Merge pull request #2412 from seefs001/pr-2372

feat: add openai video remix endpoint

Dockerfile

@@ -14,7 +14,7 @@ ENV GO111MODULE=on CGO_ENABLED=0
 ARG TARGETOS
 ARG TARGETARCH
 ENV GOOS=${TARGETOS:-linux} GOARCH=${TARGETARCH:-amd64}
-
+ENV GOEXPERIMENT=greenteagc
 
 WORKDIR /build
 
@@ -25,10 +25,11 @@ COPY . .
 COPY --from=builder /build/dist ./web/dist
 RUN go build -ldflags "-s -w -X 'github.com/QuantumNous/new-api/common.Version=$(cat VERSION)'" -o new-api
 
-FROM alpine
+FROM debian:bookworm-slim
 
-RUN apk upgrade --no-cache \
-    && apk add --no-cache ca-certificates tzdata \
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends ca-certificates tzdata libasan8 \
+    && rm -rf /var/lib/apt/lists/* \
     && update-ca-certificates
 
 COPY --from=builder2 /build/new-api /

README.en.md

@@ -238,6 +238,7 @@ docker run --name new-api -d --restart always \
 - `gemini-2.5-flash-nothinking` - Disable thinking mode
 - `gemini-2.5-pro-thinking` - Enable thinking mode
 - `gemini-2.5-pro-thinking-128` - Enable thinking mode with thinking budget of 128 tokens
+- You can also append `-low`, `-medium`, or `-high` to any Gemini model name to request the corresponding reasoning effort (no extra thinking-budget suffix needed).
 
 </details>
 

README.fr.md

@@ -234,6 +234,7 @@ docker run --name new-api -d --restart always \
 - `gemini-2.5-flash-nothinking` - Désactiver le mode de pensée
 - `gemini-2.5-pro-thinking` - Activer le mode de pensée
 - `gemini-2.5-pro-thinking-128` - Activer le mode de pensée avec budget de pensée de 128 tokens
+- Vous pouvez également ajouter les suffixes `-low`, `-medium` ou `-high` aux modèles Gemini pour fixer le niveau d’effort de raisonnement (sans suffixe de budget supplémentaire).
 
 </details>
 

README.ja.md

@@ -243,6 +243,7 @@ docker run --name new-api -d --restart always \
 - `gemini-2.5-flash-nothinking` - 思考モードを無効にする
 - `gemini-2.5-pro-thinking` - 思考モードを有効にする
 - `gemini-2.5-pro-thinking-128` - 思考モードを有効にし、思考予算を128トークンに設定する
+- Gemini モデル名の末尾に `-low` / `-medium` / `-high` を付けることで推論強度を直接指定できます（追加の思考予算サフィックスは不要です）。
 
 </details>
 

README.md

@@ -239,6 +239,7 @@ docker run --name new-api -d --restart always \
 - `gemini-2.5-flash-nothinking` - 禁用思考模式
 - `gemini-2.5-pro-thinking` - 启用思考模式
 - `gemini-2.5-pro-thinking-128` - 启用思考模式，并设置思考预算为128tokens
+- 也可以直接在 Gemini 模型名称后追加 `-low` / `-medium` / `-high` 来控制思考力度（无需再设置思考预算后缀）
 
 </details>
 

common/constants.go

@@ -121,6 +121,9 @@ var BatchUpdateInterval int
 
 var RelayTimeout int // unit is second
 
+var RelayMaxIdleConns int
+var RelayMaxIdleConnsPerHost int
+
 var GeminiSafetySetting string
 
 // https://docs.cohere.com/docs/safety-modes Type; NONE/CONTEXTUAL/STRICT

common/init.go

@@ -90,6 +90,8 @@ func InitEnv() {
 	SyncFrequency = GetEnvOrDefault("SYNC_FREQUENCY", 60)
 	BatchUpdateInterval = GetEnvOrDefault("BATCH_UPDATE_INTERVAL", 5)
 	RelayTimeout = GetEnvOrDefault("RELAY_TIMEOUT", 0)
+	RelayMaxIdleConns = GetEnvOrDefault("RELAY_MAX_IDLE_CONNS", 500)
+	RelayMaxIdleConnsPerHost = GetEnvOrDefault("RELAY_MAX_IDLE_CONNS_PER_HOST", 100)
 
 	// Initialize string variables with GetEnvOrDefaultString
 	GeminiSafetySetting = GetEnvOrDefaultString("GEMINI_SAFETY_SETTING", "BLOCK_NONE")
@@ -129,6 +131,8 @@ func initConstantEnv() {
 	constant.GenerateDefaultToken = GetEnvOrDefaultBool("GENERATE_DEFAULT_TOKEN", false)
 	// 是否启用错误日志
 	constant.ErrorLogEnabled = GetEnvOrDefaultBool("ERROR_LOG_ENABLED", false)
+	// 任务轮询时查询的最大数量
+	constant.TaskQueryLimit = GetEnvOrDefault("TASK_QUERY_LIMIT", 1000)
 
 	soraPatchStr := GetEnvOrDefaultString("TASK_PRICE_PATCH", "")
 	if soraPatchStr != "" {

common/json.go

@@ -23,11 +23,11 @@ func Marshal(v any) ([]byte, error) {
 }
 
 func GetJsonType(data json.RawMessage) string {
-	data = bytes.TrimSpace(data)
-	if len(data) == 0 {
+	trimmed := bytes.TrimSpace(data)
+	if len(trimmed) == 0 {
 		return "unknown"
 	}
-	firstChar := bytes.TrimSpace(data)[0]
+	firstChar := trimmed[0]
 	switch firstChar {
 	case '{':
 		return "object"

common/model.go

@@ -17,6 +17,13 @@ var (
 		"flux-",
 		"flux.1-",
 	}
+	OpenAITextModels = []string{
+		"gpt-",
+		"o1",
+		"o3",
+		"o4",
+		"chatgpt",
+	}
 )
 
 func IsOpenAIResponseOnlyModel(modelName string) bool {
@@ -40,3 +47,13 @@ func IsImageGenerationModel(modelName string) bool {
 	}
 	return false
 }
+
+func IsOpenAITextModel(modelName string) bool {
+	modelName = strings.ToLower(modelName)
+	for _, m := range OpenAITextModels {
+		if strings.Contains(modelName, m) {
+			return true
+		}
+	}
+	return false
+}

common/str.go

@@ -3,12 +3,19 @@ package common
 import (
 	"encoding/base64"
 	"encoding/json"
-	"math/rand"
 	"net/url"
 	"regexp"
 	"strconv"
 	"strings"
 	"unsafe"
+
+	"github.com/samber/lo"
+)
+
+var (
+	maskURLPattern    = regexp.MustCompile(`(http|https)://[^\s/$.?#].[^\s]*`)
+	maskDomainPattern = regexp.MustCompile(`\b(?:[a-zA-Z0-9](?:[a-zA-Z0-9-]{0,61}[a-zA-Z0-9])?\.)+[a-zA-Z]{2,}\b`)
+	maskIPPattern     = regexp.MustCompile(`\b(?:\d{1,3}\.){3}\d{1,3}\b`)
 )
 
 func GetStringIfEmpty(str string, defaultValue string) string {
@@ -19,12 +26,10 @@ func GetStringIfEmpty(str string, defaultValue string) string {
 }
 
 func GetRandomString(length int) string {
-	//rand.Seed(time.Now().UnixNano())
-	key := make([]byte, length)
-	for i := 0; i < length; i++ {
-		key[i] = keyChars[rand.Intn(len(keyChars))]
+	if length <= 0 {
+		return ""
 	}
-	return string(key)
+	return lo.RandomString(length, lo.AlphanumericCharset)
 }
 
 func MapToJsonStr(m map[string]interface{}) string {
@@ -170,8 +175,7 @@ func maskHostForPlainDomain(domain string) string {
 // api.openai.com -> ***.***.com
 func MaskSensitiveInfo(str string) string {
 	// Mask URLs
-	urlPattern := regexp.MustCompile(`(http|https)://[^\s/$.?#].[^\s]*`)
-	str = urlPattern.ReplaceAllStringFunc(str, func(urlStr string) string {
+	str = maskURLPattern.ReplaceAllStringFunc(str, func(urlStr string) string {
 		u, err := url.Parse(urlStr)
 		if err != nil {
 			return urlStr
@@ -224,14 +228,12 @@ func MaskSensitiveInfo(str string) string {
 	})
 
 	// Mask domain names without protocol (like openai.com, www.openai.com)
-	domainPattern := regexp.MustCompile(`\b(?:[a-zA-Z0-9](?:[a-zA-Z0-9-]{0,61}[a-zA-Z0-9])?\.)+[a-zA-Z]{2,}\b`)
-	str = domainPattern.ReplaceAllStringFunc(str, func(domain string) string {
+	str = maskDomainPattern.ReplaceAllStringFunc(str, func(domain string) string {
 		return maskHostForPlainDomain(domain)
 	})
 
 	// Mask IP addresses
-	ipPattern := regexp.MustCompile(`\b(?:\d{1,3}\.){3}\d{1,3}\b`)
-	str = ipPattern.ReplaceAllString(str, "***.***.***.***")
+	str = maskIPPattern.ReplaceAllString(str, "***.***.***.***")
 
 	return str
 }

constant/context_key.go

@@ -3,8 +3,9 @@ package constant
 type ContextKey string
 
 const (
-	ContextKeyTokenCountMeta ContextKey = "token_count_meta"
-	ContextKeyPromptTokens   ContextKey = "prompt_tokens"
+	ContextKeyTokenCountMeta  ContextKey = "token_count_meta"
+	ContextKeyPromptTokens    ContextKey = "prompt_tokens"
+	ContextKeyEstimatedTokens ContextKey = "estimated_tokens"
 
 	ContextKeyOriginalModel    ContextKey = "original_model"
 	ContextKeyRequestStartTime ContextKey = "request_start_time"

constant/env.go

@@ -15,6 +15,7 @@ var NotifyLimitCount int
 var NotificationLimitDurationMinute int
 var GenerateDefaultToken bool
 var ErrorLogEnabled bool
+var TaskQueryLimit int
 
 // temporary variable for sora patch, will be removed in future
 var TaskPricePatches []string

constant/task.go

@@ -15,6 +15,7 @@ const (
 	TaskActionTextGenerate      = "textGenerate"
 	TaskActionFirstTailGenerate = "firstTailGenerate"
 	TaskActionReferenceGenerate = "referenceGenerate"
+	TaskActionRemix             = "remixGenerate"
 )
 
 var SunoModel2Action = map[string]string{

controller/channel-test.go

@@ -351,7 +351,7 @@ func testChannel(channel *model.Channel, testModel string, endpointType string)
 			newAPIError: types.NewOpenAIError(err, types.ErrorCodeReadResponseBodyFailed, http.StatusInternalServerError),
 		}
 	}
-	info.PromptTokens = usage.PromptTokens
+	info.SetEstimatePromptTokens(usage.PromptTokens)
 
 	quota := 0
 	if !priceData.UsePrice {

controller/channel.go

@@ -165,6 +165,30 @@ func GetAllChannels(c *gin.Context) {
 	return
 }
 
+func buildFetchModelsHeaders(channel *model.Channel, key string) (http.Header, error) {
+	var headers http.Header
+	switch channel.Type {
+	case constant.ChannelTypeAnthropic:
+		headers = GetClaudeAuthHeader(key)
+	default:
+		headers = GetAuthHeader(key)
+	}
+
+	headerOverride := channel.GetHeaderOverride()
+	for k, v := range headerOverride {
+		str, ok := v.(string)
+		if !ok {
+			return nil, fmt.Errorf("invalid header override for key %s", k)
+		}
+		if strings.Contains(str, "{api_key}") {
+			str = strings.ReplaceAll(str, "{api_key}", key)
+		}
+		headers.Set(k, str)
+	}
+
+	return headers, nil
+}
+
 func FetchUpstreamModels(c *gin.Context) {
 	id, err := strconv.Atoi(c.Param("id"))
 	if err != nil {
@@ -223,14 +247,13 @@ func FetchUpstreamModels(c *gin.Context) {
 	}
 	key = strings.TrimSpace(key)
 
-	// 获取响应体 - 根据渠道类型决定是否添加 AuthHeader
-	var body []byte
-	switch channel.Type {
-	case constant.ChannelTypeAnthropic:
-		body, err = GetResponseBody("GET", url, channel, GetClaudeAuthHeader(key))
-	default:
-		body, err = GetResponseBody("GET", url, channel, GetAuthHeader(key))
+	headers, err := buildFetchModelsHeaders(channel, key)
+	if err != nil {
+		common.ApiError(c, err)
+		return
 	}
+
+	body, err := GetResponseBody("GET", url, channel, headers)
 	if err != nil {
 		common.ApiError(c, err)
 		return

controller/relay.go

@@ -125,13 +125,13 @@ func Relay(c *gin.Context, relayFormat types.RelayFormat) {
 		}
 	}
 
-	tokens, err := service.CountRequestToken(c, meta, relayInfo)
+	tokens, err := service.EstimateRequestToken(c, meta, relayInfo)
 	if err != nil {
 		newAPIError = types.NewError(err, types.ErrorCodeCountTokenFailed)
 		return
 	}
 
-	relayInfo.SetPromptTokens(tokens)
+	relayInfo.SetEstimatePromptTokens(tokens)
 
 	priceData, err := helper.ModelPriceHelper(c, relayInfo, tokens, meta)
 	if err != nil {
@@ -285,7 +285,7 @@ func processChannelError(c *gin.Context, channelError types.ChannelError, err *t
 	logger.LogError(c, fmt.Sprintf("channel error (channel #%d, status code: %d): %s", channelError.ChannelId, err.StatusCode, err.Error()))
 	// 不要使用context获取渠道信息，异步处理时可能会出现渠道信息不一致的情况
 	// do not use context to get channel info, there may be inconsistent channel info when processing asynchronously
-	if service.ShouldDisableChannel(channelError.ChannelId, err) && channelError.AutoBan {
+	if service.ShouldDisableChannel(channelError.ChannelType, err) && channelError.AutoBan {
 		gopool.Go(func() {
 			service.DisableChannel(channelError, err.Error())
 		})

controller/task.go

@@ -29,7 +29,7 @@ func UpdateTaskBulk() {
 		time.Sleep(time.Duration(15) * time.Second)
 		common.SysLog("任务进度轮询开始")
 		ctx := context.TODO()
-		allTasks := model.GetAllUnFinishSyncTasks(500)
+		allTasks := model.GetAllUnFinishSyncTasks(constant.TaskQueryLimit)
 		platformTask := make(map[constant.TaskPlatform][]*model.Task)
 		for _, t := range allTasks {
 			platformTask[t.Platform] = append(platformTask[t.Platform], t)
@@ -116,9 +116,10 @@ func updateSunoTaskAll(ctx context.Context, channelId int, taskIds []string, tas
 	if adaptor == nil {
 		return errors.New("adaptor not found")
 	}
+	proxy := channel.GetSetting().Proxy
 	resp, err := adaptor.FetchTask(*channel.BaseURL, channel.Key, map[string]any{
 		"ids": taskIds,
-	})
+	}, proxy)
 	if err != nil {
 		common.SysLog(fmt.Sprintf("Get Task Do req error: %v", err))
 		return err

controller/task_video.go

@@ -67,6 +67,7 @@ func updateVideoSingleTask(ctx context.Context, adaptor channel.TaskAdaptor, cha
 	if channel.GetBaseURL() != "" {
 		baseURL = channel.GetBaseURL()
 	}
+	proxy := channel.GetSetting().Proxy
 
 	task := taskM[taskId]
 	if task == nil {
@@ -76,7 +77,7 @@ func updateVideoSingleTask(ctx context.Context, adaptor channel.TaskAdaptor, cha
 	resp, err := adaptor.FetchTask(baseURL, channel.Key, map[string]any{
 		"task_id": taskId,
 		"action":  task.Action,
-	})
+	}, proxy)
 	if err != nil {
 		return fmt.Errorf("fetchTask failed for task %s: %w", taskId, err)
 	}

controller/token.go

@@ -142,7 +142,7 @@ func AddToken(c *gin.Context) {
 		common.ApiError(c, err)
 		return
 	}
-	if len(token.Name) > 30 {
+	if len(token.Name) > 50 {
 		c.JSON(http.StatusOK, gin.H{
 			"success": false,
 			"message": "令牌名称过长",
@@ -208,7 +208,7 @@ func UpdateToken(c *gin.Context) {
 		common.ApiError(c, err)
 		return
 	}
-	if len(token.Name) > 30 {
+	if len(token.Name) > 50 {
 		c.JSON(http.StatusOK, gin.H{
 			"success": false,
 			"message": "令牌名称过长",

controller/video_proxy.go

@@ -1,6 +1,7 @@
 package controller
 
 import (
+	"context"
 	"fmt"
 	"io"
 	"net/http"
@@ -10,6 +11,7 @@ import (
 	"github.com/QuantumNous/new-api/constant"
 	"github.com/QuantumNous/new-api/logger"
 	"github.com/QuantumNous/new-api/model"
+	"github.com/QuantumNous/new-api/service"
 
 	"github.com/gin-gonic/gin"
 )
@@ -75,11 +77,22 @@ func VideoProxy(c *gin.Context) {
 	}
 
 	var videoURL string
-	client := &http.Client{
-		Timeout: 60 * time.Second,
+	proxy := channel.GetSetting().Proxy
+	client, err := service.GetHttpClientWithProxy(proxy)
+	if err != nil {
+		logger.LogError(c.Request.Context(), fmt.Sprintf("Failed to create proxy client for task %s: %s", taskID, err.Error()))
+		c.JSON(http.StatusInternalServerError, gin.H{
+			"error": gin.H{
+				"message": "Failed to create proxy client",
+				"type":    "server_error",
+			},
+		})
+		return
 	}
 
-	req, err := http.NewRequestWithContext(c.Request.Context(), http.MethodGet, "", nil)
+	ctx, cancel := context.WithTimeout(c.Request.Context(), 60*time.Second)
+	defer cancel()
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, "", nil)
 	if err != nil {
 		logger.LogError(c.Request.Context(), fmt.Sprintf("Failed to create request: %s", err.Error()))
 		c.JSON(http.StatusInternalServerError, gin.H{

controller/video_proxy_gemini.go

@@ -35,10 +35,11 @@ func getGeminiVideoURL(channel *model.Channel, task *model.Task, apiKey string)
 		return "", fmt.Errorf("api key not available for task")
 	}
 
