perf: bill

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

@@ -1,13 +1,8 @@
 import { ChatItemType } from '@/types/chat';
 import { modelToolMap } from '@/utils/plugin';
-import type { ChatModelType } from '@/constants/model';
-import { ChatRoleEnum, sseResponseEventEnum } from '@/constants/chat';
-import { sseResponse } from '../tools';
+import { ChatRoleEnum } from '@/constants/chat';
 import { OpenAiChatEnum } from '@/constants/model';
-import { chatResponse, openAiStreamResponse } from './openai';
 import type { NextApiResponse } from 'next';
-import { textAdaptGptResponse } from '@/utils/adapt';
-import { parseStreamChunk } from '@/utils/adapt';
 
 export type ChatCompletionType = {
   apiKey: string;
@@ -36,11 +31,6 @@ export type StreamResponseReturnType = {
   finishMessages: ChatItemType[];
 };
 
-export const modelServiceToolMap = {
-  chatCompletion: chatResponse,
-  streamResponse: openAiStreamResponse
-};
-
 /* delete invalid symbol */
 const simplifyStr = (str = '') =>
   str
@@ -54,7 +44,7 @@ export const ChatContextFilter = ({
   prompts,
   maxTokens
 }: {
-  model: ChatModelType;
+  model: string;
   prompts: ChatItemType[];
   maxTokens: number;
 }) => {
@@ -111,126 +101,3 @@ export const ChatContextFilter = ({
 
   return [...systemPrompts, ...chats];
 };
-
-/* stream response */
-export const resStreamResponse = async ({
-  model,
-  res,
-  chatResponse,
-  prompts
-}: StreamResponseType & {
-  model: ChatModelType;
-}) => {
-  // 创建响应流
-  res.setHeader('Content-Type', 'text/event-stream;charset=utf-8');
-  res.setHeader('Access-Control-Allow-Origin', '*');
-  res.setHeader('X-Accel-Buffering', 'no');
-  res.setHeader('Cache-Control', 'no-cache, no-transform');
-
-  const { responseContent, totalTokens, finishMessages } = await modelServiceToolMap.streamResponse(
-    {
-      chatResponse,
-      prompts,
-      res,
-      model
-    }
-  );
-
-  return { responseContent, totalTokens, finishMessages };
-};
-
-/* stream response */
-export const V2_StreamResponse = async ({
-  model,
-  res,
-  chatResponse,
-  prompts
-}: StreamResponseType & {
-  model: ChatModelType;
-}) => {
-  let responseContent = '';
-  let error: any = null;
-  let truncateData = '';
-  const clientRes = async (data: string) => {
-    //部分代理会导致流式传输时的数据被截断，不为json格式，这里做一个兼容
-    const { content = '' } = (() => {
-      try {
-        if (truncateData) {
-          try {
-            //判断是否为json，如果是的话直接跳过后续拼装操作，注意极端情况下可能出现截断成3截以上情况也可以兼容
-            JSON.parse(data);
-          } catch (e) {
-            data = truncateData + data;
-          }
-          truncateData = '';
-        }
-        const json = JSON.parse(data);
-        const content: string = json?.choices?.[0].delta.content || '';
-        error = json.error;
-        responseContent += content;
-        return { content };
-      } catch (error) {
-        truncateData = data;
-        return {};
-      }
-    })();
-
-    if (res.closed || error) return;
-
-    if (data === '[DONE]') {
-      sseResponse({
-        res,
-        event: sseResponseEventEnum.answer,
-        data: textAdaptGptResponse({
-          text: null,
-          finish_reason: 'stop'
-        })
-      });
-      sseResponse({
-        res,
-        event: sseResponseEventEnum.answer,
-        data: '[DONE]'
-      });
-    } else {
-      sseResponse({
-        res,
-        event: sseResponseEventEnum.answer,
-        data: textAdaptGptResponse({
-          text: content
-        })
-      });
-    }
-  };
-
-  try {
-    for await (const chunk of chatResponse.data as any) {
-      if (res.closed) break;
-      const parse = parseStreamChunk(chunk);
-      parse.forEach((item) => clientRes(item.data));
-    }
-  } catch (error) {
-    console.log('pipe error', error);
-  }
-
-  if (error) {
-    console.log(error);
-    return Promise.reject(error);
-  }
-
-  // count tokens
-  const finishMessages = prompts.concat({
-    obj: ChatRoleEnum.AI,
-    value: responseContent
-  });
-
-  const totalTokens = modelToolMap.countTokens({
-    model,
-    messages: finishMessages
-  });
-
-  return {
-    responseContent,
-    totalTokens,
-    finishMessages
-  };
-};