feat: add support for multi-key channels in RelayInfo and access token caching

fix: correct Gemini channel model retrieval logic

controller/channel.go

@@ -36,11 +36,30 @@ type OpenAIModel struct {
 	Parent string `json:"parent"`
 }
 
+type GoogleOpenAICompatibleModels []struct {
+	Name                       string   `json:"name"`
+	Version                    string   `json:"version"`
+	DisplayName                string   `json:"displayName"`
+	Description                string   `json:"description,omitempty"`
+	InputTokenLimit            int      `json:"inputTokenLimit"`
+	OutputTokenLimit           int      `json:"outputTokenLimit"`
+	SupportedGenerationMethods []string `json:"supportedGenerationMethods"`
+	Temperature                float64  `json:"temperature,omitempty"`
+	TopP                       float64  `json:"topP,omitempty"`
+	TopK                       int      `json:"topK,omitempty"`
+	MaxTemperature             int      `json:"maxTemperature,omitempty"`
+}
+
 type OpenAIModelsResponse struct {
 	Data    []OpenAIModel `json:"data"`
 	Success bool          `json:"success"`
 }
 
+type GoogleOpenAICompatibleResponse struct {
+	Models        []GoogleOpenAICompatibleModels `json:"models"`
+	NextPageToken string                         `json:"nextPageToken"`
+}
+
 func parseStatusFilter(statusParam string) int {
 	switch strings.ToLower(statusParam) {
 	case "enabled", "1":
@@ -168,26 +187,59 @@ func FetchUpstreamModels(c *gin.Context) {
 	if channel.GetBaseURL() != "" {
 		baseURL = channel.GetBaseURL()
 	}
-	url := fmt.Sprintf("%s/v1/models", baseURL)
+
+	var url string
 	switch channel.Type {
 	case constant.ChannelTypeGemini:
-		url = fmt.Sprintf("%s/v1beta/openai/models", baseURL)
+		// curl https://example.com/v1beta/models?key=$GEMINI_API_KEY
+		url = fmt.Sprintf("%s/v1beta/openai/models?key=%s", baseURL, channel.Key)
 	case constant.ChannelTypeAli:
 		url = fmt.Sprintf("%s/compatible-mode/v1/models", baseURL)
+	default:
+		url = fmt.Sprintf("%s/v1/models", baseURL)
+	}
+
+	// 获取响应体 - 根据渠道类型决定是否添加 AuthHeader
+	var body []byte
+	if channel.Type == constant.ChannelTypeGemini {
+		body, err = GetResponseBody("GET", url, channel, nil) // I don't know why, but Gemini requires no AuthHeader
+	} else {
+		body, err = GetResponseBody("GET", url, channel, GetAuthHeader(channel.Key))
 	}
-	body, err := GetResponseBody("GET", url, channel, GetAuthHeader(channel.Key))
 	if err != nil {
 		common.ApiError(c, err)
 		return
 	}
 
 	var result OpenAIModelsResponse
-	if err = json.Unmarshal(body, &result); err != nil {
-		c.JSON(http.StatusOK, gin.H{
-			"success": false,
-			"message": fmt.Sprintf("解析响应失败: %s", err.Error()),
-		})
-		return
+	var parseSuccess bool
+
+	// 适配特殊格式
+	switch channel.Type {
+	case constant.ChannelTypeGemini:
+		var googleResult GoogleOpenAICompatibleResponse
+		if err = json.Unmarshal(body, &googleResult); err == nil {
+			// 转换Google格式到OpenAI格式
+			for _, model := range googleResult.Models {
+				for _, gModel := range model {
+					result.Data = append(result.Data, OpenAIModel{
+						ID: gModel.Name,
+					})
+				}
+			}
+			parseSuccess = true
+		}
+	}
+
+	// 如果解析失败，尝试OpenAI格式
+	if !parseSuccess {
+		if err = json.Unmarshal(body, &result); err != nil {
+			c.JSON(http.StatusOK, gin.H{
+				"success": false,
+				"message": fmt.Sprintf("解析响应失败: %s", err.Error()),
+			})
+			return
+		}
 	}
 
 	var ids []string

model/ability.go

@@ -284,6 +284,21 @@ func FixAbility() (int, int, error) {
 		return 0, 0, errors.New("已经有一个修复任务在运行中，请稍后再试")
 	}
 	defer fixLock.Unlock()
+
+	// truncate abilities table
+	if common.UsingSQLite {
+		err := DB.Exec("DELETE FROM abilities").Error
+		if err != nil {
+			common.SysError(fmt.Sprintf("Delete abilities failed: %s", err.Error()))
+			return 0, 0, err
+		}
+	} else {
+		err := DB.Exec("TRUNCATE TABLE abilities").Error
+		if err != nil {
+			common.SysError(fmt.Sprintf("Truncate abilities failed: %s", err.Error()))
+			return 0, 0, err
+		}
+	}
 	var channels []*Channel
 	// Find all channels
 	err := DB.Model(&Channel{}).Find(&channels).Error

