github/new-api
1
0
Fork 0
mirror of https://github.com/QuantumNous/new-api.git synced 2026-04-19 07:57:28 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'main' into alpha

Browse Source
This commit is contained in:
Seefs
2025-09-13 13:14:34 +08:00
committed by GitHub
parent a3b8a1998a 406a3e4dca
commit 6628fd9181
45 changed files with 1032 additions and 514 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
relay/channel/aws/adaptor.go
View File

@@ -60,7 +60,16 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
if request == nil {
return nil, errors.New("request is nil")
}
// 检查是否为Nova模型
if isNovaModel(request.Model) {
novaReq := convertToNovaRequest(request)
c.Set("request_model", request.Model)
c.Set("converted_request", novaReq)
c.Set("is_nova_model", true)
return novaReq, nil
}
// 原有的Claude模型处理逻辑
var claudeReq *dto.ClaudeRequest
var err error
claudeReq, err = claude.RequestOpenAI2ClaudeMessage(c, *request)
@@ -69,6 +78,7 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
}
c.Set("request_model", claudeReq.Model)
c.Set("converted_request", claudeReq)
c.Set("is_nova_model", false)
return claudeReq, err
}

34
relay/channel/aws/constants.go
View File

@@ -1,5 +1,7 @@
package aws
import "strings"
var awsModelIDMap = map[string]string{
"claude-instant-1.2": "anthropic.claude-instant-v1",
"claude-2.0": "anthropic.claude-v2",
@@ -14,6 +16,11 @@ var awsModelIDMap = map[string]string{
"claude-sonnet-4-20250514": "anthropic.claude-sonnet-4-20250514-v1:0",
"claude-opus-4-20250514": "anthropic.claude-opus-4-20250514-v1:0",
"claude-opus-4-1-20250805": "anthropic.claude-opus-4-1-20250805-v1:0",
// Nova models
"nova-micro-v1:0": "amazon.nova-micro-v1:0",
"nova-lite-v1:0": "amazon.nova-lite-v1:0",
"nova-pro-v1:0": "amazon.nova-pro-v1:0",
"nova-premier-v1:0": "amazon.nova-premier-v1:0",
}
var awsModelCanCrossRegionMap = map[string]map[string]bool{
@@ -58,7 +65,27 @@ var awsModelCanCrossRegionMap = map[string]map[string]bool{
"anthropic.claude-opus-4-1-20250805-v1:0": {
"us": true,
},
}
// Nova models - all support three major regions
"amazon.nova-micro-v1:0": {
"us": true,
"eu": true,
"apac": true,
},
"amazon.nova-lite-v1:0": {
"us": true,
"eu": true,
"apac": true,
},
"amazon.nova-pro-v1:0": {
"us": true,
"eu": true,
"apac": true,
},
"amazon.nova-premier-v1:0": {
"us": true,
"eu": true,
"apac": true,
}}
var awsRegionCrossModelPrefixMap = map[string]string{
"us": "us",
@@ -67,3 +94,8 @@ var awsRegionCrossModelPrefixMap = map[string]string{
}
var ChannelName = "aws"
// 判断是否为Nova模型
func isNovaModel(modelId string) bool {
return strings.HasPrefix(modelId, "nova-")
}

