Recently I run into a problem with collecting metrics from python app running in muliple processes (using uvicorn). When you have these multi-process applications, any of the multiple workers-processes can respond to prometheus’s scraping request. Each worker then responds with a value for a metric that it knows of.

I faced with the problem with type coercion in Union when resulting type depends on the order of types in annotation:

With Pydantic we can build full-fledged two-way mapping using aliases and allow_population_by_field_name setting in Config

For couple of years I’ve been using pip-tools for managing requirements but recently I faced with a problem of splitting requirements for dev and production environments. The problem was that if we just split requirements and compile them separately we will end up with different versions.