Low-level access

The high-level object graph should cover normal analysis. Use Images and masks for exported image data. Use raw member access when you need bytes that the SDK deliberately does not interpret.

List archive members

with open_export("project.zip") as export:
    for path in sorted(export.members):
        print(path)

Read one member

with open_export("project.zip") as export:
    artifact = export.member("manifest.json")
    print(artifact.size_bytes)
    with artifact.open() as stream:
        payload = stream.read()

Each open() call returns an independent binary stream. Directory members are read directly from their existing files; ZIP members are decompressed as the stream is read.

If a library requires a filesystem path, use as_path():

from pathlib import Path

SCRATCH = Path("analysis-scratch")
SCRATCH.mkdir(exist_ok=True)

with open_export("project.zip") as export:
    artifact = export.member("data/slices/slice-1/images/image-1__cells.tif")
    with artifact.as_path(directory=SCRATCH) as path:
        use_library_that_requires_a_path(path)

An extracted export yields its existing member path. A ZIP member is copied into a temporary directory under SCRATCH and removed when the context exits.

Keep the safety boundary

Do not use raw access to bypass validation or to load NumPy object arrays. If a supported v2 member is missing from the high-level API, open an issue rather than depending on private modules such as neuroqp._reader.