Route all prometheus widget queries through Grafana /api/ds/query instead of
direct Prom HTTP. PrometheusConfig: drop base_url, add grafana_url +
datasource_uid; secret grafana_api_key (required). PrometheusWidgetSource →
MetricSource with _gateway_query POST method. normalize_grafana_frames
recovered from 65bae95 + shared _dedup_label helper. Gateway-path status
check. Startup old-config validation. CHANGELOG migration note. All adapter
tests rewritten for POST /api/ds/query + Grafana frames mock. Backend: 331
pytest pass, ruff clean. Frontend: build green (unchanged in S1).
Route metric queries through Grafana /api/ds/query instead of direct Prom
(Prom is firewalled / unreachable from Manage; Grafana is the only path).
Partial revert of prometheus-direct-charting: keep the gauge/mean/
LineSeriesChart rendering, restore the frames->series normalizer (recovered
from git 65bae95), change prometheus service config to hold Grafana gateway
fields (url + api_key + datasource_uid). Rename widgets to neutral Metric*.
First non-additive canonical sync (prometheus-charting MODIFIED). Updates the
pending service-credential-tester proposal dependency.
On-demand per-service-type credential tester for the add/edit dialog. Model on
test_machine_ssh. POST /api/services/test dispatching to per-type routines
(qBit login+maindata, Prom instant query, Jellyfin /Users, etc.) returning
{ok, detail, evidence}. Frontend Test button + gate Create/Save on pass with
'Save anyway' override. Resolves#3 (qBit login failures visible in UI not
logs) and complements #1 (validation surfacing, already shipped).
qBittorrent was missing from SERVICE_TYPE_NAV_ENTRIES, so configured qBit
instances never appeared in the left nav (unlike jellyfin/prometheus/etc.).
Add entry with Magnet icon. navEntries.test.ts filters by configured types
(no fixed-count assertion) so it stays green.
CreateServiceDialog.save() awaited mutateAsync without a try/catch, so a
backend 422 (e.g. base_url missing http:// schema) threw uncaught and the
dialog sat silent with no feedback. Wrap in try/catch, hold the error in
local state, render a destructive Alert above the footer. Reset/onClose
only on success; on error the user can fix and retry.
Correctness fix: observability + backup hooks query globally, so multi-instance
service pages show data for the wrong instance. Tabs already accept instance
prop with TODO comments; backend mostly supports service_id already. Scope:
add serviceId to 6 hooks + fetch fns + 3 tabs; add service_id to backup
endpoints. Backward-compatible (optional params). ~250-350 lines, single slice.
Code merged to main on 2026-06-26 (01527ae) + fix passes. All 10 ACs PASS.
~8600 ins / ~3900 del across 103 files. Only SDD artifacts were untracked.
No code changes, no canonical sync. NOTE: verify-report describes a Grafana
LinksTab later removed by prometheus-direct-charting (2026-07-08); historically
accurate point-in-time record — see archive-report.md footnote.
Code merged to main on 2026-06-26 via rebase (01527ae) + bug-fix passes
(5a43894/04871bd/f7f590f). All 8 ACs PASS. Only SDD artifacts were untracked;
this archive closes the paperwork gap. No code changes, no canonical sync.
See archive-report.md for carry-over notes (iOS real-device test residual).
Wire ServiceDataHarness.cascade_delete into SettingsStore.delete_service
(best-effort try/except, logs on failure). Fix migration runner to also
catch 'no such table' on fresh DBs (ALTER TABLE before init_schema).
Integration test proves end-to-end cascade across both concerns (qBit
samples + media items) with multi-instance preservation.
Backend: 322 pytest pass, ruff clean.
Register MediaIndex as a harness concern with ALTER TABLE migration to add
service_id column (idempotent). Scope replace_items by service_id (FIXES latent
global-clear bug where building for one Jellyfin wiped another's rows). Scope
query by service_id (empty-string = all rows, backward-compat). Thread
service_id through build_media_index + worker + query_media router. New
regression test proves scoped replace preserves other services' rows.
Backend: 321 pytest pass, ruff clean. Frontend: build green.
