- /api/v2/place still called grid.in_north_america(), which the worldwide-
coverage change removed — every request raised AttributeError (500). The
failure was invisible because compare.js treats the call as best-effort,
so the instant chip-naming feature was silently dead. Drop the dead guard;
the lat/lon Query validators already bound the inputs.
- /api/v2/day was the only data route not wrapping get_history in
_weather_fetch_error, so a rate-limited cold cell returned a raw 500
instead of the clean 503 the other routes emit.
- calendar: the #calendar pointerleave handler was re-registered inside
attachHover on every render — and the comfort slider re-renders per input
tick, stacking dozens of copies. Register it once at module scope next to
the matching document-level dismiss handler.
- getJSON: check res.ok before parsing the body as JSON and fall back to the
status line, so a non-JSON error body (a proxy 502 HTML page) reads as
"Request failed (502 Bad Gateway)" instead of a JSON parse error.
- Widen the content column in steps at 1680/2400/3400px so 1440p-4K monitors
get real layout instead of a 1200px strip; header brand aligns with it.
- Calendar: months flow as a responsive multi-column grid (year-at-a-glance
on desktop, single column on phones); the time-filter panel spans the
content width as one row (Range | Seasons/Years/Weather) on desktop;
category-totals strip widened to 720px.
- Day: metric ladder cards lay out 2-up from 1000px (3-up at 4K).
- Weekly: the calendar CTA is a compact button (sub-line removed); the trend
chart's drawing width now tracks its container (~1.5 CSS px per SVG unit,
redrawn on resize) so wide screens gain plot room instead of magnification.
- Compare: scope the rank-bar segment colors to .cmp-seg — the bare
.cmp-comfort rule was painting a green box behind the comfort slider.
- Calendar/Compare From-To month inputs open the native month picker on
click (input.showPicker, ignored where unsupported).
- CLAUDE.md: require validating design changes at mobile/tablet/1080p/2K/4K.
Remove the US+Canada bounding box so every endpoint accepts any lat/lon.
The grading pipeline was already global-ready (ERA5 archive, timezone=auto,
day-of-year climatology), so opening it up is mostly deleting the guard —
plus the edge cases that only exist once the whole globe is in play:
- grid.py: snap() wraps longitude into [-180, 180) and clamps latitude, and
cell centers are normalized so the polar row and the cells straddling the
antimeridian always report valid coordinates to the weather/geocoding APIs.
snap() and from_id() now share one _cell() builder, making id round-trips
exact by construction (verified with a 300k-point global sweep).
- nav.js: neighbor-cell prefetch skips rows past the poles and wraps
longitudes across the dateline instead of sending out-of-range queries.
- Nominatim reverse geocoding requests accept-language=en so place labels
render in one script worldwide (matching the forward geocoder).
- mappicker: search suggestions are no longer filtered to US/CA, the
placeholder and default map view are worldwide.
- calendar: season filter labels flip for southern-hemisphere locations
(Dec-Feb shows as Summer); the underlying month groups are unchanged, so
saved filter selections keep meaning the same months.
Verified end-to-end on a scratch server: Tokyo and Sydney grade with real
labels, a Fiji cell on the antimeridian's east edge builds and serves warm
hits from the derived store, and prefetch=1 on a cold cell still answers
204 without spending weather-API quota.
* Picker map: smaller city labels, darker/bolder roads (split tile layers)
Split the CARTO Voyager basemap into two raster layers so label size and road
weight are independent: a labels-free base upscaled via tileSize 512 + zoomOffset
-1 (bold, prominent roads) plus a labels-only layer at natural size on top (small,
crisp city names). The darken/saturate filter now targets the base layer only
(.mp-base: brightness .92, contrast 1.12), deepening the roads while leaving label
text at full contrast.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B3Q2EkHHnTUX2BfhV5q1Zc
* Location label: show coordinates immediately, then resolve to place name
On selecting a location, write the raw coordinates to the location label right
away, so the user sees the picked spot instantly instead of a stale/blank label.
The existing post-fetch render already overwrites it with the reverse-geocoded
"neighborhood, city, region" string (data.place), so the label now reads
coords → place name live.
- app.js / calendar.js / day.js: set coords in the selection handler before the
fetch; the render/initOnce overwrite is unchanged.
- compare.js: show coordinates through the pending + loading states (instead of
"Locating…") so each chip's coords are replaced live by the place name once
scored.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B3Q2EkHHnTUX2BfhV5q1Zc
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* Revert "Merge pull request #22 from griffemi/fem-orchid-palette"
This reverts commit f7d16238f2, reversing
changes made to 34e4fed1f0.
