- add tests for notes, property key/value, hidden-property exclusion, matched-value display
- fix flatMap indentation and stray semicolon
- use stable match key instead of display_value (avoids dup-key collisions)
- keep label searchable but not repeated in matched-values column
useAutosave reset its debounce whenever the nodes/edges array reference
changed. Live status polling (setNodeStatus) returns a fresh nodes array
on every message without dirtying the canvas, so during active
monitoring the status churn kept re-arming the timer faster than the
delay and the silent save could be starved indefinitely — the user
thinks autosave ran, but it never fired. Select-only changes reset it
the same way.
Add a monotonic editSeq counter that bumps only on real user edits. The
store's set() is wrapped to increment editSeq whenever an update flips
hasUnsavedChanges to true, so it stays centralised instead of touching
every mutating action. onNodesChange/onEdgesChange now set the dirty
flag only when a genuine edit occurred (not on select/measure), so those
no longer bump the counter. App keys the autosave changeSignals on
[editSeq] instead of [nodes, edges].
Add a regression test asserting editSeq bumps on add/update but not on
setNodeStatus, select, or initial dimensions measure.
ha-relevant: maybe
Autosave saves whenever hasUnsavedChanges is set, so any source that
dirties the canvas without a user edit turns into a spurious silent
save. Two such sources:
1. Live status updates. useStatusPolling routed backend status pushes
through updateNode, which always sets hasUnsavedChanges. With the
60s status cycle, an idle opted-in tab saved itself every cycle and
could clobber edits made elsewhere with its stale in-memory canvas.
Add setNodeStatus, a live-overlay action that updates a node's
status/response_time/last_seen without dirtying (mirrors
setServiceStatuses), and route polling through it.
2. Initial dimensions measure. onNodesChange treated React Flow's
first-measure `dimensions` change as an edit, marking a freshly
loaded canvas dirty before any interaction — autosave then saved on
every load/switch. Only count a `dimensions` change as an edit when
it is a real resize (resizing === true).
Add regression tests: setNodeStatus updates status without dirtying;
onNodesChange ignores initial-measure dimensions and select-only
changes but still dirties on a user resize.
ha-relevant: maybe
The previous guard compared the pinned design against the live
activeDesignId (the selection). On a design switch the selection flips
synchronously while the new canvas loads asynchronously, so the
on-screen nodes briefly still belong to the previous design — arming the
timer with the newly-selected id would save those stale nodes under the
wrong design. Track the design the canvas was actually loaded as
(loadedDesignId provenance) and gate autosave on that instead: pin it
when the timer is armed and re-check the live provenance at fire time,
skipping the save if a different canvas has since loaded.
Also validate the persisted `enabled` flag by type (mirror of `delay`),
and add the matching regression tests (provenance switch skip, enabled
type validation).
ha-relevant: maybe
Extract the debounced autosave into a useAutosave hook. Pin the active
design id when the timer is armed and re-check it at fire time: if the
user switched designs while the timer was pending, the in-memory canvas
belongs to a different design, so skip the save instead of writing it
under the wrong design id. Also skip when there is no active design.
Add unit tests for autosaveSettings (persistence, validation, pub/sub)
and useAutosave (debounce, enable/unsaved/design guards, switch skip).
ha-relevant: maybe
Auto Layout and YAML import routed through loadCanvas, which wipes
past/future history and clears the unsaved flag (design-switch
semantics). Undo was therefore impossible after either action (#280).
Add an in-place applyLayout store action that snapshots history, marks
the canvas unsaved and re-fits the view without discarding undo history,
and wire both Auto Layout and YAML import to it.
ha-relevant: yes
Adds an opt-in autosave feature (disabled by default) that silently
saves the canvas after a configurable period of inactivity.
- frontend/src/utils/autosaveSettings.ts (new) — localStorage-backed
setting {enabled, delay} with pub/sub CustomEvent pattern (mirrors
alignmentSettings.ts). Key: homelable.autosave.
- frontend/src/App.tsx:
* handleSave accepts optional { silent?: boolean } — skips the success
toast when called from autosave to avoid noise on every periodic save
* autosave state (useState + subscribeAutosaveSettings)
* debounce useEffect: resets on nodes/edges change; when quiet for
autosave.delay seconds and hasUnsavedChanges is true, calls
handleSaveRef.current silently
- frontend/src/components/modals/SettingsModal.tsx:
* Canvas section: Autosave canvas toggle + Save after delay selector
(3/5/10/30/60 s). Changes persist immediately to localStorage via
writeAutosaveSettings + propagate cross-tab via CustomEvent.
Default: disabled — no behaviour change for existing users.
Manual Ctrl+S and the Save button continue to work normally.
