feat: add Proxy channel type and relay mode

Add the Proxy channel type and relay mode to support proxying requests to custom upstream services.

.github/workflows/docker-image-amd64-en.yml

@@ -1,4 +1,4 @@
-name: Publish Docker image (English)
+name: Publish Docker image (amd64, English)
 
 on:
   push:
@@ -34,13 +34,6 @@ jobs:
       - name: Translate
         run: |
           python ./i18n/translate.py --repository_path . --json_file_path ./i18n/en.json
-
-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@v2
-
-      - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v2
-
       - name: Log in to Docker Hub
         uses: docker/login-action@v2
         with:
@@ -58,7 +51,6 @@ jobs:
         uses: docker/build-push-action@v3
         with:
           context: .
-          platforms: linux/amd64,linux/arm64
           push: true
           tags: ${{ steps.meta.outputs.tags }}
           labels: ${{ steps.meta.outputs.labels }}

.github/workflows/docker-image-amd64.yml

@@ -0,0 +1,61 @@
+name: Publish Docker image (amd64)
+
+on:
+  push:
+    tags:
+      - 'v*.*.*'
+  workflow_dispatch:
+    inputs:
+      name:
+        description: 'reason'
+        required: false
+jobs:
+  push_to_registries:
+    name: Push Docker image to multiple registries
+    runs-on: ubuntu-latest
+    permissions:
+      packages: write
+      contents: read
+    steps:
+      - name: Check out the repo
+        uses: actions/checkout@v3
+
+      - name: Check repository URL
+        run: |
+          REPO_URL=$(git config --get remote.origin.url)
+          if [[ $REPO_URL == *"pro" ]]; then
+            exit 1
+          fi        
+
+      - name: Save version info
+        run: |
+          git describe --tags > VERSION 
+
+      - name: Log in to Docker Hub
+        uses: docker/login-action@v2
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+
+      - name: Log in to the Container registry
+        uses: docker/login-action@v2
+        with:
+          registry: ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Extract metadata (tags, labels) for Docker
+        id: meta
+        uses: docker/metadata-action@v4
+        with:
+          images: |
+            justsong/one-api
+            ghcr.io/${{ github.repository }}
+
+      - name: Build and push Docker images
+        uses: docker/build-push-action@v3
+        with:
+          context: .
+          push: true
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}

.github/workflows/docker-image-arm64.yml

@@ -1,9 +1,10 @@
-name: Publish Docker image
+name: Publish Docker image (arm64)
 
 on:
   push:
     tags:
       - 'v*.*.*'
+      - '!*-alpha*'
   workflow_dispatch:
     inputs:
       name:

Dockerfile

@@ -1,4 +1,4 @@
-FROM --platform=$BUILDPLATFORM node:16 AS builder
+FROM node:16 as builder
 
 WORKDIR /web
 COPY ./VERSION .

README.md

@@ -89,7 +89,6 @@ _✨ 通过标准的 OpenAI API 格式访问所有的大模型，开箱即用
    + [x] [DeepL](https://www.deepl.com/)
    + [x] [together.ai](https://www.together.ai/)
    + [x] [novita.ai](https://www.novita.ai/)
-   + [x] [硅基流动 SiliconCloud](https://siliconflow.cn/siliconcloud)
 2. 支持配置镜像以及众多[第三方代理服务](https://iamazing.cn/page/openai-api-third-party-services)。
 3. 支持通过**负载均衡**的方式访问多个渠道。
 4. 支持 **stream 模式**，可以通过流式传输实现打字机效果。
@@ -252,9 +251,9 @@ docker run --name chatgpt-web -d -p 3002:3002 -e OPENAI_API_BASE_URL=https://ope
 #### QChatGPT - QQ机器人
 项目主页：https://github.com/RockChinQ/QChatGPT
 
-根据[文档](https://qchatgpt.rockchin.top)完成部署后，在 `data/provider.json`设置`requester.openai-chat-completions.base-url`为 One API 实例地址，并填写 API Key 到 `keys.openai` 组中，设置 `model` 为要使用的模型名称。
+根据文档完成部署后，在`config.py`设置配置项`openai_config`的`reverse_proxy`为 One API 后端地址，设置`api_key`为 One API 生成的key，并在配置项`completion_api_params`的`model`参数设置为 One API 支持的模型名称。
 
-运行期间可以通过`!model`命令查看、切换可用模型。
+可安装 [Switcher 插件](https://github.com/RockChinQ/Switcher)在运行时切换所使用的模型。
 
