameters",
+  "管理员账号": "Admin account",
+  "请输入管理员用户名": "Please enter the admin username",
+  "请输入管理员密码": "Please enter the admin password",
+  "请确认管理员密码": "Please confirm the admin password",
+  "请选择使用模式": "Please select the usage mode",
+  "数据库警告": "Database warning",
+  "您正在使用 SQLite 数据库。如果您在容器环境中运行，请确保已正确设置数据库文件的持久化映射，否则容器重启后所有数据将丢失！": "You are using the SQLite database. If you are running in a container environment, please ensure that the database file persistence mapping is correctly set, otherwise all data will be lost after container restart!",
+  "建议在生产环境中使用 MySQL 或 PostgreSQL 数据库，或确保 SQLite 数据库文件已映射到宿主机的持久化存储。": "It is recommended to use MySQL or PostgreSQL databases in production environments, or ensure that the SQLite database file is mapped to the persistent storage of the host machine.",
+  "使用模式": "Usage mode",
+  "对外运营模式": "Default mode",
+  "密码长度至少为8个字符": "Password must be at least 8 characters long",
+  "表单引用错误，请刷新页面重试": "Form reference error, please refresh the page and try again",
+  "默认模式，适用于为多个用户提供服务的场景。": "Default mode, suitable for scenarios where multiple users are provided.",
+  "此模式下，系统将计算每次调用的用量，您需要对每个模型都设置价格，如果没有设置价格，用户将无法使用该模型。": "In this mode, the system will calculate the usage of each call, you need to set the price for each model, if the price is not set, the user will not be able to use the model.",
+  "适用于个人使用的场景。": "Suitable for personal use.",
+  "不需要设置模型价格，系统将弱化用量计算，您可专注于使用模型。": "No need to set the model price, the system will weaken the usage calculation, you can focus on using the model.",
+  "适用于展示系统功能的场景。": "Suitable for scenarios where the system functions are displayed.",
+  "可在初始化后修改": "Can be modified after initialization",
+  "初始化系统": "Initialize system"
 }

web/src/pages/Channel/EditChannel.js

@@ -556,13 +556,13 @@ const EditChannel = (props) => {
           {inputs.type !== 3 && inputs.type !== 8 && inputs.type !== 22 && inputs.type !== 36 && inputs.type !== 45 && (
             <>
               <div style={{ marginTop: 10 }}>
-                <Typography.Text strong>{t('代理站地址')}：</Typography.Text>
+                <Typography.Text strong>{t('API地址')}：</Typography.Text>
               </div>
               <Tooltip content={t('对于官方渠道，new-api已经内置地址，除非是第三方代理站点或者Azure的特殊接入地址，否则不需要填写')}>
                 <Input
-                  label={t('代理站地址')}
+                  label={t('API地址')}
                   name="base_url"
-                  placeholder={t('此项可选，用于通过代理站来进行 API 调用，末尾不要带/v1和/')}
+                  placeholder={t('此项可选，用于通过自定义API地址来进行 API 调用，末尾不要带/v1和/')}
                   onChange={(value) => {
                     handleInputChange('base_url', value);
                   }}

web/src/pages/Home/index.js

@@ -66,13 +66,9 @@ const Home = () => {
   };
 
   useEffect(() => {
-    if (statusState.status?.setup === false) {
-      window.location.href = '/setup';
-      return;
-    }
     displayNotice().then();
     displayHomePageContent().then();
-  });
+  }, []);
 
   return (
     <>
@@ -116,6 +112,7 @@ const Home = () => {
                       https://github.com/Calcium-Ion/new-api
                     </a>
                   </p>
+                  
                   <p>
                     {t('协议')}：
                     <a

