Collecting weather data from MQTT in WeeWX using a custom driver

Posted on 26th of January 2025.

I’ve been running a small weather station setup at home and using WeeWX to visualize the data. If you haven’t heard of WeeWX, it’s an open-source weather software that takes in data from various weather stations and displays it in a slick dashboard you can view in your browser. It’s written in Python and pretty easy to extend — which turned out to be a lifesaver for my use case.

The problem

My weather station doesn’t output data in a way WeeWX natively understands. Instead, I’ve got a little custom setup that listens to 433 MHz signals (you could also do 868 or 915 MHz) using a Software Defined Radio (SDR) dongle. Then, with a custom script I wrote — 433-mqtt-bridge — I publish that data to an MQTT broker.

The idea is that MQTT acts as the hub: the data comes in from the SDR, gets interpreted and published to MQTT topics, and then can be consumed by whatever I want — like Home Assistant, Grafana, or in this case, WeeWX.

But out of the box, WeeWX doesn't support pulling data from MQTT. So, I wrote a driver for it.

My Custom MQTT Driver for WeeWX

I put together a custom Python driver for WeeWX that lets it subscribe to MQTT topics and treat the incoming values like sensor readings. Here's the driver on GitHub Gist:

👉 View the driver code

It's super straightforward, and it integrates nicely with my existing MQTT setup. It reads the most recent values from specified topics, does some basic conversion, and then passes that into WeeWX like any other station hardware driver would.

What It Supports

This driver expects metric values and supports a range of weather parameters:

Temperature
Humidity
Wind speed & gusts
Wind direction
Rain & rain rate
Pressure
UV
Light

If you want to convert values (like from °F to °C or mph to km/h), you can tweak the multiplier section in the driver (check around line 70 in the code).

How to Install the MQTT Driver

Download the driver file
Grab the .py file from the Gist link and drop it into your weewx/drivers directory. On a typical Linux install, this is:
```
/usr/share/weewx/weewx/drivers/
```
Install the MQTT Python package
The driver uses the paho-mqtt library to subscribe to MQTT topics, so make sure it’s installed:
```
pip install paho-mqtt
```

Update your WeeWX config
In your weewx.conf file (typically found at /etc/weewx/weewx.conf), update the [Station] section and add a new [Mqtt] section. Here’s an example:

[Station]
   # ... your other configuration
   station_type = Mqtt

[Mqtt]
   mqtt_broker_host = localhost
   mqtt_broker_port = 1883
   temperature_topic = weatherstation/temperature
   humidity_topic = weatherstation/humidity
   gust_speed_topic = weatherstation/gustspeed
   wind_speed_topic = weatherstation/windspeed
   wind_direction_topic = weatherstation/winddirection
   uv_topic = weatherstation/uv
   light_topic = weatherstation/light
   rain_topic = weatherstation/rain
   rain_rate_topic = weatherstation/rain_rate
   pressure_topic = weatherstation/pressure
   hardware_name = My Weather Station Hardware
   driver = weewx.drivers.mqtt

Of course, update the topic names if you're using something different in your MQTT setup.

Restart WeeWX
After making all the changes, just restart the WeeWX service and it should start pulling in data from your MQTT broker.
```
sudo systemctl restart weewx
```

Why This Is Cool

What I like about this setup is that it completely decouples the weather station hardware from the visualization layer. I can use the same data stream in WeeWX, Home Assistant, and even feed it into a time-series database like InfluxDB for custom dashboards in Grafana.

This also makes it way easier to support random off-brand or legacy weather stations that don’t have native WeeWX support — as long as you can decode the signal and push it to MQTT, this driver should work.

Gotchas & Customizing

The driver expects metric units. If your MQTT data is in imperial or something else, you’ll need to adjust the conversion factors in the code.
It's a simple poll-every-X-seconds kind of model (based on WeeWX’s usual loop), not an event-driven thing. That said, it’s plenty responsive for most weather data needs.
You might need to fiddle with your loop_interval or observation timestamps if you’re seeing weird gaps or timing issues.