✨ feat: Add support for retrieving model list from providers (#188)

* ✨ feat: Add support for retrieving model list from providers

* 🔖 chore: Custom channel automatically get the model

controller/channel-model.go

@@ -0,0 +1,58 @@
+package controller
+
+import (
+	"net/http"
+	"one-api/model"
+	"one-api/providers"
+	providersBase "one-api/providers/base"
+	"strings"
+
+	"github.com/gin-gonic/gin"
+)
+
+func GetModelList(c *gin.Context) {
+	channel := model.Channel{}
+	err := c.ShouldBindJSON(&channel)
+	if err != nil {
+		c.JSON(http.StatusOK, gin.H{
+			"success": false,
+			"message": err.Error(),
+		})
+		return
+	}
+	keys := strings.Split(channel.Key, "\n")
+	channel.Key = keys[0]
+
+	provider := providers.GetProvider(&channel, c)
+	if provider == nil {
+		c.JSON(http.StatusOK, gin.H{
+			"success": false,
+			"message": "provider not found",
+		})
+		return
+	}
+
+	modelProvider, ok := provider.(providersBase.ModelListInterface)
+	if !ok {
+		c.JSON(http.StatusOK, gin.H{
+			"success": false,
+			"message": "channel not implemented",
+		})
+		return
+	}
+
+	modelList, err := modelProvider.GetModelList()
+	if err != nil {
+		c.JSON(http.StatusOK, gin.H{
+			"success": false,
+			"message": err.Error(),
+		})
+		return
+	}
+
+	c.JSON(http.StatusOK, gin.H{
+		"success": true,
+		"message": "",
+		"data":    modelList,
+	})
+}

providers/base/common.go

@@ -25,6 +25,7 @@ type ProviderConfig struct {
 	ImagesGenerations   string
 	ImagesEdit          string
 	ImagesVariations    string
+	ModelList           string
 }
 
 type BaseProvider struct {

providers/base/interface.go

@@ -99,6 +99,16 @@ type ImageVariationsInterface interface {
 	CreateImageVariations(request *types.ImageEditRequest) (*types.ImageResponse, *types.OpenAIErrorWithStatusCode)
 }
 
+// type RelayInterface interface {
+// 	ProviderInterface
+// 	CreateRelay() (*http.Response, *types.OpenAIErrorWithStatusCode)
+// }
+
+type ModelListInterface interface {
+	ProviderInterface
+	GetModelList() ([]string, error)
+}
+
 // 余额接口
 type BalanceInterface interface {
 	Balance() (float64, error)

providers/cohere/base.go

@@ -32,6 +32,7 @@ func getConfig() base.ProviderConfig {
 	return base.ProviderConfig{
 		BaseURL:         "https://api.cohere.ai/v1",
 		ChatCompletions: "/chat",
+		ModelList:       "/models",
 	}
 }
 

providers/cohere/model.go

@@ -0,0 +1,35 @@
+package cohere
+
+import (
+	"errors"
+	"net/http"
+	"net/url"
+)
+
+func (p *CohereProvider) GetModelList() ([]string, error) {
+	params := url.Values{}
+	params.Add("page_size", "1000")
+	params.Add("endpoint", "chat")
+	queryString := params.Encode()
+
+	fullRequestURL := p.GetFullRequestURL(p.Config.ModelList) + "?" + queryString
+	headers := p.GetRequestHeaders()
+
+	req, err := p.Requester.NewRequest(http.MethodGet, fullRequestURL, p.Requester.WithHeader(headers))
+	if err != nil {
+		return nil, errors.New("new_request_failed")
+	}
+
+	response := &ModelListResponse{}
+	_, errWithCode := p.Requester.SendRequest(req, response, false)
+	if errWithCode != nil {
+		return nil, errors.New(errWithCode.Message)
+	}
+
+	var modelList []string
+	for _, model := range response.Models {
+		modelList = append(modelList, model.Name)
+	}
+
+	return modelList, nil
+}