web/src/pages/Setting/Operation/ModelSettingsVisualEditor.js

@@ -1,10 +1,12 @@
 // ModelSettingsVisualEditor.js
-import React, { useEffect, useState } from 'react';
-import { Table, Button, Input, Modal, Form, Space } from '@douyinfe/semi-ui';
-import { IconDelete, IconPlus, IconSearch, IconSave } from '@douyinfe/semi-icons';
+import React, { useContext, useEffect, useState, useRef } from 'react';
+import { Table, Button, Input, Modal, Form, Space, RadioGroup, Radio, Tabs, TabPane } from '@douyinfe/semi-ui';
+import { IconDelete, IconPlus, IconSearch, IconSave, IconEdit } from '@douyinfe/semi-icons';
 import { showError, showSuccess } from '../../../helpers';
 import { API } from '../../../helpers';
 import { useTranslation } from 'react-i18next';
+import { StatusContext } from '../../../context/Status/index.js';
+import { getQuotaPerUnit } from '../../../helpers/render.js';
 
 export default function ModelSettingsVisualEditor(props) {
   const { t } = useTranslation();
@@ -14,7 +16,11 @@ export default function ModelSettingsVisualEditor(props) {
   const [searchText, setSearchText] = useState('');
   const [currentPage, setCurrentPage] = useState(1);
   const [loading, setLoading] = useState(false);
+  const [pricingMode, setPricingMode] = useState('per-token'); // 'per-token' or 'per-request'
+  const [pricingSubMode, setPricingSubMode] = useState('ratio'); // 'ratio' or 'token-price'
+  const formRef = useRef(null);
   const pageSize = 10;
+  const quotaPerUnit = getQuotaPerUnit()
 
   useEffect(() => {
     try {
@@ -171,11 +177,19 @@ export default function ModelSettingsVisualEditor(props) {
       title: t('操作'),
       key: 'action',
       render: (_, record) => (
-        <Button
-          icon={<IconDelete />}
-          type="danger"
-          onClick={() => deleteModel(record.name)}
-        />
+        <Space>
+          <Button
+            type="primary"
+            icon={<IconEdit />}
+            onClick={() => editModel(record)}
+          >
+          </Button>
+          <Button
+            icon={<IconDelete />}
+            type="danger"
+            onClick={() => deleteModel(record.name)}
+          />
+        </Space>
       )
     }
   ];
@@ -197,28 +211,171 @@ export default function ModelSettingsVisualEditor(props) {
   const deleteModel = (name) => {
     setModels(prev => prev.filter(model => model.name !== name));
   };
-  const addModel = (values) => {
-    // 检查模型名称是否存在, 如果存在则拒绝添加
-    if (models.some(model => model.name === values.name)) {
-      showError('模型名称已存在');
-      return;
+  
+  const calculateRatioFromTokenPrice = (tokenPrice) => {
+    return tokenPrice / 2;
+  };
+  
+  const calculateCompletionRatioFromPrices = (modelTokenPrice, completionTokenPrice) => {
+    if (!modelTokenPrice || modelTokenPrice === '0') {
+      showError('模型价格不能为0');
+      return '';
     }
-    setModels(prev => [{
-      name: values.name,
-      price: values.price || '',
-      ratio: values.ratio || '',
-      completionRatio: values.completionRatio || ''
-    }, ...prev]);
-    setVisible(false);
-    showSuccess('添加成功');
+    return completionTokenPrice / modelTokenPrice;
+  };
+  
+  const handleTokenPriceChange = (value) => {
+
+    // Use a temporary variable to hold the new state
+    let newState = {
+      ...(currentModel || {}),
+      tokenPrice: value,
+      ratio: 0
+    };
+    
+    if (!isNaN(value) && value !== '') {
+      const tokenPrice = parseFloat(value);
+      const ratio = calculateRatioFromTokenPrice(tokenPrice);
+      newState.ratio = ratio;
+    }
+    
+    // Set the state with the complete updated object
+    setCurrentModel(newState);
+  };
+  
+  const handleCompletionTokenPriceChange = (value) => {
+
+    // Use a temporary variable to hold the new state
+    let newState = {
+      ...(currentModel || {}),
+      completionTokenPrice: value,
+      completionRatio: 0
+    };
+    
+    if (!isNaN(value) && value !== '' && currentModel?.tokenPrice) {
+      const completionTokenPrice = parseFloat(value);
+      const modelTokenPrice = parseFloat(currentModel.tokenPrice);
+      
+      if (modelTokenPrice > 0) {
+        const completionRatio = calculateCompletionRatioFromPrices(modelTokenPrice, completionTokenPrice);
+        newState.completionRatio = completionRatio;
+      }
+    }
+    
+    // Set the state with the complete updated object
+    setCurrentModel(newState);
   };
 
