- Correct the twelve notebooks to teach visible Python/pythonnet construction, configuration, headless execution and fresh-result inspection and plotting.
- Preserve the same example cases/settings and extract observation-only fixtures.
- Replace cached notebook execution with independent raw-only kernel validation; retain earlier project/result archives solely as comparison evidence.
- Maintain plain Python notebook sources so regeneration cannot restore the old viewer workflow.
- Preserve CFA marginal uncertainty by passing both freshly fitted marginal chains; retain rendered PNGs in notebooks and audit original execution/output hashes without replacing them.
Earlier development work (the cached-viewer teaching design below is superseded):
-
Replace the original seven notebooks and loose scripts with twelve concise walkthroughs drawn from the BestFit application's saved example projects.
-
Preserve 22 complete project snapshots, original settings and seeds, and the distinction between saved results, fresh runs, and convergence diagnostics.
-
Add a locked .NET 10 / RMC.Numerics 2.2.0 runtime, checked saved-result caches, explicit full-settings reruns, source audits, and independent notebook execution.
-
Use the canonical BestFit plotting skill package. Map 45 desktop plot slots through 86 acceptance variants with independent geometry evidence and a gallery.
-
Document B17C examples 2 and 4 using the official reference, the app GMM/EMA distinction, the Nile date correction, and non-comparable model-score scopes.
-
Refresh against the public BestFit
db5807druntime and plotting utilities; retain original frozen numerical inputs and all unreviewed source descriptions. -
Correct the reviewed plot labels and regression bootstrap explanation, space dense contour labels, and match the zero-inflated mixture's desktop log range.
-
Add a pinned public plotting installer with process-local Windows long-path handling, active-review notice, and consistent author/reviewer/citation metadata.
-
Revalidate all twelve notebooks, 51 original-settings runs, and 560 saved views; preserve earlier evidence separately from the current publication review.
This development update does not create a release or certify scientific review.