Installation
============

Requirements
------------

VANE targets **Python 3.10–3.12** and builds on the scientific Python stack
(NumPy, SciPy, pandas, matplotlib, Plotly, scikit-learn). A conda environment is
recommended for ABI-consistent binary wheels.

From source (recommended during alpha)
--------------------------------------

.. code-block:: bash

   conda create -n vane python=3.11 -y
   conda activate vane

   # Core scientific stack via conda-forge for ABI consistency
   conda install -c conda-forge numpy scipy pandas matplotlib plotly pyyaml scikit-learn -y

   # Install VANE in editable mode with the development tools
   pip install -e ".[dev]"

Optional extras
---------------

VANE keeps its core dependency-light; heavier or specialised components are opt-in
through extras:

``.[ml]``
   PyTorch-based experimental components (see :doc:`limitations`). The rule-based and
   Gaussian-process layers do **not** require this extra.

``.[docs]``
   The Sphinx documentation toolchain used to build this site.

``.[dev]``
   Linting (``ruff``), type checking (``mypy``), and testing (``pytest`` and
   plugins), plus build and release tooling.

.. code-block:: bash

   pip install -e ".[ml]"      # experimental ML components
   pip install -e ".[docs]"    # documentation toolchain

Verifying the installation
--------------------------

.. code-block:: bash

   python -c "import vane; print(vane.__version__)"
   pytest -q

The data-free example exercises the full analysis path without any OpenFAST files:

.. code-block:: bash

   python examples/synthetic_quickstart.py