Viz Module API Reference

For a conceptual overview of what each widget shows and how to use it, see the Visualizations section.

The Clustergram widget accepts a parquet_data argument for efficient initialization. Use Matrix.export_viz_parquet to generate this data from a clustered matrix. Passing a JSON network object is deprecated; pass matrix or parquet_data instead.

Yearbook Selection

A Yearbook renders a grid of cell "portraits". There are three ways to choose which cells it shows, in increasing order of power:

An explicit id list — cells=["cell_1", "cell_2", ...]. The simplest path: you already know the cells you want.
A back-end selection — selection=.... Accepts a celldega.select.Selection, a JSON-ready selection dict, or a plain id list. This is the recommended way to drive the grid from a Python AnnData object, because it carries the query, sampler, scores, and provenance with the cells. See the select module docs for the full query/sampling algebra.
A stateless front-end query — front_end_query=.... Evaluated in the browser against the dataset's LandscapeFiles, so it needs only a base_url and no Python AnnData.

Passing both cells= and selection= is rejected. front_end_query= is independent of those two — when set, the browser computes the cell list itself.

Front-End Query

front_end_query is a small dict evaluated entirely in the browser. It is a deliberately narrow counterpart to the Python select module — it supports a single cluster filter and/or a single-gene ranking:

Query	Behavior
`{"cluster": {"attr": "leiden", "value": "8"}}`	Random cells from cluster `8` (capped at `num_rows * num_cols * 10` by default).
`{"gene": "BRCA1"}`	All cells ranked by `BRCA1` expression, highest first.
`{"cluster": {"attr": "leiden", "value": "8"}, "gene": "BRCA1"}`	Cells in cluster `8`, ranked by `BRCA1` expression.
`{"gene": "BRCA1", "max_cells": 50}`	As above, capped at 50 cells.

yb = dega.viz.Yearbook(
    base_url="https://path-to-dataset",
    front_end_query={"gene": "BRCA1", "max_cells": 50},
    rows=2,
    cols=2,
)

Note

The former query= argument was renamed to front_end_query= to distinguish this stateless browser query from the Python-side celldega.select query module. Passing query= still works but emits a DeprecationWarning.

Module for visualization

`CellCloud`

Bases: _SpatialWidget

3D orbit view of cell centroids (replaces Landscape(technology="point-cloud")).

Renders a pre-built point-cloud DegaFiles directory — cell positions from cell_metadata.parquet (or cell_metadata_<alignment>.parquet when alignment is set), colored by cluster/gene. Build one with :func:celldega.align.write_alignment_point_cloud.

Parameters:

Name	Type	Description	Default
`base_url`	`str or list`	DegaFiles URL(s), as in `Landscape`.	required
`adata`	`AnnData`	Source of cell attributes/metadata (clusters, colors, UMAP). Never used for spatial positions — those come from the DegaFiles — except when `use_adata_3d_centroids` is set.	required
`alignment`	`str`	Named alignment variant; cell positions are read from `cell_metadata_<alignment>.parquet`.	required
`use_adata_3d_centroids`	`bool`	When an `adata` with `obsm["spatial"]` is given, render its centroids (`obs[z_key]` for Z, falling back to 0) instead of the on-disk geometry — to preview a candidate alignment without rewriting DegaFiles. Default: True.	required
`z_key`	`str`	`obs` column holding the Z coordinate for `use_adata_3d_centroids`. Default: "Z".	required

use_adata_3d_centroids writes centroids to a small sidecar file next to base_url and fetches it over HTTP when base_url is a local celldega.viz.get_local_server() address (millions of per-cell centroids don't fit through the widget's comm channel); otherwise it falls back to syncing them through widget state, fine for smaller datasets.

`Clustergram`

Bases: AnyWidget

Minimal version of the Clustergram widget.

Frontend still gets: matrix/parquet data, row/col names, manual_cat, manual_cat_config, etc.
Manual categories are treated as a simple JSON string.
All the old DataFrame-based manual_cat plumbing is removed.

`manual_cat_dict` `property`

Convenience accessor: parsed JSON from manual_cat.

