Failing with: TypeError: Failed to fetch

[djriffle]

My setup code:

import scanpy as sc
import easy_vitessce as ev

adata = sc.read_h5ad("path_to_my_file.h5ad")

ev.configure_plots(
    enable_plots=["embedding", "violin", "spatialdata-plot", "umap"]
)

AnnData preview:

AnnData object with n_obs × n_vars = 339751 × 458
    obs: 'dataset', ...
    var: 'Gene', 'Ensembl.ID', 'Probesets', 'category', 'highly_variable'
    uns: 'log1p', 'neighbors'
    obsm: 'X_pca', 'X_spatial', 'X_umap'
    obsp: 'connectivities', 'distances'

Problematic code:

sc.pl.pca(adata, color=["dataset"])

When I run this, the PCA plot fails with:

TypeError: Failed to fetch

There’s no additional traceback or details.
The strange part is:

• The same error happens on a different h5ad file.
• Using sc.pl.pca on its own (outside the Vitessce configuration) works fine.

Has anyone encountered this or knows why sc.pl.pca() would trigger a frontend Failed to fetch error after calling ev.configure_plots?

[keller-mark]

Hi @djriffle ,
Would you be able to share a few additional details to help with debugging:

• The notebook platform (e.g., Jupyter Notebook, JupyterLab, VSCode, Colab, etc) and version
• A screenshot of the notebook
• A screenshot of the browser console showing the error traceback if possible

Also, can you double check that easy_vitessce and Jupyter packages are installed in the same environment: https://github.com/vitessce/vitessce-python?tab=readme-ov-file#ensure-that-vitessce-and-jupyter-are-installed-in-the-same-environment

---

If your h5ad file is large, you can do the following to ensure that EasyVitessce does not need to save its own copy (docs) (but this is probably unrelated to the Failed to fetch error):

import scanpy as sc
import easy_vitessce as ev

adata = sc.read_h5ad("path_to_my_file.h5ad")
ev.register_data_path(adata, "path_to_my_file.h5ad")

# ...

[djriffle]

Hi @keller-mark,

I am using browser-based VS Code with the Jupyter extension (v2025.9.1), served from our HPC environment. After further investigation, it appears the issue is related to CORS restrictions when the Vitessce widget attempts to fetch JavaScript modules from unpkg.

Access to fetch at 'https://unpkg.com/' from origin 'https://1v9at5n64hcdavl6i5tdpfcnrci8bvui6oq64teu30qbgj0nbti8.vscode-cdn.net'
has been blocked by CORS policy: No 'Access-Control-Allow-Origin' header is present on the requested resource.
Failed to access CDN https://unpkg.com/ after 0 attempt(s), TypeError: Failed to fetch

I confirmed that the required packages (easy_vitessce, vitessce, and ipywidgets) are installed in the same conda environment.

I also tested using VS Code’s “Open in VS Code” feature to launch the remote session in my local desktop app. Surprisingly, the widget still fails in that context with the same error, though I’m not entirely sure whether that flow maintains the same browser-based rendering stack behind the scenes.

Please let me know if there’s a recommended workaround for VS Code notebooks running in this setup.

[langbinga]

Has it been resolved yet? I'm also in an HPC environment and encountered the same error as you.