Building Against NumPy

Finding numpy.get_include() in setup.py, numpy in build-system.requires of pyproject.toml, dependency('numpy') in meson.build, or cimport statements in .pyx or .pyd files are a telltale sign that the package links against NumPy.

Adding numpy in the host: section is sufficient to have a compatible NumPy version at run time:

v0 (meta.yaml)

requirements:
  host:
    - numpy

v1 (recipe.yaml)

requirements:
  host:
    - numpy

At the time of writing (June 2025), the above is equivalent to the following:

v0 (meta.yaml)

requirements:
  host:
    - numpy 2.*

v1 (recipe.yaml)

requirements:
  host:
    - numpy 2.*

See the pinning repository for what the pinning corresponds to at any given time.

In either case, the actual runtime requirements are determined through numpy's run-export, which is:

• >=1.2x,<2 if you're building against numpy 1.2x
• >=1.19,<3 if you're building against numpy 2.0
• >=1.21,<3 if you're building against numpy 2.1 or 2.2
• >=1.23,<3 if you're building against numpy 2.3

If the package you are building has a higher minimum requirement for numpy, please add this under run:

v0 (meta.yaml)

requirements:
  host:
    # leave this unpinned!
    - numpy
  run:
    - numpy >={{ the_lower_bound_your_package_needs }}

v1 (recipe.yaml)

requirements:
  host:
    # leave this unpinned!
    - numpy
  run:
    - numpy >=${{ the_lower_bound_your_package_needs }}