Errors always show a toast regardless of the silent flag.
Waypoints are stored as absolute canvas coords, so they stayed put when a
connected node was dragged, kinking a previously clean route. A dragged
container moved its children on screen too (their stored parent-relative
position is unchanged), so their edges were missed entirely.
onNodesChange now translates the waypoints of every edge touching a moved
node by that node's delta, propagating a container's delta to all
descendants and translating once when both endpoints ride along.
ha-relevant: yes
Fixes#279
Editing any field on a container-mode vm/lxc/docker_host node (e.g. just
renaming it or adding a property) reset its manually-set height to
undefined, snapping the container back to auto-fit content size and
scrambling/overflowing its nested children.
Root cause: updateNode clears height on every 'properties' change and the
node modal always includes properties in its payload. Only proxmox was
excluded from the reset; vm/lxc/docker_host in container mode were not.
- export CONTAINER_MODE_TYPES from virtualEdgeParent (reuse, no re-declare)
- updateNode: skip the height reset for any container-mode host type;
proxmox stays always-excluded (legacy), plain leaf nodes still auto-fit
- regression tests: manual height preserved for container-mode
vm/lxc/docker_host and proxmox; still reset for a non-container vm
Fixes#278
ha-relevant: yes
Editing any field on a container-mode host node (Proxmox/VM/LXC/docker)
collapsed all of its nested children into the top-left corner.
Root cause: the node modal always includes container_mode in its update
payload, and handleUpdateNode gated setProxmoxContainerMode on the field's
presence rather than an actual change. So every edit (e.g. changing the
icon) re-fired the container ON transition, which re-applied the
absolute->relative position conversion to children that were already
parent-relative -- piling them up in a corner.
- App.tsx: only call setProxmoxContainerMode when container_mode actually
changes (compare against the pre-update value).
- canvasStore: make setProxmoxContainerMode per-child idempotent so a
redundant ON/OFF call can never re-convert an already-nested child's
position.
- add regression test for the idempotent ON transition.
Wires DELETE /api/v1/designs/{id} into the MCP layer.
The backend already implements the endpoint and prevents
deleting the last remaining design.
- mcp/app/tools.py: Tool schema + _dispatch case
- mcp/tests/test_tools.py: 2 new tests (dispatch + schema)
30/30 tests passing.
The backend approve endpoint already accepts design_id via NodeCreate,
but the MCP schema didn't expose the field so callers couldn't target
a specific canvas when approving a discovered device. Add _DESIGN_ID_FIELD
to approve_device so AI clients (and the MCP bridge) can pass it through.
Server uses Streamable HTTP (StreamableHTTPSessionManager), but README
told users to register the client with --transport sse / "type": "sse".
An SSE client waits for an endpoint event the server never emits, so
Claude Code reports "Failed to connect" despite a healthy container.
Switch all client config snippets to http transport and point at the
trailing-slash /mcp/ URL to avoid the 307 redirect that can drop the
auth header.
Fixes#274
ha-relevant: no
The status filter on list_pending_devices closed the leak but left the
MCP surface unable to see anything beyond pending: approved inventory
and hidden devices were no longer reachable, and a wrongly hidden device
could not be recovered through the MCP at all.
Add three tools mirroring existing backend endpoints:
- list_inventory GET /scan/pending (pending + approved, optional status filter)
- list_hidden_devices GET /scan/hidden
- restore_device POST /scan/pending/{id}/restore (undo hide_device)
Bulk approve/hide/restore and scan-config endpoints are intentionally
left out: bulk mutation is the same blast radius that caused the mass-hide
incident this PR addresses.
Tests: 5 new (47 passed total).
The backend GET /scan/pending endpoint intentionally returns the whole
inventory — approved devices stay listed so the frontend can show the
canvas-presence badge. The MCP tool proxied that response verbatim while
its description promises devices "not yet approved or hidden", so MCP
clients saw every approved device as if it were still awaiting triage
(59 "pending" entries on a fully-triaged canvas, 58 of them approved).
Filter the tool response to status == "pending", keeping legacy rows
that predate the status field.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Udrvtwi3cqcxUNHZtBHkBg
CI mypy flagged the shared payload/background locals in _run_mesh_sync
as incompatibly typed across the zigbee/zwave branches. Annotate them
with the union / generic Callable and alias the per-branch
env_import_request imports so the two bindings don't clash.
- SettingsModal: widen to sm:max-w-3xl with max-h-[90vh] scroll and a
two-column grid so it no longer overflows the viewport. Left column =
status/service checks + canvas prefs; right column groups all auto-sync
config (Zigbee, Z-Wave, Proxmox).
