What your sources measured
Distance, duration, average pace, heart rate, elevation, date, and source are displayed as supplied. Missing fields remain unavailable.
How it works
Four kinds of information. Four clearly defined jobs.
Gaman separates facts, calculations, statistical estimates, and optional language-model guidance so you can understand where every recommendation comes from.
Transparent training logic
Weekly mileage, recent-versus-baseline load, race timing, tapering, and workout adjustments use deterministic rules that can be explained in plain language.
Feedback-informed difficulty
After at least five rated runs, an inspectable linear regression uses distance, pace, elevation, and your feedback to estimate workout difficulty. Sparse data falls back safely to rules.
Expanded recovery education
When configured server-side, a language model can expand conservative recovery guidance. It does not control training calculations, diagnose injury, prescribe medication, or replace a clinician.
Safety gate
Urgent symptoms bypass ordinary guidance.
Chest pain, breathing trouble, severe pain, new numbness or weakness, deformity, or inability to bear weight triggers a prominent recommendation for prompt in-person care. If the optional service is unavailable, built-in conservative guidance remains available.
Current limits
- No diagnosis or emergency care
- No fabricated cadence, sleep, HRV, splits, or grade-adjusted pace
- No confidence interval when the data cannot support one
- Coros production authentication remains a roadmap item
- JSON storage is local and not a durable production database