Phase 3 — Migrate the widget library to @conduction/nextcloud-vue

Status: planned, not started. Phases 1 (catalog deprecation) and 2 (dead-code cleanup) are done. This document is the migration plan + parity audit for the remaining work: removing launchpad's local widget renderers/forms and consuming the shared Cn*Widget / Cn*WidgetForm library from nc-vue.

Goal

Launchpad currently ships its own copy of ~36 widget renderers + config forms under src/components/Widgets/Renderers/ and src/components/Widgets/Forms/. nc-vue has a parallel Cn*Widget / Cn*WidgetForm set. The grid engine already moved to nc-vue (CnWidgetGrid / CnWidgetWrapper); the widgets should follow so launchpad keeps only the workspace shell + glue (widgetRegistry.js, the WidgetRenderer dispatcher, the nc-widget bridge).

The reality (why this is a program, not a cleanup)

The nc-vue widget library is not finished as a public API, and not published:

• The published package launchpad consumes (@conduction/nextcloud-vue beta.102) ships exactly one of these components: CnTileWidget. Everything else is absent from the published dist.
• In nc-vue source, the components exist as files but are not wired into the barrel (src/components/index.js / src/index.js) — so they are not part of the public API even locally.
• nc-vue enforces a check:docs CI gate: every barrel export must have a docs/components/<kebab>.md. None of these widgets have docs yet.
• nc-vue also enforces a check:jsdoc ratchet (new components require 100% prop/event/slot JSDoc) and a docgen freshness gate.

So "port to nc-vue" means: finish, export, document, test, parity-audit, and publish a ~36-component library in the shared fleet lib (consumed by OpenRegister / OpenCatalogi / Procest / Pipelinq / MyDash), then migrate launchpad onto it. That is multi-day and has fleet-wide blast radius — it cannot be done in one pass, and a half-done state breaks both repos.

Parity audit — current nc-vue readiness

exists = dir present in nextcloud-vue/src/components/; export = in the barrel; test = has tests/components/<name>*; doc = has docs/components/<kebab>.md.

launchpad renderer/form	nc-vue component	exists	export	test	doc
Renderers/LabelWidget	CnLabelWidget	Y	–	–	–
Renderers/TextDisplayWidget	CnTextWidget	Y	–	Y	–
Renderers/ImageWidget	CnImageWidget	Y	–	Y	–
Renderers/LinkButtonWidget	CnLinkButtonWidget	Y	–	Y	–
Renderers/HeaderWidget	CnHeaderWidget	Y	–	Y	–
Renderers/DividerWidget	CnDividerWidget	Y	–	–	–
Renderers/FilesWidget	CnFilesWidget	Y	–	Y	–
Renderers/PeopleWidget	CnPeopleWidget	Y	–	Y	–
Renderers/NewsWidget	CnNewsWidget	Y	–	Y	–
Renderers/QuicklinksWidget	CnQuicklinksWidget	Y	–	Y	–
Renderers/LinksWidget	CnLinksWidget	Y	–	Y	–
Renderers/MenuWidget	CnMenuWidget	Y	–	Y	–
Renderers/ContainerWidget	CnContainerWidget	Y	–	Y	–
Renderers/VideoWidget	CnVideoWidget	Y	–	Y	–
Renderers/CalendarWidget	CnCalendarWidget	Y	–	Y	–
Renderers/SpendAnalyticsWidget	CnSpendAnalyticsWidget	Y	–	Y	–
Renderers/NcDashboardWidget	CnNcWidgetWidget	Y	–	Y	–
Renderers/TileWidget	CnDashTileWidget	Y	–	Y	–
Forms/LabelForm	CnLabelWidgetForm	Y	–	–	–
Forms/TextDisplayForm	CnTextWidgetForm	Y	–	–	–
Forms/ImageForm	CnImageWidgetForm	Y	–	Y	–
Forms/LinkButtonForm	CnLinkButtonWidgetForm	Y	–	–	–
Forms/HeaderForm	CnHeaderWidgetForm	Y	–	–	–
Forms/DividerForm	CnDividerWidgetForm	Y	–	–	–
Forms/FilesForm	CnFilesWidgetForm	Y	–	Y	–
Forms/NewsForm	CnNewsWidgetForm	Y	–	–	–
Forms/QuicklinksForm	CnQuicklinksWidgetForm	Y	–	–	–
Forms/LinksForm	CnLinksWidgetForm	Y	–	–	–
Forms/MenuForm	CnMenuWidgetForm	Y	–	–	–
Forms/ContainerForm	CnContainerWidgetForm	Y	–	–	–
Forms/TileForm	CnDashTileWidgetForm	Y	–	–	–
Forms/VideoForm	CnVideoWidgetForm	Y	–	–	–
Forms/PeopleForm	CnPeopleWidgetForm	–	–	–	–
Forms/CalendarForm	CnCalendarWidgetForm	–	–	–	–
Forms/SpendAnalyticsForm	CnSpendAnalyticsWidgetForm	–	–	–	–
Forms/NcDashboardForm	CnNcDashboardWidgetForm	–	–	–	–