89
relay/channel/aws/dto.go
View File

@@ -34,3 +34,92 @@ func copyRequest(req *dto.ClaudeRequest) *AwsClaudeRequest {
Thinking: req.Thinking,
}
}
// NovaMessage Nova模型使用messages-v1格式
type NovaMessage struct {
Role string `json:"role"`
Content []NovaContent `json:"content"`
}
type NovaContent struct {
Text string `json:"text"`
}
type NovaRequest struct {
SchemaVersion string `json:"schemaVersion"` // 请求版本，例如 "1.0"
Messages []NovaMessage `json:"messages"` // 对话消息列表
InferenceConfig *NovaInferenceConfig `json:"inferenceConfig,omitempty"` // 推理配置，可选
}
type NovaInferenceConfig struct {
MaxTokens int `json:"maxTokens,omitempty"` // 最大生成的 token 数
Temperature float64 `json:"temperature,omitempty"` // 随机性 (默认 0.7, 范围 0-1)
TopP float64 `json:"topP,omitempty"` // nucleus sampling (默认 0.9, 范围 0-1)
TopK int `json:"topK,omitempty"` // 限制候选 token 数 (默认 50, 范围 0-128)
StopSequences []string `json:"stopSequences,omitempty"` // 停止生成的序列
}
// 转换OpenAI请求为Nova格式
func convertToNovaRequest(req *dto.GeneralOpenAIRequest) *NovaRequest {
novaMessages := make([]NovaMessage, len(req.Messages))
for i, msg := range req.Messages {
novaMessages[i] = NovaMessage{
Role: msg.Role,
Content: []NovaContent{{Text: msg.StringContent()}},
}
}
novaReq := &NovaRequest{
SchemaVersion: "messages-v1",
Messages: novaMessages,
}
// 设置推理配置
if req.MaxTokens != 0 || (req.Temperature != nil && *req.Temperature != 0) || req.TopP != 0 || req.TopK != 0 || req.Stop != nil {
novaReq.InferenceConfig = &NovaInferenceConfig{}
if req.MaxTokens != 0 {
novaReq.InferenceConfig.MaxTokens = int(req.MaxTokens)
}
if req.Temperature != nil && *req.Temperature != 0 {
novaReq.InferenceConfig.Temperature = *req.Temperature
}
if req.TopP != 0 {
novaReq.InferenceConfig.TopP = req.TopP
}
if req.TopK != 0 {
novaReq.InferenceConfig.TopK = req.TopK
}
if req.Stop != nil {
if stopSequences := parseStopSequences(req.Stop); len(stopSequences) > 0 {
novaReq.InferenceConfig.StopSequences = stopSequences
}
}
}
return novaReq
}
// parseStopSequences 解析停止序列，支持字符串或字符串数组
func parseStopSequences(stop any) []string {
if stop == nil {
return nil
}
switch v := stop.(type) {
case string:
if v != "" {
return []string{v}
}
case []string:
return v
case []interface{}:
var sequences []string
for _, item := range v {
if str, ok := item.(string); ok && str != "" {
sequences = append(sequences, str)
}
}
return sequences
}
return nil
}

