perf: chat framwork

2023-05-03 15:28:25 +08:00
parent 91decc3683
commit 00a99261ae
23 changed files with 811 additions and 1011 deletions
--- a/src/pages/api/chat/chat.ts
+++ b/src/pages/api/chat/chat.ts
@@ -1,14 +1,15 @@
 import type { NextApiRequest, NextApiResponse } from 'next';
 import { connectToDatabase } from '@/service/mongo';
-import { getOpenAIApi, authChat } from '@/service/utils/auth';
-import { axiosConfig, openaiChatFilter } from '@/service/utils/tools';
+import { authChat } from '@/service/utils/auth';
+import { modelServiceToolMap } from '@/service/utils/chat';
 import { ChatItemSimpleType } from '@/types/chat';
 import { jsonRes } from '@/service/response';
 import { PassThrough } from 'stream';
 import { ChatModelMap, ModelVectorSearchModeMap } from '@/constants/model';
 import { pushChatBill } from '@/service/events/pushBill';
-import { gpt35StreamResponse } from '@/service/utils/openai';
-import { searchKb_openai } from '@/service/tools/searchKb';
+import { resStreamResponse } from '@/service/utils/chat';
+import { searchKb } from '@/service/plugins/searchKb';
+import { ChatRoleEnum } from '@/constants/chat';

 /* 发送提示词 */
 export default async function handler(req: NextApiRequest, res: NextApiResponse) {
@@ -41,7 +42,7 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)
    await connectToDatabase();
    let startTime = Date.now();

-    const { model, showModelDetail, content, userApiKey, systemKey, userId } = await authChat({
+    const { model, showModelDetail, content, userApiKey, systemApiKey, userId } = await authChat({
      modelId,
      chatId,
      authorization
@@ -54,9 +55,9 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)

    // 使用了知识库搜索
    if (model.chat.useKb) {
-      const { code, searchPrompt } = await searchKb_openai({
-        apiKey: userApiKey || systemKey,
-        isPay: !userApiKey,
+      const { code, searchPrompt } = await searchKb({
+        userApiKey,
+        systemApiKey,
        text: prompt.value,
        similarity: ModelVectorSearchModeMap[model.chat.searchMode]?.similarity,
        model,
@@ -73,53 +74,37 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)
      // 没有用知识库搜索，仅用系统提示词
      model.chat.systemPrompt &&
        prompts.unshift({
-          obj: 'SYSTEM',
+          obj: ChatRoleEnum.System,
          value: model.chat.systemPrompt
        });
    }

-    // 控制总 tokens 数量，防止超出
-    const filterPrompts = openaiChatFilter({
-      model: model.chat.chatModel,
-      prompts,
-      maxTokens: modelConstantsData.contextMaxToken - 300
-    });
-
    // 计算温度
    const temperature = (modelConstantsData.maxTemperature * (model.chat.temperature / 10)).toFixed(
      2
    );
    // console.log(filterPrompts);
-    // 获取 chatAPI
-    const chatAPI = getOpenAIApi(userApiKey || systemKey);
+
    // 发出请求
-    const chatResponse = await chatAPI.createChatCompletion(
-      {
-        model: model.chat.chatModel,
-        temperature: Number(temperature) || 0,
-        messages: filterPrompts,
-        frequency_penalty: 0.5, // 越大，重复内容越少
-        presence_penalty: -0.5, // 越大，越容易出现新内容
-        stream: true,
-        stop: ['.!?。']
-      },
-      {
-        timeout: 40000,
-        responseType: 'stream',
-        ...axiosConfig()
-      }
-    );
+    const { streamResponse } = await modelServiceToolMap[model.chat.chatModel].chatCompletion({
+      apiKey: userApiKey || systemApiKey,
+      temperature: +temperature,
+      messages: prompts,
+      stream: true
+    });

    console.log('api response time:', `${(Date.now() - startTime) / 1000}s`);

    step = 1;

-    const { responseContent } = await gpt35StreamResponse({
+    const { totalTokens, finishMessages } = await resStreamResponse({
+      model: model.chat.chatModel,
      res,
      stream,
-      chatResponse,
+      chatResponse: streamResponse,
+      prompts,
      systemPrompt:
-        showModelDetail && filterPrompts[0].role === 'system' ? filterPrompts[0].content : ''
+        showModelDetail && prompts[0].obj === ChatRoleEnum.System ? prompts[0].value : ''
    });

