Duckling

A modern, user-friendly browser-based graphical interface for Docling - the powerful document conversion library by IBM.

Features

Drag-and-Drop Upload: Simply drag your documents onto the interface
URL-Based Conversion: Convert documents directly from URLs with automatic image extraction for HTML pages
Multi-file and folder upload: Convert multiple files or URLs at once from one drop zone—choose a folder, use Choose files…, or drag and drop; unsupported extensions are skipped client-side, and the API rejects a batch if nothing can be converted
Multi-Format Support: Convert PDFs, Word docs, PowerPoints, Excel files, HTML, Markdown, images, and more
Multiple Export Formats: Export to Markdown, HTML, JSON, DocTags, Document Tokens, RAG Chunks, or plain text
Image & Table Extraction: Extract embedded images and tables with CSV export
Image Preview Gallery: View extracted images as thumbnails with full-size lightbox viewer
RAG-Ready Chunking: Generate document chunks optimized for RAG applications
Advanced OCR: Multiple OCR backends with GPU acceleration support
Real-Time Progress: Watch your document conversion progress in real-time
Conversion History: Access previously converted documents and reload them without re-conversion
Document Persistence: Processed documents are stored on disk and can be reloaded from history
Content Deduplication: Same file with identical settings reuses stored output (instant cache hits)
Highly Configurable: Customize OCR, tables, images, performance, and chunking settings
Beautiful Dark Theme: Easy on the eyes with a modern, professional design
Accessibility: Keyboard-friendly slide-overs named with aria-label, focus trap and restore, labeled controls and progress, focusable scroll regions for long content, link underlines (not color alone), and docs styling/scripts described in Accessibility

Supported Input Formats

Format	Extensions	Description
PDF	`.pdf`	Portable Document Format
Word	`.docx`	Microsoft Word documents
PowerPoint	`.pptx`	Microsoft PowerPoint presentations
Excel	`.xlsx`	Microsoft Excel spreadsheets
HTML	`.html`, `.htm`	Web pages
Markdown	`.md`, `.markdown`	Markdown files
Images	`.png`, `.jpg`, `.jpeg`, `.tiff`, `.gif`, `.webp`, `.bmp`	Direct image OCR
AsciiDoc	`.asciidoc`, `.adoc`	Technical documentation
PubMed XML	`.xml`	Scientific articles
USPTO XML	`.xml`	Patent documents

Export Formats

Format	Extension	Description
Markdown	`.md`	Formatted text with headers, lists, links
HTML	`.html`	Web-ready format with styling
JSON	`.json`	Full document structure
Plain Text	`.txt`	Simple text without formatting
DocTags	`.doctags`	Tagged document format
Document Tokens	`.tokens.json`	Token-level representation
RAG Chunks	`.chunks.json`	Chunks for RAG applications

Quick Start

Prerequisites

Python 3.10+ (3.13 recommended)
Node.js 20.19+ or 22+ (22 recommended for Vite 7)
npm or yarn

Installation

Clone the repository:

git clone https://github.com/duckling-ui/duckling.git
cd duckling

Set up the backend:

cd backend
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt

Set up the frontend:
```
cd ../frontend
npm install
```

Running the Application

Start the backend (in one terminal):
```
cd backend
source venv/bin/activate
python duckling.py
```
The API will be available at http://localhost:5001
Start the frontend (in another terminal):
```
cd frontend
npm run dev
```
The UI will be available at http://localhost:3000
Batch and folder uploads: Drag a folder or multiple files onto the drop zone, click the zone to pick a folder, or use Choose files… for loose files—see the Quick Start guide.

Using Docker

Documentation (MkDocs)

Duckling uses MkDocs + mkdocs-static-i18n for multilingual docs. Production docs at duckling-ui.org are built with mike for multi-version support. With Material’s integrated TOC (toc.integrate), translate section headings in localized Markdown (for example docs/fr/user-guide/features.md) so the sidebar matches each locale. Supported formats and Screenshots are mirrored under docs/{de,fr,es}/user-guide/formats.md and screenshots.md (UI strings, captions, tab titles; image paths ../../assets/...). Homepage feature tiles in docs/{de,fr,es}/index.md should link to those locale-specific heading anchors (for example #glisser-deposer on the French Features page), not English slugs. The de, fr, and es trees mirror the English layout (getting started, user guide, API, architecture, deployment, contributing, changelog): keep them in sync when you add or change docs.

