* feat(ui): add runtime English and Chinese localization
* fix(meshtastic): harden t-deck send and ack recovery
Align the T-Deck SX1262 setup with the reference firmware, switch critical radio SPI paths to blocking access, and share RX restart state between direct send paths and the radio task so TX-to-RX handoff cannot silently get stuck.
Also unify Meshtastic wire transmission through a single helper, keep wire packets for ACK-tracked sends, retry ACK timeouts up to three times, and document the issue #21 repair checklist.
Validation: local builds passed for tdeck, tlora_pager_sx1262, lilygo_twatch_s3, and gat562_mesh_evb_pro; clang-format check passed.
Testing status: not fully tested on hardware yet.
* Refactor LXMF runtime toward full Reticulum alignment
Split adapter runtime state into transport, link, and propagation domains.
Add pending path request tracking, stronger link teardown semantics, outbound link establishment, deferred link payload flushing, link keepalive/stale handling, and inbound split-resource assembly.
Document the authoritative Reticulum/LXMF runtime alignment plan and mark the older device-mode plan as historical context.
* feat(web): add pages site and web flasher release flow
- add shared LXMF wire codecs for app-data payloads, link requests/responses, resource advertisements, and propagation batches
- implement ESP LXMF app-data send/receive, active link sessions, resource chunking, propagation announces, and verified envelope delivery
- propagate want_response through team and hostlink flows, track team members for unicast fan-out when broadcast app-data is unavailable, and publish node position updates from team position events
- advertise broadcast app-data support on MeshCore/Meshtastic adapters and replace fixed-size Reticulum token buffers with dynamic storage
Revert b9e20b2662 because it causes app-side configuration reads to hang on GAT562.
This restores the previous Meshtastic BLE timing/notification behavior and also drops the cross-target compatibility adjustments that came with that commit.
Adjust the Meshtastic BLE transport timing on nRF52, retain pending FROMNUM notifications, and defer local admin/self replies to reduce request/response races during Android settings flows.
Also make the shared chat/UI code compile across non-nRF targets by gating RTOS and Arduino-only helpers and by updating the team page layout to the current page profile fields.
Note: this does not resolve the gat562 Meshtastic Android settings progress bar issue yet; the BLE/admin response path is improved and instrumented, but the root cause of the stuck loading state remains unresolved.
Refactor the GAT562 board runtime by extracting dedicated GPS and input runtime components and routing config application through a focused runtime apply service.
Narrow the nRF52 BLE manager dependencies, rework the Meshtastic BLE transport to better match the Android client handshake, and harden the FromRadio/FromNum flow with improved staging, read pacing, diagnostics, and stack-safe config snapshot generation.
Also repair repository text encoding issues in docs and UI comments, and add Meshtastic BLE timing documentation to capture the expected Android/firmware interaction model for future debugging.
Bundle the current workspace changes into a single checkpoint covering the in-progress shared UI, settings, and GAT562 runtime work.
What this includes:
- add the APRS asset source and image resource currently used by the shared UI asset set
- extend the shared System/settings flow so vibration and screen brightness can be surfaced through the device runtime and settings UI plumbing
- update the mono 128x64 runtime and the GAT562 platform bindings so the new settings controls are available in the GAT562 experience
- wire the ESP runtime/startup helpers needed to support the same device-runtime settings path on the ESP side
- persist the GAT562 self GPS position into the node store during core service updates so a device that already obtained a fix can retain its own last known position across reboot
- add an explicit node-store flush interface and implement it in both the nRF52 and ESP Meshtastic node-store backends so critical self-position updates can be forced to persistent storage immediately instead of waiting for the normal debounce window
- keep the existing debounced persistence behavior for ordinary node-store churn while allowing targeted immediate flushes for data that would otherwise be easy to lose during restart or power loss
liu weikai