Make Environment Managements works for all conda virtual environments

[Abdelrhman-Gaber]

🚀 The feature, motivation and pitch

Dear Team,

Can we please for the following script add another argument to able to install the packages inside any virtual environment

#!/bin/bash
set -eux
PROJECT_DIR=${SMUS_PROJECT_DIR:-"$HOME/src"}

Check if the .libs.json file exists

if [ -e $PROJECT_DIR/.libs.json ]; then
lib_config_json=cat $PROJECT_DIR/.libs.json

apply_change_to_space=echo $lib_config_json | jq -r '.ApplyChangeToSpace'

Extract conda channels from the config, add -c before each channel and join the strings

conda_channels=echo $lib_config_json | jq -r '.Python.CondaPackages.Channels | .[]' | sed 's/^/-c /g'

Extract conda package spec from the config and join the strings

conda_package=echo $lib_config_json | jq -r '.Python.CondaPackages.PackageSpecs | .[]'

if [ ${apply_change_to_space} == "true" -a -n "$conda_package" ]; then
# if conda package spec exists, install the packages
micromamba install --freeze-installed -y $conda_channels $conda_package
fi
fi

Use case description

No response

Alternatives

No response

Additional context

No response

[TRNWWZ]

Hi, may I ask why is this change required? What's the motivation?

[Abdelrhman-Gaber]

Dear Team.

The motivation for this change is to make the package installation process more flexible and environment specific.

Currently, the script installs Conda packages only into the default environment (typically base). However, in our workflows, we often work with multiple Conda or virtual environments within the same SageMaker space for example, separate environments for different projects, Python versions, or dependency sets.

By adding an argument to specify the target environment, we can:

• Avoid polluting the base environment with project specific dependencies.
• Maintain isolated, reproducible environments for different workloads.
• Reduce dependency conflicts between projects.
• Improve startup performance by installing only into the relevant environment.
  This change would allow us to run the same .libs.json driven installation logic but direct it to any active or specified Conda environment, making the script more reusable and aligned with best practices for environment isolation in SageMaker Studio.