Skip to content

Picking a component from the catalog (v1.9.0+)

Add component opens a chooser with two routes:

  • New component is the manual stepper: pick a type, connect it to data, configure it yourself. See Dashboard Creation.
  • Pick from catalog skips all of it and offers visualizations already built for the tool outputs Depictio found in this project.
The Add-component chooser, offering New component on the left and Pick from catalog on the right
The chooser, with a note that catalog suggestions come from the project's ingested data collections.

Where the offers come from

Nothing is inferred from the data itself. Every entry in the Tools Catalog declares how to recognise the file and which columns it must carry, and Depictio compares your ingested data collections against those declarations. A project whose outputs match no entry simply has nothing to offer.

Filename

The collection's file matches the name the entry expects, such as *.variants.tsv.

Path glob

The collection's path matches the entry's glob, such as **/ivar/*.tsv. This is how a real pipeline run is recognised.

Recipe

The collection came from a recipe the entry declares. Derived collections carry no raw pipeline path, so neither of the other two can fire.

Any one of the three is enough to surface the tool. The entry's column schema then decides which of its renders are offered, so a render binding a column your file does not carry is never proposed.

MultiQC is a special case

Every multiqc/* output keys the same multiqc.parquet, so offers are narrowed to the sections the ingested report actually holds, and the originating tool is named on each one. The General Statistics offer adds the report's General Statistics table; before v1.16.0 it added the first plot of another module.

The browser

Offers are grouped by tool on the left, one row per recognised output, with a coloured dot for each component type it can become. Selecting a row previews it on the right.

The preview is the real component, rendered in an iframe at the exact grid box it will occupy, so what you see is what lands on the dashboard.

The catalog browser, with offers grouped by tool on the left and a lollipop plot previewing on the right
iVar's variant calls, previewing as a lollipop plot.

Add saves it straight away. Edit drops it into the Design step first. Browser Back and Forward walk the picker's surfaces rather than leaving the flow.

Narrowing the list

Search matches tool, output, description, file and collection at once, so coverage finds mosdepth's own outputs and the MultiQC coverage section together.

Searching for coverage narrows the list to mosdepth's coverage outputs and the MultiQC sections that report coverage
One search box across every field an offer carries.

Filters adds two facets with counts, Component and Data collection, computed against the current search rather than the whole catalog. A violet badge on the button says how many are active.

The Filters popover, with a Component facet and a Data collection facet, both showing counts
Two facets, each counted against what the search already narrowed.

One output, several renders

Most recognised outputs support more than one visualization: iVar's variant calls can be drawn as a lollipop plot, an oncoplot, a Manhattan plot or a histogram. The tab strip switches between them, each render carrying its own configuration and its own grid box.

The render tab strip switched to Manhattan plot
The same iVar output, switched to its Manhattan render.

Settings tuned inside the preview travel with the offer, on both the Add and the Edit path.

Provenance

A component added from the catalog says so:

  • a copyable use: <tool>/<render-id> snippet, the same identifier you would write in a project YAML;
  • a catalog_source stamp carried through later edits and YAML import;
  • a violet catalog action on the tile, opening a metadata inspector with identity, resolved configuration, data source, catalog origin and raw JSON.

See also