    // 只有使用平台的 key 才计费
@@ -128,7 +113,8 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)
      chatModel: model.chat.chatModel,
      userId,
      chatId,
-      messages: filterPrompts.concat({ role: 'assistant', content: responseContent })
+      textLen: finishMessages.map((item) => item.value).join('').length,
+      tokens: totalTokens
    });
  } catch (err: any) {
    if (step === 1) {
--- a/src/pages/api/openapi/chat/chat.ts
+++ b/src/pages/api/openapi/chat/chat.ts
@@ -1,14 +1,14 @@
 import type { NextApiRequest, NextApiResponse } from 'next';
 import { connectToDatabase } from '@/service/mongo';
-import { getOpenAIApi, authOpenApiKey, authModel } from '@/service/utils/auth';
-import { axiosConfig, openaiChatFilter } from '@/service/utils/tools';
+import { authOpenApiKey, authModel } from '@/service/utils/auth';
+import { modelServiceToolMap, resStreamResponse } from '@/service/utils/chat';
 import { ChatItemSimpleType } from '@/types/chat';
 import { jsonRes } from '@/service/response';
 import { PassThrough } from 'stream';
 import { ChatModelMap, ModelVectorSearchModeMap } from '@/constants/model';
 import { pushChatBill } from '@/service/events/pushBill';
-import { gpt35StreamResponse } from '@/service/utils/openai';
-import { searchKb_openai } from '@/service/tools/searchKb';
+import { searchKb } from '@/service/plugins/searchKb';
+import { ChatRoleEnum } from '@/constants/chat';

 /* 发送提示词 */
 export default async function handler(req: NextApiRequest, res: NextApiResponse) {
@@ -64,9 +64,8 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)
    if (model.chat.useKb) {
      const similarity = ModelVectorSearchModeMap[model.chat.searchMode]?.similarity || 0.22;

-      const { code, searchPrompt } = await searchKb_openai({
-        apiKey,
-        isPay: true,
+      const { code, searchPrompt } = await searchKb({
+        systemApiKey: apiKey,
        text: prompts[prompts.length - 1].value,
        similarity,
        model,
@@ -83,69 +82,55 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)
      // 没有用知识库搜索，仅用系统提示词
      if (model.chat.systemPrompt) {
        prompts.unshift({
-          obj: 'SYSTEM',
+          obj: ChatRoleEnum.System,
          value: model.chat.systemPrompt
        });
      }
    }

-    // 控制总 tokens 数量，防止超出
-    const filterPrompts = openaiChatFilter({
-      model: model.chat.chatModel,
-      prompts,
-      maxTokens: modelConstantsData.contextMaxToken - 300
-    });
-
    // 计算温度
    const temperature = (modelConstantsData.maxTemperature * (model.chat.temperature / 10)).toFixed(
      2
    );
-    // console.log(filterPrompts);
-    // 获取 chatAPI
-    const chatAPI = getOpenAIApi(apiKey);
+
    // 发出请求
-    const chatResponse = await chatAPI.createChatCompletion(
-      {
-        model: model.chat.chatModel,
-        temperature: Number(temperature) || 0,
-        messages: filterPrompts,
-        frequency_penalty: 0.5, // 越大，重复内容越少
-        presence_penalty: -0.5, // 越大，越容易出现新内容
-        stream: isStream,
-        stop: ['.!?。']
-      },
-      {
-        timeout: 180000,
-        responseType: isStream ? 'stream' : 'json',
-        ...axiosConfig()
-      }
-    );
+    const { streamResponse, responseMessages, responseText, totalTokens } =
+      await modelServiceToolMap[model.chat.chatModel].chatCompletion({
+        apiKey,
+        temperature: +temperature,
+        messages: prompts,
+        stream: isStream
+      });

    console.log('api response time:', `${(Date.now() - startTime) / 1000}s`);

