Prepare data

Project YAML

Configure named datasets, OME-Zarr plates, channels, and server settings.

Use one .yaml or .yml file when a session needs stable dataset names, mounted image plates, or shared launch settings. Relative paths resolve from the YAML file's directory.

Mapping syntax

The mapping form is concise and supports established dataset files:

datasets:
  experiment_a:
    anndata: ./analysis/experiment-a.zarr
    hcs_plate: ./images/experiment-a.zarr
    channels:
      DAPI:
        color: 3B82F6
        contrast: [100, 4000]
        visible: true
      GFP:
        color: 22C55E
        visible: false

obs_columns:
  - treatment
  - cell_type
  - fov_name
  - t
  - x
  - y

settings:
  host: 127.0.0.1
  port: 5055

path is accepted as an alias of anndata. The plate path accepts hcs_plate, ome-zarr, ome_zarr, or plate_path. Do not specify conflicting values through several aliases.

Array syntax

The equivalent array form uses explicit name and path fields:

datasets:
  - name: experiment_a
    path: ./analysis/experiment-a.zarr
    plate_path: ./images/experiment-a.zarr
  - name: experiment_b
    path: ./analysis/experiment-b.zarr

Every entry requires a non-empty name and data path. A project must contain at least one dataset.

Field reference

FieldTypeRequiredMeaning
datasetsmapping or arrayyesnamed AnnData or supported MuData mounts
anndata / pathstringper datasetlocal Zarr path
hcs_plate / plate aliasstringnomatching OME-Zarr HCS plate
channelsmappingnoinitial image channel display settings
obs_columnsstring arraynoobservation columns to ingest
settings.hoststringnobind host; default 127.0.0.1
settings.portintegernobind port; default 5055
presetstringnoshipped preset; currently annotate

Each channel entry requires a hexadecimal color string without #. contrast, when supplied, is a two-number [min, max] pair. visible is optional and converted to a boolean.

CLI values override YAML values for host, port, preset, and observation columns.

Launch and validate

ndea view project.yaml

Configuration errors are reported before the server starts. Common failures are a missing datasets object, a missing dataset path, a non-string path, an invalid channel object, or a non-integer port.