 ### 部署到第三方平台
 <details>

common/ctxkey/key.go

@@ -3,7 +3,6 @@ package ctxkey
 const (
 	Config            = "config"
 	Id                = "id"
-	RequestId         = "X-Oneapi-Request-Id"
 	Username          = "username"
 	Role              = "role"
 	Status            = "status"
@@ -16,7 +15,6 @@ const (
 	Group             = "group"
 	ModelMapping      = "model_mapping"
 	ChannelName       = "channel_name"
-	ContentType       = "content_type"
 	TokenId           = "token_id"
 	TokenName         = "token_name"
 	BaseURL           = "base_url"

common/gin.go

@@ -2,11 +2,11 @@ package common
 
 import (
 	"bytes"
-	"io"
+	"encoding/json"
-
 	"github.com/gin-gonic/gin"
-	"github.com/pkg/errors"
 	"github.com/songquanpeng/one-api/common/ctxkey"
+	"io"
+	"strings"
 )
 
 func GetRequestBody(c *gin.Context) ([]byte, error) {
@@ -28,16 +28,18 @@ func UnmarshalBodyReusable(c *gin.Context, v any) error {
 	if err != nil {
 		return err
 	}
+	contentType := c.Request.Header.Get("Content-Type")
+	if strings.HasPrefix(contentType, "application/json") {
+		err = json.Unmarshal(requestBody, &v)
+	} else {
+		// skip for now
+		// TODO: someday non json request have variant model, we will need to implementation this
+	}
+	if err != nil {
+		return err
+	}
 	// Reset request body
 	c.Request.Body = io.NopCloser(bytes.NewBuffer(requestBody))
-	defer func() {
-		c.Request.Body = io.NopCloser(bytes.NewBuffer(requestBody))
-	}()
-
-	if err = c.Bind(v); err != nil {
-		return errors.Wrap(err, "bind request body failed")
-	}
-
 	return nil
 }
 

controller/relay.go

@@ -26,8 +26,7 @@ import (
 func relayHelper(c *gin.Context, relayMode int) *model.ErrorWithStatusCode {
 	var err *model.ErrorWithStatusCode
 	switch relayMode {
-	case relaymode.ImagesGenerations,
+	case relaymode.ImagesGenerations:
-		relaymode.ImagesEdits:
 		err = controller.RelayImageHelper(c, relayMode)
 	case relaymode.AudioSpeech:
 		fallthrough
@@ -46,6 +45,10 @@ func relayHelper(c *gin.Context, relayMode int) *model.ErrorWithStatusCode {
 func Relay(c *gin.Context) {
 	ctx := c.Request.Context()
 	relayMode := relaymode.GetByPath(c.Request.URL.Path)
+	if config.DebugEnabled {
+		requestBody, _ := common.GetRequestBody(c)
+		logger.Debugf(ctx, "request body: %s", string(requestBody))
+	}
 	channelId := c.GetInt(ctxkey.ChannelId)
 	userId := c.GetInt(ctxkey.Id)
 	bizErr := relayHelper(c, relayMode)
@@ -57,8 +60,6 @@ func Relay(c *gin.Context) {
 	channelName := c.GetString(ctxkey.ChannelName)
 	group := c.GetString(ctxkey.Group)
 	originalModel := c.GetString(ctxkey.OriginalModel)
-
-	// BUG: bizErr is shared, should not run this function in goroutine to avoid race
 	go processChannelRelayError(ctx, userId, channelId, channelName, bizErr)
 	requestId := c.GetString(helper.RequestIdKey)
 	retryTimes := config.RetryTimes
@@ -89,7 +90,6 @@ func Relay(c *gin.Context) {
 		// BUG: bizErr is in race condition
 		go processChannelRelayError(ctx, userId, channelId, channelName, bizErr)
 	}
-
 	if bizErr != nil {
 		if bizErr.StatusCode == http.StatusTooManyRequests {
 			bizErr.Error.Message = "当前分组上游负载已饱和，请稍后再试"

middleware/distributor.go

@@ -12,7 +12,7 @@ import (
 )
 
 type ModelRequest struct {
-	Model string `json:"model" form:"model"`
+	Model string `json:"model"`
 }
 
 func Distribute() func(c *gin.Context) {
@@ -61,7 +61,6 @@ func SetupContextForSelectedChannel(c *gin.Context, channel *model.Channel, mode
 	c.Set(ctxkey.Channel, channel.Type)
 	c.Set(ctxkey.ChannelId, channel.Id)
 	c.Set(ctxkey.ChannelName, channel.Name)
-	c.Set(ctxkey.ContentType, c.Request.Header.Get("Content-Type"))
 	c.Set(ctxkey.ModelMapping, channel.GetModelMapping())
 	c.Set(ctxkey.OriginalModel, modelName) // for retry
 	c.Request.Header.Set("Authorization", fmt.Sprintf("Bearer %s", channel.Key))

middleware/logger.go

@@ -2,7 +2,6 @@ package middleware
 
 import (
 	"fmt"
-
 	"github.com/gin-gonic/gin"
 	"github.com/songquanpeng/one-api/common/helper"
 )

middleware/request-id.go

@@ -2,7 +2,6 @@ package middleware
 
 import (
 	"context"
-
 	"github.com/gin-gonic/gin"
 	"github.com/songquanpeng/one-api/common/helper"
 )

middleware/utils.go

@@ -2,12 +2,11 @@ package middleware
 
 import (
 	"fmt"
-	"strings"
-
 	"github.com/gin-gonic/gin"
 	"github.com/songquanpeng/one-api/common"
 	"github.com/songquanpeng/one-api/common/helper"
 	"github.com/songquanpeng/one-api/common/logger"
+	"strings"
 )
 
