@mmmbuto/nexuscrew 0.8.35 → 0.8.36

package/CHANGELOG.md

@@ -2,6 +2,39 @@
 
 All notable changes to NexusCrew are tracked here.
 
+## 0.8.36 — 2026-07-24 — "Clean Relaunch"
+
+- Makes macOS service and Fleet LaunchAgent reinstalls wait until `launchd`
+  has actually removed the previous job before bootstrapping the replacement,
+  preventing intermittent unload/bootstrap races during upgrades.
+- Bounds the unload check to five seconds and fails closed instead of leaving a
+  partially applied activation when a successfully stopped job never
+  disappears.
+- Keeps Settings service regeneration non-activating on every platform: the
+  unit is rewritten atomically, but applying it still requires the explicit
+  service restart already shown by the UI.
+- Adds opt-in, browser-local spoken alerts for new live PWA notifications.
+  Enabling the setting runs an honest native delivery test and speech remains
+  gated on a successful test for the current page session; unsupported, silent
+  or failing engines degrade without affecting the visual toast or SSE channel.
+- Speaks only from the visible, focused NexusCrew window and cancels on blur,
+  background, opt-out or unmount. Duplicate reconnect frames are suppressed
+  for 60 seconds, the normal queue keeps at most the two newest pending alerts,
+  high urgency preempts it, and a 30-second watchdog advances past a stuck
+  browser utterance.
+- Leaves Web Push unchanged as the best-effort hidden/closed-app path. Speech
+  uses the device voice and current UI language without a server audio route;
+  credential-shaped values and private home paths are redacted first. The
+  preference is off by default, works under server READONLY and is synchronized
+  across windows for the same browser origin.
+- Documents the one-speaker boundary: focus prevents duplicate speech across
+  windows on one device, while separately opted-in devices can each speak by
+  design. Persisted questions are never replayed; only new live notification
+  frames enter the speech manager.
+- Gate: **1,059 isolated Node tests** (1,058 pass / 1 platform skip),
+  **293/293 frontend tests**, production PWA build (168 modules) and zero
+  production dependency vulnerabilities in both dependency trees.
+
 ## 0.8.35 — 2026-07-23 — "Contained Fleet"
 
 - Keeps Fleet rows inside the Settings panel on phones and narrow windows.

package/README.md

@@ -271,6 +271,34 @@ draft, size preference and bounded prompt history in the current browser, includ
 ArrowUp/ArrowDown recall at textarea boundaries. This browser-local state is not federated or
 included in Fleet backups and can be cleared from Settings → System.
 
+## Notifications and optional spoken alerts
+
+Live cell notifications appear as in-app toasts and can also use Web Push when that browser and
+operating system support it. Push delivery remains best-effort and requires the browser permission;
+on iOS it also requires an installed PWA. The service worker cannot synthesize speech, so a hidden
+or closed PWA uses only its normal system-notification path.
+
+**Settings → System → Read notifications aloud** adds browser-local speech for new live
+notifications. It is off by default, remains editable when server settings are read-only, uses the
+device's system voice in the current UI language, and sends no notification text to a speech
+service. Enabling it runs an audible test. NexusCrew reports success only after the browser emits
+both speech start and end; if the native engine stays silent or errors, the UI reports that voice
+is unavailable. Credential-shaped values and private home paths are redacted before speech. The
+test must succeed once per page session, including after a reload.
+
+Speech runs only while that document is visible and has operating-system focus. Blur, background,
+opt-out and unmount cancel the current utterance and clear pending work; skipped notifications are
+not replayed when focus returns. Duplicate live frames are suppressed for 60 seconds, at most two
+normal alerts remain pending, and a high-urgency alert interrupts the current queue. A 30-second
+watchdog cancels a stuck browser utterance and advances to the next alert.
+
+Focus provides one speaker among NexusCrew windows on the same device. Two separately opted-in
+devices may both speak when each PWA is visible and focused; this is intentional for a per-device
+preference. Spoken alerts are optional and do not replace the toast's accessibility status, so
+screen-reader users can leave speech disabled if their assistive technology already announces it.
+Depending on the operating system, an enabled Web Push notification may still coexist with the
+visible toast and spoken alert.
+
 ## Connect nodes through SSH
 
 Every installation starts as a local node. A node joins another NexusCrew installation with a
@@ -545,7 +573,7 @@ See [CHANGELOG.md](CHANGELOG.md) for released changes.
 
 ## Status
 
-The current stable release is **v0.8.35** on npm and GitHub.
+The current stable release is **v0.8.36** on npm and GitHub.
 
 ## License