+  const addOrUpdateModel = (values) => {
+    // Check if we're editing an existing model or adding a new one
+    const existingModelIndex = models.findIndex(model => model.name === values.name);
+    
+    if (existingModelIndex >= 0) {
+      // Update existing model
+      setModels(prev => prev.map((model, index) => 
+        index === existingModelIndex ? {
+          name: values.name,
+          price: values.price || '',
+          ratio: values.ratio || '',
+          completionRatio: values.completionRatio || ''
+        } : model
+      ));
+      setVisible(false);
+      showSuccess(t('更新成功'));
+    } else {
+      // Add new model
+      // Check if model name already exists
+      if (models.some(model => model.name === values.name)) {
+        showError(t('模型名称已存在'));
+        return;
+      }
+      
+      setModels(prev => [{
+        name: values.name,
+        price: values.price || '',
+        ratio: values.ratio || '',
+        completionRatio: values.completionRatio || ''
+      }, ...prev]);
+      setVisible(false);
+      showSuccess(t('添加成功'));
+    }
+  };
+
+  const calculateTokenPriceFromRatio = (ratio) => {
+    return ratio * 2;
+  };
+  
+  const resetModalState = () => {
+    setCurrentModel(null);
+    setPricingMode('per-token');
+    setPricingSubMode('ratio');
+  };
+
+  const editModel = (record) => {
+
+    // Determine which pricing mode to use based on the model's current configuration
+    let initialPricingMode = 'per-token';
+    let initialPricingSubMode = 'ratio';
+    
+    if (record.price !== '') {
+      initialPricingMode = 'per-request';
+    } else {
+      initialPricingMode = 'per-token';
+      // We default to ratio mode, but could set to token-price if needed
+    }
+    
+    // Set the pricing modes for the form
+    setPricingMode(initialPricingMode);
+    setPricingSubMode(initialPricingSubMode);
+    
+    // Create a copy of the model data to avoid modifying the original
+    const modelCopy = { ...record };
+    
+    // If the model has ratio data and we want to populate token price fields
+    if (record.ratio) {
+      modelCopy.tokenPrice = calculateTokenPriceFromRatio(parseFloat(record.ratio)).toString();
+      
+      if (record.completionRatio) {
+        modelCopy.completionTokenPrice = (parseFloat(modelCopy.tokenPrice) * parseFloat(record.completionRatio)).toString();
+      }
+    }
+    
+    // Set the current model
+    setCurrentModel(modelCopy);
+    
+    // Open the modal
+    setVisible(true);
+    
+    // Use setTimeout to ensure the form is rendered before setting values
+    setTimeout(() => {
+      if (formRef.current) {
+        // Update the form fields based on pricing mode
+        const formValues = {
+          name: modelCopy.name,
+        };
+        
+        if (initialPricingMode === 'per-request') {
+          formValues.priceInput = modelCopy.price;
+        } else if (initialPricingMode === 'per-token') {
+          formValues.ratioInput = modelCopy.ratio;
+          formValues.completionRatioInput = modelCopy.completionRatio;
+          formValues.modelTokenPrice = modelCopy.tokenPrice;
+          formValues.completionTokenPrice = modelCopy.completionTokenPrice;
+        }
+        
+        formRef.current.setValues(formValues);
+      }
+    }, 0);
+  };
 
   return (
     <>
       <Space vertical align="start" style={{ width: '100%' }}>
         <Space>
-          <Button icon={<IconPlus />} onClick={() => setVisible(true)}>
+          <Button icon={<IconPlus />} onClick={() => {
+            resetModalState();
+            setVisible(true);
+          }}>
             {t('添加模型')}
           </Button>
           <Button type="primary" icon={<IconSave />} onClick={SubmitData}>
@@ -256,56 +413,205 @@ export default function ModelSettingsVisualEditor(props) {
       </Space>
 