New canonical domain openspec/specs/prometheus-charting/spec.md with all
27 requirements (SC-101..127) as the durable post-change contract. Change-side
delta specs/prometheus-charting/spec.md + sync-report.md. web-ui canonical
untouched (different concern).
Add gauge widget (recharts RadialBarChart with configurable threshold
bands, scalar-only per SC-111) and mean widget (client-side average over
range-query window, scalar-only per SC-114). Extract shared _instant_query
helper from the metric path; _fetch_gauge and _fetch_mean dispatch in
PrometheusWidgetSource.fetch(). Both new widget kinds declared in
integrations/prometheus.py and frontend registry.
Backend: 305 pytest pass, ruff clean. Frontend: 136 vitest pass, build+lint green.
ObservabilityPage.tsx was refactored into service-tabs/; update removal
criteria to name real targets and whitelist the Dashboard.test fixture
shortcut-label collision.
5 locked design decisions: step derivation formula (max(15, round(s/200))),
shared normalize_prometheus_matrix helper in widgets/prometheus_range.py,
recharts RadialBarChart gauge w/ threshold bands, mean via client-side
avg over query_range, 3-slice plan each <=400 lines.
Source-verified: ObservabilityPage refactored into service-tabs/ (map stale).
The previous fix called client.users() unconditionally to validate the
user_id, adding an extra HTTP round-trip before the build. On a slow
Jellyfin connection this burned through the 10s timeout before the
actual libraries() call.
Restructured to try libraries(user_id) directly first. Only when that
fails does the worker resolve the username via the users API. So:
- Correctly configured user_id (internal hash): zero extra round-trips.
- Username like 'admin': libraries() fails → users() resolves it → retry.
283 backend tests pass; ruff clean.
The Jellyfin service config's user_id field accepts either the internal
Jellyfin user ID (a long hash) or a username (e.g. 'admin'). The worker
passed the raw value directly to client.libraries(user_id), but Jellyfin's
API rejects usernames with a 400.
Now validates the configured user_id against the Jellyfin users API:
1. If it matches a user's Id (internal hash), use it directly.
2. If it matches a user's Name (username like 'admin'), resolve the Id.
3. If no match, fall back to the first user and log a warning.
283 backend tests pass; ruff clean.
The worker called _resolve_jellyfin() and client.libraries() OUTSIDE the
try/except block. If Jellyfin was unreachable, the worker crashed with
an unhandled exception and NO error state was written to the DB — the
status stayed 'queued' forever with zero feedback.
Moved _resolve_jellyfin + client.libraries INSIDE the try block, and
created final_index + staging_index BEFORE the try so the except handler
can write the error state. Now any failure (connection refused, timeout,
missing config) is written to the DB as build_error and the frontend
shows it inline.
283 backend tests pass; ruff clean.
The build endpoint had Depends(get_jellyfin_client) and Depends(get_user_id)
which executed BEFORE the function body. If Jellyfin was unreachable, these
raised HTTPException(503), the function never ran, and the worker was never
started. The frontend mutation had no onError handler, so the failure was
completely silent — the button briefly showed 'Building...' then reverted
to 'Build index' with zero feedback.
Backend fix: removed the Jellyfin dependencies from post_build_index.
The worker subprocess resolves its own Jellyfin connection via
_resolve_jellyfin(service_id) — the endpoint just needs to start the
worker process. The libraries count starts at 0 and gets updated by
the worker once it connects.
Frontend fix: added onError to useBuildIndex that invalidates the status
query (so the UI reflects the non-building state). MediaTab now displays
the build error inline: 'Build failed: <message>' next to the button.
283 backend tests pass (updated build test for new no-dependency flow);
128 frontend tests pass; ruff/eslint clean.
The WidgetConfigDialog useEffect that auto-enters edit mode when
editWidgetId is set only searched owned widget instances (from
useWidgetInstances). Referenced widgets (from useWidgetReferences)
were never found, so startEdit never fired and the dialog fell through
to the list view.
Now the effect searches both owned instances and referenced widgets,
so clicking edit on any widget — owned or referenced — opens the edit
form directly.
128 tests pass; 0 lint errors; build clean.
