Web Analytics Made Easy - Statcounter
Skip to content

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Latest release: 0.0.11 (2026-03-30)

Unreleased

Planned

  • User authentication
  • Cloud storage integration
  • Conversion templates
  • API rate limiting
  • WebSocket for real-time updates
  • Dark/light theme toggle
  • Keyboard shortcuts
  • Accessibility improvements (WCAG 2.1)

0.0.11 - 2026-03-30

Changed

  • Accessibility: Contributing docs add Accessibility and MkDocs nav entry; built docs use link underlines, higher-contrast secondary text in stylesheets/extra.css, and javascripts/scrollable-focus.js for keyboard-scrollable code/tables, search dialog naming, and unique scrollable regions. See root CHANGELOG.md for web UI changes.
  • Upload UX: Toolbar batch toggle removed; one drop zone supports single file, multiple files, folders, and multi-line URLs. See root CHANGELOG.md for full technical notes.
  • Dependencies: No separate root docs requirements file; use backend/requirements.txt for MkDocs and the API (see root CHANGELOG.md).

Fixed

  • Docs: Localized home pages add aria-label on feature card links (card-link). javascripts/scrollable-focus.js names the Material search dialog, uniquely labels code toolbars and scrollable code/table regions, and hooks document$ for instant navigation.
  • Frontend tests: DocsPanel iframe-navigation test waits for the message listener to attach after mocked fetch completes and uses a longer waitFor timeout so CI stays green on slower runners.
  • Docs / CI: pymdown-extensions>=10.21.2 avoids a Pygments + highlight crash when building the site (same as root CHANGELOG.md).
  • Python / CI: markdown>=3.6 and mkdocs>=1.6 in backend/requirements.txt so pip can resolve the doc stack with pymdown 10.21.2+ (same as root CHANGELOG.md).

0.0.10a - 2026-03-23

Fixed

  • Backend dependencies: One backend/requirements.txt for API + in-app docs builds; removed duplicate backend/requirements-docs.txt.

Changed

  • Repository button: Documentation header/sidebar repo link now says "Star on GitHub" instead of "Go to repository" to encourage starring (links to the repo where users can star).
  • Documentation: Switched to mike for multi-version docs; deploys to gh-pages then rsyncs to duckling-ui.org. Root and /latest/ redirect correctly; version selector works.
  • Documentation navigation: Switched from horizontal top tabs to a single left sidebar with collapsible tree navigation; each major category (Home, Getting Started, etc.) can be expanded or collapsed.
  • Key Features tiles: Each feature tile on the docs homepage is now a clickable link to its detailed documentation (Features or Formats page).
  • CONTRIBUTING.md: Added DCO (Developer Certificate of Origin) sign-off requirement for all commits.
  • Contributing documentation: Full translations for German (de), Spanish (es), and French (fr); all locales now have consistent, complete content including DCO requirements.

Security

  • Fixed Rollup path traversal (GHSA-mw96-cpmx-2vgc) and Minimatch ReDoS (GHSA-3ppc-4f35-3m26) via npm overrides in frontend: rollup >=4.59.0, minimatch 9.0.6 for @typescript-eslint/typescript-estree.
  • Fixed Werkzeug safe_join Windows device names in multi-segment paths (CVE-2026-27199, GHSA-29vq-49wr-vm6x): upgraded werkzeug 3.1.4 → 3.1.6.
  • Fixed Flask session Vary: Cookie header when using in operator (CVE-2026-27205): upgraded flask 3.0.0 → 3.1.3.
  • SSRF prevention: URL validation before outbound requests in download_from_url, download_from_url_with_images, and download_image; blocks loopback, private IPs, link-local, metadata, and dangerous schemes.
  • CodeQL security fixes:
  • SSRF: validate_url_safe_for_request now returns the validated URL; all requests.get calls use the returned value to satisfy data-flow analysis.
  • ReDoS: HTML image extraction limited to 5MB before regex processing to mitigate polynomial regex on user-controlled content.
  • Path traversal: delete_output_folder now uses validate_job_id and get_validated_output_dir from security utils instead of manual checks.
  • Information exposure: Settings API error responses sanitized via _sanitize_error_for_client to prevent stack trace or sensitive data leakage.

