xinyin025/one-api
1
0
Fork 0
mirror of https://github.com/songquanpeng/one-api.git synced 2025-11-16 21:23:44 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge branch 'songquanpeng:main' into origin-main

Browse Source
This commit is contained in:
取梦为饮
2024-06-04 15:53:19 +08:00
committed by GitHub
parent 1d402cd2bb 9321427c6e
commit 168a5f1f31
32 changed files with 710 additions and 519 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
relay/adaptor/baidu/main.go
View File

@@ -7,9 +7,9 @@ import (
"fmt"
"github.com/gin-gonic/gin"
"github.com/songquanpeng/one-api/common"
"github.com/songquanpeng/one-api/common/client"
"github.com/songquanpeng/one-api/common/logger"
"github.com/songquanpeng/one-api/relay/adaptor/openai"
"github.com/songquanpeng/one-api/relay/client"
"github.com/songquanpeng/one-api/relay/constant"
"github.com/songquanpeng/one-api/relay/model"
"io"

2
relay/adaptor/common.go
View File

@@ -4,7 +4,7 @@ import (
"errors"
"fmt"
"github.com/gin-gonic/gin"
"github.com/songquanpeng/one-api/relay/client"
"github.com/songquanpeng/one-api/common/client"
"github.com/songquanpeng/one-api/relay/meta"
"io"
"net/http"

26
relay/adaptor/gemini/adaptor.go
View File

@@ -13,6 +13,7 @@ import (
"github.com/songquanpeng/one-api/relay/adaptor/openai"
"github.com/songquanpeng/one-api/relay/meta"
"github.com/songquanpeng/one-api/relay/model"
"github.com/songquanpeng/one-api/relay/relaymode"
)
type Adaptor struct {
@@ -24,7 +25,14 @@ func (a *Adaptor) Init(meta *meta.Meta) {
func (a *Adaptor) GetRequestURL(meta *meta.Meta) (string, error) {
version := helper.AssignOrDefault(meta.Config.APIVersion, config.GeminiVersion)
action := "generateContent"
action := ""
switch meta.Mode {
case relaymode.Embeddings:
action = "batchEmbedContents"
default:
action = "generateContent"
}
if meta.IsStream {
action = "streamGenerateContent?alt=sse"
}
@@ -41,7 +49,14 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *model.G
if request == nil {
return nil, errors.New("request is nil")
}
return ConvertRequest(*request), nil
switch relayMode {
case relaymode.Embeddings:
geminiEmbeddingRequest := ConvertEmbeddingRequest(*request)
return geminiEmbeddingRequest, nil
default:
geminiRequest := ConvertRequest(*request)
return geminiRequest, nil
}
}
func (a *Adaptor) ConvertImageRequest(request *model.ImageRequest) (any, error) {
@@ -61,7 +76,12 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, meta *meta.Met
err, responseText = StreamHandler(c, resp)
usage = openai.ResponseText2Usage(responseText, meta.ActualModelName, meta.PromptTokens)
} else {
err, usage = Handler(c, resp, meta.PromptTokens, meta.ActualModelName)
switch meta.Mode {
case relaymode.Embeddings:
err, usage = EmbeddingHandler(c, resp)
default:
err, usage = Handler(c, resp, meta.PromptTokens, meta.ActualModelName)
}
}
return
}

2
relay/adaptor/gemini/constants.go
View File

@@ -4,5 +4,5 @@ package gemini
var ModelList = []string{
"gemini-pro", "gemini-1.0-pro-001", "gemini-1.5-pro",
"gemini-pro-vision", "gemini-1.0-pro-vision-001",
"gemini-pro-vision", "gemini-1.0-pro-vision-001", "embedding-001", "text-embedding-004",
}

