Repository navigation
v1.5.2
Changelog - Version 1.5.2
Release Date
November 22, 2025
Overview
Version 1.5.2 is a major stability and control release that fixes critical bugs in automatic processing behavior. The script now properly respects user configuration settings and provides three distinct behavior modes based on the autoDetectOnGeneration and autoRegenerate settings.
Critical Bug Fixes
1. Automatic Processing Bypass Fixed
Problem: Script was processing all chapters automatically when pasting multi-chapter stories, completely ignoring the autoDetectOnGeneration and autoRegenerate configuration settings.
Solution:
- Removed automatic multi-chapter batch processing loop from
onResponsehook - Added proper configuration checks before any automatic processing
- Script now respects all three behavior modes (see Configuration Modes below)
Impact: Users now have full control over when and how chapter summaries are generated.
2. Generation Counter Limit During Condensation
Problem: When condensing chapters, the script would hit the "Non-interactive generation limit reached (5 calls)" error because the generation counter wasn't being reset between different operations.
Solution:
- Reset
generationCounter = 0at the start ofautoRegenerateChanges()batch processing - Reset counter before Level 1 (normal) condensation
- Reset counter before Level 2 (aggressive) condensation
- Reset counter before Level 3 (emergency) condensation
Impact: Condensation now works reliably without hitting generation limits, even during aggressive multi-level condensation scenarios.
3. Wrong Chapter Generation on Story Paste
Problem: When pasting a story with 10 chapter breaks, the script would generate "Chapter 10" first using Chapter 1's text content.
Solution:
- Skip single-chapter generation in full auto mode when no fingerprints exist
- Let the automatic batch detection handle all chapters at once
- Only process single chapters when appropriate based on configuration mode
Impact: Pasted multi-chapter stories are now processed correctly with proper chapter numbering and content.
4. Chapter Numbering Off-By-One Error
Problem: Script was determining chapter readiness incorrectly, requiring N+1 breaks for chapter N to be ready.
Solution:
- Changed
chapterToSummarizecalculation fromchapterBreakCount - 1tochapterBreakCount - Logic now correctly recognizes: N breaks = N complete chapters, chapter N+1 in progress
- Chapter N is ready to summarize when the Nth break exists
Impact: Chapters are now detected and processed at the correct time based on break count.
5. Incomplete Chapters Being Detected
Problem: When no fingerprints existed, the script would detect chapters that were still in progress (not yet complete).
Solution:
- Modified
detectChangedChapters()to only detect complete chapters when no fingerprints exist - Set
completeChapterCount = chapterBreakCount(not total chapter count) - In-progress chapters are no longer treated as "new" chapters
Impact: Only complete chapters trigger detection and generation, preventing premature processing.
6. Lorebook Entry Format Inconsistency
Problem: generateChapterSummary() used Format B while regenerateChapter() used Format A, causing inconsistent lorebook entries.
Format A (old):
Chapter 5: Title Here
Type: chapter
Summary: Text here
Format B (standardized):
Chapter 5
Type: chapter
Title: Title Here
Summary: Text here
Solution:
- Standardized both functions to use Format B
- Fixed variable name typo in
regenerateChapter()(messages → message) - All lorebook entries now have consistent structure
Impact: All chapter summaries have uniform formatting regardless of how they were created.
7. Chapter 2 Title Extraction Bug
Problem: Chapter 2's lorebook entry would display Chapter 1's title instead of its own title (e.g., showing "Starship Hijinks" instead of "Bridge").
Root Cause: The isFirstChapter flag wasn't being cleared after chapter 1 was created via the auto-regeneration code path (regenerateChapter()), causing scanForPreviousChapter() to always return Chapter 1's text even when generating Chapter 2.
Solution:
- Added universal flag clearing logic to
storeChapterFingerprint() - When chapter 1's fingerprint is stored, set
isFirstChapter = false - This fix applies to both code paths:
generateChapterSummary()andregenerateChapter()
Impact: All chapters now extract their correct titles regardless of which code path created them.
Configuration Modes
Version 1.5.2 properly implements three distinct behavior modes:
Mode 1: Full Manual (Both OFF)
autoDetectOnGeneration: falseautoRegenerate: false
Behavior: Nothing happens automatically. User must click "Check for Changes" button to detect and regenerate chapters manually.
Exception: First chapter of a brand new story is still auto-generated when user adds the first chapter break.
Mode 2: Semi-Automatic (Detect ON, Regenerate OFF)
autoDetectOnGeneration: trueautoRegenerate: false
Behavior:
- Script automatically detects changed/new chapters after each generation
- Shows notification modal with list of detected changes
- User manually clicks "Regenerate" or "Dismiss" for each chapter
Mode 3: Fully Automatic (Both ON)
autoDetectOnGeneration: trueautoRegenerate: true
Behavior:
- Script automatically detects changed/new chapters after each generation
- Automatically regenerates all detected changes in batch
- No user intervention required (fully hands-off)
Technical Improvements
Code Quality
- Removed obsolete
multiChapterBatchInProgressflag - Cleaned up redundant configuration checks
- Improved error handling and logging throughout
- Better separation of concerns between detection and generation
Timing & Logic
- More precise chapter completion detection
- Proper sequencing of automatic operations
- Better handling of edge cases (pasted stories, mixed scenarios)
Debugging
- Enhanced debug logging for flag state changes
- Better visibility into which code paths are executing
- Clear log messages for configuration mode detection
Migration Notes
No breaking changes - Existing configurations will continue to work. However, users may notice different behavior if they previously relied on the buggy automatic processing that bypassed configuration settings.
Recommended Action: Review your autoDetectOnGeneration and autoRegenerate settings to ensure they match your desired workflow (see Configuration Modes above).
Testing Coverage
All fixes have been tested with:
- ✅ Manual mode (both settings OFF)
- ✅ Semi-automatic mode (detect ON, regenerate OFF)
- ✅ Fully automatic mode (both ON)
- ✅ Pasted multi-chapter stories (10+ chapters)
- ✅ Writing chapters incrementally from scratch
- ✅ Mixed scenarios (existing chapters + adding new ones)
- ✅ Condensation triggering during various workflows
- ✅ Title extraction for all chapters
- ✅ Format consistency across all entry types
Known Limitations
None identified in v1.5.2 testing. All critical bugs from v1.5.1 and earlier have been resolved.
Credits
Special thanks to the testing that identified these edge cases and helped refine the three configuration modes into truly distinct, predictable behaviors.
Files Changed
Automatic_Chapter_Summaries_v1.5.2.ts- Main script file with all fixes appliedFEATURE_EXPLANATION_AUTO_VS_MANUAL.md- Documentation of the three behavior modesCHANGELOG_v1.5.2.md- This changelog
Upgrade Instructions
- Replace your current script with
Automatic_Chapter_Summaries_v1.5.2.ts - Review your configuration settings (
autoDetectOnGenerationandautoRegenerate) - Read
FEATURE_EXPLANATION_AUTO_VS_MANUAL.mdto understand the three modes - Test with your current story to verify behavior matches expectations
Version 1.5.2 represents a major leap forward in reliability and user control. Enjoy the bug-free experience! 🎉