The store had three structural problems flagged in user
feedback:
1. "Lernkurve zu hoch — alle bridge, debug, alpha, planned,
keiner weiß was das heißt." Wire-level enum values bled
through into the UI: `bridge`, `debug`, `published`,
`alpha`, `planned`. Operators don't share the data model.
2. "Filter / Kategorien sollten oben in die Leiste neben Store
zum ausklappen, der Body soll übersichtlicher werden." A
chip-row plus category strip plus result-count line ate a
full row of viewport on every screen.
3. "Suche passt besser an den unteren Bildschirmrand." The
chat-style ask bar belongs at the bottom — Claude / ChatGPT
/ Slack pattern — not at the top where it competes with
the Today hero.
Plus: "Es gibt noch ganz viele overflows in den Beschreibungen."
Changes:
- Plain-language display names. New ARB keys for category
labels (Connectors / Sample modules / AI models / Storage /
Channels / Authentication / Orchestration / …) and updated
status values (stable / experimental / coming soon). Wire
ids stay in StoreEntry, RPCs, seed.yaml — only the rendered
pills and dropdowns get translated. Helpers
`_categoryDisplayName` and `_statusDisplayName` keep the
mapping in one place.
- Toolbar moves into the AppBar. The new `_CategoryDropdown`
hosts the category picker as a popup-menu with localised
labels; the existing `_FilterButton` lives next to it; the
reload icon stays where it always was. The body no longer
carries any chip strip or result-count line.
- Bottom-anchored `_AskBar`. The Scaffold body becomes a
Column of `Expanded(scrollable content)` plus a pinned
composer row at the foot of the viewport, with a top
divider matching the chat-input pattern of modern AI
assistants. AI answers now render at the top of the scroll
(above the Today-Hero) so the operator sees them right
after submitting the question at the bottom.
- Overflow sweep on descriptions. Today-Hero header row
becomes a Wrap (badge + deck flow naturally on narrow
windows), the title is bound to 3 lines, the body to 6.
Card category text gets `maxLines: 1, ellipsis` in both
StoreCard and FeaturedTile. Recommended-source card title
gets the same treatment.
- Better toast for the "added but zero capabilities" path.
When MCP discovery succeeds but returns no tools (typical
Streamable-HTTP servers without Mcp-Session-Id support, or
servers that need `notifications/initialized`), the toast
now explains the situation in plain language and points
the operator at Settings → MCP Clients to retry. Old
pluralised toast still fires when N > 0.
Hub-side `notifications/initialized` for HTTP MCP and
Mcp-Session-Id support are out of scope for this commit;
tracked as a separate fai_hub follow-up.
Signed-off-by: flemming-it <stefan.a.flemming@googlemail.com>
Three changes against the user feedback "der text ist zu
zerhackt, die filter-leiste nimmt zu viel sicht weg, llm-suche
wenn das geht":
1. Today-Hero copy rewritten in flowing prose. The fallback
story drops the code-fence templates ("`mcp.<server>.<tool>`")
and reads as a single arc per language. Same trust model,
gentler register.
2. Filter row collapses behind a single `_FilterButton` with a
"Filter · N active" count badge. Status / Source / Installed
chips now live inside `_FilterDialog` — invisible until the
operator wants them, recovers a full row of viewport. The
category strip stays inline because category is the most
common cut and benefits from being one click away.
3. New `_AskBar` replaces the old single-line `_SearchField`:
- Multi-line input (1-4 lines auto-grow) so questions don't
overflow horizontally.
- Cmd+Enter / Ctrl+Enter submits; bare Enter inserts a
newline so the operator can write multi-line questions
naturally.
- Heuristic question detector (ends in `?` or ≥4 words)
routes the submit to the System-AI's askAi RPC instead of
the substring search. Live keystrokes still keyword-search
for short queries; question-shaped input holds the grid
steady until submit so the visible result set doesn't
wipe with every space.
- The LLM is asked for a strict-JSON ranking of up to 5
modules with a one-phrase reason each. Unknown module
names are dropped silently — operators must not see
hallucinated entries.
- The new `_AiAnswerCard` renders the LLM's answer plus the
ranked match list above the grid; the grid filters to
just those names so the answer and the visible cards
stay coherent. Clearing the question collapses everything
back.
- Falls back to plain keyword search when System AI is off
(with an explanatory hint instead of the AI hint).
The AI path uses the System-AI the operator already configured
in Settings; same privacy mode, same audit-log trail. Nothing
new leaves the hub.
Signed-off-by: flemming-it <stefan.a.flemming@googlemail.com>
App-store research shows that the most loved discovery
surfaces — Apple's Today tab being the canonical example —
work because editors tell stories instead of stacking
algorithmic recommendations. F∆I has no surveillance budget
and no monetization pressure, so we lean fully into editorial
curation.
This commit adds `_StoreTodayHero`, the first surface a
browsing operator sees:
- Single curated story per release, kept const in
[_kCurrentTodayStory] so the narrative is reviewable in code
review and ships in the audit log via the binary hash. No
CMS, no network, no surveillance.
- Bilingual content shipped inline (`titleEn`/`titleDe`,
`bodyEn`/`bodyDe`) so KRITIS deployments don't need a
translation backend.
- Gradient backdrop with hero icon, deck, narrative paragraph,
optional CTA. Visual treatment matches Apple Today's
hierarchy: badge → headline → body → action.
- Auto-hides whenever a filter is active so a purposeful
search isn't pushed below the fold.
- Per-session dismiss button — no permanent suppression, the
next Studio launch shows it again so a release-bumped story
has a chance to be seen.
Current story (Studio v0.30.x) directs operators to the new
recommended-sources strip, closing the loop between editorial
context and one-click action.
