I used to work with Java for many years but in the last project I need to use python and I learn the ecosystem these days. We use UV and have a project with a bunch of dependencies. On my development machine I run uv run main.py. However I was wondering how should we prepare a deployment artifact for our application? We'll use python and uv stack in 2 different kinds of applications: Run in Docker in k8s Run as a process So My question is: Is it a good practice to install uv on production machine (or in docker) and still run the production application like I do in the development or there are other alternatives? What is the best way to run the application with multiple python modules and having many dependencies? For example here is what I do now in docker: Dockerfile: FROM python:3.13 RUN pip3 install uv WORKDIR /app # I'll optimize the image later COPY run.sh /app COPY main.py /app COPY pyproject.toml /app COPY uv.lock /app RUN uv sync --all-groups ENTRYPOINT ["/bin/bash", "/app/run.sh"] My /app/run.sh looks like this: #!/bin/bash uv run main.py But I'm not sure this is how the apps should be really deployed especially for the "run-without-docker" type of deployments - should I install uv on the prod server and run the app like this as well? In java world we used maven to build stuff but we didn't need to install it on production servers and used it solely as a build tool... Continue reading...