-    let responseContent = '';
+    let textLen = 0;
+    let tokens = totalTokens;

    if (isStream) {
      step = 1;
-      const streamResponse = await gpt35StreamResponse({
+      const { finishMessages, totalTokens } = await resStreamResponse({
+        model: model.chat.chatModel,
        res,
        stream,
-        chatResponse
+        chatResponse: streamResponse,
+        prompts
      });
-      responseContent = streamResponse.responseContent;
+      textLen = finishMessages.map((item) => item.value).join('').length;
+      tokens = totalTokens;
    } else {
-      responseContent = chatResponse.data.choices?.[0]?.message?.content || '';
+      textLen = responseMessages.map((item) => item.value).join('').length;
      jsonRes(res, {
-        data: responseContent
+        data: responseText
      });
    }

-    // 只有使用平台的 key 才计费
    pushChatBill({
      isPay: true,
      chatModel: model.chat.chatModel,
      userId,
-      messages: filterPrompts.concat({ role: 'assistant', content: responseContent })
+      textLen,
+      tokens
    });
  } catch (err: any) {
    if (step === 1) {
--- a/src/pages/api/openapi/chat/chatGpt.ts
+++ b/src/pages/api/openapi/chat/chatGpt.ts
@@ -1,144 +0,0 @@
-import type { NextApiRequest, NextApiResponse } from 'next';
-import { connectToDatabase, Model } from '@/service/mongo';
-import { getOpenAIApi, authOpenApiKey } from '@/service/utils/auth';
-import { axiosConfig, openaiChatFilter } from '@/service/utils/tools';
-import { ChatItemSimpleType } from '@/types/chat';
-import { jsonRes } from '@/service/response';
-import { PassThrough } from 'stream';
-import { ChatModelMap } from '@/constants/model';
-import { pushChatBill } from '@/service/events/pushBill';
-import { gpt35StreamResponse } from '@/service/utils/openai';
-
-/* 发送提示词 */
-export default async function handler(req: NextApiRequest, res: NextApiResponse) {
-  let step = 0; // step=1时，表示开始了流响应
-  const stream = new PassThrough();
-  stream.on('error', () => {
-    console.log('error: ', 'stream error');
-    stream.destroy();
-  });
-  res.on('close', () => {
-    stream.destroy();
-  });
-  res.on('error', () => {
-    console.log('error: ', 'request error');
-    stream.destroy();
-  });
-
-  try {
-    const {
-      prompts,
-      modelId,
-      isStream = true
-    } = req.body as {
-      prompts: ChatItemSimpleType[];
-      modelId: string;
-      isStream: boolean;
-    };
-
-    if (!prompts || !modelId) {
-      throw new Error('缺少参数');
-    }
-    if (!Array.isArray(prompts)) {
-      throw new Error('prompts is not array');
-    }
-    if (prompts.length > 30 || prompts.length === 0) {
-      throw new Error('prompts length range 1-30');
-    }
-
-    await connectToDatabase();
-    let startTime = Date.now();
-
-    const { apiKey, userId } = await authOpenApiKey(req);
-
-    const model = await Model.findOne({
-      _id: modelId,
-      userId
-    });
-
-    if (!model) {
-      throw new Error('无权使用该模型');
-    }
-
-    const modelConstantsData = ChatModelMap[model.chat.chatModel];
-
-    // 如果有系统提示词，自动插入
-    if (model.chat.systemPrompt) {
-      prompts.unshift({
-        obj: 'SYSTEM',
-        value: model.chat.systemPrompt
-      });
-    }
-
-    // 控制在 tokens 数量，防止超出
-    const filterPrompts = openaiChatFilter({
-      model: model.chat.chatModel,
-      prompts,
-      maxTokens: modelConstantsData.contextMaxToken - 300
-    });
-
-    // console.log(filterPrompts);
-    // 计算温度
-    const temperature = (modelConstantsData.maxTemperature * (model.chat.temperature / 10)).toFixed(
-      2
-    );
-    // 获取 chatAPI
-    const chatAPI = getOpenAIApi(apiKey);
-    // 发出请求
-    const chatResponse = await chatAPI.createChatCompletion(
-      {
-        model: model.chat.chatModel,
-        temperature: Number(temperature) || 0,
-        messages: filterPrompts,
-        frequency_penalty: 0.5, // 越大，重复内容越少
-        presence_penalty: -0.5, // 越大，越容易出现新内容
-        stream: isStream,
-        stop: ['.!?。']
-      },
-      {
-        timeout: 40000,
-        responseType: isStream ? 'stream' : 'json',
-        ...axiosConfig()
-      }
-    );
-
-    console.log('api response time:', `${(Date.now() - startTime) / 1000}s`);
-
-    let responseContent = '';
-
-    if (isStream) {
-      step = 1;
-      const streamResponse = await gpt35StreamResponse({
-        res,
-        stream,
-        chatResponse
-      });
-      responseContent = streamResponse.responseContent;
-    } else {
-      responseContent = chatResponse.data.choices?.[0]?.message?.content || '';
-      jsonRes(res, {
-        data: responseContent
-      });
-    }
-
-    // 只有使用平台的 key 才计费
-    pushChatBill({
-      isPay: true,
-      chatModel: model.chat.chatModel,
-      userId,
-      messages: filterPrompts.concat({ role: 'assistant', content: responseContent })
-    });
-  } catch (err: any) {
-    if (step === 1) {
-      // 直接结束流
-      console.log('error，结束');
-      stream.destroy();
-    } else {
-      res.status(500);
-      jsonRes(res, {
-        code: 500,
-        error: err
-      });
-    }
-  }
-}
--- a/src/pages/api/openapi/chat/lafGpt.ts
+++ b/src/pages/api/openapi/chat/lafGpt.ts
@@ -1,14 +1,14 @@
 import type { NextApiRequest, NextApiResponse } from 'next';
 import { connectToDatabase, Model } from '@/service/mongo';
-import { getOpenAIApi, authOpenApiKey } from '@/service/utils/auth';
-import { axiosConfig, openaiChatFilter } from '@/service/utils/tools';
+import { authOpenApiKey } from '@/service/utils/auth';
+import { resStreamResponse, modelServiceToolMap } from '@/service/utils/chat';
 import { ChatItemSimpleType } from '@/types/chat';
 import { jsonRes } from '@/service/response';
 import { PassThrough } from 'stream';
-import { ChatModelMap, ModelVectorSearchModeMap, OpenAiChatEnum } from '@/constants/model';
+import { ChatModelMap, ModelVectorSearchModeMap } from '@/constants/model';
 import { pushChatBill } from '@/service/events/pushBill';
-import { gpt35StreamResponse } from '@/service/utils/openai';
-import { searchKb_openai } from '@/service/tools/searchKb';
+import { searchKb } from '@/service/plugins/searchKb';
+import { ChatRoleEnum } from '@/constants/chat';

