Fixed the duplicate data check problem, history filter and add tts stream (#477)

2025-07-30 02:12:38 +00:00 · 2023-11-16 16:22:08 +08:00
parent 16103029f5
commit fbe1d8cfed
31 changed files with 359 additions and 187 deletions
--- a/projects/app/src/pages/api/core/chat/item/getSpeech.ts
+++ b/projects/app/src/pages/api/core/chat/item/getSpeech.ts
@@ -1,12 +1,13 @@
 import type { NextApiRequest, NextApiResponse } from 'next';
 import { jsonRes } from '@fastgpt/service/common/response';
 import { connectToDatabase } from '@/service/mongo';
-import { MongoChatItem } from '@fastgpt/service/core/chat/chatItemSchema';
 import { GetChatSpeechProps } from '@/global/core/chat/api.d';
 import { text2Speech } from '@fastgpt/service/core/ai/audio/speech';
 import { pushAudioSpeechBill } from '@/service/support/wallet/bill/push';
 import { authCert } from '@fastgpt/service/support/permission/auth/common';
 import { authType2BillSource } from '@/service/support/wallet/bill/utils';
+import { getAudioSpeechModel } from '@/service/core/ai/model';
+import { MongoTTSBuffer } from '@fastgpt/service/common/buffer/tts/schema';

 /* 
 1. get tts from chatItem store
@@ -18,50 +19,66 @@ import { authType2BillSource } from '@/service/support/wallet/bill/utils';
 export default async function handler(req: NextApiRequest, res: NextApiResponse) {
  try {
    await connectToDatabase();
-    const { chatItemId, ttsConfig, input } = req.body as GetChatSpeechProps;
+    const { ttsConfig, input } = req.body as GetChatSpeechProps;
+
+    if (!ttsConfig.model || !ttsConfig.voice) {
+      throw new Error('model or voice not found');
+    }

    const { teamId, tmbId, authType } = await authCert({ req, authToken: true });

-    const chatItem = await (async () => {
-      if (!chatItemId) return null;
-      return await MongoChatItem.findOne(
-        {
-          dataId: chatItemId
-        },
-        'tts'
-      );
-    })();
+    const ttsModel = getAudioSpeechModel(ttsConfig.model);
+    const voiceData = ttsModel.voices.find((item) => item.value === ttsConfig.voice);

-    if (chatItem?.tts) {
-      return jsonRes(res, {
-        data: chatItem.tts
-      });
+    if (!voiceData) {
+      throw new Error('voice not found');
    }

-    const { tts, model } = await text2Speech({
+    const ttsBuffer = await MongoTTSBuffer.findOne(
+      {
+        bufferId: voiceData.bufferId,
+        text: input
+      },
+      'buffer'
+    );
+
+    if (ttsBuffer?.buffer) {
+      return res.end(new Uint8Array(ttsBuffer.buffer.buffer));
+    }
+
+    await text2Speech({
+      res,
+      input,
      model: ttsConfig.model,
      voice: ttsConfig.voice,
-      input
-    });
+      props: {
+        // temp code
+        baseUrl: ttsModel.baseUrl || '',
+        key: ttsModel.key || ''
+      },
+      onSuccess: async ({ model, buffer }) => {
+        try {
+          pushAudioSpeechBill({
+            model: model,
+            textLength: input.length,
+            tmbId,
+            teamId,
+            source: authType2BillSource({ authType })
+          });

