The WeatherFlow PiConsole is a Python console that displays the data collected by a WeatherFlow Tempest or Smart Home Weather Station. The console uses either the WeatherFlow REST API and websocket service or the local UDP connection to stream data from your station in real time, including the 3-second rapid wind updates. In UDP only mode, the console requires no connection to the internet once installation is complete.
The console is fully supported for Raspberry Pi 3 Model B/B+, Raspberry Pi 4 and Raspberry Pi 5 running the 64 bit version of Raspberry Pi OS Trixie or the 32/64 bit version of Raspberry Pi OS Bookworm. It can be run on earlier models, but no direct support is provided for these systems. It is not compatible with Raspberry Pi Zero or Zero W. For full compatibility details, see below.
For a list of supported features and screenshots of the console in action, please checkout the WeatherFlow community forums: https://community.weatherflow.com/t/weatherflow-piconsole/20083
https://weatherflow.com/tempest-weather-system/
https://community.weatherflow.com/
This fork adds an optional Almanac interface: a glance-first redesign of the same station data, built to be read from across a room on a wall-mounted 7-inch screen. It runs on the same Raspberry Pi off the same feed and leaves the classic six-panel console untouched.
The Almanac (this fork) reads at a glance and finds room for things the classic layout can't. Here, heavy rain is falling through the intensity gauge, a 7-day outlook runs along the foot of the page on one shared temperature scale, the falling barometer has already switched its outlook to rain — and a quiet "Rain tomorrow" sits under the headline (that line appears only when tomorrow is a story: thunderstorms, snow, rain, wind, or fog; ordinary days say nothing). Thursday in the band wears the wind glyph — a dry day whose story is its gusts:
While it rains, etched rain falls into the rate gauge — hatching at two depths, its speed and density following the measured rate, into a sine-wave water surface whose swell and drift also track the intensity:
Everything live moves the way an instrument should: the vane swings the short way around the dial, and readings count to their new values instead of snapping. Forecast furniture holds still — on this console, motion means "now":
Winter is a first-class citizen. A pre-dawn January morning at −3.8 °F: the Winter Storm Warning shares the page with a compressed 7-day outlook (snow glyphs, sub-zero lows on the shared axis), and the rainfall panel reads "Snow Likely" — the Tempest's haptic sensor cannot register snowfall, so in freezing weather with snow forecast, the console says so instead of pretending it is dry:
The classic six-panel console (the upstream default) for comparison:
What it changes:
- One dominant temperature and a plain-language forecast line, in place of six equal-weight panels.
- A barometer zone bar (Stormy / Change / Fair / Dry) and a 24-hour pressure barograph labelled with the day's high and low.
- Active weather alerts (US National Weather Service) in a single strip below the masthead, coloured by severity and collapsed to one line when several are active. The 7-day outlook stays on screen through an alert (compressed to highs and bars) — a winter storm warning can run for days, exactly when the week ahead matters most.
- Air quality (AQI) by the station's own latitude and longitude, with a short forecast so a rising smoke event shows before the number climbs.
- A 7-day outlook band: each day's low–high drawn as a bar on ONE shared axis for the week (so a cool-down is visibly a shorter, lower bar), condition glyphs beside the highs — sun, cloud, fog, rain, snow, thunderstorm, wind, and the compound marks wind-driven rain and blowing snow (a gusty wet day keeps its water but the drops slant to the engraver's driving-rain angle behind a wind curl) — precipitation probability only when it matters, the day's expected amount beside it in the station's own unit (dry days print nothing), and today's bar carrying a dot at the observed temperature.
- A one-line hint about tomorrow under the conditions headline ("Rain tomorrow", "Wind-driven rain tomorrow", "Freezing rain tomorrow", "Blowing snow tomorrow"…), led by tomorrow's own band glyph, that appears only when tomorrow is a story — its absence is the fair-weather signal.