Totals: 0/36 exported, 0/36 documented, ~15/18 renderers tested, ~3/18 forms tested, 4 forms missing entirely (PeopleForm, CalendarForm, SpendAnalyticsForm, NcDashboardForm).

Launchpad-specific behaviour to preserve during the port (not assume "identical"):

• Renderers/TileWidget supports the legacy oc_launchpad_tiles schema shape.
• Renderers/NcDashboardWidget is the NC Dashboard API bridge (v1/v2 polling + the hideHeader path used by WidgetWrapper).
• i18n domain differs (mydash/launchpad vs nextcloud-vue).
• Helper sub-components with no standalone nc-vue export: MenuItemIcon, MenuTreeNode, MenuItemEditor, TextTableEditor, ContainerChild.

Migration plan (publishable slices)

Do not big-bang this. Each slice is independently shippable.

0. Author the migration OpenSpec change (per-app, in openspec/changes/) + a parity checklist. Decide the nc-vue branch (currently the widgets live on a feature branch release/cn-table-widget-borderless, not development).

1. nc-vue: finish the library (per component or small group):

• Parity-diff launchpad vs nc-vue source; reconcile drift (props, defaults, i18n, styling). Treat launchpad as the behavioural reference.
• Add the 4 missing forms (CnPeopleWidgetForm, CnCalendarWidgetForm, CnSpendAnalyticsWidgetForm, CnNcDashboardWidgetForm).
• Wire each into the barrel (src/components/index.js, src/index.js).
• Write a docs/components/<kebab>.md per export (satisfies check:docs).
• Add/port tests; bump the jsdoc baselines; run npm test, check:docs, check:jsdoc, docgen freshness.

2. nc-vue: publish a beta (Codeberg CI) and note the version.

3. launchpad: consume + delete:

• Bump @conduction/nextcloud-vue to the new beta; npm install.
• Rewire src/constants/widgetRegistry.js to import Cn*Widget / Cn*WidgetForm from the barrel instead of the local files.
• Keep the registry, the WidgetRenderer dispatcher, and the nc-widget bridge wiring (these stay in launchpad).
• Delete src/components/Widgets/Renderers/* and src/components/Widgets/Forms/* that now resolve to nc-vue (keep launchpad-only helpers if any remain).
• Update/move the corresponding __tests__.
• Build + full vitest + live-verify every widget type renders and edits.

4. Optional follow-ups: promote Dashboard/IconPicker + IconRenderer and WidgetStyleEditor to nc-vue; fold WidgetEditCog into CnWidgetWrapper.

Risks / gotchas

• Fleet blast radius: adding ~36 exports + publishing affects every nc-vue consumer. Parity drift that "works in launchpad" could regress another app.
• CI gates: check:docs hard-fails without a doc per export; budget ~36 doc files. check:jsdoc ratchet requires 100% on new components.
• Publish dependency: launchpad cannot verify the deletion until the beta is live; never bump launchpad to an unpublished version on development.
• No silent parity assumptions: earlier analysis called the components "identical" from excerpts — they must be diffed properly before deletion.