providers/cohere/type.go

@@ -15,28 +15,34 @@ type CohereConnector struct {
 }
 
 type CohereRequest struct {
-	Message          string                          `json:"message"`
+	Message           string                          `json:"message"`
-	Model            string                          `json:"model,omitempty"`
+	Model             string                          `json:"model,omitempty"`
-	Stream           bool                            `json:"stream,omitempty"`
+	Stream            bool                            `json:"stream,omitempty"`
-	Preamble         string                          `json:"preamble,omitempty"`
+	Preamble          string                          `json:"preamble,omitempty"`
-	ChatHistory      []ChatHistory                   `json:"chat_history,omitempty"`
+	ChatHistory       []ChatHistory                   `json:"chat_history,omitempty"`
-	ConversationId   string                          `json:"conversation_id,omitempty"`
+	ConversationId    string                          `json:"conversation_id,omitempty"`
-	PromptTruncation string                          `json:"prompt_truncation,omitempty"`
+	PromptTruncation  string                          `json:"prompt_truncation,omitempty"`
-	Connectors       []CohereConnector               `json:"connectors,omitempty"`
+	Connectors        []CohereConnector               `json:"connectors,omitempty"`
-	Temperature      float64                         `json:"temperature,omitempty"`
+	Temperature       float64                         `json:"temperature,omitempty"`
-	MaxTokens        int                             `json:"max_tokens,omitempty"`
+	MaxTokens         int                             `json:"max_tokens,omitempty"`
-	MaxInputTokens   int                             `json:"max_input_tokens,omitempty"`
+	MaxInputTokens    int                             `json:"max_input_tokens,omitempty"`
-	K                int                             `json:"k,omitempty"`
+	K                 int                             `json:"k,omitempty"`
-	P                float64                         `json:"p,omitempty"`
+	P                 float64                         `json:"p,omitempty"`
-	Seed             *int                            `json:"seed,omitempty"`
+	Seed              *int                            `json:"seed,omitempty"`
-	StopSequences    any                             `json:"stop_sequences,omitempty"`
+	StopSequences     any                             `json:"stop_sequences,omitempty"`
-	FrequencyPenalty float64                         `json:"frequency_penalty,omitempty"`
+	FrequencyPenalty  float64                         `json:"frequency_penalty,omitempty"`
-	PresencePenalty  float64                         `json:"presence_penalty,omitempty"`
+	PresencePenalty   float64                         `json:"presence_penalty,omitempty"`
-	Tools            []*types.ChatCompletionFunction `json:"tools,omitempty"`
+	Tools             []*types.ChatCompletionFunction `json:"tools,omitempty"`
-	ToolResults      any                             `json:"tool_results,omitempty"`
+	ToolResults       any                             `json:"tool_results,omitempty"`
-	// SearchQueriesOnly bool              `json:"search_queries_only,omitempty"`
+	SearchQueriesOnly *bool                           `json:"search_queries_only,omitempty"`
+	Documents         []ChatDocument                  `json:"documents,omitempty"`
+	CitationQuality   *string                         `json:"citation_quality,omitempty"`
+	RawPrompting      *bool                           `json:"raw_prompting,omitempty"`
+	ReturnPrompt      *bool                           `json:"return_prompt,omitempty"`
 }
 
+type ChatDocument = map[string]string
+
 type APIVersion struct {
 	Version string `json:"version"`
 }
@@ -60,16 +66,46 @@ type CohereToolCall struct {
 }
 
 type CohereResponse struct {
-	Text         string           `json:"text,omitempty"`
+	Text             string              `json:"text,omitempty"`
-	ResponseID   string           `json:"response_id,omitempty"`
+	ResponseID       string              `json:"response_id,omitempty"`
-	GenerationID string           `json:"generation_id,omitempty"`
+	Citations        []*ChatCitation     `json:"citations,omitempty"`
-	ChatHistory  []ChatHistory    `json:"chat_history,omitempty"`
+	Documents        []ChatDocument      `json:"documents,omitempty"`
-	FinishReason string           `json:"finish_reason,omitempty"`
+	IsSearchRequired *bool               `json:"is_search_required,omitempty"`
-	ToolCalls    []CohereToolCall `json:"tool_calls,omitempty"`
+	SearchQueries    []*ChatSearchQuery  `json:"search_queries,omitempty"`
-	Meta         Meta             `json:"meta,omitempty"`
+	SearchResults    []*ChatSearchResult `json:"search_results,omitempty"`
+	GenerationID     string              `json:"generation_id,omitempty"`
+	ChatHistory      []ChatHistory       `json:"chat_history,omitempty"`
+	Prompt           *string             `json:"prompt,omitempty"`
+	FinishReason     string              `json:"finish_reason,omitempty"`
+	ToolCalls        []CohereToolCall    `json:"tool_calls,omitempty"`
+	Meta             Meta                `json:"meta,omitempty"`
 	CohereError
 }
 
