Ben MeadorsandGitHub
bbe9a0d5cd
* feat(protobufs): sync to firmware-current and consume workspace package Sync the vendored .proto sources to firmware-current (v2.7.25+48), regenerate the v2 TS bindings, and consume the workspace @meshtastic/protobufs (workspace:*) in place of the stale JSR 2.7.20 — finishing the monorepo migration (core was already workspace:*). Includes the one required breaking-change fix: admin nodedb_reset changed int32 to bool, so resetNodes() now sends value: true. * build(protobufs): vendor generated bindings for workspace consumers The package is consumed via workspace:* — its exports point at the TS source, which imports ./dist/meshtastic/*_pb.ts — so the generated output must exist at build time. CI builds web/core with no codegen step and the runners have no buf CLI, so the bindings are vendored here (kept gitignored; lint/format skip them). Regenerate with: pnpm --filter @meshtastic/protobufs gen * fix(protobufs): clean script removes the actual generated output dir buf writes bindings to packages/ts/dist, but clean was removing a non-existent root dist — so it never cleaned stale output. Addresses Copilot review feedback. * refactor: move web app packages/web -> apps/web Aligns the web app with the apps/web layout (matching the Vercel web-test Root Directory and the SDK-migration direction). Pure directory move plus root config: pnpm-workspace (adds apps/*), vitest projects, root tsconfig reference, and the pr/release-web/nightly workflows. vercel.json moved with the app. Build + 36 validation tests green. * feat: config fields, module pages, key verification, telemetry capture Incorporates the firmware-current feature work onto the protobuf foundation: new config fields (Display message bubbles; LoRa fem_lna_mode + serial_hal_only; Telemetry air_quality_screen_enabled); 4 new ModuleConfig pages (TrafficManagement, StatusMessage, TAK, RemoteHardware); the manual Key Verification flow (sendKeyVerification + ClientNotificationDialog stages + Verify Key button); live telemetry capture (nodeDB addDeviceMetrics) and admin hardening (toggleMutedNode, graceful PortNum default); plus the sdk-preview ConfigEditor demo and store/config tests. Build + lint + format + 131 tests green. * chore: drop #1062 (unsaved-change-detection) to match upstream revert #1062 was merged to main by accident (per @danditomaso) and is being reverted. Reverse-applied its diff here via 3-way so #1097 stays consistent with where main is headed, while keeping the feature changes layered on the same files (deviceStore/changeRegistry). Build + 131 tests + lint + format green. * fix(nodes): clean up SNR display in node table and map popup SNR is a ratio measured in dB, not dBm (which is absolute power); the node table and map popup both mislabeled it and crammed three values together: '0dBm/50%/50raw'. The trailing '%/raw' pair was the same heuristic ((snr+10)*5) shown twice — once clamped, once not. Render SNR in dB rounded to one decimal, color-coded by a 0-100% signal-quality heuristic (green/yellow/red), with the quality percentage as a muted secondary. Drop the redundant raw value. Adds unit.db; this matches the existing SNRTooltip, which already renders dB.
3.2 KiB
sdk-preview — a proof-of-direction slice
This folder is a small, self-contained demonstration of the architecture from
meshtastic/web#1050 (Dan's
@meshtastic/sdk migration), built against the firmware-current protobufs
synced in this branch.
It is not wired into the live app and replaces nothing. The goal is to show
the shape of the SDK direction — domain-driven slices, signals-backed reactive
state, Result-returning use-cases, and ports/adapters — without pulling in the
full SDK + sqlocal/OPFS persistence stack.
What it mirrors
| This slice | @meshtastic/sdk (PR #1050) |
|---|---|
core/signals (createStore, SignalMap, ReadonlySignal) |
core/signals |
core/errors (MeshError, toMeshError) |
core/errors |
core/client/MeshClientPort (minimal port) |
core/client/MeshClient |
features/config/domain (RadioConfig, ModuleConfig, ConfigEditor) |
same |
features/config/application/ConfigUseCases (Result<number, MeshError>) |
same |
features/config/infrastructure (ConfigMapper, configBuilders) |
same |
react/useSignal, react/useConfigEditor |
@meshtastic/sdk-react |
adapters/fromMeshDevice |
the SDK's transport/registry wiring |
The pattern on display: ConfigEditor
ConfigEditor is the piece that replaces today's Zustand changeRegistry. It
keeps two reactive views per config section:
- baseline — device truth, updated automatically from
onConfigPacket/onModuleConfigPacket. - working — the UI's in-flight edits.
A mid-edit packet from the device updates the baseline but does not clobber a
section the user is editing; commit() sends only the dirty sections inside a
beginEdit → setConfig…/setModuleConfig… → commitEdit envelope and returns a
Result; disconnect resets everything so a stale working copy can't leak across
reconnects. See features/config/__tests__/ConfigEditor.test.ts.
Why this ties the branch together
RadioConfigSection / ModuleConfigSection are derived from the protobufs
(Exclude<…payloadVariant["case"]>), and ConfigMapper is fully generic. Because
this branch already synced the firmware-current protobufs, the new modules
(trafficManagement, statusmessage, tak, remoteHardware) and new config
fields flow through the domain, mapper, editor, and dirty-tracking with zero
per-section code — which is exactly the leverage the SDK migration is built on.
Wiring it into a page (illustrative)
import { useConfigEditor, useSignal } from "@app/sdk-preview";
function TrafficManagementPanel() {
const editor = useConfigEditor();
const modules = useSignal(editor?.modules ?? /* empty signal */);
const isDirty = useSignal(editor?.isDirty ?? /* false signal */);
// editor.setModuleSection("trafficManagement", next); … editor.commit();
}
Deliberately out of scope
- Channels / owner / queued-admin editing (need the channels + nodes slices).
@meshtastic/sdk-storage-sqlocal(OPFS SQLite) persistence.- The full
MeshClient/MeshRegistrymulti-device client. - Live UI adoption — that's the slice-by-slice migration tracked in PR #1050's follow-ups.