feat: Enhance ConvertClaudeRequest method to set request model and handle vertex-specific request conversion

feat: support AWS Model CrossRegion

README.md

@@ -36,8 +36,8 @@
 > 本项目为开源项目，在[One API](https://github.com/songquanpeng/one-api)的基础上进行二次开发
 
 > [!IMPORTANT]  
-> - 使用者必须在遵循 OpenAI 的[使用条款](https://openai.com/policies/terms-of-use)以及**法律法规**的情况下使用，不得用于非法用途。
 > - 本项目仅供个人学习使用，不保证稳定性，且不提供任何技术支持。
+> - 使用者必须在遵循 OpenAI 的[使用条款](https://openai.com/policies/terms-of-use)以及**法律法规**的情况下使用，不得用于非法用途。
 > - 根据[《生成式人工智能服务管理暂行办法》](http://www.cac.gov.cn/2023-07/13/c_1690898327029107.htm)的要求，请勿对中国地区公众提供一切未经备案的生成式人工智能服务。
 
 ## 📚 文档
@@ -46,35 +46,32 @@
 
 ## ✨ 主要特性
 
-New API提供了丰富的功能，详细特性请参考[维基百科-特性说明](https://docs.newapi.pro/wiki/features-introduction)：
+New API提供了丰富的功能，详细特性请参考[特性说明](https://docs.newapi.pro/wiki/features-introduction)：
 
 1. 🎨 全新的UI界面
 2. 🌍 多语言支持
-3. 🎨 支持[Midjourney-Proxy(Plus)](https://github.com/novicezk/midjourney-proxy)接口，[对接文档](https://docs.newapi.pro/api/relay/image/midjourney)
-4. 💰 支持在线充值功能（易支付）
-5. 🔍 支持用key查询使用额度（配合[neko-api-key-tool](https://github.com/Calcium-Ion/neko-api-key-tool)）
-6. 📑 分页支持选择每页显示数量
-7. 🔄 兼容原版One API的数据库
-8. 💵 支持模型按次数收费
-9. ⚖️ 支持渠道加权随机
-10. 📈 数据看板（控制台）
-11. 🔒 可设置令牌能调用的模型
-12. 🤖 支持Telegram授权登录
-13. 🎵 支持[Suno API](https://github.com/Suno-API/Suno-API)接口，[接口文档](https://docs.newapi.pro/api/suno-music)
-14. 🔄 支持Rerank模型（Cohere和Jina），[接口文档](https://docs.newapi.pro/api/jinaai-rerank)
-15. ⚡ 支持OpenAI Realtime API（包括Azure渠道），[接口文档](https://docs.newapi.pro/api/openai-realtime)
-16. ⚡ 支持Claude Messages 格式，[接口文档](https://docs.newapi.pro/api/anthropic-chat)
-17. 支持使用路由/chat2link进入聊天界面
-18. 🧠 支持通过模型名称后缀设置 reasoning effort：
+3. 💰 支持在线充值功能（易支付）
+4. 🔍 支持用key查询使用额度（配合[neko-api-key-tool](https://github.com/Calcium-Ion/neko-api-key-tool)）
+5. 🔄 兼容原版One API的数据库
+6. 💵 支持模型按次数收费
+7. ⚖️ 支持渠道加权随机
+8. 📈 数据看板（控制台）
+9. 🔒 令牌分组、模型限制
+10. 🤖 支持更多授权登陆方式（LinuxDO,Telegram、OIDC）
+11. 🔄 支持Rerank模型（Cohere和Jina），[接口文档](https://docs.newapi.pro/api/jinaai-rerank)
+12. ⚡ 支持OpenAI Realtime API（包括Azure渠道），[接口文档](https://docs.newapi.pro/api/openai-realtime)
+13. ⚡ 支持Claude Messages 格式，[接口文档](https://docs.newapi.pro/api/anthropic-chat)
+14. 支持使用路由/chat2link进入聊天界面
+15. 🧠 支持通过模型名称后缀设置 reasoning effort：
     1. OpenAI o系列模型
         - 添加后缀 `-high` 设置为 high reasoning effort (例如: `o3-mini-high`)
         - 添加后缀 `-medium` 设置为 medium reasoning effort (例如: `o3-mini-medium`)
         - 添加后缀 `-low` 设置为 low reasoning effort (例如: `o3-mini-low`)
     2. Claude 思考模型
         - 添加后缀 `-thinking` 启用思考模式 (例如: `claude-3-7-sonnet-20250219-thinking`)
-19. 🔄 思考转内容功能
-20. 🔄 模型限流功能
-20. 💰 缓存计费支持，开启后可以在缓存命中时按照设定的比例计费：
+16. 🔄 思考转内容功能
+17. 🔄 针对用户的模型限流功能
+18. 💰 缓存计费支持，开启后可以在缓存命中时按照设定的比例计费：
     1. 在 `系统设置-运营设置` 中设置 `提示缓存倍率` 选项
     2. 在渠道中设置 `提示缓存倍率`，范围 0-1，例如设置为 0.5 表示缓存命中时按照 50% 计费
     3. 支持的渠道：
@@ -88,12 +85,12 @@ New API提供了丰富的功能，详细特性请参考[维基百科-特性说
 此版本支持多种模型，详情请参考[接口文档-中继接口](https://docs.newapi.pro/api)：
 