+type ChatCitation struct {
+	Start       int      `json:"start"`
+	End         int      `json:"end"`
+	Text        string   `json:"text"`
+	DocumentIds []string `json:"document_ids,omitempty"`
+}
+
+type ChatSearchQuery struct {
+	Text         string `json:"text"`
+	GenerationId string `json:"generation_id"`
+}
+
+type ChatSearchResult struct {
+	SearchQuery       *ChatSearchQuery           `json:"search_query,omitempty" url:"search_query,omitempty"`
+	Connector         *ChatSearchResultConnector `json:"connector,omitempty" url:"connector,omitempty"`
+	DocumentIds       []string                   `json:"document_ids,omitempty" url:"document_ids,omitempty"`
+	ErrorMessage      *string                    `json:"error_message,omitempty" url:"error_message,omitempty"`
+	ContinueOnFailure *bool                      `json:"continue_on_failure,omitempty" url:"continue_on_failure,omitempty"`
+}
+
+type ChatSearchResultConnector struct {
+	Id string `json:"id" url:"id"`
+}
+
 type CohereError struct {
 	Message string `json:"message,omitempty"`
 }
@@ -83,3 +119,77 @@ type CohereStreamResponse struct {
 	FinishReason string           `json:"finish_reason,omitempty"`
 	ToolCalls    []CohereToolCall `json:"tool_calls,omitempty"`
 }
+
+type RerankRequest struct {
+	Model           *string                       `json:"model,omitempty"`
+	Query           string                        `json:"query" url:"query"`
+	Documents       []*RerankRequestDocumentsItem `json:"documents,omitempty"`
+	TopN            *int                          `json:"top_n,omitempty"`
+	RankFields      []string                      `json:"rank_fields,omitempty"`
+	ReturnDocuments *bool                         `json:"return_documents,omitempty"`
+	MaxChunksPerDoc *int                          `json:"max_chunks_per_doc,omitempty"`
+}
+
+type RerankRequestDocumentsItem struct {
+	String                         string
+	RerankRequestDocumentsItemText *RerankDocumentsItemText
+}
+type RerankDocumentsItemText struct {
+	Text string `json:"text"`
+}
+
+type RerankResponse struct {
+	Id      *string                      `json:"id,omitempty"`
+	Results []*RerankResponseResultsItem `json:"results,omitempty"`
+	Meta    *Meta                        `json:"meta,omitempty"`
+}
+
+type RerankResponseResultsItem struct {
+	Document       *RerankDocumentsItemText `json:"document,omitempty"`
+	Index          int                      `json:"index"`
+	RelevanceScore float64                  `json:"relevance_score"`
+}
+
+type EmbedRequest struct {
+	Texts          any      `json:"texts,omitempty"`
+	Model          *string  `json:"model,omitempty"`
+	InputType      *string  `json:"input_type,omitempty"`
+	EmbeddingTypes []string `json:"embedding_types,omitempty"`
+	Truncate       *string  `json:"truncate,omitempty"`
+}
+
+type EmbedResponse struct {
+	ResponseType string `json:"response_type"`
+	Embeddings   any    `json:"embeddings"`
+}
+
+type EmbedFloatsResponse struct {
+	Id         string      `json:"id"`
+	Embeddings [][]float64 `json:"embeddings,omitempty"`
+	Texts      []string    `json:"texts,omitempty"`
+	Meta       *Meta       `json:"meta,omitempty"`
+}
+
+type EmbedByTypeResponse struct {
+	Id         string                         `json:"id"`
+	Embeddings *EmbedByTypeResponseEmbeddings `json:"embeddings,omitempty"`
+	Texts      []string                       `json:"texts,omitempty"`
+	Meta       *Meta                          `json:"meta,omitempty"`
+}
+
+type EmbedByTypeResponseEmbeddings struct {
+	Float   [][]float64 `json:"float,omitempty"`
+	Int8    [][]int     `json:"int8,omitempty"`
+	Uint8   [][]int     `json:"uint8,omitempty"`
+	Binary  [][]int     `json:"binary,omitempty"`
+	Ubinary [][]int     `json:"ubinary,omitempty"`
+}
+
+type ModelListResponse struct {
+	Models []ModelDetails `json:"models"`
+}
+
+type ModelDetails struct {
+	Name      string   `json:"name"`
+	Endpoints []string `json:"endpoints"`
+}

