[Alexander-Mitrofanov]

Summary

vaRRI now ships a reviewed set of 12 contextual WebM demonstrations for complex viewer workflows. The clips are integrated into the composed documentation and the canvas question-mark help for Select, rectangle selection, Move, Rotate, Release, Reset, FASTA input, probability profiles, text annotations, JSON sessions, PNG export, and paired-region highlights.

Video help uses progressive disclosure: focus or hover reveals a poster teaser, while click, Enter, or touch explicitly expands a native player. Sources and captions attach lazily, playback never autostarts or loops, only one clip plays at a time, dialog close resets playback, and every clip includes English captions, a written description, a step-by-step transcript, and a download link. Previews are Escape-dismissible and kept inside the viewport/dialog, including short windows; controls have task-specific accessible names.

Recording and publication

• Expands the Playwright/FFmpeg recorder from 3 to 12 behavior-asserting scenes with canvas, full-UI, dialog, and export-dialog stage profiles.
• Adds real pointer, drag, fill, select, wheel, and key capture helpers.
• Adds validated publication into doc/demos/ after manual review; publishing requires a clean, revision-matched recording. Source provenance is checked before and after recording, and Git failures fail closed.
• Publishes only WebM, poster, and VTT assets plus a checksummed catalog. QA review frames remain ignored.
• Verifies original recording hashes before promotion and the staged catalog again after copying; rejects changed bytes, missing/duplicate scenes, invalid assets, caption/transcript mismatches, and stale metadata.
• Adds catalog/package integrity checks to CI and documents the recording, review, and publication workflow. Automated validation does not replace human review.

Council review and fixes

Three reviewers audited the player/accessibility, documentation/scene semantics, and recording/publication pipeline, then cross-reviewed each other's findings and fixes.

• Fixed publication accepting altered recordings, incomplete catalogs, and unverifiable source state; added 15 regression tests.
• Fixed clipped previews, Escape dismissal, task-specific accessible names, and keyboard focus hidden behind the dialog footer.
• Corrected the documented rotation gesture, required scene placements/captions, and the probability demo's Start 1 prerequisite.
• Added persistent browser regressions for delayed posters, placement, real playback/captions, single-active playback, collapse, tooltip bounds, and short-height keyboard access.

Independent media QA

• Every clip was generated and reviewed in a separate recorder invocation before integration.
• Caption-readability findings were corrected and all affected clips were independently rerun.
• After integration, separate reviewers audited video behavior and written explanations.
• Final catalog recheck passed 12/12 clips in Chromium and Firefox: 960x640 decode, duration/cue match, posters, MIME types, lazy loading, mouse/keyboard/touch access, one-active playback, dialog reset, responsive layout, and reduced motion.
• Regenerated all clips from clean source e011a9fae3812193f1e29eac20941eadb64610d6; independent integrity and explanation audits passed 12/12.
• Final catalog: 37 production files, no orphans or QA artifacts, 3064.6 KiB total.

Validation

Both GitHub checks passed on final commit b26fc208d9ece788a7a9df1cc135e3b8662b8ee0: Tests and Standalone core bundle. The council has no unresolved findings.

• npm run demos — 12 clips recorded and fully decoded with VP8.
• npm run demos:check-published — 12 clips and all hashes/bytes/assets passed.
• npm run test:architecture — passed.
• npm run test:ci — 47 suites / 552 tests passed.
• npm run test:package — packed/installed assets and entries passed.
• npm run test:browser — native/core bundle and export contracts passed.
• npm run test:ui — Chromium passed.
• npm run test:ui:firefox — Firefox passed.
• npm run test:pages — composed help, lazy videos, anchors, images, and downloads passed.
• npm run test:visual — five fixtures passed; maximum changed-pixel ratio 0.0253% (limit 0.2%); no baseline updates.