Import step history
Bring in an export file, check how each row will be read, then save it as one undoable batch.
Opening your local journal…
Checked before anything is saved
Every count must be a whole number of zero or more and below a plausibility bound; anything else is listed as rejected with the reason. Dates such as 03/04/2026 are ambiguous, so you choose day-first or month-first explicitly. Timestamps without a time-zone offset are read in the zone you pick, with daylight-saving changes handled.
One rule for overlaps
A daily total always covers the whole day, so sessions on the same day are not added to it. Sessions that overlap in time are resolved by your source ranking. Sessions that do not overlap are added. The original file can be kept on this device for re-checking, and deleted separately later.
Questions
Which files can I import?
Any CSV (comma, semicolon or tab separated) or JSON file with one record per row: a date or start/end times plus a step count. Many phone, watch and spreadsheet exports look like this. The file is read in your browser and never uploaded.
How are duplicates detected?
By provenance, not by matching numbers. A row is a duplicate when the same source already has a record for that date or interval (or the same record ID). Two different days with the same count are not duplicates.
What if my watch and phone exports both cover the same day?
Both are kept, but only the source ranked higher in your source list is counted for that day. The ranking table shows every day where this happens and which record was set aside.
Can I undo an import?
Yes. Each import is one batch. Undoing it removes exactly the records and corrections it added and leaves everything else alone.
My file has hourly rows. Will they be added up?
Only if you tick the option that says the rows are partial-day buckets from one device. Otherwise repeated dates are flagged, because adding rows that overlap would double count.