 func abortWithMessage(c *gin.Context, statusCode int, message string) {

relay/adaptor/cloudflare/constant.go

@@ -1,7 +1,6 @@
 package cloudflare
 
 var ModelList = []string{
-	"@cf/meta/llama-3.1-8b-instruct",
 	"@cf/meta/llama-2-7b-chat-fp16",
 	"@cf/meta/llama-2-7b-chat-int8",
 	"@cf/mistral/mistral-7b-instruct-v0.1",

relay/adaptor/common.go

@@ -3,13 +3,11 @@ package adaptor
 import (
 	"errors"
 	"fmt"
-	"io"
-	"net/http"
-
 	"github.com/gin-gonic/gin"
 	"github.com/songquanpeng/one-api/common/client"
-	"github.com/songquanpeng/one-api/common/ctxkey"
 	"github.com/songquanpeng/one-api/relay/meta"
+	"io"
+	"net/http"
 )
 
 func SetupCommonRequestHeader(c *gin.Context, req *http.Request, meta *meta.Meta) {
@@ -29,9 +27,6 @@ func DoRequestHelper(a Adaptor, c *gin.Context, meta *meta.Meta, requestBody io.
 	if err != nil {
 		return nil, fmt.Errorf("new request failed: %w", err)
 	}
-
-	req.Header.Set("Content-Type", c.GetString(ctxkey.ContentType))
-
 	err = a.SetupRequestHeader(c, req, meta)
 	if err != nil {
 		return nil, fmt.Errorf("setup request header failed: %w", err)

relay/adaptor/doubao/main.go

@@ -7,12 +7,8 @@ import (
 )
 
 func GetRequestURL(meta *meta.Meta) (string, error) {
-	switch meta.Mode {
+	if meta.Mode == relaymode.ChatCompletions {
-	case relaymode.ChatCompletions:
 		return fmt.Sprintf("%s/api/v3/chat/completions", meta.BaseURL), nil
-	case relaymode.Embeddings:
-		return fmt.Sprintf("%s/api/v3/embeddings", meta.BaseURL), nil
-	default:
 	}
 	return "", fmt.Errorf("unsupported relay mode %d for doubao", meta.Mode)
 }

relay/adaptor/gemini/constants.go

@@ -3,5 +3,6 @@ package gemini
 // https://ai.google.dev/models/gemini
 
 var ModelList = []string{
-	"gemini-pro", "gemini-1.0-pro", "gemini-1.5-flash", "gemini-1.5-pro", "text-embedding-004", "aqa",
+	"gemini-pro", "gemini-1.0-pro-001", "gemini-1.5-pro",
+	"gemini-pro-vision", "gemini-1.0-pro-vision-001", "embedding-001", "text-embedding-004",
 }

relay/adaptor/groq/constants.go

@@ -4,14 +4,9 @@ package groq
 
 var ModelList = []string{
 	"gemma-7b-it",
+	"llama2-7b-2048",
+	"llama2-70b-4096",
 	"mixtral-8x7b-32768",
 	"llama3-8b-8192",
 	"llama3-70b-8192",
-	"gemma2-9b-it",
-	"llama-3.1-405b-reasoning",
-	"llama-3.1-70b-versatile",
-	"llama-3.1-8b-instant",
-	"llama3-groq-70b-8192-tool-use-preview",
-	"llama3-groq-8b-8192-tool-use-preview",
-	"whisper-large-v3",
 }

relay/adaptor/ollama/adaptor.go

@@ -24,7 +24,7 @@ func (a *Adaptor) GetRequestURL(meta *meta.Meta) (string, error) {
 	// https://github.com/ollama/ollama/blob/main/docs/api.md
 	fullRequestURL := fmt.Sprintf("%s/api/chat", meta.BaseURL)
 	if meta.Mode == relaymode.Embeddings {
-		fullRequestURL = fmt.Sprintf("%s/api/embed", meta.BaseURL)
+		fullRequestURL = fmt.Sprintf("%s/api/embeddings", meta.BaseURL)
 	}
 	return fullRequestURL, nil
 }

relay/adaptor/ollama/main.go

@@ -31,8 +31,6 @@ func ConvertRequest(request model.GeneralOpenAIRequest) *ChatRequest {
 			TopP:             request.TopP,
 			FrequencyPenalty: request.FrequencyPenalty,
 			PresencePenalty:  request.PresencePenalty,
-			NumPredict:  	  request.MaxTokens,
-			NumCtx:  	  request.NumCtx,
 		},
 		Stream: request.Stream,
 	}
@@ -120,10 +118,8 @@ func StreamHandler(c *gin.Context, resp *http.Response) (*model.ErrorWithStatusC
 	common.SetEventStreamHeaders(c)
 
 	for scanner.Scan() {
-		data := scanner.Text()
+		data := strings.TrimPrefix(scanner.Text(), "}")
-		if strings.HasPrefix(data, "}") {
+		data = data + "}"
-		    data = strings.TrimPrefix(data, "}") + "}"
-		}
 
