Use juv when there's no pyproject.toml file

[kolibril13]

I’d love to use juv in my everyday projects, especially now that Cursor AI supports notebooks.
It would be great if the juv VS Code extension could add packages to the project even when no pyproject.toml is present.
At the moment, this only works when the file exists, as shown in this screen recording (juv version 0.1.0, latest VS Code):

Screen_Recording_2025-05-03_at_15.08.40.mp4

[m-z-b]

This has been confusing me for months. juv add produces a confusing error / backtrace if pyproject.toml does not exist when adding a dependency to a script, even though the pyproject.toml file is not being used and is not updated by the operation.

I've only just discovered that adding an unused pyproject.toml to a directory is a workaround.

[manzt]

Is this within VS Code extension or from the CLI?

[m-z-b]

This is in the CLI version.
I've just updated to 0.4.2. Previously I was getting a confusing stack trace and error message. Now I'm getting a more helpful:

$ uv add test.ipynb numpy
error: No `pyproject.toml` found in current directory or any parent directory

But then if I do:

$ juv init fred.ipynb
Initialized notebook at `fred.ipynb`

$ juv add fred.ipynb numpy
Updated `fred.ipynb`

$ juv add test.ipynb scipy
Updated `test.ipynb`

So it then starts working without a pyproject.toml in the directory, which is extremely puzzling. (Is there a cache somewhere?)

I've mainly encountered the issue converting old notebooks to use juv. Possibly on first juv use after boot.

(Thanks for simplifying notebook dependency management, by the way!)

[manzt]

Thanks for the report, and glad you're finding juv helpful!

No caching involved (other than uv's environment cache). Please be mindful of the CLI you're using — as you noted, $ uv add test.ipynb numpy is using uv (not juv). The:

# error: No `pyproject.toml` found in current directory or any parent directory

is coming from uv add which requires a pyproject.toml unless you are using the --script flag.

[m-z-b]

I keep on making that uv / juv typo! However, I'm certain in the past I was not consistently making that mistake, and was getting a Python stack trace if pyproject.toml didn't exist.

After updating uv and juv to the latest version and purging the uv cache, I am no longer able to reproduce the issue.

Thanks for your help / patience... From my (CLI) perspective the issue can be closed.

However, I've no idea if the VS Code issue reported by kolibril13 is due to a similar version problem.