Skip to content

Session Log: AV13 corrected (real Kaltura player id), AV15 shipped as the real live accordion shape

Date: 2026-09-17 Driver: Yuji Shinozaki (with Claude Code) Outcome: Sprint 3's last two solo-workable items are done. AV13 corrected a wrong uiconf_id already live on dev-0 since the previous session, tracing D7's real primary-playback value through two independent sources. AV15 shipped a technical-metadata presentation for audio/video nodes — built once against Images' Sprint 2 B2 modal precedent, then rebuilt the same day after checking the real live production page revealed that precedent didn't apply. A real AV4 migration gap was found and filed along the way. Sprint 3 now has nothing left that isn't deferred to Than's return (week of 2026-09-22).

PRs merged: #219 (AV13), #220 (AV15, including the AV4 deferred note).

1. Session start: local DB had drifted from the 09-15 afternoon session

git pull brought in the previous session's config changes (AV9's gallery thumbnail, the /av//images// landing pages) that the local DDEV DB hadn't imported yet. drush config:status showed the expected drift (core.extension, system.site, the new views/displays); drush config:import caught it up cleanly, verified against dev-0's known counts (1,543 users, 4,187 audio, 7,396 video). Standing per-session check, not a new finding.

2. AV13: the shipped Kaltura player id was wrong

AV13 was scoped to migrate D7's per-view-mode player config into AV10's preset registry. Traced the real primary-playback uiconf_id two independent ways rather than trusting either alone:

  • D7 Drupal's own field display config (field_config_instance, read from the real d7_av database): field_video's default view mode configures entry_widget: 31832371. Traced all the way through field_kaltura_build_embed() (the actual function building the JS kWidget.embed() call) to confirm this value is genuinely wired into the live page render, not dead config.
  • The live React app (mandala-om/kmaps-app/src/legacy/audiovideo.js, AudioVideo.DrawPlayer() — confirmed the only reachable code path; a sibling Draw() method is dead, starting with a bare return; from an August 2020 refactor): hardcodes uiConfId='31832371' for both audio and video, no per-bundle branch.

Both independently confirmed 31832371. What AV9/AV10 had actually shipped — 24762821 — turned out to be MB_MAIN_PLAYER_ID, consumed only by mb_services_node_player(), a separate share/embed-redirect endpoint that has nothing to do with primary in-page playback. AV4's migration had picked it up by mistake, and it had been live on dev-0 since the previous session.

Fixed by correcting mandala_kaltura.settings' single default preset in place (no per-bundle split needed — the live evidence converges on one value, simplifying the row's original premise rather than expanding it). delivery stays HTTP, never RTMP (confirmed dead in D7's own embed builder regardless, and off the table per team decision). Verified live in DDEV against real video and audio nodes; deployed to dev-0 same session (webhook auto-triggered, no manual pipeline call needed).

Also checked, since it mattered for confidence in the fix: mb_kaltura's real transcript-sync JS (transcripts_ui.js) depends on the standard Kaltura Player Toolkit API (kWidget.addReadyCallback + kBind/sendNotification), generic to any compatible uiconf — and React's live player already uses 31832371 with exactly that API successfully. So the fix de-risks Sprint 4's future transcript-sync work rather than complicating it. Separately noted: the generic transcripts_ui module (the files Spike 11's live-evidence section cites) expects a plain <video>/<audio> DOM element — dead for Kaltura content specifically, since mb_kaltura's override is what's actually live. Worth Spike 11 building against the real mechanism when that work starts, not re-derived here.

3. AV15: built once, then rebuilt after checking the real page

Scoped from the sprint doc's own instruction to mirror Sprint 2 B2's Images precedent (a Bootstrap modal, shanti_images_carousel's node--shanti-image.html.twig). Built a new module, mandala_av_metadata, as an extra field (pseudo-field) rather than a full node template override — AV9 already owns the rest of the default view mode via display config. Rendered paragraph fields generically, introspecting each paragraph's field definitions at runtime rather than hand-listing the ~85 sub-fields across AV3's 14 relevant paragraph types, recursing for the two nested cases (av_pbcore_instantiation → av_pbcore_format_id, av_workflow → three av_workflow_note streams). Checked AV's actual field permissions before assuming none applied, as the sprint doc flagged: field_workflow is D7 field_permissions type 2/CUSTOM, and the real role_permission table confirms anonymous/general-authenticated roles have no view grant at all — gated on $node->access('update') as the D11 stand-in, same reasoning as Images' field_private_note gate. First version verified live in DDEV and committed as PR #220.