0.0.10a - 2026-02-24

Added

  • Docker image publishing workflow: GitHub Action runs when PRs are merged to main, building multi-platform images and pushing to Docker Hub and GitHub Container Registry (requires DOCKERHUB_USERNAME and DOCKERHUB_TOKEN secrets).
  • Generate Chunks Now: Button in RAG Chunks tab to generate chunks on demand for completed documents (POST /api/history/{job_id}/generate-chunks)
  • Content-addressed deduplication: Same file + same document-affecting settings reuse stored content instead of re-converting
  • Cache hit: create symlink, load metadata, complete immediately (no Docling run)
  • Cache miss: run conversion, move output to content store, create symlink
  • Database migration scripts/migrate_add_content_hash.py adds content_hash column
  • Conversion statistics and metrics: Extended history stats for Docling and Duckling usage analytics
  • GET /api/history/stats returns avg_processing_seconds, ocr_backend_breakdown, output_format_breakdown, performance_device_breakdown, chunking_enabled_count, error_category_breakdown, source_type_breakdown, and queue_depth
  • Database migration scripts/migrate_add_stats_columns.py adds stats columns to conversions table
  • History panel displays average processing time and queue depth when available
  • Statistics panel: Dedicated viewer for conversion statistics (header button, "View full statistics" from History)
  • Extended statistics: Hardware and performance metrics in the Statistics panel
  • System section: hardware type (CPU/CUDA/MPS), CPU count, current CPU usage, GPU info
  • Average pages/sec and pages/sec per CPU
  • Conversion time distribution (median, 95th, 99th percentile)
  • Pages/sec over time chart
  • CPU usage averaged during each conversion (stored in DB)
  • Database migration scripts/migrate_add_cpu_usage_column.py adds cpu_usage_avg_during_conversion column
  • CPU usage is now process-specific (Duckling backend process, runs Docling), not system-wide
  • Per-conversion config stored: performance_device_used (resolved from "auto" at completion), images_classify_enabled
  • Database migration scripts/migrate_add_config_columns.py adds these columns
  • Stats breakdown by hardware, OCR backend, image classifier (pages/sec, conversion time per config)
  • UI language support (English en, Spanish es, French fr, German de) with a language switcher.
  • Multilingual MkDocs documentation (English, Spanish, French, German) served under /api/docs/site/<locale>/.
  • Dropzone panel category labels (Documents, Web, Images, Data) now fully internationalized.
  • Docling docs section in MkDocs (curated, vendored subset of upstream Docling documentation + sync script).
  • Session-Based User Settings: User settings stored per-session in the database instead of a shared file.

Security

  • Fixed frontend security vulnerabilities (esbuild GHSA-67mh-4wv8-2f99): Updated Vite 5→7, Vitest 1→4, and related dependencies.

Changed

  • Backend entry point renamed from app.py to duckling.py for better clarity.
  • Flask application name changed to "duckling" (displays as "Serving Flask app 'duckling'").

Fixed

  • Documentation navigation now displays fully localized page names in all supported languages.
  • Dropzone file format category labels now properly translate based on selected language.
  • Improved documentation page title extraction with better fallback to translated names.
  • In-app docs panel footer prev/next links now stay within the current sidebar category, and navigating inside the embedded docs keeps the sidebar selection in sync.
  • Fixed in-app docs rebuild failing with cannot access local variable 'shutil' when building the MkDocs site.
  • Backend docs rebuild now prefers the repo-local ./venv MkDocs environment to ensure required plugins (like i18n) are available.
  • Fix clicking a history entry not loading the document; now uses the history load endpoint (disk) instead of the in-memory result endpoint.
  • When document_json_path is missing in the DB, history load now finds and loads *.document.json from the output directory so all history items load, not just the first.
  • Document viewing panel now refreshes when loading a different history item (uses component key to remount with fresh state).
  • Updated vitest.config.ts for Vitest 4 compatibility.
  • Updated CI/CD Node.js version requirement to 22 (required for Vite 7).

0.0.9 - 2026-01-08

Added

  • Custom Branding: Duckling logo and version display in header.
  • URL-Based Document Conversion: Convert documents from URLs with automatic image extraction for HTML.
  • Document Enrichment Options: Code enrichment, formula enrichment, picture classification, picture description.
  • Enrichment Model Pre-Download: Download AI models before processing.
  • Image Preview Gallery: Visual thumbnails with lightbox viewer.
  • OCR Backend Auto-Installation: One-click installation for pip-installable backends.
  • Format-Specific Preview: Preview panel shows content in selected export format.
  • Rendered vs Raw Preview Mode: Toggle for HTML and Markdown.
  • Enhanced Docker Support: Multi-stage Dockerfiles, docker-compose variants, multi-platform builds.

Fixed

  • Multi-worker content retrieval (images, tables, results).
  • HTML preview in UI.
  • URL image extraction for unquoted src attributes.
  • Documentation panel now serves pre-built MkDocs site.
  • Environment variables and .env loading.
  • Case-insensitive file extensions.
  • Confidence score and OCR backend selection.

