harshankur/officeParser

Overview

officeParser is a Node‑compatible library that can read many common office‑document formats (like DOCX, PPTX, XLSX, PDF …​) and turn them into a strictly‑typed abstract syntax tree (AST). The same AST can then be written back out to a wide range of target formats – DOCX, ODT, LaTeX, Markdown, HTML, CSV, RTF, PDF, EPUB‑…​including RAG‑ready chunks for retrieval‑augmented generation pipelines.

The project ships both a programmatic API (for use in JavaScript/TypeScript code) and a CLI (npx officeparser …) for quick conversions on the command line. A browser‑based visualiser lets you upload a file, see the generated AST, tweak options and preview the output instantly.

---

Key Features

Feature	What it does
Multi‑format parsing	Reads DOCX, PPTX, XLSX, ODT/ODP/ODS/ODG, PDF, RTF, CSV, Markdown, HTML, EPUB, LaTeX (including Overleaf zip projects) and produces a unified AST.
Multi‑format generation	From the AST you can emit DOCX, ODT, LaTeX, Markdown, HTML, CSV, RTF, PDF (native via pdf‑lib or Puppeteer), EPUB, plain text, or RAG‑chunks.
Native PDF engine	No headless browser needed – pdf-lib can create PDF bytes directly in Node or the browser.
Password handling	Single password/onPassword option works for encrypted PDFs, OOXML and ODF files.
Template / mail‑merge	OfficeTemplate.render fills {{placeholders}} in a DOCX/ODT template, supporting batch rendering.
OCR integration	Optional Tesseract‑based OCR on extracted images (useful for scanned PDFs or PPTX slides).
RAG chunking	Built‑in chunks output splits a document into retrieval‑ready pieces, with configurable size and overlap.
Fine‑grained config	Hundreds of CLI flags / TypeScript options control things like heading detection, column layout, colour extraction, image handling, LaTeX document class, etc.
Browser visualiser	Interactive web UI (/officeParser/) shows the AST tree, lets you edit config and preview generated files live.
Type‑safe API	Written in TypeScript; the AST and all public functions have strict typings.

---

Typical Use‑Cases

Scenario	How you would use officeParser
Extract plain text for indexing	await officeParser.parseOffice('file.pdf'); then await ast.to('text'); – the text respects original layout (columns, tables, headings).
Convert a Word doc to Markdown	CLI: npx officeparser report.docx --to=md --output=report.md or programmatically: await OfficeConverter.convert('report.docx', 'md');
Generate a PDF from a DOCX without a headless browser	Set pdfConfig.engine: 'native' and call ast.to('pdf').
Create a mail‑merge batch	Load a DOCX template, call OfficeTemplate.render([{name:'Alice'}, {name:'Bob'}]) to produce separate docs.
Feed documents into a Retrieval‑Augmented Generation pipeline	npx officeparser big_report.pdf --to=chunks or OfficeConverter.convert(file, 'chunks', {chunkingConfig:{maxTokens:500}}).
Process scanned PDFs	Enable --ocr (or ocr: true in code) together with extractAttachments – OCR runs on each extracted image and the recognised text becomes part of the AST.

---

Getting Started

# Install the library
npm i officeparser   # requires Node >= 22.13

Quick programmatic example (Node)

import { OfficeParser, OfficeConverter } from 'officeparser';

// Parse a DOCX file
const ast = await OfficeParser.parseOffice('proposal.docx', {
  extractAttachments: true,
  ocr: true,               // run OCR on any embedded images
});

// Convert the same document to HTML
const html = await ast.to('html');
console.log(html.value);

// Or do a one‑step conversion
await OfficeConverter.convert('proposal.docx', 'pdf', {
  pdfConfig: { engine: 'native' },
  output: 'proposal.pdf',
});

CLI one‑liner

npx officeparser presentation.pptx \
  --ocr --extractAttachments \
  --to=html --output=preview.html

---

Limitations & Known Issues

• Very large PDFs can hit the default memory limits of the separate pdf.js process; increase pdfParserConfig.processMemoryMb if needed.
• Text‑based formats (MD, HTML, CSV, TEX) need an explicit fileType when supplied as a raw buffer because they lack magic bytes.
• The native PDF engine does not support advanced CSS‑based layout – for complex styling fall back to the html (Puppeteer) engine.
• OCR is CPU‑intensive and may be slow on large documents; it can be cancelled via an AbortSignal.

---

License

MIT – free for commercial and personal use.

---

officeParser aims to be a one‑stop solution for developers who need to ingest, transform, or generate office‑style documents, especially when those documents feed downstream AI or RAG workflows.