Ensuring Runtime Reproducibility in Python
summary
Claims a specific solution: conda-store (built on conda) produces pinned YAMLs, lockfiles, Docker images and archives, offers a UI, version control, role‑based access, and policy enforcement so teams can rebuild exact environments. Concrete failure example: a Streamlit+Polars app breaks because requirements.txt omits Python version and platform and the hosted domain is down. Workflow advice includes enumerating components, threat modeling, and six levels of reproducibility from lockfiles to full hardware images; local builds can use mamba as the installer.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:09Runtime reproducibility defined
- 3:16Threat modeling for reproducibility
- 7:08Reproducibility issues example
- 10:14Six levels of trust
- 12:39Data science challenges
- 14:50Cond store tool
- 19:51Key takeaways
- 21:03Q&A
related talks
30mFrom Chaos to Confidence: Solving Python's Environment Reprodu...
41mEcosystem for Reliable Python Programming
40mMaking packages and packaging "just work"
01h 22mLearn how to Make Life Easier with Anaconda
34mModern development environments for Pythonistas
28mBuilding reproducible Python applications for secured environments
37mEmbrace conda packages.
23mCleaning Up Your Python Environment Superfund Site