Installation

uv add mlschema

MLschema is a pure-Python package on PyPI. Any standard installer works:

pip install mlschema
poetry add mlschema

Inside a Conda environment, activate it and use pip.

Requirements

Dependency	Supported range
Python	>=3.14,<3.15
pandas	>=3.0.3,<4.0.0
pydantic	>=2.13.4,<3.0.0

pandas and pydantic are installed with the package. Install them explicitly only when your own code imports them, which it usually does for pandas.

Pin a version

Inference rules and the field contract can change between minor versions while MLschema is pre-1.0. Pin the version you tested and upgrade deliberately after reading the changelog:

uv add "mlschema==0.3.0"
pip install "mlschema==0.3.0"

Verify the installation

from importlib.metadata import version

import pandas as pd

from mlschema import infer_schema

print(version("mlschema"))
print(infer_schema(pd.DataFrame({"score": [98.5, 86.0]})))

The second line prints a list with one field:

[{'label': 'score', 'required': True, 'mappedTo': 'score', 'kind': 'number', 'step': 0.1}]

Troubleshooting

Symptom	Likely cause	Resolution
No matching distribution found for mlschema	Python is outside >=3.14,<3.15, or PyPI is unreachable.	Check python --version and your package index.
ModuleNotFoundError: No module named 'mlschema'	The package is in a different environment.	Activate the intended environment and reinstall.
Most fields come back as text	Columns are stored as object, str, or a nullable extension dtype.	Convert columns to the dtypes listed in Field kinds.
EmptyDataFrameError	The DataFrame has no rows or no columns.	Check the data before inference.
pydantic.ValidationError	An override, builder, or custom kind produced invalid constraints.	See Errors.

Working on MLschema itself

The library uses uv, Ruff, and Pyright:

uv sync
uv run ruff check .
uv run ruff format --check .
uv run pyright
uv run pytest

This documentation is an Astro Starlight project under docs/, using Vite+ (vp) and the repository’s pnpm workspace. Use Node.js 22.18+, 24.11+, or 26+:

vp install --frozen-lockfile
vp run docs#dev
vp run docs#build