Installation

Requirements

scikit-agent requires Python 3.10 or higher. Its main runtime dependencies are:

  • NumPy, SciPy, SymPy, Numba, and xarray

  • PyTorch

  • Matplotlib

  • cairosvg, pydot, PyYAML, and IPython (model visualization)

See pyproject.toml for the authoritative dependency list; installing the package pulls these in automatically.

System Dependencies

The cairosvg and pydot packages link against the Cairo and Graphviz system libraries, which must be installed separately for model visualization.

On macOS:

brew install cairo libffi graphviz

And link Cairo by setting the environment variable:

DYLD_LIBRARY_PATH=/opt/homebrew/lib:$DYLD_LIBRARY_PATH

On Debian/Ubuntu Linux:

sudo apt-get install libcairo2 graphviz

Install from PyPI

To add scikit-agent to a uv project:

uv add scikit-agent

To install it into a standalone environment, use uv pip install scikit-agent (or plain pip install scikit-agent).

Install from Source

Source and development installs use uv. To install the latest development version:

git clone https://github.com/scikit-agent/scikit-agent.git
cd scikit-agent
uv sync

Development Installation

For development, sync with the test and docs extras (uv sync is exact, so include every extra you need in one command):

git clone https://github.com/scikit-agent/scikit-agent.git
cd scikit-agent
uv sync --extra test --extra docs

Verify Installation

import skagent

print(skagent.__version__)

For a source install managed by uv, you can instead run:

uv run python -c "import skagent; print(skagent.__version__)"