+	proxy := channel.GetSetting().Proxy
 	resp, err := adaptor.FetchTask(baseURL, apiKey, map[string]any{
 		"task_id": task.TaskID,
 		"action":  task.Action,
-	})
+	}, proxy)
 	if err != nil {
 		return "", fmt.Errorf("fetch task failed: %w", err)
 	}

docs/openapi/relay.json

@@ -16,13 +16,13 @@
       "name": "OpenAI格式(Responses)"
     },
     {
-      "name": "图片生成/编辑(Images)"
+      "name": "图片生成"
     },
     {
-      "name": "图片生成/编辑(Images)/OpenAI兼容格式"
+      "name": "图片生成/OpenAI兼容格式"
     },
     {
-      "name": "图片生成/编辑(Images)/Qwen千问"
+      "name": "图片生成/Qwen千问"
     },
     {
       "name": "视频生成"
@@ -292,7 +292,7 @@
         "description": " 百炼qwen-image系列图片生成",
         "operationId": "createImage",
         "tags": [
-          "图片生成/编辑(Images)/Qwen千问"
+          "图片生成/Qwen千问"
         ],
         "parameters": [],
         "requestBody": {
@@ -408,7 +408,7 @@
         "description": " 百炼qwen-image系列图片生成",
         "operationId": "createImage",
         "tags": [
-          "图片生成/编辑(Images)/Qwen千问"
+          "图片生成/Qwen千问"
         ],
         "parameters": [],
         "requestBody": {

dto/gemini.go

@@ -142,7 +142,7 @@ type GeminiThinkingConfig struct {
 	IncludeThoughts bool `json:"includeThoughts,omitempty"`
 	ThinkingBudget  *int `json:"thinkingBudget,omitempty"`
 	// TODO Conflict with thinkingbudget.
-	ThinkingLevel json.RawMessage `json:"thinkingLevel,omitempty"`
+	ThinkingLevel string `json:"thinkingLevel,omitempty"`
 }
 
 // UnmarshalJSON allows GeminiThinkingConfig to accept both snake_case and camelCase fields.
@@ -150,9 +150,9 @@ func (c *GeminiThinkingConfig) UnmarshalJSON(data []byte) error {
 	type Alias GeminiThinkingConfig
 	var aux struct {
 		Alias
-		IncludeThoughtsSnake *bool           `json:"include_thoughts,omitempty"`
-		ThinkingBudgetSnake  *int            `json:"thinking_budget,omitempty"`
-		ThinkingLevelSnake   json.RawMessage `json:"thinking_level,omitempty"`
+		IncludeThoughtsSnake *bool  `json:"include_thoughts,omitempty"`
+		ThinkingBudgetSnake  *int   `json:"thinking_budget,omitempty"`
+		ThinkingLevelSnake   string `json:"thinking_level,omitempty"`
 	}
 
 	if err := common.Unmarshal(data, &aux); err != nil {
@@ -169,7 +169,7 @@ func (c *GeminiThinkingConfig) UnmarshalJSON(data []byte) error {
 		c.ThinkingBudget = aux.ThinkingBudgetSnake
 	}
 
-	if len(aux.ThinkingLevelSnake) > 0 {
+	if aux.ThinkingLevelSnake != "" {
 		c.ThinkingLevel = aux.ThinkingLevelSnake
 	}
 

dto/openai_image.go

@@ -27,8 +27,11 @@ type ImageRequest struct {
 	OutputCompression json.RawMessage `json:"output_compression,omitempty"`
 	PartialImages     json.RawMessage `json:"partial_images,omitempty"`
 	// Stream            bool            `json:"stream,omitempty"`
-	Watermark *bool           `json:"watermark,omitempty"`
-	Image     json.RawMessage `json:"image,omitempty"`
+	Watermark *bool `json:"watermark,omitempty"`
+	// zhipu 4v
+	WatermarkEnabled json.RawMessage `json:"watermark_enabled,omitempty"`
+	UserId           json.RawMessage `json:"user_id,omitempty"`
+	Image            json.RawMessage `json:"image,omitempty"`
 	// 用匿名参数接收额外参数
 	Extra map[string]json.RawMessage `json:"-"`
 }

dto/openai_request.go

@@ -83,6 +83,7 @@ type GeneralOpenAIRequest struct {
 	// Ali Qwen Params
 	VlHighResolutionImages json.RawMessage `json:"vl_high_resolution_images,omitempty"`
 	EnableThinking         any             `json:"enable_thinking,omitempty"`
+	ChatTemplateKwargs     json.RawMessage `json:"chat_template_kwargs,omitempty"`
 	// ollama Params
 	Think json.RawMessage `json:"think,omitempty"`
 	// baidu v2

go.mod

@@ -33,7 +33,7 @@ require (
 	github.com/mewkiz/flac v1.0.13
 	github.com/pkg/errors v0.9.1
 	github.com/pquerna/otp v1.5.0
-	github.com/samber/lo v1.39.0
+	github.com/samber/lo v1.52.0
 	github.com/shirou/gopsutil v3.21.11+incompatible
 	github.com/shopspring/decimal v1.4.0
 	github.com/stripe/stripe-go/v81 v81.4.0
@@ -99,6 +99,7 @@ require (
 	github.com/mitchellh/mapstructure v1.5.0 // indirect
 	github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd // indirect
 	github.com/modern-go/reflect2 v1.0.2 // indirect
+	github.com/ncruces/go-strftime v0.1.9 // indirect
 	github.com/pelletier/go-toml/v2 v2.2.1 // indirect
 	github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec // indirect
 	github.com/tidwall/match v1.1.1 // indirect
@@ -110,13 +111,13 @@ require (
 	github.com/x448/float16 v0.8.4 // indirect
 	github.com/yusufpapurcu/wmi v1.2.3 // indirect
 	golang.org/x/arch v0.21.0 // indirect
-	golang.org/x/exp v0.0.0-20240404231335-c0f41cb1a7a0 // indirect
+	golang.org/x/exp v0.0.0-20250620022241-b7579e27df2b // indirect
 	golang.org/x/sys v0.38.0 // indirect
 	golang.org/x/text v0.31.0 // indirect
 	google.golang.org/protobuf v1.34.2 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect
-	modernc.org/libc v1.22.5 // indirect
-	modernc.org/mathutil v1.5.0 // indirect
-	modernc.org/memory v1.5.0 // indirect
-	modernc.org/sqlite v1.23.1 // indirect
+	modernc.org/libc v1.66.10 // indirect
+	modernc.org/mathutil v1.7.1 // indirect
+	modernc.org/memory v1.11.0 // indirect
+	modernc.org/sqlite v1.40.1 // indirect
 )

go.sum

@@ -120,6 +120,7 @@ github.com/google/go-tpm v0.9.5/go.mod h1:h9jEsEECg7gtLis0upRBQU+GhYVH6jMjrFxI8u
 github.com/google/gofuzz v1.0.0/go.mod h1:dBl0BpW6vV/+mYPU4Po3pmUjxk6FQPldtuIdl/M65Eg=
 github.com/google/pprof v0.0.0-20221118152302-e6195bd50e26 h1:Xim43kblpZXfIBQsbuBVKCudVG457BR2GZFIz3uw3hQ=
 github.com/google/pprof v0.0.0-20221118152302-e6195bd50e26/go.mod h1:dDKJzRmX4S37WGHujM7tX//fmj1uioxKzKxz3lo4HJo=
+github.com/google/pprof v0.0.0-20250317173921-a4b03ec1a45e h1:ijClszYn+mADRFY17kjQEVQ1XRhq2/JR1M3sGqeJoxs=
 github.com/google/uuid v1.1.2/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
 github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
 github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
@@ -193,6 +194,8 @@ github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd/go.mod h1:6dJ
 github.com/modern-go/reflect2 v0.0.0-20180701023420-4b7aa43c6742/go.mod h1:bx2lNnkwVCuqBIxFjflWJWanXIb3RllmbCylyMrvgv0=
 github.com/modern-go/reflect2 v1.0.2 h1:xBagoLtFs94CBntxluKeaWgTMpvLxC4ur3nMaC9Gz0M=
 github.com/modern-go/reflect2 v1.0.2/go.mod h1:yWuevngMOJpCy52FWWMvUC8ws7m/LJsjYzDa0/r8luk=
+github.com/ncruces/go-strftime v0.1.9 h1:bY0MQC28UADQmHmaF5dgpLmImcShSi2kHU9XLdhx/f4=
+github.com/ncruces/go-strftime v0.1.9/go.mod h1:Fwc5htZGVVkseilnfgOVb9mKy6w1naJmn9CehxcKcls=
 github.com/nxadm/tail v1.4.8 h1:nPr65rt6Y5JFSKQO7qToXr7pePgD6Gwiw05lkbyAQTE=
 github.com/nxadm/tail v1.4.8/go.mod h1:+ncqLTQzXmGhMZNUePPaPqPvBxHAIsmXswZKocGu+AU=
 github.com/onsi/ginkgo v1.16.5 h1:8xi0RTUf59SOSfEtZMvwTvXYMzG4gV23XVHOZiXNtnE=
@@ -219,6 +222,8 @@ github.com/rogpeppe/go-internal v1.8.0 h1:FCbCCtXNOY3UtUuHUYaghJg4y7Fd14rXifAYUA
 github.com/rogpeppe/go-internal v1.8.0/go.mod h1:WmiCO8CzOY8rg0OYDC4/i/2WRWAB6poM+XZ2dLUbcbE=
 github.com/samber/lo v1.39.0 h1:4gTz1wUhNYLhFSKl6O+8peW0v2F4BCY034GRpU9WnuA=
 github.com/samber/lo v1.39.0/go.mod h1:+m/ZKRl6ClXCE2Lgf3MsQlWfh4bn1bz6CXEOxnEXnEA=
+github.com/samber/lo v1.52.0 h1:Rvi+3BFHES3A8meP33VPAxiBZX/Aws5RxrschYGjomw=
+github.com/samber/lo v1.52.0/go.mod h1:4+MXEGsJzbKGaUEQFKBq2xtfuznW9oz/WrgyzMzRoM0=
 github.com/shirou/gopsutil v3.21.11+incompatible h1:+1+c1VGhc88SSonWP6foOcLhvnKlUeu/erjjvaPEYiI=
 github.com/shirou/gopsutil v3.21.11+incompatible/go.mod h1:5b4v6he4MtMOwMlS0TUMTu2PcXUg8+E1lC7eC3UO/RA=
 github.com/shopspring/decimal v1.4.0 h1:bxl37RwXBklmTi0C79JfXCEBD1cqqHt0bbgBAGFp81k=
@@ -285,6 +290,8 @@ golang.org/x/crypto v0.45.0 h1:jMBrvKuj23MTlT0bQEOBcAE0mjg8mK9RXFhRH6nyF3Q=
 golang.org/x/crypto v0.45.0/go.mod h1:XTGrrkGJve7CYK7J8PEww4aY7gM3qMCElcJQ8n8JdX4=
 golang.org/x/exp v0.0.0-20240404231335-c0f41cb1a7a0 h1:985EYyeCOxTpcgOTJpflJUwOeEz0CQOdPt73OzpE9F8=
 golang.org/x/exp v0.0.0-20240404231335-c0f41cb1a7a0/go.mod h1:/lliqkxwWAhPjf5oSOIJup2XcqJaw8RGS6k3TGEc7GI=
+golang.org/x/exp v0.0.0-20250620022241-b7579e27df2b h1:M2rDM6z3Fhozi9O7NWsxAkg/yqS/lQJ6PmkyIV3YP+o=
+golang.org/x/exp v0.0.0-20250620022241-b7579e27df2b/go.mod h1:3//PLf8L/X+8b4vuAfHzxeRUl04Adcb341+IGKfnqS8=
 golang.org/x/image v0.23.0 h1:HseQ7c2OpPKTPVzNjG5fwJsOTCiiwS4QdsYi5XU6H68=
 golang.org/x/image v0.23.0/go.mod h1:wJJBTdLfCCf3tiHa1fNxpZmUI4mmoZvwMCPP0ddoNKY=
 golang.org/x/net v0.0.0-20210226172049-e18ecbb05110/go.mod h1:m0MpNAwzfU5UDzcl9v0D8zg8gWTRqZa9RBIspLL5mdg=
@@ -345,9 +352,17 @@ gorm.io/gorm v1.25.2 h1:gs1o6Vsa+oVKG/a9ElL3XgyGfghFfkKA2SInQaCyMho=
 gorm.io/gorm v1.25.2/go.mod h1:L4uxeKpfBml98NYqVqwAdmV1a2nBtAec/cf3fpucW/k=
 modernc.org/libc v1.22.5 h1:91BNch/e5B0uPbJFgqbxXuOnxBQjlS//icfQEGmvyjE=
 modernc.org/libc v1.22.5/go.mod h1:jj+Z7dTNX8fBScMVNRAYZ/jF91K8fdT2hYMThc3YjBY=
+modernc.org/libc v1.66.10 h1:yZkb3YeLx4oynyR+iUsXsybsX4Ubx7MQlSYEw4yj59A=
+modernc.org/libc v1.66.10/go.mod h1:8vGSEwvoUoltr4dlywvHqjtAqHBaw0j1jI7iFBTAr2I=
 modernc.org/mathutil v1.5.0 h1:rV0Ko/6SfM+8G+yKiyI830l3Wuz1zRutdslNoQ0kfiQ=
 modernc.org/mathutil v1.5.0/go.mod h1:mZW8CKdRPY1v87qxC/wUdX5O1qDzXMP5TH3wjfpga6E=
+modernc.org/mathutil v1.7.1 h1:GCZVGXdaN8gTqB1Mf/usp1Y/hSqgI2vAGGP4jZMCxOU=
+modernc.org/mathutil v1.7.1/go.mod h1:4p5IwJITfppl0G4sUEDtCr4DthTaT47/N3aT6MhfgJg=
 modernc.org/memory v1.5.0 h1:N+/8c5rE6EqugZwHii4IFsaJ7MUhoWX07J5tC/iI5Ds=
 modernc.org/memory v1.5.0/go.mod h1:PkUhL0Mugw21sHPeskwZW4D6VscE/GQJOnIpCnW6pSU=
+modernc.org/memory v1.11.0 h1:o4QC8aMQzmcwCK3t3Ux/ZHmwFPzE6hf2Y5LbkRs+hbI=
+modernc.org/memory v1.11.0/go.mod h1:/JP4VbVC+K5sU2wZi9bHoq2MAkCnrt2r98UGeSK7Mjw=
 modernc.org/sqlite v1.23.1 h1:nrSBg4aRQQwq59JpvGEQ15tNxoO5pX/kUjcRNwSAGQM=
 modernc.org/sqlite v1.23.1/go.mod h1:OrDj17Mggn6MhE+iPbBNf7RGKODDE9NFT0f3EwDzJqk=
+modernc.org/sqlite v1.40.1 h1:VfuXcxcUWWKRBuP8+BR9L7VnmusMgBNNnBYGEe9w/iY=
+modernc.org/sqlite v1.40.1/go.mod h1:9fjQZ0mB1LLP0GYrp39oOJXx/I2sxEnZtzCmEQIKvGE=

middleware/distributor.go

@@ -181,6 +181,10 @@ func getModelRequest(c *gin.Context) (*ModelRequest, bool, error) {
 		}
 		c.Set("platform", string(constant.TaskPlatformSuno))
 		c.Set("relay_mode", relayMode)
+	} else if strings.Contains(c.Request.URL.Path, "/v1/videos/") && strings.HasSuffix(c.Request.URL.Path, "/remix") {
+		relayMode := relayconstant.RelayModeVideoSubmit
+		c.Set("relay_mode", relayMode)
+		shouldSelectChannel = false
 	} else if strings.Contains(c.Request.URL.Path, "/v1/videos") {
 		//curl https://api.openai.com/v1/videos \
 		//  -H "Authorization: Bearer $OPENAI_API_KEY" \

relay/channel/adapter.go

@@ -47,7 +47,7 @@ type TaskAdaptor interface {
 	GetChannelName() string
 
 	// FetchTask
-	FetchTask(baseUrl, key string, body map[string]any) (*http.Response, error)
+	FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error)
 
 	ParseTaskResult(respBody []byte) (*relaycommon.TaskInfo, error)
 }

relay/channel/claude/constants.go

@@ -9,6 +9,7 @@ var ModelList = []string{
 	"claude-3-opus-20240229",
 	"claude-3-haiku-20240307",
 	"claude-3-5-haiku-20241022",
+	"claude-haiku-4-5-20251001",
 	"claude-3-5-sonnet-20240620",
 	"claude-3-5-sonnet-20241022",
 	"claude-3-7-sonnet-20250219",

relay/channel/claude/relay-claude.go

@@ -673,7 +673,7 @@ func HandleStreamResponseData(c *gin.Context, info *relaycommon.RelayInfo, claud
 func HandleStreamFinalResponse(c *gin.Context, info *relaycommon.RelayInfo, claudeInfo *ClaudeResponseInfo, requestMode int) {
 
 	if requestMode == RequestModeCompletion {
-		claudeInfo.Usage = service.ResponseText2Usage(c, claudeInfo.ResponseText.String(), info.UpstreamModelName, info.PromptTokens)
+		claudeInfo.Usage = service.ResponseText2Usage(c, claudeInfo.ResponseText.String(), info.UpstreamModelName, info.GetEstimatePromptTokens())
 	} else {
 		if claudeInfo.Usage.PromptTokens == 0 {
 			//上游出错
@@ -734,10 +734,7 @@ func HandleClaudeResponseData(c *gin.Context, info *relaycommon.RelayInfo, claud
 		return types.WithClaudeError(*claudeError, http.StatusInternalServerError)
 	}
 	if requestMode == RequestModeCompletion {
-		completionTokens := service.CountTextToken(claudeResponse.Completion, info.OriginModelName)
-		claudeInfo.Usage.PromptTokens = info.PromptTokens
-		claudeInfo.Usage.CompletionTokens = completionTokens
-		claudeInfo.Usage.TotalTokens = info.PromptTokens + completionTokens
+		claudeInfo.Usage = service.ResponseText2Usage(c, claudeResponse.Completion, info.UpstreamModelName, info.GetEstimatePromptTokens())
 	} else {
 		claudeInfo.Usage.PromptTokens = claudeResponse.Usage.InputTokens
 		claudeInfo.Usage.CompletionTokens = claudeResponse.Usage.OutputTokens

relay/channel/cloudflare/relay_cloudflare.go

@@ -74,7 +74,7 @@ func cfStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Res
 	if err := scanner.Err(); err != nil {
 		logger.LogError(c, "error_scanning_stream_response: "+err.Error())
 	}
-	usage := service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.PromptTokens)
+	usage := service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.GetEstimatePromptTokens())
 	if info.ShouldIncludeUsage {
 		response := helper.GenerateFinalUsageResponse(id, info.StartTime.Unix(), info.UpstreamModelName, *usage)
 		err := helper.ObjectData(c, response)
@@ -105,7 +105,7 @@ func cfHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response)
 	for _, choice := range response.Choices {
 		responseText += choice.Message.StringContent()
 	}
-	usage := service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.PromptTokens)
+	usage := service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.GetEstimatePromptTokens())
 	response.Usage = *usage
 	response.Id = helper.GetResponseID(c)
 	jsonResponse, err := json.Marshal(response)
@@ -142,10 +142,6 @@ func cfSTTHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Respon
 	c.Writer.WriteHeader(resp.StatusCode)
 	_, _ = c.Writer.Write(jsonResponse)
 