76
relay/adaptor/gemini/main.go
View File

@@ -134,6 +134,29 @@ func ConvertRequest(textRequest model.GeneralOpenAIRequest) *ChatRequest {
return &geminiRequest
}
func ConvertEmbeddingRequest(request model.GeneralOpenAIRequest) *BatchEmbeddingRequest {
inputs := request.ParseInput()
requests := make([]EmbeddingRequest, len(inputs))
model := fmt.Sprintf("models/%s", request.Model)
for i, input := range inputs {
requests[i] = EmbeddingRequest{
Model: model,
Content: ChatContent{
Parts: []Part{
{
Text: input,
},
},
},
}
}
return &BatchEmbeddingRequest{
Requests: requests,
}
}
type ChatResponse struct {
Candidates []ChatCandidate `json:"candidates"`
PromptFeedback ChatPromptFeedback `json:"promptFeedback"`
@@ -230,6 +253,23 @@ func streamResponseGeminiChat2OpenAI(geminiResponse *ChatResponse) *openai.ChatC
return &response
}
func embeddingResponseGemini2OpenAI(response *EmbeddingResponse) *openai.EmbeddingResponse {
openAIEmbeddingResponse := openai.EmbeddingResponse{
Object: "list",
Data: make([]openai.EmbeddingResponseItem, 0, len(response.Embeddings)),
Model: "gemini-embedding",
Usage: model.Usage{TotalTokens: 0},
}
for _, item := range response.Embeddings {
openAIEmbeddingResponse.Data = append(openAIEmbeddingResponse.Data, openai.EmbeddingResponseItem{
Object: `embedding`,
Index: 0,
Embedding: item.Values,
})
}
return &openAIEmbeddingResponse
}
func StreamHandler(c *gin.Context, resp *http.Response) (*model.ErrorWithStatusCode, string) {
responseText := ""
scanner := bufio.NewScanner(resp.Body)
@@ -337,3 +377,39 @@ func Handler(c *gin.Context, resp *http.Response, promptTokens int, modelName st
_, err = c.Writer.Write(jsonResponse)
return nil, &usage
}
func EmbeddingHandler(c *gin.Context, resp *http.Response) (*model.ErrorWithStatusCode, *model.Usage) {
var geminiEmbeddingResponse EmbeddingResponse
responseBody, err := io.ReadAll(resp.Body)
if err != nil {
return openai.ErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
}
err = resp.Body.Close()
if err != nil {
return openai.ErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
}
err = json.Unmarshal(responseBody, &geminiEmbeddingResponse)
if err != nil {
return openai.ErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
}
if geminiEmbeddingResponse.Error != nil {
return &model.ErrorWithStatusCode{
Error: model.Error{
Message: geminiEmbeddingResponse.Error.Message,
Type: "gemini_error",
Param: "",
Code: geminiEmbeddingResponse.Error.Code,
},
StatusCode: resp.StatusCode,
}, nil
}
fullTextResponse := embeddingResponseGemini2OpenAI(&geminiEmbeddingResponse)
jsonResponse, err := json.Marshal(fullTextResponse)
if err != nil {
return openai.ErrorWrapper(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)
return nil, &fullTextResponse.Usage
}

27
relay/adaptor/gemini/model.go
View File

@@ -7,6 +7,33 @@ type ChatRequest struct {
Tools []ChatTools `json:"tools,omitempty"`
}
type EmbeddingRequest struct {
Model string `json:"model"`
Content ChatContent `json:"content"`
TaskType string `json:"taskType,omitempty"`
Title string `json:"title,omitempty"`
OutputDimensionality int `json:"outputDimensionality,omitempty"`
}
type BatchEmbeddingRequest struct {
Requests []EmbeddingRequest `json:"requests"`
}
type EmbeddingData struct {
Values []float64 `json:"values"`
}
type EmbeddingResponse struct {
Embeddings []EmbeddingData `json:"embeddings"`
Error *Error `json:"error,omitempty"`
}
type Error struct {
Code int `json:"code,omitempty"`
Message string `json:"message,omitempty"`
Status string `json:"status,omitempty"`
}
type InlineData struct {
MimeType string `json:"mimeType"`
Data string `json:"data"`