91
relay/channel/aws/relay-aws.go
View File

@@ -1,6 +1,7 @@
package aws
import (
"encoding/json"
"fmt"
"net/http"
"one-api/common"
@@ -93,7 +94,19 @@ func awsHandler(c *gin.Context, info *relaycommon.RelayInfo, requestMode int) (*
}
awsModelId := awsModelID(c.GetString("request_model"))
// 检查是否为Nova模型
isNova, _ := c.Get("is_nova_model")
if isNova == true {
// Nova模型也支持跨区域
awsRegionPrefix := awsRegionPrefix(awsCli.Options().Region)
canCrossRegion := awsModelCanCrossRegion(awsModelId, awsRegionPrefix)
if canCrossRegion {
awsModelId = awsModelCrossRegion(awsModelId, awsRegionPrefix)
}
return handleNovaRequest(c, awsCli, info, awsModelId)
}
// 原有的Claude处理逻辑
awsRegionPrefix := awsRegionPrefix(awsCli.Options().Region)
canCrossRegion := awsModelCanCrossRegion(awsModelId, awsRegionPrefix)
if canCrossRegion {
@@ -130,7 +143,12 @@ func awsHandler(c *gin.Context, info *relaycommon.RelayInfo, requestMode int) (*
Usage: &dto.Usage{},
}
handlerErr := claude.HandleClaudeResponseData(c, info, claudeInfo, awsResp.Body, RequestModeMessage)
// 复制上游 Content-Type 到客户端响应头
if awsResp.ContentType != nil && *awsResp.ContentType != "" {
c.Writer.Header().Set("Content-Type", *awsResp.ContentType)
}
handlerErr := claude.HandleClaudeResponseData(c, info, claudeInfo, nil, awsResp.Body, RequestModeMessage)
if handlerErr != nil {
return handlerErr, nil
}
@@ -204,3 +222,74 @@ func awsStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.Rel
claude.HandleStreamFinalResponse(c, info, claudeInfo, RequestModeMessage)
return nil, claudeInfo.Usage
}
// Nova模型处理函数
func handleNovaRequest(c *gin.Context, awsCli *bedrockruntime.Client, info *relaycommon.RelayInfo, awsModelId string) (*types.NewAPIError, *dto.Usage) {
novaReq_, ok := c.Get("converted_request")
if !ok {
return types.NewError(errors.New("nova request not found"), types.ErrorCodeInvalidRequest), nil
}
novaReq := novaReq_.(*NovaRequest)
// 使用InvokeModel API，但使用Nova格式的请求体
awsReq := &bedrockruntime.InvokeModelInput{
ModelId: aws.String(awsModelId),
Accept: aws.String("application/json"),
ContentType: aws.String("application/json"),
}
reqBody, err := json.Marshal(novaReq)
if err != nil {
return types.NewError(errors.Wrap(err, "marshal nova request"), types.ErrorCodeBadResponseBody), nil
}
awsReq.Body = reqBody
awsResp, err := awsCli.InvokeModel(c.Request.Context(), awsReq)
if err != nil {
return types.NewError(errors.Wrap(err, "InvokeModel"), types.ErrorCodeChannelAwsClientError), nil
}
// 解析Nova响应
var novaResp struct {
Output struct {
Message struct {
Content []struct {
Text string `json:"text"`
} `json:"content"`
} `json:"message"`
} `json:"output"`
Usage struct {
InputTokens int `json:"inputTokens"`
OutputTokens int `json:"outputTokens"`
TotalTokens int `json:"totalTokens"`
} `json:"usage"`
}
if err := json.Unmarshal(awsResp.Body, &novaResp); err != nil {
return types.NewError(errors.Wrap(err, "unmarshal nova response"), types.ErrorCodeBadResponseBody), nil
}
// 构造OpenAI格式响应
response := dto.OpenAITextResponse{
Id: helper.GetResponseID(c),
Object: "chat.completion",
Created: common.GetTimestamp(),
Model: info.UpstreamModelName,
Choices: []dto.OpenAITextResponseChoice{{
Index: 0,
Message: dto.Message{
Role: "assistant",
Content: novaResp.Output.Message.Content[0].Text,
},
FinishReason: "stop",
}},
Usage: dto.Usage{
PromptTokens: novaResp.Usage.InputTokens,
CompletionTokens: novaResp.Usage.OutputTokens,
TotalTokens: novaResp.Usage.TotalTokens,
},
}
c.JSON(http.StatusOK, response)
return nil, &response.Usage
}

12
relay/channel/claude/relay-claude.go
View File

@@ -276,7 +276,7 @@ func RequestOpenAI2ClaudeMessage(c *gin.Context, textRequest dto.GeneralOpenAIRe
isFirstMessage := true
// 初始化system消息数组，用于累积多个system消息
var systemMessages []dto.ClaudeMediaMessage
for _, message := range formatMessages {
if message.Role == "system" {
// 根据Claude API规范，system字段使用数组格式更有通用性
@@ -401,12 +401,12 @@ func RequestOpenAI2ClaudeMessage(c *gin.Context, textRequest dto.GeneralOpenAIRe
claudeMessages = append(claudeMessages, claudeMessage)
}
}
// 设置累积的system消息
if len(systemMessages) > 0 {
claudeRequest.System = systemMessages
}
claudeRequest.Prompt = ""
claudeRequest.Messages = claudeMessages
return &claudeRequest, nil
@@ -716,7 +716,7 @@ func ClaudeStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.
return claudeInfo.Usage, nil
}
func HandleClaudeResponseData(c *gin.Context, info *relaycommon.RelayInfo, claudeInfo *ClaudeResponseInfo, data []byte, requestMode int) *types.NewAPIError {
func HandleClaudeResponseData(c *gin.Context, info *relaycommon.RelayInfo, claudeInfo *ClaudeResponseInfo, httpResp *http.Response, data []byte, requestMode int) *types.NewAPIError {
var claudeResponse dto.ClaudeResponse
err := common.Unmarshal(data, &claudeResponse)
if err != nil {
@@ -754,7 +754,7 @@ func HandleClaudeResponseData(c *gin.Context, info *relaycommon.RelayInfo, claud
c.Set("claude_web_search_requests", claudeResponse.Usage.ServerToolUse.WebSearchRequests)
}
service.IOCopyBytesGracefully(c, nil, responseData)
service.IOCopyBytesGracefully(c, httpResp, responseData)
return nil
}
@@ -775,7 +775,7 @@ func ClaudeHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayI
if common.DebugEnabled {
println("responseBody: ", string(responseBody))
}
handleErr := HandleClaudeResponseData(c, info, claudeInfo, responseBody, requestMode)
handleErr := HandleClaudeResponseData(c, info, claudeInfo, resp, responseBody, requestMode)
if handleErr != nil {
return nil, handleErr
}