Build (strict):

./scripts/docs-build.sh

Serve locally:

./scripts/docs-serve.sh

Quick Start (Build Locally):

docker-compose up --build

Using Pre-built Images:

# Download and run pre-built images
curl -O https://raw.githubusercontent.com/duckling-ui/duckling/main/docker-compose.prebuilt.yml
docker-compose -f docker-compose.prebuilt.yml up -d

Production Deployment:

docker-compose -f docker-compose.yml -f docker-compose.prod.yml up -d

Access the application at http://localhost:3000

Build Docker Images:

# Build images locally
./scripts/docker-build.sh

# Build and push to registry
./scripts/docker-build.sh --push --registry yourusername

When PRs are merged to main, images are automatically published to Docker Hub and GitHub Container Registry. See Docker Deployment Guide for details and required secrets.

Configuration

Backend Environment Variables

Create a .env file in the backend directory:

FLASK_ENV=development
SECRET_KEY=your-secret-key
DEBUG=True
MAX_CONTENT_LENGTH=104857600  # 100MB

Conversion Settings

Settings can be configured through the UI or via the API. All settings are organized into categories:

OCR Settings

Setting	Description	Default
`enabled`	Enable/disable OCR	`true`
`backend`	OCR engine (easyocr, tesseract, ocrmac, rapidocr)	`easyocr`
`language`	Primary language for recognition	`en`
`force_full_page_ocr`	OCR entire page vs detected regions	`false`
`use_gpu`	Enable GPU acceleration (EasyOCR)	`false`
`confidence_threshold`	Minimum confidence for results	`0.5`

Supported Languages: English, German, French, Spanish, Italian, Portuguese, Dutch, Polish, Russian, Japanese, Chinese (Simplified/Traditional), Korean, Arabic, Hindi, Thai, Vietnamese, Turkish, Ukrainian, Czech, Greek, Hebrew, Indonesian, Malay, Swedish, Danish, Finnish, Norwegian

Table Settings

Setting	Description	Default
`enabled`	Enable table detection	`true`
`mode`	Detection mode (fast/accurate)	`accurate`
`do_cell_matching`	Match cell content to structure	`true`
`structure_extraction`	Preserve table structure	`true`

Image Settings

Setting	Description	Default
`extract`	Extract embedded images	`true`
`classify`	Classify and tag images	`true`
`generate_page_images`	Create page images	`false`
`generate_picture_images`	Extract pictures as files	`true`
`generate_table_images`	Extract tables as images	`true`
`images_scale`	Scale factor (0.1-4.0)	`1.0`

Performance Settings

Setting	Description	Default
`device`	Processing device (auto/cpu/cuda/mps)	`auto`
`num_threads`	CPU threads (1-32)	`4`
`document_timeout`	Max processing time (seconds)	`null`

RAG Chunking Settings

Setting	Description	Default
`enabled`	Enable document chunking	`false`
`max_tokens`	Maximum tokens per chunk	`512`
`merge_peers`	Merge undersized chunks	`true`

API Endpoints

Conversion

Endpoint	Method	Description
`/api/convert`	POST	Upload and convert a document
`/api/convert/batch`	POST	Batch convert multiple documents
`/api/convert/<job_id>/status`	GET	Get conversion status
`/api/convert/<job_id>/result`	GET	Get conversion result
`/api/convert/<job_id>/images`	GET	List extracted images
`/api/convert/<job_id>/images/<id>`	GET	Download extracted image
`/api/convert/<job_id>/tables`	GET	List extracted tables
`/api/convert/<job_id>/tables/<id>/csv`	GET	Download table as CSV
`/api/convert/<job_id>/chunks`	GET	Get document chunks
`/api/export/<job_id>/<format>`	GET	Download converted file

Settings

Endpoint	Method	Description
`/api/settings`	GET/PUT	Get/update all settings
`/api/settings/reset`	POST	Reset to defaults
`/api/settings/formats`	GET	List supported formats
`/api/settings/ocr`	GET/PUT	OCR settings
`/api/settings/tables`	GET/PUT	Table settings
`/api/settings/images`	GET/PUT	Image settings
`/api/settings/performance`	GET/PUT	Performance settings
`/api/settings/chunking`	GET/PUT	Chunking settings
`/api/settings/output`	GET/PUT	Output settings

History

Endpoint	Method	Description
`/api/history`	GET	List conversion history
`/api/history/<job_id>`	GET	Get history entry
`/api/history/stats`	GET	Get conversion statistics (includes processing time, OCR/format breakdowns, queue depth)
`/api/history/search`	GET	Search history

Development

Running Tests

Backend tests:

cd backend
pytest

Frontend tests:

cd frontend
npm test

CI runs the same frontend test command; for Vitest + React components that listen on window after async data loads, see Frontend Tests in CONTRIBUTING.md (effect scheduling / act).

