@fastgpt-sdk/anydoc

Convert Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV, and PDF files into clean GitHub-Flavored Markdown. Node.js bindings for the anydoc Rust crate, built by Firecrawl. Also available as a hosted API through Firecrawl Parse, which adds our OCR models for the scanned pages anydoc can't read on its own.

Every format parses into one shared document model and renders through a single Markdown serializer, so headings, tables, lists, and footnotes come out the same no matter which format goes in. Conversion runs on the libuv thread pool and never blocks the event loop. TypeScript types ship with the package.

npm install @fastgpt-sdk/anydoc

Supported formats

CLI

The package ships an anydoc command, so npx converts a document with no install:

npx @fastgpt-sdk/anydoc report.docx               # Markdown to stdout
npx @fastgpt-sdk/anydoc slides.pptx -o slides.md  # or to a file
npx @fastgpt-sdk/anydoc - --format csv < data.csv # read stdin
npx @fastgpt-sdk/anydoc scan.pdf --ocr hosted     # scanned pages via Firecrawl Parse

Markdown goes to stdout, errors to stderr, and anydoc --help covers the rest.

Usage

import { toDocument, toMarkdown, toMarkdownBytes } from '@fastgpt-sdk/anydoc';

// From a file path:
const markdown = await toMarkdown('report.docx');

// From bytes, with the format detected from the content:
const fromBytes = await toMarkdownBytes(bytes);

// Or name it, which signature-less formats (CSV) need:
const fromCsv = await toMarkdownBytes(bytes, 'csv');

// Preserve embedded images and their positions in the Markdown:
const { markdown, assets } = await toMarkdownBytes(bytes, 'doc', {
  embeddedImageMode: 'reference',
  maxImageBytes: 32 * 1024 * 1024,
  maxImageTotalBytes: 128 * 1024 * 1024,
});
// markdown contains ![alt](asset:0); assets contains { id: 0, mediaType, data, ... }.

// Or stop at the document model, which also carries embedded assets:
const document = await toDocument(bytes);

Scanned pages

anydoc converts locally and does not do OCR, so a PDF with scanned or image-only pages rejects with needsOcr. Opt in with ocr: 'hosted' to send that document to Firecrawl Parse. No signup needed. Set apiKey or FIRECRAWL_API_KEY for higher limits.

const markdown = await toMarkdown('scan.pdf', { ocr: 'hosted' });

On the CLI, anydoc scan.pdf --ocr hosted.

Errors

A conversion rejects only when no complete Markdown could come out of the file. The rejection is an Error whose code names what went wrong:

try {
  return await toMarkdown(path);
} catch (error) {
  // No document comes out of these, so record the file and take the next one.
  if (error.code === 'encrypted' || error.code === 'unsupported') {
    unconverted.push({ path, reason: error.code });
    return null;
  }
  throw error;
}

error.message carries the detail, naming the package part at fault where the format identifies one. TypeScript gets the union as ConvertErrorCode.

Format detection

The format is read from the file content, using the marker its specification designates: the PDF header, the RTF open group, OLE stream names, the ZIP package mimetype and content types. CSV has no such marker, so detection returns null for it and the extension, or an explicit format, names it instead.

formatFromBytes(bytes); // 'docx', or null when nothing matches
formatFromExtension('.pptm'); // 'pptx'
formatFromPath('report.odt'); // 'odt'

Images and embedded objects

By default, an embedded image renders as its alt text while the bytes stay on document.assets. Pass embeddedImageMode: 'reference' to toMarkdown or toMarkdownBytes to receive { markdown, assets }: the Markdown contains asset:<id> image URLs and assets contains the corresponding image/* bytes. maxImageBytes and maxImageTotalBytes cap one image and all images respectively; both default to 128 MiB and reject with resourceLimit when crossed. Non-image embedded objects keep the default alt-text behavior. Images that carry an external URL always render as ordinary Markdown images.

Full behavior notes and benchmarks live in the repository README.

License

MIT