Upgrade sopa version in the dependencies

CHANGELOG.md

@@ -6,6 +6,9 @@
 ### Changed
 - Cleanup: removed unused `pixel_size` argument from `novae.spatial_neighbors`
 
+### Fixed
+- Upgrade sopa to import `sopa.constants` instead of `sopa._constants` (#38)
+
 ## [1.0.1] - 2025-12-10
 
 🎉 Novae is now [published in Nature Methods](https://www.nature.com/articles/s41592-025-02899-6)!

docs/advice.md

@@ -37,7 +37,7 @@ In that case, you can use other backends (two options shown below).
 
 === "Backed H5"
 
-    You can read your data in "backed mode" to keep `adata.X` on disk and load only the requested slices into memory when needed. It enables scalable operations on large datasets without fully materializing them in RAM.
+    You can read your data in "backed mode" to keep `adata.X` on disk and load only the requested slices into memory when needed. It enables scalable operations on large datasets without fully materializing them in RAM. **Important:** for now, the backed mode only supports dense matrices (see [this issue](https://github.com/MICS-Lab/novae/issues/37#issuecomment-3748420274)).
 
     For instance, you can read an `h5ad` file as below:
     ```python

docs/tutorials/main_usage.ipynb

@@ -209,14 +209,16 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "To assign domains, you can use the [`assign_domains`](../../api/Novae/#novae.Novae.assign_domains) method, as below. By default, it creates 7 domains, but you can choose the number of domains you want with the `level` argument.\n",
+    "To assign domains, you can use the [`assign_domains`](../../api/Novae/#novae.Novae.assign_domains) method, as below.\n",
     "\n",
-    "The function will save the domains in `adata.obs`, and return the name of the column in which it was saved (in this case, `adata.obs[\"novae_domains_7\"]`)\n"
+    "We recommend to **try different number of domains**, depending on your biological questions. For that, just run this function multiple times with a different `level` (or check the others [API parameters](../../api/Novae/#novae.Novae.assign_domains)).\n",
+    "\n",
+    "The function adds a new column to `adata.obs`, and returns the column name. In this case, it saved the results in `adata.obs[\"novae_domains_7\"]`.\n"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 9,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -231,7 +233,7 @@
     }
    ],
    "source": [
-    "model.assign_domains(adata)"
+    "model.assign_domains(adata, level=7)"
    ]
   },
   {
@@ -891,7 +893,7 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "novae-ezkWKrh6-py3.10",
+   "display_name": "spatial-dev",
    "language": "python",
    "name": "python3"
   },
@@ -905,7 +907,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.10.14"
+   "version": "3.12.12"
   }
  },
  "nbformat": 4,

novae/data/_embeddings/_histo.py

@@ -1,5 +1,7 @@
+from __future__ import annotations
+
 import logging
-from typing import TYPE_CHECKING, Callable, Union
+from typing import TYPE_CHECKING, Callable
 
 import numpy as np
 from anndata import AnnData
@@ -16,7 +18,7 @@
 
 
 def compute_histo_embeddings(
-    sdata: "SpatialData",
+    sdata: SpatialData,
     model: str | Callable = "conch",
     table_key: str = "table",
     patch_overlap_ratio: float = 0.5,
@@ -43,7 +45,7 @@ def compute_histo_embeddings(
     """
     try:
         import sopa
-        from sopa._constants import SopaAttrs, SopaKeys
+        from sopa.constants import SopaAttrs, SopaKeys
         from spatialdata.models import get_table_keys
     except ImportError:
         raise ImportError(
@@ -105,7 +107,9 @@ def _quality_control_join(distances: np.ndarray):
 
 
 def compute_histo_pca(
-    sdatas: Union["SpatialData", list["SpatialData"]], n_components: int = 50, table_key: str = "table"
+    sdatas: SpatialData | list[SpatialData],
+    n_components: int = 50,
+    table_key: str = "table",
 ) -> None:
     """Run PCA on the histology embeddings associated to each cell (from the closest patch).
     The embedding is stored in `adata.obsm["histo_embeddings"]`, where `adata` is the table of cell expression.
@@ -123,7 +127,7 @@ def compute_histo_pca(
     if isinstance(sdatas, SpatialData):
         sdatas = [sdatas]
 
-    def _histo_emb(sdata: "SpatialData") -> np.ndarray:
+    def _histo_emb(sdata: SpatialData) -> np.ndarray:
         _table: AnnData = sdata.tables[table_key]
 
         assert "embedding_key" in _table.obs, (

pyproject.toml

@@ -27,7 +27,7 @@ dependencies = [
 ]
 
 [project.optional-dependencies]
-multimodal = ["sopa>=2.1.0", "spatialdata>=0.5.0", "timm>=1.0.15"]
+multimodal = ["sopa>=2.1.11", "timm>=1.0.15"]
 conch = ["einops-exts>=0.0.4", "transformers>=4.52.4"]
 
 [dependency-groups]