uv pip install cannot be used with non-default venv names created by uv venv without venv activation

[SnoopJ]

uv pip install does not have a workflow for installing directly into a venv previously created by uv venv (and there is no uv entrypoint in the venv that could be used instead).

This is inconvenient when using a workflow that benefits from not activating the venv whose site is being manipulated (e.g. multiple Python versions/target environments).

uv version: (0.1.4 @ fef1956)
Platform: Ubuntu 20.04
Python version: 3.9.16 provided by pyenv

Reproduction

NOTE: this invocation points directly at the uv entrypoint as a workaround for #1623.

$ ~/.pyenv/versions/3.9.16/bin/uv venv test_venv
Using Python 3.9.16 interpreter at /home/snoopjedi/.pyenv/versions/3.9.16/bin/python3
Creating virtualenv at: test_venv
$ ~/.pyenv/versions/3.9.16/bin/uv pip install numpy
error: Failed to locate a virtualenv or Conda environment (checked: `VIRTUAL_ENV`, `CONDA_PREFIX`, and `.venv`). Run `uv venv` to create a virtualenv.
$ ~/.pyenv/versions/3.9.16/bin/uv pip install --help | grep venv  # --help does not mention any venv functionality
$ source test_venv/bin/activate
$ ~/.pyenv/versions/3.9.16/bin/uv pip install numpy  # once venv is activated, install works as expected
Resolved 1 package in 41ms
Installed 1 package in 35ms
 + numpy==1.26.4
$ ls test_venv/lib/python3.9/site-packages/
__pycache__  _virtualenv.pth  _virtualenv.py  numpy  numpy-1.26.4.dist-info  numpy.libs

[zanieb]

Related

• Support custom virtual environment directory names #1422
• Enhancement Request: uv pip install with explicit venv path #1598 (comment)
• venv must be activated even when running via venv/bin/python -m uv #1501
• Flag to allow uv pip install to install into specific environments (that do not necessarily contain uv) #1396
• Create stub for uv in virtual environments #1632

[zanieb]

Closing in favor of #1632 and #1422

[zanieb]

Thanks for the clear issue by the way! It's nicely written