Installation¶

Requirements¶

makeshift requires Python ≥ 3.8. Its runtime dependencies are all mainstream scientific-Python packages: numpy, pandas, scipy, scikit-learn, matplotlib, seaborn, nmrglue, and tqdm.

Install from GitHub¶

pip install makeshift-nmr

Install for local development¶

git clone https://github.com/WaymentSteeleLab/makeshift.git
cd makeshift
pip install -e .

Optional extras¶

The demo notebooks need Jupyter:

pip install "makeshift[demos]"   # adds jupyter

NumPy 2.x and nmrglue

The makeshift.spectra and makeshift.relaxation modules import nmrglue, whose current release (0.11) is not compatible with NumPy 2.x (TypeError: data type 'a8' not understood). If you hit this, pin an older NumPy alongside a compatible SciPy:

pip install "numpy<2" "scipy<1.14"

The core modules (NMRStarEntry, ChemicalShifts, PeakList) do not import nmrglue and work fine on NumPy 2.x.

Component-specific downloads¶

Two subsystems fetch external assets on demand rather than bundling them:

TALOS-N — the NIH binary and its database are downloaded into a specified directory. See TALOS-N prediction.
Structures & datasets — Use PDB or AlphaFoldDB structure and example datasets are fetched and cached under ~/.makeshift/. See Datasets & structures.