numpydantic

What it is and what it does

Numpydantic bridges the gap between pydantic's data validation and array libraries by letting you specify arrays with precise shape and dtype constraints in type annotations. It solves the problem that pydantic cannot natively validate numpy arrays or other array types without workarounds—numpydantic provides an NDArray type that works seamlessly in pydantic models while also supporting static type checking via a mypy plugin.

The package works with multiple array backends (numpy, dask, zarr, hdf5, video files) through a pluggable interface, so the same type annotation can validate arrays from different libraries. You can use NDArray outside pydantic too, for isinstance checks or as a callable validator. Serialization to JSON-compatible formats is built in, and the package generates correct JSON Schema for your models, making them interoperable.

Use it for:

• Validate image or video data in pydantic models with exact shape and dtype requirements before processing.
• Define reusable array type aliases (e.g., GRAYSCALE) and catch shape mismatches at type-check time with mypy.
• Accept arrays from multiple backends (numpy, dask, zarr) in the same model without writing separate validators.
• Serialize array-containing models to JSON with metadata sufficient to reconstruct them in their native format.
• Lazy-load large arrays from hdf5 or zarr without full validation overhead by leveraging backend metadata.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Numpydantic adds type annotations and validation for arrays with shape and dtype constraints in pydantic models, supporting numpy, dask, zarr, hdf5, and video files through a pluggable interface.

Yes. Numpydantic is actively maintained, has no known vulnerabilities, and solves a real problem—pydantic's inability to validate arrays with shape and dtype constraints. Low install friction, permissive license, and support for multiple array backends make it a practical choice for data science and scientific computing workflows that use pydantic for validation.

Install

numpydantic on PyPI

pip install numpydantic

uv add numpydantic

poetry add numpydantic

Installing numpydantic

Before you install

Low friction: pure Python wheel with only three runtime dependencies (pydantic, numpy, typing-extensions). Actively maintained with a recent release 53 days ago and no known vulnerabilities.

License in practice

MIT license is permissive; you can use, modify, and distribute numpydantic freely in commercial and private projects with minimal restrictions.

Quickstart

pip install numpydantic

from pydantic import BaseModel
from numpydantic import NDArray, Shape
import numpy as np

class MyModel(BaseModel):
    array: NDArray[Shape["3 x, 4 y, * z"], int]

model = MyModel(array=np.zeros((3, 4, 5), dtype=int))

Requires Python 3.10 or later; optional array-library extras (dask, hdf5, zarr, video) needed for those backends.

Verify before relying

• Whether the mypy plugin (introduced in 1.9.0) is production-ready and widely adopted.
• Performance overhead of validation on large arrays compared to raw numpy operations.
• Stability of the extensible specification syntax mentioned as planned for v2.

Package facts

Evidence: numpydantic-1.10.0-py3-none-any.whl

Keywords: arrays, validation, serialization, numpy, pydantic

Tags

More Utilities packages