What goes into the weekly report
A weekly report pulls the last seven days of data from Apple Health and turns it into a single, readable narrative. The metrics that drive it are the ones the Apple Watch is best at:
- Sleep: total time, consistency, and a sense of how stable your nights have been.
- Heart rate variability (HRV): the overnight readings your watch records while you sleep, compared against your own rolling 60-day baseline.
- Resting heart rate: trend across the week, not a single day.
- Workouts: sessions logged on the watch, including duration, heart-rate zones, and distance where it applies.
- Active energy and exercise minutes: the load side of the recovery-load picture.
Cross-metric reading is where it gets interesting. The report does not just list these numbers, it connects them. If HRV dropped the day after a long Zone 2 ride, that shows up. If sleep was short on three of the seven nights and resting heart rate ticked up, the report names it.
Why a weekly cadence
Daily numbers are noisy. One bad night does not mean a trend, and one strong workout does not mean fitness. Monthly summaries lose the granularity, you cannot act on something from three weeks ago. A week is the smallest window where patterns become reliable, and it is also short enough that you can do something with what you read.
That is why the report covers seven days and ends on Sunday. The score for the closed week appears on its own at the start of the next one. The written analysis is generated when you ask for it, so a week you do not care about does not cost you a report.
How VitalBrief generates it
The app reads from Apple Health locally, builds a structured summary of the week, sends it through an encrypted proxy to an external AI service, and gets back a 200 to 350 word narrative. No name, email, Apple ID, or device identifier is added by the app. Under the provider's API terms, the summary is not used to train models and is deleted within 30 days.
The Wellness Score on top of the report is calculated locally on your device, not by the AI. It combines a recovery component (HRV and resting heart rate against your own baseline) with a training component (volume relative to your usual and to a population reference), and it carries its own confidence level, so a score built on thin data is downgraded rather than presented as certain. The AI is told the score and the matching commentary, and it cannot contradict either.
Wellness, not medical. The weekly report is built for context and self-awareness. It does not diagnose anything and is not medical advice.
