App preview · Synthetic sample data
Focus
Health & strength trends
Built with
Python · SQLite
Data sources
Apple Health · Gravl · Hevy · Hume
Approach
Local-first · Data ownership
01 / WHY I BUILT IT
I had the data.
I needed more.
I could see pieces of my training and health, but not the whole picture. Switching training apps meant losing the history I’d built up. Even when I stayed with an app, it didn’t always give me the analytics I wanted.
Apple Health could store most of the data, but it didn’t give me the analysis I was looking for.
So I built Training Atlas to bring that history together and explore it on my own terms. Apple Health provides the long view, while Gravl workouts and Hevy history add the detail of exercises, sets, and personal records. Hume provides body composition and sleep. There’s an event database to tag injuries, vacations, supplement changes, and more.
As long as your data source has an export, an API, or a screenshot it can be normalized and saved.
Your training has a history. Training Atlas helps me see the whole map. It runs locally; it isn’t a hosted service, and it remains in active development.
Follow the work
Explore exercise progression, personal records, training volume, and muscle-group trends.
Connect the context
See weight, body composition, activity, and logged nutrition alongside your training.
Keep the nuance
Missing food logs stay unknown. Exploratory correlations never claim cause and effect.
02 / INSIDE THE APP
Strength and context, side by side.
The dashboard connects strength progression with body composition, activity, and logged nutrition. These previews use synthetic sample data.
More than the last workout.
View progress over time, explore the exercises behind it, and see how training volume changes across muscle groups.

The gaps matter, too.
Food logging is an incomplete record. Training Atlas preserves that distinction, showing logged nutrition without assuming that an untracked day means no intake.

03 / HOW I’M BUILDING IT
A history I can keep.
I’m keeping the app local with a lightweight Python application and SQLite database. It isn’t a hosted service. Historical imports and immutable monthly snapshots preserve the records behind the trends. Conflicting records are flagged instead of silently rewriting history. Training Atlas is in active development, one iteration at a time.