36
relay/channel/gemini/relay-gemini-native.go
View File

@@ -46,32 +46,6 @@ func GeminiTextGenerationHandler(c *gin.Context, info *relaycommon.RelayInfo, re
usage.CompletionTokenDetails.ReasoningTokens = geminiResponse.UsageMetadata.ThoughtsTokenCount
if strings.HasPrefix(info.UpstreamModelName, "gemini-2.5-flash-image-preview") {
imageOutputCounts := 0
for _, candidate := range geminiResponse.Candidates {
for _, part := range candidate.Content.Parts {
if part.InlineData != nil && strings.HasPrefix(part.InlineData.MimeType, "image/") {
imageOutputCounts++
}
}
}
if imageOutputCounts != 0 {
usage.CompletionTokens = usage.CompletionTokens - imageOutputCounts*1290
usage.TotalTokens = usage.TotalTokens - imageOutputCounts*1290
c.Set("gemini_image_tokens", imageOutputCounts*1290)
}
}
// if strings.HasPrefix(info.UpstreamModelName, "gemini-2.5-flash-image-preview") {
// for _, detail := range geminiResponse.UsageMetadata.CandidatesTokensDetails {
// if detail.Modality == "IMAGE" {
// usage.CompletionTokens = usage.CompletionTokens - detail.TokenCount
// usage.TotalTokens = usage.TotalTokens - detail.TokenCount
// c.Set("gemini_image_tokens", detail.TokenCount)
// }
// }
// }
for _, detail := range geminiResponse.UsageMetadata.PromptTokensDetails {
if detail.Modality == "AUDIO" {
usage.PromptTokensDetails.AudioTokens = detail.TokenCount
@@ -162,16 +136,6 @@ func GeminiTextGenerationStreamHandler(c *gin.Context, info *relaycommon.RelayIn
usage.PromptTokensDetails.TextTokens = detail.TokenCount
}
}
if strings.HasPrefix(info.UpstreamModelName, "gemini-2.5-flash-image-preview") {
for _, detail := range geminiResponse.UsageMetadata.CandidatesTokensDetails {
if detail.Modality == "IMAGE" {
usage.CompletionTokens = usage.CompletionTokens - detail.TokenCount
usage.TotalTokens = usage.TotalTokens - detail.TokenCount
c.Set("gemini_image_tokens", detail.TokenCount)
}
}
}
}
// 直接发送 GeminiChatResponse 响应

26
relay/channel/task/jimeng/adaptor.go
View File

@@ -18,7 +18,6 @@ import (
"github.com/gin-gonic/gin"
"github.com/pkg/errors"
"one-api/common"
"one-api/constant"
"one-api/dto"
"one-api/relay/channel"
@@ -89,22 +88,7 @@ func (a *TaskAdaptor) Init(info *relaycommon.RelayInfo) {
// ValidateRequestAndSetAction parses body, validates fields and sets default action.
func (a *TaskAdaptor) ValidateRequestAndSetAction(c *gin.Context, info *relaycommon.RelayInfo) (taskErr *dto.TaskError) {
// Accept only POST /v1/video/generations as "generate" action.
action := constant.TaskActionGenerate
info.Action = action
req := relaycommon.TaskSubmitReq{}
if err := common.UnmarshalBodyReusable(c, &req); err != nil {
taskErr = service.TaskErrorWrapperLocal(err, "invalid_request", http.StatusBadRequest)
return
}
if strings.TrimSpace(req.Prompt) == "" {
taskErr = service.TaskErrorWrapperLocal(fmt.Errorf("prompt is required"), "invalid_request", http.StatusBadRequest)
return
}
// Store into context for later usage
c.Set("task_request", req)
return nil
return relaycommon.ValidateBasicTaskRequest(c, info, constant.TaskActionGenerate)
}
// BuildRequestURL constructs the upstream URL.
@@ -334,11 +318,11 @@ func (a *TaskAdaptor) convertToRequestPayload(req *relaycommon.TaskSubmitReq) (*
}
// Handle one-of image_urls or binary_data_base64
if req.Image != "" {
if strings.HasPrefix(req.Image, "http") {
r.ImageUrls = []string{req.Image}
if req.HasImage() {
if strings.HasPrefix(req.Images[0], "http") {
r.ImageUrls = req.Images
} else {
r.BinaryDataBase64 = []string{req.Image}
r.BinaryDataBase64 = req.Images
}
}
metadata := req.Metadata