`init(**kwargs)`

Parameters

parquet_data : dict, optional Pre-exported parquet payload from your matrix object. matrix : object, optional If provided and has .export_viz_parquet(), we'll call that. network : dict, optional Deprecated path, kept only for backwards-compatibility.

`close()`

Close the widget and notify the frontend to release resources.

`to_cluster(axis='row', n_clusters=None, threshold=None, criterion=None)`

Cut the dendrogram into flat cluster labels via the underlying Matrix.

Thin wrapper over :meth:celldega.clust.Matrix.to_cluster. When neither n_clusters nor threshold is passed, the cut is read from the front-end dendrogram slider state in dendro_cut[axis] — a dict of {"n_clusters": int} or {"threshold": float} that the JS widget writes as the user drags the slider. Passing an explicit value overrides the slider.

Parameters:

Name	Type	Description	Default
`axis`	`str`	`"row"` or `"col"` — which dendrogram to cut.	`'row'`
`n_clusters`	`int \| None`	Target number of flat clusters (overrides the slider).	`None`
`threshold`	`float \| None`	Linkage-distance cutoff (overrides the slider).	`None`
`criterion`	`str \| None`	Explicit `scipy` `fcluster` criterion.	`None`

Returns:

Type	Description
`Series`	A `pd.Series` of cluster labels indexed by the axis names.

Raises:

Type	Description
`ValueError`	If no Matrix is attached, or no cut is available from either the arguments or the front-end slider.

`Composition`

Bases: Clustergram

Composition view: count/proportion of categories compared across groups.

A Clustergram subclass with viz_mode="composition". The body draws each group (dataset/sample) as a stacked bar whose segments are populations (cell types), reusing the Clustergram's column-attribute tracks, reorder buttons (ini / sum / clust), and the control-panel PROP/COUNTS normalization toggle.

"Composition shows the count or relative proportion of categories within each group, and compares those compositions across groups."

Example::

dset = dega.DatasetCollection(adata, dataset_col="sample_id",
                               obs_columns=["condition"])
dset.calc_population(adata, category="cell_type")
dega.viz.Composition(
    dset, category="cell_type", group_attrs=["condition"]
)

Note: calc_population already copies adata.uns[f"{category}_colors"] onto the population modality it builds, so Composition picks up the same colors from dset alone — passing adata= is only needed as a fallback (e.g. a plain DataFrame input, or an AnnData/modality that has no color palette of its own).

`init(data, modality='population', *, category=None, colors=None, adata=None, group_attrs=None, normalized=None, col_weights=None, cluster=True, name='composition', width=700, height=450, **kwargs)`

Parameters:

Name	Type	Description	Default
`data`	`Any`	A Celldega collection (`DatasetCollection` / `SetCollection`), a `MuData`, an `AnnData` (obs = groups, var = populations), or a `DataFrame` (rows = groups, columns = populations) — typically the output of `calc_population`.	required
`modality`	`str`	Modality key on a collection/MuData (default `"population"`).	`'population'`
`category`	`str \| None`	Population `obs` column name used to resolve colors from `adata.uns[f"{category}_colors"]` when the modality has none.	`None`
`colors`	`dict[str, str] \| None`	Optional `{population: hex}` overrides.	`None`
`adata`	`Any`	Optional source cell-level `AnnData` to fall back to for its color palette. Usually unnecessary: `calc_population` already copies the category's colors onto the modality it builds, so a `DatasetCollection`/`SetCollection` that has already run it carries its own colors.	`None`
`group_attrs`	`list[str] \| None`	Dataset/set `obs` columns to show as Clustergram column attribute tracks (e.g. `["condition", "timepoint"]`).	`None`
`normalized`	`bool \| None`	Column-normalize each bar to 100%. Defaults to `True` for proportion matrices and `False` for count matrices.	`None`
`col_weights`	`dict[str, float] \| None`	Optional `{group: n_cells}` true per-group magnitude, used to scale bar height in non-normalized ("counts") mode. Defaults to `DatasetCollection`/`calc_population`'s own `n_cells` obs column when available — pass explicitly to override, e.g. for a plain `DataFrame` input.	`None`
`cluster`	`bool`	Run hierarchical clustering before display (default `True`).	`True`
`name`	`str`	Clustergram registry name.	`'composition'`
`width / height`		Widget size in pixels.	required
`**kwargs`	`Any`	Forwarded to :class:`Clustergram`.	`{}`

`Enrich`

Bases: AnyWidget

A widget for interactive enrichment analysis using the Enrichr API.