 1. 第三方模型 **gpts** （gpt-4-gizmo-*）
-2. [Midjourney-Proxy(Plus)](https://github.com/novicezk/midjourney-proxy)接口，[接口文档](https://docs.newapi.pro/api/midjourney-proxy-image)
-3. 自定义渠道，支持填入完整调用地址
-4. [Suno API](https://github.com/Suno-API/Suno-API)接口，[接口文档](https://docs.newapi.pro/api/suno-music)
+2. 第三方渠道[Midjourney-Proxy(Plus)](https://github.com/novicezk/midjourney-proxy)接口，[接口文档](https://docs.newapi.pro/api/midjourney-proxy-image)
+3. 第三方渠道[Suno API](https://github.com/Suno-API/Suno-API)接口，[接口文档](https://docs.newapi.pro/api/suno-music)
+4. 自定义渠道，支持填入完整调用地址
 5. Rerank模型（[Cohere](https://cohere.ai/)和[Jina](https://jina.ai/)），[接口文档](https://docs.newapi.pro/api/jinaai-rerank)
 6. Claude Messages 格式，[接口文档](https://docs.newapi.pro/api/anthropic-chat)
-7. Dify
+7. Dify，当前仅支持chatflow
 
 ## 环境变量配置
 
@@ -168,9 +165,6 @@ docker run --name new-api -d --restart always -p 3000:3000 -e SQL_DSN="root:1234
 
 - [聊天接口（Chat）](https://docs.newapi.pro/api/openai-chat)
 - [图像接口（Image）](https://docs.newapi.pro/api/openai-image)
-- [Midjourney接口](https://docs.newapi.pro/api/midjourney-proxy-image)
-- [音乐接口（Music）](https://docs.newapi.pro/api/relay/music)
-- [Suno接口](https://docs.newapi.pro/api/suno-music)
 - [重排序接口（Rerank）](https://docs.newapi.pro/api/jinaai-rerank)
 - [实时对话接口（Realtime）](https://docs.newapi.pro/api/openai-realtime)
 - [Claude聊天接口（messages）](https://docs.newapi.pro/api/anthropic-chat)

dto/claude.go

@@ -183,7 +183,7 @@ type ClaudeResponse struct {
 	Completion   string               `json:"completion,omitempty"`
 	StopReason   string               `json:"stop_reason,omitempty"`
 	Model        string               `json:"model,omitempty"`
-	Error        ClaudeError          `json:"error,omitempty"`
+	Error        *ClaudeError         `json:"error,omitempty"`
 	Usage        *ClaudeUsage         `json:"usage,omitempty"`
 	Index        *int                 `json:"index,omitempty"`
 	ContentBlock *ClaudeMediaMessage  `json:"content_block,omitempty"`

dto/rerank.go

@@ -5,18 +5,29 @@ type RerankRequest struct {
 	Query           string `json:"query"`
 	Model           string `json:"model"`
 	TopN            int    `json:"top_n"`
-	ReturnDocuments bool   `json:"return_documents,omitempty"`
+	ReturnDocuments *bool  `json:"return_documents,omitempty"`
 	MaxChunkPerDoc  int    `json:"max_chunk_per_doc,omitempty"`
 	OverLapTokens   int    `json:"overlap_tokens,omitempty"`
 }
 
-type RerankResponseDocument struct {
+func (r *RerankRequest) GetReturnDocuments() bool {
+	if r.ReturnDocuments == nil {
+		return false
+	}
+	return *r.ReturnDocuments
+}
+
+type RerankResponseResult struct {
 	Document       any     `json:"document,omitempty"`
 	Index          int     `json:"index"`
 	RelevanceScore float64 `json:"relevance_score"`
 }
 
