Demo data

Every slide and session the showcases use is described here, and this page is generated from the two files that actually drive them — test/fixtures/data/manifest.json and test/fixtures/sessions/index.json. Nothing here is maintained by hand, so it cannot describe a file that is not fetchable.

Getting it

npm run fixtures:fetch            # download + verify every slide below
npm run fixtures:fetch -- --list  # what is present, what is missing
npm run fixtures:derive           # build the derived overlays
npm run fixtures:serve            # serve it on :9100, with byte ranges

Each slide carries a SHA-256. A mismatch deletes the download and fails rather than warning: a truncated TIFF opens far enough to render a plausible-looking wrong demo. Set XOPAT_FIXTURE_BASE to fetch from a mirror.

There is also a slide that needs no download at all — npm run fixtures:synthetic generates a DeepZoom pyramid from tracked code, which is what makes npm run up:dev -- synthetic the one deployment with zero external dependencies.

Slides

File	What it is	Size	Licence	Provenance	Used by
LuCa-7color_Scan1.ome.tiff	Fluorescence, 7-channel OME-TIFF	1989 MiB	OME public sample data	Open Microscopy Environment sample data, 2016-06 sub-resolutions set	viz-flex-multichannel, fluorescence-background, channel-series, all-shaders
slide.tif	H&E slide, the cancer-prediction subject	652 MiB	not stated	RationAI; the slide every prediction overlay below was inferred over	basic-overlay, time-series, all-shaders, multi-view-goals, viz-flex-multichannel, viz-flex-geojson, viz-flex-mvt, viz-flex-grid, viz-flex-mask-coarse, viz-flex-mask-preview
cancer-inference.tif	Cancer prediction over slide.tif	44 MiB	not stated	RationAI model output; 512x512 prediction squares	basic-overlay, time-series, all-shaders, multi-view-goals, derive:vector, derive:mvt, derive:grid, derive:pyramid
detection.tiff	Graded detection map over slide.tif	14 MiB	not stated	RationAI model output; the third class in the derived vector layers	derive:vector, derive:mvt, derive:grid, derive:pyramid
annotations.tif	Rasterized annotations for slide.tif (white = annotation)	11 MiB	not stated	RationAI ground-truth annotations, rasterized	basic-overlay, time-series, all-shaders
explainability.tif	Explainability map for the prediction over slide.tif	16 MiB	not stated	RationAI model output	basic-overlay, time-series, all-shaders
slide-errors.tif	H&E slide used for the error-overlay sessions	646 MiB	not stated	RationAI	grouped-errors, multi-view-goals, all-shaders
channel-error.tif	Inference error caused by a channel offset, over slide-errors.tif	48 MiB	not stated	RationAI; a real failure mode, kept because the error UI needs one	grouped-errors, multi-view-goals, all-shaders
other-error.tif	Inference error of a second kind, over slide-errors.tif	32 MiB	not stated	RationAI	grouped-errors, multi-view-goals, all-shaders

Sessions

Each session is a plain JSON file that opens in the viewer's URL hash. Print an openable link for one with npm run fixtures:urls, or paste the JSON into /dev_setup.

visualization

Session	Demonstrates	Deployment	Needs first
basic-overlay	the ordinary case: an H&E background with prediction, annotation and explainability layers stacked over it	npm run up:dev -- webtiff	fixtures:fetch
two-backgrounds	per-background visualizationIndex, activeBackgroundIndex opening two at once, and a third entry reusing a data reference	npm run up:dev -- webtiff	fixtures:fetch
multi-view-goals	the multi-viewport grid: two viewers, different slides, independent visualization selection	npm run up:dev -- webtiff	fixtures:fetch
grouped-errors	shader groups, over two genuine inference failure modes rather than invented ones	npm run up:dev -- webtiff	fixtures:fetch
all-shaders	the shader registry: every registered layer type except interaction-debug (a debug aid, not a rendering), so a type that fails to register shows up here first	npm run up:dev -- webtiff	fixtures:fetch
time-series	the time-series wrapper and its source resolver — the broker owns swap-vs-append	npm run up:dev -- webtiff	fixtures:fetch

multichannel

Session	Demonstrates	Deployment	Needs first
fluorescence-background	shaders declared on a background rather than in a visualization, one per channel	npm run up:dev -- webtiff	fixtures:fetch
fluorescence-cross-source	per-shader dataReferences retargeting inside a background array, plus use_gamma	npm run up:dev -- webtiff	fixtures:fetch
channel-series	the channel-series wrapper: one data entry, channels browsed as a series	npm run up:dev -- webtiff	fixtures:fetch

viz-flex

Session	Demonstrates	Deployment	Needs first
viz-flex-multichannel	channel fan-out, plus one layer retargeted at a different data entry	npm run up:dev -- viz-flex-demo	fixtures:fetch
viz-flex-geojson	autodetected GeoJSONTileSource — a vector layer needs no protocol declaration	npm run up:dev -- viz-flex-demo	fixtures:fetch, fixtures:derive
viz-flex-mvt	a sparse MVT pyramid over the slide's own pixel space, via the demo-mvt factory protocol	npm run up:dev -- viz-flex-demo	fixtures:fetch, fixtures:derive
viz-flex-grid	imageSmoothingEnabled: false with pixelScale: 512 — without the latter the overlay is squeezed	npm run up:dev -- viz-flex-demo	fixtures:fetch, fixtures:derive
viz-flex-mask-coarse	coarsest level 1025 px, below PREVIEW_LEVEL_MIN_COARSEST_PX — the decline path	npm run up:dev -- viz-flex-demo	fixtures:fetch, fixtures:derive
viz-flex-mask-preview	coarsest level 4100 px — injection fires on the overlay layer, not the background	npm run up:dev -- viz-flex-demo	fixtures:fetch, fixtures:derive

errors

Session	Demonstrates	Deployment	Needs first
errors-partial-viz	partial degradation: a broken shader must not take its visualization or the slide down with it	npm run up:dev -- webtiff	fixtures:fetch
errors-all-invalid	the fully-failed-open path — every background faulty, and the viewer still has to say so rather than hang	npm run up:dev -- webtiff	nothing
empty-session	empty background and visualizations: activeBackgroundIndex: [] means nothing open, not everything	npm run up:dev -- webtiff	fixtures:fetch

annotations

Session	Demonstrates	Deployment	Needs first
annotations-lab	the subject of the annotations/measurements/pathology loop: one stained background with microns declared, so measured areas come back in µm² instead of pixels	npm run up:dev -- annotations-lab	fixtures:fetch
annotations-lab-overlay	measuring with source: "rendered" — the sampler reads the composited canvas, so what a threshold selects changes with the overlay, unlike the raw-data sources	npm run up:dev -- annotations-lab	fixtures:fetch

dicom

Session	Demonstrates	Deployment	Needs first
dicom-google-standalone	derived: "auto" — the plugin finds SEG/parametric series for the source series itself	npm run up:dev -- googledicom	env GOOGLE_DICOM_SERVICE_URL
dicom-google-standalone-explicit	the explicit {role, sourceSeriesUID} shape; the seriesUID is a placeholder you fill in from your own store	npm run up:dev -- googledicom	env GOOGLE_DICOM_SERVICE_URL