providers/deepseek/base.go

@@ -28,6 +28,7 @@ func getDeepseekConfig() base.ProviderConfig {
 	return base.ProviderConfig{
 		BaseURL:         "https://api.deepseek.com",
 		ChatCompletions: "/v1/chat/completions",
+		ModelList:       "/v1/models",
 	}
 }
 

providers/gemini/base.go

@@ -32,6 +32,7 @@ func getConfig() base.ProviderConfig {
 	return base.ProviderConfig{
 		BaseURL:         "https://generativelanguage.googleapis.com",
 		ChatCompletions: "/",
+		ModelList:       "/models",
 	}
 }
 

providers/gemini/model.go

@@ -0,0 +1,45 @@
+package gemini
+
+import (
+	"errors"
+	"fmt"
+	"net/http"
+	"net/url"
+	"strings"
+)
+
+func (p *GeminiProvider) GetModelList() ([]string, error) {
+	params := url.Values{}
+	params.Add("page_size", "1000")
+	queryString := params.Encode()
+
+	baseURL := strings.TrimSuffix(p.GetBaseURL(), "/")
+	version := "v1beta"
+	fullRequestURL := fmt.Sprintf("%s/%s%s?%s", baseURL, version, p.Config.ModelList, queryString)
+
+	headers := p.GetRequestHeaders()
+
+	req, err := p.Requester.NewRequest(http.MethodGet, fullRequestURL, p.Requester.WithHeader(headers))
+	if err != nil {
+		return nil, errors.New("new_request_failed")
+	}
+
+	response := &ModelListResponse{}
+	_, errWithCode := p.Requester.SendRequest(req, response, false)
+	if errWithCode != nil {
+		return nil, errors.New(errWithCode.Message)
+	}
+
+	var modelList []string
+	for _, model := range response.Models {
+		for _, modelType := range model.SupportedGenerationMethods {
+			if modelType == "generateContent" {
+				modelName := strings.TrimPrefix(model.Name, "models/")
+				modelList = append(modelList, modelName)
+				break
+			}
+		}
+	}
+
+	return modelList, nil
+}

providers/gemini/type.go

@@ -218,3 +218,12 @@ func ConvertRole(roleName string) string {
 		return types.ChatMessageRoleUser
 	}
 }
+
+type ModelListResponse struct {
+	Models []ModelDetails `json:"models"`
+}
+
+type ModelDetails struct {
+	Name                       string   `json:"name"`
+	SupportedGenerationMethods []string `json:"supportedGenerationMethods"`
+}

providers/groq/base.go

@@ -27,6 +27,7 @@ func getConfig() base.ProviderConfig {
 	return base.ProviderConfig{
 		BaseURL:         "https://api.groq.com/openai",
 		ChatCompletions: "/v1/chat/completions",
+		ModelList:       "/v1/models",
 	}
 }
 

providers/mistral/base.go

@@ -41,6 +41,7 @@ func getMistralConfig(baseURL string) base.ProviderConfig {
 		BaseURL:         baseURL,
 		ChatCompletions: "/v1/chat/completions",
 		Embeddings:      "/v1/embeddings",
+		ModelList:       "/v1/models",
 	}
 }
 

providers/mistral/model.go

@@ -0,0 +1,29 @@
+package mistral
+
+import (
+	"errors"
+	"net/http"
+)
+
+func (p *MistralProvider) GetModelList() ([]string, error) {
+	fullRequestURL := p.GetFullRequestURL(p.Config.ModelList, "")
+	headers := p.GetRequestHeaders()
+
+	req, err := p.Requester.NewRequest(http.MethodGet, fullRequestURL, p.Requester.WithHeader(headers))
+	if err != nil {
+		return nil, errors.New("new_request_failed")
+	}
+
+	response := &ModelListResponse{}
+	_, errWithCode := p.Requester.SendRequest(req, response, false)
+	if errWithCode != nil {
+		return nil, errors.New(errWithCode.Message)
+	}
+
+	var modelList []string
+	for _, model := range response.Data {
+		modelList = append(modelList, model.Id)
+	}
+
+	return modelList, nil
+}

