Python conda env not activated in CI

conda activation requires shell initialization that does not carry across CI steps by default. Without it, the job runs the base Python and cannot see packages installed into your named env.

What this error means

Commands fail with ModuleNotFoundError or "command not found" because the conda env was created but never activated in the running shell.

python

CommandNotFoundError: Your shell has not been properly configured to use
'conda activate'. To initialize your shell, run $ conda init <SHELL_NAME>

Common causes

conda not initialized for the step shell

Each step is a fresh shell; without sourcing conda or a login shell, conda activate is unavailable.

Activation that does not persist across steps

Even when activation works in one step, the env is gone in the next unless you re-activate or use a login shell.

How to fix it

Use a login shell with the conda action

setup-miniconda configures a bash -l shell so conda activate works and persists per step.

.github/workflows/ci.yml

- uses: conda-incubator/setup-miniconda@v3
  with:
    activate-environment: myenv
- run: python -c "import numpy"
  shell: bash -l {0}

Run via conda run if you avoid activation

Execute commands inside the env without activating the shell.

Terminal

conda run -n myenv python -c "import numpy"

How to prevent it

Use setup-miniconda with a login shell for conda jobs.
Activate or conda run in every step that needs the env.
Pin the environment name and verify it before running commands.