Allows users to select a gene list, choose an enrichment library, and specify the number of terms to display. Automatically replaces older widgets with the same name to prevent notebook bloat.

`Landmark`

Bases: AnyWidget

A widget for interactively marking corresponding landmark points across dataset slices, for procrustes/thin-plate-spline alignment.

Parameters:

Name	Type	Description	Default
`adatas`	`AnnData \| list[AnnData] \| None`	A single `AnnData` with `slice_attr` given, or a list of per-slice `AnnData` (list order is slice order) — the same two input shapes :func:`~celldega.align.landmarks.calc_landmarks` accepts. Must resolve to at least 2 slices; any number beyond 2 is fine — both viewports can swap to any slice in the resolved set after construction (driven by the front-end dropdowns), not just the initial pair.	`None`
`slice_attr`	`str \| None`	For a single combined `AnnData`, the `obs` column identifying each cell's slice.	`None`
`slices`	`Sequence[Any] \| None`	The initial `(slice_id_a, slice_id_b)` pair to show. Defaults to the first two resolved slice ids.	`None`
`cluster_key`	`str \| None`	Optional `obs` column to color centroids by, for visual context while marking. Purely cosmetic — has no effect on the resulting landmark table.	`None`
`slice_labels`	`dict[Any, str] \| None`	Optional `{slice_id: display_name}` overrides for the dropdown/panel labels. Slices not present default to `str(slice_id)`.	`None`
`landmarks`	`DataFrame \| None`	An optional pre-existing landmark table to load — :func:`~celldega.align.landmarks.calc_landmarks`'s output shape (columns `label`, `x`, `y`, and a slice-tagging column matching `slice_attr`), or this widget's own `.landmarks` from a previous session. Immediately visible on whichever of the initial pair already has points, reviewable/editable via MARK and MODIFY exactly like anything marked in this session, and safe to append new landmark sets onto (the auto-numbered label counter starts past whatever numeric labels are already present).	`None`
`cell_radius`		Cell-point radius in data-space (micron) units — the points scale with zoom, exactly like Landscape's cell layer (which defaults to `5.0`), with a 1px on-screen floor. `width`/`height`/`cell_radius` are all plain synced traits, so this can also be set as `Landmark(..., cell_radius=8.0)` or afterward via `lm.cell_radius = 8.0`. The CELL control has a runtime opacity slider instead — radius rarely needs mid-session adjustment once set.	required
`landscapes`	`Any`	Not implemented yet — planned future alternative to `adatas` that would point Landmark directly at two :class:`~celldega.viz.widget.Landscape` instances.	`None`

Raises:

Type	Description
`NotImplementedError`	If `landscapes` is given.
`ValueError`	If `adatas` is missing, resolves to fewer than 2 slices, if a `slices` id isn't found among them, if a selected slice is missing `obsm["spatial"]`, or if `landmarks` is missing a required column or references a slice id not among `adatas`.

`calc_alignment_transform(**kwargs)`

Fit a transform directly from the currently-marked landmarks.

Convenience wrapper over :func:~celldega.align.serial_slices.calc_alignment_transform — equivalent to calling it on :attr:landmarks directly.

`close()`

Close the widget and notify the frontend to release resources.

`Landscape`

Bases: AnyWidget

A widget for interactive visualization of spatial omics data. This widget currently supports segmented spatial transcriptomics data (Xenium, MERSCOPE, Visium HD) and H&E image data.