6
relay/adaptor/openai/token.go
View File

@@ -24,6 +24,10 @@ func InitTokenEncoders() {
logger.FatalLog(fmt.Sprintf("failed to get gpt-3.5-turbo token encoder: %s", err.Error()))
}
defaultTokenEncoder = gpt35TokenEncoder
gpt4oTokenEncoder, err := tiktoken.EncodingForModel("gpt-4o")
if err != nil {
logger.FatalLog(fmt.Sprintf("failed to get gpt-4o token encoder: %s", err.Error()))
}
gpt4TokenEncoder, err := tiktoken.EncodingForModel("gpt-4")
if err != nil {
logger.FatalLog(fmt.Sprintf("failed to get gpt-4 token encoder: %s", err.Error()))
@@ -31,6 +35,8 @@ func InitTokenEncoders() {
for model := range billingratio.ModelRatio {
if strings.HasPrefix(model, "gpt-3.5") {
tokenEncoderMap[model] = gpt35TokenEncoder
} else if strings.HasPrefix(model, "gpt-4o") {
tokenEncoderMap[model] = gpt4oTokenEncoder
} else if strings.HasPrefix(model, "gpt-4") {
tokenEncoderMap[model] = gpt4TokenEncoder
} else {

5
relay/billing/ratio/image.go
View File

@@ -49,3 +49,8 @@ var ImagePromptLengthLimitations = map[string]int{
"wanx-v1": 4000,
"cogview-3": 833,
}
var ImageOriginModelName = map[string]string{
"ali-stable-diffusion-xl": "stable-diffusion-xl",
"ali-stable-diffusion-v1.5": "stable-diffusion-v1.5",
}

24
relay/client/init.go
View File

@@ -1,24 +0,0 @@
package client
import (
"github.com/songquanpeng/one-api/common/config"
"net/http"
"time"
)
var HTTPClient *http.Client
var ImpatientHTTPClient *http.Client
func init() {
if config.RelayTimeout == 0 {
HTTPClient = &http.Client{}
} else {
HTTPClient = &http.Client{
Timeout: time.Duration(config.RelayTimeout) * time.Second,
}
}
ImpatientHTTPClient = &http.Client{
Timeout: 5 * time.Second,
}
}

2
relay/controller/audio.go
View File

@@ -9,6 +9,7 @@ import (
"fmt"
"github.com/gin-gonic/gin"
"github.com/songquanpeng/one-api/common"
"github.com/songquanpeng/one-api/common/client"
"github.com/songquanpeng/one-api/common/config"
"github.com/songquanpeng/one-api/common/ctxkey"
"github.com/songquanpeng/one-api/common/logger"
@@ -17,7 +18,6 @@ import (
"github.com/songquanpeng/one-api/relay/billing"
billingratio "github.com/songquanpeng/one-api/relay/billing/ratio"
"github.com/songquanpeng/one-api/relay/channeltype"
"github.com/songquanpeng/one-api/relay/client"
"github.com/songquanpeng/one-api/relay/meta"
relaymodel "github.com/songquanpeng/one-api/relay/model"
"github.com/songquanpeng/one-api/relay/relaymode"

7
relay/controller/image.go
View File

@@ -55,6 +55,11 @@ func RelayImageHelper(c *gin.Context, relayMode int) *relaymodel.ErrorWithStatus
return openai.ErrorWrapper(err, "get_image_cost_ratio_failed", http.StatusInternalServerError)
}
imageModel := imageRequest.Model
// Convert the original image model
imageRequest.Model, _ = getMappedModelName(imageRequest.Model, billingratio.ImageOriginModelName)
c.Set("response_format", imageRequest.ResponseFormat)
var requestBody io.Reader
if isModelMapped || meta.ChannelType == channeltype.Azure { // make Azure channel request body
jsonStr, err := json.Marshal(imageRequest)
@@ -89,7 +94,7 @@ func RelayImageHelper(c *gin.Context, relayMode int) *relaymodel.ErrorWithStatus
requestBody = bytes.NewBuffer(jsonStr)
}
modelRatio := billingratio.GetModelRatio(imageRequest.Model)
modelRatio := billingratio.GetModelRatio(imageModel)
groupRatio := billingratio.GetGroupRatio(meta.Group)
ratio := modelRatio * groupRatio
userQuota, err := model.CacheGetUserQuota(ctx, meta.UserId)