-	usage := &dto.Usage{}
-	usage.PromptTokens = info.PromptTokens
-	usage.CompletionTokens = service.CountTextToken(cfResp.Result.Text, info.UpstreamModelName)
-	usage.TotalTokens = usage.PromptTokens + usage.CompletionTokens
-
+	usage := service.ResponseText2Usage(c, cfResp.Result.Text, info.UpstreamModelName, info.GetEstimatePromptTokens())
 	return nil, usage
 }

relay/channel/cohere/relay-cohere.go

@@ -165,7 +165,7 @@ func cohereStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http
 		}
 	})
 	if usage.PromptTokens == 0 {
-		usage = service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.PromptTokens)
+		usage = service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.GetEstimatePromptTokens())
 	}
 	return usage, nil
 }
@@ -225,9 +225,9 @@ func cohereRerankHandler(c *gin.Context, resp *http.Response, info *relaycommon.
 	}
 	usage := dto.Usage{}
 	if cohereResp.Meta.BilledUnits.InputTokens == 0 {
-		usage.PromptTokens = info.PromptTokens
+		usage.PromptTokens = info.GetEstimatePromptTokens()
 		usage.CompletionTokens = 0
-		usage.TotalTokens = info.PromptTokens
+		usage.TotalTokens = info.GetEstimatePromptTokens()
 	} else {
 		usage.PromptTokens = cohereResp.Meta.BilledUnits.InputTokens
 		usage.CompletionTokens = cohereResp.Meta.BilledUnits.OutputTokens

relay/channel/dify/relay-dify.go

@@ -246,7 +246,7 @@ func difyStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.R
 	})
 	helper.Done(c)
 	if usage.TotalTokens == 0 {
-		usage = service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.PromptTokens)
+		usage = service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.GetEstimatePromptTokens())
 	}
 	usage.CompletionTokens += nodeToken
 	return usage, nil

relay/channel/gemini/adaptor.go

@@ -137,6 +137,8 @@ func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
 			info.UpstreamModelName = strings.TrimSuffix(info.UpstreamModelName, "-thinking")
 		} else if strings.HasSuffix(info.UpstreamModelName, "-nothinking") {
 			info.UpstreamModelName = strings.TrimSuffix(info.UpstreamModelName, "-nothinking")
+		} else if baseModel, level := parseThinkingLevelSuffix(info.UpstreamModelName); level != "" {
+			info.UpstreamModelName = baseModel
 		}
 	}
 

relay/channel/gemini/relay-gemini-native.go

@@ -5,7 +5,6 @@ import (
 	"net/http"
 
 	"github.com/QuantumNous/new-api/common"
-	"github.com/QuantumNous/new-api/constant"
 	"github.com/QuantumNous/new-api/dto"
 	"github.com/QuantumNous/new-api/logger"
 	relaycommon "github.com/QuantumNous/new-api/relay/common"
@@ -70,12 +69,7 @@ func NativeGeminiEmbeddingHandler(c *gin.Context, resp *http.Response, info *rel
 		println(string(responseBody))
 	}
 
-	usage := &dto.Usage{
-		PromptTokens: info.PromptTokens,
-		TotalTokens:  info.PromptTokens,
-	}
-
-	common.SetContextKey(c, constant.ContextKeyLocalCountTokens, true)
+	usage := service.ResponseText2Usage(c, "", info.UpstreamModelName, info.GetEstimatePromptTokens())
 
 	if info.IsGeminiBatchEmbedding {
 		var geminiResponse dto.GeminiBatchEmbeddingResponse

relay/channel/gemini/relay-gemini.go

@@ -19,8 +19,8 @@ import (
 	"github.com/QuantumNous/new-api/relay/helper"
 	"github.com/QuantumNous/new-api/service"
 	"github.com/QuantumNous/new-api/setting/model_setting"
+	"github.com/QuantumNous/new-api/setting/reasoning"
 	"github.com/QuantumNous/new-api/types"
-
 	"github.com/gin-gonic/gin"
 )
 
@@ -122,6 +122,14 @@ func clampThinkingBudgetByEffort(modelName string, effort string) int {
 	return clampThinkingBudget(modelName, maxBudget)
 }
 
+func parseThinkingLevelSuffix(modelName string) (string, string) {
+	base, level, ok := reasoning.TrimEffortSuffix(modelName)
+	if !ok {
+		return modelName, ""
+	}
+	return base, level
+}
+
 func ThinkingAdaptor(geminiRequest *dto.GeminiChatRequest, info *relaycommon.RelayInfo, oaiRequest ...dto.GeneralOpenAIRequest) {
 	if model_setting.GetGeminiSettings().ThinkingAdapterEnabled {
 		modelName := info.UpstreamModelName
@@ -178,6 +186,12 @@ func ThinkingAdaptor(geminiRequest *dto.GeminiChatRequest, info *relaycommon.Rel
 					ThinkingBudget: common.GetPointer(0),
 				}
 			}
+		} else if _, level := parseThinkingLevelSuffix(modelName); level != "" {
+			geminiRequest.GenerationConfig.ThinkingConfig = &dto.GeminiThinkingConfig{
+				IncludeThoughts: true,
+				ThinkingLevel:   level,
+			}
+			info.ReasoningEffort = level
 		}
 	}
 }
@@ -1115,7 +1129,7 @@ func geminiStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http
 	if usage.CompletionTokens <= 0 {
 		str := responseText.String()
 		if len(str) > 0 {
-			usage = service.ResponseText2Usage(c, responseText.String(), info.UpstreamModelName, info.PromptTokens)
+			usage = service.ResponseText2Usage(c, responseText.String(), info.UpstreamModelName, info.GetEstimatePromptTokens())
 		} else {
 			usage = &dto.Usage{}
 		}
@@ -1288,11 +1302,7 @@ func GeminiEmbeddingHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *h
 	// Google has not yet clarified how embedding models will be billed
 	// refer to openai billing method to use input tokens billing
 	// https://platform.openai.com/docs/guides/embeddings#what-are-embeddings
-	usage := &dto.Usage{
-		PromptTokens:     info.PromptTokens,
-		CompletionTokens: 0,
-		TotalTokens:      info.PromptTokens,
-	}
+	usage := service.ResponseText2Usage(c, "", info.UpstreamModelName, info.GetEstimatePromptTokens())
 	openAIResponse.Usage = *usage
 
 	jsonResponse, jsonErr := common.Marshal(openAIResponse)

relay/channel/minimax/tts.go

@@ -163,7 +163,7 @@ func handleTTSResponse(c *gin.Context, resp *http.Response, info *relaycommon.Re
 	}
 
 	usage = &dto.Usage{
-		PromptTokens:     info.PromptTokens,
+		PromptTokens:     info.GetEstimatePromptTokens(),
 		CompletionTokens: 0,
 		TotalTokens:      int(minimaxResp.ExtraInfo.UsageCharacters),
 	}

relay/channel/openai/adaptor.go

@@ -306,10 +306,11 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
 			request.Temperature = nil
 		}
 
+		// gpt-5系列模型适配 归零不再支持的参数
 		if strings.HasPrefix(info.UpstreamModelName, "gpt-5") {
-			if info.UpstreamModelName != "gpt-5-chat-latest" {
-				request.Temperature = nil
-			}
+			request.Temperature = nil
+			request.TopP = 0 // oai 的 top_p 默认值是 1.0，但是为了 omitempty 属性直接不传，这里显式设置为 0
+			request.LogProbs = false
 		}
 
 		// 转换模型推理力度后缀

relay/channel/openai/relay-openai.go

@@ -183,7 +183,7 @@ func OaiStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Re
 	}
 
 	if !containStreamUsage {
-		usage = service.ResponseText2Usage(c, responseTextBuilder.String(), info.UpstreamModelName, info.PromptTokens)
+		usage = service.ResponseText2Usage(c, responseTextBuilder.String(), info.UpstreamModelName, info.GetEstimatePromptTokens())
 		usage.CompletionTokens += toolCount * 7
 	}
 
@@ -245,9 +245,9 @@ func OpenaiHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Respo
 			}
 		}
 		simpleResponse.Usage = dto.Usage{
-			PromptTokens:     info.PromptTokens,
+			PromptTokens:     info.GetEstimatePromptTokens(),
 			CompletionTokens: completionTokens,
-			TotalTokens:      info.PromptTokens + completionTokens,
+			TotalTokens:      info.GetEstimatePromptTokens() + completionTokens,
 		}
 		usageModified = true
 	}
@@ -336,8 +336,8 @@ func OpenaiTTSHandler(c *gin.Context, resp *http.Response, info *relaycommon.Rel
 	// and can be terminated directly.
 	defer service.CloseResponseBodyGracefully(resp)
 	usage := &dto.Usage{}
-	usage.PromptTokens = info.PromptTokens
-	usage.TotalTokens = info.PromptTokens
+	usage.PromptTokens = info.GetEstimatePromptTokens()
+	usage.TotalTokens = info.GetEstimatePromptTokens()
 	for k, v := range resp.Header {
 		c.Writer.Header().Set(k, v[0])
 	}
@@ -383,7 +383,7 @@ func OpenaiSTTHandler(c *gin.Context, resp *http.Response, info *relaycommon.Rel
 	}
 
 	usage := &dto.Usage{}
-	usage.PromptTokens = info.PromptTokens
+	usage.PromptTokens = info.GetEstimatePromptTokens()
 	usage.CompletionTokens = 0
 	usage.TotalTokens = usage.PromptTokens + usage.CompletionTokens
 	return nil, usage

relay/channel/openai/relay_responses.go

@@ -141,7 +141,7 @@ func OaiResponsesStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp
 	}
 
 	if usage.PromptTokens == 0 && usage.CompletionTokens != 0 {
-		usage.PromptTokens = info.PromptTokens
+		usage.PromptTokens = info.GetEstimatePromptTokens()
 	}
 
 	usage.TotalTokens = usage.PromptTokens + usage.CompletionTokens

relay/channel/palm/adaptor.go

@@ -81,7 +81,7 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
 	if info.IsStream {
 		var responseText string
 		err, responseText = palmStreamHandler(c, resp)
-		usage = service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.PromptTokens)
+		usage = service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.GetEstimatePromptTokens())
 	} else {
 		usage, err = palmHandler(c, info, resp)
 	}

relay/channel/palm/relay-palm.go

@@ -121,13 +121,8 @@ func palmHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Respons
 		}, resp.StatusCode)
 	}
 	fullTextResponse := responsePaLM2OpenAI(&palmResponse)
-	completionTokens := service.CountTextToken(palmResponse.Candidates[0].Content, info.UpstreamModelName)
-	usage := dto.Usage{
-		PromptTokens:     info.PromptTokens,
-		CompletionTokens: completionTokens,
-		TotalTokens:      info.PromptTokens + completionTokens,
-	}
-	fullTextResponse.Usage = usage
+	usage := service.ResponseText2Usage(c, palmResponse.Candidates[0].Content, info.UpstreamModelName, info.GetEstimatePromptTokens())
+	fullTextResponse.Usage = *usage
 	jsonResponse, err := common.Marshal(fullTextResponse)
 	if err != nil {
 		return nil, types.NewError(err, types.ErrorCodeBadResponseBody)
@@ -135,5 +130,5 @@ func palmHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Respons
 	c.Writer.Header().Set("Content-Type", "application/json")
 	c.Writer.WriteHeader(resp.StatusCode)
 	service.IOCopyBytesGracefully(c, resp, jsonResponse)
-	return &usage, nil
+	return usage, nil
 }

relay/channel/task/ali/adaptor.go

@@ -393,7 +393,7 @@ func (a *TaskAdaptor) DoResponse(c *gin.Context, resp *http.Response, info *rela
 }
 
 // FetchTask 查询任务状态
-func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http.Response, error) {
+func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
 	taskID, ok := body["task_id"].(string)
 	if !ok {
 		return nil, fmt.Errorf("invalid task_id")
@@ -408,7 +408,11 @@ func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http
 
 	req.Header.Set("Authorization", "Bearer "+key)
 
-	return service.GetHttpClient().Do(req)
+	client, err := service.GetHttpClientWithProxy(proxy)
+	if err != nil {
+		return nil, fmt.Errorf("new proxy http client failed: %w", err)
+	}
+	return client.Do(req)
 }
 
 func (a *TaskAdaptor) GetModelList() []string {

relay/channel/task/doubao/adaptor.go

@@ -146,7 +146,7 @@ func (a *TaskAdaptor) DoResponse(c *gin.Context, resp *http.Response, info *rela
 }
 
 // FetchTask fetch task status
-func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http.Response, error) {
+func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
 	taskID, ok := body["task_id"].(string)
 	if !ok {
 		return nil, fmt.Errorf("invalid task_id")
@@ -163,7 +163,11 @@ func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http
 	req.Header.Set("Content-Type", "application/json")
 	req.Header.Set("Authorization", "Bearer "+key)
 
-	return service.GetHttpClient().Do(req)
+	client, err := service.GetHttpClientWithProxy(proxy)
+	if err != nil {
+		return nil, fmt.Errorf("new proxy http client failed: %w", err)
+	}
+	return client.Do(req)
 }
 
 func (a *TaskAdaptor) GetModelList() []string {

relay/channel/task/gemini/adaptor.go

@@ -200,7 +200,7 @@ func (a *TaskAdaptor) GetChannelName() string {
 }
 
 // FetchTask fetch task status
-func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http.Response, error) {
+func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
 	taskID, ok := body["task_id"].(string)
 	if !ok {
 		return nil, fmt.Errorf("invalid task_id")
@@ -223,7 +223,11 @@ func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http
 	req.Header.Set("Accept", "application/json")
 	req.Header.Set("x-goog-api-key", key)
 
-	return service.GetHttpClient().Do(req)
+	client, err := service.GetHttpClientWithProxy(proxy)
+	if err != nil {
+		return nil, fmt.Errorf("new proxy http client failed: %w", err)
+	}
+	return client.Do(req)
 }
 
 func (a *TaskAdaptor) ParseTaskResult(respBody []byte) (*relaycommon.TaskInfo, error) {

relay/channel/task/hailuo/adaptor.go

@@ -110,7 +110,7 @@ func (a *TaskAdaptor) DoResponse(c *gin.Context, resp *http.Response, info *rela
 	return hResp.TaskID, responseBody, nil
 }
 
-func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http.Response, error) {
+func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
 	taskID, ok := body["task_id"].(string)
 	if !ok {
 		return nil, fmt.Errorf("invalid task_id")
@@ -126,7 +126,11 @@ func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http
 	req.Header.Set("Accept", "application/json")
 	req.Header.Set("Authorization", "Bearer "+key)
 
-	return service.GetHttpClient().Do(req)
+	client, err := service.GetHttpClientWithProxy(proxy)
+	if err != nil {
+		return nil, fmt.Errorf("new proxy http client failed: %w", err)
+	}
+	return client.Do(req)
 }
 
 func (a *TaskAdaptor) GetModelList() []string {

relay/channel/task/jimeng/adaptor.go

@@ -210,7 +210,7 @@ func (a *TaskAdaptor) DoResponse(c *gin.Context, resp *http.Response, info *rela
 }
 
 // FetchTask fetch task status
-func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http.Response, error) {
+func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
 	taskID, ok := body["task_id"].(string)
 	if !ok {
 		return nil, fmt.Errorf("invalid task_id")
@@ -251,7 +251,11 @@ func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http
 			return nil, errors.Wrap(err, "sign request failed")
 		}
 	}
-	return service.GetHttpClient().Do(req)
+	client, err := service.GetHttpClientWithProxy(proxy)
+	if err != nil {
+		return nil, fmt.Errorf("new proxy http client failed: %w", err)
+	}
+	return client.Do(req)
 }
 
 func (a *TaskAdaptor) GetModelList() []string {

relay/channel/task/kling/adaptor.go

@@ -199,7 +199,7 @@ func (a *TaskAdaptor) DoResponse(c *gin.Context, resp *http.Response, info *rela
 }
 
 // FetchTask fetch task status
-func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http.Response, error) {
+func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
 	taskID, ok := body["task_id"].(string)
 	if !ok {
 		return nil, fmt.Errorf("invalid task_id")
@@ -228,7 +228,11 @@ func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http
 	req.Header.Set("Authorization", "Bearer "+token)
 	req.Header.Set("User-Agent", "kling-sdk/1.0")
 
-	return service.GetHttpClient().Do(req)
+	client, err := service.GetHttpClientWithProxy(proxy)
+	if err != nil {
+		return nil, fmt.Errorf("new proxy http client failed: %w", err)
+	}
+	return client.Do(req)
 }
 
 func (a *TaskAdaptor) GetModelList() []string {

relay/channel/task/sora/adaptor.go

@@ -5,8 +5,10 @@ import (
 	"fmt"
 	"io"
 	"net/http"
+	"strings"
 
 	"github.com/QuantumNous/new-api/common"
+	"github.com/QuantumNous/new-api/constant"
 	"github.com/QuantumNous/new-api/dto"
 	"github.com/QuantumNous/new-api/model"
 	"github.com/QuantumNous/new-api/relay/channel"
@@ -67,11 +69,30 @@ func (a *TaskAdaptor) Init(info *relaycommon.RelayInfo) {
 	a.apiKey = info.ApiKey
 }
 
