feat: add gemini-2.0-flash-exp and fix race condition in processChannelRelayError

controller/relay.go

@@ -60,7 +60,7 @@ func Relay(c *gin.Context) {
 	channelName := c.GetString(ctxkey.ChannelName)
 	group := c.GetString(ctxkey.Group)
 	originalModel := c.GetString(ctxkey.OriginalModel)
-	go processChannelRelayError(ctx, userId, channelId, channelName, bizErr)
+	go processChannelRelayError(ctx, userId, channelId, channelName, *bizErr)
 	requestId := c.GetString(helper.RequestIdKey)
 	retryTimes := config.RetryTimes
 	if !shouldRetry(c, bizErr.StatusCode) {
@@ -87,8 +87,7 @@ func Relay(c *gin.Context) {
 		channelId := c.GetInt(ctxkey.ChannelId)
 		lastFailedChannelId = channelId
 		channelName := c.GetString(ctxkey.ChannelName)
-		// BUG: bizErr is in race condition
+		go processChannelRelayError(ctx, userId, channelId, channelName, *bizErr)
-		go processChannelRelayError(ctx, userId, channelId, channelName, bizErr)
 	}
 	if bizErr != nil {
 		if bizErr.StatusCode == http.StatusTooManyRequests {
@@ -122,7 +121,7 @@ func shouldRetry(c *gin.Context, statusCode int) bool {
 	return true
 }
 
-func processChannelRelayError(ctx context.Context, userId int, channelId int, channelName string, err *model.ErrorWithStatusCode) {
+func processChannelRelayError(ctx context.Context, userId int, channelId int, channelName string, err model.ErrorWithStatusCode) {
 	logger.Errorf(ctx, "relay error (channel id %d, user id: %d): %s", channelId, userId, err.Message)
 	// https://platform.openai.com/docs/guides/error-codes/api-errors
 	if monitor.ShouldDisableChannel(&err.Error, err.StatusCode) {

monitor/manage.go

@@ -34,7 +34,7 @@ func ShouldDisableChannel(err *model.Error, statusCode int) bool {
 		strings.Contains(lowerMessage, "credit") ||
 		strings.Contains(lowerMessage, "balance") ||
 		strings.Contains(lowerMessage, "permission denied") ||
-  	strings.Contains(lowerMessage, "organization has been restricted") || // groq
+		strings.Contains(lowerMessage, "organization has been restricted") || // groq
 		strings.Contains(lowerMessage, "已欠费") {
 		return true
 	}

relay/adaptor/gemini/adaptor.go

@@ -24,7 +24,12 @@ func (a *Adaptor) Init(meta *meta.Meta) {
 }
 
 func (a *Adaptor) GetRequestURL(meta *meta.Meta) (string, error) {
-	version := helper.AssignOrDefault(meta.Config.APIVersion, config.GeminiVersion)
+	defaultVersion := config.GeminiVersion
+	if meta.ActualModelName == "gemini-2.0-flash-exp" {
+		defaultVersion = "v1beta"
+	}
+
+	version := helper.AssignOrDefault(meta.Config.APIVersion, defaultVersion)
 	action := ""
 	switch meta.Mode {
 	case relaymode.Embeddings:
@@ -36,6 +41,7 @@ func (a *Adaptor) GetRequestURL(meta *meta.Meta) (string, error) {
 	if meta.IsStream {
 		action = "streamGenerateContent?alt=sse"
 	}
+
 	return fmt.Sprintf("%s/%s/models/%s:%s", meta.BaseURL, version, meta.ActualModelName, action), nil
 }
 

relay/adaptor/gemini/constants.go

@@ -3,5 +3,8 @@ package gemini
 // https://ai.google.dev/models/gemini
 
 var ModelList = []string{
-	"gemini-pro", "gemini-1.0-pro", "gemini-1.5-flash", "gemini-1.5-pro", "text-embedding-004", "aqa",
+	"gemini-pro", "gemini-1.0-pro",
+	"gemini-1.5-flash", "gemini-1.5-pro",
+	"text-embedding-004", "aqa",
+	"gemini-2.0-flash-exp",
 }

relay/adaptor/ollama/main.go

@@ -31,8 +31,8 @@ func ConvertRequest(request model.GeneralOpenAIRequest) *ChatRequest {
 			TopP:             request.TopP,
 			FrequencyPenalty: request.FrequencyPenalty,
 			PresencePenalty:  request.PresencePenalty,
-			NumPredict:  	  request.MaxTokens,
+			NumPredict:       request.MaxTokens,
-			NumCtx:  	  request.NumCtx,
+			NumCtx:           request.NumCtx,
 		},
 		Stream: request.Stream,
 	}
@@ -122,7 +122,7 @@ func StreamHandler(c *gin.Context, resp *http.Response) (*model.ErrorWithStatusC
 	for scanner.Scan() {
 		data := scanner.Text()
 		if strings.HasPrefix(data, "}") {
-		    data = strings.TrimPrefix(data, "}") + "}"
+			data = strings.TrimPrefix(data, "}") + "}"
 		}
 
 		var ollamaResponse ChatResponse

relay/adaptor/vertexai/gemini/adapter.go

@@ -15,7 +15,10 @@ import (
 )
 
 var ModelList = []string{
-	"gemini-1.5-pro-001", "gemini-1.5-flash-001", "gemini-pro", "gemini-pro-vision", "gemini-1.5-pro-002", "gemini-1.5-flash-002", 
+	"gemini-pro", "gemini-pro-vision",
+	"gemini-1.5-pro-001", "gemini-1.5-flash-001",
+	"gemini-1.5-pro-002", "gemini-1.5-flash-002",
+	"gemini-2.0-flash-exp",
 }
 
 type Adaptor struct {

relay/billing/ratio/model.go

@@ -102,11 +102,14 @@ var ModelRatio = map[string]float64{
 	"bge-large-en":       0.002 * RMB,
 	"tao-8k":             0.002 * RMB,
 	// https://ai.google.dev/pricing
-	"gemini-pro":       1, // $0.00025 / 1k characters -> $0.001 / 1k tokens
+	"gemini-pro":           1, // $0.00025 / 1k characters -> $0.001 / 1k tokens
-	"gemini-1.0-pro":   1,
+	"gemini-1.0-pro":       1,
-	"gemini-1.5-flash": 1,
+	"gemini-1.5-pro":       1,
-	"gemini-1.5-pro":   1,
+	"gemini-1.5-pro-001":   1,
-	"aqa":              1,
+	"gemini-1.5-flash":     1,
+	"gemini-1.5-flash-001": 1,
+	"gemini-2.0-flash-exp": 1,
+	"aqa":                  1,
 	// https://open.bigmodel.cn/pricing
 	"glm-4":         0.1 * RMB,
 	"glm-4v":        0.1 * RMB,