- A rainfall rate gauge scaled by intensity rather than linearly — the five named bands (Very Light through Very Heavy) each take an equal fifth of the tube, so drizzle registers and a downpour doesn't pin the needle. Light rain never reads as dry: the Tempest's haptic sensor logs drizzle as an occasional trace minute with zeros between, so the rate carries a 10-minute window that bridges those gaps. Rain arrives and leaves like weather, not like a switch — the volume of falling drops counts up quickly and down slowly, column by column, toward the measured rate (so the sensor's wet/dry flip between minutes reads as a swell and a settle), the drops shorten and slow as it eases, and the last column leaves through a fade.
- A wind panel that resolves to one current reading, with a bolder compass.
- Animated updates: values count up, the vane swings — and while rain falls, etched rain falls through the gauge into a waving water surface, with fall speed, density, swell, and drift all tracking the measured rate.
- Snow-aware: in freezing weather with snow in the forecast, a dry rain sensor reads "Snow Likely" rather than "Currently Dry" (the Tempest's haptic sensor cannot register snowfall).
- Day/night aware. After sunset the Sun & Sky panel becomes Moon & Sky (phase, illumination, moonrise/set).
Both extra data sources degrade quietly. Weather alerts come from the US National Weather Service, so outside the US the strip simply stays hidden. Air quality is worldwide (Open-Meteo), and its panel hides itself wherever a reading isn't available. Neither one can stall the display: they are fetched off the main thread, keep their last good value through a network blip, and are marked stale rather than shown as current if the connection stays down.
Once the kiosk is running, the page is also reachable from any device on your local network — phone, laptop, tablet — without any extra software:
http://weather.local:8137
Replace weather with your Pi's hostname if it differs. The page polls
/wx.json every two seconds and renders the same live data the wall display
shows. A status endpoint is also available:
http://weather.local:8137/health
This is enabled by default in the systemd service via Environment=WFP_BIND=0.0.0.0.
If you want to restrict the server back to the Pi only (no LAN access), remove
that line from ~/.config/systemd/user/almanac-kiosk.service and run
systemctl --user daemon-reload && systemctl --user restart almanac-kiosk.
For live weather readings (temperature, wind, rain, pressure) yes — the console is built entirely around WeatherFlow's data formats and has no support for other hardware brands (Ecowitt, Davis, Ambient, etc.).
The three connection modes are all WeatherFlow-only:
| Mode | What you need |
|---|---|
| Websocket + REST API (default) | WeatherFlow Personal Access Token + any Tempest, AIR, or SKY device |
| UDP + REST API | WeatherFlow device broadcasting on your local network |
| UDP only | WeatherFlow device + serial number; no internet required after setup |
No hardware? Use a nearby public station. Many WeatherFlow owners share their stations publicly, and the console can read one of them. During first-run setup (Websocket + REST mode), when you say you don't own hardware, the wizard offers to find nearby public stations: enter an approximate latitude and longitude and it lists the closest ones by distance, each with its hardware type. Pick one and it fills in the station and device IDs for you, so you get full live readings with no hardware of your own. All you need is a free WeatherFlow account and a Personal Access Token (see below). The station belongs to someone else, so the feed stops if its owner makes it private or takes it offline; re-run the wizard to choose another.
What works without a Tempest: the Almanac's supplementary panels — air
quality, weather forecasts, and astronomy — pull from public APIs keyed only on
latitude and longitude. If you set those manually in wfpiconsole.ini, those
panels display correctly even with no hardware attached.
Trying it before you buy: the first-run wizard offers a "blank config"
option that installs a minimal wfpiconsole.ini and starts the console
immediately. The weather readings will be empty, but you can see the layout and
verify the Pi setup is working. Run wfpiconsole start, choose the blank
config option at the prompt, then edit ~/wfpiconsole/wfpiconsole.ini manually
to add your station details once your hardware arrives.
To get a WeatherFlow Personal Access Token, go to tempestwx.com/settings/tokens. Your station ID and device IDs are under the WeatherFlow app: gear icon → Stations → [your station] → Status.