+func validateRemixRequest(c *gin.Context) *dto.TaskError {
+	var req struct {
+		Prompt string `json:"prompt"`
+	}
+	if err := common.UnmarshalBodyReusable(c, &req); err != nil {
+		return service.TaskErrorWrapperLocal(err, "invalid_request", http.StatusBadRequest)
+	}
+	if strings.TrimSpace(req.Prompt) == "" {
+		return service.TaskErrorWrapperLocal(fmt.Errorf("field prompt is required"), "invalid_request", http.StatusBadRequest)
+	}
+	return nil
+}
+
 func (a *TaskAdaptor) ValidateRequestAndSetAction(c *gin.Context, info *relaycommon.RelayInfo) (taskErr *dto.TaskError) {
+	if info.Action == constant.TaskActionRemix {
+		return validateRemixRequest(c)
+	}
 	return relaycommon.ValidateMultipartDirect(c, info)
 }
 
 func (a *TaskAdaptor) BuildRequestURL(info *relaycommon.RelayInfo) (string, error) {
+	if info.Action == constant.TaskActionRemix {
+		return fmt.Sprintf("%s/v1/videos/%s/remix", a.baseURL, info.OriginTaskID), nil
+	}
 	return fmt.Sprintf("%s/v1/videos", a.baseURL), nil
 }
 
@@ -125,7 +146,7 @@ func (a *TaskAdaptor) DoResponse(c *gin.Context, resp *http.Response, _ *relayco
 }
 
 // FetchTask fetch task status
-func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http.Response, error) {
+func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
 	taskID, ok := body["task_id"].(string)
 	if !ok {
 		return nil, fmt.Errorf("invalid task_id")
@@ -140,7 +161,11 @@ func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http
 
 	req.Header.Set("Authorization", "Bearer "+key)
 
-	return service.GetHttpClient().Do(req)
+	client, err := service.GetHttpClientWithProxy(proxy)
+	if err != nil {
+		return nil, fmt.Errorf("new proxy http client failed: %w", err)
+	}
+	return client.Do(req)
 }
 
 func (a *TaskAdaptor) GetModelList() []string {

relay/channel/task/suno/adaptor.go

@@ -132,7 +132,7 @@ func (a *TaskAdaptor) GetChannelName() string {
 	return ChannelName
 }
 
-func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http.Response, error) {
+func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
 	requestUrl := fmt.Sprintf("%s/suno/fetch", baseUrl)
 	byteBody, err := json.Marshal(body)
 	if err != nil {
@@ -153,11 +153,11 @@ func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http
 	req = req.WithContext(ctx)
 	req.Header.Set("Content-Type", "application/json")
 	req.Header.Set("Authorization", "Bearer "+key)
-	resp, err := service.GetHttpClient().Do(req)
+	client, err := service.GetHttpClientWithProxy(proxy)
 	if err != nil {
-		return nil, err
+		return nil, fmt.Errorf("new proxy http client failed: %w", err)
 	}
-	return resp, nil
+	return client.Do(req)
 }
 
 func actionValidate(c *gin.Context, sunoRequest *dto.SunoSubmitReq, action string) (err error) {

relay/channel/task/vertex/adaptor.go

@@ -120,7 +120,11 @@ func (a *TaskAdaptor) BuildRequestHeader(c *gin.Context, req *http.Request, info
 		return fmt.Errorf("failed to decode credentials: %w", err)
 	}
 
-	token, err := vertexcore.AcquireAccessToken(*adc, "")
+	proxy := ""
+	if info != nil {
+		proxy = info.ChannelSetting.Proxy
+	}
+	token, err := vertexcore.AcquireAccessToken(*adc, proxy)
 	if err != nil {
 		return fmt.Errorf("failed to acquire access token: %w", err)
 	}
@@ -216,7 +220,7 @@ func (a *TaskAdaptor) GetModelList() []string { return []string{"veo-3.0-generat
 func (a *TaskAdaptor) GetChannelName() string { return "vertex" }
 
 // FetchTask fetch task status
-func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http.Response, error) {
+func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
 	taskID, ok := body["task_id"].(string)
 	if !ok {
 		return nil, fmt.Errorf("invalid task_id")
@@ -249,7 +253,7 @@ func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http
 	if err := json.Unmarshal([]byte(key), adc); err != nil {
 		return nil, fmt.Errorf("failed to decode credentials: %w", err)
 	}
-	token, err := vertexcore.AcquireAccessToken(*adc, "")
+	token, err := vertexcore.AcquireAccessToken(*adc, proxy)
 	if err != nil {
 		return nil, fmt.Errorf("failed to acquire access token: %w", err)
 	}
@@ -261,7 +265,11 @@ func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http
 	req.Header.Set("Accept", "application/json")
 	req.Header.Set("Authorization", "Bearer "+token)
 	req.Header.Set("x-goog-user-project", adc.ProjectID)
-	return service.GetHttpClient().Do(req)
+	client, err := service.GetHttpClientWithProxy(proxy)
+	if err != nil {
+		return nil, fmt.Errorf("new proxy http client failed: %w", err)
+	}
+	return client.Do(req)
 }
 
 func (a *TaskAdaptor) ParseTaskResult(respBody []byte) (*relaycommon.TaskInfo, error) {

relay/channel/task/vidu/adaptor.go

@@ -188,7 +188,7 @@ func (a *TaskAdaptor) DoResponse(c *gin.Context, resp *http.Response, info *rela
 	return vResp.TaskId, responseBody, nil
 }
 
-func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http.Response, error) {
+func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
 	taskID, ok := body["task_id"].(string)
 	if !ok {
 		return nil, fmt.Errorf("invalid task_id")
@@ -204,7 +204,11 @@ func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http
 	req.Header.Set("Accept", "application/json")
 	req.Header.Set("Authorization", "Token "+key)
 
-	return service.GetHttpClient().Do(req)
+	client, err := service.GetHttpClientWithProxy(proxy)
+	if err != nil {
+		return nil, fmt.Errorf("new proxy http client failed: %w", err)
+	}
+	return client.Do(req)
 }
 
 func (a *TaskAdaptor) GetModelList() []string {

relay/channel/tencent/relay-tencent.go

@@ -105,7 +105,7 @@ func tencentStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *htt
 		data = strings.TrimPrefix(data, "data:")
 
 		var tencentResponse TencentChatResponse
-		err := json.Unmarshal([]byte(data), &tencentResponse)
+		err := common.Unmarshal([]byte(data), &tencentResponse)
 		if err != nil {
 			common.SysLog("error unmarshalling stream response: " + err.Error())
 			continue
@@ -130,7 +130,7 @@ func tencentStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *htt
 
 	service.CloseResponseBodyGracefully(resp)
 
-	return service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.PromptTokens), nil
+	return service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.GetEstimatePromptTokens()), nil
 }
 
 func tencentHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.Usage, *types.NewAPIError) {

relay/channel/vertex/adaptor.go

@@ -17,6 +17,7 @@ import (
 	relaycommon "github.com/QuantumNous/new-api/relay/common"
 	"github.com/QuantumNous/new-api/relay/constant"
 	"github.com/QuantumNous/new-api/setting/model_setting"
+	"github.com/QuantumNous/new-api/setting/reasoning"
 	"github.com/QuantumNous/new-api/types"
 
 	"github.com/gin-gonic/gin"
@@ -181,6 +182,8 @@ func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
 				info.UpstreamModelName = strings.TrimSuffix(info.UpstreamModelName, "-thinking")
 			} else if strings.HasSuffix(info.UpstreamModelName, "-nothinking") {
 				info.UpstreamModelName = strings.TrimSuffix(info.UpstreamModelName, "-nothinking")
+			} else if baseModel, level, ok := reasoning.TrimEffortSuffix(info.UpstreamModelName); ok && level != "" {
+				info.UpstreamModelName = baseModel
 			}
 		}
 

relay/channel/volcengine/tts.go

@@ -184,9 +184,9 @@ func handleTTSResponse(c *gin.Context, resp *http.Response, info *relaycommon.Re
 	c.Data(http.StatusOK, contentType, audioData)
 
 	usage = &dto.Usage{
-		PromptTokens:     info.PromptTokens,
+		PromptTokens:     info.GetEstimatePromptTokens(),
 		CompletionTokens: 0,
-		TotalTokens:      info.PromptTokens,
+		TotalTokens:      info.GetEstimatePromptTokens(),
 	}
 
 	return usage, nil
@@ -284,9 +284,9 @@ func handleTTSWebSocketResponse(c *gin.Context, requestURL string, volcRequest V
 			if msg.Sequence < 0 {
 				c.Status(http.StatusOK)
 				usage = &dto.Usage{
-					PromptTokens:     info.PromptTokens,
+					PromptTokens:     info.GetEstimatePromptTokens(),
 					CompletionTokens: 0,
-					TotalTokens:      info.PromptTokens,
+					TotalTokens:      info.GetEstimatePromptTokens(),
 				}
 				return usage, nil
 			}
@@ -297,9 +297,9 @@ func handleTTSWebSocketResponse(c *gin.Context, requestURL string, volcRequest V
 
 	c.Status(http.StatusOK)
 	usage = &dto.Usage{
-		PromptTokens:     info.PromptTokens,
+		PromptTokens:     info.GetEstimatePromptTokens(),
 		CompletionTokens: 0,
-		TotalTokens:      info.PromptTokens,
+		TotalTokens:      info.GetEstimatePromptTokens(),
 	}
 	return usage, nil
 }

relay/channel/xai/text.go

@@ -70,7 +70,7 @@ func xAIStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Re
 	})
 
 	if !containStreamUsage {
-		usage = service.ResponseText2Usage(c, responseTextBuilder.String(), info.UpstreamModelName, info.PromptTokens)
+		usage = service.ResponseText2Usage(c, responseTextBuilder.String(), info.UpstreamModelName, info.GetEstimatePromptTokens())
 		usage.CompletionTokens += toolCount * 7
 	}
 

relay/channel/zhipu_4v/adaptor.go

@@ -36,8 +36,7 @@ func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInf
 }
 
 func (a *Adaptor) ConvertImageRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.ImageRequest) (any, error) {
-	//TODO implement me
-	return nil, errors.New("not implemented")
+	return request, nil
 }
 
 func (a *Adaptor) Init(info *relaycommon.RelayInfo) {
@@ -63,6 +62,8 @@ func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
 				return fmt.Sprintf("%s/embeddings", specialPlan.OpenAIBaseURL), nil
 			}
 			return fmt.Sprintf("%s/api/paas/v4/embeddings", baseURL), nil
+		case relayconstant.RelayModeImagesGenerations:
+			return fmt.Sprintf("%s/api/paas/v4/images/generations", baseURL), nil
 		default:
 			if hasSpecialPlan && specialPlan.OpenAIBaseURL != "" {
 				return fmt.Sprintf("%s/chat/completions", specialPlan.OpenAIBaseURL), nil
@@ -114,6 +115,9 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
 			return claude.ClaudeHandler(c, resp, info, claude.RequestModeMessage)
 		}
 	default:
+		if info.RelayMode == relayconstant.RelayModeImagesGenerations {
+			return zhipu4vImageHandler(c, resp, info)
+		}
 		adaptor := openai.Adaptor{}
 		return adaptor.DoResponse(c, resp, info)
 	}

relay/channel/zhipu_4v/image.go

@@ -0,0 +1,127 @@
+package zhipu_4v
+
+import (
+	"io"
+	"net/http"
+
+	"github.com/QuantumNous/new-api/common"
+	"github.com/QuantumNous/new-api/dto"
+	"github.com/QuantumNous/new-api/logger"
+	relaycommon "github.com/QuantumNous/new-api/relay/common"
+	"github.com/QuantumNous/new-api/service"
+	"github.com/QuantumNous/new-api/types"
+
+	"github.com/gin-gonic/gin"
+)
+
+type zhipuImageRequest struct {
+	Model            string `json:"model"`
+	Prompt           string `json:"prompt"`
+	Quality          string `json:"quality,omitempty"`
+	Size             string `json:"size,omitempty"`
+	WatermarkEnabled *bool  `json:"watermark_enabled,omitempty"`
+	UserID           string `json:"user_id,omitempty"`
+}
+
+type zhipuImageResponse struct {
+	Created       *int64            `json:"created,omitempty"`
+	Data          []zhipuImageData  `json:"data,omitempty"`
+	ContentFilter any               `json:"content_filter,omitempty"`
+	Usage         *dto.Usage        `json:"usage,omitempty"`
+	Error         *zhipuImageError  `json:"error,omitempty"`
+	RequestID     string            `json:"request_id,omitempty"`
+	ExtendParam   map[string]string `json:"extendParam,omitempty"`
+}
+
+type zhipuImageError struct {
+	Code    string `json:"code"`
+	Message string `json:"message"`
+}
+
+type zhipuImageData struct {
+	Url      string `json:"url,omitempty"`
+	ImageUrl string `json:"image_url,omitempty"`
+	B64Json  string `json:"b64_json,omitempty"`
+	B64Image string `json:"b64_image,omitempty"`
+}
+
+type openAIImagePayload struct {
+	Created int64             `json:"created"`
+	Data    []openAIImageData `json:"data"`
+}
+
+type openAIImageData struct {
+	B64Json string `json:"b64_json"`
+}
+
+func zhipu4vImageHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.Usage, *types.NewAPIError) {
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return nil, types.NewOpenAIError(err, types.ErrorCodeReadResponseBodyFailed, http.StatusInternalServerError)
+	}
+	service.CloseResponseBodyGracefully(resp)
+
+	var zhipuResp zhipuImageResponse
+	if err := common.Unmarshal(responseBody, &zhipuResp); err != nil {
+		return nil, types.NewOpenAIError(err, types.ErrorCodeBadResponseBody, http.StatusInternalServerError)
+	}
+
+	if zhipuResp.Error != nil && zhipuResp.Error.Message != "" {
+		return nil, types.WithOpenAIError(types.OpenAIError{
+			Message: zhipuResp.Error.Message,
+			Type:    "zhipu_image_error",
+			Code:    zhipuResp.Error.Code,
+		}, resp.StatusCode)
+	}
+
+	payload := openAIImagePayload{}
+	if zhipuResp.Created != nil && *zhipuResp.Created != 0 {
+		payload.Created = *zhipuResp.Created
+	} else {
+		payload.Created = info.StartTime.Unix()
+	}
+	for _, data := range zhipuResp.Data {
+		url := data.Url
+		if url == "" {
+			url = data.ImageUrl
+		}
+		if url == "" {
+			logger.LogWarn(c, "zhipu_image_missing_url")
+			continue
+		}
+
+		var b64 string
+		switch {
+		case data.B64Json != "":
+			b64 = data.B64Json
+		case data.B64Image != "":
+			b64 = data.B64Image
+		default:
+			_, downloaded, err := service.GetImageFromUrl(url)
+			if err != nil {
+				logger.LogError(c, "zhipu_image_get_b64_failed: "+err.Error())
+				continue
+			}
+			b64 = downloaded
+		}
+
+		if b64 == "" {
+			logger.LogWarn(c, "zhipu_image_empty_b64")
+			continue
+		}
+
+		imageData := openAIImageData{
+			B64Json: b64,
+		}
+		payload.Data = append(payload.Data, imageData)
+	}
+
+	jsonResp, err := common.Marshal(payload)
+	if err != nil {
+		return nil, types.NewError(err, types.ErrorCodeBadResponseBody)
+	}
+
+	service.IOCopyBytesGracefully(c, resp, jsonResp)
+
+	return &dto.Usage{}, nil
+}

relay/common/override.go

@@ -11,6 +11,8 @@ import (
 	"github.com/tidwall/sjson"
 )
 
+var negativeIndexRegexp = regexp.MustCompile(`\.(-\d+)`)
+
 type ConditionOperation struct {
 	Path           string      `json:"path"`             // JSON路径
 	Mode           string      `json:"mode"`             // full, prefix, suffix, contains, gt, gte, lt, lte
@@ -186,8 +188,7 @@ func checkSingleCondition(jsonStr, contextJSON string, condition ConditionOperat
 }
 
 func processNegativeIndex(jsonStr string, path string) string {
-	re := regexp.MustCompile(`\.(-\d+)`)
-	matches := re.FindAllStringSubmatch(path, -1)
+	matches := negativeIndexRegexp.FindAllStringSubmatch(path, -1)
 
 	if len(matches) == 0 {
 		return path

relay/common/relay_info.go

@@ -73,6 +73,11 @@ type ChannelMeta struct {
 	SupportStreamOptions bool // 是否支持流式选项
 }
 
+type TokenCountMeta struct {
+	//promptTokens int
+	estimatePromptTokens int
+}
+
 type RelayInfo struct {
 	TokenId           int
 	TokenKey          string
@@ -91,7 +96,6 @@ type RelayInfo struct {
 	RelayMode              int
 	OriginModelName        string
 	RequestURLPath         string
-	PromptTokens           int
 	ShouldIncludeUsage     bool
 	DisablePing            bool // 是否禁止向下游发送自定义 Ping
 	ClientWs               *websocket.Conn
@@ -115,6 +119,7 @@ type RelayInfo struct {
 	Request dto.Request
 
 	ThinkingContentInfo
+	TokenCountMeta
 	*ClaudeConvertInfo
 	*RerankerInfo
 	*ResponsesUsageInfo
@@ -189,7 +194,7 @@ func (info *RelayInfo) ToString() string {
 	fmt.Fprintf(b, "IsPlayground: %t, ", info.IsPlayground)
 	fmt.Fprintf(b, "RequestURLPath: %q, ", info.RequestURLPath)
 	fmt.Fprintf(b, "OriginModelName: %q, ", info.OriginModelName)
-	fmt.Fprintf(b, "PromptTokens: %d, ", info.PromptTokens)
+	fmt.Fprintf(b, "EstimatePromptTokens: %d, ", info.estimatePromptTokens)
 	fmt.Fprintf(b, "ShouldIncludeUsage: %t, ", info.ShouldIncludeUsage)
 	fmt.Fprintf(b, "DisablePing: %t, ", info.DisablePing)
 	fmt.Fprintf(b, "SendResponseCount: %d, ", info.SendResponseCount)
@@ -391,7 +396,6 @@ func genBaseRelayInfo(c *gin.Context, request dto.Request) *RelayInfo {
 		UserEmail:  common.GetContextKeyString(c, constant.ContextKeyUserEmail),
 
 		OriginModelName: common.GetContextKeyString(c, constant.ContextKeyOriginalModel),
-		PromptTokens:    common.GetContextKeyInt(c, constant.ContextKeyPromptTokens),
 
