Collaborative and efficient scientific software support architecture
Kick-off workshop
December 5, 2025
CAES³AR project team
Institutes
- Geophysical Imaging and Monitoring (GIM)
- Research Process and Data Management, IT Center (ITC)
Principal investigators
- Prof. Dr. sc. Florian Wagner (GIM)
- Dr. rer. nat. Marius Politze (ITC)
Project team
- Andrea Balza Morales, M.Sc. (GIM)
- Camilla Lüttgens, M.Sc. (ITC)
- Dr. rer. nat. Marc Boxberg (GIM)
- Marcel Nellesen (ITC)
pyGIMLi is a versatile open-source toolbox:
- Open-source library for modeling and inversion in geophysics
- Management tools for structured and unstructured meshes in 2D & 3D
- Computationally efficient finite-element and finite-volume solvers
- Various geophysical forward operators: ERT/IP, Seismics/GPR traveltime, Gravity, Magnetics, SP, EM
- Versatile inversion frameworks
- Open-source, platform compatible, documented & tested code
- Suited for teaching & reproducible research
- v1.0 published in Computers and Geosciences (Rücker et al. 2017) (among 5 Most Downloaded papers and more than 550 citations) and broadly used since
Growing user base, growing support requests
Motivation of the project
Collaborative and efficient scientific software support architecture
Scientific open-source software is often
- developed by scientists, not software developers
- maintained by a small team of core developers
- built with limited resources
- a user-sided library, not a cloud service
- depending on third-party packages
Software support is additionally challenged due to
- differences in system hardware, dependencies, and input data
- limited knowledge of minimal working examples
- lacking possibility of interactive and collaborative programming in issues and mails
General statistics
What is the career stage distribution of participants?
- A total of 41 participants submitted responses
- Outreach was conducted through:
- mailing lists
- personal networks
- JOSS publications
- online message boards
- GitHub
Variety of different disciplines
Figure 2
What roles do participants have and what programs do they use?
--------------------------------------------------------------------------- NameError Traceback (most recent call last) Cell In[11], line 25 13 rwth_cmap = LinearSegmentedColormap.from_list("rwth_colors", rwth_palette) 15 # Generate word cloud with RWTH colors 16 wordcloud = WordCloud( 17 width=1600, 18 height=1600, 19 background_color='white', 20 colormap=rwth_cmap, 21 max_words=100, 22 relative_scaling=0.2, 23 min_font_size=10, 24 stopwords=custom_stopwords ---> 25 ).generate(all_text) 27 # Display word cloud 28 ax.imshow(wordcloud, interpolation='bilinear') NameError: name 'all_text' is not defined
Which programming languages are most commonly used?
Figure 5
Which research tools or topics are you most interested in?
Figure 6
Research tools and topics by user type
How do interests differ between users and developers?
- Different perspectives between user and developer roles
- Some overlap in key areas of interest
As a developer, what is the biggest challenge you encounter in maintaining open-source software?
Figure 8
As a developer, how do you currently deal with support requests / issues ?
- Community platform refers to message boards
- A notable number of contribution related to ticket systems
Ice Breaker
Groups of about five people (randomly distributed)
What you could talk about
- What is your experience with scientific software?
- What do you expect from this workshop?
- What was your most significant support request experience?
Time
5 minutes
Architecture Overview
- User creates an issue in an SCM platform
- A CAESAR bot handles orchestration of tasks and communication in the issue
- Actions/Jobs try to reproduce the issue in different environments in the background
- An interactive code execution environment is provided
Prerequisite: Executable Issues
- Minimal working example as code block or attached
- Additional data
- Environment details
Issue Forms
Bot Communication on Issue Creation
- CAESAR bot listens on issue creation
- More interaction is possible, e.g., explaining the process, asking for missing details, assigning labels,…
Reproduce Workflow in the Background
- A workflow is started automatically in the background
- Using the issue2docker tool, it tries to reproduce the described issue
- For the environment, a file is given in the repository in
.issue-environment/requirements.txt
Reproduce Workflow Results
- Issue could not be reproduced, meaning the described
Actual behavioris not happening when executing the provided code - In the example, it is because the prototype only uses the newest pyGIMLi version and there the bug is fixed
Workflow for Interactive Environment
Interactive Environment
