Skip to main content
Script: scripts/data_migrations/relabel_oura_hrv_sdnn_to_rmssd.py

​
Problem

Oura reports RMSSD-based HRV, but the ingestion path historically stored those values under the SDNN series type (id=3) instead of RMSSD (id=7). The numeric values are correct — only the metric label (and therefore its unit semantics) was wrong. The ingestion path was fixed separately in backend/app/services/providers/oura/data_247.py, so new Oura HRV samples are now stored as RMSSD. This script corrects pre-existing rows in data_point_series and data_point_series_archive.

​
What it does

For provider='oura' rows only, it relabels every HRV series from SDNN (id=3) to RMSSD (id=7). The scope is strict — other providers’ genuine SDNN data (e.g. Apple, Ultrahuman) is left untouched. Both data_point_series and data_point_series_archive are updated.

​
Conflict handling

After the ingestion fix, a re-sync can write a correct RMSSD row at the same key as an old SDNN row, which would collide with the unique constraint on relabel:
  • data_point_series — unique on (data_source_id, series_type_definition_id, recorded_at)
  • data_point_series_archive — unique on (data_source_id, series_type_definition_id, bucket_start_at, aggregation_type)
In that case the stale SDNN duplicate is deleted (the RMSSD row already holds the same Oura value) rather than relabeled.

​
Idempotency

Once relabeled, rows no longer match the SDNN filter, so re-runs are no-ops. The script is wired into container startup (scripts/start/app.sh) and runs automatically on every deploy until removed.
This script is scheduled for removal after 2026-09-01, once all deployments have migrated.

​
Usage

​
Options

​
Dry run output

  • Commit 3bcf2973 — the ingestion fix (oura/data_247.py) that stores new Oura HRV as RMSSD.
  • Coverage Matrix — Oura HRV is listed under heart_rate_variability_rmssd.