 		TokenId:        common.GetContextKeyInt(c, constant.ContextKeyTokenId),
 		TokenKey:       common.GetContextKeyString(c, constant.ContextKeyTokenKey),
@@ -408,6 +412,10 @@ func genBaseRelayInfo(c *gin.Context, request dto.Request) *RelayInfo {
 			IsFirstThinkingContent:  true,
 			SendLastThinkingContent: false,
 		},
+		TokenCountMeta: TokenCountMeta{
+			//promptTokens: common.GetContextKeyInt(c, constant.ContextKeyPromptTokens),
+			estimatePromptTokens: common.GetContextKeyInt(c, constant.ContextKeyEstimatedTokens),
+		},
 	}
 
 	if info.RelayMode == relayconstant.RelayModeUnknown {
@@ -463,8 +471,16 @@ func GenRelayInfo(c *gin.Context, relayFormat types.RelayFormat, request dto.Req
 	}
 }
 
-func (info *RelayInfo) SetPromptTokens(promptTokens int) {
-	info.PromptTokens = promptTokens
+//func (info *RelayInfo) SetPromptTokens(promptTokens int) {
+//	info.promptTokens = promptTokens
+//}
+
+func (info *RelayInfo) SetEstimatePromptTokens(promptTokens int) {
+	info.estimatePromptTokens = promptTokens
+}
+
+func (info *RelayInfo) GetEstimatePromptTokens() int {
+	return info.estimatePromptTokens
 }
 
 func (info *RelayInfo) SetFirstResponseTime() {

relay/common_handler/rerank.go

@@ -57,8 +57,8 @@ func RerankHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Respo
 		jinaResp = dto.RerankResponse{
 			Results: jinaRespResults,
 			Usage: dto.Usage{
-				PromptTokens: info.PromptTokens,
-				TotalTokens:  info.PromptTokens,
+				PromptTokens: info.GetEstimatePromptTokens(),
+				TotalTokens:  info.GetEstimatePromptTokens(),
 			},
 		}
 	} else {

relay/compatible_handler.go

@@ -192,9 +192,9 @@ func TextHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *types
 func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usage *dto.Usage, extraContent string) {
 	if usage == nil {
 		usage = &dto.Usage{
-			PromptTokens:     relayInfo.PromptTokens,
+			PromptTokens:     relayInfo.GetEstimatePromptTokens(),
 			CompletionTokens: 0,
-			TotalTokens:      relayInfo.PromptTokens,
+			TotalTokens:      relayInfo.GetEstimatePromptTokens(),
 		}
 		extraContent += "（可能是请求出错）"
 	}

relay/helper/price.go

@@ -99,7 +99,10 @@ func ModelPriceHelper(c *gin.Context, info *relaycommon.RelayInfo, promptTokens
 	// check if free model pre-consume is disabled
 	if !operation_setting.GetQuotaSetting().EnableFreeModelPreConsume {
 		// if model price or ratio is 0, do not pre-consume quota
-		if usePrice {
+		if groupRatioInfo.GroupRatio == 0 {
+			preConsumedQuota = 0
+			freeModel = true
+		} else if usePrice {
 			if modelPrice == 0 {
 				preConsumedQuota = 0
 				freeModel = true

relay/helper/stream_scanner.go

@@ -72,6 +72,8 @@ func StreamScannerHandler(c *gin.Context, resp *http.Response, info *relaycommon
 	if common.DebugEnabled {
 		// print timeout and ping interval for debugging
 		println("relay timeout seconds:", common.RelayTimeout)
+		println("relay max idle conns:", common.RelayMaxIdleConns)
+		println("relay max idle conns per host:", common.RelayMaxIdleConnsPerHost)
 		println("streaming timeout seconds:", int64(streamingTimeout.Seconds()))
 		println("ping interval seconds:", int64(pingInterval.Seconds()))
 	}

relay/relay_task.go

@@ -32,7 +32,94 @@ func RelayTaskSubmit(c *gin.Context, info *relaycommon.RelayInfo) (taskErr *dto.
 	if info.TaskRelayInfo == nil {
 		info.TaskRelayInfo = &relaycommon.TaskRelayInfo{}
 	}
+	path := c.Request.URL.Path
+	if strings.Contains(path, "/v1/videos/") && strings.HasSuffix(path, "/remix") {
+		info.Action = constant.TaskActionRemix
+	}
+
+	// 提取 remix 任务的 video_id
+	if info.Action == constant.TaskActionRemix {
+		videoID := c.Param("video_id")
+		if strings.TrimSpace(videoID) == "" {
+			return service.TaskErrorWrapperLocal(fmt.Errorf("video_id is required"), "invalid_request", http.StatusBadRequest)
+		}
+		info.OriginTaskID = videoID
+	}
+
 	platform := constant.TaskPlatform(c.GetString("platform"))
+
+	// 获取原始任务信息
+	if info.OriginTaskID != "" {
+		originTask, exist, err := model.GetByTaskId(info.UserId, info.OriginTaskID)
+		if err != nil {
+			taskErr = service.TaskErrorWrapper(err, "get_origin_task_failed", http.StatusInternalServerError)
+			return
+		}
+		if !exist {
+			taskErr = service.TaskErrorWrapperLocal(errors.New("task_origin_not_exist"), "task_not_exist", http.StatusBadRequest)
+			return
+		}
+		if info.OriginModelName == "" {
+			if originTask.Properties.OriginModelName != "" {
+				info.OriginModelName = originTask.Properties.OriginModelName
+			} else if originTask.Properties.UpstreamModelName != "" {
+				info.OriginModelName = originTask.Properties.UpstreamModelName
+			} else {
+				var taskData map[string]interface{}
+				_ = json.Unmarshal(originTask.Data, &taskData)
+				if m, ok := taskData["model"].(string); ok && m != "" {
+					info.OriginModelName = m
+					platform = originTask.Platform
+				}
+			}
+		}
+		if originTask.ChannelId != info.ChannelId {
+			channel, err := model.GetChannelById(originTask.ChannelId, true)
+			if err != nil {
+				taskErr = service.TaskErrorWrapperLocal(err, "channel_not_found", http.StatusBadRequest)
+				return
+			}
+			if channel.Status != common.ChannelStatusEnabled {
+				taskErr = service.TaskErrorWrapperLocal(errors.New("the channel of the origin task is disabled"), "task_channel_disable", http.StatusBadRequest)
+				return
+			}
+			key, _, newAPIError := channel.GetNextEnabledKey()
+			if newAPIError != nil {
+				taskErr = service.TaskErrorWrapper(newAPIError, "channel_no_available_key", newAPIError.StatusCode)
+				return
+			}
+			common.SetContextKey(c, constant.ContextKeyChannelKey, key)
+			common.SetContextKey(c, constant.ContextKeyChannelType, channel.Type)
+			common.SetContextKey(c, constant.ContextKeyChannelBaseUrl, channel.GetBaseURL())
+			common.SetContextKey(c, constant.ContextKeyChannelId, originTask.ChannelId)
+
+			info.ChannelBaseUrl = channel.GetBaseURL()
+			info.ChannelId = originTask.ChannelId
+			info.ChannelType = channel.Type
+			info.ApiKey = key
+			platform = originTask.Platform
+		}
+
+		// 使用原始任务的参数
+		if info.Action == constant.TaskActionRemix {
+			var taskData map[string]interface{}
+			_ = json.Unmarshal(originTask.Data, &taskData)
+			secondsStr, _ := taskData["seconds"].(string)
+			seconds, _ := strconv.Atoi(secondsStr)
+			if seconds <= 0 {
+				seconds = 4
+			}
+			sizeStr, _ := taskData["size"].(string)
+			if info.PriceData.OtherRatios == nil {
+				info.PriceData.OtherRatios = map[string]float64{}
+			}
+			info.PriceData.OtherRatios["seconds"] = float64(seconds)
+			info.PriceData.OtherRatios["size"] = 1
+			if sizeStr == "1792x1024" || sizeStr == "1024x1792" {
+				info.PriceData.OtherRatios["size"] = 1.666667
+			}
+		}
+	}
 	if platform == "" {
 		platform = GetTaskPlatform(c)
 	}
@@ -94,34 +181,6 @@ func RelayTaskSubmit(c *gin.Context, info *relaycommon.RelayInfo) (taskErr *dto.
 		return
 	}
 
-	if info.OriginTaskID != "" {
-		originTask, exist, err := model.GetByTaskId(info.UserId, info.OriginTaskID)
-		if err != nil {
-			taskErr = service.TaskErrorWrapper(err, "get_origin_task_failed", http.StatusInternalServerError)
-			return
-		}
-		if !exist {
-			taskErr = service.TaskErrorWrapperLocal(errors.New("task_origin_not_exist"), "task_not_exist", http.StatusBadRequest)
-			return
-		}
-		if originTask.ChannelId != info.ChannelId {
-			channel, err := model.GetChannelById(originTask.ChannelId, true)
-			if err != nil {
-				taskErr = service.TaskErrorWrapperLocal(err, "channel_not_found", http.StatusBadRequest)
-				return
-			}
-			if channel.Status != common.ChannelStatusEnabled {
-				return service.TaskErrorWrapperLocal(errors.New("该任务所属渠道已被禁用"), "task_channel_disable", http.StatusBadRequest)
-			}
-			c.Set("base_url", channel.GetBaseURL())
-			c.Set("channel_id", originTask.ChannelId)
-			c.Request.Header.Set("Authorization", fmt.Sprintf("Bearer %s", channel.Key))
-
-			info.ChannelBaseUrl = channel.GetBaseURL()
-			info.ChannelId = originTask.ChannelId
-		}
-	}
-
 	// build body
 	requestBody, err := adaptor.BuildRequestBody(c, info)
 	if err != nil {
@@ -326,6 +385,7 @@ func videoFetchByIDRespBodyBuilder(c *gin.Context) (respBody []byte, taskResp *d
 		if channelModel.GetBaseURL() != "" {
 			baseURL = channelModel.GetBaseURL()
 		}
+		proxy := channelModel.GetSetting().Proxy
 		adaptor := GetTaskAdaptor(constant.TaskPlatform(strconv.Itoa(channelModel.Type)))
 		if adaptor == nil {
 			return
@@ -333,7 +393,7 @@ func videoFetchByIDRespBodyBuilder(c *gin.Context) (respBody []byte, taskResp *d
 		resp, err2 := adaptor.FetchTask(baseURL, channelModel.Key, map[string]any{
 			"task_id": originTask.TaskID,
 			"action":  originTask.Action,
-		})
+		}, proxy)
 		if err2 != nil || resp == nil {
 			return
 		}

router/video-router.go

@@ -14,6 +14,7 @@ func SetVideoRouter(router *gin.Engine) {
 		videoV1Router.GET("/videos/:task_id/content", controller.VideoProxy)
 		videoV1Router.POST("/video/generations", controller.RelayTask)
 		videoV1Router.GET("/video/generations/:task_id", controller.RelayTask)
+		videoV1Router.POST("/videos/:video_id/remix", controller.RelayTask)
 	}
 	// openai compatible API video routes
 	// docs: https://platform.openai.com/docs/api-reference/videos/create

service/convert.go

@@ -201,6 +201,10 @@ func generateStopBlock(index int) *dto.ClaudeResponse {
 }
 
 func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamResponse, info *relaycommon.RelayInfo) []*dto.ClaudeResponse {
+	if info.ClaudeConvertInfo.Done {
+		return nil
+	}
+
 	var claudeResponses []*dto.ClaudeResponse
 	if info.SendResponseCount == 1 {
 		msg := &dto.ClaudeMediaMessage{
@@ -209,7 +213,7 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon
 			Type:  "message",
 			Role:  "assistant",
 			Usage: &dto.ClaudeUsage{
-				InputTokens:  info.PromptTokens,
+				InputTokens:  info.GetEstimatePromptTokens(),
 				OutputTokens: 0,
 			},
 		}
@@ -218,45 +222,117 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon
 			Type:    "message_start",
 			Message: msg,
 		})
-		claudeResponses = append(claudeResponses)
 		//claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
 		//	Type: "ping",
 		//})
 		if openAIResponse.IsToolCall() {
 			info.ClaudeConvertInfo.LastMessagesType = relaycommon.LastMessageTypeTools
+			var toolCall dto.ToolCallResponse
+			if len(openAIResponse.Choices) > 0 && len(openAIResponse.Choices[0].Delta.ToolCalls) > 0 {
+				toolCall = openAIResponse.Choices[0].Delta.ToolCalls[0]
+			} else {
+				first := openAIResponse.GetFirstToolCall()
+				if first != nil {
+					toolCall = *first
+				} else {
+					toolCall = dto.ToolCallResponse{}
+				}
+			}
 			resp := &dto.ClaudeResponse{
 				Type: "content_block_start",
 				ContentBlock: &dto.ClaudeMediaMessage{
-					Id:    openAIResponse.GetFirstToolCall().ID,
+					Id:    toolCall.ID,
 					Type:  "tool_use",
-					Name:  openAIResponse.GetFirstToolCall().Function.Name,
+					Name:  toolCall.Function.Name,
 					Input: map[string]interface{}{},
 				},
 			}
 			resp.SetIndex(0)
 			claudeResponses = append(claudeResponses, resp)
+			// 首块包含工具 delta，则追加 input_json_delta
+			if toolCall.Function.Arguments != "" {
+				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
+					Index: &info.ClaudeConvertInfo.Index,
+					Type:  "content_block_delta",
+					Delta: &dto.ClaudeMediaMessage{
+						Type:        "input_json_delta",
+						PartialJson: &toolCall.Function.Arguments,
+					},
+				})
+			}
 		} else {
 