Parameters:

Name	Type	Description	Default
`ini_x`	`float`	The initial x-coordinate of the view.	required
`ini_y`	`float`	The initial y-coordinate of the view.	required
`ini_zoom`	`float`	The initial zoom level of the view.	required
`rotation_orbit`	`float`	Rotating angle around orbit axis for point-cloud views.	required
`rotation_x`	`float`	Rotating angle around X axis for point-cloud views.	required
`token`	`str`	The token traitlet.	required
`base_url`	`str or list`	The base URL(s) for the widget. Can be a single string or a list of dicts with 'url' and 'label' keys for multiple datasets. Example: [{'url': 'http://...', 'label': 'Dataset1'}, ...] You can also pass a simple list of URL strings.	required
`dataset_names`	`list`	Short names for the datasets to display in the dropdown selector. Should match the length of base_urls. Example: ['Brain', 'Kidney'] for two datasets.	required
`rotate`	`float`	Degrees to rotate the 2D landscape visualization.	required
`AnnData`	`AnnData`	AnnData object to derive metadata from.	required
`dataset_name`	`str`	The name of the dataset to visualize. This will show up in the user interface bar.	required
`cell_name_prefix`	`bool`	If True, cell names in adata.obs.index are assumed to have a dataset prefix (e.g., "dataset-name_cell-name") that should be trimmed when mapping to LandscapeFiles. Default: False.	required
`use_adata_3d_centroids`	`bool`	For `technology="point-cloud"` views given an `adata`, render that AnnData's `obsm["spatial"]`/`obs[z_key]` centroids instead of the geometry baked into `cell_metadata.parquet` — no DegaFiles rewrite needed to preview a candidate alignment. Has no effect on 2D (non point-cloud) views, which always use the on-disk x/y. Default: True.	required
`z_key`	`str`	`adata.obs` column holding the Z coordinate used for `use_adata_3d_centroids` (falls back to 0 if absent). Default: "Z".	required

use_adata_3d_centroids writes centroids to a small file next to base_url and fetches it over HTTP when base_url is a local celldega.viz.get_local_server() address (millions of per-cell centroids don't fit through the widget's comm channel); otherwise it falls back to syncing them directly through the widget state, which is fine for smaller datasets.

A point-cloud (3D) view requires a real, pre-built DegaFiles base_url like any other technology — build one with the celldega.pre module (e.g. after running an alignment with :func:~celldega.align.serial_slices.align_serial_slices, regenerate LandscapeFiles from the aligned AnnData before visualizing it). adata here is only ever used for cell attributes/metadata, never for spatial positions.

The AnnData input automatically extracts cell attributes (e.g., leiden clusters), the corresponding colors (or derives them when missing), and any available UMAP coordinates.

`close()`

Close the widget and notify the frontend to release resources.

`highlight_cells(cell_ids)`

Highlight specific cells by their identifiers.

`trigger_update(new_value)`

Update the update_trigger traitlet with a new value.

`update_cell_clusters(new_clusters)`

Update cell clusters with new data.

`NeighborhoodCloud`

Bases: _SpatialWidget

3D orbit view of neighborhood alpha shapes (replaces Landscape(technology="neighborhood-cloud")).

Renders a pre-built neighborhood-cloud DegaFiles directory: one precomputed alpha-shape polygon per (cluster, slice), cheap to display regardless of dataset size, with real cell centroids loaded on demand when a cluster is selected. Build one with :func:celldega.align.write_nbhd_cloud (or :func:celldega.pre.write_nbhd_cloud_dataset).

The neighborhood geometry lives entirely on disk under the DegaFiles nbhd_cloud/ tree and is fetched by the front-end; this widget carries only the shared spatial trait surface plus the 3D orbit camera. (The 2D neighborhood drawing editor is a Landscape feature and intentionally not part of NeighborhoodCloud.)

Parameters:

Name	Type	Description	Default
`base_url`	`str or list`	DegaFiles URL(s), as in `Landscape`.	required
`adata`	`AnnData`	Source of cell attributes/metadata.	required

`Yearbook`

Bases: AnyWidget

A widget for visualizing cell portraits in a yearbook-style grid layout.

This widget creates a grid of cell "portraits" - zoomed-in views centered on selected cells. All portraits share synchronized zoom state but display different spatial regions. The control panel works similarly to Landscape, showing gene and cell bars based on visible content.

