Migrate admin UI from FastUI to Svelte

[Gudsfile]

Tracking issue for the admin UI migration. Origin: ADR 0003 (#309), FastUI was archived with no maintenance and the project had hit real limits with it. Decision: replace it with a Svelte single-page app served at /ui, talking to the existing /api/v1 REST API, migrated incrementally alongside the FastUI app until full parity, then FastUI removed.

Work is landing as a stack of PRs, each a self-contained layer:

• 📝 Decide: migrate admin UI from FastUI to Svelte #309: Decide: migrate admin UI from FastUI to Svelte (the ADR itself)
• 🎉 Scaffold the Svelte admin UI, coexisting with FastUI #310: Scaffold the Svelte admin UI, coexisting with FastUI
• ✨ Expose current-tag status as SSE under /api/v1 #311: Expose current-tag status as SSE under /api/v1
• ✨ Implement the Library page (list, create, edit, delete) #312: Implement the Library page (list, create, edit, delete)
• ✨ Implement the Settings page #313: Implement the Settings page
• ✨ Add the live current-tag banner #314: Add the live current-tag banner
• 💄 Bring Library visually in line with the old FastUI page #315: Bring Library visually in line with the old FastUI page
• ✨ Implement the Sonos page #316: Implement the Sonos page
• 👷 Build and package the admin-ui bundle in CI/releases #317: Build and package the admin-ui bundle in CI/releases
• 💄 Library UI polish: palette, icons, footer, click-to-copy #318: Library UI polish: palette, icons, footer, click-to-copy
• ✅ Add frontend test tooling and cover the whole admin-ui #320: Add frontend test tooling and cover the whole admin-ui
• Merge and create a development release
• Fix bugs and regressions and re-release
• Update documentation (README.md, AGENTS.md)
• Remove FastUI once the above is merged and running in production without regressions
• Release

Further UI polish and follow-ups beyond this list will be tracked separately as they come up.

[Gudsfile]

State as of September 10 (e3c94cb commit, #320)

Light Mode	Dark Mode

[Gudsfile]

Built a dev release and published it to TestPyPI to validate the whole install/run path end-to-end, outside this dev machine:

cd admin-ui ; npm install ; npm run build ; cp -r dist/. ../jukebox/adapters/inbound/admin/static/ ; cd ..
uv version --bump minor   # 1.5.1 (I forgot `--bump dev`)
uv build
uv publish --publish-url https://test.pypi.org/legacy/

Then, on a Raspberry Pi with no local checkout of this repo and no node/npm:

uvx --index https://test.pypi.org/simple/ --index-strategy unsafe-best-match --prerelease allow --from gukebox[ui]==1.5.1jukebox-admin ui

Works end-to-end: confirms the core premise of ADR 0003 (Pi installs a plain Python package, no JS toolchain on-device, compiled Svelte bundle served correctly).

Also noticed the package got noticeably heavier even for non-ui installs (sdist 78.2 kB → 107.7 kB, wheel 117.7 kB → 147.7 kB).

• 📝 Document the package-weight trade-off #322: Document why and the decision to accept it as-is for now, in ADR 0003.

[Gudsfile]

Some regressions found:

• ✨ Show Default and Persisted values in the Settings table #324: [Settings] Restore Default / Persisted / Effective / Source details
• ✨ Show Configured, Pinned default, Restart required and Advanced badges in Settings #325: [Settings] Restore Configured / Pinned default / Restart required / Advanced badges
• 🐛 Degrade Sonos selection status instead of 502 on total discovery failure #323: [Sonos] Sonos selection fails closed on discovery
• [Library, Sonos, Settings] Visual confirmation after save or reset (to be decided)
• [Settings] jukebox.player.sonos.selected_group (to be decided)

Not regression but to be added before release:

• JS tooling (lint, format, audit, vitest leak detector)

[Gudsfile]

Ideas for the future (issues to be creates):

• [All] Current-tag player widget (3 states)
• [Library] Search
• [Library] Sort
• [Library] Filter
• [Library] Cover art (POC w/ external sources, manual assignment)
• [Library] Mobile layout