FAQ

Is this a modeling library or a benchmark library?

It is primarily a benchmark-and-workflow library. It includes some built-in baselines, but its main value is the shared protocol, reproducibility surface, and surrounding workflow tools.

Do I need to use the canonical studies?

No. They are there because they are recognizable and easy to teach. If you have your own data, start with PanelCase or ImpactCase.

Why does the package talk about agents and tokens?

Because more research workflows now include coding agents. Large benchmark outputs are hard to move through a chat or tool-calling runtime. The bundle/context-plan layer exists to keep those workflows structured and efficient.

Should beginners start with external adapters?

Usually no. Start with the built-in baselines and the benchmark protocol. Add optional ecosystems only after you understand what comparison you want to make.

When should I use the CLI instead of the Python API?

Use the Python API when you are still shaping a custom workflow. Use the CLI when you want reproducibility, easier onboarding, or a command sequence that maps directly to docs and CI.