By default the AQI panel pulls from Open-Meteo (a CAMS satellite model — no account needed). For readings that match airnow.gov exactly, configure a free WAQI token: it switches the source to the nearest EPA/AirNow monitoring station.
During install or update the script asks for the token automatically. If you skipped it or want to change it later, run:
wfpiconsole configure-aqi
That command prompts for a token, validates it against the WAQI API, and saves
it to wfpiconsole.ini. Leave the field blank to clear an existing token and
revert to Open-Meteo. A free token takes under a minute to obtain at
aqicn.org/data-platform/token.
Two ways to run it:
- HTML kiosk (recommended). The console runs headless as a data engine and
chromium --kioskrenders the interface, pixel-identical and low-power (~½ core total on a Pi). A systemd-supervised watchdog keeps the whole chain alive: it relaunches any component that dies, restarts the engine if the feed goes stale (a hung websocket the classic UI would sit in), and re-checks for a wedged render. It self-heals: I confirmed that by killing the watchdog and watching systemd bring the display back on its own. Setup, management, and revert steps are indesign/almanac/kiosk/README.md. - Native Kivy layout. Set
[Display] LayoutStyle = almanacin the config.
For the architecture, the wx.json data contract, and upgrade notes, see
design/almanac/ARCHITECTURE.md.
The fork's data pipeline (the observation parser, the wx.json emitter, and a
merge-safety guard) has an offline pytest suite that needs no Kivy, display, or
network, so paths like a lightning strike are verified without waiting for real
weather:
python3 -m venv venv-test
venv-test/bin/pip install -r requirements-dev.txt
venv-test/bin/pytestCI runs the same suite on GitHub Actions (Python 3.11) on every push and pull
request. One guard test compares this fork's CurrentConditions class against
the upstream project and fails on any divergence, so upstream fixes keep merging
cleanly.
The Almanac UI (this fork)
Viewing the Almanac remotely
Do I need a WeatherFlow Tempest?
Air quality source
Compatibility
Installation Instructions
Update Instructions
Auto-Start Instructions
Advanced: Custom Panels
Advanced: Device Replacement
Advanced: Windows Installation
Credits
The console is fully supported for Raspberry Pi 3 Model B/B+, Raspberry Pi 4 and Raspberry Pi 5 running the 64 bit version of Raspberry Pi OS Trixie or the 32/64 bit legacy version of Raspberry Pi OS Bookworm. It can be run on earlier models, but no direct support is provided for these systems. It is not compatible with the 32 bit version of Raspberry Pi OS Trixie or Raspberry Pi Zero or Zero W. While the console is compatible with Raspberry Pi 3, the graphics hardware on this model is ageing and performance of the console can be sluggish. It is recommended to use a Pi 4 or above. The console is not compatible with Raspberry Pi OS Buster.
The console is compatible with the Raspberry Pi Official 7 inch Touchscreen or other HDMI equivalents. Note, screens that attach solely to the GPIO pins (SPI) are not compatible and the console will not start.
The console is fully supported on laptops and PCs running Ubuntu 22.04 LTS or later, or the desktop version of Raspberry Pi OS. It will run on other debian-based operating systems with Python version 3.9 or above, but no direct support is provided for these environments.
The installation of the WeatherFlow PiConsole is fully automated, and can be started from the terminal with a single command. The automated installation should take no longer than 10 minutes.
The automated installer assumes you have already successfully setup your Raspberry Pi and have installed Raspberry Pi OS with Desktop, or you ar running on a PC with Ubuntu 20.04 or later or Raspberry Pi OS installed. For a Raspberry Pi you should have also attached the touch screen, and have either a keyboard and mouse attached directly to the Pi, or have accessesd the Pi remotely through SSH/VNC. If you are starting from scratch with a Raspberry Pi, the documentation should help get you started:
The WeatherFlow PiConsole can be installed quickly and conveniently with the following command:
curl -sSL https://peted-davis.github.io/wfpiconsole | bash
Piping a command directly to bash is controversial, as it prevents the
user from reading code that is about to run on their system. If you are worried
about the contents of the installer, please examine the first
and second
installation files in the Git repository before running the install command.
