Completes milestone 3: frames render over the map, play as a loop, and can be
scrubbed, with the legend drawn from the manifest rather than a constant that
could drift from what the worker actually rendered.
The overlay alternates two MapLibre image sources. Updating one source in place
flickers, because the layer briefly shows a half-written texture; adding every
frame as its own layer avoids that but pins them all in GPU memory, and twenty
512x512 RGBA frames is about 80 MB. Two buffers cost the same whether the
timeline holds six frames or sixty. Platform-channel work is serialised because
`show` is called faster than the round trip completes during playback, and
overlapping updates would swap visibility out of order and strobe.
Prefetching loads a window around the playhead, nearest first and forward
before backward, since playback moves forward and that frame is needed
soonest. `FrameCache` is byte-budgeted rather than entry-counted because frame
size tracks how much precipitation is on screen, and it evicts by distance from
the playhead: plain LRU would keep frames the prefetcher touched a moment ago
even after the playhead moved to the far end of the timeline.
Also adds DpcRadarSource, which reads published frames from our CDN and never
from the DPC API. Without it, the `dpc` adapter would have had to fall back to
mock, putting demo frames on screen under the label of live data — exactly the
confusion the adapter split exists to prevent. It now fails naming the missing
setting instead.
Running it on the emulator caught three things the tests had not:
- The notifier wrote to `state` from inside `build()`, which Riverpod rejects
as an uninitialised provider. That broke startup, not just tests.
- Eight-month-old demo frames rendered as "Aggiornato 342535 minuti fa". The
age formatter now steps up to hours and days.
- Demo mode sat permanently behind a stale-data warning and so never showed the
working state it exists to demonstrate. MockRadarSource now shifts the
bundled timestamps onto the present, leaving images, order and spacing
untouched, so the timeline behaves exactly as it would on live data.
Corrects docs/stack-decisions.md, which described a disk cache that was not
built: mock frames already live in the asset bundle, so a disk layer belongs
with the network adapter where it would save a real request.
Verified: analyze clean, 129 tests passing, and on the emulator the loop
advances, wraps, and reports "Aggiornato ora".
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The AVD's quickboot snapshot had grown to 2.6 GB, which took this machine down
to 1.4 GB free. Saving a minute of boot time is not worth that here, so cold
boot is now the default and the snapshot directory is gone.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Running the app on an emulator for the first time showed two problems that the
unit tests could not have caught, because both are about what the pixels
actually do.
The configured initial zoom of 7.2 puts the viewport entirely inside the
region on a 411dp-wide screen, so the map rendered as a flat expanse of fill
colour with no outline, border or reference point anywhere in view — it read as
a failure to load. A single zoom number cannot frame a region on both a phone
and a tablet, so the camera now fits the region bounds once the map is ready,
which is correct at any screen size. The configured zoom stays as the starting
position before the fit.
The attribution bar sat behind the system gesture pill. A credit that is
covered by the navigation bar is a credit that is not being displayed, which is
the thing ODbL and CC BY-SA actually require, so it now sits inside a SafeArea.
Also records how to run and screenshot the app on the `nuvolari` AVD, and notes
the pre-existing broken `pixel_7_-_api_35` AVD that is holding 4.8 GB.
Verified on the emulator: the region outline is framed with margin, the
attribution bar is clear of the pill, and the Sources screen shows the
disclaimer, all four credits with their licences, the "no licence stated" line
for ARPA and the share-alike note.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Puts a MapLibre map on screen framed on the region, with the credit
obligations that come with showing third-party data satisfied structurally
rather than by remembering to add a label.
The base map style comes from MAP_STYLE_URL, and with no key configured the
app generates a fallback style from the region's own bounding box: a flat
background and the extent outline, no network sources at all. That keeps a
fresh clone runnable offline without pulling in a boundary dataset that would
carry its own licence, and it deliberately looks like a placeholder so it is
not mistaken for a finished map.
Attribution is driven by what is actually rendered. The bar lists the credits
for the active sources only, because crediting OpenStreetMap while showing the
fallback style would be a false attribution, and it says plainly when no base
map is configured. It sits below the map rather than floating over it so no
map control or gesture overlay can occlude a credit that the ODbL and CC BY-SA
terms require to be visible.
The Sources screen leads with the independence disclaimer, before the sources
it qualifies, so a reader who stops after the first screenful has still seen
it. Sources with no stated licence — ARPA publishes none for the alert
bulletin — say so explicitly rather than being shown bare or given an invented
one. A separate note explains that the rendered radar frames inherit CC BY-SA
from the DPC source data.
Also declares the INTERNET permission in the main manifest: Flutter injects it
into the debug and profile manifests only, so a release build would otherwise
fail every request on device.
Verified: analyze clean, 66 tests passing, appbundle builds with the native
MapLibre plugin. Not verified visually — this machine has no Android device or
emulator image, so nobody has watched the map render.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Creates the app skeleton and the seam everything else in the project hangs
off: region-specific data lives in an asset file, not in code, so adding a
region later is a new JSON file rather than a refactor.
Notable choices:
- applicationId and namespace are it.nuvolari.app rather than the doubled
it.nuvolari.nuvolari that `flutter create` produces, with the Kotlin package
and the iOS bundle identifiers moved to match.
- SDK levels are pinned instead of inherited from `flutter.*`. Google Play
requires API 36, and that is a release blocker rather than something to let a
Flutter upgrade change silently. minSdk 24 is the highest floor the planned
dependencies impose.
- Riverpod and Dio-free for now, no code generation: see docs/stack-decisions.md.
- Italian is the l10n source language, so app_it.arb is the template rather
than a translation of an English original.
The region parser rejects rather than repairs. An inverted bounding box, a map
centre outside its own bounds, an unknown adapter name, a duplicate zone code
or an empty attribution list all throw with the offending field named. Each of
those would otherwise fail silently and visibly wrong: a swapped latitude and
longitude renders the radar in the wrong place, an unknown adapter falling back
to mock would show demo frames where live data was expected, and a missing
attribution is a licence violation rather than a cosmetic gap.
Tests run against the asset that actually ships and against a captured copy of
the live ARPA CAP feed, so the eleven zone codes in the config are checked
against the eleven the feed really emits rather than against a list retyped
from documentation.
Verified: dart format clean, flutter analyze 0 issues, 50 tests passing,
flutter build appbundle --debug produces an AAB with applicationId
it.nuvolari.app, minSdk 24, targetSdk 36.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Sets up the Nuvolari monorepo layout (app/, backend/, docs/, tool/) with the
groundwork that does not depend on the Flutter toolchain: gitignore covering
Flutter, Python and every secret file shape; env.example.json as the template
for --dart-define-from-file; a verification script that skips stages whose
target does not exist yet so it is runnable from day one; and a CI workflow in
GitHub Actions syntax so it runs unchanged on Gitea or GitHub.
The documentation records facts verified against the live services rather than
restated from the brief. Two of them change the design:
- DPC VMI rasters are 1200x1400 Float32 on a 1 km grid in a custom projection
centred on Italy, not EPSG:4326 or EPSG:3857, and their GeoKeys are
internally inconsistent. Reprojection is mandatory and the source CRS must be
read from each file rather than hardcoded.
- The ARPA CAP feed carries six level values, not four: BIANCO for the
avalanche scale out of season and "-" for no data. Collapsing either into
VERDE would report "no alert" where the bulletin reports "not assessed".
Also documents why the frames we publish inherit CC BY-SA from the DPC source,
and why rain notifications subscribe to cell topics from the device so no user
location ever reaches a server.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Alby96andClaude Opus 5