34
relay/channel/task/kling/adaptor.go
View File

@@ -16,7 +16,6 @@ import (
"github.com/golang-jwt/jwt"
"github.com/pkg/errors"
"one-api/common"
"one-api/constant"
"one-api/dto"
"one-api/relay/channel"
@@ -28,16 +27,6 @@ import (
// Request / Response structures
// ============================
type SubmitReq struct {
Prompt string `json:"prompt"`
Model string `json:"model,omitempty"`
Mode string `json:"mode,omitempty"`
Image string `json:"image,omitempty"`
Size string `json:"size,omitempty"`
Duration int `json:"duration,omitempty"`
Metadata map[string]interface{} `json:"metadata,omitempty"`
}
type TrajectoryPoint struct {
X int `json:"x"`
Y int `json:"y"`
@@ -121,23 +110,8 @@ func (a *TaskAdaptor) Init(info *relaycommon.RelayInfo) {
// ValidateRequestAndSetAction parses body, validates fields and sets default action.
func (a *TaskAdaptor) ValidateRequestAndSetAction(c *gin.Context, info *relaycommon.RelayInfo) (taskErr *dto.TaskError) {
// Accept only POST /v1/video/generations as "generate" action.
action := constant.TaskActionGenerate
info.Action = action
var req SubmitReq
if err := common.UnmarshalBodyReusable(c, &req); err != nil {
taskErr = service.TaskErrorWrapperLocal(err, "invalid_request", http.StatusBadRequest)
return
}
if strings.TrimSpace(req.Prompt) == "" {
taskErr = service.TaskErrorWrapperLocal(fmt.Errorf("prompt is required"), "invalid_request", http.StatusBadRequest)
return
}
// Store into context for later usage
c.Set("task_request", req)
return nil
// Use the standard validation method for TaskSubmitReq
return relaycommon.ValidateBasicTaskRequest(c, info, constant.TaskActionGenerate)
}
// BuildRequestURL constructs the upstream URL.
@@ -166,7 +140,7 @@ func (a *TaskAdaptor) BuildRequestBody(c *gin.Context, info *relaycommon.RelayIn
if !exists {
return nil, fmt.Errorf("request not found in context")
}
req := v.(SubmitReq)
req := v.(relaycommon.TaskSubmitReq)
body, err := a.convertToRequestPayload(&req)
if err != nil {
@@ -255,7 +229,7 @@ func (a *TaskAdaptor) GetChannelName() string {
// helpers
// ============================
func (a *TaskAdaptor) convertToRequestPayload(req *SubmitReq) (*requestPayload, error) {
func (a *TaskAdaptor) convertToRequestPayload(req *relaycommon.TaskSubmitReq) (*requestPayload, error) {
r := requestPayload{
Prompt: req.Prompt,
Image: req.Image,

