SomviaWhy daily summaries are not enough
Two runs can have the same distance, duration, and average heart rate while feeling completely different. One may be evenly paced; the other may contain four hard repeats and full recoveries. An average hides that shape.
For training questions, a useful agent should start with a compact workout list, open the relevant sessions, and fetch a downsampled time series only when necessary. Somvia exposes those layers separately so the model does not need every recorded point for every question.
Workout data available to an agent
| Layer | Possible fields | Example use |
|---|---|---|
| Session | Activity, start time, duration, distance, energy, pace, elevation | Find comparable workouts |
| Splits | Intervals, automatic kilometers or miles, laps, segments, average heart rate when derivable | Compare work reps and pacing consistency |
| Heart rate | Timestamped samples, downsampled for a bounded response | Inspect effort rise and recovery shape |
| Movement | Speed, route, power, cadence, or stroke rate when Apple Health records them | Relate output, terrain, and technique to effort |
The exact data varies by activity, watch model, workout app, sensors, and permissions. Somvia reports what Apple Health contains; it does not synthesize absent measurements.
Questions that use the deeper data
Intervals and pacing
- Compare the pace range and average heart rate of each one-kilometer work rep across my last four interval runs.
- Did I start too hard in the latest session, or did pace and heart rate build progressively?
Endurance and drift
- Compare the first and final thirds of my recent long runs at similar speed. Is heart rate drifting more or less?
- Which run held the steadiest cadence after the 60-minute mark?
Context across days
- How did sleep duration, HRV, and resting heart rate differ before my strongest and weakest workouts this month?
- After which sessions did my next-day resting heart rate rise the most?
A good prompt names the comparison and the evidence. Ask for specific sessions, ranges, or series. “Analyze my health” is too broad; “compare work-rep pace and recovery heart rate across four interval runs” gives the agent a tractable job.
How Apple Watch data reaches the agent
-
Apple Watch records the workout
The workout and available measurements are saved into Apple Health through Apple's Workout app or another compatible workout app.
-
Somvia reads authorized categories
On iPhone, you choose which Health categories Somvia may read. The app syncs selected daily history plus recent workout sessions and their available details.
-
Your agent queries the remote MCP
Connect Claude, ChatGPT, OpenClaw, Hermes, Grok, Muse, or another compatible client. The client chooses which Somvia tools and date ranges it needs.
Limits and responsible use
- The first sync reads back 180 days of workouts and available series, not your entire lifetime archive. Everything after that keeps syncing.
- Downsampled series preserve the useful shape while keeping MCP responses bounded; they are not a replacement for raw research exports.
- GPS routes are sensitive personal data. Review the Apple Health permissions and connected agents you trust.
- Somvia and the connected agent do not provide medical diagnosis, treatment, or emergency support.
Frequently asked questions
Can AI analyze Apple Watch workout intervals?
Yes, when Apple Health records the relevant splits or time series. Somvia can expose structured session details and available heart-rate, speed, power, or cadence series to a compatible agent.
Does Somvia read directly from Apple Watch?
No. Apple Watch contributes workouts and measurements to Apple Health on the iPhone. Somvia reads only the Health categories authorized on that iPhone.
Can an agent diagnose a health condition from workout data?
No. Somvia is not a medical device, and an agent's workout analysis is not a diagnosis. Use qualified medical care for symptoms, treatment, or clinical decisions.