Effort estimate

Roughly: 1 day reconcile/parity + write 4 forms, ~1 day docs (×36) + tests + baselines, publish/verify cycle, then ~0.5 day launchpad rewire + delete + verify. Realistically a multi-session program gated on an nc-vue release.

---

Slice 1 — Parity audit (DONE)

Key finding: nc-vue's widgets were deliberately refactored to be app-agnostic — data comes from dataSource props or cn*Source injections / *Endpoint builders, NOT hardcoded /apps/mydash/... calls. So the port is "drop-in reuse" only for the static/presentational widgets; the data-driven ones need launchpad to wire an adapter that points the nc-vue component at launchpad's existing endpoints. Nothing here blocks the migration, but it changes the work from "delete + import" to "delete + import + wire data source" for ~6 widgets.

Drop-in reuse (identical except i18n domain / icon-wrapper)

Renderers: CnLabelWidget, CnTextWidget, CnImageWidget, CnHeaderWidget, CnDividerWidget, CnVideoWidget, CnQuicklinksWidget, CnLinksWidget, CnMenuWidget. Forms: Label, Text, Image, Header, Divider, Quicklinks, Links, Menu, Container, Video, and TileForm → CnDashTileWidgetForm (rename).

Reuse + wire a data adapter (launchpad supplies the source)

widget	nc-vue hook	launchpad endpoint to wire
PeopleWidget → CnPeopleWidget	cnPeopleSource inject / dataSource	/apps/mydash/api/people
CalendarWidget → CnCalendarWidget	cnCalendarSource inject	/api/widgets/calendar/events ⚠ view-mode regression: nc-vue is agenda/upcoming only — month/week grid views are missing
SpendAnalyticsWidget → CnSpendAnalyticsWidget	cnSpendSource inject	/api/widgets/spend-analytics/...
NewsWidget → CnNewsWidget	itemsEndpoint prop	/api/widgets/news/{id}/items
FilesWidget → CnFilesWidget	folder-picker/list endpoint	path moves /apps/mydash/... → /apps/files/... (verify)
NcDashboardWidget → CnNcWidgetWidget	widgetId + itemsEndpoint	NC Dashboard API bridge route

Needs an nc-vue enhancement before parity

• CnCalendarWidget: add month/week grid views (or accept the regression).
• CnContainerWidget: enforce REQ-CONT max nesting depth (=3) — launchpad guards it, nc-vue doesn't.
• CnLinkButtonWidget: internal-action registry + create-file modal is launchpad-specific; nc-vue emits internal-action/create-file events instead → launchpad keeps that handler host-side.

Port from launchpad (no nc-vue counterpart yet) → add as Cn*WidgetForm

• PeopleForm → CnPeopleWidgetForm (props: layout, filters, excludeDisabled, showBirthdays, birthdayWindowDays, sortBy, columns)
• CalendarForm → CnCalendarWidgetForm (viewMode, daysAhead, eventSources, excludeNcalendars)
• SpendAnalyticsForm → CnSpendAnalyticsWidgetForm (viewMode, defaultPeriod, colorByVendor)
• NcDashboardForm → CnNcDashboardWidgetForm (widgetId, itemLimit, displayMode)

Recommended next slices

2. nc-vue: enhancements (Calendar views, Container depth) + the 4 forms, each with barrel export + doc + test.
3. nc-vue: wire the static drop-ins into the barrel + docs; publish a beta.
4. launchpad: add the data-source adapters (people/calendar/spend/news/files/nc), rewire widgetRegistry.js, delete local copies, verify.