When the console is run for the first time, you'll first be asked whether you want to install a blank configuration file for demonstration purposes or advanced setup. You can use this option if you wish to try out the console before your WeatherFlow hardware has arrived, or if you are a power user and wish to configure the console manually rather than using the configuration wizard. For most users, the advanced installation option is not appropriate and the default option of 'no' should be selected at this prompt. If you choose to install a blank configuration file, the console will start but no data will show unless you edit the configuration file manually.
You will also be prompted to specify your preferred connection type: Websocket and REST API (default), UDP and REST API, or UDP only. For UDP only you will be required to manually enter futher information about your station (location, name, elevation etc.). For Websocket and REST API or UDP and REST API you will be asked to enter a WeatherFlow Personal Access Token and a CheckWX Aviation Weather API key. The Personal Access Token is required for the PiConsole to access the data from your station, and the CheckWX API key is required to download the closest METAR information to your station location.
A Personal Access Token can be generated, viewed, and deleted here: https://tempestwx.com/settings/tokens, and a CheckWX API key can be obtained by registering for a free account here: https://www.checkwxapi.com/auth/signup
Once you have a Personal Access Token and registered with CheckWX (if required), go ahead and run the console for the first time using:
wfpiconsole start
Depending on the connection type you select, you'll be asked to enter the API keys you have just generated above, as well as information about your station. This includes your station ID and device IDs for your AIR, SKY, or TEMPEST modules. To find this information either open the WeatherFlow app or view your station page in a web browser. Click on the gear (settings) icon -> Stations -> [Station name] -> Status.
If you don't own WeatherFlow hardware, the station-ID step (Websocket + REST mode) instead offers to search for a nearby public station. Answer yes, give an approximate latitude and longitude, and pick from the distance-sorted list; the console fills in the station and device IDs for you. See "Do I need a WeatherFlow Tempest?" above for the details and caveats.
If all goes smoothly the console should automatically add the extra information it needs to your configuration file and then start running. You should not need to enter this configuration information again.
Congratulations, you have installed the PiConsole for the Weather Flow Tempest and Smart Home Weather Stations.
By default the PiConsole will run in full screen mode. Fullscreen mode can be
disabled in Menu -> Settings -> Display. In this case the console will use the
dimensions specified in the configuration file (wfpiconsole.ini), which
can be changed manually. Please note that extreme changes to the aspect ratio
will result in text fields running into one another. Under Settings -> Display
there are also settings to show/hide the cursor and show/hide the window border.
Please note that you cannot use SSH to start the console remotely. Instead for remote access it is recommended to setup VNC (https://www.raspberrypi.org/documentation/remote-access/vnc/). Note there are currently issues using Real VNC (the default VNC provider on Raspberry Pis) with the latest version of Raspberry Pi OS (Bookworm): https://help.realvnc.com/hc/en-us/articles/14110635000221-Raspberry-Pi-5-Bookworm-and-RealVNC-Connect
The WeatherFlow PiConsole can be updated quickly and easily with the following command:
wfpiconsole update
The update process will retain your existing user settings, but may prompt for input from time to time in order to add new functionality. Once the update has finished, restart the console using:
wfpiconsole start
The WeatherFlow PiConsole can be configured to run automatically when the Raspberry Pi powers up. To enable the console to start automatically, run
wfpiconsole autostart-enable
To stop the WeatherFlow PiConsole from starting automatically, run
wfpiconsole autostart-disable
If you are going to use the auto-start method, it is highly recommended that you can SSH into your Raspberry Pi, as the console can only be stopped using the stop command or a hard shutdown:
wfpiconsole stop
The console is distributed with 7 built-in panels to display weather, forecast and astronomical information. For advanced users, custom panels can be created allowing the data display to be customised, or additional data sources to be integrated into the console. Custom panels should not be overwritten when the console is updated.