Parameters:

Name	Type	Description	Default
`base_url`	`str`	The base URL for the dataset.	required
`cells`	`list`	List of cell identifiers to display as portraits. If not provided and no query is given, random cells will be selected.	required
`selection`	`Selection or dict or list`	Ordered selection of cells to display as portraits. Accepts a `celldega.select.Selection` returned by `dega.select.Selector.select`, a JSON-ready selection dict, or a plain list of cell ids. Yearbook uses its ids as the portrait cell order and stores the JSON-ready payload for provenance. Pass either `selection` or `cells`, not both.	required
`front_end_query`	`dict`	Stateless query evaluated in the browser against LandscapeFiles (no Python/AnnData required). This is separate from the Python-side `celldega.select` query module. Supports the following formats: Cluster only: `{"cluster": {"attr": "leiden", "value": "8"}}` Returns random cells from the specified cluster. Gene only: `{"gene": "BRCA1"}` Returns cells ranked by gene expression (highest first). Cluster + Gene: `{"cluster": {"attr": "leiden", "value": "8"}, "gene": "BRCA1"}` Returns cells from the cluster ranked by gene expression. Max cells: `{"max_cells": 100}` Limits the number of cells returned (default: num_rows * num_cols * 10). (The former `query` argument is deprecated; it now maps to `front_end_query`.)	required
`num_rows`	`int`	Number of rows in the portrait grid. Alias: `rows`.	required
`num_cols`	`int`	Number of columns in the portrait grid. Alias: `cols`.	required
`portrait_size_um`	`float`	Size of each portrait in micrometers.	required
`portrait_gap`	`int`	Gap between portraits in pixels. Default is 4.	required
`pixel_width`	`float`	Pixel width for scale bar calculation. If provided, a scale bar will be displayed.	required
`token`	`str`	Authentication token for data access.	required
`dataset_name`	`str`	Name to display in the UI.	required
`width`	`int`	Widget width in pixels. 0 means 100%.	required
`height`	`int`	Widget height in pixels.	required
`segmentation`	`str`	Segmentation version to use. Default is "default".	required
`adata`	`AnnData`	AnnData object for cell metadata.	required
`cell_attr`	`list`	List of cell attributes to extract from adata.	required

Example::

# Using an explicit list of cell ids
yb = Yearbook(
    base_url="https://path-to-dataset",
    cells=["cell_1", "cell_2", "cell_3", "cell_4"],
    rows=2,
    cols=2,
    portrait_size_um=100,
)

# Using a Python selector result
selector = dega.select.Selector(adata)
selection = selector.select(query=selector.attr("leiden") == "5")
yb = Yearbook(
    base_url="https://path-to-dataset",
    selection=selection,
    rows=2,
    cols=2,
)

# Using a stateless front-end query (no AnnData needed)
yb = Yearbook(
    base_url="https://path-to-dataset",
    front_end_query={"gene": "BRCA1", "max_cells": 50},
    rows=2,
    cols=2,
    portrait_size_um=100,
)

`total_pages` `property`

Calculate total number of pages based on cells and grid size.

`close()`

Close the widget and notify the frontend to release resources.

`go_to_page(page)`

Navigate to a specific page.

`next_page()`

Navigate to next page.

`prev_page()`

Navigate to previous page.

`clustergram_enrich(cgm, *, row_enrich=True, col_enrich=False)`

Display a Clustergram widget and an Enrich widget side by side.

Parameters:

Name	Type	Description	Default
`cgm`	`Clustergram`	A `Clustergram` widget.	required
`row_enrich`	`bool`	If True (default), run enrichment analysis when row dendrogram clusters are selected.	`True`
`col_enrich`	`bool`	If True, run enrichment analysis when column dendrogram clusters are selected.	`False`

Returns:

Name	Type	Description
`HBox`	`HBox`	Visualization display containing both widgets.

`get_local_server()`

Start a local HTTP server with CORS support and return the port number.

Returns:

Name	Type	Description
`int`	`int`	The port number on which the server is running.

`get_proxy_server(remote_base_url=None, verbose=False)`

Start a local proxy server that forwards requests to a remote URL.

This is useful for bypassing CORS restrictions when the remote server (like Hugging Face) doesn't support CORS for Range requests.

Security: The proxy validates all URLs to prevent SSRF attacks: - Only http/https schemes are allowed - Private/loopback IP addresses are blocked - When remote_base_url is set, /proxy/ requests are constrained to that host

Parameters:

Name	Type	Description	Default
`remote_base_url`	`str \| None`	Optional base URL for the remote server. If provided, requests to the proxy will be forwarded to this URL. If not provided, use /proxy/FULL_URL format.	`None`
`verbose`	`bool`	If True, print log messages.	`False`

Returns:

Name	Type	Description
`int`	`int`	The port number on which the proxy server is running.

