Merge c44928b3a7 into f9774698e9

update groq add gemma2-9b-it llama3.1 family fixup price k/token -> m/token

README.md

@@ -89,6 +89,7 @@ _✨ 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。
@@ -251,9 +252,9 @@ 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。
+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。
 
-可安装 [Switcher 插件](https://github.com/RockChinQ/Switcher)在运行时切换所使用的模型。
+运行期间可以通过`!model`命令查看、切换可用模型。
 
 ### 部署到第三方平台
 <details>

relay/adaptor/cloudflare/constant.go

@@ -1,6 +1,7 @@
 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/gemini/constants.go

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

relay/adaptor/groq/constants.go

@@ -4,9 +4,14 @@ 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/embeddings", meta.BaseURL)
+		fullRequestURL = fmt.Sprintf("%s/api/embed", meta.BaseURL)
 	}
 	return fullRequestURL, nil
 }

relay/adaptor/ollama/main.go

@@ -31,6 +31,8 @@ func ConvertRequest(request model.GeneralOpenAIRequest) *ChatRequest {
 			TopP:             request.TopP,
 			FrequencyPenalty: request.FrequencyPenalty,
 			PresencePenalty:  request.PresencePenalty,
+			NumPredict:  	  request.MaxTokens,
+			NumCtx:  	  request.NumCtx,
 		},
 		Stream: request.Stream,
 	}
@@ -118,8 +120,10 @@ func StreamHandler(c *gin.Context, resp *http.Response) (*model.ErrorWithStatusC
 	common.SetEventStreamHeaders(c)
 
 	for scanner.Scan() {
-		data := strings.TrimPrefix(scanner.Text(), "}")
-		data = data + "}"
+		data := scanner.Text()
+		if strings.HasPrefix(data, "}") {
+		    data = strings.TrimPrefix(data, "}") + "}"
+		}
 
 		var ollamaResponse ChatResponse
 		err := json.Unmarshal([]byte(data), &ollamaResponse)
@@ -157,8 +161,15 @@ func StreamHandler(c *gin.Context, resp *http.Response) (*model.ErrorWithStatusC
 
 func ConvertEmbeddingRequest(request model.GeneralOpenAIRequest) *EmbeddingRequest {
 	return &EmbeddingRequest{
-		Model:  request.Model,
-		Prompt: strings.Join(request.ParseInput(), " "),
+		Model: request.Model,
+		Input: request.ParseInput(),
+		Options: &Options{
+			Seed:             int(request.Seed),
+			Temperature:      request.Temperature,
+			TopP:             request.TopP,
+			FrequencyPenalty: request.FrequencyPenalty,
+			PresencePenalty:  request.PresencePenalty,
+		},
 	}
 }
 
@@ -201,15 +212,17 @@ func embeddingResponseOllama2OpenAI(response *EmbeddingResponse) *openai.Embeddi
 	openAIEmbeddingResponse := openai.EmbeddingResponse{
 		Object: "list",
 		Data:   make([]openai.EmbeddingResponseItem, 0, 1),
-		Model:  "text-embedding-v1",
+		Model:  response.Model,
 		Usage:  model.Usage{TotalTokens: 0},
 	}
 
-	openAIEmbeddingResponse.Data = append(openAIEmbeddingResponse.Data, openai.EmbeddingResponseItem{
-		Object:    `embedding`,
-		Index:     0,
-		Embedding: response.Embedding,
-	})
+	for i, embedding := range response.Embeddings {
+		openAIEmbeddingResponse.Data = append(openAIEmbeddingResponse.Data, openai.EmbeddingResponseItem{
+			Object:    `embedding`,
+			Index:     i,
+			Embedding: embedding,
+		})
+	}
 	return &openAIEmbeddingResponse
 }
 

relay/adaptor/ollama/model.go

@@ -7,6 +7,8 @@ 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 {
@@ -37,11 +39,15 @@ type ChatResponse struct {
 }
 
 type EmbeddingRequest struct {
-	Model  string `json:"model"`
-	Prompt string `json:"prompt"`
+	Model string   `json:"model"`
+	Input []string `json:"input"`
+	// Truncate  bool     `json:"truncate,omitempty"`
+	Options *Options `json:"options,omitempty"`
+	// KeepAlive string   `json:"keep_alive,omitempty"`
 }
 
 type EmbeddingResponse struct {
-	Error     string    `json:"error,omitempty"`
-	Embedding []float64 `json:"embedding,omitempty"`
+	Error      string      `json:"error,omitempty"`
+	Model      string      `json:"model"`
+	Embeddings [][]float64 `json:"embeddings"`
 }

relay/adaptor/openai/compatible.go

@@ -13,6 +13,7 @@ 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"
 )
 
@@ -30,6 +31,7 @@ var CompatibleChannels = []int{
 	channeltype.DeepSeek,
 	channeltype.TogetherAI,
 	channeltype.Novita,
+	channeltype.SiliconFlow,
 }
 
 func GetCompatibleChannelMeta(channelType int) (string, []string) {
@@ -60,6 +62,8 @@ 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/siliconflow/constants.go

@@ -0,0 +1,36 @@
+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

@@ -98,12 +98,11 @@ 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-pro-vision":         1, // $0.00025 / 1k characters -> $0.001 / 1k tokens
-	"gemini-1.0-pro-vision-001": 1,
-	"gemini-1.0-pro-001":        1,
-	"gemini-1.5-pro":            1,
+	"gemini-pro":       1, // $0.00025 / 1k characters -> $0.001 / 1k tokens
+	"gemini-1.0-pro":   1,
+	"gemini-1.5-flash": 1,
+	"gemini-1.5-pro":   1,
+	"aqa":              1,
 	// https://open.bigmodel.cn/pricing
 	"glm-4":         0.1 * RMB,
 	"glm-4v":        0.1 * RMB,
@@ -158,12 +157,16 @@ 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
-	"llama3-70b-8192":    0.59 / 1000 * USD,
-	"mixtral-8x7b-32768": 0.27 / 1000 * USD,
-	"llama3-8b-8192":     0.05 / 1000 * USD,
-	"gemma-7b-it":        0.1 / 1000 * USD,
-	"llama2-70b-4096":    0.64 / 1000 * USD,
-	"llama2-7b-2048":     0.1 / 1000 * USD,
+	"gemma-7b-it":                           0.07 / 1000000 * USD,
+	"mixtral-8x7b-32768":                    0.24 / 1000000 * USD,
+	"llama3-8b-8192":                        0.05 / 1000000 * USD,
+	"llama3-70b-8192":                       0.59 / 1000000 * USD,
+	"gemma2-9b-it":                          0.20 / 1000000 * USD,
+	"llama-3.1-405b-reasoning":              0.89 / 1000000 * 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,

relay/channeltype/define.go

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

relay/channeltype/url.go

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

relay/model/general.go

@@ -29,6 +29,7 @@ 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 {

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

@@ -29,6 +29,7 @@ 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

@@ -173,6 +173,12 @@ 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

@@ -29,6 +29,7 @@ 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' },