undms

Text and metadata extraction library for document files with text similarity comparison, built with napi-rs.

Installation

bun add undms

Features

• Extracts text and metadata from document files
• Computes similarity between extracted documents and reference texts
• Works in Node.js and Bun (via N-API)

Supported formats

• Text: text/*, plus JSON/XML/JS/TS MIME variants
• DOCX: application/vnd.openxmlformats-officedocument.wordprocessingml.document
• XLSX: application/vnd.openxmlformats-officedocument.spreadsheetml.sheet
• PDF: application/pdf
• Images: image/jpeg, image/png, image/gif, image/bmp, image/tiff, image/webp

Metadata schema

Each extracted document may include a metadata payload with these optional fields:

type MetadataPayload = {
  text?: {
    lineCount: number;
    wordCount: number;
    characterCount: number;
    nonWhitespaceCharacterCount: number;
  };
  docx?: {
    paragraphCount: number;
    tableCount: number;
    imageCount: number;
    hyperlinkCount: number;
  };
  xlsx?: {
    sheetCount: number;
    sheetNames: string[];
    rowCount: number;
    columnCount: number;
    cellCount: number;
  };
  pdf?: {
    title?: string;
    author?: string;
    subject?: string;
    producer?: string;
    pageSize?: { width: number; height: number };
    pageCount: number;
  };
  image?: {
    width: number;
    height: number;
    format?: string;
    cameraMake?: string;
    cameraModel?: string;
    datetimeOriginal?: string;
    location: {
      latitude?: number;
      longitude?: number;
    };
  };
};

Handler details

• Text: decodes content (UTF-8 by default) and provides text metadata.
• DOCX: extracts paragraphs, plus paragraph/table/image/hyperlink counts.
• XLSX: extracts cell text and reports sheet/row/column/cell counts.
• PDF: extracts text and document info (title/author/subject/producer) and page size/count when available.
• Images: runs OCR to extract text and reads EXIF for camera details and GPS location when present.

Troubleshooting

• OCR is CPU-intensive; large images can be slow.
• EXIF GPS fields depend on the source image; if absent, location still exists but latitude/longitude are undefined.
• PDF metadata fields are optional and may be empty when the source file does not include them.
• If OCR returns empty text, check that the image has legible, high-contrast text.

Usage

import { extract, computeDocumentSimilarity, computeTextSimilarity } from 'undms';

const result = extract([
  {
    name: 'note.txt',
    size: 12,
    type: 'text/plain',
    lastModified: Date.now(),
    webkitRelativePath: '',
    buffer: Buffer.from('hello world!'),
  },
]);

const matches = computeDocumentSimilarity(
  result[0].documents.map((doc) => ({
    name: doc.name,
    size: doc.size,
    type: result[0].mimeType,
    lastModified: Date.now(),
    webkitRelativePath: '',
    buffer: Buffer.from(doc.content),
  })),
  ['reference text A', 'reference text B'],
  70,
  'hybrid',
);

const textMatches = computeTextSimilarity('alpha beta gamma', ['alpha beta gamma'], 99, 'hybrid');

API

extract(documents)

Extracts text and metadata from input documents. Output is grouped by MIME type.

computeDocumentSimilarity(documents, referenceTexts, threshold?, method?)

Extracts documents and computes similarity against the reference texts.

computeTextSimilarity(sourceText, referenceTexts, threshold?, method?)

Computes similarity for raw text without file extraction.

Development

Requirements

• Rust (latest stable)
• Node.js 12+ (for Node-API)
• Bun

Build

bun run build

Test

bun run test

Benchmarks

bun run bench
bun run bench:sweep