 		var ollamaResponse ChatResponse
 		err := json.Unmarshal([]byte(data), &ollamaResponse)
@@ -162,14 +158,7 @@ func StreamHandler(c *gin.Context, resp *http.Response) (*model.ErrorWithStatusC
 func ConvertEmbeddingRequest(request model.GeneralOpenAIRequest) *EmbeddingRequest {
 	return &EmbeddingRequest{
 		Model:  request.Model,
-		Input: request.ParseInput(),
+		Prompt: strings.Join(request.ParseInput(), " "),
-		Options: &Options{
-			Seed:             int(request.Seed),
-			Temperature:      request.Temperature,
-			TopP:             request.TopP,
-			FrequencyPenalty: request.FrequencyPenalty,
-			PresencePenalty:  request.PresencePenalty,
-		},
 	}
 }
 
@@ -212,17 +201,15 @@ func embeddingResponseOllama2OpenAI(response *EmbeddingResponse) *openai.Embeddi
 	openAIEmbeddingResponse := openai.EmbeddingResponse{
 		Object: "list",
 		Data:   make([]openai.EmbeddingResponseItem, 0, 1),
-		Model:  response.Model,
+		Model:  "text-embedding-v1",
 		Usage:  model.Usage{TotalTokens: 0},
 	}
 
-	for i, embedding := range response.Embeddings {
 	openAIEmbeddingResponse.Data = append(openAIEmbeddingResponse.Data, openai.EmbeddingResponseItem{
 		Object:    `embedding`,
-			Index:     i,
+		Index:     0,
-			Embedding: embedding,
+		Embedding: response.Embedding,
 	})
-	}
 	return &openAIEmbeddingResponse
 }
 

relay/adaptor/ollama/model.go

@@ -7,8 +7,6 @@ type Options struct {
 	TopP             float64 `json:"top_p,omitempty"`
 	FrequencyPenalty float64 `json:"frequency_penalty,omitempty"`
 	PresencePenalty  float64 `json:"presence_penalty,omitempty"`
-	NumPredict  	 int 	 `json:"num_predict,omitempty"`
-	NumCtx  	 int 	 `json:"num_ctx,omitempty"`
 }
 
 type Message struct {
@@ -40,14 +38,10 @@ type ChatResponse struct {
 
 type EmbeddingRequest struct {
 	Model  string `json:"model"`
-	Input []string `json:"input"`
+	Prompt string `json:"prompt"`
-	// Truncate  bool     `json:"truncate,omitempty"`
-	Options *Options `json:"options,omitempty"`
-	// KeepAlive string   `json:"keep_alive,omitempty"`
 }
 
 type EmbeddingResponse struct {
 	Error     string    `json:"error,omitempty"`
-	Model      string      `json:"model"`
+	Embedding []float64 `json:"embedding,omitempty"`
-	Embeddings [][]float64 `json:"embeddings"`
 }

relay/adaptor/openai/adaptor.go

@@ -104,13 +104,10 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, meta *meta.Met
 		switch meta.Mode {
 		case relaymode.ImagesGenerations:
 			err, _ = ImageHandler(c, resp)
-		case relaymode.ImagesEdits:
-			err, _ = ImagesEditsHandler(c, resp)
 		default:
 			err, usage = Handler(c, resp, meta.PromptTokens, meta.ActualModelName)
 		}
 	}
-
 	return
 }
 

relay/adaptor/openai/compatible.go

@@ -13,7 +13,6 @@ import (
 	"github.com/songquanpeng/one-api/relay/adaptor/novita"
 	"github.com/songquanpeng/one-api/relay/adaptor/stepfun"
 	"github.com/songquanpeng/one-api/relay/adaptor/togetherai"
-	"github.com/songquanpeng/one-api/relay/adaptor/siliconflow"
 	"github.com/songquanpeng/one-api/relay/channeltype"
 )
 
@@ -31,7 +30,6 @@ var CompatibleChannels = []int{
 	channeltype.DeepSeek,
 	channeltype.TogetherAI,
 	channeltype.Novita,
-	channeltype.SiliconFlow,
 }
 
 func GetCompatibleChannelMeta(channelType int) (string, []string) {
@@ -62,8 +60,6 @@ func GetCompatibleChannelMeta(channelType int) (string, []string) {
 		return "doubao", doubao.ModelList
 	case channeltype.Novita:
 		return "novita", novita.ModelList
-	case channeltype.SiliconFlow:
-		return "siliconflow", siliconflow.ModelList
 	default:
 		return "openai", ModelList
 	}

relay/adaptor/openai/constants.go

@@ -8,7 +8,6 @@ var ModelList = []string{
 	"gpt-4-32k", "gpt-4-32k-0314", "gpt-4-32k-0613",
 	"gpt-4-turbo-preview", "gpt-4-turbo", "gpt-4-turbo-2024-04-09",
 	"gpt-4o", "gpt-4o-2024-05-13",
-	"gpt-4o-mini", "gpt-4o-mini-2024-07-18",
 	"gpt-4-vision-preview",
 	"text-embedding-ada-002", "text-embedding-3-small", "text-embedding-3-large",
 	"text-curie-001", "text-babbage-001", "text-ada-001", "text-davinci-002", "text-davinci-003",

relay/adaptor/openai/image.go

@@ -3,30 +3,12 @@ package openai
 import (
 	"bytes"
 	"encoding/json"
-	"io"
-	"net/http"
-
 	"github.com/gin-gonic/gin"
 	"github.com/songquanpeng/one-api/relay/model"
+	"io"
+	"net/http"
 )
 