 		}
 		// 判断首个响应是否存在内容（非标准的 OpenAI 响应）
-		if len(openAIResponse.Choices) > 0 && len(openAIResponse.Choices[0].Delta.GetContentString()) > 0 {
+		if len(openAIResponse.Choices) > 0 {
+			reasoning := openAIResponse.Choices[0].Delta.GetReasoningContent()
+			content := openAIResponse.Choices[0].Delta.GetContentString()
+
+			if reasoning != "" {
+				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
+					Index: &info.ClaudeConvertInfo.Index,
+					Type:  "content_block_start",
+					ContentBlock: &dto.ClaudeMediaMessage{
+						Type:     "thinking",
+						Thinking: common.GetPointer[string](""),
+					},
+				})
+				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
+					Index: &info.ClaudeConvertInfo.Index,
+					Type:  "content_block_delta",
+					Delta: &dto.ClaudeMediaMessage{
+						Type:     "thinking_delta",
+						Thinking: &reasoning,
+					},
+				})
+				info.ClaudeConvertInfo.LastMessagesType = relaycommon.LastMessageTypeThinking
+			} else if content != "" {
+				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
+					Index: &info.ClaudeConvertInfo.Index,
+					Type:  "content_block_start",
+					ContentBlock: &dto.ClaudeMediaMessage{
+						Type: "text",
+						Text: common.GetPointer[string](""),
+					},
+				})
+				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
+					Index: &info.ClaudeConvertInfo.Index,
+					Type:  "content_block_delta",
+					Delta: &dto.ClaudeMediaMessage{
+						Type: "text_delta",
+						Text: common.GetPointer[string](content),
+					},
+				})
+				info.ClaudeConvertInfo.LastMessagesType = relaycommon.LastMessageTypeText
+			}
+		}
+
+		// 如果首块就带 finish_reason，需要立即发送停止块
+		if len(openAIResponse.Choices) > 0 && openAIResponse.Choices[0].FinishReason != nil && *openAIResponse.Choices[0].FinishReason != "" {
+			info.FinishReason = *openAIResponse.Choices[0].FinishReason
+			claudeResponses = append(claudeResponses, generateStopBlock(info.ClaudeConvertInfo.Index))
+			oaiUsage := openAIResponse.Usage
+			if oaiUsage == nil {
+				oaiUsage = info.ClaudeConvertInfo.Usage
+			}
+			if oaiUsage != nil {
+				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
+					Type: "message_delta",
+					Usage: &dto.ClaudeUsage{
+						InputTokens:              oaiUsage.PromptTokens,
+						OutputTokens:             oaiUsage.CompletionTokens,
+						CacheCreationInputTokens: oaiUsage.PromptTokensDetails.CachedCreationTokens,
+						CacheReadInputTokens:     oaiUsage.PromptTokensDetails.CachedTokens,
+					},
+					Delta: &dto.ClaudeMediaMessage{
+						StopReason: common.GetPointer[string](stopReasonOpenAI2Claude(info.FinishReason)),
+					},
+				})
+			}
 			claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
-				Index: &info.ClaudeConvertInfo.Index,
-				Type:  "content_block_start",
-				ContentBlock: &dto.ClaudeMediaMessage{
-					Type: "text",
-					Text: common.GetPointer[string](""),
-				},
+				Type: "message_stop",
 			})
-			claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
-				Index: &info.ClaudeConvertInfo.Index,
-				Type:  "content_block_delta",
-				Delta: &dto.ClaudeMediaMessage{
-					Type: "text_delta",
-					Text: common.GetPointer[string](openAIResponse.Choices[0].Delta.GetContentString()),
-				},
-			})
-			info.ClaudeConvertInfo.LastMessagesType = relaycommon.LastMessageTypeText
+			info.ClaudeConvertInfo.Done = true
 		}
 		return claudeResponses
 	}
@@ -264,7 +340,7 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon
 	if len(openAIResponse.Choices) == 0 {
 		// no choices
 		// 可能为非标准的 OpenAI 响应，判断是否已经完成
-		if info.Done {
+		if info.ClaudeConvertInfo.Done {
 			claudeResponses = append(claudeResponses, generateStopBlock(info.ClaudeConvertInfo.Index))
 			oaiUsage := info.ClaudeConvertInfo.Usage
 			if oaiUsage != nil {
@@ -288,16 +364,110 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon
 		return claudeResponses
 	} else {
 		chosenChoice := openAIResponse.Choices[0]
-		if chosenChoice.FinishReason != nil && *chosenChoice.FinishReason != "" {
-			// should be done
+		doneChunk := chosenChoice.FinishReason != nil && *chosenChoice.FinishReason != ""
+		if doneChunk {
 			info.FinishReason = *chosenChoice.FinishReason
-			if !info.Done {
-				return claudeResponses
+		}
+
+		var claudeResponse dto.ClaudeResponse
+		var isEmpty bool
+		claudeResponse.Type = "content_block_delta"
+		if len(chosenChoice.Delta.ToolCalls) > 0 {
+			toolCalls := chosenChoice.Delta.ToolCalls
+			if info.ClaudeConvertInfo.LastMessagesType != relaycommon.LastMessageTypeTools {
+				claudeResponses = append(claudeResponses, generateStopBlock(info.ClaudeConvertInfo.Index))
+				info.ClaudeConvertInfo.Index++
+			}
+			info.ClaudeConvertInfo.LastMessagesType = relaycommon.LastMessageTypeTools
+
+			for i, toolCall := range toolCalls {
+				blockIndex := info.ClaudeConvertInfo.Index
+				if toolCall.Index != nil {
+					blockIndex = *toolCall.Index
+				} else if len(toolCalls) > 1 {
+					blockIndex = info.ClaudeConvertInfo.Index + i
+				}
+
+				idx := blockIndex
+				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
+					Index: &idx,
+					Type:  "content_block_start",
+					ContentBlock: &dto.ClaudeMediaMessage{
+						Id:    toolCall.ID,
+						Type:  "tool_use",
+						Name:  toolCall.Function.Name,
+						Input: map[string]interface{}{},
+					},
+				})
+
+				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
+					Index: &idx,
+					Type:  "content_block_delta",
+					Delta: &dto.ClaudeMediaMessage{
+						Type:        "input_json_delta",
+						PartialJson: &toolCall.Function.Arguments,
+					},
+				})
+
+				info.ClaudeConvertInfo.Index = blockIndex
+			}
+		} else {
+			reasoning := chosenChoice.Delta.GetReasoningContent()
+			textContent := chosenChoice.Delta.GetContentString()
+			if reasoning != "" || textContent != "" {
+				if reasoning != "" {
+					if info.ClaudeConvertInfo.LastMessagesType != relaycommon.LastMessageTypeThinking {
+						claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
+							Index: &info.ClaudeConvertInfo.Index,
+							Type:  "content_block_start",
+							ContentBlock: &dto.ClaudeMediaMessage{
+								Type:     "thinking",
+								Thinking: common.GetPointer[string](""),
+							},
+						})
+					}
+					info.ClaudeConvertInfo.LastMessagesType = relaycommon.LastMessageTypeThinking
+					claudeResponse.Delta = &dto.ClaudeMediaMessage{
+						Type:     "thinking_delta",
+						Thinking: &reasoning,
+					}
+				} else {
+					if info.ClaudeConvertInfo.LastMessagesType != relaycommon.LastMessageTypeText {
+						if info.ClaudeConvertInfo.LastMessagesType == relaycommon.LastMessageTypeThinking || info.ClaudeConvertInfo.LastMessagesType == relaycommon.LastMessageTypeTools {
+							claudeResponses = append(claudeResponses, generateStopBlock(info.ClaudeConvertInfo.Index))
+							info.ClaudeConvertInfo.Index++
+						}
+						claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
+							Index: &info.ClaudeConvertInfo.Index,
+							Type:  "content_block_start",
+							ContentBlock: &dto.ClaudeMediaMessage{
+								Type: "text",
+								Text: common.GetPointer[string](""),
+							},
+						})
+					}
+					info.ClaudeConvertInfo.LastMessagesType = relaycommon.LastMessageTypeText
+					claudeResponse.Delta = &dto.ClaudeMediaMessage{
+						Type: "text_delta",
+						Text: common.GetPointer[string](textContent),
+					}
+				}
+			} else {
+				isEmpty = true
 			}
 		}
-		if info.Done {
+
+		claudeResponse.Index = &info.ClaudeConvertInfo.Index
+		if !isEmpty && claudeResponse.Delta != nil {
+			claudeResponses = append(claudeResponses, &claudeResponse)
+		}
+
+		if doneChunk || info.ClaudeConvertInfo.Done {
 			claudeResponses = append(claudeResponses, generateStopBlock(info.ClaudeConvertInfo.Index))
-			oaiUsage := info.ClaudeConvertInfo.Usage
+			oaiUsage := openAIResponse.Usage
+			if oaiUsage == nil {
+				oaiUsage = info.ClaudeConvertInfo.Usage
+			}
 			if oaiUsage != nil {
 				claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
 					Type: "message_delta",
@@ -315,83 +485,8 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon
 			claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
 				Type: "message_stop",
 			})
-		} else {
-			var claudeResponse dto.ClaudeResponse
-			var isEmpty bool
-			claudeResponse.Type = "content_block_delta"
-			if len(chosenChoice.Delta.ToolCalls) > 0 {
-				if info.ClaudeConvertInfo.LastMessagesType != relaycommon.LastMessageTypeTools {
-					claudeResponses = append(claudeResponses, generateStopBlock(info.ClaudeConvertInfo.Index))
-					info.ClaudeConvertInfo.Index++
-					claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
-						Index: &info.ClaudeConvertInfo.Index,
-						Type:  "content_block_start",
-						ContentBlock: &dto.ClaudeMediaMessage{
-							Id:    openAIResponse.GetFirstToolCall().ID,
-							Type:  "tool_use",
-							Name:  openAIResponse.GetFirstToolCall().Function.Name,
-							Input: map[string]interface{}{},
-						},
-					})
-				}
-				info.ClaudeConvertInfo.LastMessagesType = relaycommon.LastMessageTypeTools
-				// tools delta
-				claudeResponse.Delta = &dto.ClaudeMediaMessage{
-					Type:        "input_json_delta",
-					PartialJson: &chosenChoice.Delta.ToolCalls[0].Function.Arguments,
-				}
-			} else {
-				reasoning := chosenChoice.Delta.GetReasoningContent()
-				textContent := chosenChoice.Delta.GetContentString()
-				if reasoning != "" || textContent != "" {
-					if reasoning != "" {
-						if info.ClaudeConvertInfo.LastMessagesType != relaycommon.LastMessageTypeThinking {
-							//info.ClaudeConvertInfo.Index++
-							claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
-								Index: &info.ClaudeConvertInfo.Index,
-								Type:  "content_block_start",
-								ContentBlock: &dto.ClaudeMediaMessage{
-									Type:     "thinking",
-									Thinking: common.GetPointer[string](""),
-								},
-							})
-						}
-						info.ClaudeConvertInfo.LastMessagesType = relaycommon.LastMessageTypeThinking
-						// text delta
-						claudeResponse.Delta = &dto.ClaudeMediaMessage{
-							Type:     "thinking_delta",
-							Thinking: &reasoning,
-						}
-					} else {
-						if info.ClaudeConvertInfo.LastMessagesType != relaycommon.LastMessageTypeText {
-							if info.LastMessagesType == relaycommon.LastMessageTypeThinking || info.LastMessagesType == relaycommon.LastMessageTypeTools {
-								claudeResponses = append(claudeResponses, generateStopBlock(info.ClaudeConvertInfo.Index))
-								info.ClaudeConvertInfo.Index++
-							}
-							claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
-								Index: &info.ClaudeConvertInfo.Index,
-								Type:  "content_block_start",
-								ContentBlock: &dto.ClaudeMediaMessage{
-									Type: "text",
-									Text: common.GetPointer[string](""),
-								},
-							})
-						}
-						info.ClaudeConvertInfo.LastMessagesType = relaycommon.LastMessageTypeText
-						// text delta
-						claudeResponse.Delta = &dto.ClaudeMediaMessage{
-							Type: "text_delta",
-							Text: common.GetPointer[string](textContent),
-						}
-					}
-				} else {
-					isEmpty = true
-				}
-			}
-			claudeResponse.Index = &info.ClaudeConvertInfo.Index
-			if !isEmpty {
-				claudeResponses = append(claudeResponses, &claudeResponse)
-			}
+			info.ClaudeConvertInfo.Done = true
+			return claudeResponses
 		}
 	}
 
@@ -734,12 +829,18 @@ func StreamResponseOpenAI2Gemini(openAIResponse *dto.ChatCompletionsStreamRespon
 	geminiResponse := &dto.GeminiChatResponse{
 		Candidates: make([]dto.GeminiChatCandidate, 0, len(openAIResponse.Choices)),
 		UsageMetadata: dto.GeminiUsageMetadata{
-			PromptTokenCount:     info.PromptTokens,
+			PromptTokenCount:     info.GetEstimatePromptTokens(),
 			CandidatesTokenCount: 0, // 流式响应中可能没有完整的 usage 信息
-			TotalTokenCount:      info.PromptTokens,
+			TotalTokenCount:      info.GetEstimatePromptTokens(),
 		},
 	}
 
+	if openAIResponse.Usage != nil {
+		geminiResponse.UsageMetadata.PromptTokenCount = openAIResponse.Usage.PromptTokens
+		geminiResponse.UsageMetadata.CandidatesTokenCount = openAIResponse.Usage.CompletionTokens
+		geminiResponse.UsageMetadata.TotalTokenCount = openAIResponse.Usage.TotalTokens
+	}
+
 	for _, choice := range openAIResponse.Choices {
 		candidate := dto.GeminiChatCandidate{
 			Index:         int64(choice.Index),

service/http_client.go

@@ -34,12 +34,20 @@ func checkRedirect(req *http.Request, via []*http.Request) error {
 }
 
 func InitHttpClient() {
+	transport := &http.Transport{
+		MaxIdleConns:        common.RelayMaxIdleConns,
+		MaxIdleConnsPerHost: common.RelayMaxIdleConnsPerHost,
+		ForceAttemptHTTP2:   true,
+	}
+
 	if common.RelayTimeout == 0 {
 		httpClient = &http.Client{
+			Transport:     transport,
 			CheckRedirect: checkRedirect,
 		}
 	} else {
 		httpClient = &http.Client{
+			Transport:     transport,
 			Timeout:       time.Duration(common.RelayTimeout) * time.Second,
 			CheckRedirect: checkRedirect,
 		}
@@ -50,6 +58,14 @@ func GetHttpClient() *http.Client {
 	return httpClient
 }
 
+// GetHttpClientWithProxy returns the default client or a proxy-enabled one when proxyURL is provided.
+func GetHttpClientWithProxy(proxyURL string) (*http.Client, error) {
+	if proxyURL == "" {
+		return GetHttpClient(), nil
+	}
+	return NewProxyHttpClient(proxyURL)
+}
+
 // ResetProxyClientCache 清空代理客户端缓存，确保下次使用时重新初始化
 func ResetProxyClientCache() {
 	proxyClientLock.Lock()
@@ -84,7 +100,10 @@ func NewProxyHttpClient(proxyURL string) (*http.Client, error) {
 	case "http", "https":
 		client := &http.Client{
 			Transport: &http.Transport{
-				Proxy: http.ProxyURL(parsedURL),
+				MaxIdleConns:        common.RelayMaxIdleConns,
+				MaxIdleConnsPerHost: common.RelayMaxIdleConnsPerHost,
+				ForceAttemptHTTP2:   true,
+				Proxy:               http.ProxyURL(parsedURL),
 			},
 			CheckRedirect: checkRedirect,
 		}
@@ -116,6 +135,9 @@ func NewProxyHttpClient(proxyURL string) (*http.Client, error) {
 
 		client := &http.Client{
 			Transport: &http.Transport{
+				MaxIdleConns:        common.RelayMaxIdleConns,
+				MaxIdleConnsPerHost: common.RelayMaxIdleConnsPerHost,
+				ForceAttemptHTTP2:   true,
 				DialContext: func(ctx context.Context, network, addr string) (net.Conn, error) {
 					return dialer.Dial(network, addr)
 				},

service/token_counter.go

@@ -1,7 +1,6 @@
 package service
 
 import (
-	"encoding/json"
 	"errors"
 	"fmt"
 	"image"
@@ -12,7 +11,6 @@ import (
 	"math"
 	"path/filepath"
 	"strings"
-	"sync"
 	"unicode/utf8"
 
 	"github.com/QuantumNous/new-api/common"
@@ -23,64 +21,8 @@ import (
 	"github.com/QuantumNous/new-api/types"
 
 	"github.com/gin-gonic/gin"
-	"github.com/tiktoken-go/tokenizer"
-	"github.com/tiktoken-go/tokenizer/codec"
 )
 
-// tokenEncoderMap won't grow after initialization
-var defaultTokenEncoder tokenizer.Codec
-
-// tokenEncoderMap is used to store token encoders for different models
-var tokenEncoderMap = make(map[string]tokenizer.Codec)
-
-// tokenEncoderMutex protects tokenEncoderMap for concurrent access
-var tokenEncoderMutex sync.RWMutex
-
-func InitTokenEncoders() {
-	common.SysLog("initializing token encoders")
-	defaultTokenEncoder = codec.NewCl100kBase()
-	common.SysLog("token encoders initialized")
-}
-
-func getTokenEncoder(model string) tokenizer.Codec {
-	// First, try to get the encoder from cache with read lock
-	tokenEncoderMutex.RLock()
-	if encoder, exists := tokenEncoderMap[model]; exists {
-		tokenEncoderMutex.RUnlock()
-		return encoder
-	}
-	tokenEncoderMutex.RUnlock()
-
-	// If not in cache, create new encoder with write lock
-	tokenEncoderMutex.Lock()
-	defer tokenEncoderMutex.Unlock()
-
-	// Double-check if another goroutine already created the encoder
-	if encoder, exists := tokenEncoderMap[model]; exists {
-		return encoder
-	}
-
-	// Create new encoder
-	modelCodec, err := tokenizer.ForModel(tokenizer.Model(model))
-	if err != nil {
-		// Cache the default encoder for this model to avoid repeated failures
-		tokenEncoderMap[model] = defaultTokenEncoder
-		return defaultTokenEncoder
-	}
-
-	// Cache the new encoder
-	tokenEncoderMap[model] = modelCodec
-	return modelCodec
-}
-
-func getTokenNum(tokenEncoder tokenizer.Codec, text string) int {
-	if text == "" {
-		return 0
-	}
-	tkm, _ := tokenEncoder.Count(text)
-	return tkm
-}
-
 func getImageToken(fileMeta *types.FileMeta, model string, stream bool) (int, error) {
 	if fileMeta == nil {
 		return 0, fmt.Errorf("image_url_is_nil")
@@ -257,7 +199,7 @@ func getImageToken(fileMeta *types.FileMeta, model string, stream bool) (int, er
 	return tiles*tileTokens + baseTokens, nil
 }
 
-func CountRequestToken(c *gin.Context, meta *types.TokenCountMeta, info *relaycommon.RelayInfo) (int, error) {
+func EstimateRequestToken(c *gin.Context, meta *types.TokenCountMeta, info *relaycommon.RelayInfo) (int, error) {
 	// 是否统计token
 	if !constant.CountToken {
 		return 0, nil
@@ -375,14 +317,14 @@ func CountRequestToken(c *gin.Context, meta *types.TokenCountMeta, info *relayco
 	for i, file := range meta.Files {
 		switch file.FileType {
 		case types.FileTypeImage:
-			if info.RelayFormat == types.RelayFormatGemini {
-				tkm += 520 // gemini per input image tokens
-			} else {
+			if common.IsOpenAITextModel(info.OriginModelName) {
 				token, err := getImageToken(file, model, info.IsStream)
 				if err != nil {
 					return 0, fmt.Errorf("error counting image token, media index[%d], original data[%s], err: %v", i, file.OriginData, err)
 				}
 				tkm += token
+			} else {
+				tkm += 520
 			}
 		case types.FileTypeAudio:
 			tkm += 256
@@ -399,111 +341,6 @@ func CountRequestToken(c *gin.Context, meta *types.TokenCountMeta, info *relayco
 	return tkm, nil
 }
 
-func CountTokenClaudeRequest(request dto.ClaudeRequest, model string) (int, error) {
-	tkm := 0
-
-	// Count tokens in messages
-	msgTokens, err := CountTokenClaudeMessages(request.Messages, model, request.Stream)
-	if err != nil {
-		return 0, err
-	}
-	tkm += msgTokens
-
-	// Count tokens in system message
-	if request.System != "" {
-		systemTokens := CountTokenInput(request.System, model)
-		tkm += systemTokens
-	}
-
-	if request.Tools != nil {
-		// check is array
-		if tools, ok := request.Tools.([]any); ok {
-			if len(tools) > 0 {
-				parsedTools, err1 := common.Any2Type[[]dto.Tool](request.Tools)
-				if err1 != nil {
-					return 0, fmt.Errorf("tools: Input should be a valid list: %v", err)
-				}
-				toolTokens, err2 := CountTokenClaudeTools(parsedTools, model)
-				if err2 != nil {
-					return 0, fmt.Errorf("tools: %v", err)
-				}
-				tkm += toolTokens
-			}
-		} else {
-			return 0, errors.New("tools: Input should be a valid list")
-		}
-	}
-
-	return tkm, nil
-}
-
-func CountTokenClaudeMessages(messages []dto.ClaudeMessage, model string, stream bool) (int, error) {
-	tokenEncoder := getTokenEncoder(model)
-	tokenNum := 0
-
-	for _, message := range messages {
-		// Count tokens for role
-		tokenNum += getTokenNum(tokenEncoder, message.Role)
-		if message.IsStringContent() {
-			tokenNum += getTokenNum(tokenEncoder, message.GetStringContent())
-		} else {
-			content, err := message.ParseContent()
-			if err != nil {
-				return 0, err
-			}
-			for _, mediaMessage := range content {
-				switch mediaMessage.Type {
-				case "text":
-					tokenNum += getTokenNum(tokenEncoder, mediaMessage.GetText())
-				case "image":
-					//imageTokenNum, err := getClaudeImageToken(mediaMsg.Source, model, stream)
-					//if err != nil {
-					//	return 0, err
-					//}
-					tokenNum += 1000
-				case "tool_use":
-					if mediaMessage.Input != nil {
-						tokenNum += getTokenNum(tokenEncoder, mediaMessage.Name)
-						inputJSON, _ := json.Marshal(mediaMessage.Input)
-						tokenNum += getTokenNum(tokenEncoder, string(inputJSON))
-					}
-				case "tool_result":
-					if mediaMessage.Content != nil {
-						contentJSON, _ := json.Marshal(mediaMessage.Content)
-						tokenNum += getTokenNum(tokenEncoder, string(contentJSON))
-					}
-				}
-			}
-		}
-	}
-
-	// Add a constant for message formatting (this may need adjustment based on Claude's exact formatting)
-	tokenNum += len(messages) * 2 // Assuming 2 tokens per message for formatting
-
-	return tokenNum, nil
-}
-
-func CountTokenClaudeTools(tools []dto.Tool, model string) (int, error) {
-	tokenEncoder := getTokenEncoder(model)
-	tokenNum := 0
-
-	for _, tool := range tools {
-		tokenNum += getTokenNum(tokenEncoder, tool.Name)
-		tokenNum += getTokenNum(tokenEncoder, tool.Description)
-
-		schemaJSON, err := json.Marshal(tool.InputSchema)
-		if err != nil {
-			return 0, errors.New(fmt.Sprintf("marshal_tool_schema_fail: %s", err.Error()))
-		}
-		tokenNum += getTokenNum(tokenEncoder, string(schemaJSON))
-	}
-
-	// Add a constant for tool formatting (this may need adjustment based on Claude's exact formatting)
-	tokenNum += len(tools) * 3 // Assuming 3 tokens per tool for formatting
-
-	return tokenNum, nil
-}
-
 func CountTokenRealtime(info *relaycommon.RelayInfo, request dto.RealtimeEvent, model string) (int, int, error) {
 	audioToken := 0
 	textToken := 0
@@ -578,31 +415,6 @@ func CountTokenInput(input any, model string) int {
 	return CountTokenInput(fmt.Sprintf("%v", input), model)
 }
 