33
relay/channel/task/vidu/adaptor.go
View File

@@ -23,16 +23,6 @@ import (
// Request / Response structures
// ============================
type SubmitReq struct {
Prompt string `json:"prompt"`
Model string `json:"model,omitempty"`
Mode string `json:"mode,omitempty"`
Image string `json:"image,omitempty"`
Size string `json:"size,omitempty"`
Duration int `json:"duration,omitempty"`
Metadata map[string]interface{} `json:"metadata,omitempty"`
}
type requestPayload struct {
Model string `json:"model"`
Images []string `json:"images"`
@@ -90,23 +80,8 @@ func (a *TaskAdaptor) Init(info *relaycommon.RelayInfo) {
}
func (a *TaskAdaptor) ValidateRequestAndSetAction(c *gin.Context, info *relaycommon.RelayInfo) *dto.TaskError {
var req SubmitReq
if err := c.ShouldBindJSON(&req); err != nil {
return service.TaskErrorWrapper(err, "invalid_request_body", http.StatusBadRequest)
}
if req.Prompt == "" {
return service.TaskErrorWrapperLocal(fmt.Errorf("prompt is required"), "missing_prompt", http.StatusBadRequest)
}
if req.Image != "" {
info.Action = constant.TaskActionGenerate
} else {
info.Action = constant.TaskActionTextGenerate
}
c.Set("task_request", req)
return nil
// Use the unified validation method for TaskSubmitReq with image-based action determination
return relaycommon.ValidateTaskRequestWithImageBinding(c, info)
}
func (a *TaskAdaptor) BuildRequestBody(c *gin.Context, _ *relaycommon.RelayInfo) (io.Reader, error) {
@@ -114,7 +89,7 @@ func (a *TaskAdaptor) BuildRequestBody(c *gin.Context, _ *relaycommon.RelayInfo)
if !exists {
return nil, fmt.Errorf("request not found in context")
}
req := v.(SubmitReq)
req := v.(relaycommon.TaskSubmitReq)
body, err := a.convertToRequestPayload(&req)
if err != nil {
@@ -211,7 +186,7 @@ func (a *TaskAdaptor) GetChannelName() string {
// helpers
// ============================
func (a *TaskAdaptor) convertToRequestPayload(req *SubmitReq) (*requestPayload, error) {
func (a *TaskAdaptor) convertToRequestPayload(req *relaycommon.TaskSubmitReq) (*requestPayload, error) {
var images []string
if req.Image != "" {
images = []string{req.Image}

121
relay/channel/vertex/adaptor.go
View File

@@ -6,6 +6,7 @@ import (
"fmt"
"io"
"net/http"
"one-api/common"
"one-api/dto"
"one-api/relay/channel"
"one-api/relay/channel/claude"
@@ -80,16 +81,64 @@ func (a *Adaptor) Init(info *relaycommon.RelayInfo) {
}
}
func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
adc := &Credentials{}
if err := json.Unmarshal([]byte(info.ApiKey), adc); err != nil {
return "", fmt.Errorf("failed to decode credentials file: %w", err)
}
func (a *Adaptor) getRequestUrl(info *relaycommon.RelayInfo, modelName, suffix string) (string, error) {
region := GetModelRegion(info.ApiVersion, info.OriginModelName)
a.AccountCredentials = *adc
if info.ChannelOtherSettings.VertexKeyType != dto.VertexKeyTypeAPIKey {
adc := &Credentials{}
if err := common.Unmarshal([]byte(info.ApiKey), adc); err != nil {
return "", fmt.Errorf("failed to decode credentials file: %w", err)
}
a.AccountCredentials = *adc
if a.RequestMode == RequestModeLlama {
return fmt.Sprintf(
"https://%s-aiplatform.googleapis.com/v1beta1/projects/%s/locations/%s/endpoints/openapi/chat/completions",
region,
adc.ProjectID,