-// ImagesEditsHandler just copy response body to client
-//
-// https://platform.openai.com/docs/api-reference/images/createEdit
-func ImagesEditsHandler(c *gin.Context, resp *http.Response) (*model.ErrorWithStatusCode, *model.Usage) {
-	c.Writer.WriteHeader(resp.StatusCode)
-	for k, v := range resp.Header {
-		c.Writer.Header().Set(k, v[0])
-	}
-
-	if _, err := io.Copy(c.Writer, resp.Body); err != nil {
-		return ErrorWrapper(err, "copy_response_body_failed", http.StatusInternalServerError), nil
-	}
-	defer resp.Body.Close()
-
-	return nil, nil
-}
-
 func ImageHandler(c *gin.Context, resp *http.Response) (*model.ErrorWithStatusCode, *model.Usage) {
 	var imageResponse ImageResponse
 	responseBody, err := io.ReadAll(resp.Body)

relay/adaptor/openai/token.go

@@ -110,7 +110,7 @@ func CountTokenMessages(messages []model.Message, model string) int {
 						if imageUrl["detail"] != nil {
 							detail = imageUrl["detail"].(string)
 						}
-						imageTokens, err := countImageTokens(url, detail, model)
+						imageTokens, err := countImageTokens(url, detail)
 						if err != nil {
 							logger.SysError("error counting image tokens: " + err.Error())
 						} else {
@@ -134,15 +134,11 @@ const (
 	lowDetailCost         = 85
 	highDetailCostPerTile = 170
 	additionalCost        = 85
-	// gpt-4o-mini cost higher than other model
-	gpt4oMiniLowDetailCost  = 2833
-	gpt4oMiniHighDetailCost = 5667
-	gpt4oMiniAdditionalCost = 2833
 )
 
 // https://platform.openai.com/docs/guides/vision/calculating-costs
 // https://github.com/openai/openai-cookbook/blob/05e3f9be4c7a2ae7ecf029a7c32065b024730ebe/examples/How_to_count_tokens_with_tiktoken.ipynb
-func countImageTokens(url string, detail string, model string) (_ int, err error) {
+func countImageTokens(url string, detail string) (_ int, err error) {
 	var fetchSize = true
 	var width, height int
 	// Reference: https://platform.openai.com/docs/guides/vision/low-or-high-fidelity-image-understanding
@@ -176,9 +172,6 @@ func countImageTokens(url string, detail string, model string) (_ int, err error
 	}
 	switch detail {
 	case "low":
-		if strings.HasPrefix(model, "gpt-4o-mini") {
-			return gpt4oMiniLowDetailCost, nil
-		}
 		return lowDetailCost, nil
 	case "high":
 		if fetchSize {
@@ -198,9 +191,6 @@ func countImageTokens(url string, detail string, model string) (_ int, err error
 			height = int(float64(height) * ratio)
 		}
 		numSquares := int(math.Ceil(float64(width)/512) * math.Ceil(float64(height)/512))
-		if strings.HasPrefix(model, "gpt-4o-mini") {
-			return numSquares*gpt4oMiniHighDetailCost + gpt4oMiniAdditionalCost, nil
-		}
 		result := numSquares*highDetailCostPerTile + additionalCost
 		return result, nil
 	default:

relay/adaptor/siliconflow/constants.go

@@ -1,36 +0,0 @@
-package siliconflow
-
-// https://docs.siliconflow.cn/docs/getting-started
-
-var ModelList = []string{
-	"deepseek-ai/deepseek-llm-67b-chat",
-	"Qwen/Qwen1.5-14B-Chat",
-	"Qwen/Qwen1.5-7B-Chat",
-	"Qwen/Qwen1.5-110B-Chat",
-	"Qwen/Qwen1.5-32B-Chat",
-	"01-ai/Yi-1.5-6B-Chat",
-	"01-ai/Yi-1.5-9B-Chat-16K",
-	"01-ai/Yi-1.5-34B-Chat-16K",
-	"THUDM/chatglm3-6b",
-	"deepseek-ai/DeepSeek-V2-Chat",
-	"THUDM/glm-4-9b-chat",
-	"Qwen/Qwen2-72B-Instruct",
-	"Qwen/Qwen2-7B-Instruct",
-	"Qwen/Qwen2-57B-A14B-Instruct",
-	"deepseek-ai/DeepSeek-Coder-V2-Instruct",
-	"Qwen/Qwen2-1.5B-Instruct",
-	"internlm/internlm2_5-7b-chat",
-	"BAAI/bge-large-en-v1.5",
-	"BAAI/bge-large-zh-v1.5",
-	"Pro/Qwen/Qwen2-7B-Instruct",
-	"Pro/Qwen/Qwen2-1.5B-Instruct",
-	"Pro/Qwen/Qwen1.5-7B-Chat",
-	"Pro/THUDM/glm-4-9b-chat",
-	"Pro/THUDM/chatglm3-6b",
-	"Pro/01-ai/Yi-1.5-9B-Chat-16K",
-	"Pro/01-ai/Yi-1.5-6B-Chat",
-	"Pro/google/gemma-2-9b-it",
-	"Pro/internlm/internlm2_5-7b-chat",
-	"Pro/meta-llama/Meta-Llama-3-8B-Instruct",
-	"Pro/mistralai/Mistral-7B-Instruct-v0.2",
-}

