You can now customize the Reader experience to your liking with settings for: - Typography, fonts, line height and width - Themes, light/dark appearance - Custom CSS Improved: - Reader: Comments on sites like Reddit and Hacker News now have a threaded interactive UI. - Reader: Added to context menu. - Translations: Khmer, Hungarian, Portuguese, Slovak. - Upgraded to Defuddle 0.15]. - New Settings UI to match Obsidian . - Safari: Fix copy to clipboard. - Safari: Improve recovery after extension upgrade. - Fix error message on browser extension store pages.

Upgrade to Defuddle 0.14

- Add {{language}} variable to get page language - Add {{transcript}} variable (only for YouTube currently) - Update to Defuddle 0.13

- Add Reader mode - Add transcripts to YouTube in the {{content}} and new {{transcript}} variable - Switch to Defuddle's built-in Markdown converter

Fix quote-stripping regex in filters to handle newline characters. Fixes the issue with using map filter in highlights example.

New - Add message if clipboard doesn't work - Add `decode_uri` filter Improved - String literal maps now return plain strings instead of {str: ...} wrapper objects. The template filter still supports ${str} for backwards compatibility. - Treat unquoted filter args as string literals when no matching variable exists, support unquoted date format in date filter #694 - Make `src` URLs on video & audio elements absolute Fixes - Fix for multiple replacements - Fix normalize ISO 8601 duration components correctly

New - New template logic for conditionals, loops, and variable assignment. - New template validation in the template editor checks for syntax errors. - Add context menu option to copy Markdown to clipboard. Improved - Update to Defuddle 0.7.0 with better support for X articles and more.

New - Import multiple templates at once Improved - Fix for null Schema.org data - Force opening via popup (not embedded mode) on mobile Safari - Fix issue with bad provider names and possible corruption of settings - Updated Japanese translation - Another attempt at fixing the default locales bug on Safari - Remove uses of innerHTML

New {{selection}} and {{selectionHtml}} variables for selected text Improved - Update to Defuddle 0.6.6 - Fix issue with srcset images (e.g. on Substack) - Fix Properties section remembering the collapsed state - Fix Interpreter context field not consistently appearing - Fix for temperature parameter issue - Wrap property names that contain special characters to ensure compliant YAML

This update mostly reverts back to the old highlighter behavior (before 0.11.6) due to some issues with the new approach. Improvements - Localization for shortcuts - Fix editing custom providers - Fix remove_attr does not remove the whole attribute if it includes spaces - Upgraded to Defuddle 0.6.4 with fixes for Substack and other sites

- Upgraded parsing to Defuddle 0.6.2. - If image filter input is an array the output is also returned as an array.

- Revamped highlighter mode, now works much more easily for highlighting text. Highlights are saved in a more robust format similar to text fragments URLs. - Added Link button in highlighter mode that allows sharing a URL that includes the highlighted text as a text fragment parameters. Improved - Fix getting extractors for certain sites, e.g. ChatGPT.

Upgrade to Defuddle 0.6.0, improves capturing websites, images, and adds better support for Gemini chats

- Added built-in support for saving ChatGPT and Claude chats - Improved code block and footnote extraction

- Added {{words}} preset variable for word count. Improved - Fixed primary button not triggering in side panel mode. - Upgraded to Defuddle 0.3.3 which includes many improvements for accurately getting the content of pages, with many improvements for code blocks and math. - Fixed indenting nested list items.

- Added new save behavior setting that allows default save action to be changed to copy to clipboard or save to file - Added support for converting callouts on GitHub - Added extractor for Hacker News threads - Added Perplexity as a preset provider Improved - Fixed missing domain variable - Upgraded Defuddle with fixes for code blocks

New - Added translations for Bengali, Czech, Danish, Finnish, Greek, Hebrew, Hungarian, Norwegian, Romanian, Swedish, Tagalog, Thai, Ukrainian, Vietnamese. Improved - Fixed Interpreter bug with newlines.

- Added new Overwrite note behavior for templates. - Added Activity chart to Settings page. Improved - Revamped Interpreter configuration with popular models and links to model lists. - Ollama 403 error text gives more useful instructions. - Ollama no longer requires API key. - Fixed issue with certain Anthropic prompts. - safe_name should not exceed maximum file name length.

New - Added Hugging Face and Deepseek API support for Interpreter - Added KaTeX support - Added Bengali and Catalan localizations Improvements - Longer context length for Ollama - Improved language detection for code blocks - Improved handling line breaks in code blocks - Improved handling footnotes and citations - Fixed capturing code blocks in WeChat articles - Fixed issue with URLs that contain x.com - Refined German UI strings

New - New duration filter to convert duration formats. - Added Turkish and Hindi translations. Improvements - Fixed issue with \n in replace filter, should now work as both as a search and replacement value. - Fixed Anthropic models not always using the right authentication. - Fixed extension window when current model has been disabled. - Fixed duplicate images in Twitter extractor. - Added YouTube extractor.

A new experimental approach has been added to improve results with the default template. Extractors override the Readability library for popular sites where Readability does not work well. This populates the {{content}} and {{contentHtml}} variables with more useful content. This approach doesn't replace the need for templates but it makes it easier to clip sites that have complex DOM structures that are not easy to parse with templates. So far Twitter (X) and Reddit have been added but PRs are welcomed for other sites. New - New unique filter returns a deduplicated array or object. - New remove_tags and remove_attr to remove only specific HTML tags and attributes (whereas strip filters remove all with exceptions). - Added Google Gemini preset provider. Improved - replace filter now supports regex. - split filter now splits letters in a string when no parameter is provided. - Fixed syntax inconsistencies for strip filters. - Fixed regression with Ollama. - Fixed "save" feature now saves to .md file instead of .txt. - Various improvements for mobile UX in Settings pages.

Improved - Added "Reset models" option in Interpreter - Fixed bug for checking version - Fixed Azure models - Improved migration to new Providers settings - Fixed detecting Chinese and Portuguese locales

Important: this update changes the way Interpreter providers are stored, to provide more flexibility with a variety of providers. Your existing configurations should be automatically migrated, but if they aren't you may need to update them to the new format. New - New Providers section in Interpreter settings with preset provider configurations - Show current version in General settings - New language options: Polski, Nederlands, Italiano - New replace_tags filter Improved - Fixed issue with some toggles - Fixed issues with enabling and deleting models - Improved support for Azure OpenAI models - Fixed localization of various text throughout the UI

New - Adds Interpreter features to extract text using natural language, using the syntax {{"a summary of the page"}}. Requires being enabled and configured in the Web Clipper settings. - New length filter returns a number of items or characters. - New calc filter allows basic mathematical operations e.g. {{number|calc:"/2"}} — can be combined with length filter to count words, or estimate reading time, e.g. {{content|split:" "|length|calc:"/200"|round}} minutes. - Added translations for Brazilian Portuguese, Bahasa Indonesia, Korean, and Traditional Chinese. Improved - Fixed issue with fonts in Japanese. - Fixed issue with {{publish}} variable sometimes returning multiple dates. - Export all settings now decompresses templates so they are human readable. Import all settings compresses them again to maximize sync storage space.

New - Highlights can now be saved inline by going to Settings → Highlights. - UI is mirrored for right-to-left languages. - Added translations for Arabic and German. - Added {{favicon}} to the preset variables to get favicon URL. Improved - Fixed issue with {{published}} and {{author}} variables under some conditions.