-func CountTokenStreamChoices(messages []dto.ChatCompletionsStreamResponseChoice, model string) int {
-	tokens := 0
-	for _, message := range messages {
-		tkm := CountTokenInput(message.Delta.GetContentString(), model)
-		tokens += tkm
-		if message.Delta.ToolCalls != nil {
-			for _, tool := range message.Delta.ToolCalls {
-				tkm := CountTokenInput(tool.Function.Name, model)
-				tokens += tkm
-				tkm = CountTokenInput(tool.Function.Arguments, model)
-				tokens += tkm
-			}
-		}
-	}
-	return tokens
-}
-
-func CountTTSToken(text string, model string) int {
-	if strings.HasPrefix(model, "tts") {
-		return utf8.RuneCountInString(text)
-	} else {
-		return CountTextToken(text, model)
-	}
-}
-
 func CountAudioTokenInput(audioBase64 string, audioFormat string) (int, error) {
 	if audioBase64 == "" {
 		return 0, nil
@@ -625,17 +437,16 @@ func CountAudioTokenOutput(audioBase64 string, audioFormat string) (int, error)
 	return int(duration / 60 * 200 / 0.24), nil
 }
 
-//func CountAudioToken(sec float64, audioType string) {
-//	if audioType == "input" {
-//
-//	}
-//}
-
-// CountTextToken 统计文本的token数量，仅当文本包含敏感词，返回错误，同时返回token数量
+// CountTextToken 统计文本的token数量，仅OpenAI模型使用tokenizer，其余模型使用估算
 func CountTextToken(text string, model string) int {
 	if text == "" {
 		return 0
 	}
-	tokenEncoder := getTokenEncoder(model)
-	return getTokenNum(tokenEncoder, text)
+	if common.IsOpenAITextModel(model) {
+		tokenEncoder := getTokenEncoder(model)
+		return getTokenNum(tokenEncoder, text)
+	} else {
+		// 非openai模型，使用tiktoken-go计算没有意义，使用估算节省资源
+		return EstimateTokenByModel(model, text)
+	}
 }

service/token_estimator.go

@@ -0,0 +1,230 @@
+package service
+
+import (
+	"math"
+	"strings"
+	"sync"
+	"unicode"
+)
+
+// Provider 定义模型厂商大类
+type Provider string
+
+const (
+	OpenAI  Provider = "openai"  // 代表 GPT-3.5, GPT-4, GPT-4o
+	Gemini  Provider = "gemini"  // 代表 Gemini 1.0, 1.5 Pro/Flash
+	Claude  Provider = "claude"  // 代表 Claude 3, 3.5 Sonnet
+	Unknown Provider = "unknown" // 兜底默认
+)
+
+// multipliers 定义不同厂商的计费权重
+type multipliers struct {
+	Word       float64 // 英文单词 (每词)
+	Number     float64 // 数字 (每连续数字串)
+	CJK        float64 // 中日韩字符 (每字)
+	Symbol     float64 // 普通标点符号 (每个)
+	MathSymbol float64 // 数学符号 (∑,∫,∂,√等，每个)
+	URLDelim   float64 // URL分隔符 (/,:,?,&,=,#,%) - tokenizer优化好
+	AtSign     float64 // @符号 - 导致单词切分，消耗较高
+	Emoji      float64 // Emoji表情 (每个)
+	Newline    float64 // 换行符/制表符 (每个)
+	Space      float64 // 空格 (每个)
+	BasePad    int     // 基础起步消耗 (Start/End tokens)
+}
+
+var (
+	multipliersMap = map[Provider]multipliers{
+		Gemini: {
+			Word: 1.15, Number: 2.8, CJK: 0.68, Symbol: 0.38, MathSymbol: 1.05, URLDelim: 1.2, AtSign: 2.5, Emoji: 1.08, Newline: 1.15, Space: 0.2, BasePad: 0,
+		},
+		Claude: {
+			Word: 1.13, Number: 1.63, CJK: 1.21, Symbol: 0.4, MathSymbol: 4.52, URLDelim: 1.26, AtSign: 2.82, Emoji: 2.6, Newline: 0.89, Space: 0.39, BasePad: 0,
+		},
+		OpenAI: {
+			Word: 1.02, Number: 1.55, CJK: 0.85, Symbol: 0.4, MathSymbol: 2.68, URLDelim: 1.0, AtSign: 2.0, Emoji: 2.12, Newline: 0.5, Space: 0.42, BasePad: 0,
+		},
+	}
+	multipliersLock sync.RWMutex
+)
+
+// getMultipliers 根据厂商获取权重配置
+func getMultipliers(p Provider) multipliers {
+	multipliersLock.RLock()
+	defer multipliersLock.RUnlock()
+
+	switch p {
+	case Gemini:
+		return multipliersMap[Gemini]
+	case Claude:
+		return multipliersMap[Claude]
+	case OpenAI:
+		return multipliersMap[OpenAI]
+	default:
+		// 默认兜底 (按 OpenAI 的算)
+		return multipliersMap[OpenAI]
+	}
+}
+
+// EstimateToken 计算 Token 数量
+func EstimateToken(provider Provider, text string) int {
+	m := getMultipliers(provider)
+	var count float64
+
+	// 状态机变量
+	type WordType int
+	const (
+		None WordType = iota
+		Latin
+		Number
+	)
+	currentWordType := None
+
+	for _, r := range text {
+		// 1. 处理空格和换行符
+		if unicode.IsSpace(r) {
+			currentWordType = None
+			// 换行符和制表符使用Newline权重
+			if r == '\n' || r == '\t' {
+				count += m.Newline
+			} else {
+				// 普通空格使用Space权重
+				count += m.Space
+			}
+			continue
+		}
+
+		// 2. 处理 CJK (中日韩) - 按字符计费
+		if isCJK(r) {
+			currentWordType = None
+			count += m.CJK
+			continue
+		}
+
+		// 3. 处理Emoji - 使用专门的Emoji权重
+		if isEmoji(r) {
+			currentWordType = None
+			count += m.Emoji
+			continue
+		}
+
+		// 4. 处理拉丁字母/数字 (英文单词)
+		if isLatinOrNumber(r) {
+			isNum := unicode.IsNumber(r)
+			newType := Latin
+			if isNum {
+				newType = Number
+			}
+
+			// 如果之前不在单词中，或者类型发生变化（字母<->数字），则视为新token
+			// 注意：对于OpenAI，通常"version 3.5"会切分，"abc123xyz"有时也会切分
+			// 这里简单起见，字母和数字切换时增加权重
+			if currentWordType == None || currentWordType != newType {
+				if newType == Number {
+					count += m.Number
+				} else {
+					count += m.Word
+				}
+				currentWordType = newType
+			}
+			// 单词中间的字符不额外计费
+			continue
+		}
+
+		// 5. 处理标点符号/特殊字符 - 按类型使用不同权重
+		currentWordType = None
+		if isMathSymbol(r) {
+			count += m.MathSymbol
+		} else if r == '@' {
+			count += m.AtSign
+		} else if isURLDelim(r) {
+			count += m.URLDelim
+		} else {
+			count += m.Symbol
+		}
+	}
+
+	// 向上取整并加上基础 padding
+	return int(math.Ceil(count)) + m.BasePad
+}
+
+// 辅助：判断是否为 CJK 字符
+func isCJK(r rune) bool {
+	return unicode.Is(unicode.Han, r) ||
+		(r >= 0x3040 && r <= 0x30FF) || // 日文
+		(r >= 0xAC00 && r <= 0xD7A3) // 韩文
+}
+
+// 辅助：判断是否为单词主体 (字母或数字)
+func isLatinOrNumber(r rune) bool {
+	return unicode.IsLetter(r) || unicode.IsNumber(r)
+}
+
+// 辅助：判断是否为Emoji字符
+func isEmoji(r rune) bool {
+	// Emoji的Unicode范围
+	// 基本范围：0x1F300-0x1F9FF (Emoticons, Symbols, Pictographs)
+	// 补充范围：0x2600-0x26FF (Misc Symbols), 0x2700-0x27BF (Dingbats)
+	// 表情符号：0x1F600-0x1F64F (Emoticons)
+	// 其他：0x1F900-0x1F9FF (Supplemental Symbols and Pictographs)
+	return (r >= 0x1F300 && r <= 0x1F9FF) ||
+		(r >= 0x2600 && r <= 0x26FF) ||
+		(r >= 0x2700 && r <= 0x27BF) ||
+		(r >= 0x1F600 && r <= 0x1F64F) ||
+		(r >= 0x1F900 && r <= 0x1F9FF) ||
+		(r >= 0x1FA00 && r <= 0x1FAFF) // Symbols and Pictographs Extended-A
+}
+
+// 辅助：判断是否为数学符号
+func isMathSymbol(r rune) bool {
+	// 数学运算符和符号
+	// 基本数学符号：∑ ∫ ∂ √ ∞ ≤ ≥ ≠ ≈ ± × ÷
+	// 上下标数字：² ³ ¹ ⁴ ⁵ ⁶ ⁷ ⁸ ⁹ ⁰
+	// 希腊字母等也常用于数学
+	mathSymbols := "∑∫∂√∞≤≥≠≈±×÷∈∉∋∌⊂⊃⊆⊇∪∩∧∨¬∀∃∄∅∆∇∝∟∠∡∢°′″‴⁺⁻⁼⁽⁾ⁿ₀₁₂₃₄₅₆₇₈₉₊₋₌₍₎²³¹⁴⁵⁶⁷⁸⁹⁰"
+	for _, m := range mathSymbols {
+		if r == m {
+			return true
+		}
+	}
+	// Mathematical Operators (U+2200–U+22FF)
+	if r >= 0x2200 && r <= 0x22FF {
+		return true
+	}
+	// Supplemental Mathematical Operators (U+2A00–U+2AFF)
+	if r >= 0x2A00 && r <= 0x2AFF {
+		return true
+	}
+	// Mathematical Alphanumeric Symbols (U+1D400–U+1D7FF)
+	if r >= 0x1D400 && r <= 0x1D7FF {
+		return true
+	}
+	return false
+}
+
+// 辅助：判断是否为URL分隔符（tokenizer对这些优化较好）
+func isURLDelim(r rune) bool {
+	// URL中常见的分隔符，tokenizer通常优化处理
+	urlDelims := "/:?&=;#%"
+	for _, d := range urlDelims {
+		if r == d {
+			return true
+		}
+	}
+	return false
+}
+
+func EstimateTokenByModel(model, text string) int {
+	// strings.Contains(model, "gpt-4o")
+	if text == "" {
+		return 0
+	}
+
+	model = strings.ToLower(model)
+	if strings.Contains(model, "gemini") {
+		return EstimateToken(Gemini, text)
+	} else if strings.Contains(model, "claude") {
+		return EstimateToken(Claude, text)
+	} else {
+		return EstimateToken(OpenAI, text)
+	}
+}

service/tokenizer.go

@@ -0,0 +1,63 @@
+package service
+
+import (
+	"sync"
+
+	"github.com/QuantumNous/new-api/common"
+	"github.com/tiktoken-go/tokenizer"
+	"github.com/tiktoken-go/tokenizer/codec"
+)
+
+// tokenEncoderMap won't grow after initialization
+var defaultTokenEncoder tokenizer.Codec
+
+// tokenEncoderMap is used to store token encoders for different models
+var tokenEncoderMap = make(map[string]tokenizer.Codec)
+
+// tokenEncoderMutex protects tokenEncoderMap for concurrent access
+var tokenEncoderMutex sync.RWMutex
+
+func InitTokenEncoders() {
+	common.SysLog("initializing token encoders")
+	defaultTokenEncoder = codec.NewCl100kBase()
+	common.SysLog("token encoders initialized")
+}
+
+func getTokenEncoder(model string) tokenizer.Codec {
+	// First, try to get the encoder from cache with read lock
+	tokenEncoderMutex.RLock()
+	if encoder, exists := tokenEncoderMap[model]; exists {
+		tokenEncoderMutex.RUnlock()
+		return encoder
+	}
+	tokenEncoderMutex.RUnlock()
+
+	// If not in cache, create new encoder with write lock
+	tokenEncoderMutex.Lock()
+	defer tokenEncoderMutex.Unlock()
+
+	// Double-check if another goroutine already created the encoder
+	if encoder, exists := tokenEncoderMap[model]; exists {
+		return encoder
+	}
+
+	// Create new encoder
+	modelCodec, err := tokenizer.ForModel(tokenizer.Model(model))
+	if err != nil {
+		// Cache the default encoder for this model to avoid repeated failures
+		tokenEncoderMap[model] = defaultTokenEncoder
+		return defaultTokenEncoder
+	}
+
+	// Cache the new encoder
+	tokenEncoderMap[model] = modelCodec
+	return modelCodec
+}
+
+func getTokenNum(tokenEncoder tokenizer.Codec, text string) int {
+	if text == "" {
+		return 0
+	}
+	tkm, _ := tokenEncoder.Count(text)
+	return tkm
+}

service/usage_helpr.go

@@ -23,8 +23,7 @@ func ResponseText2Usage(c *gin.Context, responseText string, modeName string, pr
 	common.SetContextKey(c, constant.ContextKeyLocalCountTokens, true)
 	usage := &dto.Usage{}
 	usage.PromptTokens = promptTokens
-	ctkm := CountTextToken(responseText, modeName)
-	usage.CompletionTokens = ctkm
+	usage.CompletionTokens = EstimateTokenByModel(modeName, responseText)
 	usage.TotalTokens = usage.PromptTokens + usage.CompletionTokens
 	return usage
 }

setting/model_setting/global.go

@@ -32,7 +32,7 @@ func GetGlobalSettings() *GlobalSettings {
 	return &globalSettings
 }
 
-// ShouldPreserveThinkingSuffix 判断模型是否配置为保留 thinking/-nothinking 后缀
+// ShouldPreserveThinkingSuffix 判断模型是否配置为保留 thinking/-nothinking/-low/-high/-medium 后缀
 func ShouldPreserveThinkingSuffix(modelName string) bool {
 	target := strings.TrimSpace(modelName)
 	if target == "" {

setting/ratio_setting/cache_ratio.go

@@ -43,6 +43,7 @@ var defaultCacheRatio = map[string]float64{
 	"claude-3-opus-20240229":              0.1,
 	"claude-3-haiku-20240307":             0.1,
 	"claude-3-5-haiku-20241022":           0.1,
+	"claude-haiku-4-5-20251001":           0.1,
 	"claude-3-5-sonnet-20240620":          0.1,
 	"claude-3-5-sonnet-20241022":          0.1,
 	"claude-3-7-sonnet-20250219":          0.1,
@@ -64,6 +65,7 @@ var defaultCreateCacheRatio = map[string]float64{
 	"claude-3-opus-20240229":              1.25,
 	"claude-3-haiku-20240307":             1.25,
 	"claude-3-5-haiku-20241022":           1.25,
+	"claude-haiku-4-5-20251001":           1.25,
 	"claude-3-5-sonnet-20240620":          1.25,
 	"claude-3-5-sonnet-20241022":          1.25,
 	"claude-3-7-sonnet-20250219":          1.25,

setting/ratio_setting/model_ratio.go

@@ -7,6 +7,7 @@ import (
 
 	"github.com/QuantumNous/new-api/common"
 	"github.com/QuantumNous/new-api/setting/operation_setting"
+	"github.com/QuantumNous/new-api/setting/reasoning"
 )
 
 // from songquanpeng/one-api
@@ -136,6 +137,7 @@ var defaultModelRatio = map[string]float64{
 	"claude-2.1":                                4,     // $8 / 1M tokens
 	"claude-3-haiku-20240307":                   0.125, // $0.25 / 1M tokens
 	"claude-3-5-haiku-20241022":                 0.5,   // $1 / 1M tokens
+	"claude-haiku-4-5-20251001":                 0.5,   // $1 / 1M tokens
 	"claude-3-sonnet-20240229":                  1.5,   // $3 / 1M tokens
 	"claude-3-5-sonnet-20240620":                1.5,
 	"claude-3-5-sonnet-20241022":                1.5,
@@ -559,7 +561,7 @@ func getHardcodedCompletionModelRatio(name string) (float64, bool) {
 
 	if strings.Contains(name, "claude-3") {
 		return 5, true
-	} else if strings.Contains(name, "claude-sonnet-4") || strings.Contains(name, "claude-opus-4") {
+	} else if strings.Contains(name, "claude-sonnet-4") || strings.Contains(name, "claude-opus-4") || strings.Contains(name, "claude-haiku-4") {
 		return 5, true
 	} else if strings.Contains(name, "claude-instant-1") || strings.Contains(name, "claude-2") {
 		return 3, true
@@ -821,6 +823,10 @@ func FormatMatchingModelName(name string) string {
 		name = handleThinkingBudgetModel(name, "gemini-2.5-pro", "gemini-2.5-pro-thinking-*")
 	}
 
