Release Notes v0.13.0
Reading Features: X-Ray, Recap, and Analyze Highlights
Three new book-level actions that leverage your reading position, highlights, and book text (By request, with help from implementation in https://github.com/omer-faruq/assistant.koplugin):
- X-Ray: Structured reference guide (characters, locations, themes, timeline) spoiler-free up to your current position
- Recap: "Previously on..." style summary to refresh your memory when returning to a book
- Analyze Highlights: Discover patterns and connections in your annotations
New context extractor supports:
- EPUB text extraction via XPointers
- PDF text extraction page-by-page
- Highlights and annotations from your document
- Reading progress, chapter info, and time-since-last-read stats
New placeholders for custom actions:
{reading_progress},{progress_decimal}- current position{highlights},{highlights_section}- document highlights{annotations},{annotations_section}- highlights with notes{book_text},{book_text_section}- extracted book content{chapter_title},{chapters_read},{time_since_last_read}
Access: Settings → Reading Features (visible when document is open), or bind to gestures via Dispatcher
Settings: Book text extraction is off by default. Enable in Advanced → Book Text Extraction with configurable limits (max characters, max PDF pages).
Reliability
- Metadata fallback: When book metadata is missing, filename is used as title fallback (e.g.,
Crime_and_Punishment.epub→ "Crime and Punishment") - Safety instructions: X-Ray, Recap, and Analyze Highlights now ask AI to admit uncertainty rather than hallucinate when book context is unclear
Translation Improvements
New translation tooling (work in progress):
scripts/translate.py: Unified CLI for extraction and applicationscripts/validate_translations.py: Validates placeholders, brand names, duplicates
Quality fixes: Re-translated German, Portuguese, and Japanese
Web Inspector Enhancements
- Action testing system - Select any action to load its prompt template with placeholders resolved. Edit prompts live to test variations.
- Action overrides panel - Dynamically loaded checkboxes for all action flags (
skip_language_instruction,skip_domain,use_book_text,use_highlights,use_annotations,use_reading_progress,use_reading_stats) - Sample data viewer - "View Sample Data" button shows the fake book/highlight data being used for testing context-aware actions
- Expanded editor modal - Click ⤢ on any textarea to open a full-screen editor with placeholder picker
- Placeholder picker - Insert placeholders by category (Standard, Reading Progress, Context Extraction, Reading Stats) directly into prompts
- Max tokens slider (256-16,384) for controlling response length
- Per-output copy buttons - Copy request body, response, or chat content independently
- Reset button to restore all settings to defaults
- Sample context now properly merges based on action extraction flags (fixes X-Ray/Recap testing)
Bug Fixes
- Settings reset functions now correctly reset all settings (context extraction, reasoning settings, poll interval)
- Web inspector sample context properly merges for book-level actions
Documentation
- README reorganized with improved structure and navigation
- Test suite documentation updated (clone requirement, web UI features)
Full Changelog: v0.12.0...v0.13.0