-    (async () => {
-      if (!chatItem) return;
-      try {
-        chatItem.tts = tts;
-        await chatItem.save();
-      } catch (error) {}
-    })();
-
-    jsonRes(res, {
-      data: tts
-    });
-
-    pushAudioSpeechBill({
-      model: model,
-      textLength: input.length,
-      tmbId,
-      teamId,
-      source: authType2BillSource({ authType })
+          await MongoTTSBuffer.create({
+            bufferId: voiceData.bufferId,
+            text: input,
+            buffer
+          });
+        } catch (error) {}
+      },
+      onError: (err) => {
+        jsonRes(res, {
+          code: 500,
+          error: err
+        });
+      }
    });
  } catch (err) {
    jsonRes(res, {
--- a/projects/app/src/pages/api/system/getInitData.ts
+++ b/projects/app/src/pages/api/system/getInitData.ts
@@ -34,8 +34,12 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)
      qaModels: global.qaModels,
      cqModels: global.cqModels,
      extractModels: global.extractModels,
-      qgModels: global.qgModels,
      vectorModels: global.vectorModels,
+      audioSpeechModels: global.audioSpeechModels.map((item) => ({
+        ...item,
+        baseUrl: undefined,
+        key: undefined
+      })),
      priceMd: global.priceMd,
      systemVersion: global.systemVersion || '0.0.0'
    }
--- a/projects/app/src/pages/api/v1/chat/completions.ts
+++ b/projects/app/src/pages/api/v1/chat/completions.ts
@@ -86,6 +86,7 @@ export default withNextCors(async function handler(req: NextApiRequest, res: Nex
    if (chatMessages[chatMessages.length - 1].obj !== ChatRoleEnum.Human) {
      chatMessages.pop();
    }
+
    // user question
    const question = chatMessages.pop();
    if (!question) {
@@ -173,15 +174,7 @@ export default withNextCors(async function handler(req: NextApiRequest, res: Nex
    const isAppOwner = !shareId && String(user.team.tmbId) === String(app.tmbId);

    /* format prompts */
-    const prompts = history.concat(gptMessage2ChatType(messages));
-
-    // set sse response headers
-    if (stream) {
-      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 concatHistory = history.concat(chatMessages);

    /* start flow controller */
    const { responseData, answerText } = await dispatchModules({
@@ -193,7 +186,7 @@ export default withNextCors(async function handler(req: NextApiRequest, res: Nex
      tmbId: user.team.tmbId,
      variables,
      params: {
-        history: prompts,
+        history: concatHistory,
        userChatInput: question.value
      },
      stream,
--- a/projects/app/src/pages/app/detail/components/TTSSelect.tsx
+++ b/projects/app/src/pages/app/detail/components/TTSSelect.tsx
@@ -6,10 +6,10 @@ import React, { useCallback, useMemo } from 'react';
 import { useTranslation } from 'next-i18next';
 import MySelect from '@/components/Select';
 import { TTSTypeEnum } from '@/constants/app';
-import { Text2SpeechVoiceEnum, openaiTTSModel } from '@fastgpt/global/core/ai/speech/constant';
 import { AppTTSConfigType } from '@/types/app';
 import { useAudioPlay } from '@/web/common/utils/voice';
 import { useLoading } from '@/web/common/hooks/useLoading';
+import { audioSpeechModels } from '@/web/common/system/staticData';

 const TTSSelect = ({
  value,
@@ -37,10 +37,16 @@ const TTSSelect = ({
      if (e === TTSTypeEnum.none || e === TTSTypeEnum.web) {
        onChange({ type: e as `${TTSTypeEnum}` });
      } else {
+        const audioModel = audioSpeechModels.find((item) =>
+          item.voices.find((voice) => voice.value === e)
+        );
+        if (!audioModel) {
+          return;
+        }
        onChange({
          type: TTSTypeEnum.model,
-          model: openaiTTSModel,
-          voice: e as `${Text2SpeechVoiceEnum}`,
+          model: audioModel.model,
+          voice: e,
          speed: 1
        });
      }
@@ -77,12 +83,7 @@ const TTSSelect = ({
        list={[
          { label: t('core.app.tts.Close'), value: TTSTypeEnum.none },
          { label: t('core.app.tts.Web'), value: TTSTypeEnum.web },
-          { label: 'Alloy', value: Text2SpeechVoiceEnum.alloy },
-          { label: 'Echo', value: Text2SpeechVoiceEnum.echo },
-          { label: 'Fable', value: Text2SpeechVoiceEnum.fable },
-          { label: 'Onyx', value: Text2SpeechVoiceEnum.onyx },
-          { label: 'Nova', value: Text2SpeechVoiceEnum.nova },
-          { label: 'Shimmer', value: Text2SpeechVoiceEnum.shimmer }
+          ...audioSpeechModels.map((item) => item.voices).flat()
        ]}
        onchange={onclickChange}
      />
--- a/projects/app/src/pages/dataset/detail/components/Import/Csv.tsx
+++ b/projects/app/src/pages/dataset/detail/components/Import/Csv.tsx
@@ -2,11 +2,13 @@ import React from 'react';
 import { Box, Flex, Button } from '@chakra-ui/react';
 import { useConfirm } from '@/web/common/hooks/useConfirm';
 import { useImportStore, SelectorContainer, PreviewFileOrChunk } from './Provider';
+import { useTranslation } from 'next-i18next';

