DLE version: (e.g., v1.0.0-beta)
SillyTavern version: (e.g., 1.18.x)
Browser: (Chrome 150.0.7871.182 )
Connection mode: (profile)
Obsidian REST API version: (Local BEARER)
Multi-vault configuration: (single vault)
AI features enabled: (Librarian / Scribe / Auto-Suggest /)
What happened?
The system has stopped flagging anything. I need to ask Emma to read the chat at the end of the night. She invariably finds 2-4 major flags that didn't trigger, and several smaller files that need to be cleaned up or corrected. It usually takes thirty minutes to an hour for her to clean everything. No exaggeration.
Emma explained it as: The threshold for flagging seems calibrated for factual discrepancies, not for new information density. When a single scene drops five vault-relevant character insights and the system stays silent, that's a detection gap.
What did you expect?
I expect major changes to be flagged without needing to prompt Emma to read everything after each chat. Something has degraded on the detection side.
Steps to reproduce
No idea.
Diagnostic report (recommended)
dle-diagnostics-2026-07-24T02-44-52-731Z.md
Console errors
Additional context
DLE version: (e.g., v1.0.0-beta)
SillyTavern version: (e.g., 1.18.x)
Browser: (Chrome 150.0.7871.182 )
Connection mode: (profile)
Obsidian REST API version: (Local BEARER)
Multi-vault configuration: (single vault)
AI features enabled: (Librarian / Scribe / Auto-Suggest /)
What happened?
The system has stopped flagging anything. I need to ask Emma to read the chat at the end of the night. She invariably finds 2-4 major flags that didn't trigger, and several smaller files that need to be cleaned up or corrected. It usually takes thirty minutes to an hour for her to clean everything. No exaggeration.
Emma explained it as: The threshold for flagging seems calibrated for factual discrepancies, not for new information density. When a single scene drops five vault-relevant character insights and the system stays silent, that's a detection gap.
What did you expect?
I expect major changes to be flagged without needing to prompt Emma to read everything after each chat. Something has degraded on the detection side.
Steps to reproduce
No idea.
Diagnostic report (recommended)
dle-diagnostics-2026-07-24T02-44-52-731Z.md
Console errors
Additional context