-type RerankResponse struct {
-	Results []RerankResponseDocument `json:"results"`
-	Usage   Usage                    `json:"usage"`
+type RerankDocument struct {
+	Text any `json:"text"`
+}
+
+type RerankResponse struct {
+	Results []RerankResponseResult `json:"results"`
+	Usage   Usage                  `json:"usage"`
 }

relay/channel/aws/adaptor.go

@@ -21,6 +21,8 @@ type Adaptor struct {
 }
 
 func (a *Adaptor) ConvertClaudeRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.ClaudeRequest) (any, error) {
+	c.Set("request_model", request.Model)
+	c.Set("converted_request", request)
 	return request, nil
 }
 

relay/channel/aws/constants.go

@@ -13,4 +13,41 @@ var awsModelIDMap = map[string]string{
 	"claude-3-7-sonnet-20250219": "anthropic.claude-3-7-sonnet-20250219-v1:0",
 }
 
+var awsModelCanCrossRegionMap = map[string]map[string]bool{
+	"anthropic.claude-3-sonnet-20240229-v1:0": {
+		"us": true,
+		"eu": true,
+		"ap": true,
+	},
+	"anthropic.claude-3-opus-20240229-v1:0": {
+		"us": true,
+	},
+	"anthropic.claude-3-haiku-20240307-v1:0": {
+		"us": true,
+		"eu": true,
+		"ap": true,
+	},
+	"anthropic.claude-3-5-sonnet-20240620-v1:0": {
+		"us": true,
+		"eu": true,
+		"ap": true,
+	},
+	"anthropic.claude-3-5-sonnet-20241022-v2:0": {
+		"us": true,
+		"ap": true,
+	},
+	"anthropic.claude-3-5-haiku-20241022-v1:0": {
+		"us": true,
+	},
+	"anthropic.claude-3-7-sonnet-20250219-v1:0": {
+		"us": true,
+	},
+}
+
+var awsRegionCrossModelPrefixMap = map[string]string{
+	"us": "us",
+	"eu": "eu",
+	"ap": "apac",
+}
+
 var ChannelName = "aws"

relay/channel/aws/relay-aws.go

@@ -43,6 +43,28 @@ func wrapErr(err error) *dto.OpenAIErrorWithStatusCode {
 	}
 }
 
+func awsRegionPrefix(awsRegionId string) string {
+	parts := strings.Split(awsRegionId, "-")
+	regionPrefix := ""
+	if len(parts) > 0 {
+		regionPrefix = parts[0]
+	}
+	return regionPrefix
+}
+
+func awsModelCanCrossRegion(awsModelId, awsRegionPrefix string) bool {
+	regionSet, exists := awsModelCanCrossRegionMap[awsModelId]
+	return exists && regionSet[awsRegionPrefix]
+}
+
+func awsModelCrossRegion(awsModelId, awsRegionPrefix string) string {
+	modelPrefix, find := awsRegionCrossModelPrefixMap[awsRegionPrefix]
+	if !find {
+		return awsModelId
+	}
+	return modelPrefix + "." + awsModelId
+}
+
 func awsModelID(requestModel string) (string, error) {
 	if awsModelID, ok := awsModelIDMap[requestModel]; ok {
 		return awsModelID, nil
@@ -62,6 +84,12 @@ func awsHandler(c *gin.Context, info *relaycommon.RelayInfo, requestMode int) (*
 		return wrapErr(errors.Wrap(err, "awsModelID")), nil
 	}
 
+	awsRegionPrefix := awsRegionPrefix(awsCli.Options().Region)
+	canCrossRegion := awsModelCanCrossRegion(awsModelId, awsRegionPrefix)
+	if canCrossRegion {
+		awsModelId = awsModelCrossRegion(awsModelId, awsRegionPrefix)
+	}
+
 	awsReq := &bedrockruntime.InvokeModelInput{
 		ModelId:     aws.String(awsModelId),
 		Accept:      aws.String("application/json"),