 const fileExtension = '.csv';
 const csvTemplate = `index,content\n"被索引的内容","对应的答案。CSV 中请注意内容不能包含双引号，双引号是列分割符号"\n"什么是 laf","laf 是一个云函数开发平台……",""\n"什么是 sealos","Sealos 是以 kubernetes 为内核的云操作系统发行版,可以……"`;

 const CsvImport = () => {
+  const { t } = useTranslation();
  const { successChunks, totalChunks, isUnselectedFile, onclickUpload, uploading } =
    useImportStore();

@@ -24,6 +26,7 @@ const CsvImport = () => {
          value: csvTemplate,
          type: 'text/csv'
        }}
+        tip={t('dataset.import csv tip')}
      >
        <Flex mt={3}>
          <Button isDisabled={uploading} onClick={openConfirm(onclickUpload)}>
--- a/projects/app/src/pages/dataset/detail/components/Import/FileSelect.tsx
+++ b/projects/app/src/pages/dataset/detail/components/Import/FileSelect.tsx
@@ -55,6 +55,7 @@ export interface Props extends BoxProps {
  };
  showUrlFetch?: boolean;
  showCreateFile?: boolean;
+  tip?: string;
 }

 const FileSelect = ({
@@ -65,6 +66,7 @@ const FileSelect = ({
  fileTemplate,
  showUrlFetch = true,
  showCreateFile = true,
+  tip,
  ...props
 }: Props) => {
  const { datasetDetail } = useDatasetStore();
@@ -423,6 +425,7 @@ const FileSelect = ({
          {t('file.Click to download file template', { name: fileTemplate.filename })}
        </Box>
      )}
+      {!!tip && <Box color={'myGray.500'}>{tip}</Box>}
      {selectingText !== undefined && (
        <FileSelectLoading loading text={selectingText} fixed={false} />
      )}
--- a/projects/app/src/pages/dataset/detail/components/Import/Provider.tsx
+++ b/projects/app/src/pages/dataset/detail/components/Import/Provider.tsx
@@ -403,12 +403,14 @@ export const SelectorContainer = ({
  showUrlFetch,
  showCreateFile,
  fileTemplate,
+  tip,
  children
 }: {
  fileExtension: string;
  showUrlFetch?: boolean;
  showCreateFile?: boolean;
  fileTemplate?: FileSelectProps['fileTemplate'];
+  tip?: string;
  children: React.ReactNode;
 }) => {
  const { files, setPreviewFile, isUnselectedFile, setFiles, chunkLen } = useImportStore();
@@ -433,6 +435,7 @@ export const SelectorContainer = ({
        showUrlFetch={showUrlFetch}
        showCreateFile={showCreateFile}
        fileTemplate={fileTemplate}
+        tip={tip}
        py={isUnselectedFile ? '100px' : 5}
      />
      {!isUnselectedFile && (
--- a/projects/app/src/pages/dataset/detail/components/InputDataModal.tsx
+++ b/projects/app/src/pages/dataset/detail/components/InputDataModal.tsx
@@ -117,10 +117,8 @@ const InputDataModal = ({
  const { mutate: sureImportData, isLoading: isImporting } = useRequest({
    mutationFn: async (e: InputDataType) => {
      if (!e.q) {
-        return toast({
-          title: '匹配的知识点不能为空',
-          status: 'warning'
-        });
+        setCurrentTab(TabEnum.content);
+        return Promise.reject(t('dataset.data.input is empty'));
      }
      if (countPromptTokens(e.q) >= maxToken) {
        return toast({