Installation

Note

ChemGraph requires Python 3.10+.

Install from PyPI (recommended)

pip install chemgraph

Default installation does not require tblite.

To include optional calculator extras (including tblite):

pip install "chemgraph[calculators]"

To use the experimental Codex subscription provider with an existing ChatGPT login, first follow the official Codex CLI installation guide. The Python SDK does not install the codex shell command. Then run:

pip install "chemgraph[codex]"
codex login

See Experimental Codex subscription support for usage and authentication constraints.

Warning

On platforms without a prebuilt tblite wheel, installing calculators may require a local Fortran toolchain.

Install from source

pip/venv

git clone https://github.com/argonne-lcf/ChemGraph
cd ChemGraph
python -m venv chemgraph-env
source chemgraph-env/bin/activate  # Windows: .\chemgraph-env\Scripts\activate
pip install -e .

For experimental Codex subscription support, install the Codex CLI as shown above, then include the optional extra in the editable install:

pip install -e ".[codex]"
codex login

conda

git clone --depth 1 https://github.com/argonne-lcf/ChemGraph
cd ChemGraph
conda env create -f environment.yml
conda activate chemgraph

uv

git clone https://github.com/argonne-lcf/ChemGraph
cd ChemGraph
uv venv --python 3.11 chemgraph-env
source chemgraph-env/bin/activate  # Windows: .\chemgraph-env\Scripts\activate
uv pip install -e .

Optional UMA install

uma and mace-torch can conflict through different e3nn requirements. Use separate environments if you need both MACE and UMA.

PyPI attempt:

pip install "chemgraph[uma]"

From source:

pip install -e ".[uma]"

If resolution fails, install UMA in a separate environment dedicated to UMA workflows.