       <Modal
-        title={t('添加模型')}
+        title={currentModel && currentModel.name && models.some(model => model.name === currentModel.name) ? t('编辑模型') : t('添加模型')}
         visible={visible}
-        onCancel={() => setVisible(false)}
+        onCancel={() => {
+          resetModalState();
+          setVisible(false);
+        }}
         onOk={() => {
-          currentModel && addModel(currentModel);
+          if (currentModel) {
+            // If we're in token price mode, make sure ratio values are properly set
+            const valuesToSave = { ...currentModel };
+            
+            if (pricingMode === 'per-token' && pricingSubMode === 'token-price' && currentModel.tokenPrice) {
+              // Calculate and set ratio from token price
+              const tokenPrice = parseFloat(currentModel.tokenPrice);
+              valuesToSave.ratio = (tokenPrice / 2).toString();
+              
+              // Calculate and set completion ratio if both token prices are available
+              if (currentModel.completionTokenPrice && currentModel.tokenPrice) {
+                const completionPrice = parseFloat(currentModel.completionTokenPrice);
+                const modelPrice = parseFloat(currentModel.tokenPrice);
+                if (modelPrice > 0) {
+                  valuesToSave.completionRatio = (completionPrice / modelPrice).toString();
+                }
+              }
+            }
+            
+            // Clear price if we're in per-token mode
+            if (pricingMode === 'per-token') {
+              valuesToSave.price = '';
+            } else {
+              // Clear ratios if we're in per-request mode
+              valuesToSave.ratio = '';
+              valuesToSave.completionRatio = '';
+            }
+            
+            addOrUpdateModel(valuesToSave);
+          }
         }}
       >
-        <Form>
+        <Form getFormApi={api => formRef.current = api}>
           <Form.Input
             field="name"
             label={t('模型名称')}
             placeholder="strawberry"
             required
+            disabled={currentModel && currentModel.name && models.some(model => model.name === currentModel.name)}
             onChange={value => setCurrentModel(prev => ({ ...prev, name: value }))}
           />
-          <Form.Switch
-            field="priceMode"
-            label={<>{t('定价模式')}：{currentModel?.priceMode ? t("固定价格") : t("倍率模式")}</>}
-            onChange={checked => {
-              setCurrentModel(prev => ({
-                ...prev,
-                price: '',
-                ratio: '',
-                completionRatio: '',
-                priceMode: checked
-              }));
-            }}
-          />
-          {currentModel?.priceMode ? (
+          
+          <Form.Section text={t('定价模式')}>
+            <div style={{ marginBottom: '16px' }}>
+              <RadioGroup type="button" value={pricingMode} onChange={(e) => {
+                const newMode = e.target.value;
+                const oldMode = pricingMode;
+                setPricingMode(newMode);
+                
+                // Instead of resetting all values, convert between modes
+                if (currentModel) {
+                  const updatedModel = { ...currentModel };
+                  
+                  // Update formRef with converted values
+                  if (formRef.current) {
+                    const formValues = {
+                      name: updatedModel.name
+                    };
+                    
+                    if (newMode === 'per-request') {
+                      formValues.priceInput = updatedModel.price || '';
+                    } else if (newMode === 'per-token') {
+                      formValues.ratioInput = updatedModel.ratio || '';
+                      formValues.completionRatioInput = updatedModel.completionRatio || '';
+                      formValues.modelTokenPrice = updatedModel.tokenPrice || '';
+                      formValues.completionTokenPrice = updatedModel.completionTokenPrice || '';
+                    }
+                    
+                    formRef.current.setValues(formValues);
+                  }
+                  
+                  // Update the model state
+                  setCurrentModel(updatedModel);
+                }
+              }}>
+                <Radio value="per-token">{t('按量计费')}</Radio>
+                <Radio value="per-request">{t('按次计费')}</Radio>
+              </RadioGroup>
+            </div>
+          </Form.Section>
+          
+          {pricingMode === 'per-token' && (
+            <>
+              <Form.Section text={t('价格设置方式')}>
+                <div style={{ marginBottom: '16px' }}>
+                  <RadioGroup type="button" value={pricingSubMode} onChange={(e) => {
+                    const newSubMode = e.target.value;
+                    const oldSubMode = pricingSubMode;
+                    setPricingSubMode(newSubMode);
+                    
+                    // Handle conversion between submodes
+                    if (currentModel) {