 /* 发送提示词 */
 export default async function handler(req: NextApiRequest, res: NextApiResponse) {
@@ -57,20 +57,16 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)

    console.log('laf gpt start');

-    // 获取 chatAPI
-    const chatAPI = getOpenAIApi(apiKey);
-
    // 请求一次 chatgpt 拆解需求
-    const promptResponse = await chatAPI.createChatCompletion(
-      {
-        model: OpenAiChatEnum.GPT35,
-        temperature: 0,
-        frequency_penalty: 0.5, // 越大，重复内容越少
-        presence_penalty: -0.5, // 越大，越容易出现新内容
-        messages: [
-          {
-            role: 'system',
-            content: `服务端逻辑生成器.根据用户输入的需求,拆解成 laf 云函数实现的步骤,只返回步骤,按格式返回步骤: 1.\n2.\n3.\n ......
+    const { responseText: resolveText, totalTokens: resolveTokens } = await modelServiceToolMap[
+      model.chat.chatModel
+    ].chatCompletion({
+      apiKey,
+      temperature: 0,
+      messages: [
+        {
+          obj: ChatRoleEnum.System,
+          value: `服务端逻辑生成器.根据用户输入的需求,拆解成 laf 云函数实现的步骤,只返回步骤,按格式返回步骤: 1.\n2.\n3.\n ......
 下面是一些例子:
 一个 hello world 例子
 1. 返回字符串: "hello world"
@@ -103,35 +99,25 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)
 5. 获取当前时间,记录为 updateTime.
 6. 更新数据库数据,表为"blogs",更新符合 blogId 的记录的内容为{blogText, tags, updateTime}.
 7. 返回结果 "更新博客记录成功"`
-          },
-          {
-            role: 'user',
-            content: prompt.value
-          }
-        ]
-      },
-      {
-        timeout: 180000,
-        ...axiosConfig()
-      }
-    );
+        },
+        {
+          obj: ChatRoleEnum.Human,
+          value: prompt.value
+        }
+      ],
+      stream: false
+    });