Then asked to replicate a specific real production page (av.mandala.library.virginia.edu/video/oral-culture-riddles-39-68) — and the Images-modal precedent turned out not to match reality at all. That domain serves D7's own live theme directly (shanti_sarvaka, jQuery, no React), confirmed from the page's actual script tags — a different, newer presentation than the React SPA (mandala.library.virginia.edu) or anything in the locally-checked-out mandala-om repo (searched all branches, found nothing — the AV site's live theme has evolved past what's cloned locally). The real "technical metadata" surface is a Field Group accordion with 5 named panels (Details, People, Rights & Licensing, Availability & Access, Technical Metadata) plus a "Video Overview" block above the fold. Pulled the exact field-to-panel mapping from the live DOM via browser automation, since the local d7_av dump's own field_group table only has edit-form groups (mode=default) — confirming that dump predates this accordion entirely, a real drift between the migration reference data and current production.

Rebuilt same session to match: Bootstrap 5 accordion instead of a modal, fields correctly split per the live mapping rather than lumped together (Details: coverage + relation; People: creator/contributor/publisher/sponsor with the role value used as the row label — "Cinematographer: Wende Khar", matching production exactly instead of generic "Creator:"/"Creator role:" pairs; Rights & Licensing: rights_summary + copyright_owner + year_published; Technical Metadata: identifier + extension + instantiation's sub-fields flattened in, not nested). Explicitly demo-priority per direction ("shape and a glimpse of the data," not full fidelity): "Availability & Access" is an empty placeholder (its real fields were empty on the live example checked, and it may belong to AV7's Group-access territory rather than AV15); no "Related Audio-Video" tab; the transcript panel is stubbed (Sprint 4/Spike 11 territory).

A real rendering bug, and a real (separate) AV4 data gap

Asked to fix the Details panel after it showed no data on several demo nodes. Two different things, chased both rather than assuming one explained the other:

  1. Real bug in this module, fixed: field_relation_identifier (on av_pbcore_relation) is a plain entity_reference, unlike every other paragraph-to-paragraph link this module walks (entity_reference_revisions). The generic renderer only special-cased the latter and silently produced nothing for the former (->value doesn't exist on an entity-reference item) instead of erroring — verified the fix with a synthetic in-memory paragraph, confirmed it resolves to the referenced entity's real label.
  2. Real AV4 migration gap, not fixed here, filed separately: even after the rendering fix, the field is genuinely empty — paragraph__field_relation_identifier has zero rows across all 836 migrated av_pbcore_relation paragraphs, while D7's real source has it populated (spot-checked node 116965 / legacy nid 24621: all 9 relation items have a real field_relation_identifier_target_id in D7, confirmed live on production's own "Related Media" list). This is exactly the risk AV3's own paragraph-model doc already flagged — "field_relation_identifier... needs a second pass or stub-and-backfill" — that pass evidently never happened. Filed as docs/deferred/av4-relation-identifier-not-migrated.md: Medium priority, since nothing shipped depends on it yet, but real data loss versus D7 that needs its own backfill pass (D7 nid → D11 nid via field_legacy_nid, not a fresh migration run).

Verified live in DDEV against 5+ real nodes across both bundles: correct panel population, the role-as-label pairing, nested instantiation/workflow-note data, and the workflow gate confirmed both ways (present for an admin account, absent for anonymous) on the same node. No new watchdog errors.

4. Sprint 3 status

Everything solo-workable is now done — AV2, AV3, AV4, AV5, AV6, AV8, AV9, AV13, AV14, AV15, AV16 all shipped. What remains (AV7's group_access_uva_member/ mb_collection_admin realms, AV11→AV12 uploads) stays deferred to Than's return, week of 2026-09-22 — a capacity call from the previous session, not a technical blocker.

What's next

Resuming this afternoon to address av4-relation-identifier-not-migrated.md — the backfill pass for field_relation_identifier (D7 nid → D11 nid via field_legacy_nid, not a fresh migration run; see that note's Recommendation section for the approach).