Project Structure

duckling/
├── backend/
│   ├── duckling.py         # Flask application
│   ├── config.py           # Configuration & defaults
│   ├── models/             # Database models
│   ├── routes/             # API endpoints
│   │   ├── convert.py      # Conversion endpoints
│   │   ├── settings.py     # Settings endpoints
│   │   └── history.py      # History endpoints
│   ├── services/           # Business logic
│   │   ├── converter.py    # Docling integration
│   │   ├── file_manager.py # File handling
│   │   └── history.py      # History management
│   └── tests/              # Backend tests
├── frontend/
│   ├── src/
│   │   ├── components/     # React components
│   │   │   ├── DropZone.tsx
│   │   │   ├── ConversionProgress.tsx
│   │   │   ├── ExportOptions.tsx
│   │   │   ├── SettingsPanel.tsx
│   │   │   └── HistoryPanel.tsx
│   │   ├── hooks/          # Custom hooks
│   │   │   ├── useConversion.ts
│   │   │   └── useSettings.ts
│   │   ├── services/       # API client
│   │   └── types/          # TypeScript types
│   └── tests/              # Frontend tests
├── uploads/                # Temporary uploads
├── outputs/                # Converted files
└── docker-compose.yml      # Docker configuration

Documentation

Comprehensive documentation is available using MkDocs:

In-App Documentation

The documentation is available directly in the Duckling UI. Click the Documentation button in the header to open the docs panel.

UI + Docs Languages

UI: English (en), Spanish (es), French (fr), German (de)
Docs: Built and served with language prefixes at /api/docs/site/en/, /api/docs/site/es/, /api/docs/site/fr/, /api/docs/site/de/

To add a new language:

UI: add frontend/src/locales/<lang>/common.json and wire it into frontend/src/i18n.ts
Docs: add docs/<lang>/... (mirroring the English structure), then add a language entry under the i18n plugin in mkdocs.yml

Note: The documentation must be built before it can be viewed in the UI. You can either:

Build manually (recommended for development): Install backend/requirements.txt, then from the repo root run mkdocs build. That file includes the MkDocs stack ( pymdown-extensions 10.21.2+ ) so Pygments does not fail on fenced code without a title.
Build from the UI: Click the "Build Documentation" button in the docs panel if the site hasn't been built yet.
Auto-build: The backend will automatically attempt to build the docs if MkDocs is installed and the site doesn't exist.

View Documentation Locally

# MkDocs is included in backend/requirements.txt — activate that venv.

# Serve documentation with live reload (for editing)
mkdocs serve

Then open http://localhost:8000 in your browser.

Build Documentation

mkdocs build

The built documentation will be in the site/ directory and served at /api/docs/site/ by the backend.

Documentation Sections

Getting Started - Installation and quick start
User Guide - Features and configuration
Docling docs - Curated upstream Docling documentation (vendored subset)
API Reference - Complete API documentation
Architecture - System design and components
Deployment - Production deployment guide
Contributing - How to contribute

Contributing

See CONTRIBUTING.md for guidelines.

Security

See SECURITY.md for security policies.

API hardening: History document reload (GET /api/history/{job_id}/load) validates job_id using a strict allowlist and enforces safe output directory resolution to prevent path traversal.

License

This project is licensed under the MIT License - see LICENSE for details.

Acknowledgments

Docling by IBM for the powerful document conversion engine
React for the frontend framework
Flask for the backend framework
Tailwind CSS for styling
Framer Motion for animations

Name		Name	Last commit message	Last commit date
Latest commit History 184 Commits
.cursor/rules		.cursor/rules
.github		.github
backend		backend
docs		docs
frontend		frontend
scripts		scripts
tests		tests
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
MULTI_USER_IMPLEMENTATION.md		MULTI_USER_IMPLEMENTATION.md
README.md		README.md
SECURITY.md		SECURITY.md
arch.md		arch.md
diagrams.md		diagrams.md
docker-compose.prebuilt.yml		docker-compose.prebuilt.yml
docker-compose.prod.yml		docker-compose.prod.yml
docker-compose.yml		docker-compose.yml
duckling-old.png		duckling-old.png
duckling.png		duckling.png
mkdocs.yml		mkdocs.yml

Folders and files

Latest commit

History

Repository files navigation

Duckling

Features

Supported Input Formats

Export Formats

Quick Start

Prerequisites

Installation

Running the Application

Using Docker

Documentation (MkDocs)

Configuration

Backend Environment Variables

Conversion Settings

OCR Settings

Table Settings

Image Settings

Performance Settings

RAG Chunking Settings

API Endpoints

Conversion

Settings

History

Development

Running Tests

Project Structure

Documentation

In-App Documentation

UI + Docs Languages

View Documentation Locally

Build Documentation

Documentation Sections

Contributing

Security

License

Acknowledgments

About

Resources

License

Code of conduct

Contributing

Security policy

Uh oh!

Stars

Watchers

Forks

Releases 12

Packages 0

Uh oh!

Uh oh!

Contributors

Uh oh!

Languages

Packages