Added the demo set generator and the EMBER upload workflow

[CodyCBakerPhD]

scripts/generate-demos.sh derives 28 short videos from a source recording (supplementary Video S1 of PLOS ONE 10.1371/journal.pone.0025390), each varying exactly one thing the app surfaces: atom layout (faststart, fragmented MP4), container (MOV, MKV, WebM, AVI), codec and profile (H.264 Baseline / High 10 / High 4:4:4, HEVC, AV1, VP8/VP9, MJPEG, lossless FFV1), GOP length, B-frame structure, bitrate behaviour (true CBR, starved cap), track properties (display-matrix rotation, VFR, half resolution, no audio) and metadata tags (rich vs stripped). The output is a BEP047 dataset (bids-specification#2231, as clip-extractor#41 writes it): one session per variant at sub-01/ses-<label>/beh/sub-01_ses-<label>_video.<ext>, a JSON sidecar per video carrying ffprobe-extracted technical keys (including ImageBitDepth and VideoCodecRFC6381 where determinable), and a dataset_description.json at the root recording the CC-BY-4.0 license, the PLOS article as SourceDatasets, the generation date, and the per-session index under an encoding-helper key. Each file also explains itself in its title/comment tags for the planned replacement of the Load Sample button.

The unmodified recording is ses-original, the first of the 29 sessions: the dataset is raw at its outer level, so it needs no sourcedata/ layer of its own. It is not in this repository — a video blob would sit in the git history forever — and the generator fetches it from that archived session when there is no local copy.

scripts/upload-demos.sh copies the dataset into EMBER dandiset 000527 with the dandi CLI (validation skipped, since these are plain videos); dandiset, instance and layout are fixed. The manual Upload demos to EMBER workflow dispatches with no parameters and runs both, keeping the set as a run artifact. It needs an EMBER_DANDI_API_KEY repository secret with write access to the dandiset.

No version bump: nothing under src/, configs/ or the package files changed.

Prompts

I'd like to setup a range of demos showcasing the importance of each aspect of the inspect tab; can you use the attached source video in a simple bash script using ffmpeg to create a range of file structures, both H264 and otherwise (also useful for interoperability testing) and a separate script that pushes them via a workflow dispatch github action to ember dandiset https://dandi.emberarchive.org/dandiset/000527 ?

A follow up would then expose direct integration that would replace teh load sample button

@"/root/.claude/uploads/205159ee-a1e4-5de0-a73a-2e48a585909c/a6f51fbb-Video_S1.m4v"

rephrase away from "Inspect Demos" to "Upload demos" (and just 'demos' elsewhere, they are not unique to the inspect tab

the dandiset ID will not change, no need to expose that as a parameter

the dandi instance will not change, no need to expose that as a parameter

the point of the workflow is to upload so remove that checkbox option too

might as well clamp the directory structure; follow BEP047 like you do for output of clip-extractor, refer to this as 'sub-01' and add different desc- to each

for BEP047 (again refer to brain-bbqs/clip-extractor#41 for specifics) compatability in this output, replace manifest.json with a dataset_description.json and extract metadata sidecars for each video. Actually, maybe replaces 'desc-' with 'ses-' as if each video case was a different session rather than a variant recording/acquisition of a single session and therefore nest each under its own level

don't forget "ImageBitDepth" and "VideoCodecRFC6381" fields

also drop the section [encoding-helper sidecar block] from each video sidecar, that info belongs more in the dataset_description.json

should probably include the date the dataset was generated on in the "Description" of "GeneratedBy"

also make license CC-BY (or copy w/e license is from source https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0025390#s5) also list that as source dataset

can you add the name for the source dataset? any other useful metadata to include for that?

"from the ses-reference recording" -> "from the original video"

IDT the version should be the year; also order keys by "Name" -> "Description" then as-is

ok last thing; now that it's all on the archive, remove the video from this branch so it doesn't get injected into the git blob history

yeah i mean if its not already on there somewhere in the sessions go ahead and add it and adjust the upload script to operate on that remote blob (ill make it persistent)

unpack the sourcedata/rawbids, the entire dataset at the outer level is raw. just include that one as the 'ses-original'

🤖 Generated with Claude Code

https://claude.ai/code/session_01ErEUUQ4hdFhkcGptkwGgAn

[github-actions]

PR Preview Action v1.8.1
🚀 View preview at https://brain-bbqs.github.io/encoding-helper/pr-preview/pr-27/
Built to branch gh-pages at 2026-08-23 20:09 UTC. Preview will be ready when the GitHub Pages deployment is complete.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.