0.0.8 - 2026-01-07

Changed

  • Renamed: Project renamed from "Docling UI" to "Duckling"
  • Updated all documentation, code, and configuration files
  • Branding updated throughout the application

0.0.7 - 2026-01-07

Added

  • MkDocs Documentation: Migrated documentation to MkDocs with Material theme
  • Modern, searchable documentation site
  • Dark/light theme toggle
  • Mermaid diagram support
  • Improved navigation and organization

Changed

  • Documentation structure reorganized for better navigation
  • All diagrams converted to Mermaid format for live rendering

0.0.6 - 2025-12-11

Security

  • CRITICAL: Fixed Flask debug mode enabled by default in production
  • Debug mode now controlled by FLASK_DEBUG environment variable (default: false)
  • Host binding defaults to 127.0.0.1 instead of 0.0.0.0
  • HIGH: Updated vulnerable dependencies
  • flask-cors: 4.0.0 → 6.0.0 (CVE-2024-1681, CVE-2024-6844, CVE-2024-6866, CVE-2024-6839)
  • gunicorn: 21.2.0 → 23.0.0 (CVE-2024-1135, CVE-2024-6827)
  • werkzeug: 3.0.1 → 3.1.4 (CVE-2024-34069, CVE-2024-49766, CVE-2024-49767, CVE-2025-66221)
  • MEDIUM: Added path traversal protection to file serving endpoints
  • Image serving validates paths don't escape allowed directories
  • Blocks directory traversal sequences (..)
  • MEDIUM: Enhanced SQL query sanitization
  • Search queries now escape LIKE wildcards
  • Added query length limits
  • Added comprehensive SECURITY.md with:
  • Security audit summary
  • Production deployment checklist
  • Environment variable documentation
  • Vulnerability reporting guidelines

Changed

  • Backend now uses environment variables for all security-sensitive configuration
  • Default host changed from 0.0.0.0 to 127.0.0.1 for safer defaults

0.0.5 - 2025-12-10

Added

  • Batch Processing: Upload and convert multiple files at once
  • Toggle batch mode in the header

  • Process multiple documents simultaneously

  • Image & Table Extraction:

  • Extract embedded images from documents
  • Download individual images
  • Extract tables with full data preservation
  • Export tables to CSV format

  • View table previews in the UI

  • RAG/Chunking Support:

  • Document chunking for RAG applications
  • Configurable max tokens per chunk (64-8192)
  • Merge peers option for undersized chunks

  • Download chunks as JSON

  • Additional Export Formats:

  • Document Tokens (.tokens.json)
  • RAG Chunks (.chunks.json)

  • Enhanced DocTags export

  • Advanced OCR Options:

  • Multiple OCR backends: EasyOCR, Tesseract, macOS Vision, RapidOCR
  • GPU acceleration support (EasyOCR)
  • Configurable confidence threshold (0-1)
  • Bitmap area threshold control

  • Support for 28+ languages

  • Table Structure Options:

  • Fast vs Accurate detection modes
  • Cell matching configuration

  • Structure extraction toggle

  • Image Generation Options:

  • Generate page images
  • Extract picture images
  • Extract table images

  • Configurable image scale (0.1x - 4.0x)

  • Performance/Accelerator Options:

  • Device selection: Auto, CPU, CUDA, MPS (Apple Silicon)
  • Thread count configuration (1-32)

  • Document timeout setting

  • New API Endpoints:

  • POST /api/convert/batch - Batch conversion
  • GET /api/convert/<job_id>/images - List extracted images
  • GET /api/convert/<job_id>/images/<id> - Download image
  • GET /api/convert/<job_id>/tables - List extracted tables
  • GET /api/convert/<job_id>/tables/<id>/csv - Download table CSV
  • GET /api/convert/<job_id>/tables/<id>/image - Download table image
  • GET /api/convert/<job_id>/chunks - Get document chunks
  • GET/PUT /api/settings/performance - Performance settings
  • GET/PUT /api/settings/chunking - Chunking settings
  • GET /api/settings/formats - List all supported formats

Changed

  • Settings Panel: Completely redesigned with all new options
  • Export Options: Enhanced with tabs for different content types
  • DropZone: Updated with format categories and batch mode support
  • Converter Service: Major refactoring for dynamic pipeline options

Fixed

  • Confidence score calculation now uses cluster-level predictions
  • Better handling of partial conversion success

0.0.4 - 2025-12-10

Added

  • OCR Support: Full OCR integration using EasyOCR
  • Support for 14+ languages
  • Force Full Page OCR option
  • Configurable OCR settings
  • Improved Confidence Calculation: Average confidence from layout predictions

Changed

  • Updated converter service to use configurable pipeline options
  • Enhanced settings panel with OCR options

0.0.3 - 2025-12-10

Added

  • Initial release of Duckling
  • Frontend Features:
  • Drag-and-drop file upload
  • Real-time conversion progress
  • Multi-format export options
  • Settings panel
  • Conversion history panel
  • Dark theme with teal accent
  • Responsive design

  • Animated transitions

  • Backend Features:

  • Flask REST API with CORS
  • Async document conversion
  • SQLite-based history
  • File upload management
  • Configurable settings

  • Health check endpoint

  • Supported Input Formats:

  • PDF, Word, PowerPoint, Excel
  • HTML, Markdown, CSV
  • Images (PNG, JPG, TIFF, etc.)

  • AsciiDoc, XML

  • Export Formats:

  • Markdown, HTML, JSON

  • DocTags, Plain Text

  • Developer Experience:

  • Comprehensive test suites
  • Docker support
  • TypeScript
  • ESLint configuration

Security

  • Input validation for file uploads
  • File type restrictions
  • Maximum file size limits
  • Secure filename handling