region,
), nil
}
if region == "global" {
return fmt.Sprintf(
"https://aiplatform.googleapis.com/v1/projects/%s/locations/global/publishers/google/models/%s:%s",
adc.ProjectID,
modelName,
suffix,
), nil
} else {
return fmt.Sprintf(
"https://%s-aiplatform.googleapis.com/v1/projects/%s/locations/%s/publishers/google/models/%s:%s",
region,
adc.ProjectID,
region,
modelName,
suffix,
), nil
}
} else {
if region == "global" {
return fmt.Sprintf(
"https://aiplatform.googleapis.com/v1/publishers/google/models/%s:%s?key=%s",
modelName,
suffix,
info.ApiKey,
), nil
} else {
return fmt.Sprintf(
"https://%s-aiplatform.googleapis.com/v1/publishers/google/models/%s:%s?key=%s",
region,
modelName,
suffix,
info.ApiKey,
), nil
}
}
}
func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
suffix := ""
if a.RequestMode == RequestModeGemini {
if model_setting.GetGeminiSettings().ThinkingAdapterEnabled {
// 新增逻辑：处理 -thinking-<budget> 格式
if strings.Contains(info.UpstreamModelName, "-thinking-") {
@@ -111,24 +160,7 @@ func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
if strings.HasPrefix(info.UpstreamModelName, "imagen") {
suffix = "predict"
}
if region == "global" {
return fmt.Sprintf(
"https://aiplatform.googleapis.com/v1/projects/%s/locations/global/publishers/google/models/%s:%s",
adc.ProjectID,
info.UpstreamModelName,
suffix,
), nil
} else {
return fmt.Sprintf(
"https://%s-aiplatform.googleapis.com/v1/projects/%s/locations/%s/publishers/google/models/%s:%s",
region,
adc.ProjectID,
region,
info.UpstreamModelName,
suffix,
), nil
}
return a.getRequestUrl(info, info.UpstreamModelName, suffix)
} else if a.RequestMode == RequestModeClaude {
if info.IsStream {
suffix = "streamRawPredict?alt=sse"
@@ -139,42 +171,25 @@ func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
if v, ok := claudeModelMap[info.UpstreamModelName]; ok {
model = v
}
if region == "global" {
return fmt.Sprintf(
"https://aiplatform.googleapis.com/v1/projects/%s/locations/global/publishers/anthropic/models/%s:%s",
adc.ProjectID,
model,
suffix,
), nil
} else {
return fmt.Sprintf(
"https://%s-aiplatform.googleapis.com/v1/projects/%s/locations/%s/publishers/anthropic/models/%s:%s",
region,
adc.ProjectID,
region,
model,
suffix,
), nil
}
return a.getRequestUrl(info, model, suffix)
} else if a.RequestMode == RequestModeLlama {
return fmt.Sprintf(
"https://%s-aiplatform.googleapis.com/v1beta1/projects/%s/locations/%s/endpoints/openapi/chat/completions",
region,
adc.ProjectID,
region,
), nil
return a.getRequestUrl(info, "", "")
}
return "", errors.New("unsupported request mode")
}
func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *relaycommon.RelayInfo) error {
channel.SetupApiRequestHeader(info, c, req)
accessToken, err := getAccessToken(a, info)
if err != nil {
return err
if info.ChannelOtherSettings.VertexKeyType != dto.VertexKeyTypeAPIKey {
accessToken, err := getAccessToken(a, info)
if err != nil {
return err
}
req.Set("Authorization", "Bearer "+accessToken)
}
if a.AccountCredentials.ProjectID != "" {
req.Set("x-goog-user-project", a.AccountCredentials.ProjectID)
}
req.Set("Authorization", "Bearer "+accessToken)
req.Set("x-goog-user-project", a.AccountCredentials.ProjectID)
return nil
}

