perf: token params

src/service/plugins/searchKb.ts

@@ -29,6 +29,11 @@ export const searchKb = async ({
   }[];
 }> => {
   async function search(textArr: string[] = []) {
+    const limitMap: Record<ModelVectorSearchModeEnum, number> = {
+      [ModelVectorSearchModeEnum.hightSimilarity]: 15,
+      [ModelVectorSearchModeEnum.noContext]: 15,
+      [ModelVectorSearchModeEnum.lowSimilarity]: 20
+    };
     // 获取提示词的向量
     const { vectors: promptVectors } = await openaiCreateEmbedding({
       userOpenAiKey,
@@ -48,7 +53,7 @@ export const searchKb = async ({
             `vector <=> '[${promptVector}]' < ${similarity}`
           ],
           order: [{ field: 'vector', mode: `<=> '[${promptVector}]'` }],
-          limit: 20
+          limit: limitMap[model.chat.searchMode]
         }).then((res) => res.rows)
       )
     );

src/service/utils/chat/claude.ts

@@ -40,7 +40,7 @@ export const lafClaudChat = async ({
       headers: {
         Authorization: apiKey
       },
-      timeout: stream ? 40000 : 240000,
+      timeout: stream ? 60000 : 240000,
       responseType: stream ? 'stream' : 'json'
     }
   );

src/service/utils/chat/openai.ts

@@ -73,7 +73,7 @@ export const chatResponse = async ({
   const filterMessages = ChatContextFilter({
     model,
     prompts: messages,
-    maxTokens: Math.ceil(ChatModelMap[model].contextMaxToken * 0.9)
+    maxTokens: Math.ceil(ChatModelMap[model].contextMaxToken * 0.85)
   });
 
   const adaptMessages = adaptChatItem_openAI({ messages: filterMessages });
@@ -90,7 +90,7 @@ export const chatResponse = async ({
       stop: ['.!?。']
     },
     {
-      timeout: stream ? 40000 : 240000,
+      timeout: stream ? 60000 : 240000,
       responseType: stream ? 'stream' : 'json',
       ...axiosConfig()
     }