evidence · 5.8 KiB

README.md

Committed at docs/evidence/work-card-media/README.md. This is what grounds the requirements that cite it.

Open raw source

Work card media and evidence access

Host browser proof recorded 2026-09-24 at source 4133cdf5398db99f58fd0b3398ebfec22f288245.

The work board discovers committed images, videos, reports and logs from change citations and the associated evidence records. A reviewed override selects the coherent-shell card’s QEMU screenshot; it remains explicitly labelled QEMU. The gallery and evidence menu do not alter completion or physical verification state. New evidence is discovered on the next normal site build, including before a proposal completes or archives.

python3 tests/test_work_board.py
python3 tests/test_render_specs.py
python3 scripts/build_site.py
python3 tests/work_card_media_browser.py

The browser command serves the local build and maps revision-pinned raw media URLs to the corresponding local repository files. It actually decodes and advances the existing system-controls MP4. This is host website behavior, not remote deployment or a new board interaction. Reports render Markdown; raw logs, tables and media link directly to their pinned revision to keep the static bundle bounded. Dialog media stays inert until a card opens; videos have controls and never autoplay.

The coordinator visually inspected these public browser captures:

Desktop evidence menu on a whole clickable card

Mobile inline gallery with the QEMU provenance label

The initial card/media implementation was published at https://knewter.github.io/tdisplay-k230/work/: Pages run 35958300081 succeeded and the page contained source revision 1005d4f5a538.

Additional host browser proof recorded 2026-09-24 at source 046ae6082a92b281d0a0e4e05343dfa29b524027 using:

python3 scripts/build_site.py
python3 -u tests/work_card_media_browser.py

The build passed: 238 pages, 8,043,984 bytes, 12.99 seconds and all 12 output checks. Desktop (1440×1000) and mobile (390×844) Chromium both passed:

The coordinator visually reviewed the desktop gallery and mobile document captures below. The browser routes pinned raw artifacts to local files; this is host website proof, not new device evidence or remote deployment proof. The explicit original-file action may open the original separately; ordinary card evidence clicks stay in the modal. Publication of this extension is recorded separately after CI.

Largest-fit desktop gallery

Rendered Markdown in the mobile document modal

Viewer playback update

At source 28cc0fbde32475f59a00f5e48737faa6c9a4c183, the viewer starts video playback when a reader opens or selects a video. Card previews remain paused. Space toggles playback with focus on Close or on native video controls, without dismissing the gallery or toggling twice. Native controls, Escape/Close, stop-on-close, and card focus restoration remain available. This supersedes the earlier no-autoplay viewer behavior above.

python3 scripts/build_site.py passed: 239 pages, 8,060,379 bytes, 28.60s. python3 -u tests/work_card_media_browser.py passed desktop and mobile, including automatic video advance, Space pause/resume from Close and native video focus, and all prior media/document modal checks. Browser autoplay restrictions may still require pressing Play; controls remain usable.

Published autoplay and keyboard proof

At 2026-09-24 05:39 UTC, CI run 35960674736 completed successfully for exact source b2478b5dcb53d742a0c5a2523c74668a2e15901c. The public https://knewter.github.io/tdisplay-k230/work/ contained that full revision. An independent headless Chromium check visited this public URL directly, with no asset routing override. Desktop 1440×1000 and mobile 390×844 both passed: the actual system-controls MP4 decoded and advanced on opening; Space paused and resumed it with focus on Close, then paused and resumed it with focus on the native video; the media modal remained open; Close stopped the video, removed its source and returned focus to the trigger. Neither viewport reported a page JavaScript error. This is published website/browser evidence, not a physical K230 panel observation.

The public probe used Python Playwright Chromium against the published URL, with its real pinned media requests and no mocked responses. The source build and local browser regression are recorded above; this paragraph records the additional deployed-revision check required before archive. The archive’s later published revision must be checked separately after it is pushed.

python3 tests/work_card_media_public_browser.py --revision b2478b5dcb53d742a0c5a2523c74668a2e15901c