9
relay/common/relay_info.go
View File

@@ -481,11 +481,20 @@ type TaskSubmitReq struct {
Model string `json:"model,omitempty"`
Mode string `json:"mode,omitempty"`
Image string `json:"image,omitempty"`
Images []string `json:"images,omitempty"`
Size string `json:"size,omitempty"`
Duration int `json:"duration,omitempty"`
Metadata map[string]interface{} `json:"metadata,omitempty"`
}
func (t TaskSubmitReq) GetPrompt() string {
return t.Prompt
}
func (t TaskSubmitReq) HasImage() bool {
return len(t.Images) > 0
}
type TaskInfo struct {
Code int `json:"code"`
TaskID string `json:"task_id"`

80
relay/common/relay_utils.go
View File

@@ -2,12 +2,23 @@ package common
import (
"fmt"
"net/http"
"one-api/common"
"one-api/constant"
"one-api/dto"
"strings"
"github.com/gin-gonic/gin"
)
type HasPrompt interface {
GetPrompt() string
}
type HasImage interface {
HasImage() bool
}
func GetFullRequestURL(baseURL string, requestURL string, channelType int) string {
fullRequestURL := fmt.Sprintf("%s%s", baseURL, requestURL)
@@ -30,3 +41,72 @@ func GetAPIVersion(c *gin.Context) string {
}
return apiVersion
}
func createTaskError(err error, code string, statusCode int, localError bool) *dto.TaskError {
return &dto.TaskError{
Code: code,
Message: err.Error(),
StatusCode: statusCode,
LocalError: localError,
Error: err,
}
}
func storeTaskRequest(c *gin.Context, info *RelayInfo, action string, requestObj interface{}) {
info.Action = action
c.Set("task_request", requestObj)
}
func validatePrompt(prompt string) *dto.TaskError {
if strings.TrimSpace(prompt) == "" {
return createTaskError(fmt.Errorf("prompt is required"), "invalid_request", http.StatusBadRequest, true)
}
return nil
}
func ValidateBasicTaskRequest(c *gin.Context, info *RelayInfo, action string) *dto.TaskError {
var req TaskSubmitReq
if err := common.UnmarshalBodyReusable(c, &req); err != nil {
return createTaskError(err, "invalid_request", http.StatusBadRequest, true)
}
if taskErr := validatePrompt(req.Prompt); taskErr != nil {
return taskErr
}
if len(req.Images) == 0 && strings.TrimSpace(req.Image) != "" {
// 兼容单图上传
req.Images = []string{req.Image}
}
storeTaskRequest(c, info, action, req)
return nil
}
func ValidateTaskRequestWithImage(c *gin.Context, info *RelayInfo, requestObj interface{}) *dto.TaskError {
hasPrompt, ok := requestObj.(HasPrompt)
if !ok {
return createTaskError(fmt.Errorf("request must have prompt"), "invalid_request", http.StatusBadRequest, true)
}
if taskErr := validatePrompt(hasPrompt.GetPrompt()); taskErr != nil {
return taskErr
}
action := constant.TaskActionTextGenerate
if hasImage, ok := requestObj.(HasImage); ok && hasImage.HasImage() {
action = constant.TaskActionGenerate
}
storeTaskRequest(c, info, action, requestObj)
return nil
}
func ValidateTaskRequestWithImageBinding(c *gin.Context, info *RelayInfo) *dto.TaskError {
var req TaskSubmitReq
if err := c.ShouldBindJSON(&req); err != nil {
return createTaskError(err, "invalid_request_body", http.StatusBadRequest, false)
}
return ValidateTaskRequestWithImage(c, info, req)
}

15
relay/compatible_handler.go
View File

@@ -326,22 +326,11 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usage
} else {
quotaCalculateDecimal = dModelPrice.Mul(dQuotaPerUnit).Mul(dGroupRatio)
}
var dGeminiImageOutputQuota decimal.Decimal
var imageOutputPrice float64
if strings.HasPrefix(modelName, "gemini-2.5-flash-image-preview") {
imageOutputPrice = operation_setting.GetGeminiImageOutputPricePerMillionTokens(modelName)
if imageOutputPrice > 0 {
dImageOutputTokens := decimal.NewFromInt(int64(ctx.GetInt("gemini_image_tokens")))
dGeminiImageOutputQuota = decimal.NewFromFloat(imageOutputPrice).Div(decimal.NewFromInt(1000000)).Mul(dImageOutputTokens).Mul(dGroupRatio).Mul(dQuotaPerUnit)
}
}
// 添加 responses tools call 调用的配额
quotaCalculateDecimal = quotaCalculateDecimal.Add(dWebSearchQuota)
quotaCalculateDecimal = quotaCalculateDecimal.Add(dFileSearchQuota)
// 添加 audio input 独立计费
quotaCalculateDecimal = quotaCalculateDecimal.Add(audioInputQuota)
// 添加 Gemini image output 计费
quotaCalculateDecimal = quotaCalculateDecimal.Add(dGeminiImageOutputQuota)
quota := int(quotaCalculateDecimal.Round(0).IntPart())
totalTokens := promptTokens + completionTokens
@@ -440,10 +429,6 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usage
other["audio_input_token_count"] = audioTokens
other["audio_input_price"] = audioInputPrice
}
if !dGeminiImageOutputQuota.IsZero() {
other["image_output_token_count"] = ctx.GetInt("gemini_image_tokens")
other["image_output_price"] = imageOutputPrice
}
model.RecordConsumeLog(ctx, relayInfo.UserId, model.RecordConsumeLogParams{
ChannelId: relayInfo.ChannelId,
PromptTokens: promptTokens,

2
relay/responses_handler.go
View File

@@ -41,7 +41,7 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *
}
adaptor.Init(info)
var requestBody io.Reader
if model_setting.GetGlobalSettings().PassThroughRequestEnabled {
if model_setting.GetGlobalSettings().PassThroughRequestEnabled || info.ChannelSetting.PassThroughBodyEnabled {
body, err := common.GetRequestBody(c)
if err != nil {
return types.NewError(err, types.ErrorCodeReadRequestBodyFailed, types.ErrOptionWithSkipRetry())