Need way of setting mirrors / alternate download locations for install and packages

[Axel-Shark]

My org blocks https://www.python.org/ftp/python so reticulate commands like install_python() and py_install() fail since they're hitting https://www.python.org/ftp/python by default.

We have internal mirrors for all of the sources but I don't see any mention in the reticulate MAN for how to define alternate sources / mirrors

[t-kalinowski]

You can point pip at your org's mirror by setting an env var like

PIP_INDEX_URL=https://<custom-mirror-url>

or by creating a global pip.conf file with the index-url set. That should work for py_install() and virtualenv_create()/virtualenv_install(), but I'm not sure about install_python().

You might also be able to use the Posit package manager, using the setup instructions here:
https://packagemanager.posit.co/client/#/repos/pypi/setup

[Axel-Shark]

Thank you! I'm new to Python and hadn't come across any of these options while looking for an answer. This looks promising!

EDIT: the solution offered by @t-kalinowski may be an option, but this is what I ended up doing in case someone finds themselves in a similar situation. I found that Posit mirrors PyPi and isn't blocked by my company's firewall.

system('wget <URL to Miniconda3-latest-Linux-x86_64.sh> -O ~/miniconda.sh') # download latest version of miniconda
system('bash ~/miniconda.sh -b -u -p ~/miniconda3') # install miniconda
system('rm -rf ~/miniconda.sh') # remove miniconda installer
system('~/miniconda3/bin/conda init bash') # activate conda for bash
system('~/miniconda3/bin/conda init zsh') # activate conda for zsh

library(reticulate)

# confirm that miniconda is installed
conda_list()
use_python(<use path returned by conda_list()>)

# use Posit Package Manager instead of PyPi / Python official repo
system('pip config set global.index-url https://packagemanager.posit.co/pypi/latest/simple')
system('pip config set global.trusted-host packagemanager.posit.co')

py_install("numpy", pip = TRUE) # confirm install is possible now. pip = TRUE is essential!!
# NB all Python libraries will be installed to your miniconda environment, so make sure you set the right one with use_python()

# run Python in R
np <- import("numpy", convert = TRUE)
np1 <- np$array(c(1:4))
np1