Skip to content

Grid Power card release — 26 October 2026

Server release accepted at 22:21 UTC. Native app distribution is tracked separately below.

The Current Conditions section of Power gains a Grid Power card above its charts on desktop and mobile web. Native apps receive the same card through the optional currentSignals field on Current, Observed and All Power responses. Forecast responses are unchanged. App build and store availability remain separate from the server release.

The card reports signed GridWatts, the recorded AC input measurement. Positive input is green; zero and negative readings remain neutral and do not claim that the grid is disconnected. It requires its own valid, nonfuture measurement time. Missing or invalid values, and samples older than 30 minutes, show Unknown. The browser clock updates age and expires cached readings locally, including while disconnected and after returning to the foreground.

The existing cloud operations collector reads the final GridWatts and MultiPlusDCWatts row together from its existing local power product. It neither fills gaps from older finite readings nor queries field equipment. Failure of either optional field leaves a valid sibling measurement intact. API and browser card reads use the cached operations JSON; no polling cadence or device control changes are part of this release.

Candidate and verification

Dashboard PR #35 merged as 7399c08a60e37fa2b7b48b21369b1bfb6ac54c1d, with an identical source tree to candidate 7f1fc05ecc2956fef19164b155f64e1d8851892d. Development acceptance passed before the annotated production tag prod-20261026.3 was pushed at that exact revision. CI run 40409496192 passed 812 tests and 55 subtests, public-module import checks and the strict documentation build.

Local checks passed 114 web/display/shell/session tests and two subtests. The actual ReactiveHTML scripts were exercised for positive, zero and negative grid power, charger behavior, expiry without network access, foreground events, and repeated-render/removal cleanup. After isolating test view state, all 20 grid and browser-clock tests passed again. Collector/API regression coverage verifies the single cached-product open, final-row gaps, own timestamps, invalid values and optional-field failure isolation. All 231 focused checks and two subtests passed again on the merged revision.

Local browser acceptance of the candidate passed through a proxy using the exact deployed content security policy. Both desktop and mobile Current Power rendered the styled Grid Power card above the charts, showing 1.93 kW and its own sample age. Desktop Forecast hid the card; returning to Current restored it. After stopping only the local Panel test server, both disconnected pages changed to Unknown with Last sample 30m ago at the next local clock tick. The test servers and tabs were then removed. Neither public host nor field equipment was interrupted for this check.

Deployment acceptance

The focused development release completed with 23 tasks successful, three changed and no failures. Its actual browser rendered the styled card, existing charts and Development live-mirror banner without browser errors or warnings. The public manifest reported schema 3, the exact merged revision and live-mirror provenance. currentSignals correctly returned Unknown with a null sample before the primary collector update reached the snapshot.

The staged helper read only the existing development Zarr, without invoking the full collector: GridWatts was 1838 W and MultiPlusDCWatts was 1679.58 W at 21:56:03.244806 UTC. The shared card returned 1.84 kW with an age of 16 minutes. This checked the data path without an extra device request or collection run.

The combined primary presentation/collector release completed with 29 tasks successful, five changed and no failures. The operational API-only mirror release completed with 11 tasks successful, two changed and no failures; its nginx task was not applied. Primary desktop/mobile and development desktop rendered the fresh Grid Power card above their charts without browser errors.

At 22:21 UTC the primary, development and isolated operational mirror APIs all reported 1.88 kW from the same sample at 22:15:28.226902 UTC. Their manifests reported the release revision, using the full SHA on the ordinary endpoints and its short form on the isolated mirror. The API response evidence is retained with the local release packet as primary-api.json, development-api.json and operational-mirror-api.json.

The primary collector's normal 22:17:00–22:17:19 UTC run exited successfully. Its snapshot, timestamped 22:17:01.115014 UTC, contained 1884 W GridWatts and 1701.290039 W MultiPlusDCWatts, both stamped 22:15:28.226902 UTC and matching the raw product's final row. Normal development replication succeeded at 22:17:54 UTC and carried those fields to both mirror APIs. No additional collector run was started.

The release used the focused presentation role on both hosts, the primary-only operations collector role and the existing operational mirror role. It retained /opt/aurora-presentations/runtime-20261024, environment and output paths, and the existing *:01/5 collector timer. Only the reviewed reader/collector source pins changed. Scientific writer HEADs remained primary 84527ce46626e98961f8484d80dc42fa22fe9e2c and development 433cdfa9ab643f043f68ac7c76313946077d2d21. Field acquisition/settings, writer commands/runtimes, other timers, observation paths, network settings and nginx were unchanged.

Native app distribution

The corresponding native work is iOS 0.13.3 (build 30), PR #16, and Android 0.14.2 (version code 7), PR #8. At server acceptance, final release artifacts and store uploads remained pending. The iOS upload required the Mac to be unlocked; Google Play required sign-in to the existing owner developer account. Server acceptance does not establish that either new app build is available to testers.

Rollback

Restore both presentation reader pins to 71eef853ad4d5ca3617e8f856a269e393f187a0d. Restore the primary collector pin and isolated operational mirror source/revision to c900dd85e71d1a4f81331f92740cbc5049074bf3, using their existing focused roles and the same runtime. Preserve all observations and existing schedules; wait for the next scheduled collector run rather than starting an extra collection.