perf: template

client/src/service/moduleDispatch/agent/classifyQuestion.ts

@@ -51,7 +51,7 @@ export const dispatchClassifyQuestion = async (props: Record<string, any>): Prom
   //   function body
   const agentFunction = {
     name: agentFunName,
-    description: '判断用户问题的类型，并返回指定值',
+    description: '判断用户问题的类型属于哪方面，返回对应的枚举字段',
     parameters: {
       type: 'object',
       properties: {
@@ -81,13 +81,9 @@ export const dispatchClassifyQuestion = async (props: Record<string, any>): Prom
 
   const arg = JSON.parse(response.data.choices?.[0]?.message?.function_call?.arguments || '');
 
-  if (!arg.type) {
-    throw new Error('');
-  }
-
   const tokens = response.data.usage?.total_tokens || 0;
 
-  const result = agents.find((item) => item.key === arg.type) || agents[0];
+  const result = agents.find((item) => item.key === arg?.type) || agents[0];
 
   return {
     [result.key]: 1,

client/src/service/utils/chat/index.ts

@@ -31,7 +31,7 @@ const simplifyStr = (str = '') =>
     .replace(/[^\S\r\n]+/g, ' ') // 连续空白内容
     .trim();
 
-/* 聊天上下文 tokens 截断 */
+/* slice chat context by tokens */
 export const ChatContextFilter = ({
   model,
   prompts,
@@ -41,33 +41,19 @@ export const ChatContextFilter = ({
   prompts: ChatItemType[];
   maxTokens: number;
 }) => {
-  const systemPrompts: ChatItemType[] = [];
-  const chatPrompts: ChatItemType[] = [];
+  const rawTextLen = prompts.reduce((sum, item) => sum + item.value.length, 0);
 
-  let rawTextLen = 0;
-  prompts.forEach((item) => {
-    const val = simplifyStr(item.value);
-    rawTextLen += val.length;
-
-    const data = {
-      _id: item._id,
-      obj: item.obj,
-      value: val
-    };
-
-    if (item.obj === ChatRoleEnum.System) {
-      systemPrompts.push(data);
-    } else {
-      chatPrompts.push(data);
-    }
-  });
-
-  // 长度太小时，不需要进行 token 截断
+  // If the text length is less than half of the maximum token, no calculation is required
   if (rawTextLen < maxTokens * 0.5) {
-    return [...systemPrompts, ...chatPrompts];
+    return prompts;
   }
 
-  // 去掉 system 的 token
+  // filter startWith system prompt
+  const chatStartIndex = prompts.findIndex((item) => item.obj !== ChatRoleEnum.System);
+  const systemPrompts: ChatItemType[] = prompts.slice(0, chatStartIndex);
+  const chatPrompts: ChatItemType[] = prompts.slice(chatStartIndex);
+
+  // reduce  token of systemPrompt
   maxTokens -= modelToolMap.countTokens({
     model,
     messages: systemPrompts