A bugfix release. If HACS could not download v2026.10.8 for you: that was a change on GitHub's side, reverted since, and updates work again (#933).
Fixed
- An evapotranspiration provided by Open-Meteo did not follow its hours. Since v2026.10.2, a zone whose ET comes from the weather service was meant to sum Open-Meteo's own hourly ET0 when the hourly calculation is on. The check that should have stepped aside only for an ET from a sensor stepped aside for the weather service too, so those zones kept the daily figure spread evenly, and their live bucket fell in a straight line, day and night alike. They follow the day now. Thanks @Megalos for the graphs (#878).
- The live bucket dropped to last night's value after a restart. It is meant to come back with the estimate it showed, as long as no calculation ran since; a comparison between two ways of writing the same time never matched, so it fell back after every restart until the first new estimate. Thanks @Megalos (#869).
Clarified
- Continuous updates work in a sensor group that also uses a weather service. The group's own sensors are recorded at every change, a rain gauge included, while the service is still read once per update, so it costs no extra requests. The setting said it "cannot be used" there, which had not been true for a while. Thanks @Megalos (#935).