# Triton v3 doc-driven dev ## Python ### Run our own pre-made environments from terminal ### Self-contained applications in containers - Basically like executables. Everything in a container. Users can ask for modifications, but they cannot and should not try to modify them. - Should contain graphical applications, tools etc. for OOD use in them. - When running in OOD, the application should run **inside** the container. - module load the relevant application - For running stuff, one of: - bin-directory added to path with wrappers which run the corresponding commands inside the wrapper - singularity_wrapper - This will print out what is happening so you understand what is going on. - Basically like: https://appimage.org/ ### Develop your own code based on our pre-made environments - Don't do this. Make your own environment - Want to base on our own? Run `singularty_wrapper pip list` or `singularity_wrapper pip freeze` to see the relevant versions from our container. Take the only relevant packages from there as a basis of your own enviroment - See [link] to make the environment ### Make your own environment - ~~`module load miniconda`~~ `module load conda` and maybe `module load r`. - Same as existing instructions we have - Mark these as development tools or installation tools. ### Add a few packages to our environments - Don't do this, make your own - See above to see the versions we have and make your own easily. ## Create your own containers - We should have a lot more instructions on how to create apptainer containers. ### Translate other installation instructions to our own system (e.g. you see a readme that says `pip install .`). - . ### Connect a container to Jupyter/OOD for usage that way ~/.local/share/jupyter/kernels/*.json - Current (RD) `module load envkernel` `envkernel conda --name=my-python $WRKDIR/conda-dir/` Only needs ipykernel installed inside JupyterLab+extensions managed by us - Proposed (ST): OOD jupyter app, in the OOD spawner, specify environment path (which has jupyterlab/etc. installed in it) - User manages all jupyter stuff - ~~OOD desktop application, activate environment, Run `jupyterlab`.~~