The custom panel templates are contained within the ~\wfpiconsole\user folder.
To use the custom panel feature, you first need to rename customPanels.kv.tmpl
to customPanels.kv and customPanels.py.tmpl to customPanels.py. An example
panel called "BigTemperature" is included as an example, and will be loaded the
next time you start the console.
In the customPanels.py file you must create two classes per custom panel called:
[panel_name]Panel and [panel_name]Button. "panel_name" can be whatever you want,
but you must add the two classes that end with Panel and Button per custom panel.
The classes should be empty (just add pass under the class name), unless you want
to add methods to your custom panel to control its behaviour. The classes required
for the "BigTemperature" panel can be used as examples.
In the customPanels.kv file you can define the layout of the panel. You need to
add the two class names that you defined in customPanels.py surrounded by left and
right angled brackets: <>. Again, you can see the "BigTemperature" panel in
customPanels.kv.tmpl as an example. For the Button class, you can change the text
attribute under PanelButton: to set the name of the panel that will be displayed in
the bottom bar of the PiConsole. Otherwise leave this class unchanged. For the Panel
class, the panel title is defined by the _title attribute under PanelTitle:. This can
be different to the name of the panel that is displayed in the bottom bar. Otherwise
you are free to define the layout however you want using in-built or custom Kivy
widgets (https://kivy.org/doc/stable/api-kivy.uix.html).
Occasionally it may be necessary to replace your Tempest device due to hardware
failure. Depending on how the replacement Tempest is added to your existing station,
the Tempest device ID and serial number may change. If this is the case, the
wfpiconsole.ini file needs to be updated with the new device ID and serial number.
The .ini file can either be edited directly, or if you are not comfortable editing
the .ini file, you can delete it and then restart the console. You will be taken
through the steps to generate a new .ini file with the updated device ID and serial
number.
When a device is replaced, the total monthly/yearly rain accumulation displayed in the
console will also reset to zero as these fields are calculated directly from the
total rain accumuluation recorded by the new device (which is naturally zero as
the device is brand new). To retain the correct values, it is necessary to switch
the console to use the Tempest Statistics API endpoint using Menu -> Settings ->
System -> Statistics API endpoint. By default this option is disbaled as it
results in a small loss of accuracy through rounding errors. Therefore it is not
recommended for use unless you have replaced a device within the last calendar
year. At the end of a calendar year, the Statistics endoint can be switched off.
Although not officially supported, use the following step-by-step instructions to install and run the WeatherFlow PiConsole on Windows.
-
Download and install the Python 3.11.5 version of Miniconda for Windows (a lightweight Python interpreter): https://conda.io/miniconda.html
-
Once Miniconda is installed open the ‘Anaconda Prompt’ program.
-
In the Anaconda prompt, run:
python -m pip install --upgrade pip
- Once that process has finished, run:
python -m pip install websockets numpy pytz tzlocal ephem packaging pyOpenSSL certifi
- Once that has finished, install Kivy using
python -m pip install kivy[base]
This is the GUI library that drives the console.
- Once Kivy is installed, run the following commands in order in the Anaconda Prompt. This will install the WeatherFlow PiConsole.
cd && mkdir wfpiconsole && cd wfpiconsole
curl -sL https://api.github.com/repos/peted-davis/WeatherFlow_PiConsole/tarball -o PiConsole.tar.gz
tar -xvf PiConsole.tar.gz --strip 1
del /f PiConsole.tar.gz
- You’re almost there now! You can start the console using
python main.py. As this is the first time you have run the console, you’ll be asked for some API keys. Details of what you need can be found under "Configure and Run WeatherFlow PiConsole" in the Installation Instructions.
Many of the graphical elements in the console are based on the Weather34 Home Weather Station Template (https://www.weather34.com/homeweatherstation/) copyright 2015-2021 Brian Underdown. The Weather34 Home Weather Station Template is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.