-    const promptResolve = promptResponse.data.choices?.[0]?.message?.content || '';
-    if (!promptResolve) {
-      throw new Error('gpt 异常');
-    }
-
-    prompt.value += ` ${promptResolve}`;
+    prompt.value += ` ${resolveText}`;
    console.log('prompt resolve success, time:', `${(Date.now() - startTime) / 1000}s`);

    // 读取对话内容
    const prompts = [prompt];

    // 获取向量匹配到的提示词
-    const { searchPrompt } = await searchKb_openai({
-      isPay: true,
-      apiKey,
-      similarity: ModelVectorSearchModeMap[model.chat.searchMode]?.similarity || 0.22,
+    const { searchPrompt } = await searchKb({
+      systemApiKey: apiKey,
+      similarity: ModelVectorSearchModeMap[model.chat.searchMode]?.similarity,
      text: prompt.value,
      model,
      userId
@@ -139,49 +125,41 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)

    searchPrompt && prompts.unshift(searchPrompt);

-    // 控制上下文 tokens 数量，防止超出
-    const filterPrompts = openaiChatFilter({
-      model: model.chat.chatModel,
-      prompts,
-      maxTokens: modelConstantsData.contextMaxToken - 300
-    });
-
-    // console.log(filterPrompts);
    // 计算温度
    const temperature = (modelConstantsData.maxTemperature * (model.chat.temperature / 10)).toFixed(
      2
    );
-    // 发出请求
-    const chatResponse = await chatAPI.createChatCompletion(
-      {
-        model: model.chat.chatModel,
-        temperature: Number(temperature) || 0,
-        messages: filterPrompts,
-        frequency_penalty: 0.5, // 越大，重复内容越少
-        presence_penalty: -0.5, // 越大，越容易出现新内容
-        stream: isStream
-      },
-      {
-        timeout: 180000,
-        responseType: isStream ? 'stream' : 'json',
-        ...axiosConfig()
-      }
-    );

-    let responseContent = '';
+    // 发出请求
+    const { streamResponse, responseMessages, responseText, totalTokens } =
+      await modelServiceToolMap[model.chat.chatModel].chatCompletion({
+        apiKey,
+        temperature: +temperature,
+        messages: prompts,
+        stream: isStream
+      });
+
+    console.log('api response time:', `${(Date.now() - startTime) / 1000}s`);
+
+    let textLen = resolveText.length;
+    let tokens = resolveTokens;

    if (isStream) {
      step = 1;
-      const streamResponse = await gpt35StreamResponse({
+      const { finishMessages, totalTokens } = await resStreamResponse({
+        model: model.chat.chatModel,
        res,
        stream,
-        chatResponse
+        chatResponse: streamResponse,
+        prompts
      });
-      responseContent = streamResponse.responseContent;
+      textLen += finishMessages.map((item) => item.value).join('').length;
+      tokens += totalTokens;
    } else {
-      responseContent = chatResponse.data.choices?.[0]?.message?.content || '';
+      textLen += responseMessages.map((item) => item.value).join('').length;
+      tokens += totalTokens;
      jsonRes(res, {
-        data: responseContent
+        data: responseText
      });
    }