Four fixes:
1. Mobile edit fullscreen scroll: the Sheet primitive's
data-[side=bottom]:h-auto was overriding our h-[100dvh] on
SheetForm, preventing scroll. Added data-[side=bottom]:h-[100dvh]
to the SheetForm className to win the specificity battle.
2. Direct-edit close showed list view: when opened via editWidgetId
(the hover edit button), saving or canceling called reset() which
showed the widget list instead of closing the dialog. Now derives
directEdit from editWidgetId — when true, reset() calls onClose()
to close entirely.
3. Copy button missing on mobile: MobileWidgetSections didn't pass
onCopy to its WidgetInstanceCard instances. Now accepts and wires
onCopyWidget, so referenced widgets show the copy/detach button on
mobile too.
4. Service enabled badge stale: ServiceConfigEditor showed
instance.enabled (the initial prop) instead of the local enabled
state. Now reads the local enabled variable so the badge updates
when the user toggles the switch.
128 tests pass; 0 lint errors; build clean.
Five fixes:
1. Dialog mobile scroll: DialogContent now has max-h-[calc(100dvh-2rem)]
overflow-y-auto so dialogs that don't fit on screen can scroll
instead of clipping their footer (and Cancel button) off-screen.
2. isDirty false positive: WidgetConfigDialog's SheetForm used
isDirty={draft !== null} which was true the moment you opened edit
mode, even with no changes. Now stores a draftBaseline at startEdit
time and compares JSON.stringify(draft) !== JSON.stringify(baseline).
The discard-confirmation only appears when something actually changed.
3. Mobile edit button always visible: the widget card's edit button was
opacity-0 group-hover:opacity-100 (hover-only). Changed to
md:opacity-0 md:group-hover:opacity-100 — always visible below md,
hover-reveal at md+.
4. Copy button for referenced widgets: WidgetInstanceCard gains an onCopy
prop. On the Dashboard, referenced widgets get a Copy icon button that
triggers detachRef (creates an independent clone). The edit button on
referenced widgets edits the original (shared config).
5. ConfirmDialog Cancel: fixed by #1 (the Cancel button was off-screen
on mobile dialogs that couldn't scroll).
128 tests pass; lint/build green.
WidgetConfigDialog fetched useWidgetInstances(serviceId) with no scope
filter. On the main dashboard (serviceId undefined, dashboardScope='main'),
this returned ALL widget instances including service-scoped ones from
service overviews — so editing the main dashboard showed widgets that
were never added there.
Fix: pass scope='dashboard' when serviceId is empty and dashboardScope
is set. This fetches only dashboard-scoped widgets (service_id IS NULL).
The 'Add existing' picker (allWidgets) stays unscoped so users can still
reference service widgets onto the dashboard.
128 tests pass; lint/build green.
Four fixes:
1. Settings > Services tab: master/detail layout. Replaced the vertical
stack of ServiceConfigEditor cards with a SelectionRailCard (list on
left) + SectionCard (details on right) — same pattern as Machines
and SSH Keys tabs. Click a service in the rail to edit it.
2. Service Overview widget restriction. When adding widgets on a
service's Overview, the dialog now only shows built-in widgets +
widgets for THAT service type (not all services). Dashboard/named
dashboards (no serviceId) still see all.
3. Reorder buttons fixed. The swap-sort_order approach was a no-op when
both items had sort_order=0 (the default). Now moveInstance
renumbers ALL items by their new index position (i * 10) after the
swap, guaranteeing values change. References use updateRef, owned
widgets use saveWidget, both sequential.
4. Media index worker resolution. The subprocess worker called
get_jellyfin_client/get_user_id (FastAPI request dependencies) which
don't work outside request context. Now accepts a service_id
parameter (passed from post_build_index) and resolves the Jellyfin
instance directly from the settings store via _resolve_jellyfin.
Raises RuntimeError (not HTTPException) on failure.
283 backend tests pass; 128 frontend tests pass; ruff/eslint clean.
Each widget card now shows a settings icon in the top-right corner on
hover (desktop) or always-visible (mobile via mobile-touch-target).
Clicking it opens the WidgetConfigDialog directly in edit mode for that
widget (via a new editWidgetId prop on WidgetConfigDialog that
auto-enters the draft-edit path via useEffect).
