G-Cubed Python runtime and dependency inventory
Table of contents
This page is for client IT and security teams that need to approve, mirror and maintain the Python artefacts used to run G-Cubed in an internal package repository, restricted network or air-gapped environment.
The inventory applies to G-Cubed package 5.0.1.4.
Required Python version
G-Cubed 5.0.1.4 requires CPython 3.14.5 exactly. Its package metadata uses:
Requires-Python: ==3.14.5
This is an exact patch-version requirement, not “3.14 or later”.
Python is not installed from a Python package repository. Approve the CPython interpreter or the base image that supplies it separately, including its provider, operating system, CPU architecture, checksum, provenance and patching policy.
Dependency scope
The gcubed wheel declares 11 direct runtime dependencies. Those packages introduce another 10 indirect, or transitive, dependencies in the reference Linux environment. There are no optional runtime extras in the G-Cubed package metadata.
The versions below are the complete resolved runtime set for G-Cubed 5.0.1.4 on CPython 3.14.5 in the Linux dev-container environment. The exact direct pins come from the gcubed package metadata. The indirect packages use the exact versions in the verified environment; their parent packages declare the minimum-version constraints shown in the table.
Direct dependencies
| Package | Exact version | Purpose in the runtime | Distribution |
|---|---|---|---|
| numpy | 2.4.6 | Arrays and numerical operations | Platform-specific binary wheel |
| pandas | 3.0.3 | Tabular model data and time series | Platform-specific binary wheel |
| pyarrow | 24.0.0 | Arrow and Parquet data interchange | Platform-specific binary wheel |
| regex | 2026.5.9 | Extended regular-expression processing | Platform-specific binary wheel |
| scipy | 1.17.1 | Optimisation, linear algebra and sparse matrices | Platform-specific binary wheel |
| Jinja2 | 3.1.6 | Report templating | Universal Python wheel |
| joblib | 1.5.3 | Serialisation of solved models and baselines | Universal Python wheel |
| fastparquet | 2026.5.0 | Parquet file support | Platform-specific binary wheel |
| PyYAML | 6.0.3 | YAML configuration files | Platform-specific binary wheel |
| jsonschema | 4.26.0 | Configuration-schema validation | Universal Python wheel |
| threadpoolctl | 3.6.0 | Inspection and control of native numerical thread pools | Universal Python wheel |
Indirect dependencies
| Package | Exact version | Introduced by and declared constraint | Distribution |
|---|---|---|---|
| attrs | 26.1.0 | jsonschema → attrs >=22.2.0; referencing → attrs >=22.2.0 | Universal Python wheel |
| cramjam | 2.11.0 | fastparquet → cramjam >=2.3 | Platform-specific binary wheel |
| fsspec | 2026.4.0 | fastparquet → fsspec (no version constraint) | Universal Python wheel |
| jsonschema-specifications | 2025.9.1 | jsonschema → jsonschema-specifications >=2023.03.6 | Universal Python wheel |
| MarkupSafe | 3.0.3 | Jinja2 → MarkupSafe >=2.0 | Platform-specific binary wheel |
| packaging | 26.2 | fastparquet → packaging (no version constraint) | Universal Python wheel |
| python-dateutil | 2.9.0.post0 | pandas → python-dateutil >=2.8.2 | Universal Python wheel |
| referencing | 0.37.0 | jsonschema → referencing >=0.28.4; jsonschema-specifications → referencing >=0.31.0 | Universal Python wheel |
| rpds-py | 2026.6.3 | jsonschema → rpds-py >=0.25.0; referencing → rpds-py >=0.7.0 | Platform-specific binary wheel |
| six | 1.17.0 | python-dateutil → six >=1.5 | Universal Python wheel |
Dependencies that are shared between direct packages are listed once. For example, numpy is required directly by G-Cubed and is also required by pandas, scipy and fastparquet; pandas is both direct and required by fastparquet.