relay/channel/claude/relay-claude.go

@@ -485,7 +485,7 @@ func HandleStreamResponseData(c *gin.Context, info *relaycommon.RelayInfo, claud
 		common.SysError("error unmarshalling stream response: " + err.Error())
 		return service.OpenAIErrorWrapper(err, "stream_response_error", http.StatusInternalServerError)
 	}
-	if claudeResponse.Error.Type != "" {
+	if claudeResponse.Error != nil && claudeResponse.Error.Type != "" {
 		return &dto.OpenAIErrorWithStatusCode{
 			Error: dto.OpenAIError{
 				Code:    "stream_response_error",
@@ -598,7 +598,7 @@ func HandleClaudeResponseData(c *gin.Context, info *relaycommon.RelayInfo, claud
 	if err != nil {
 		return service.OpenAIErrorWrapper(err, "unmarshal_claude_response_failed", http.StatusInternalServerError)
 	}
-	if claudeResponse.Error.Type != "" {
+	if claudeResponse.Error != nil && claudeResponse.Error.Type != "" {
 		return &dto.OpenAIErrorWithStatusCode{
 			Error: dto.OpenAIError{
 				Message: claudeResponse.Error.Message,

relay/channel/cohere/dto.go

@@ -40,8 +40,8 @@ type CohereRerankRequest struct {
 }
 
 type CohereRerankResponseResult struct {
-	Results []dto.RerankResponseDocument `json:"results"`
-	Meta    CohereMeta                   `json:"meta"`
+	Results []dto.RerankResponseResult `json:"results"`
+	Meta    CohereMeta                 `json:"meta"`
 }
 
 type CohereMeta struct {

relay/channel/jina/adaptor.go

@@ -69,7 +69,7 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 
 func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage any, err *dto.OpenAIErrorWithStatusCode) {
 	if info.RelayMode == constant.RelayModeRerank {
-		err, usage = common_handler.RerankHandler(c, resp)
+		err, usage = common_handler.RerankHandler(c, info, resp)
 	} else if info.RelayMode == constant.RelayModeEmbeddings {
 		err, usage = openai.OpenaiHandler(c, resp, info)
 	}

relay/channel/openai/adaptor.go

@@ -262,7 +262,7 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
 	case constant.RelayModeImagesGenerations:
 		err, usage = OpenaiTTSHandler(c, resp, info)
 	case constant.RelayModeRerank:
-		err, usage = common_handler.RerankHandler(c, resp)
+		err, usage = common_handler.RerankHandler(c, info, resp)
 	default:
 		if info.IsStream {
 			err, usage = OaiStreamHandler(c, resp, info)

relay/channel/siliconflow/dto.go

@@ -12,6 +12,6 @@ type SFMeta struct {
 }
 
 type SFRerankResponse struct {
-	Results []dto.RerankResponseDocument `json:"results"`
-	Meta    SFMeta                       `json:"meta"`
+	Results []dto.RerankResponseResult `json:"results"`
+	Meta    SFMeta                     `json:"meta"`
 }

relay/channel/vertex/adaptor.go

@@ -39,8 +39,15 @@ type Adaptor struct {
 }
 
 func (a *Adaptor) ConvertClaudeRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.ClaudeRequest) (any, error) {
-	return request, nil
+	if v, ok := claudeModelMap[info.UpstreamModelName]; ok {
+		c.Set("request_model", v)
+	} else {
+		c.Set("request_model", request.Model)
+	}
+	vertexClaudeReq := copyRequest(request, anthropicVersion)
+	return vertexClaudeReq, nil
 }
+
 func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
 	//TODO implement me
 	return nil, errors.New("not implemented")

relay/channel/xinference/dto.go

@@ -0,0 +1,11 @@
+package xinference
+
+type XinRerankResponseDocument struct {
+	Document       string  `json:"document,omitempty"`
+	Index          int     `json:"index"`
+	RelevanceScore float64 `json:"relevance_score"`
+}
+
+type XinRerankResponse struct {
+	Results []XinRerankResponseDocument `json:"results"`
+}

relay/common/relay_info.go

@@ -33,6 +33,11 @@ const (
 	RelayFormatClaude = "claude"
 )
 
+type RerankerInfo struct {
+	Documents       []any
+	ReturnDocuments bool
+}
+
 type RelayInfo struct {
 	ChannelType       int
 	ChannelId         int
@@ -78,6 +83,7 @@ type RelayInfo struct {
 	SendResponseCount    int
 	ThinkingContentInfo
 	ClaudeConvertInfo
+	*RerankerInfo
 }
 
 // 定义支持流式选项的通道类型
@@ -111,6 +117,16 @@ func GenRelayInfoClaude(c *gin.Context) *RelayInfo {
 	return info
 }
 
