Local app (no B-Fabric)

The standalone queue app lets you generate an instrument queue from a sample table you upload — no FGCZ/B-Fabric account, no LIMS. It shares the full generation pipeline, QC patterns, randomization, and visualizations with the portal app; only the sample source (a file upload instead of a B-Fabric order) and the output (a local download instead of a workunit upload) differ.

Launch

make app-local
# or
qg-app-local
# or
uv run marimo run src/qg/apps/queue_app_local.py

No QG_ALLOW_UNAUTHENTICATED and no B-Fabric packages are needed — this works on a core pip install qg.

1. Upload a sample table — or load a bundled example

Drop a .csv or .xlsx file on the landing page, or pick one from the bundled-example dropdown (single-project vial/plate, an 80-sample run, and a 3-project multi-container set) to load it directly — no file hunting needed. A download button next to the dropdown saves the selected example as an editable template for your own data. An uploaded file takes precedence over a selected example, and the page shows which source is active. The same examples also live in docs/examples/ for browsing.

The mode is inferred from the columns — plate when both plate_id and grid_position are present, otherwise vial.

Vial — required sample_name, sample_id, container_id; optional tube_id, grouping_var.

Plate — required sample_name, sample_id, container_id, plate_id, grid_position; optional tray, grouping_var.

Header spellings such as Sample Name, Sample ID, Tube ID, Container ID (and the B-Fabric export names) are accepted and normalized; unknown columns are ignored. sample_id must be unique whole numbers. container_id is a generic group id — samples are grouped and shuffled within each container_id, and a separation QC block is inserted between containers, so a single upload can span several projects.

Clear errors are shown for unreadable files, unknown extensions, missing required columns, and duplicate sample_ids.

2. Configure and generate

Use the sidebar exactly as in the portal app: Tech Area → Instrument → Sampler → Queue Type → Plate Layout → QC Layout → Pattern, plus polarity, methods, randomization, date, user, injection volume, and QC frequency. The Sample Selection tab lets you uncheck samples or reorder them.

• Randomization (no / random / blocked / blocked_uniform) shuffles user samples within each container. Set a seed (in a downloaded parameters JSON) to reproduce a run; otherwise one is drawn and recorded.
• The Visualizations tab shows the plate layout and the acquisition-order timeline, each annotated with an η² balance score (0 = group evenly interleaved, 1 = fully separated). Uploads with a grouping_var make this meaningful; ungrouped tables show N/A.

3. Download

On the Queue Preview tab the Download Queue File button is enabled as soon as a queue is generated (no upload step). The Parameters tab offers Download Params JSON — keep it for reproducibility; feeding it back through the CLI (qg params.json -o queue.csv) regenerates the same queue.

Differences from the portal app

	Local app	Portal app
Sample source	CSV/XLSX upload	B-Fabric order browser
Auth	none	B-Fabric session
Output	direct download	workunit upload (download gated behind it)
Install	core qg	qg[bfabric]