fix: gpt35 4k

2025-07-23 21:13:50 +00:00 · 2023-06-14 20:54:34 +08:00
parent e4aeee7be3
commit 07f8e18c10
16 changed files with 74 additions and 22 deletions
--- a/client/public/docs/chatProblem.md
+++ b/client/public/docs/chatProblem.md
@@ -15,7 +15,8 @@
 | 计费项 | 价格: 元/ 1K tokens（包含上下文）|
 | --- | --- |
 | 知识库 - 索引 | 0.001 |
-| chatgpt - 对话 | 0.025 |
+| chatgpt - 对话 | 0.022 |
+| chatgpt16K - 对话 | 0.025 |
 | gpt4 - 对话 | 0.5 |
 | 文件拆分 | 0.025 |

--- a/client/public/docs/intro.md
+++ b/client/public/docs/intro.md
@@ -19,7 +19,8 @@ FastGpt 项目完全开源，可随意私有化部署，去除平台风险忧虑
 | 计费项 | 价格: 元/ 1K tokens（包含上下文）|
 | --- | --- |
 | 知识库 - 索引 | 0.001 |
-| chatgpt - 对话 | 0.025 |
+| chatgpt - 对话 | 0.022 |
+| chatgpt16K - 对话 | 0.025 |
 | gpt4 - 对话 | 0.5 |
 | 文件拆分 | 0.025 |

--- a/client/src/constants/model.ts
+++ b/client/src/constants/model.ts
@@ -7,7 +7,8 @@ export const embeddingPrice = 0.1;
 export type EmbeddingModelType = 'text-embedding-ada-002';

 export enum OpenAiChatEnum {
-  'GPT35' = 'gpt-3.5-turbo-16k',
+  'GPT35' = 'gpt-3.5-turbo',
+  'GPT3516k' = 'gpt-3.5-turbo-16k',
  'GPT4' = 'gpt-4',
  'GPT432k' = 'gpt-4-32k'
 }
@@ -29,7 +30,15 @@ export type ChatModelItemType = {
 export const ChatModelMap = {
  [OpenAiChatEnum.GPT35]: {
    chatModel: OpenAiChatEnum.GPT35,
-    name: 'ChatGpt',
+    name: 'Gpt35-4k',
+    contextMaxToken: 4096,
+    systemMaxToken: 2400,
+    maxTemperature: 1.2,
+    price: 2.2
+  },
+  [OpenAiChatEnum.GPT3516k]: {
+    chatModel: OpenAiChatEnum.GPT3516k,
+    name: 'Gpt35-16k',
    contextMaxToken: 16000,
    systemMaxToken: 8000,
    maxTemperature: 1.2,
--- a/client/src/pages/_app.tsx
+++ b/client/src/pages/_app.tsx
@@ -49,14 +49,20 @@ export default function App({ Component, pageProps }: AppProps) {
        />
        <link rel="icon" href="/favicon.ico" />
      </Head>
-      <Script src="/js/qrcode.min.js" strategy="lazyOnload"></Script>
-      <Script src="/js/pdf.js" strategy="lazyOnload"></Script>
-      <Script src="/js/html2pdf.bundle.min.js" strategy="lazyOnload"></Script>
+      <Script src="/js/qrcode.min.js" strategy="afterInteractive"></Script>
+      <Script src="/js/pdf.js" strategy="afterInteractive"></Script>
+      <Script src="/js/html2pdf.bundle.min.js" strategy="afterInteractive"></Script>
      {googleVerKey && (
-        <Script
-          src={`https://www.recaptcha.net/recaptcha/api.js?render=${googleVerKey}`}
-          strategy="lazyOnload"
-        ></Script>
+        <>
+          <Script
+            src={`https://www.recaptcha.net/recaptcha/api.js?render=${googleVerKey}`}
+            strategy="afterInteractive"
+          ></Script>
+          <Script
+            src={`https://www.google.com/recaptcha/api.js?render=${googleVerKey}`}
+            strategy="afterInteractive"
+          ></Script>
+        </>
      )}
      <Script src="/js/particles.js"></Script>
      <QueryClientProvider client={queryClient}>
--- a/client/src/pages/api/openapi/kb/pushData.ts
+++ b/client/src/pages/api/openapi/kb/pushData.ts
@@ -78,7 +78,7 @@ export async function pushDataToKb({