* Restyle location-picker map: themed CARTO basemap + branded pin
Replace the plain gray OpenStreetMap raster in the location picker with a
theme-aware CARTO basemap (no API key): sleek "dark matter" tiles under the
dark theme, colorful "voyager" tiles under the light theme, so the map
matches whichever is active. Swap Leaflet's default blue marker for an
accent-colored teardrop pin (the same glyph the Find button uses), and theme
the zoom controls + attribution to sit in the app's surfaces.
- mappicker.js: theme-aware CARTO tileLayer (retina, CARTO attribution);
L.divIcon "mp-pin" used for the marker
- style.css: .mp-pin styling, themed .leaflet-bar / attribution, map shadow
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B3Q2EkHHnTUX2BfhV5q1Zc
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
A shared units module in nav.js holds the active unit (persisted), converts
Fahrenheit API values for display, and notifies pages to repaint on a switch.
A segmented °F/°C control is injected into the top-right of every header
(left of the view tabs); Weekly, Day and Calendar convert every temperature —
cards, dense timeline, trend chart axis/labels, tooltips and the PNG export —
while wind, humidity and precip stay as-is. Compare keeps its Fahrenheit
comfort slider, so the toggle is hidden there.
Replace the clinical blue→green→red diverging temperature scale, the
green→navy rain ramp, and the orange accent with a warmer, more vibrant
"Orchid Sunset" palette: violet/lavender cool end, soft blush middle,
rose/magenta warm end, hot-pink accent.
Keeps the diverging structure grading relies on (cool reads cold, warm
reads hot) and updates every mirror of the literals so PNG export and the
self-contained subpages stay in sync:
- style.css :root tier + precip vars, accent, cta-cal gradient/shadows,
dry-metric toggle; light-theme overrides for the near-white middle tiers
so "Normal" stays legible on white
- app.js TIER_COLORS, chartPalette() fallbacks, dry-streak ramp
- calendar.js TIER_COLORS, PRECIP_COLORS, dry-streak ramp
- day.js TIER_COLORS, PRECIP_COLORS, DRY_COLOR
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B3Q2EkHHnTUX2BfhV5q1Zc
* Compress API responses and revalidate static assets instead of re-downloading
- Add GZipMiddleware (min 1 KB): the 2-year calendar JSON shrinks ~6-8x.
- Serve pages/assets with Cache-Control: no-cache instead of no-store, so
browsers revalidate via the ETag/Last-Modified that FileResponse and
StaticFiles already emit. Unchanged assets now cost an empty 304 rather
than a full transfer on every page navigation, while deploys still show
up immediately.
* Persist derived responses in SQLite so grading is computed once per cell, not per request
New data/thermograph.sqlite (WAL) holds what's derived from the raw parquet
records — finished grade/calendar/day/forecast payloads and reverse-geocode
labels — so the expensive work (notably the 2-year calendar grade_range, ~270ms)
becomes a ~5ms database read that survives restarts and is shared across views.
Raw parquet stays the source of truth; the store is a pure accelerator (every
reader falls back to recomputing on a miss, and deleting the db is a safe reset).
Freshness is token-driven, not clock-driven: each cached payload is validated by
a token encoding what it was computed from (payload schema version, the archive
record's end date, the recent-fetch stamp). The existing freshness drivers are
untouched — get_history still tops up the tail hourly and get_recent_forecast
still refetches hourly — and tokens are derived from what they return, so cached
payloads expire exactly when their inputs change. The same tokens double as weak
ETags: If-None-Match answers with an empty 304 without touching the payload.
- backend/store.py: derived-payload + revgeo tables, thread-local WAL conns,
every helper fail-soft.
- app.py: endpoints split into pure payload builders + HTTP/caching shells; the
in-memory 10-minute _CAL_CACHE is retired (superseded by the persistent store).
- climate.py: revgeo persisted through the store; recent_stamp() and
load_cached_history() (no-network read) helpers.
- backend/migrate.py + make migrate: idempotent, resumable backfill of the store
from existing parquet caches (default calendar span + latest-day detail +
revgeo, ≤1 throttled Nominatim call per unlabeled cell). Never fetches weather.
- grid.from_id(): rebuild a cell from its cache filename (migrate tooling).
