I went with Zola just because it's so much easier to setup imo

And for testing in general it will almost always default to testing the code instead of testing the implementation, you need to give a ton of context to get useful tests. I'd prefer it doesn't generate code at all in those cases.

Sugar and butter, as the elders willed it.

nedbatchelder.com/blog/202204/... !

Happens with every single headline being "This strange bird was found in a famous capital" when it's actually "raven was seen in London yesterday".

I haven't and don't plan to unless it becomes more popular, can't justify the time investment. For the schema problem, I parse my configuration into valid Python objects and also apply custom logic when needed, the end result can be a request to an API or another data file with sane defaults.

I'll be honest I don't really see how that setup with now another language thrown into the mix is better than plain data via YAML/JSON and Python.

If you find yourself moved and asking, how can I be part of this organization?! One answer is joining as a Supporting Member: psfmember.org/civicrm/cont...

Yes github.com/tox-dev/pypr...

Trust me I know, I need to write Jinja templates for Ansible and GH actions for other people. My go-to right now is to add CI/CD for the templates where I render them and run the rendered operations (when possible). Not amazing but catches a lot of issues that end-users would face.

Always and forever: mikehadlow.blogspot.com/2012/05/conf...

Nowadays I think using YAML is fine, just quote things. Use Python and restrict yourself to a sensible set of features, when you need more power. I can't bother with learning another config language tbh.

I remember using Vue and doing some custom things to update or remove components on un-mount but that's maybe a Vue thing.

How does that work when you have a large amount of nesting in components? Do you usually work from the leaf component calling its parent or the parent calling all the children?

$work is a Python shop but I'm interested in basically exposing inference to end users for translation and fast document lookup for example. There's a lot of integration for monitoring and safety but I could see lightweight LLM being served without a ton of Azure wrappers around it.

Is there any current advice to deploy with large cloud providers? Most cloud platforms will guide you using Docker or Kubernetes but they're pretty light on details otherwise.