I looked into buying a Tempest but then saw the local map of existing installations at for example here, and there are already hundreds if not thousands of devices installed in my area. It makes no sense to buy one if I can get enough data from the existing ones.
As a thought experiment on using ChatGPT to do most of the programming for me, I just spent a few hours prompting and driving ChatGPT4o to create a python script to obtain the data for any weather station in the world, without requiring a login or token. It gets the same data as other approaches do.
It formats it in json then uploads to an MQTT server for use in HA. The script handles parameters and config files, interactively lets you select a station if there are several to choose from, and a few other things.
I then used ChatGPT to create a myriad of MQTT sensors and templates and a weather dashboard. I also wanted a windrose, so the script generates windrose data as well for use in Grafana. The grafana graphic is then iframed into my weather dashboard.
Obviously the main problem with this approach is that it depends on a website remaining unchanged. Any major revisions will likely break the script. But for now, it works, generates the same 20 or so sensors, and works fine.