relay/billing/ratio/model.go

@@ -35,8 +35,6 @@ var ModelRatio = map[string]float64{
 	"gpt-4-turbo-2024-04-09":  5,    // $0.01 / 1K tokens
 	"gpt-4o":                  2.5,  // $0.005 / 1K tokens
 	"gpt-4o-2024-05-13":       2.5,  // $0.005 / 1K tokens
-	"gpt-4o-mini":             0.075, // $0.00015 / 1K tokens
-	"gpt-4o-mini-2024-07-18":  0.075, // $0.00015 / 1K tokens
 	"gpt-4-vision-preview":    5,    // $0.01 / 1K tokens
 	"gpt-3.5-turbo":           0.25, // $0.0005 / 1K tokens
 	"gpt-3.5-turbo-0301":      0.75,
@@ -98,11 +96,12 @@ var ModelRatio = map[string]float64{
 	"bge-large-en":       0.002 * RMB,
 	"tao-8k":             0.002 * RMB,
 	// https://ai.google.dev/pricing
+	"PaLM-2":                    1,
 	"gemini-pro":                1, // $0.00025 / 1k characters -> $0.001 / 1k tokens
-	"gemini-1.0-pro":   1,
+	"gemini-pro-vision":         1, // $0.00025 / 1k characters -> $0.001 / 1k tokens
-	"gemini-1.5-flash": 1,
+	"gemini-1.0-pro-vision-001": 1,
+	"gemini-1.0-pro-001":        1,
 	"gemini-1.5-pro":            1,
-	"aqa":              1,
 	// https://open.bigmodel.cn/pricing
 	"glm-4":         0.1 * RMB,
 	"glm-4v":        0.1 * RMB,
@@ -157,16 +156,12 @@ var ModelRatio = map[string]float64{
 	"mistral-large-latest":  8.0 / 1000 * USD,
 	"mistral-embed":         0.1 / 1000 * USD,
 	// https://wow.groq.com/#:~:text=inquiries%C2%A0here.-,Model,-Current%20Speed
-	"gemma-7b-it":                           0.07 / 1000000 * USD,
+	"llama3-70b-8192":    0.59 / 1000 * USD,
-	"mixtral-8x7b-32768":                    0.24 / 1000000 * USD,
+	"mixtral-8x7b-32768": 0.27 / 1000 * USD,
-	"llama3-8b-8192":                        0.05 / 1000000 * USD,
+	"llama3-8b-8192":     0.05 / 1000 * USD,
-	"llama3-70b-8192":                       0.59 / 1000000 * USD,
+	"gemma-7b-it":        0.1 / 1000 * USD,
-	"gemma2-9b-it":                          0.20 / 1000000 * USD,
+	"llama2-70b-4096":    0.64 / 1000 * USD,
-	"llama-3.1-405b-reasoning":              0.89 / 1000000 * USD,
+	"llama2-7b-2048":     0.1 / 1000 * USD,
-	"llama-3.1-70b-versatile":               0.59 / 1000000 * USD,
-	"llama-3.1-8b-instant":                  0.05 / 1000000 * USD,
-	"llama3-groq-70b-8192-tool-use-preview": 0.89 / 1000000 * USD,
-	"llama3-groq-8b-8192-tool-use-preview":  0.19 / 1000000 * USD,
 	// https://platform.lingyiwanwu.com/docs#-计费单元
 	"yi-34b-chat-0205": 2.5 / 1000 * RMB,
 	"yi-34b-chat-200k": 12.0 / 1000 * RMB,
@@ -313,9 +308,6 @@ func GetCompletionRatio(name string, channelType int) float64 {
 		return 4.0 / 3.0
 	}
 	if strings.HasPrefix(name, "gpt-4") {
-		if strings.HasPrefix(name, "gpt-4o-mini") {
-			return 4
-		}
 		if strings.HasPrefix(name, "gpt-4-turbo") ||
 			strings.HasPrefix(name, "gpt-4o") ||
 			strings.HasSuffix(name, "preview") {

relay/channeltype/define.go

@@ -45,6 +45,5 @@ const (
 	Novita
 	VertextAI
 	Proxy
-	SiliconFlow
 	Dummy
 )

relay/channeltype/url.go

@@ -45,7 +45,6 @@ var ChannelBaseURLs = []string{
 	"https://api.novita.ai/v3/openai",           // 41
 	"",                                          // 42
 	"",                                          // 43
-	"https://api.siliconflow.cn",                 // 44
 }
 
 func init() {

relay/controller/image.go

@@ -8,7 +8,6 @@ import (
 	"fmt"
 	"io"
 	"net/http"
-	"strings"
 