---

Slices 2–3 — DONE (committed on branches)

nc-vue branch feat/widget-library-exports (commit e91b1ab1):

• Exported the 17 renderers + 18 forms via src/components/index.js + src/index.js.
• Created the 4 missing forms (CnPeopleWidgetForm, CnCalendarWidgetForm, CnSpendAnalyticsWidgetForm, CnNcDashboardWidgetForm).
• A doc page per export (check:docs green incl. prop coverage), regenerated docgen partials, updated jsdoc baselines (check:jsdoc green), rollup build green.
• Fixed pre-existing doc debt on CnDetailPage / CnActionsBar.

launchpad branch feat/widget-library-consumption (commit fda6bd06):

• widgetRegistry.js now sources the 9 presentational renderers + all 18 forms from @conduction/nextcloud-vue (aliased to the historical names).
• Deleted the 9 local renderers + 18 forms + 5 orphaned helpers + their tests (~64 files); updated the vitest stub + AddWidgetModal/ContainerWidget specs.
• Full vitest green (571), build green, live-verified end-to-end: adding a Label widget renders via nc-vue's CnLabelWidgetForm → CnLabelWidget.
• Verified locally by injecting the freshly-built nc-vue dist into node_modules.

Remaining (release mechanics + a follow-up slice)

1. Publish (CI/release — cannot be done from a dev shell): merge the nc-vue branch → publish a @conduction/nextcloud-vue beta containing the widget library.
2. Bump + merge launchpad: set launchpad's @conduction/nextcloud-vue dep to that beta, then merge feat/widget-library-consumption to development. (Its CI build fails until the dep is bumped — that's expected; do not merge before the publish.)
3. Data-widget adapter slice — DONE for the cleanly-adaptable renderers: people / news / spend-analytics now use nc-vue components, bridged via WidgetRenderer's provide() (cnPeopleSource, cnSpendAnalyticsSource) + the news itemsEndpoint prop (commit b67cd6a9).

Migration progress — 16 of 18 renderers + all 18 forms done

Resolved since the audit (committed on the two feature branches):

• calendar — added month/week views to CnCalendarWidget (nc-vue) + wired cnCalendarSource.
• files — added a configurable apiBase prop to CnFilesWidget (nc-vue) + pointed it at /apps/mydash.
• link — LinkButtonHost wrapper renders CnLinkButtonWidget, keeps the internal-action registry + create-file modal host-side.
• people / news / spend-analytics — cnPeopleSource/cnSpendAnalyticsSource provide() + news itemsEndpoint.
• nc-widget — CnNcWidgetWidget (native OCA.Dashboard fast-path + core OCS items API, no launchpad proxy). Live-verified: all 6 NC widgets on the dashboard render with 0 console errors.

All widget types migrated — what stays local is orchestration, by design

Every entry in the widget registry now resolves to an nc-vue renderer. The two items previously flagged as "blocked" turned out not to be widget-renderer migrations at all — they are launchpad-specific orchestration that correctly stays in-app:

item	resolution
container	The container host (Renderers/ContainerWidget.vue) is a GridStack-nesting grid + the REQ-CONT-006 max-depth=3 guard — app-specific orchestration, the direct parallel of the top-level DashboardGrid, which is also local. Its children render through the nc-vue renderers: ContainerChild.vue dispatches via launchpad's own registry and now provide()s the same per-placement cn*Source data adapters as WidgetRenderer (extracted to src/services/widgetDataAdapters.js), so nested people/calendar/spend/news/files widgets work identically to the top level. No nc-vue change was needed — the host stays local, the children are already on nc-vue.
tile	Two distinct components were conflated. (1) The registry tile widget type (content-based, {title, icon, iconType, backgroundColor, textColor, linkType, linkValue}) now renders via nc-vue's CnDashTileWidget — same content shape, no nldesign hacks; the old Renderers/TileWidget.vue + its spec are deleted. (2) The app-launcher tile row (src/components/TileWidget.vue, with the runtime nldesign <style> override + !important colour rules + icon brightness/invert + the edit cog) is the LaunchPad launcher itself — not a generic dashboard widget — so it correctly stays local, exactly like DashboardGrid/WorkspaceApp.

Net: 18/18 widget types on nc-vue; container host and the app-launcher tile are launchpad orchestration that belongs in the app.

Release handoff (CI/review — not doable from a dev shell)

1. Merge nc-vue feat/widget-library-exports → publish a @conduction/nextcloud-vue beta.
2. Bump launchpad's dep to that beta; merge feat/widget-library-consumption → development. (The launchpad branch's CI build fails until the dep is bumped — by design.)