@@ -191,7 +169,8 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)
      isPay: true,
      chatModel: model.chat.chatModel,
      userId,
-      messages: filterPrompts.concat({ role: 'assistant', content: responseContent })
+      textLen,
+      tokens
    });
  } catch (err: any) {
    if (step === 1) {
--- a/src/pages/api/openapi/chat/vectorGpt.ts
+++ b/src/pages/api/openapi/chat/vectorGpt.ts
@@ -1,159 +0,0 @@
-import type { NextApiRequest, NextApiResponse } from 'next';
-import { connectToDatabase, Model } from '@/service/mongo';
-import { axiosConfig, openaiChatFilter } from '@/service/utils/tools';
-import { getOpenAIApi, authOpenApiKey } from '@/service/utils/auth';
-import { ChatItemSimpleType } from '@/types/chat';
-import { jsonRes } from '@/service/response';
-import { PassThrough } from 'stream';
-import {
-  ChatModelMap,
-  ModelVectorSearchModeMap,
-  ModelVectorSearchModeEnum
-} from '@/constants/model';
-import { pushChatBill } from '@/service/events/pushBill';
-import { gpt35StreamResponse } from '@/service/utils/openai';
-import { searchKb_openai } from '@/service/tools/searchKb';
-
-/* 发送提示词 */
-export default async function handler(req: NextApiRequest, res: NextApiResponse) {
-  let step = 0; // step=1时，表示开始了流响应
-  const stream = new PassThrough();
-  stream.on('error', () => {
-    console.log('error: ', 'stream error');
-    stream.destroy();
-  });
-  res.on('close', () => {
-    stream.destroy();
-  });
-  res.on('error', () => {
-    console.log('error: ', 'request error');
-    stream.destroy();
-  });
-
-  try {
-    const {
-      prompts,
-      modelId,
-      isStream = true
-    } = req.body as {
-      prompts: ChatItemSimpleType[];
-      modelId: string;
-      isStream: boolean;
-    };
-
-    if (!prompts || !modelId) {
-      throw new Error('缺少参数');
-    }
-    if (!Array.isArray(prompts)) {
-      throw new Error('prompts is not array');
-    }
-    if (prompts.length > 30 || prompts.length === 0) {
-      throw new Error('prompts length range 1-30');
-    }
-
-    await connectToDatabase();
-    let startTime = Date.now();
-
-    /* 凭证校验 */
-    const { apiKey, userId } = await authOpenApiKey(req);
-
-    const model = await Model.findOne({
-      _id: modelId,
-      userId
-    });
-
-    if (!model) {
-      throw new Error('无权使用该模型');
-    }
-
-    const modelConstantsData = ChatModelMap[model.chat.chatModel];
-
-    // 获取向量匹配到的提示词
-    const { code, searchPrompt } = await searchKb_openai({
-      isPay: true,
-      apiKey,
-      similarity: ModelVectorSearchModeMap[model.chat.searchMode]?.similarity || 0.22,
-      text: prompts[prompts.length - 1].value,
-      model,
-      userId
-    });
-
-    // search result is empty
-    if (code === 201) {
-      return res.send(searchPrompt?.value);
-    }
-
-    searchPrompt && prompts.unshift(searchPrompt);
-
-    // 控制在 tokens 数量，防止超出
-    const filterPrompts = openaiChatFilter({
-      model: model.chat.chatModel,
-      prompts,
-      maxTokens: modelConstantsData.contextMaxToken - 300
-    });
-
-    // console.log(filterPrompts);
-    // 计算温度
-    const temperature = (modelConstantsData.maxTemperature * (model.chat.temperature / 10)).toFixed(
-      2
-    );
-    const chatAPI = getOpenAIApi(apiKey);
-
-    // 发出请求
-    const chatResponse = await chatAPI.createChatCompletion(
-      {
-        model: model.chat.chatModel,
-        temperature: Number(temperature) || 0,
-        messages: filterPrompts,
-        frequency_penalty: 0.5, // 越大，重复内容越少
-        presence_penalty: -0.5, // 越大，越容易出现新内容
-        stream: isStream,
-        stop: ['.!?。']
-      },
-      {
-        timeout: 180000,
-        responseType: isStream ? 'stream' : 'json',
-        ...axiosConfig()
-      }
-    );
-
-    console.log('api response time:', `${(Date.now() - startTime) / 1000}s`);
-
-    let responseContent = '';
-
-    if (isStream) {
-      step = 1;
-      const streamResponse = await gpt35StreamResponse({
-        res,
-        stream,
-        chatResponse
-      });
-      responseContent = streamResponse.responseContent;
-    } else {
-      responseContent = chatResponse.data.choices?.[0]?.message?.content || '';
-      jsonRes(res, {
-        data: responseContent
-      });
-    }
-
-    pushChatBill({
-      isPay: true,
-      chatModel: model.chat.chatModel,
-      userId,
-      messages: filterPrompts.concat({ role: 'assistant', content: responseContent })
-    });
-    // jsonRes(res);
-  } catch (err: any) {
-    if (step === 1) {
-      // 直接结束流
-      console.log('error，结束');
-      stream.destroy();
-    } else {
-      res.status(500);
-      jsonRes(res, {
-        code: 500,
-        error: err
-      });
-    }
-  }
-}