Example

port = get_proxy_server("https://huggingface.co/datasets/user/repo/resolve/main/folder")

Now use http://localhost:{port}/file.parquet

Or use http://localhost:{port}/proxy/https://huggingface.co/.../other/file.parquet

`landscape_clustergram(landscape, mat, width='600px', height='700px', *, enrich=False, row_enrich=True, col_enrich=False, enrich_kwargs=None)`

Deprecated alias for :func:spatial_clustergram, kept for backward compatibility. Prefer spatial_clustergram, which also works with CellCloud, NeighborhoodCloud, and Yearbook.

`landscape_yearbook(landscape, yearbook, width='100%', height='400px', cluster_attr='leiden')`

Display a Landscape widget above a Yearbook widget with linked queries.

When the user clicks on a cluster in the Landscape, the Yearbook automatically updates to show cells from that cluster. When a gene is selected, cells are ranked by gene expression.

Parameters:

Name	Type	Description	Default
`landscape`	`Landscape`	A `Landscape` widget.	required
`yearbook`	`Yearbook`	A `Yearbook` widget.	required
`width`	`str`	The width of the widgets.	`'100%'`
`height`	`str`	The height of each widget.	`'400px'`
`cluster_attr`	`str`	The cell attribute (`adata.obs` column) a clicked cluster value refers to (default `"leiden"`). If the click payload carries its own `attr` it takes precedence, so linked Clustergrams over non-leiden sets (cell types, domains) color the right cells.	`'leiden'`

Returns:

Name	Type	Description
`VBox`	`VBox`	Visualization display containing both widgets stacked vertically.

Example::

landscape = dega.viz.Landscape(base_url="...", adata=adata)
yearbook = dega.viz.Yearbook(base_url="...", rows=2, cols=4)
display = dega.viz.landscape_yearbook(landscape, yearbook)

`landscape_yearbook_clustergram(landscape, yearbook, cgm, width='600px', height='400px', cluster_attr=None)`

Display a Landscape and Clustergram side by side, with a Yearbook below.

All three widgets are linked: - Clustergram clicks update both Landscape and Yearbook - Gene selections rank cells in Yearbook by expression - Cluster selections filter cells in Yearbook

Parameters:

Name	Type	Description	Default
`landscape`	`Landscape`	A `Landscape` widget.	required
`yearbook`	`Yearbook`	A `Yearbook` widget.	required
`cgm`	`Clustergram`	A `Clustergram` widget.	required
`width`	`str`	The width of each widget in the top row.	`'600px'`
`height`	`str`	The height of each widget.	`'400px'`

Returns:

Name	Type	Description
`VBox`	`VBox`	Visualization display with Landscape+Clustergram on top, Yearbook below.

Example::

landscape = dega.viz.Landscape(base_url="...", adata=adata)
yearbook = dega.viz.Yearbook(base_url="...", rows=2, cols=4)
cgm = dega.viz.Clustergram(matrix=mat)
display = dega.viz.landscape_yearbook_clustergram(landscape, yearbook, cgm)

`spatial_clustergram(spatial, mat, width='600px', height='700px', *, enrich=False, row_enrich=True, col_enrich=False, enrich_kwargs=None, cluster_attr=None)`

Display a spatial widget and a Clustergram widget side by side, linked so that clicking a Clustergram row/column updates the spatial widget.

Works with any of celldega's spatial widgets: Landscape, CellCloud, NeighborhoodCloud, or Yearbook. Landscape/CellCloud/ NeighborhoodCloud share an update_trigger trait and are linked via a front-end jslink (no round-trip through Python); Yearbook has no such trait and is instead linked by observing the Clustergram's click_info in Python and translating it into a front_end_query (same mechanism as landscape_yearbook_clustergram).

