Support Nifti feature type (bump datasets dependency)

[The-Obstacle-Is-The-Way]

Problem

The Dataset Viewer fails on datasets that use the Nifti feature type with:

ValueError: Feature type 'Nifti' not found. Available feature types: ['Value', 'ClassLabel', 'Translation', 'TranslationVariableLanguages', 'LargeList', 'List', 'Array2D', 'Array3D', 'Array4D', 'Array5D', 'Audio', 'Image', 'Video', 'Pdf']

Affected dataset: hugging-science/arc-aphasia-bids

Root Cause

1. huggingface/datasets added Nifti support in PR #7815 (merged 2025-10-24)
2. dataset-viewer pins datasets==4.1.1 in libs/libcommon/pyproject.toml
3. Nifti is only in the datasets main branch (will be in next PyPI release)

When the Dataset Viewer parses a README with dtype: nifti in the YAML frontmatter, it crashes because 4.1.1 doesn't know what Nifti is.

Full Traceback

File "/src/services/worker/src/worker/job_runners/dataset/config_names.py", line 66, in compute_config_names_response
  config_names = get_dataset_config_names(
File "/usr/local/lib/python3.12/site-packages/datasets/inspect.py", line 161, in get_dataset_config_names
  dataset_module = dataset_module_factory(
File "/usr/local/lib/python3.12/site-packages/datasets/load.py", line 1031, in dataset_module_factory
  raise e1 from None
File "/usr/local/lib/python3.12/site-packages/datasets/load.py", line 605, in get_module
  dataset_infos = DatasetInfosDict.from_dataset_card_data(dataset_card_data)
File "/usr/local/lib/python3.12/site-packages/datasets/info.py", line 386, in from_dataset_card_data
  dataset_info = DatasetInfo._from_yaml_dict(dataset_card_data["dataset_info"])
File "/usr/local/lib/python3.12/site-packages/datasets/info.py", line 317, in _from_yaml_dict
  yaml_data["features"] = Features._from_yaml_list(yaml_data["features"])
File "/usr/local/lib/python3.12/site-packages/datasets/features/features.py", line 2031, in _from_yaml_list
  return cls.from_dict(from_yaml_inner(yaml_data))
File "/usr/local/lib/python3.12/site-packages/datasets/features/features.py", line 1876, in from_dict
  obj = generate_from_dict(dic)
File "/usr/local/lib/python3.12/site-packages/datasets/features/features.py", line 1463, in generate_from_dict
  return {key: generate_from_dict(value) for key, value in obj.items()}
File "/usr/local/lib/python3.12/site-packages/datasets/features/features.py", line 1469, in generate_from_dict
  raise ValueError(f"Feature type '{_type}' not found. Available feature types: {list(_FEATURE_TYPES.keys())}")
ValueError: Feature type 'Nifti' not found.

Proposed Fix

Bump datasets dependency in libs/libcommon/pyproject.toml to a version that includes Nifti:

# Current
datasets = {version = "4.1.1", extras = ["vision"]}

# After next datasets release (or use git)
datasets = {version = ">=4.5.0", extras = ["vision"]}

Context

• Nifti is for NIfTI neuroimaging files (.nii, .nii.gz) - standard format for MRI/fMRI data
• The datasets library now has full support including:
  • Nifti() feature type (#7815)
  • embed_storage for uploads (#7853)
  • Visualization via niivue (#7878)
• This enables ML workflows on medical imaging datasets

Related

• Add nifti support huggingface/datasets#7815 - Add nifti support
• Fix embed storage nifti huggingface/datasets#7853 - Fix embed storage nifti
• Replace papaya with niivue huggingface/datasets#7878 - Nifti visualization with niivue