feat: chat content use tiktoken count

src/service/events/generateAbstract.ts

@@ -4,7 +4,7 @@ import { httpsAgent } from '@/service/utils/tools';
 import { getOpenApiKey } from '../utils/openai';
 import type { ChatCompletionRequestMessage } from 'openai';
 import { DataItemSchema } from '@/types/mongoSchema';
-import { ChatModelNameEnum } from '@/constants/model';
+import { ChatModelEnum } from '@/constants/model';
 import { pushSplitDataBill } from '@/service/events/pushBill';
 
 export async function generateAbstract(next = false): Promise<any> {
@@ -68,7 +68,7 @@ export async function generateAbstract(next = false): Promise<any> {
     // 请求 chatgpt 获取摘要
     const abstractResponse = await chatAPI.createChatCompletion(
       {
-        model: ChatModelNameEnum.GPT35,
+        model: ChatModelEnum.GPT35,
         temperature: 0.8,
         n: 1,
         messages: [

src/service/events/generateQA.ts

@@ -3,7 +3,7 @@ import { getOpenAIApi } from '@/service/utils/auth';
 import { httpsAgent } from '@/service/utils/tools';
 import { getOpenApiKey } from '../utils/openai';
 import type { ChatCompletionRequestMessage } from 'openai';
-import { ChatModelNameEnum } from '@/constants/model';
+import { ChatModelEnum } from '@/constants/model';
 import { pushSplitDataBill } from '@/service/events/pushBill';
 import { generateVector } from './generateVector';
 import { openaiError2 } from '../errorCode';
@@ -84,7 +84,7 @@ A2:
         chatAPI
           .createChatCompletion(
             {
-              model: ChatModelNameEnum.GPT35,
+              model: ChatModelEnum.GPT35,
               temperature: 0.8,
               n: 1,
               messages: [

src/service/events/pushBill.ts

@@ -1,27 +1,34 @@
 import { connectToDatabase, Bill, User } from '../mongo';
-import { modelList, ChatModelNameEnum } from '@/constants/model';
-import { encode } from 'gpt-token-utils';
+import {
+  modelList,
+  ChatModelEnum,
+  ModelNameEnum,
+  Model2ChatModelMap,
+  embeddingModel
+} from '@/constants/model';
 import { BillTypeEnum } from '@/constants/user';
 import type { DataType } from '@/types/data';
+import { countChatTokens } from '@/utils/tools';
 
 export const pushChatBill = async ({
   isPay,
   modelName,
   userId,
   chatId,
-  text
+  messages
 }: {
   isPay: boolean;
-  modelName: string;
+  modelName: `${ModelNameEnum}`;
   userId: string;
   chatId?: '' | string;
-  text: string;
+  messages: { role: 'system' | 'user' | 'assistant'; content: string }[];
 }) => {
-  let billId;
+  let billId = '';
 
   try {
     // 计算 token 数量
-    const tokens = Math.floor(encode(text).length * 0.75);
+    const tokens = countChatTokens({ model: Model2ChatModelMap[modelName] as any, messages });
+    const text = messages.map((item) => item.content).join('');
 
     console.log(
       `chat generate success. text len: ${text.length}. token len: ${tokens}. pay:${isPay}`
@@ -88,7 +95,7 @@ export const pushSplitDataBill = async ({
     if (isPay) {
       try {
         // 获取模型单价格, 都是用 gpt35 拆分
-        const modelItem = modelList.find((item) => item.model === ChatModelNameEnum.GPT35);
+        const modelItem = modelList.find((item) => item.model === ChatModelEnum.GPT35);
         const unitPrice = modelItem?.price || 3;
         // 计算价格
         const price = unitPrice * tokenLen;
@@ -97,7 +104,7 @@ export const pushSplitDataBill = async ({
         const res = await Bill.create({
           userId,
           type,
-          modelName: ChatModelNameEnum.GPT35,
+          modelName: ChatModelEnum.GPT35,
           textLen: text.length,
           tokenLen,
           price
@@ -149,7 +156,7 @@ export const pushGenerateVectorBill = async ({
         const res = await Bill.create({
           userId,
           type: BillTypeEnum.vector,
-          modelName: ChatModelNameEnum.VECTOR,
+          modelName: embeddingModel,
           textLen: text.length,
           tokenLen,
           price