model/channel.go

@@ -46,6 +46,9 @@ type Channel struct {
 	ParamOverride     *string `json:"param_override" gorm:"type:text"`
 	// add after v0.8.5
 	ChannelInfo ChannelInfo `json:"channel_info" gorm:"type:json"`
+
+	// cache info
+	Keys []string `json:"-" gorm:"-"`
 }
 
 type ChannelInfo struct {
@@ -71,6 +74,9 @@ func (channel *Channel) getKeys() []string {
 	if channel.Key == "" {
 		return []string{}
 	}
+	if len(channel.Keys) > 0 {
+		return channel.Keys
+	}
 	trimmed := strings.TrimSpace(channel.Key)
 	// If the key starts with '[', try to parse it as a JSON array (e.g., for Vertex AI scenarios)
 	if strings.HasPrefix(trimmed, "[") {

model/channel_cache.go

@@ -69,10 +69,15 @@ func InitChannelCache() {
 	group2model2channels = newGroup2model2channels
 	//channelsIDM = newChannelId2channel
 	for i, channel := range newChannelId2channel {
-		if oldChannel, ok := channelsIDM[i]; ok {
-			// 存在旧的渠道，如果是多key且轮询，保留轮询索引信息
-			if oldChannel.ChannelInfo.IsMultiKey && oldChannel.ChannelInfo.MultiKeyMode == constant.MultiKeyModePolling {
-				channel.ChannelInfo.MultiKeyPollingIndex = oldChannel.ChannelInfo.MultiKeyPollingIndex
+		if channel.ChannelInfo.IsMultiKey {
+			channel.Keys = channel.getKeys()
+			if channel.ChannelInfo.MultiKeyMode == constant.MultiKeyModePolling {
+				if oldChannel, ok := channelsIDM[i]; ok {
+					// 存在旧的渠道，如果是多key且轮询，保留轮询索引信息
+					if oldChannel.ChannelInfo.IsMultiKey && oldChannel.ChannelInfo.MultiKeyMode == constant.MultiKeyModePolling {
+						channel.ChannelInfo.MultiKeyPollingIndex = oldChannel.ChannelInfo.MultiKeyPollingIndex
+					}
+				}
 			}
 		}
 	}

relay/channel/vertex/service_account.go

@@ -36,7 +36,12 @@ var Cache = asynccache.NewAsyncCache(asynccache.Options{
 })
 
 func getAccessToken(a *Adaptor, info *relaycommon.RelayInfo) (string, error) {
-	cacheKey := fmt.Sprintf("access-token-%d", info.ChannelId)
+	var cacheKey string
+	if info.ChannelIsMultiKey {
+		cacheKey = fmt.Sprintf("access-token-%d-%d", info.ChannelId, info.ChannelMultiKeyIndex)
+	} else {
+		cacheKey = fmt.Sprintf("access-token-%d", info.ChannelId)
+	}
 	val, err := Cache.Get(cacheKey)
 	if err == nil {
 		return val.(string), nil

relay/common/relay_info.go

@@ -60,17 +60,19 @@ type ResponsesUsageInfo struct {
 }
 
 type RelayInfo struct {
-	ChannelType       int
-	ChannelId         int
-	TokenId           int
-	TokenKey          string
-	UserId            int
-	UsingGroup        string // 使用的分组
-	UserGroup         string // 用户所在分组
-	TokenUnlimited    bool
-	StartTime         time.Time
-	FirstResponseTime time.Time
-	isFirstResponse   bool
+	ChannelType          int
+	ChannelId            int
+	ChannelIsMultiKey    bool // 是否多密钥
+	ChannelMultiKeyIndex int  // 多密钥索引
+	TokenId              int
+	TokenKey             string
+	UserId               int
+	UsingGroup           string // 使用的分组
+	UserGroup            string // 用户所在分组
+	TokenUnlimited       bool
+	StartTime            time.Time
+	FirstResponseTime    time.Time
+	isFirstResponse      bool
 	//SendLastReasoningResponse bool
 	ApiType           int
 	IsStream          bool
@@ -260,6 +262,9 @@ func GenRelayInfo(c *gin.Context) *RelayInfo {
 			IsFirstThinkingContent:  true,
 			SendLastThinkingContent: false,
 		},
+
+		ChannelIsMultiKey:    common.GetContextKeyBool(c, constant.ContextKeyChannelIsMultiKey),
+		ChannelMultiKeyIndex: common.GetContextKeyInt(c, constant.ContextKeyChannelMultiKeyIndex),
 	}
 	if strings.HasPrefix(c.Request.URL.Path, "/pg") {
 		info.IsPlayground = true

relay/gemini_handler.go

@@ -80,7 +80,11 @@ func getGeminiInputTokens(req *dto.GeminiChatRequest, info *relaycommon.RelayInf
 
 func isNoThinkingRequest(req *dto.GeminiChatRequest) bool {
 	if req.GenerationConfig.ThinkingConfig != nil && req.GenerationConfig.ThinkingConfig.ThinkingBudget != nil {
-		return *req.GenerationConfig.ThinkingConfig.ThinkingBudget == 0
+		configBudget := req.GenerationConfig.ThinkingConfig.ThinkingBudget
+		if configBudget != nil && *configBudget == 0 {
+			// 如果思考预算为 0，则认为是非思考请求
+			return true
+		}
 	}
 	return false
 }