Features¶

Duckling provides a comprehensive set of features for document conversion.

Document Upload¶

Drag-and-Drop¶

Simply drag files onto the drop zone for instant upload. The interface validates file types and shows upload progress.

URL Input¶

Convert documents directly from URLs without downloading them first:

1. Click the URLs toggle button above the drop zone
2. Enter a document URL (must be HTTP or HTTPS)
3. Click Convert or press Enter
4. The document is downloaded and converted automatically

Supported URL features:

• Automatic file type detection from URL path
• Content-Type header detection for files without extensions
• Content-Disposition header support for filename extraction
• Same file type restrictions as local uploads
• Automatic image extraction for HTML pages: When converting HTML from URLs, Duckling automatically downloads all images referenced in the page and makes them available in the Image Preview Gallery

Batch Processing¶

Toggle batch mode to upload and convert multiple files at once:

1. Click Batch Mode toggle in the header
2. Drag multiple files onto the drop zone, or switch to URL mode and enter multiple URLs (one per line)
3. Monitor individual progress for each file
4. Download results separately or together

Batch URL Mode¶

In batch mode, the URL input changes to a multi-line textarea:

1. Enable Batch Mode in the header
2. Switch to URLs input mode
3. Paste multiple URLs, one per line
4. Click Convert All

OCR (Optical Character Recognition)¶

Extract text from scanned documents and images.

Supported Backends¶

Automatic Backend Installation¶

Duckling can automatically install OCR backends when you select them:

1. Open Settings panel
2. Select an OCR backend from the dropdown
3. If the backend is not installed, you'll see an Install button
4. Click to automatically install via pip

The Settings panel shows the status of each backend:

• ✓ Installed and ready - Backend is available for use
• ⚠ Not installed - Click to install (pip-installable backends)
• ℹ Requires system installation - Follow manual installation instructions

Language Support¶

28+ languages including:

• European: English, German, French, Spanish, Italian, Portuguese, Dutch, Polish, Russian
• Asian: Japanese, Chinese (Simplified/Traditional), Korean, Thai, Vietnamese
• Middle Eastern: Arabic, Hebrew, Turkish
• South Asian: Hindi

OCR Options¶

Table Extraction¶

Automatically detect and extract tables from documents.

Detection Modes¶

Export Options¶

• CSV: Download individual tables as CSV files
• Image: Download table as PNG image
• JSON: Full table structure in API response

Image Extraction¶

Extract embedded images from documents.

Options¶

Image Preview Gallery¶

After conversion, extracted images are displayed in a visual gallery:

• Thumbnail Grid: View all images as thumbnails in a responsive grid
• Hover Actions: Quick access to view and download buttons on hover
• Lightbox Viewer: Click any image to view full-size in a modal
• Navigation: Use arrow buttons to browse through multiple images
• Download: Download individual images directly from the gallery or lightbox

Document Enrichment¶

Enhance your converted documents with advanced AI-powered features.

Available Enrichments¶

Configuration¶

Enable enrichments in the Settings panel under Document Enrichment:

1. Open Settings (gear icon)
2. Scroll to "Document Enrichment" section
3. Toggle desired features on/off
4. Settings are saved automatically

Code Enrichment¶

When enabled, code blocks in your documents are enhanced with:

• Automatic programming language detection
• Syntax highlighting metadata
• Improved code structure recognition

Formula Enrichment¶

Extracts mathematical formulas and converts them to LaTeX:

• Inline equations: $E = mc^2$
• Display equations with proper formatting
• Better rendering in HTML and Markdown exports

Picture Classification¶

Automatically tags images with semantic types:

• Figure: Diagrams, illustrations, schematics
• Chart: Bar charts, line graphs, pie charts
• Photo: Photographs, screenshots
• Logo: Brand logos, icons
• Table: Table images (separate from table extraction)

Picture Description¶

Uses vision-language AI models to generate descriptive captions:

• Natural language descriptions of image content
• Useful for accessibility (alt text)
• Enhances searchability of documents
• Requires model download on first use

Pre-Downloading Models¶

To avoid delays during document processing, you can pre-download enrichment models:

1. Open Settings panel
2. Scroll to Document Enrichment section
3. Find the Pre-Download Models area at the bottom
4. Click Download next to any model you want to pre-download

RAG Chunking¶

Generate document chunks optimized for Retrieval-Augmented Generation.

How It Works¶

1. Document is split into semantic chunks
2. Each chunk respects document structure
3. Chunks include metadata (headings, page numbers)
4. Undersized chunks can be merged

Configuration¶

Output Format¶

{
  "chunks": [
    {
      "id": 1,
      "text": "Introduction to machine learning...",
      "meta": {
        "headings": ["Chapter 1", "Introduction"],
        "page": 1
      }
    }
  ]
}

Export Formats¶

Available Formats¶

Preview¶

The export panel shows a live preview of your converted content that updates as you switch between export formats.

Format-Specific Preview¶

• Dynamic Content: Preview automatically loads content for the selected export format
• Format Badge: Shows which format you're currently previewing
• Content Caching: Previously loaded formats are cached for instant switching

Rendered vs Raw Mode¶

For HTML and Markdown formats, toggle between rendered and raw views:

Other Formats¶

• JSON: Automatically pretty-printed with proper indentation
• Plain Text: Displayed as-is
• DocTags/Tokens: Raw format display

Conversion History¶

Access previously converted documents:

• View conversion status and metadata
• Re-download converted files
• Search history by filename
• View conversion statistics

History Features¶

• Search: Find documents by filename
• Filter: Filter by status (completed, failed)
• Statistics: View success rates and format breakdown
• Export: Download history as JSON