Signed-off-by: flemming-it <stefan.a.flemming@googlemail.com>
The Store is now the centre of capability discovery — provider
configuration moves out of Settings into the Store itself.
Three changes work together:
1. Source filter chips next to the existing status chips:
All sources / Native / MCP / n8n. Applied client-side after
the hub returns results so toggling is instant. The result
count and the "Clear filters" reset both account for the
source filter too.
2. Per-card provenance pill (`_ProvenancePill`): shows whether
each entry is native, mcp · <provider>, or n8n · <provider>
so the operator can triage the source at a glance — same
role the "verified" badge plays in commercial app stores.
3. `_RecommendedSourcesStrip` replaces the older
`_FederationNudge`. Renders curated public MCP servers
(DeepWiki, Semgrep) as one-click cards with an inline
`[+ Add]` button — no Settings detour, no form. Both servers
are HTTPS Streamable-HTTP, no API key, no subprocess.
Useful AI was considered but its 340+ tools would drown the
index — kept out of the curated list. The strip auto-hides
the moment any federated entry exists.
Removed the now-orphaned storeFederationNudge* ARB keys.
Signed-off-by: flemming-it <stefan.a.flemming@googlemail.com>
- Localize Flows page: app-bar title and reload tooltip,
hub-unreachable / no-saved-flows empty states, run-flow input
dialog (title with flow name, description, hint, Cancel /
Run), running dialog (title, "Flow running…" status, no-output
message, Close button), flow-card Run button.
- Localize the module-sheet bottom sheet: failed-to-load text,
Capabilities and "Declared permissions" section headers,
no-permissions placeholder copy.
- Localize the System-AI configuration dialog: title, intro
paragraph, provider dropdown label, endpoint label, API-key
env-var label (required vs optional) and disclaimer, privacy
mode header and three options (Off / Redacted / Full) plus
their descriptions, test-result panel ("Connection ok" /
"Connection failed", "Reply: …" prefix), model picker (label,
hint fallback, helper text with Ollama variant, Refresh /
Pull / Pulling… buttons, list errors and empty states),
hardware banner ("Detected: …" + " · curation reviewed …"
suffix), suitability legend (recommended / balanced / small /
large / huge / unknown), cache row with pluralized count and
Clear button, cache cleared / clear-failed toasts, pull-empty
/ pull-failed errors. Suitability label moved off a getter
onto a `labelFor(BuildContext)` method so it can read the
current locale.
Provider preset descriptions and modelHint strings stay in
English in `_ProviderPreset.all` — they're tightly coupled to
the wire-protocol identifiers and would require a runtime
factory rebuild rather than a const list.
Signed-off-by: flemming-it <stefan.a.flemming@googlemail.com>
Two threads.
Translation expansion: every visible string on the Doctor
page (section headers, status text, button labels, daemon-
control card, daemon-files panel, update banner) and the
Modules page (recent-activity strip, capabilities label,
uninstall dialog + toast) flips between DE and EN with the
sidebar toggle. Adds ~50 ARB keys split between
`app_en.arb` / `app_de.arb`. Pluralised + parametrised
strings (`{n} events verified`, `Running on {name}: {endpoint}`)
use the standard ICU placeholder syntax so future locales
slot in without code changes.
Sparse-store federation nudge: the Store page renders an
inline banner above the grid when the operator has zero
federated entries — single biggest store-fullness lever is
configuring an MCP server / n8n endpoint, so the banner
says exactly that and the button opens Settings straight
to the editor. Banner dismisses automatically the next
render after a federated entry appears.
Coverage stand for translation: Doctor + Modules + sidebar
+ page titles + nav + Cmd+K labels are bilingual. Settings
dialog, Approvals cards, Audit drilldown, MCP/n8n editors,
Store search hint + filter chips remain English-literal.
The infrastructure (ARB plumbing, generator hookup,
locale-notifier) means each follow-up surface is a
one-line ARB edit + one call-site swap.
Signed-off-by: flemming-it <stefan.a.flemming@googlemail.com>
Replaces the Store-only DE/EN toggle with an app-wide one
parked in the sidebar footer next to the theme button.
Pressing it flips every translated string at once: nav
labels, page titles, common buttons, the bilingual
store-index content.
Implementation:
- Adds `flutter_localizations` + `intl` to pubspec, plus
`flutter.generate: true` so `flutter gen-l10n` runs in the
build pipeline.
- ARB sources at `lib/l10n/app_en.arb` and `app_de.arb`. The
EN file is the template; DE carries the German strings.
Initial coverage: navigation, common buttons, page titles,
channels / store / audit / modules / approvals headers,
hub-unreachable copy, MCP + n8n panel headers + hints.
Rest of the UI strings are still English-literal — those
fall in incrementally as we touch each surface.
- Generated `AppLocalizations` lives at
`lib/l10n/app_localizations*.dart` (regenerated via
`flutter gen-l10n` on every ARB edit).
- `StudioAppState` gains `localeNotifier` alongside
`modeNotifier`; persisted via SharedPreferences key
`locale.code`.
- Sidebar `_LanguageToggle` reads/writes through the
notifier. The Store's per-page locale state is gone:
`_locale` now reads `Localizations.localeOf(context)
.languageCode`, so the bilingual store-index content
follows the global setting without a second toggle.
- `_NavPage.label` becomes `_NavPage.id` + `labelOf(context)`;
Cmd+K palette and Sidebar both read the localized label.
Out of scope this iteration: localizing the remaining
~80% of UI strings (Settings dialog labels, Store search
hint, error messages). Those land incrementally — the i18n
infrastructure now means each is a one-line ARB edit + one
call-site swap.
Signed-off-by: flemming-it <stefan.a.flemming@googlemail.com>
