AI Health Insights
Personalized patterns and recommendations from your check-in history.
AI Health Insights
After you've built up a few days of check-in history, AVYCENNA's insight engine starts analyzing your data and surfacing patterns you might not notice on your own. There are two engines working together.
Two Engines, One Feed
The rule engine runs on every check-in, immediately. It applies deterministic logic to your data — thresholds, streaks, trends — and generates insights that are always available, even on day one.
Rules run server-side in milliseconds. You'll see rule-based insights the moment you complete a check-in.
When AI insights are enabled (the default), AVYCENNA sends a structured summary of your recent check-ins to Claude. Claude analyzes patterns across all your metrics together — something rule engines can't do — and generates personalized narrative insights.
AI insights are cached for 6 hours to avoid redundant analysis. A new check-in invalidates the cache.
Rule Engine Examples
The rule engine covers the most common and important patterns:
Low Sleep Warning — Triggered when your average sleep over the past 3 days drops below 6 hours. Sleep deprivation compounds quickly; this rule fires early.
Mood Trend — If your mood score has declined for 5 or more consecutive days, the engine flags a downward trend and suggests reviewing recent life events or medication changes.
Step Streak — When you've logged 7,500+ steps for 5 consecutive days, the engine celebrates your activity streak and notes the correlation with energy scores.
Low Energy + Low Mood Cluster — When both energy and mood average below 3 for the past 3 days, the engine flags a combined low-wellbeing pattern and recommends discussing it with your doctor.
Other rules include: focus decline, sleep improvement trends, high mood streaks, step count drops, and irregular check-in patterns.
AI Insights
Claude analyzes your last 30 days of check-ins holistically. It can surface:
- Cross-metric correlations — "Your energy scores are consistently 1 point lower on days following less than 6.5 hours of sleep."
- Weekly rhythm patterns — "Your mood scores are reliably higher on Saturdays and Sundays compared to weekdays."
- Recovery observations — "After your energy dip in the first week of this month, your scores have been steadily recovering."
- Positive reinforcement — "You've logged more than 8,000 steps on 18 of the last 30 days — a meaningful activity level for your age range."
Sample AI Insights
"Over the past three weeks, your focus scores average 3.8 on days when you sleep more than 7.5 hours, compared to 2.6 on shorter nights — a gap of 1.2 points. Prioritizing sleep appears to have a measurable effect on your cognitive clarity."
"Your mood has been trending upward over the past 10 days, from an average of 2.9 to 4.1. This is a meaningful improvement. If you've made any lifestyle changes recently, this data supports continuing them."
"Your step count dropped significantly in the second week of the month (average 3,200 steps/day, versus 7,800 the week before). This coincided with a 0.8-point dip in energy scores. Whether the reduced activity caused the fatigue or vice versa isn't clear from the data alone, but the correlation is worth noting."
Insight Structure
Every insight — whether from the rule engine or Claude — has the same structure:
| Field | Values | Description |
|---|---|---|
| Severity | positive, neutral, warning, critical | How urgent the finding is |
| Confidence | 0.0 – 1.0 | How strongly the data supports this insight |
| Title | Short string | The main finding in a sentence |
| Detail | Longer text | Explanation and context |
| Recommendation | Optional text | What to do about it |
Severity Levels
- Positive — a good trend worth reinforcing (green)
- Neutral — an observation with no clear valence (gray)
- Warning — a pattern worth paying attention to (yellow)
- Critical — something that warrants prompt attention from you or your doctor (red)
Insights are informational, not medical advice. A "critical" severity insight means the data pattern is significant — it does not replace clinical judgment. If you're concerned about anything in your insights, discuss it with your doctor.
Streak Detection
The insight engine also tracks behavioral streaks that go beyond the simple day-count on your home screen:
- High mood streak — 5+ consecutive days with mood ≥ 4
- Active week — 7 consecutive days with steps ≥ 7,500
- Sleep consistency — 7 consecutive days within 1 hour of your average sleep time
- Low energy streak — 4+ consecutive days with energy ≤ 2 (triggers a warning)
Privacy
All insight computation happens server-side. Your check-in data is sent to Claude's API over HTTPS and is not used for model training. Insights are stored in your account and are never shared with your doctor or anyone else without your explicit permission via the grant system.