    if (mode === TrainingModeEnum.qa) {
      // count token
-      const token = modelToolMap[OpenAiChatEnum.GPT35].countTokens({
+      const token = modelToolMap[OpenAiChatEnum.GPT3516k].countTokens({
        messages: [{ obj: 'System', value: item.q }]
      });
      if (token > modeMaxToken[TrainingModeEnum.qa]) {
--- a/client/src/pages/api/system/getModels.ts
+++ b/client/src/pages/api/system/getModels.ts
@@ -8,6 +8,7 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)
  const chatModelList: ChatModelItemType[] = [];

  if (global.systemEnv.openAIKeys) {
+    chatModelList.push(ChatModelMap[OpenAiChatEnum.GPT3516k]);
    chatModelList.push(ChatModelMap[OpenAiChatEnum.GPT35]);
  }
  if (global.systemEnv.gpt4Key) {
--- a/client/src/pages/kb/components/SelectFileModal.tsx
+++ b/client/src/pages/kb/components/SelectFileModal.tsx
@@ -31,12 +31,12 @@ const modeMap = {
  [TrainingModeEnum.qa]: {
    maxLen: 8000,
    slideLen: 3000,
-    price: ChatModelMap[OpenAiChatEnum.GPT35].price,
+    price: ChatModelMap[OpenAiChatEnum.GPT3516k].price,
    isPrompt: true
  },
  [TrainingModeEnum.index]: {
-    maxLen: 1400,
-    slideLen: 700,
+    maxLen: 1000,
+    slideLen: 500,
    price: embeddingPrice,
    isPrompt: false
  }
--- a/client/src/pages/number/components/PayModal.tsx
+++ b/client/src/pages/number/components/PayModal.tsx
@@ -114,7 +114,8 @@ const PayModal = ({ onClose }: { onClose: () => void }) => {
 | 计费项 | 价格: 元/ 1K tokens(包含上下文)|
 | --- | --- |
 | 知识库 - 索引 | 0.001 |
-| chatgpt - 对话 | 0.025 |
+| chatgpt - 对话 | 0.022 |
+| chatgpt16K - 对话 | 0.025 |
 | gpt4 - 对话 | 0.5 |
 | 文件拆分 | 0.025 |`}
                />
--- a/client/src/service/events/generateQA.ts
+++ b/client/src/service/events/generateQA.ts
@@ -87,7 +87,7 @@ export async function generateQA(): Promise<any> {
    // 请求 chatgpt 获取回答
    const response = await Promise.all(
      [data.q].map((text) =>
-        modelServiceToolMap[OpenAiChatEnum.GPT35]
+        modelServiceToolMap[OpenAiChatEnum.GPT3516k]
          .chatCompletion({
            apiKey: systemAuthKey,
            temperature: 0.8,
--- a/client/src/service/events/pushBill.ts
+++ b/client/src/service/events/pushBill.ts
@@ -104,7 +104,7 @@ export const pushSplitDataBill = async ({
    await connectToDatabase();

    // 获取模型单价格, 都是用 gpt35 拆分
-    const unitPrice = ChatModelMap[OpenAiChatEnum.GPT35].price || 3;
+    const unitPrice = ChatModelMap[OpenAiChatEnum.GPT3516k].price || 3;
    // 计算价格
    const price = unitPrice * totalTokens;