 	"github.com/gin-gonic/gin"
 	"github.com/songquanpeng/one-api/common"
@@ -135,8 +134,7 @@ func RelayImageHelper(c *gin.Context, relayMode int) *relaymodel.ErrorWithStatus
 	c.Set("response_format", imageRequest.ResponseFormat)
 
 	var requestBody io.Reader
-	if strings.ToLower(c.GetString(ctxkey.ContentType)) == "application/json" &&
+	if isModelMapped || meta.ChannelType == channeltype.Azure { // make Azure channel request body
-		isModelMapped || meta.ChannelType == channeltype.Azure { // make Azure channel request body
 		jsonStr, err := json.Marshal(imageRequest)
 		if err != nil {
 			return openai.ErrorWrapper(err, "marshal_image_request_failed", http.StatusInternalServerError)

relay/controller/text.go

@@ -10,7 +10,6 @@ import (
 	"github.com/gin-gonic/gin"
 	"github.com/songquanpeng/one-api/common/logger"
 	"github.com/songquanpeng/one-api/relay"
-	"github.com/songquanpeng/one-api/relay/adaptor"
 	"github.com/songquanpeng/one-api/relay/adaptor/openai"
 	"github.com/songquanpeng/one-api/relay/apitype"
 	"github.com/songquanpeng/one-api/relay/billing"
@@ -32,8 +31,9 @@ func RelayTextHelper(c *gin.Context) *model.ErrorWithStatusCode {
 	meta.IsStream = textRequest.Stream
 
 	// map model name
+	var isModelMapped bool
 	meta.OriginModelName = textRequest.Model
-	textRequest.Model, _ = getMappedModelName(textRequest.Model, meta.ModelMapping)
+	textRequest.Model, isModelMapped = getMappedModelName(textRequest.Model, meta.ModelMapping)
 	meta.ActualModelName = textRequest.Model
 	// get model ratio & group ratio
 	modelRatio := billingratio.GetModelRatio(textRequest.Model, meta.ChannelType)
@@ -55,10 +55,31 @@ func RelayTextHelper(c *gin.Context) *model.ErrorWithStatusCode {
 	adaptor.Init(meta)
 
 	// get request body
-	requestBody, err := getRequestBody(c, meta, textRequest, adaptor)
+	var requestBody io.Reader
+	if meta.APIType == apitype.OpenAI {
+		// no need to convert request for openai
+		shouldResetRequestBody := isModelMapped || meta.ChannelType == channeltype.Baichuan // frequency_penalty 0 is not acceptable for baichuan
+		if shouldResetRequestBody {
+			jsonStr, err := json.Marshal(textRequest)
+			if err != nil {
+				return openai.ErrorWrapper(err, "json_marshal_failed", http.StatusInternalServerError)
+			}
+			requestBody = bytes.NewBuffer(jsonStr)
+		} else {
+			requestBody = c.Request.Body
+		}
+	} else {
+		convertedRequest, err := adaptor.ConvertRequest(c, meta.Mode, textRequest)
 		if err != nil {
 			return openai.ErrorWrapper(err, "convert_request_failed", http.StatusInternalServerError)
 		}
+		jsonData, err := json.Marshal(convertedRequest)
+		if err != nil {
+			return openai.ErrorWrapper(err, "json_marshal_failed", http.StatusInternalServerError)
+		}
+		logger.Debugf(ctx, "converted request: \n%s", string(jsonData))
+		requestBody = bytes.NewBuffer(jsonData)
+	}
 
 	// do request
 	resp, err := adaptor.DoRequest(c, meta, requestBody)
@@ -82,26 +103,3 @@ func RelayTextHelper(c *gin.Context) *model.ErrorWithStatusCode {
 	go postConsumeQuota(ctx, usage, meta, textRequest, ratio, preConsumedQuota, modelRatio, groupRatio)
 	return nil
 }
-
-func getRequestBody(c *gin.Context, meta *meta.Meta, textRequest *model.GeneralOpenAIRequest, adaptor adaptor.Adaptor) (io.Reader, error) {
-	if meta.APIType == apitype.OpenAI && meta.OriginModelName == meta.ActualModelName && meta.ChannelType != channeltype.Baichuan {
-		// no need to convert request for openai
-		return c.Request.Body, nil
-	}
-
-	// get request body
-	var requestBody io.Reader
-	convertedRequest, err := adaptor.ConvertRequest(c, meta.Mode, textRequest)
-	if err != nil {
-		logger.Debugf(c.Request.Context(), "converted request failed: %s\n", err.Error())
-		return nil, err
-	}
-	jsonData, err := json.Marshal(convertedRequest)
-	if err != nil {
-		logger.Debugf(c.Request.Context(), "converted request json_marshal_failed: %s\n", err.Error())
-		return nil, err
-	}
-	logger.Debugf(c.Request.Context(), "converted request: \n%s", string(jsonData))
-	requestBody = bytes.NewBuffer(jsonData)
-	return requestBody, nil
-}