providers/mistral/type.go

@@ -53,3 +53,15 @@ type ChatCompletionStreamResponse struct {
 	types.ChatCompletionStreamResponse
 	Usage *types.Usage `json:"usage,omitempty"`
 }
+
+type ModelListResponse struct {
+	Object string         `json:"object"`
+	Data   []ModelDetails `json:"data"`
+}
+
+type ModelDetails struct {
+	Id      string `json:"id"`
+	Object  string `json:"object"`
+	Created int64  `json:"created"`
+	OwnedBy string `json:"owned_by"`
+}

providers/openai/base.go

@@ -57,6 +57,7 @@ func getOpenAIConfig(baseURL string) base.ProviderConfig {
 		ImagesGenerations:   "/v1/images/generations",
 		ImagesEdit:          "/v1/images/edits",
 		ImagesVariations:    "/v1/images/variations",
+		ModelList:           "/v1/models",
 	}
 }
 

providers/openai/model.go

@@ -0,0 +1,29 @@
+package openai
+
+import (
+	"errors"
+	"net/http"
+)
+
+func (p *OpenAIProvider) GetModelList() ([]string, error) {
+	fullRequestURL := p.GetFullRequestURL(p.Config.ModelList, "")
+	headers := p.GetRequestHeaders()
+
+	req, err := p.Requester.NewRequest(http.MethodGet, fullRequestURL, p.Requester.WithHeader(headers))
+	if err != nil {
+		return nil, errors.New("new_request_failed")
+	}
+
+	response := &ModelListResponse{}
+	_, errWithCode := p.Requester.SendRequest(req, response, false)
+	if errWithCode != nil {
+		return nil, errors.New(errWithCode.Message)
+	}
+
+	var modelList []string
+	for _, model := range response.Data {
+		modelList = append(modelList, model.Id)
+	}
+
+	return modelList, nil
+}

providers/openai/type.go

@@ -73,3 +73,15 @@ type OpenAIUsageResponse struct {
 	//DailyCosts []OpenAIUsageDailyCost `json:"daily_costs"`
 	TotalUsage float64 `json:"total_usage"` // unit: 0.01 dollar
 }
+
+type ModelListResponse struct {
+	Object string         `json:"object"`
+	Data   []ModelDetails `json:"data"`
+}
+
+type ModelDetails struct {
+	Id      string `json:"id"`
+	Object  string `json:"object"`
+	Created int64  `json:"created"`
+	OwnedBy string `json:"owned_by"`
+}

router/api-router.go

@@ -78,6 +78,7 @@ func SetApiRouter(router *gin.Engine) {
 		{
 			channelRoute.GET("/", controller.GetChannelsList)
 			channelRoute.GET("/models", relay.ListModelsForAdmin)
+			channelRoute.POST("/provider_models_list", controller.GetModelList)
 			channelRoute.GET("/:id", controller.GetChannel)
 			channelRoute.GET("/test", controller.TestAllChannels)
 			channelRoute.GET("/test/:id", controller.TestChannel)

web/src/views/Channel/component/EditModal.js

@@ -24,8 +24,10 @@ import {
   Checkbox,
   Switch,
   FormControlLabel,
-  Typography
+  Typography,
+  Tooltip
 } from '@mui/material';
+import LoadingButton from '@mui/lab/LoadingButton';
 
 import { Formik } from 'formik';
 import * as Yup from 'yup';
@@ -78,6 +80,7 @@ const EditModal = ({ open, channelId, onCancel, onOk, groupOptions }) => {
   const [inputPrompt, setInputPrompt] = useState(defaultConfig.prompt);
   const [modelOptions, setModelOptions] = useState([]);
   const [batchAdd, setBatchAdd] = useState(false);
+  const [providerModelsLoad, setProviderModelsLoad] = useState(false);
 
   const initChannel = (typeValue) => {
     if (typeConfig[typeValue]?.inputLabel) {
@@ -144,6 +147,22 @@ const EditModal = ({ open, channelId, onCancel, onOk, groupOptions }) => {
     return modelList;
   };
 
