You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Energy meter undercounting (severe). The parser was reading kWh from consumption.values.quantity (a DPI/chart-scaled helper), not consumption.quantity (the real meter read). This undercounted the Energy
dashboard by 4-73% per day with no consistent ratio. Confirmed against the
AGL portal "MyUsageData" CSV across 11 mitm captures, 2026-05-12. Cost
values were already read correctly from consumption.amount and are
unaffected. Anyone running v0.1.0 / v0.2.0-beta.{1,2,3} should wipe the haggle:* rows from statistics / statistics_short_term after upgrading
and let the 30-day backfill rebuild from scratch.
AGL placeholder days no longer create phantom zero rows. On days where
AEMO hasn't delivered the meter reads yet, AGL returns intervals with a
non-none type but quantity == 0 && amount == 0. The parser now drops
these instead of writing 24 zero-kWh hourly rows that the resume logic
would never re-check.
Trailing rewindow self-heals AGL backfills. Every poll now re-fetches
the last REWINDOW_DAYS (default 7) so a slot first returned as a
placeholder is overwritten when AGL has the real read. The cumulative-sum
baseline for the rewindow is looked up via statistics_during_period
(sum at the hour right before fetch_start UTC midnight), not the latest
stored sum.