+func GenRelayInfoRerank(c *gin.Context, req *dto.RerankRequest) *RelayInfo {
+	info := GenRelayInfo(c)
+	info.RelayMode = relayconstant.RelayModeRerank
+	info.RerankerInfo = &RerankerInfo{
+		Documents:       req.Documents,
+		ReturnDocuments: req.GetReturnDocuments(),
+	}
+	return info
+}
+
 func GenRelayInfo(c *gin.Context) *RelayInfo {
 	channelType := c.GetInt("channel_type")
 	channelId := c.GetInt("channel_id")

relay/common_handler/rerank.go

@@ -1,15 +1,17 @@
 package common_handler
 
 import (
-	"encoding/json"
 	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
+	"one-api/common"
 	"one-api/dto"
+	"one-api/relay/channel/xinference"
+	relaycommon "one-api/relay/common"
 	"one-api/service"
 )
 
-func RerankHandler(c *gin.Context, resp *http.Response) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
+func RerankHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
 	responseBody, err := io.ReadAll(resp.Body)
 	if err != nil {
 		return service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
@@ -18,18 +20,49 @@ func RerankHandler(c *gin.Context, resp *http.Response) (*dto.OpenAIErrorWithSta
 	if err != nil {
 		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
 	}
+	if common.DebugEnabled {
+		println("reranker response body: ", string(responseBody))
+	}
 	var jinaResp dto.RerankResponse
-	err = json.Unmarshal(responseBody, &jinaResp)
-	if err != nil {
-		return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	if info.ChannelType == common.ChannelTypeXinference {
+		var xinRerankResponse xinference.XinRerankResponse
+		err = common.DecodeJson(responseBody, &xinRerankResponse)
+		if err != nil {
+			return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+		}
+		jinaRespResults := make([]dto.RerankResponseResult, len(xinRerankResponse.Results))
+		for i, result := range xinRerankResponse.Results {
+			respResult := dto.RerankResponseResult{
+				Index:          result.Index,
+				RelevanceScore: result.RelevanceScore,
+			}
+			if info.ReturnDocuments {
+				var document any
+				if result.Document == "" {
+					document = info.Documents[result.Index]
+				} else {
+					document = result.Document
+				}
+				respResult.Document = document
+			}
+			jinaRespResults[i] = respResult
+		}
+		jinaResp = dto.RerankResponse{
+			Results: jinaRespResults,
+			Usage: dto.Usage{
+				PromptTokens: info.PromptTokens,
+				TotalTokens:  info.PromptTokens,
+			},
+		}
+	} else {
+		err = common.DecodeJson(responseBody, &jinaResp)
+		if err != nil {
+			return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+		}
+		jinaResp.Usage.PromptTokens = jinaResp.Usage.TotalTokens
 	}
 
-	jsonResponse, err := json.Marshal(jinaResp)
-	if err != nil {
-		return service.OpenAIErrorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
-	}
 	c.Writer.Header().Set("Content-Type", "application/json")
-	c.Writer.WriteHeader(resp.StatusCode)
-	_, err = c.Writer.Write(jsonResponse)
+	c.JSON(http.StatusOK, jinaResp)
 	return nil, &jinaResp.Usage
 }

relay/relay_rerank.go

@@ -25,7 +25,6 @@ func getRerankPromptToken(rerankRequest dto.RerankRequest) int {
 }
 
 func RerankHelper(c *gin.Context, relayMode int) (openaiErr *dto.OpenAIErrorWithStatusCode) {
-	relayInfo := relaycommon.GenRelayInfo(c)
 
 	var rerankRequest *dto.RerankRequest
 	err := common.UnmarshalBodyReusable(c, &rerankRequest)
@@ -33,6 +32,9 @@ func RerankHelper(c *gin.Context, relayMode int) (openaiErr *dto.OpenAIErrorWith
 		common.LogError(c, fmt.Sprintf("getAndValidateTextRequest failed: %s", err.Error()))
 		return service.OpenAIErrorWrapperLocal(err, "invalid_text_request", http.StatusBadRequest)
 	}
+
+	relayInfo := relaycommon.GenRelayInfoRerank(c, rerankRequest)
+
 	if rerankRequest.Query == "" {
 		return service.OpenAIErrorWrapperLocal(fmt.Errorf("query is empty"), "invalid_query", http.StatusBadRequest)
 	}