What problem does it solve?
Help users see their week as a coherent whole by reading diary entries and check-ins, surfacing repeating emotional themes, frequent people/trigger mentions, and lightweight growth signals so insights are offered gently rather than as a report.
Core Features & Use Cases
- Triggered by a Sunday 20:00 Heartbeat, but only runs when there is sufficient diary data and the user hasn't opted out.
- Aggregates diary/YYYY/MM/YYYY-MM-DD.md and memory check-in data, performs semantic emotion grouping, repeated-theme detection (emotion/person/trigger), cross-week pattern matching, and lightweight growth-signal detection.
- Produces conversational summaries by default and can output JSON for agent consumption or Canvas-ready HTML on supportive platforms; includes caching (weekly_cache), downgrade rules for repeated refusals, and clear safety fallbacks to pause if the user becomes distressed.
- Real-world use: automatically send a friendly Sunday message highlighting a recurring worry and asking if the user wants to talk about this week's events.
Quick Start
Ask the agent to run a weekly reflection and present a brief conversational summary of this week's diary entries and check-ins.