- .env.example: drop ZIGBEE_/ZWAVE_ SYNC_ENABLED + SYNC_INTERVAL — the
auto-sync activation is configured in the Settings modal (persisted to
scan_config.json), same as Proxmox; connection config stays env-only.
ha-relevant: no
Mirror the Proxmox auto-sync pattern for the Zigbee2MQTT and Z-Wave JS
UI mesh imports. Connection config + MQTT credentials live in .env only,
are never persisted to scan_config.json, and are never returned by any
API or shown in the UI (single source of truth).
- config: ZIGBEE_* / ZWAVE_* env settings; only sync_enabled+interval
are persisted, connection/credentials stay env-only
- routes: GET/POST /config, POST /sync-now; auto-sync reuses the exact
manual _background_*_import + _persist_pending_import path (fresh
import when empty, update-in-place when nodes exist, ScanRun trace)
- scheduler: zigbee_sync / zwave_sync jobs with live enable + reschedule
- frontend: reusable MeshAutoSync section in Settings (Zigbee, Z-Wave)
- .env.example: documented both blocks
- tests: scheduler jobs, router config/sync-now/auth, credential-never-
persisted, SettingsModal sections
Manual Zigbee/Z-Wave import behaviour is unchanged.
ha-relevant: no
The scheduled auto-sync job called _persist_pending_import directly and
never created a ScanRun, so auto-imports left no trace in Scan history
(unlike manual /sync-now and /import-pending, which both record a run).
Create a ScanRun(kind=proxmox) and delegate to the shared
_background_proxmox_import flow — fetch, persist, mark the run
done/error, and broadcast the inventory-reload signal.
ha-relevant: maybe
- Replace 10 inline sonner toast mocks with the shared mockSonner() builder.
- Toolbar: add tests for the store-driven undo/redo disabled state and the
unsaved-changes dot — behaviour the full-store mock previously left
unasserted.
ha-relevant: yes
- Split the 1430-line canvasStore.test.ts into a canvasStore/ subpackage by
concern (nodes, containers, sizing, edges, selection, grouping, history,
clipboard, customStyle, floorMap); each file <350 lines. Test bodies are
unchanged; local makeNode/makeEdge replaced by the shared factories.
- Move panels/DetailPanel.test.ts -> utils/serviceUrl.test.ts: it tested
getServiceUrl, not DetailPanel.
Same 1396 tests, all green.
ha-relevant: yes
Add backend API tests for the two issue #265 features that shipped
untested: auto-positioning root nodes into a free grid slot (first at
origin, collision avoidance, child origin default, explicit coords and
explicit zero preserved) and auto-assigning edge handles from absolute
canvas Y (source above/below/equal target, partial handle fill, child
abs-Y resolved through parent).
Drop the max(0, round()) clamp in _find_free_position: it folded
negative-positioned nodes onto cell (0,0), falsely blocking or freeing
the origin slot. Negative nodes now keep their true cells and never
intersect the positive search space.
Wrap POST /api/v1/designs so AI clients can create a new design (canvas) and get its id back for use as design_id. Completes the create-and-target canvas workflow alongside list_designs.
Co-authored-by: Cursor <cursoragent@cursor.com>
Add an optional design_id argument to create_node, create_edge and get_canvas so AI clients can read and populate a canvas other than the first design. Add a list_designs tool wrapping GET /api/v1/designs so those IDs are discoverable.
Backward compatible: omitting design_id preserves the existing first-design fallback in the backend (nodes.py / edges.py / canvas.py).
Co-authored-by: Cursor <cursoragent@cursor.com>
Audit of the heavy-mock suites (scanner/status_checker/scheduler)
confirmed the mocking is sound layered seam-mocking, not brittle
internal patching. It did hide real gaps where a primitive was always
mocked and never exercised:
- status_checker._http_get: add status-code interpretation tests
(2xx/4xx online, 5xx offline) + service-check exception -> offline.
Module coverage 95% -> 100%.
- scheduler.reschedule_status_checks / reschedule_service_checks were
wholly untested; add validation + guard tests. Add _run_proxmox_sync
happy-path and error-swallow tests. Module coverage 76% -> 91%.
639 passed.
ha-relevant: no
Group the three upgrade-path suites (legacy->designs, hardware->
properties, pre-migration DB backup) under tests/migrations/ with
clearer names. Pure relocation via git mv; each keeps its own
self-contained fixtures. No behavior change; 627 passed.
ha-relevant: no
Break the 1897-line test_scan.py into topic modules under tests/scan/
(routes, approve, properties, run) sharing fixtures via a package
conftest.py and pure builders via helpers.py. Same 80 tests, no
behavior change; full suite 627 passed.
ha-relevant: no
Pouzor