The AI Time Agent is an autonomous scheduling and accountability system built to do more than manage time it stewards it. Powered by a finite-state machine (FSM) and APScheduler jobs, it integrates Google Calendar for real-time event governance, PostgreSQL for persistent context, and an LLM tool-calling layer for intelligent intervention.

Through a conversational Telegram interface, the agent guides each segment of your day from proactive reminders and midpoint check-ins to post-event recovery nudges while dynamically adjusting its tone (Gentle → Coach → Drill Sergeant) based on recent behaviour. Current live capabilities include Live Session Stewardship, Proactive Time Reminders, Conversational Tone Handling, Covey Quadrant Classification, Evening Reviews (21:30 daily, Europe/London), Weekly Audits (Sundays at 21:00), and partial Real-Time Accountability follow-ups for missed Quadrant II commitments.

Planned modules on the roadmap expand this foundation into a full 18-capability personal operating system, aligning work, wellness, and values into a unified daily rhythm.

• Designed the 18-module capability roadmap and implemented the current operational set.

• Built the FSM and tone escalation engine with cooldown logic and state-based messaging.

• Created tool-calling prompts for calendar edits, reschedules, and coaching interactions.
• Integrated Google Calendar API with APScheduler for timed daily/weekly flows.
• Developed weekly audit analytics to visualize quadrant distribution and provide corrective suggestions.

• Dual-path integrity — Keeping the calendar authoritative when deterministic APScheduler jobs and AI-initiated adjustments both operate in real time.
• FSM stewardship design — Managing both planned events and opportunistic free-time windows with tailored prompts, pauses, and recovery flows.
• Tone escalation without fatigue — Implementing behavior-driven progression (Gentle → Coach → DS) with a 30-minute cooldown to preserve trust and motivation.
• Quadrant-aware interventions — Building lightweight NLP rules to classify commitments into Covey Quadrants and link them to coaching and audit insights.
• Early-stage accountability loop — Designing follow-ups that encourage recovery without overwhelming the user.

• Database persistence is enough — PostgreSQL event/segment logging delivers contextual nudges without the overhead of Redis for this scale.
• Tone cool-downs matter — Regulating escalation frequency keeps the agent constructive, not punitive.
• End-of-day and weekly audits drive change — Quadrant breakdowns and reflection questions shift behaviour more than raw stats alone.
• Manual overrides + keyword tagging — Explicit #q1/#q2/#q3 tags offer reliable quadrant overrides and make NLP classification more accurate.