04 · Guided Verification Journey

Theme 04 · Modernization · data: synthetic · reimagines METplotpy + METviewer

A cohesive walkthrough for the question “is my forecast good?”: Overview → Error vs lead → Categorical skill → Reliability. Every view carries a provenance panel (data source, selection, method), sessions serialize into a shareable URL hash, and a deterministic rule-based narrator describes what the numbers show — grounded in the on-screen values, not generated text.

Structure

apps/modernization/
  index.html          entry
  src/router.js       step routing + URL-hash state
  src/steps/          step0_overview · step1_leadtime · step2_categorical · step3_reliability
  src/components/     chart.js · narrator.js · provenance.js
  src/data.js         synthetic verification dataset accessors

Status & notes

• Migrated to the shared lib (cnt(), cts(), reliability() → lib functions) with 0 differential diffs; the shared Brier decomposition also added UNC/BS/BSS that the local version lacked.
• Values already route through an esc() helper — the app is ahead of the pack on the injection-hardening checklist for real data.
• Real-data wiring is queued (second in sequence): SL1L2 + CTC swap in cleanly; step 3 (reliability) will show an honest “no probabilistic data” card.
• Single-file build: dist/modernization.html (~800 KB).