+	if base, _, ok := reasoning.TrimEffortSuffix(name); ok {
+		name = base
+	}
+
 	if strings.HasPrefix(name, "gpt-4-gizmo") {
 		name = "gpt-4-gizmo-*"
 	}

setting/reasoning/suffix.go

@@ -0,0 +1,20 @@
+package reasoning
+
+import (
+	"strings"
+
+	"github.com/samber/lo"
+)
+
+var EffortSuffixes = []string{"-high", "-medium", "-low"}
+
+// TrimEffortSuffix -> modelName level(low) exists
+func TrimEffortSuffix(modelName string) (string, string, bool) {
+	suffix, found := lo.Find(EffortSuffixes, func(s string) bool {
+		return strings.HasSuffix(modelName, s)
+	})
+	if !found {
+		return modelName, "", false
+	}
+	return strings.TrimSuffix(modelName, suffix), strings.TrimPrefix(suffix, "-"), true
+}

web/src/components/auth/LoginForm.jsx

@@ -294,7 +294,7 @@ const LoginForm = () => {
       setGithubButtonDisabled(true);
     }, 20000);
     try {
-      onGitHubOAuthClicked(status.github_client_id);
+      onGitHubOAuthClicked(status.github_client_id, { shouldLogout: true });
     } finally {
       // 由于重定向，这里不会执行到，但为了完整性添加
       setTimeout(() => setGithubLoading(false), 3000);
@@ -309,7 +309,7 @@ const LoginForm = () => {
     }
     setDiscordLoading(true);
     try {
-      onDiscordOAuthClicked(status.discord_client_id);
+      onDiscordOAuthClicked(status.discord_client_id, { shouldLogout: true });
     } finally {
       // 由于重定向，这里不会执行到，但为了完整性添加
       setTimeout(() => setDiscordLoading(false), 3000);
@@ -324,7 +324,12 @@ const LoginForm = () => {
     }
     setOidcLoading(true);
     try {
-      onOIDCClicked(status.oidc_authorization_endpoint, status.oidc_client_id);
+      onOIDCClicked(
+        status.oidc_authorization_endpoint,
+        status.oidc_client_id,
+        false,
+        { shouldLogout: true },
+      );
     } finally {
       // 由于重定向，这里不会执行到，但为了完整性添加
       setTimeout(() => setOidcLoading(false), 3000);
@@ -339,7 +344,7 @@ const LoginForm = () => {
     }
     setLinuxdoLoading(true);
     try {
-      onLinuxDOOAuthClicked(status.linuxdo_client_id);
+      onLinuxDOOAuthClicked(status.linuxdo_client_id, { shouldLogout: true });
     } finally {
       // 由于重定向，这里不会执行到，但为了完整性添加
       setTimeout(() => setLinuxdoLoading(false), 3000);

web/src/components/auth/RegisterForm.jsx

@@ -261,7 +261,7 @@ const RegisterForm = () => {
       setGithubButtonDisabled(true);
     }, 20000);
     try {
-      onGitHubOAuthClicked(status.github_client_id);
+      onGitHubOAuthClicked(status.github_client_id, { shouldLogout: true });
     } finally {
       setTimeout(() => setGithubLoading(false), 3000);
     }
@@ -270,7 +270,7 @@ const RegisterForm = () => {
   const handleDiscordClick = () => {
     setDiscordLoading(true);
     try {
-      onDiscordOAuthClicked(status.discord_client_id);
+      onDiscordOAuthClicked(status.discord_client_id, { shouldLogout: true });
     } finally {
       setTimeout(() => setDiscordLoading(false), 3000);
     }
@@ -279,7 +279,12 @@ const RegisterForm = () => {
   const handleOIDCClick = () => {
     setOidcLoading(true);
     try {
-      onOIDCClicked(status.oidc_authorization_endpoint, status.oidc_client_id);
+      onOIDCClicked(
+        status.oidc_authorization_endpoint,
+        status.oidc_client_id,
+        false,
+        { shouldLogout: true },
+      );
     } finally {
       setTimeout(() => setOidcLoading(false), 3000);
     }
@@ -288,7 +293,7 @@ const RegisterForm = () => {
   const handleLinuxDOClick = () => {
     setLinuxdoLoading(true);
     try {
-      onLinuxDOOAuthClicked(status.linuxdo_client_id);
+      onLinuxDOOAuthClicked(status.linuxdo_client_id, { shouldLogout: true });
     } finally {
       setTimeout(() => setLinuxdoLoading(false), 3000);
     }

web/src/components/layout/SiderBar.jsx

@@ -377,7 +377,6 @@ const SiderBar = ({ onNavigate = () => {} }) => {
       className='sidebar-container'
       style={{
         width: 'var(--sidebar-current-width)',
-        background: 'var(--semi-color-bg-0)',
       }}
     >
       <SkeletonWrapper

web/src/components/table/task-logs/TaskLogsColumnDefs.jsx

@@ -39,6 +39,7 @@ import {
   TASK_ACTION_GENERATE,
   TASK_ACTION_REFERENCE_GENERATE,
   TASK_ACTION_TEXT_GENERATE,
+  TASK_ACTION_REMIX_GENERATE,
 } from '../../../constants/common.constant';
 import { CHANNEL_OPTIONS } from '../../../constants/channel.constants';
 
@@ -125,6 +126,12 @@ const renderType = (type, t) => {
           {t('参照生视频')}
         </Tag>
       );
+    case TASK_ACTION_REMIX_GENERATE:
+      return (
+        <Tag color='blue' shape='circle' prefixIcon={<Sparkles size={14} />}>
+          {t('视频Remix')}
+        </Tag>
+      );
     default:
       return (
         <Tag color='white' shape='circle' prefixIcon={<HelpCircle size={14} />}>
@@ -359,7 +366,8 @@ export const getTaskLogsColumns = ({
           record.action === TASK_ACTION_GENERATE ||
           record.action === TASK_ACTION_TEXT_GENERATE ||
           record.action === TASK_ACTION_FIRST_TAIL_GENERATE ||
-          record.action === TASK_ACTION_REFERENCE_GENERATE;
+          record.action === TASK_ACTION_REFERENCE_GENERATE ||
+          record.action === TASK_ACTION_REMIX_GENERATE;
         const isSuccess = record.status === 'SUCCESS';
         const isUrl = typeof text === 'string' && /^https?:\/\//.test(text);
         if (isSuccess && isVideoTask && isUrl) {

web/src/constants/common.constant.js

@@ -42,3 +42,4 @@ export const TASK_ACTION_GENERATE = 'generate';
 export const TASK_ACTION_TEXT_GENERATE = 'textGenerate';
 export const TASK_ACTION_FIRST_TAIL_GENERATE = 'firstTailGenerate';
 export const TASK_ACTION_REFERENCE_GENERATE = 'referenceGenerate';
+export const TASK_ACTION_REMIX_GENERATE = 'remixGenerate';

web/src/helpers/api.js

@@ -231,8 +231,22 @@ export async function getOAuthState() {
   }
 }
 
-export async function onDiscordOAuthClicked(client_id) {
-  const state = await getOAuthState();
+async function prepareOAuthState(options = {}) {
+  const { shouldLogout = false } = options;
+  if (shouldLogout) {
+    try {
+      await API.get('/api/user/logout', { skipErrorHandler: true });
+    } catch (err) {
+
+    }
+    localStorage.removeItem('user');
+    updateAPI();
+  }
+  return await getOAuthState();
+}
+
+export async function onDiscordOAuthClicked(client_id, options = {}) {
+  const state = await prepareOAuthState(options);
   if (!state) return;
   const redirect_uri = `${window.location.origin}/oauth/discord`;
   const response_type = 'code';
@@ -242,8 +256,13 @@ export async function onDiscordOAuthClicked(client_id) {
   );
 }
 
-export async function onOIDCClicked(auth_url, client_id, openInNewTab = false) {
-  const state = await getOAuthState();
+export async function onOIDCClicked(
+  auth_url,
+  client_id,
+  openInNewTab = false,
+  options = {},
+) {
+  const state = await prepareOAuthState(options);
   if (!state) return;
   const url = new URL(auth_url);
   url.searchParams.set('client_id', client_id);
@@ -258,16 +277,19 @@ export async function onOIDCClicked(auth_url, client_id, openInNewTab = false) {
   }
 }
 
-export async function onGitHubOAuthClicked(github_client_id) {
-  const state = await getOAuthState();
+export async function onGitHubOAuthClicked(github_client_id, options = {}) {
+  const state = await prepareOAuthState(options);
   if (!state) return;
   window.open(
     `https://github.com/login/oauth/authorize?client_id=${github_client_id}&state=${state}&scope=user:email`,
   );
 }
 
-export async function onLinuxDOOAuthClicked(linuxdo_client_id) {
-  const state = await getOAuthState();
+export async function onLinuxDOOAuthClicked(
+  linuxdo_client_id,
+  options = { shouldLogout: false },
+) {
+  const state = await prepareOAuthState(options);
   if (!state) return;
   window.open(
     `https://connect.linux.do/oauth2/authorize?response_type=code&client_id=${linuxdo_client_id}&state=${state}`,

web/src/helpers/render.jsx

@@ -1086,9 +1086,12 @@ function renderPriceSimpleCore({
   );
   const finalGroupRatio = effectiveGroupRatio;
 
+  const { symbol, rate } = getCurrencyConfig();
   if (modelPrice !== -1) {
-    return i18next.t('价格：${{price}} * {{ratioType}}：{{ratio}}', {
-      price: modelPrice,
+    const displayPrice = (modelPrice * rate).toFixed(6);
+    return i18next.t('价格：{{symbol}}{{price}} * {{ratioType}}：{{ratio}}', {
+      symbol: symbol,
+      price: displayPrice,
       ratioType: ratioLabel,
       ratio: finalGroupRatio,
     });

web/src/i18n/locales/en.json

@@ -548,6 +548,7 @@
     "参数值": "Parameter value",
     "参数覆盖": "Parameters override",
     "参照生视频": "Reference video generation",
+    "视频Remix": "Video remix",
     "友情链接": "Friendly links",
     "发布日期": "Publish Date",
     "发布时间": "Publish Time",
@@ -1996,7 +1997,7 @@
     "适用于个人使用的场景，不需要设置模型价格": "Suitable for personal use, no need to set model price.",
     "适用于为多个用户提供服务的场景": "Suitable for scenarios where multiple users are provided.",
     "适用于展示系统功能的场景，提供基础功能演示": "Suitable for scenarios where the system functions are displayed, providing basic feature demonstrations.",
-    "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀": "Adapt to -thinking, -thinking-budget number, and -nothinking suffixes",
+    "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀": "Adapt to -thinking, -thinking-budget number, -nothinking, and -low/-medium/-high suffixes",
     "选择充值套餐": "Choose a top-up package",
     "选择充值额度": "Select recharge amount",
     "选择分组": "Select group",
@@ -2178,4 +2179,4 @@
     "默认测试模型": "Default Test Model",
     "默认补全倍率": "Default completion ratio"
   }
-}
+}

web/src/i18n/locales/fr.json

@@ -551,6 +551,7 @@
     "参数值": "Valeur du paramètre",
     "参数覆盖": "Remplacement des paramètres",
     "参照生视频": "Générer une vidéo par référence",
+    "视频Remix": "Remix vidéo",
     "友情链接": "Liens amicaux",
     "发布日期": "Date de publication",
     "发布时间": "Heure de publication",
@@ -2006,7 +2007,7 @@
     "适用于个人使用的场景，不需要设置模型价格": "Adapté à un usage personnel, pas besoin de définir le prix du modèle.",
     "适用于为多个用户提供服务的场景": "Adapté aux scénarios où plusieurs utilisateurs sont fournis.",
     "适用于展示系统功能的场景，提供基础功能演示": "Adapté aux scénarios où les fonctions du système sont affichées, fournissant des démonstrations de fonctionnalités de base.",
-    "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀": "Adapter les suffixes -thinking, -thinking-budget et -nothinking",
+    "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀": "Adapter les suffixes -thinking, -thinking-budget, -nothinking et -low/-medium/-high",
     "选择充值额度": "Sélectionner le montant de la recharge",
     "选择分组": "Sélectionner un groupe",
     "选择同步来源": "Sélectionner la source de synchronisation",
@@ -2227,4 +2228,4 @@
     "随机种子 (留空为随机)": "Graine aléatoire (laisser vide pour aléatoire)",
     "默认补全倍率": "Taux de complétion par défaut"
   }
-}
+}

web/src/i18n/locales/ja.json

@@ -510,6 +510,7 @@
     "参数值": "パラメータ値",
     "参数覆盖": "パラメータの上書き",
     "参照生视频": "参照動画生成",
+    "视频Remix": "動画リミックス",
     "友情链接": "関連リンク",
     "发布日期": "公開日",
     "发布时间": "公開日時",
@@ -1903,7 +1904,7 @@
     "适用于个人使用的场景，不需要设置模型价格": "個人利用のシナリオに適しており、モデル料金の設定は不要です",
     "适用于为多个用户提供服务的场景": "複数のユーザーにサービスを提供するシナリオに適しています",
     "适用于展示系统功能的场景，提供基础功能演示": "システムの機能を紹介するシナリオに適しており、基本的な機能のデモンストレーションを提供します",
-    "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀": "-thinking、-thinking-予算数値、-nothinkingサフィックスに対応",
+    "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀": "-thinking、-thinking-予算数値、-nothinking、および -low/-medium/-high サフィックスに対応",
     "选择充值额度": "チャージ額を選択",
     "选择分组": "グループを選択",
     "选择同步来源": "同期ソースを選択",
@@ -2126,4 +2127,4 @@
     "可选，用于复现结果": "オプション、結果の再現用",
     "随机种子 (留空为随机)": "ランダムシード（空欄でランダム）"
   }
-}
+}

web/src/i18n/locales/ru.json

@@ -555,6 +555,7 @@
     "参数值": "Значение параметра",
     "参数覆盖": "Переопределение параметров",
     "参照生视频": "Ссылка на генерацию видео",
+    "视频Remix": "Видео ремикс",
     "友情链接": "Дружественные ссылки",
     "发布日期": "Дата публикации",
     "发布时间": "Время публикации",
@@ -2017,7 +2018,7 @@
     "适用于个人使用的场景，不需要设置模型价格": "Подходит для сценариев личного использования, не требует установки цен на модели",
     "适用于为多个用户提供服务的场景": "Подходит для сценариев предоставления услуг нескольким пользователям",
     "适用于展示系统功能的场景，提供基础功能演示": "Подходит для сценариев демонстрации системных функций, предоставляет демонстрацию базовых функций",
-    "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀": "Адаптация суффиксов -thinking, -thinking-бюджетные-цифры и -nothinking",
+    "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀": "Адаптация суффиксов -thinking, -thinking-бюджетные-цифры, -nothinking и -low/-medium/-high",
     "选择充值额度": "Выберите сумму пополнения",
     "选择分组": "Выберите группу",
     "选择同步来源": "Выберите источник синхронизации",
@@ -2237,4 +2238,4 @@
     "可选，用于复现结果": "Необязательно, для воспроизводимых результатов",
     "随机种子 (留空为随机)": "Случайное зерно (оставьте пустым для случайного)"
   }
-}
+}

web/src/i18n/locales/vi.json

@@ -510,6 +510,7 @@
     "参数值": "Giá trị tham số",
     "参数覆盖": "Ghi đè tham số",
     "参照生视频": "Tạo video tham chiếu",
+    "视频Remix": "Remix video",
     "友情链接": "Liên kết thân thiện",
     "发布日期": "Ngày xuất bản",
     "发布时间": "Thời gian xuất bản",
@@ -2197,7 +2198,7 @@
     "适用于个人使用的场景，不需要设置模型价格": "Phù hợp cho mục đích sử dụng cá nhân, không cần đặt giá mô hình.",
     "适用于为多个用户提供服务的场景": "Phù hợp cho các kịch bản cung cấp dịch vụ cho nhiều người dùng.",
     "适用于展示系统功能的场景，提供基础功能演示": "Phù hợp cho các kịch bản hiển thị chức năng hệ thống, cung cấp bản demo chức năng cơ bản.",
-    "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀": "Thích ứng với các hậu tố -thinking, -thinking-budget number và -nothinking",
+    "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀": "Thích ứng với các hậu tố -thinking, -thinking-budget number, -nothinking và -low/-medium/-high",
     "选择充值额度": "Chọn hạn ngạch nạp tiền",
     "选择同步来源": "Chọn nguồn đồng bộ",
     "选择同步渠道": "Chọn kênh đồng bộ",
@@ -2737,4 +2738,4 @@
     "可选，用于复现结果": "Tùy chọn, để tái tạo kết quả",
     "随机种子 (留空为随机)": "Hạt giống ngẫu nhiên (để trống cho ngẫu nhiên)"
   }
-}
+}

web/src/i18n/locales/zh.json

@@ -543,6 +543,7 @@
     "参数值": "参数值",
     "参数覆盖": "参数覆盖",
     "参照生视频": "参照生视频",
+    "视频Remix": "视频 Remix",
     "友情链接": "友情链接",
     "发布日期": "发布日期",
     "发布时间": "发布时间",
@@ -1984,7 +1985,7 @@
     "适用于个人使用的场景，不需要设置模型价格": "适用于个人使用的场景，不需要设置模型价格",
     "适用于为多个用户提供服务的场景": "适用于为多个用户提供服务的场景",
     "适用于展示系统功能的场景，提供基础功能演示": "适用于展示系统功能的场景，提供基础功能演示",
-    "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀": "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀",
+    "适配 -thinking、-thinking-预算数字 和 -nothinking 后缀": "适配 -thinking、-thinking-预算数字、-nothinking 以及 -low/-medium/-high 后缀",
     "选择充值额度": "选择充值额度",
     "选择分组": "选择分组",
     "选择同步来源": "选择同步来源",
@@ -2204,4 +2205,4 @@
     "可选，用于复现结果": "可选，用于复现结果",
     "随机种子 (留空为随机)": "随机种子 (留空为随机)"
   }
-}
+}

web/src/index.css

@@ -108,6 +108,7 @@ code {
   display: flex;
   flex-direction: column;
   transition: width 0.3s ease;
+  background: var(--semi-color-bg-0);
 }
 
 .sidebar-nav {
@@ -221,6 +222,22 @@ code {
   padding-top: 12px;
 }
 
+@media (max-width: 767px) {
+  .sidebar-container {
+    background: var(--semi-color-bg-1);
+    border-right: 1px solid var(--semi-color-border);
+  }
+
+  .sidebar-nav {
+    background: var(--semi-color-bg-1);
+  }
+
+  .sidebar-collapse-button {
+    background-color: var(--semi-color-bg-1);
+    box-shadow: 0 -10px 10px -5px var(--semi-color-bg-1);
+  }
+}
+
 /* ==================== 聊天界面样式 ==================== */
 .semi-chat {
   padding-top: 0 !important;