+  const getProviderModels = async (values, setFieldValue) => {
+    setProviderModelsLoad(true);
+    try {
+      const res = await API.post(`/api/channel/provider_models_list`, { ...values, models: '' });
+      const { success, message, data } = res.data;
+      if (success && data) {
+        setFieldValue('models', data);
+      } else {
+        showError(message || '获取模型列表失败');
+      }
+    } catch (error) {
+      showError(error.message);
+    }
+    setProviderModelsLoad(false);
+  };
+
   const fetchModels = async () => {
     try {
       let res = await API.get(`/api/channel/models`);
@@ -505,6 +524,18 @@ const EditModal = ({ open, channelId, onCancel, onOk, groupOptions }) => {
                   >
                     填入所有模型
                   </Button>
+                  {inputLabel.provider_models_list && (
+                    <Tooltip title={inputPrompt.provider_models_list} placement="top">
+                      <LoadingButton
+                        loading={providerModelsLoad}
+                        onClick={() => {
+                          getProviderModels(values, setFieldValue);
+                        }}
+                      >
+                        {inputLabel.provider_models_list}
+                      </LoadingButton>
+                    </Tooltip>
+                  )}
                 </ButtonGroup>
               </Container>
               <FormControl fullWidth error={Boolean(touched.key && errors.key)} sx={{ ...theme.typography.otherInput }}>

web/src/views/Channel/type/Config.js

@@ -24,7 +24,8 @@ const defaultConfig = {
     models: '模型',
     model_mapping: '模型映射关系',
     groups: '用户组',
-    only_chat: '仅支持聊天'
+    only_chat: '仅支持聊天',
+    provider_models_list: ''
   },
   prompt: {
     type: '请选择渠道类型',
@@ -39,12 +40,23 @@ const defaultConfig = {
     model_mapping:
       '请输入要修改的模型映射关系，格式为：api请求模型ID:实际转发给渠道的模型ID，使用JSON数组表示，例如：{"gpt-3.5": "gpt-35"}',
     groups: '请选择该渠道所支持的用户组',
-    only_chat: '如果选择了仅支持聊天，那么遇到有函数调用的请求会跳过该渠道'
+    only_chat: '如果选择了仅支持聊天，那么遇到有函数调用的请求会跳过该渠道',
+    provider_models_list: '必须填写所有数据后才能获取模型列表'
   },
   modelGroup: 'OpenAI'
 };
 
 const typeConfig = {
+  1: {
+    inputLabel: {
+      provider_models_list: '从OpenAI获取模型列表'
+    }
+  },
+  8: {
+    inputLabel: {
+      provider_models_list: '从渠道获取模型列表'
+    }
+  },
   3: {
     inputLabel: {
       base_url: 'AZURE_OPENAI_ENDPOINT',
@@ -143,7 +155,8 @@ const typeConfig = {
   },
   25: {
     inputLabel: {
-      other: '版本号'
+      other: '版本号',
+      provider_models_list: '从Gemini获取模型列表'
     },
     input: {
       models: ['gemini-pro', 'gemini-pro-vision', 'gemini-1.0-pro', 'gemini-1.5-pro'],
@@ -189,6 +202,9 @@ const typeConfig = {
       models: ['deepseek-coder', 'deepseek-chat'],
       test_model: 'deepseek-chat'
     },
+    inputLabel: {
+      provider_models_list: '从Deepseek获取模型列表'
+    },
     modelGroup: 'Deepseek'
   },
   29: {
@@ -210,6 +226,9 @@ const typeConfig = {
       ],
       test_model: 'open-mistral-7b'
     },
+    inputLabel: {
+      provider_models_list: '从Mistral获取模型列表'
+    },
     modelGroup: 'Mistral'
   },
   31: {
@@ -217,6 +236,9 @@ const typeConfig = {
       models: ['llama2-7b-2048', 'llama2-70b-4096', 'mixtral-8x7b-32768', 'gemma-7b-it'],
       test_model: 'llama2-7b-2048'
     },
+    inputLabel: {
+      provider_models_list: '从Groq获取模型列表'
+    },
     modelGroup: 'Groq'
   },
   32: {
@@ -297,6 +319,9 @@ const typeConfig = {
       models: ['command-r', 'command-r-plus'],
       test_model: 'command-r'
     },
+    inputLabel: {
+      provider_models_list: '从Cohere获取模型列表'
+    },
     modelGroup: 'Cohere'
   },
   37: {