Merge dabaa795b9 into 8df4a2670b

README.md

@@ -72,7 +72,7 @@ _✨ 通过标准的 OpenAI API 格式访问所有的大模型，开箱即用
    + [x] [Anthropic Claude 系列模型](https://anthropic.com) (支持 AWS Claude)
    + [x] [Google PaLM2/Gemini 系列模型](https://developers.generativeai.google)
    + [x] [Mistral 系列模型](https://mistral.ai/)
-   + [x] [字节跳动豆包大模型](https://console.volcengine.com/ark/region:ark+cn-beijing/model)
+   + [x] [字节跳动豆包大模型（火山引擎）](https://www.volcengine.com/experience/ark?utm_term=202502dsinvite&ac=DSASUQY5&rc=2QXCA1VI)
    + [x] [百度文心一言系列模型](https://cloud.baidu.com/doc/WENXINWORKSHOP/index.html)
    + [x] [阿里通义千问系列模型](https://help.aliyun.com/document_detail/2400395.html)
    + [x] [讯飞星火认知大模型](https://www.xfyun.cn/doc/spark/Web.html)

relay/adaptor/openai/adaptor.go

@@ -33,16 +33,24 @@ func (a *Adaptor) Init(meta *meta.Meta) {
 func (a *Adaptor) GetRequestURL(meta *meta.Meta) (string, error) {
 	switch meta.ChannelType {
 	case channeltype.Azure:
+		defaultVersion := meta.Config.APIVersion
+
+		// https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/reasoning?tabs=python#api--feature-support
+		if strings.HasPrefix(meta.ActualModelName, "o1") ||
+			strings.HasPrefix(meta.ActualModelName, "o3") {
+			defaultVersion = "2024-12-01-preview"
+		}
+
 		if meta.Mode == relaymode.ImagesGenerations {
 			// https://learn.microsoft.com/en-us/azure/ai-services/openai/dall-e-quickstart?tabs=dalle3%2Ccommand-line&pivots=rest-api
 			// https://{resource_name}.openai.azure.com/openai/deployments/dall-e-3/images/generations?api-version=2024-03-01-preview
-			fullRequestURL := fmt.Sprintf("%s/openai/deployments/%s/images/generations?api-version=%s", meta.BaseURL, meta.ActualModelName, meta.Config.APIVersion)
+			fullRequestURL := fmt.Sprintf("%s/openai/deployments/%s/images/generations?api-version=%s", meta.BaseURL, meta.ActualModelName, defaultVersion)
 			return fullRequestURL, nil
 		}
 
 		// https://learn.microsoft.com/en-us/azure/cognitive-services/openai/chatgpt-quickstart?pivots=rest-api&tabs=command-line#rest-api
 		requestURL := strings.Split(meta.RequestURLPath, "?")[0]
-		requestURL = fmt.Sprintf("%s?api-version=%s", requestURL, meta.Config.APIVersion)
+		requestURL = fmt.Sprintf("%s?api-version=%s", requestURL, defaultVersion)
 		task := strings.TrimPrefix(requestURL, "/v1/")
 		model_ := meta.ActualModelName
 		model_ = strings.Replace(model_, ".", "", -1)
@@ -92,6 +100,23 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *model.G
 		}
 		request.StreamOptions.IncludeUsage = true
 	}
+
+	// o1/o1-mini/o1-preview do not support system prompt/max_tokens/temperature
+	if strings.HasPrefix(request.Model, "o1") {
+		temperature := float64(1)
+		request.Temperature = &temperature // Only the default (1) value is supported
+		request.MaxTokens = 0
+		request.Messages = func(raw []model.Message) (filtered []model.Message) {
+			for i := range raw {
+				if raw[i].Role != "system" {
+					filtered = append(filtered, raw[i])
+				}
+			}
+
+			return
+		}(request.Messages)
+	}
+
 	return request, nil
 }
 

relay/billing/ratio/model.go

@@ -763,7 +763,7 @@ func GetCompletionRatio(name string, channelType int) float64 {
 		}
 		return 2
 	}
-	// including o1, o1-preview, o1-mini
+	// including o1/o1-preview/o1-mini
 	if strings.HasPrefix(name, "o1") {
 		return 4
 	}