relay/model/general.go

@@ -29,7 +29,6 @@ type GeneralOpenAIRequest struct {
 	Dimensions       int             `json:"dimensions,omitempty"`
 	Instruction      string          `json:"instruction,omitempty"`
 	Size             string          `json:"size,omitempty"`
-	NumCtx           int         	 `json:"num_ctx,omitempty"`
 }
 
 func (r GeneralOpenAIRequest) ParseInput() []string {

relay/model/image.go

@@ -1,12 +1,12 @@
 package model
 
 type ImageRequest struct {
-	Model          string `json:"model" form:"model"`
+	Model          string `json:"model"`
-	Prompt         string `json:"prompt" binding:"required" form:"prompt"`
+	Prompt         string `json:"prompt" binding:"required"`
-	N              int    `json:"n,omitempty" form:"n"`
+	N              int    `json:"n,omitempty"`
-	Size           string `json:"size,omitempty" form:"size"`
+	Size           string `json:"size,omitempty"`
-	Quality        string `json:"quality,omitempty" form:"quality"`
+	Quality        string `json:"quality,omitempty"`
-	ResponseFormat string `json:"response_format,omitempty" form:"response_format"`
+	ResponseFormat string `json:"response_format,omitempty"`
-	Style          string `json:"style,omitempty" form:"style"`
+	Style          string `json:"style,omitempty"`
-	User           string `json:"user,omitempty" form:"user"`
+	User           string `json:"user,omitempty"`
 }

relay/relaymode/define.go

@@ -11,7 +11,6 @@ const (
 	AudioSpeech
 	AudioTranscription
 	AudioTranslation
-	ImagesEdits
 	// Proxy is a special relay mode for proxying requests to custom upstream
 	Proxy
 )

relay/relaymode/helper.go

@@ -24,11 +24,8 @@ func GetByPath(path string) int {
 		relayMode = AudioTranscription
 	} else if strings.HasPrefix(path, "/v1/audio/translations") {
 		relayMode = AudioTranslation
-	} else if strings.HasPrefix(path, "/v1/images/edits") {
-		relayMode = ImagesEdits
 	} else if strings.HasPrefix(path, "/v1/oneapi/proxy") {
 		relayMode = Proxy
 	}
-
 	return relayMode
 }

router/relay.go

@@ -24,7 +24,7 @@ func SetRelayRouter(router *gin.Engine) {
 		relayV1Router.POST("/chat/completions", controller.Relay)
 		relayV1Router.POST("/edits", controller.Relay)
 		relayV1Router.POST("/images/generations", controller.Relay)
-		relayV1Router.POST("/images/edits", controller.Relay)
+		relayV1Router.POST("/images/edits", controller.RelayNotImplemented)
 		relayV1Router.POST("/images/variations", controller.RelayNotImplemented)
 		relayV1Router.POST("/embeddings", controller.Relay)
 		relayV1Router.POST("/engines/:model/embeddings", controller.Relay)

web/air/src/constants/channel.constants.js

@@ -29,7 +29,6 @@ export const CHANNEL_OPTIONS = [
   { key: 39, text: 'together.ai', value: 39, color: 'blue' },
   { key: 42, text: 'VertexAI', value: 42, color: 'blue' },
   { key: 43, text: 'Proxy', value: 43, color: 'blue' },
-  { key: 44, text: 'SiliconFlow', value: 44, color: 'blue' },
   { key: 8, text: '自定义渠道', value: 8, color: 'pink' },
   { key: 22, text: '知识库：FastGPT', value: 22, color: 'blue' },
   { key: 21, text: '知识库：AI Proxy', value: 21, color: 'purple' },

web/berry/src/constants/ChannelConstants.js

@@ -173,12 +173,6 @@ export const CHANNEL_OPTIONS = {
     value: 43,
     color: 'primary'
   },
-  44: {
-    key: 44,
-    text: 'SiliconFlow',
-    value: 44,
-    color: 'primary'
-  },
   41: {
     key: 41,
     text: 'Novita',

web/default/src/constants/channel.constants.js

@@ -29,7 +29,6 @@ export const CHANNEL_OPTIONS = [
     { key: 39, text: 'together.ai', value: 39, color: 'blue' },
     { key: 42, text: 'VertexAI', value: 42, color: 'blue' },
     { key: 43, text: 'Proxy', value: 43, color: 'blue' },
-    { key: 44, text: 'SiliconFlow', value: 44, color: 'blue' },
     { key: 8, text: '自定义渠道', value: 8, color: 'pink' },
     { key: 22, text: '知识库：FastGPT', value: 22, color: 'blue' },
     { key: 21, text: '知识库：AI Proxy', value: 21, color: 'purple' },