Merge 761ee32d19 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/gemini/main.go

@@ -8,19 +8,18 @@ import (
 	"net/http"
 	"strings"
 
-	"github.com/songquanpeng/one-api/common/render"
-
+	"github.com/gin-gonic/gin"
 	"github.com/songquanpeng/one-api/common"
 	"github.com/songquanpeng/one-api/common/config"
 	"github.com/songquanpeng/one-api/common/helper"
 	"github.com/songquanpeng/one-api/common/image"
 	"github.com/songquanpeng/one-api/common/logger"
 	"github.com/songquanpeng/one-api/common/random"
+	"github.com/songquanpeng/one-api/common/render"
+	"github.com/songquanpeng/one-api/relay/adaptor/geminiv2"
 	"github.com/songquanpeng/one-api/relay/adaptor/openai"
 	"github.com/songquanpeng/one-api/relay/constant"
 	"github.com/songquanpeng/one-api/relay/model"
-
-	"github.com/gin-gonic/gin"
 )
 
 // https://ai.google.dev/docs/gemini_api_overview?hl=zh-cn
@@ -61,12 +60,10 @@ func ConvertRequest(textRequest model.GeneralOpenAIRequest) *ChatRequest {
 			},
 		},
 		GenerationConfig: ChatGenerationConfig{
-			Temperature:     textRequest.Temperature,
-			TopP:            textRequest.TopP,
-			MaxOutputTokens: textRequest.MaxTokens,
-			ResponseModalities: []string{
-				"TEXT", "IMAGE",
-			},
+			Temperature:        textRequest.Temperature,
+			TopP:               textRequest.TopP,
+			MaxOutputTokens:    textRequest.MaxTokens,
+			ResponseModalities: geminiv2.GetModelModalities(textRequest.Model),
 		},
 	}
 	if textRequest.ResponseFormat != nil {

relay/adaptor/geminiv2/constants.go

@@ -1,5 +1,7 @@
 package geminiv2
 
+import "strings"
+
 // https://ai.google.dev/models/gemini
 
 var ModelList = []string{
@@ -14,3 +16,17 @@ var ModelList = []string{
 	"gemini-2.0-pro-exp-02-05",
 	"gemini-2.0-flash-exp-image-generation",
 }
+
+const (
+	ModalityText  = "TEXT"
+	ModalityImage = "IMAGE"
+)
+
+// GetModelModalities returns the modalities of the model.
+func GetModelModalities(model string) []string {
+	if strings.Contains(model, "-image-generation") {
+		return []string{ModalityText, ModalityImage}
+	}
+
+	return []string{ModalityText}
+}

relay/model/general.go

@@ -40,7 +40,7 @@ type GeneralOpenAIRequest struct {
 	MaxCompletionTokens *int     `json:"max_completion_tokens,omitempty"`
 	// N is how many chat completion choices to generate for each input message,
 	// default to 1.
-	N *int `json:"n,omitempty" binding:"omitempty,min=1"`
+	N *int `json:"n,omitempty" binding:"omitempty,min=0"`
 	// ReasoningEffort constrains effort on reasoning for reasoning models, reasoning models only.
 	ReasoningEffort *string `json:"reasoning_effort,omitempty" binding:"omitempty,oneof=low medium high"`
 	// Modalities currently the model only programmatically allows modalities = [“text”, “audio”]