- DB: tasks table with scheduled_datetime, reminder_minutes_before, advance_notified, status
- API: CRUD routes GET/POST /api/tasks, PATCH/DELETE /api/tasks/<id>
- Scheduler: check_task_reminders() fires advance + at-time notifications, tracks advance_notified to prevent double-fire
- Bot: handle_task() with add/list/done/cancel/delete actions + datetime resolution helper
- AI: task interaction type + examples added to command_parser
- Web: task list page with overdue/notified color coding + new task form with datetime-local picker
- Nav: replaced Templates with Tasks in bottom nav
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Web client: trustDevice now defaults to true so a refresh token is always
issued on login, preventing deauth after the 1-hour access token expiry.
Users can still uncheck the box on shared devices.
Bot: cache file path is now env-configurable (BOT_CACHE_FILE) and
defaults to /app/cache/user_cache.pkl. Docker Compose mounts a named
volume at /app/cache so the session cache survives container restarts.
saveCache() now creates the directory if it doesn't exist.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Fix TIME object vs string comparison in scheduler preventing adaptive med
reminders from ever firing (#12, #6)
- Add frequency filtering to midnight schedule creation for every_n_days meds
- Require start_date and interval_days for every_n_days medications
- Add refresh token support (30-day) to API and bot for persistent sessions (#13)
- Add "trusted device" checkbox to frontend login for long-lived sessions (#7)
- Auto-refresh expired tokens in both bot (apiRequest) and frontend (api.ts)
- Restore bot sessions from cache on restart using refresh tokens
- Duration-aware routine scheduling conflict detection (#11)
- Add conflict check when starting routine sessions against medication times
- Add diagnostic logging to notification delivery channels
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Create knowledge.py handler with dynamic book selection
- Support list/select/query actions for multiple books
- Implement vector search with cosine similarity
- Add knowledge detection to AI parser config
- Cache embeddings per-book for performance
- Enhanced AI prompts with time/frequency conversion rules and 20+ examples
- Added smart medication name resolution and confirmation flows
- New medication commands: today, refills, snooze, adherence
- New routine commands: create_with_steps, add_steps, steps, schedule
- Added active session awareness with shortcuts (done, skip, pause, resume)
- Confirmation handling for destructive/create actions
- Improved help message with natural language examples
