New service: serve output OME-Zarr

[tcompa]

This would be a nice-to-have prototype of a feature that we will eventually include in Fractal.

After the demos service is complete, we could startup a image-viewer service so that the user can directly go to localhost:something and view the output image.

[tcompa]

Here is a log of a preliminary exploration.

Serve the zarr array via http

I could not find whether vizarr can work with local zarr arrays. If not, we have to serve the array via http.
I used https://github.com/manzt/simple-zarr-server, with a serve.py script like

from simple_zarr_server import serve
import zarr
  
group = zarr.open_group("/.../fractal-tasks-core/tests/data/10_5281_zenodo_8091756/plate.zarr/B/03/0")
serve(group, allowed_origins=["*"])

Run vizarr locally

I cloned the https://github.com/hms-dbmi/vizarr repo, and I modified vite.config.js as in

const source = process.env.VIZARR_DATA || 'http://127.0.0.1:8000'

Other than that, a combination of npm install and npm start already works.

The first view is empty:

But what is missing is simply to select the DAPI channel (maybe there is an option to provide a default):

(note: one could even start by using the existing vizarr instance, e.g. with https://hms-dbmi.github.io/vizarr/?source=http://127.0.0.1:8000)

[jluethi]

Oh, that's super promising! Will want to try this locally as well, thanks for sharing the steps!

[tcompa]

Interesting link: hms-dbmi/vizarr#176

EDIT:

group = zarr.open("/somewhere.zarr")
viewer = vizarr.Viewer()
viewer.add_image(group)

[tcompa]

Closing (old issue)