Make uv primary tool in our documentation and READMEs?

[astrojuanlu]

The landscape of tools for Python project management is changing fast, and one particular tool has been making a lot of noise, which is uv https://docs.astral.sh/uv/

Just in the last few hours, a couple of people have asked questions related to setting up Kedro with uv:

• https://kedro.hall.community/trouble-getting-kedro-viz-to-run-r081WrBQ0cjq
• https://kedro.hall.community/updating-kedro-project-from-conda-to-uv-FRnf5Pm3xWVD

And this is on top of some more questions asked over the past few weeks:

• https://kedro.hall.community/kedro-and-uv-typical-initialization-and-working-pattern-YfYR3TQrr0Cb
• https://kedro.hall.community/question-on-project-setup-KMcGdpBR5Wsr
• https://kedro.hall.community/kedro-run-executes-nodes-out-of-order-despite-expected-dag-behavior-NpNgB3pWhM6n

In addition, MLflow adopted uv in the 2.2.0 version to manage their model lifecycle https://github.com/mlflow/mlflow/releases/tag/v2.20.0

And finally, in a bit less than a year uv surpassed Hatch, PDM, Pipenv in monthly downloads, and is only behind Poetry https://clickpy.clickhouse.com/dashboard/uv

Edit: The download count of GitHub releases is actually public (see example), and looks like they've had on the ~10s of millions of downloads

Should we primarily use uv as the workflow tool of choice, rather than the current mix of pip + venv + conda?

This should be more about just using uv pip install, but rather using uv to its full potential:

• uvx kedro new for project creation (no more "double venv", see Run kedro new without creating a new directory #681)
• uv run kedro run, no more venv activation/deactivation

Thoughts? (Please upvote this discussion if you generally agreee)

[datajoely]

Yes, yes, yes 1000% yes.

[datajoely]

Literally the whole workflow top to bottom should be uv

[astrojuanlu]

For folks that need conda packages, we can explore mentioning pixi as well. I've heard many good things about it, although I haven't used it personally. It uses uv under the hood.

[datajoely]

It uses uv under the hood for pypi stuff, but implements its own mamba solution for conda packages. We can ask our friend nichita to help us think through this if we need.

[tynandebold]

As someone who always had to use virtual envs for Kedro-Viz, uv has made jumping back into contributing infinitely easier. I'm in full agreement here.

[astrojuanlu]

meltano SDK also migrates from Poetry to uv https://sdk.meltano.com/en/v0.44.2/guides/migrate-to-uv.html