Wired across all three widget surfaces:
- Dashboard (both mobile section + desktop grid)
- Service OverviewTab
- NamedDashboardPage
128 tests pass; build/lint green.
Extract the duplicated _resolve_service_record helper (identical in
routers/monitoring.py and routers/authentik_users.py) into a shared
services/service_resolution.py module. Both routers now import
resolve_service_record from the shared module.
The authentik router previously hardcoded service_type='authentik' in
its local copy; the shared helper takes service_type as a param (same
as monitoring's did).
Tests updated: test_api.py patches now target the correct module paths
(resolve_service_record on the monitoring module where it's imported,
build_service_record on the service_resolution module).
283 backend tests pass; ruff clean.
Three reusable-widget follow-up fixes:
1. Reference sort_order independently reorderable. Reordering a
referenced widget now updates the widget_references.sort_order (per-
dashboard), not the shared widget instance sort_order. New backend
update_widget_reference method + PUT /api/widgets/references/{id}
endpoint. Frontend moveInstance checks _ref_id to choose the right
mutation (updateRef for references, saveWidget for owned).
2. Detach preserves service_id. detach_widget_reference now copies the
original widget's service_id into the clone, so service-bound widgets
(Grafana chart, Jellyfin activity) continue to render after detach.
3. Named dashboards support widget references. NamedDashboardPage
fetches useWidgetReferences('named:<slug>') and renders them via
WidgetInstanceCard alongside pinned links. 'Edit widgets' button
opens WidgetConfigDialog with dashboardScope='named:<slug>'.
Also: removed useMemo on combinedWidgets in WidgetConfigDialog to fix
a react-hooks/preserve-manual-memoization lint error (the React Compiler
ESLint plugin couldn't verify the spread+sort memoization).
283 backend tests pass (+1 update_reference test); 128 frontend tests
pass; ruff clean; 0 lint errors.
Widgets configured on one dashboard (e.g., a Grafana service's Overview)
can now be live-referenced on other dashboards. Editing the widget config
updates it everywhere it's referenced. References can be detached into
independent clones.
Backend: new widget_references table (dashboard_scope, widget_id,
sort_order) with ON DELETE CASCADE. CRUD methods + 4 endpoints:
GET/POST /api/widgets/references, DELETE /api/widgets/references/{id},
POST /api/widgets/references/{id}/detach (clones the widget into a
standalone instance, then removes the reference).
Frontend: WidgetConfigDialog gains a dashboardScope prop. When set
(the main Dashboard passes 'main'), the dialog shows:
- Owned + referenced widgets in a combined list, with a link badge on
references.
- 'Add existing widget' picker: searchable list of ALL widget instances
not already on this dashboard. Click to create a reference.
- Detach button on references: clones the widget (service_id=NULL) and
removes the reference.
- Delete on a reference removes the REFERENCE (not the original widget).
Dashboard renders referenced widgets alongside owned widgets.
Detaching a service-bound widget clones it with service_id=NULL — the
clone may need re-binding to a service to render correctly. Named
dashboards don't pass dashboardScope yet (pinned-links-only); when they
gain widget support, the backend already handles any scope string.
282 backend tests pass (+2 reference lifecycle); 127 frontend tests
pass; ruff/eslint/tsc/vite all green.
Two fixes for the Grafana chart widget:
1. Invisible lines: the CHART_COLORS used 'hsl(var(--chart-1))' but the
CSS variable is named '--color-chart-1' and already contains a hex
color (#4f8cff). The hsl() wrapper produced invalid CSS, making
every stroke invisible. Fixed to var(--color-chart-1).
2. Multiple series collision: the backend labeled all Prometheus series
with the value field name (often just 'Value'), so multiple time
series collided on the same recharts dataKey and overwrote each
other. Now extracts meaningful labels from the Grafana frame metadata:
prefers displayName, then Prometheus metric labels (e.g.
'instance=server1:9100 mode=iowait'), then falls back to the field
name. Duplicate labels get a numeric suffix for uniqueness.
Multi-series queries now render correctly: each Prometheus time series
gets its own colored line with a unique label in the legend/tooltip.
280 backend tests pass (+1 labels test); 127 frontend tests pass; ruff/
eslint clean.
Developer