Perf: read file woker (#1337)

* perf: read file worker

* fix: Http node url input

* fix: htm2md

* fix: html2md

* fix: ts

* perf: Problem classification increases the matching order

* feat: tool response answer

packages/service/worker/file/extension/csv.ts

@@ -0,0 +1,21 @@
+import Papa from 'papaparse';
+import { ReadRawTextByBuffer, ReadFileResponse } from '../type';
+import { readFileRawText } from './rawText';
+
+// 加载源文件内容
+export const readCsvRawText = async (params: ReadRawTextByBuffer): Promise<ReadFileResponse> => {
+  const { rawText } = readFileRawText(params);
+
+  const csvArr = Papa.parse(rawText).data as string[][];
+
+  const header = csvArr[0];
+
+  const formatText = header
+    ? csvArr.map((item) => item.map((item, i) => `${header[i]}:${item}`).join('\n')).join('\n')
+    : '';
+
+  return {
+    rawText,
+    formatText
+  };
+};

packages/service/worker/file/extension/docx.ts

@@ -0,0 +1,23 @@
+import mammoth from 'mammoth';
+import { ReadRawTextByBuffer, ReadFileResponse } from '../type';
+import { html2md } from '../../htmlStr2Md/utils';
+
+/**
+ * read docx to markdown
+ */
+export const readDocsFile = async ({ buffer }: ReadRawTextByBuffer): Promise<ReadFileResponse> => {
+  try {
+    const { value: html } = await mammoth.convertToHtml({
+      buffer
+    });
+
+    const rawText = html2md(html);
+
+    return {
+      rawText
+    };
+  } catch (error) {
+    console.log('error doc read:', error);
+    return Promise.reject('Can not read doc file, please convert to PDF');
+  }
+};

packages/service/worker/file/extension/html.ts

@@ -0,0 +1,13 @@
+import { ReadRawTextByBuffer, ReadFileResponse } from '../type';
+import { readFileRawText } from './rawText';
+import { html2md } from '../../htmlStr2Md/utils';
+
+export const readHtmlRawText = async (params: ReadRawTextByBuffer): Promise<ReadFileResponse> => {
+  const { rawText: html } = readFileRawText(params);
+
+  const rawText = html2md(html);
+
+  return {
+    rawText
+  };
+};

packages/service/worker/file/extension/pdf.ts

@@ -0,0 +1,68 @@
+import * as pdfjs from 'pdfjs-dist/legacy/build/pdf.mjs';
+// @ts-ignore
+import('pdfjs-dist/legacy/build/pdf.worker.min.mjs');
+import { ReadRawTextByBuffer, ReadFileResponse } from '../type';
+
+type TokenType = {
+  str: string;
+  dir: string;
+  width: number;
+  height: number;
+  transform: number[];
+  fontName: string;
+  hasEOL: boolean;
+};
+
+export const readPdfFile = async ({ buffer }: ReadRawTextByBuffer): Promise<ReadFileResponse> => {
+  const readPDFPage = async (doc: any, pageNo: number) => {
+    const page = await doc.getPage(pageNo);
+    const tokenizedText = await page.getTextContent();
+
+    const viewport = page.getViewport({ scale: 1 });
+    const pageHeight = viewport.height;
+    const headerThreshold = pageHeight * 0.95;
+    const footerThreshold = pageHeight * 0.05;
+
+    const pageTexts: TokenType[] = tokenizedText.items.filter((token: TokenType) => {
+      return (
+        !token.transform ||
+        (token.transform[5] < headerThreshold && token.transform[5] > footerThreshold)
+      );
+    });
+
+    // concat empty string 'hasEOL'
+    for (let i = 0; i < pageTexts.length; i++) {
+      const item = pageTexts[i];
+      if (item.str === '' && pageTexts[i - 1]) {
+        pageTexts[i - 1].hasEOL = item.hasEOL;
+        pageTexts.splice(i, 1);
+        i--;
+      }
+    }
+
+    page.cleanup();
+
+    return pageTexts
+      .map((token) => {
+        const paragraphEnd = token.hasEOL && /([。？！.?!\n\r]|(\r\n))$/.test(token.str);
+
+        return paragraphEnd ? `${token.str}\n` : token.str;
+      })
+      .join('');
+  };
+
+  const loadingTask = pdfjs.getDocument(buffer.buffer);
+  const doc = await loadingTask.promise;
+
+  const pageTextPromises = [];
+  for (let pageNo = 1; pageNo <= doc.numPages; pageNo++) {
+    pageTextPromises.push(readPDFPage(doc, pageNo));
+  }
+  const pageTexts = await Promise.all(pageTextPromises);
+
+  loadingTask.destroy();
+
+  return {
+    rawText: pageTexts.join('')
+  };
+};

packages/service/worker/file/extension/pptx.ts

@@ -0,0 +1,18 @@
+import { ReadRawTextByBuffer, ReadFileResponse } from '../type';
+// import { parseOfficeAsync } from 'officeparser';
+import { parseOffice } from '../parseOffice';
+
+export const readPptxRawText = async ({
+  buffer,
+  encoding
+}: ReadRawTextByBuffer): Promise<ReadFileResponse> => {
+  const result = await parseOffice({
+    buffer,
+    encoding: encoding as BufferEncoding,
+    extension: 'pptx'
+  });
+
+  return {
+    rawText: result
+  };
+};

packages/service/worker/file/extension/rawText.ts

@@ -0,0 +1,28 @@
+import iconv from 'iconv-lite';
+import { ReadRawTextByBuffer, ReadFileResponse } from '../type';
+
+const rawEncodingList = [
+  'ascii',
+  'utf8',
+  'utf-8',
+  'utf16le',
+  'utf-16le',
+  'ucs2',
+  'ucs-2',
+  'base64',
+  'base64url',
+  'latin1',
+  'binary',
+  'hex'
+];
+
+// 加载源文件内容
+export const readFileRawText = ({ buffer, encoding }: ReadRawTextByBuffer): ReadFileResponse => {
+  const content = rawEncodingList.includes(encoding)
+    ? buffer.toString(encoding as BufferEncoding)
+    : iconv.decode(buffer, 'gbk');
+
+  return {
+    rawText: content
+  };
+};

packages/service/worker/file/extension/xlsx.ts

@@ -0,0 +1,45 @@
+import { ReadRawTextByBuffer, ReadFileResponse } from '../type';
+import xlsx from 'node-xlsx';
+import Papa from 'papaparse';
+
+export const readXlsxRawText = async ({
+  buffer
+}: ReadRawTextByBuffer): Promise<ReadFileResponse> => {
+  const result = xlsx.parse(buffer, {
+    skipHidden: false,
+    defval: ''
+  });
+
+  const format2Csv = result.map(({ name, data }) => {
+    return {
+      title: `#${name}`,
+      csvText: data.map((item) => item.join(',')).join('\n')
+    };
+  });
+
+  const rawText = format2Csv.map((item) => item.csvText).join('\n');
+  const formatText = format2Csv
+    .map((item) => {
+      const csvArr = Papa.parse(item.csvText).data as string[][];
+      const header = csvArr[0];
+
+      const formatText = header
+        ? csvArr
+            .map((item) =>
+              item
+                .map((item, i) => (item ? `${header[i]}:${item}` : ''))
+                .filter(Boolean)
+                .join('\n')
+            )
+            .join('\n')
+        : '';
+
+      return `${item.title}\n${formatText}`;
+    })
+    .join('\n');
+
+  return {
+    rawText: rawText,
+    formatText
+  };
+};