@@ -112,7 +112,7 @@ export const pushSplitDataBill = async ({
    const res = await Bill.create({
      userId,
      type,
-      modelName: OpenAiChatEnum.GPT35,
+      modelName: OpenAiChatEnum.GPT3516k,
      textLen,
      tokenLen: totalTokens,
      price
--- a/client/src/service/models/model.ts
+++ b/client/src/service/models/model.ts
@@ -57,7 +57,7 @@ const ModelSchema = new Schema({
      // 聊天时使用的模型
      type: String,
      enum: Object.keys(ChatModelMap),
-      default: OpenAiChatEnum.GPT35
+      default: OpenAiChatEnum.GPT3516k
    }
  },
  share: {
--- a/client/src/service/utils/auth.ts
+++ b/client/src/service/utils/auth.ts
@@ -162,6 +162,10 @@ export const getApiKey = async ({
      userOpenAiKey: user.openaiKey || '',
      systemAuthKey: getSystemOpenAiKey(type) as string
    },
+    [OpenAiChatEnum.GPT3516k]: {
+      userOpenAiKey: user.openaiKey || '',
+      systemAuthKey: getSystemOpenAiKey(type) as string
+    },
    [OpenAiChatEnum.GPT4]: {
      userOpenAiKey: user.openaiKey || '',
      systemAuthKey: getGpt4Key() as string
--- a/client/src/service/utils/chat/index.ts
+++ b/client/src/service/utils/chat/index.ts
@@ -48,6 +48,15 @@ export const modelServiceToolMap: Record<
        ...data
      })
  },
+  [OpenAiChatEnum.GPT3516k]: {
+    chatCompletion: (data: ChatCompletionType) =>
+      chatResponse({ model: OpenAiChatEnum.GPT3516k, ...data }),
+    streamResponse: (data: StreamResponseType) =>
+      openAiStreamResponse({
+        model: OpenAiChatEnum.GPT3516k,
+        ...data
+      })
+  },
  [OpenAiChatEnum.GPT4]: {
    chatCompletion: (data: ChatCompletionType) =>
      chatResponse({ model: OpenAiChatEnum.GPT4, ...data }),
--- a/client/src/store/chat.ts
+++ b/client/src/store/chat.ts
@@ -53,7 +53,7 @@ const defaultChatData = {
    intro: '',
    canUse: false
  },
-  chatModel: OpenAiChatEnum.GPT35,
+  chatModel: OpenAiChatEnum.GPT3516k,
  history: []
 };
 const defaultShareChatData: ShareChatType = {
@@ -64,7 +64,7 @@ const defaultShareChatData: ShareChatType = {
    avatar: '/icon/logo.png',
    intro: ''
  },
-  chatModel: OpenAiChatEnum.GPT35,
+  chatModel: OpenAiChatEnum.GPT3516k,
  history: []
 };

--- a/client/src/utils/plugin/index.ts
+++ b/client/src/utils/plugin/index.ts
@@ -20,6 +20,11 @@ export const modelToolMap: Record<
    sliceText: (data) => openAiSliceTextByToken({ model: OpenAiChatEnum.GPT35, ...data }),
    tokenSlice: (data) => gpt_chatItemTokenSlice({ model: OpenAiChatEnum.GPT35, ...data })
  },
+  [OpenAiChatEnum.GPT3516k]: {
+    countTokens: ({ messages }) => countOpenAIToken({ model: OpenAiChatEnum.GPT3516k, messages }),
+    sliceText: (data) => openAiSliceTextByToken({ model: OpenAiChatEnum.GPT3516k, ...data }),
+    tokenSlice: (data) => gpt_chatItemTokenSlice({ model: OpenAiChatEnum.GPT3516k, ...data })
+  },
  [OpenAiChatEnum.GPT4]: {
    countTokens: ({ messages }) => countOpenAIToken({ model: OpenAiChatEnum.GPT4, messages }),
    sliceText: (data) => openAiSliceTextByToken({ model: OpenAiChatEnum.GPT4, ...data }),
--- a/client/src/utils/plugin/openai.ts
+++ b/client/src/utils/plugin/openai.ts
@@ -16,6 +16,11 @@ export const getOpenAiEncMap = () => {
        '<|im_end|>': 100265,
        '<|im_sep|>': 100266
      }),
+      [OpenAiChatEnum.GPT3516k]: encoding_for_model('gpt-3.5-turbo', {
+        '<|im_start|>': 100264,
+        '<|im_end|>': 100265,
+        '<|im_sep|>': 100266
+      }),
      [OpenAiChatEnum.GPT4]: encoding_for_model('gpt-4', {
        '<|im_start|>': 100264,
        '<|im_end|>': 100265,
@@ -36,6 +41,11 @@ export const getOpenAiEncMap = () => {
        '<|im_end|>': 100265,
        '<|im_sep|>': 100266
      }),
+      [OpenAiChatEnum.GPT3516k]: encoding_for_model('gpt-3.5-turbo', {
+        '<|im_start|>': 100264,
+        '<|im_end|>': 100265,
+        '<|im_sep|>': 100266
+      }),
      [OpenAiChatEnum.GPT4]: encoding_for_model('gpt-4', {
        '<|im_start|>': 100264,
        '<|im_end|>': 100265,
@@ -55,6 +65,11 @@ export const getOpenAiEncMap = () => {
      '<|im_end|>': 100265,
      '<|im_sep|>': 100266
    }),
+    [OpenAiChatEnum.GPT3516k]: encoding_for_model('gpt-3.5-turbo', {
+      '<|im_start|>': 100264,
+      '<|im_end|>': 100265,
+      '<|im_sep|>': 100266
+    }),
    [OpenAiChatEnum.GPT4]: encoding_for_model('gpt-4', {
      '<|im_start|>': 100264,
      '<|im_end|>': 100265,