Parameters:

Name	Type	Description	Default
`spatial`	`Landscape \| CellCloud \| NeighborhoodCloud \| Yearbook`	The spatial widget to link.	required
`mat`	`Clustergram`	A `Clustergram` widget.	required
`width`	`str`	The width of the widgets.	`'600px'`
`height`	`str`	The height of the widgets.	`'700px'`
`enrich`	`bool \| Enrich`	If True, create an `Enrich` widget; if an `Enrich` instance is provided, use it directly. If False, no enrichment widget is shown. Ignored for a `Yearbook` `spatial`.	`False`
`row_enrich`	`bool`	If True (default), run enrichment analysis when row dendrogram clusters are selected.	`True`
`col_enrich`	`bool`	If True, run enrichment analysis when column dendrogram clusters are selected.	`False`
`enrich_kwargs`	`dict \| None`	Optional kwargs passed to `Enrich` when `enrich=True`.	`None`
`cluster_attr`	`str \| None`	The cell attribute (`adata.obs` column) a clicked cluster refers to. Only used when `spatial` is a `Yearbook`; defaults to the Clustergram's own `col_entity` attribute (see `_clustergram_col_attr`).	`None`

Returns:

Name	Type	Description
`HBox`	`HBox`	Visualization display containing the widgets.

Viz Module API Reference

Widget Classes

Yearbook Selection

Front-End Query

CellCloud

Clustergram

manual_cat_dict property

__init__(**kwargs)

Parameters

close()

to_cluster(axis='row', n_clusters=None, threshold=None, criterion=None)

Composition

__init__(data, modality='population', *, category=None, colors=None, adata=None, group_attrs=None, normalized=None, col_weights=None, cluster=True, name='composition', width=700, height=450, **kwargs)

Enrich

Landmark

calc_alignment_transform(**kwargs)

close()

Landscape

close()

highlight_cells(cell_ids)

trigger_update(new_value)

update_cell_clusters(new_clusters)

NeighborhoodCloud

Yearbook

total_pages property

close()

go_to_page(page)

next_page()

prev_page()

clustergram_enrich(cgm, *, row_enrich=True, col_enrich=False)

get_local_server()

get_proxy_server(remote_base_url=None, verbose=False)

Now use http://localhost:{port}/file.parquet

Or use http://localhost:{port}/proxy/https://huggingface.co/.../other/file.parquet

landscape_clustergram(landscape, mat, width='600px', height='700px', *, enrich=False, row_enrich=True, col_enrich=False, enrich_kwargs=None)

landscape_yearbook(landscape, yearbook, width='100%', height='400px', cluster_attr='leiden')

landscape_yearbook_clustergram(landscape, yearbook, cgm, width='600px', height='400px', cluster_attr=None)

spatial_clustergram(spatial, mat, width='600px', height='700px', *, enrich=False, row_enrich=True, col_enrich=False, enrich_kwargs=None, cluster_attr=None)

`CellCloud`

`Clustergram`

`manual_cat_dict` `property`

`init(**kwargs)`

`close()`

`to_cluster(axis='row', n_clusters=None, threshold=None, criterion=None)`

`Composition`

`init(data, modality='population', *, category=None, colors=None, adata=None, group_attrs=None, normalized=None, col_weights=None, cluster=True, name='composition', width=700, height=450, **kwargs)`

`Enrich`

`Landmark`

`calc_alignment_transform(**kwargs)`

`close()`

`Landscape`

`close()`

`highlight_cells(cell_ids)`

`trigger_update(new_value)`

`update_cell_clusters(new_clusters)`

`NeighborhoodCloud`

`Yearbook`

`total_pages` `property`

`close()`

`go_to_page(page)`

`next_page()`

`prev_page()`

`clustergram_enrich(cgm, *, row_enrich=True, col_enrich=False)`

`get_local_server()`

`get_proxy_server(remote_base_url=None, verbose=False)`

`landscape_clustergram(landscape, mat, width='600px', height='700px', *, enrich=False, row_enrich=True, col_enrich=False, enrich_kwargs=None)`

`landscape_yearbook(landscape, yearbook, width='100%', height='400px', cluster_attr='leiden')`

`landscape_yearbook_clustergram(landscape, yearbook, cgm, width='600px', height='400px', cluster_attr=None)`

`spatial_clustergram(spatial, mat, width='600px', height='700px', *, enrich=False, row_enrich=True, col_enrich=False, enrich_kwargs=None, cluster_attr=None)`