+                      const updatedModel = { ...currentModel };
+                      
+                      // Convert between ratio and token price
+                      if (oldSubMode === 'ratio' && newSubMode === 'token-price') {
+                        if (updatedModel.ratio) {
+                          updatedModel.tokenPrice = calculateTokenPriceFromRatio(parseFloat(updatedModel.ratio)).toString();
+                          
+                          if (updatedModel.completionRatio) {
+                            updatedModel.completionTokenPrice = (parseFloat(updatedModel.tokenPrice) * parseFloat(updatedModel.completionRatio)).toString();
+                          }
+                        }
+                      } else if (oldSubMode === 'token-price' && newSubMode === 'ratio') {
+                        // Ratio values should already be calculated by the handlers
+                      }
+                      
+                      // Update the form values
+                      if (formRef.current) {
+                        const formValues = {};
+                        
+                        if (newSubMode === 'ratio') {
+                          formValues.ratioInput = updatedModel.ratio || '';
+                          formValues.completionRatioInput = updatedModel.completionRatio || '';
+                        } else if (newSubMode === 'token-price') {
+                          formValues.modelTokenPrice = updatedModel.tokenPrice || '';
+                          formValues.completionTokenPrice = updatedModel.completionTokenPrice || '';
+                        }
+                        
+                        formRef.current.setValues(formValues);
+                      }
+                      
+                      setCurrentModel(updatedModel);
+                    }
+                  }}>
+                    <Radio value="ratio">{t('按倍率设置')}</Radio>
+                    <Radio value="token-price">{t('按价格设置')}</Radio>
+                  </RadioGroup>
+                </div>
+              </Form.Section>
+              
+              {pricingSubMode === 'ratio' && (
+                <>
+                  <Form.Input
+                    field="ratioInput"
+                    label={t('模型倍率')}
+                    placeholder={t('输入模型倍率')}
+                    onChange={value => setCurrentModel(prev => ({ 
+                      ...prev || {}, 
+                      ratio: value 
+                    }))}
+                    initValue={currentModel?.ratio || ''}
+                  />
+                  <Form.Input
+                    field="completionRatioInput"
+                    label={t('补全倍率')}
+                    placeholder={t('输入补全倍率')}
+                    onChange={value => setCurrentModel(prev => ({ 
+                      ...prev || {}, 
+                      completionRatio: value 
+                    }))}
+                    initValue={currentModel?.completionRatio || ''}
+                  />
+                </>
+              )}
+              
+              {pricingSubMode === 'token-price' && (
+                <>
+                  <Form.Input
+                    field="modelTokenPrice"
+                    label={t('输入价格')}
+                    onChange={(value) => {
+                      handleTokenPriceChange(value);
+                    }}
+                    initValue={currentModel?.tokenPrice || ''}
+                    suffix={t('$/1M tokens')}
+                  />
+                  <Form.Input
+                    field="completionTokenPrice"
+                    label={t('输出价格')}
+                    onChange={(value) => {
+                      handleCompletionTokenPriceChange(value);
+                    }}
+                    initValue={currentModel?.completionTokenPrice || ''}
+                    suffix={t('$/1M tokens')}
+                  />
+                </>
+              )}
+            </>
+          )}
+          
+          {pricingMode === 'per-request' && (
             <Form.Input
-              field="price"
+              field="priceInput"
               label={t('固定价格(每次)')}
               placeholder={t('输入每次价格')}
-              onChange={value => setCurrentModel(prev => ({ ...prev, price: value }))}
+              onChange={value => setCurrentModel(prev => ({ 
+                ...prev || {}, 
+                price: value 
+              }))}
+              initValue={currentModel?.price || ''}
             />
-          ) : (
-            <>
-              <Form.Input
-                field="ratio"
-                label={t('模型倍率')}
-                placeholder={t('输入模型倍率')}
-                onChange={value => setCurrentModel(prev => ({ ...prev, ratio: value }))}
-              />
-              <Form.Input
-                field="completionRatio"
-                label={t('补全倍率')}
-                placeholder={t('输入补全价格')}
-                onChange={value => setCurrentModel(prev => ({ ...prev, completionRatio: value }))}
-              />
-            </>
           )}
         </Form>
       </Modal>