What changed: The in-app feedback survey mechanism was refactored for improved accuracy and maintainability
Key improvements:
- Switched from tracking user message counts to tracking actual submission counts, preventing inconsistencies from message editing or restoration
- Moved survey-specific state from mixed local/global storage to centralized app-level state management
- Simplified conditional logic with early returns for better maintainability
- Adjusted triggering configuration:
- Enabled for all models (was disabled by default)
- Reduced probability from 100% to 0.5% when conditions are met
- Reduced turn requirement between surveys from 15 to 10
- Effectively disabled global cooldown (increased from 1 hour to ~1157 days)
- Removed per-conversation cooldown (30 minutes)
Net effect: Surveys appear less frequently but with more accurate tracking across sessions
Evidence
Function Ky2() at line 445844 (renamed from mS2), configuration ku5 at line 445950, state structure at line 298044
Strings lifted out of the shipped bundle, so the claim above can be checked against them.