Devops

My previous article recommended that one should reconsider using Python in production. However, there’s one category of use case where Python is the dominant option for running production workloads. And that’s data analysis and machine learning. Almost all bleeding-edge work in data analysis and machine learning, especially around LLMs, happens in Python. So, here are some of my learnings on how to run Python in production. Project quality Package manager Python has a fragmented ecosystem of package managers. The only ones I can recommend are poetry and uv . After learning about uv on Hacker News , I decided to give it a try . uv is blazingly fast and manages the Python binary as well. It even supports migrations from other package managers. The only downside is that uv is still not on a stable release yet. ...

The ultimate guide to using Poetry inside Docker

Update: As of Mar 2022, I recommend everyone to use GitHub Actions I maintain a somewhat popular Android developer tool ( adb-enhanced ). The tool is written in Python, supporting both Python 2 and 3. Testing the tool requires both Python runtime as well a running Android emulator. I, initially, used Travis CI for setting up continuous testing of this tool. Later, I felt that Travis CI was too slow and when I came across Circle CI, I decided to give it a try. As of now, both Travis and Circle CI are used for testing. Here is what I learned from my experience.