# Copyright 2026 Kemal Yaylali # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """The dashboard: five tabs, one viewport, no scrolling. Layout contract. The page is a fixed three-row grid pinned to the viewport height: header, tab bar, then a content region that takes the remaining space and never overflows the fold. Each tab lays its panels out on an internal grid sized in fractions of that region, so nothing depends on content height. Where a panel genuinely holds more than fits (the daily records table, the methods prose) that individual panel scrolls internally while the page frame stays put. Below 1024 px the constraint is released, because pinning five panels into a phone viewport produces unreadable eight-pixel type, and a phone user expects to scroll anyway. Visual language carries over unchanged from the previous station page: slate-950 ground, glass panels, Jakarta for prose and JetBrains Mono for anything numeric. The one new structural device is the tab bar, and it earns its place. Five distinct questions (what is it doing, what will it do, what did it do, is the model any good, how does it work) were previously one long scroll where the important things sat below the fold. The signature element is the estimator internals panel on the Live tab. Most weather dashboards show numbers. This one shows the state estimator working: self-heating coefficient, Kalman innovation, novelty distance and drift pressure, all ticking at 2 Hz. It is the part of the system that is normally invisible, and watching a filter converge is the most honest possible demonstration that there is real machinery underneath. """ DASHBOARD_HTML = r""" Ashvale Station

Ashvale Station

- · 0 d logged · k=-

health --:--:--
Temperature KALMAN
--°C
rate--
raw / cpu--
Humidity HTS221
--%
dew point--
depression--
Barometer MSL
--hPa
tendency--
character--
Sky
--clr
cloud index--
sun / cct--

Rolling window

2 s stream

Estimator internals

what the filter is doing right now

self-heating k--
cpu offset--
Removes the SoC bias. Calibrate it on the Models tab.
novelty d²--
drift pressure--
Novelty is a multivariate departure from the recent norm. Drift reaching 100% queues a retrain.
wet bulb
--
vpd
--
abs hum
--
heat idx
--
solar el
--
pitch
--
roll
--
yaw
--
compass
--
accel z
--

Observed and forecast

shaded band is the 90% conformal interval

Conditions ahead

Zambretti prior + online logistic

--
Z=- · -
rain probability--
prior -learner -trust -
Was it wet in the last hour?
0 confirmed observations
pressure, last 24 h --

The only signal here that sees past your walls. Its slope, not its level, is what drives the forecast above.

Seven day outlook

climatology plus decaying anomaly, not a synoptic forecast

warming up
Export CSV

-

Records

Verification scorecard

skill above zero means it beats persistence

targetleadMAE persistskillcover np/c/l

No matured forecasts yet. Rows appear as each horizon reaches its validity time: 15 minutes first, 24 hours tomorrow. The p/c/l column is the ensemble weight on persistence, climatology and the learned model.

Calibration and state

Trusted thermometer reading
Recursive least squares on the self-heating coefficient. One good reading is enough.
Storage tiers
Precipitation coefficients

Station log

How it is wired

select a stage to read its rationale

"""