kaafil-react-uikit 0.11.2 → 0.11.4

package/CHANGELOG.md

@@ -6,6 +6,49 @@ The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this package uses [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 While on `0.x` a minor bump may carry a breaking change; the entry will say so.
 
+## [0.11.4] — 2026-09-15
+
+### Changed
+
+- Peer floor moves to `kaafil-js@^0.7.0`, which coalesces identical in-flight
+  GETs at the transport. 0.11.3 fixed the duplicate capability read in this
+  package; the SDK change covers the same shape across every read hook, which
+  is where it belongs — `execute()` is one funnel, this package has thirty-odd
+  hooks. `core/internal/coalesceRead.ts` stays as the in-package seam for reads
+  that do not go through a resource method.
+
+## [0.11.3] — 2026-09-15
+
+### Fixed
+
+- **The manager's Now tab asked the same question four times.** A hook is an
+  implementation shared by its callers, not an instance shared by them: every
+  component calling a read hook gets its own state, its own effect and its own
+  request. Four components on that tab each needed `journey/capabilities`, so
+  four identical requests went out together.
+
+  `core/internal/coalesceRead.ts` is now the one place this package collapses
+  identical concurrent reads into one. Whoever asks first issues the request,
+  anyone asking while it is in flight awaits the same promise, and the entry is
+  dropped the moment it settles — success or failure, so a rejection is never
+  replayed to a caller who would have retried.
+
+  It is deliberately NOT a result cache: nothing survives the promise settling,
+  so this changes only how many requests a single moment produces, never how
+  fresh an answer is or when a refetch happens. Caching results would move
+  staleness decisions out of the hooks that own them.
+
+  Measured against a live engine, the tab's opening burst: **10 requests →
+  7, wall 2213–3552ms → 1845–2027ms**. The saving is larger than three
+  requests' worth because they were also competing with each other.
+
+- **Two more empty time columns.** `ManagerNowSection`'s timeline and
+  `ManagerOverviewSection`'s itinerary preview both reserved a fixed,
+  non-shrinking time column and rendered it empty when no block carried a
+  `startTime` — the same defect 0.10.2 fixed on the traveller surface, which
+  that release's note predicted might exist elsewhere. It did, twice. A sweep
+  of every fixed-width non-shrinking column in the package found no others.
+
 ## [0.11.2] — 2026-09-15
 
 Needs an engine build carrying the widened `deleteFile` rule; against an older