* Add /api/v2/cell: one bundle carrying every view's payload
GET /api/v2/cell?lat&lon returns the grade, forecast, calendar (last 24 months)
and day (today) payloads in one response, each the exact payload its per-view
endpoint returns — built by the same builders and cached under the same
derived-store keys/tokens — paired with the etag that endpoint would emit. The
frontend can warm all views with a single request, seed its per-view cache from
the slices, and later revalidate each view individually with If-None-Match. The
bundle's own etag combines the slices', so an unchanged bundle is an empty 304.
prefetch=1 is a warm-only mode for neighbor-cell prefetching with a hard
guarantee: it never spends weather-API quota. A cell with no cached archive
answers 204, and only the history-derived slices (calendar + latest-day detail)
are built. At most one Nominatim lookup for a never-labeled cell.
* Frontend: IndexedDB response cache with stale-while-revalidate + bundle prefetch
The per-URL response cache moves from localStorage/sessionStorage (~5 MB quota,
which multi-year calendar payloads regularly blew through) to IndexedDB, with an
in-memory map in front. Entries carry the server's ETag, so anything stale
revalidates conditionally — unchanged data costs an empty 304 and a re-stamp,
never a re-transfer. On network failure the stale copy is served over an error.
getJSON gains an optional onUpdate callback opting into stale-while-revalidate:
the three views (weekly, day, calendar) now render a cached copy immediately —
spinners are delayed 150ms so warm loads never flash-blank — and repaint only if
background revalidation finds changed data. New data shows up the moment it
exists instead of waiting out a TTL.
Cross-view prefetch collapses from one request per view to a single /api/v2/cell
bundle, whose slices (exact per-view payloads + their etags) are seeded under the
URLs each view actually requests; the bundle call itself is conditional via a
remembered etag. Afterwards the 8 surrounding grid cells are warmed server-side
with prefetch=1 (never spends weather-API quota; staggered ≥1.1s for the one
possible Nominatim lookup each), so tapping nearby lands on already-graded data.
Legacy tg:* storage entries are cleared once; cache entries untouched for two
weeks are pruned on page load.
The Weekly location card joined coordinates, cell size, climatology range
and sample count into one dot-separated meta line. Stack them as one fact
per line so the important context reads top-down instead of as a run of
tags. Drop the '~N sq mi cell' wording from the Weekly/Day/Calendar meta
lines and the map hint.
Compare page (frontend/compare.{html,js} + /thermograph/compare route):
line up several places over a date range and rank which best matches a
comfort temperature. Per location it pulls the same daily record the
Calendar uses and, per day, takes a chosen temperature (daytime high /
daily mean / overnight low / feels-like) against the comfort target. A
day "hits comfort" when it lands within an adjustable band; otherwise it
counts as colder or warmer and the average miss is tracked. Results are
ranked by comfort-day share (tie-broken by the smaller typical miss),
with a diverging below/comfort/above bar and per-bucket stats. Comfort,
band and judged temperature re-rank instantly client-side; only the
location set or date range trigger a data load (shared calendar cache).
Feels calendar filter: drop the comfort-temperature slider and the
client-side felt-high/felt-low re-pick. The Feels metric now colors by
the server's combined feels-like value like the other metrics, so its
tab matches the rest of the metric selector.
Caption every category column with its name (two reserved lines so the
value figures stay aligned whether a name wraps or not), and add a
"Show count" checkbox in the strip header that swaps every share figure
for the raw day count (state persists in localStorage, re-renders in
place via a delegated listener on #cal-totals).
Also plot the neutral "Normal" median category as a real bar instead of
a bare baseline tick, so the strip reads as a full distribution; bar
heights now scale to the tallest category (Normal included).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Widen the per-category deviation bars (3px→10px) so each reads
clearly, and print sub-1% shares as e.g. "0.4%" instead of "<1%".
Claude-Session: https://claude.ai/code/session_019VP23wKmjS2ozk1g5a9g1Z
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Replace the stacked share bar + labeled percentage chips above the
calendar grid with a compact deviation strip: one thin status-colored
line per category (height scaled to the tallest non-median category),
with each share printed above it in that category's own status color.
The median tier — the neutral "Normal" center of the diverging
temperature scale — draws no line, just a faint baseline tick, so it
reads as the reference the other categories deviate from. Precip and
dry-streak have no natural median, so every category there keeps a line.
Percentage text is lifted toward the theme text color (color-mix) so
even the darkest/lightest tiers stay legible in both light and dark.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
