Installation

Requirements

Python 3.10, 3.11, 3.12, or 3.13. Older Python versions are not supported.
Core dependencies are installed automatically: pandas >= 1.5, numpy >= 1.23, scipy >= 1.10, statsmodels >= 0.14, scikit-learn >= 1.2.

From PyPI

pip install openpls-engine

Inside a virtual environment is recommended:

python3 -m venv .venv
source .venv/bin/activate
pip install openpls-engine

Pin a specific version

For reproducible analyses, pin the exact version you used in your paper or notebook:

pip install openpls-engine==1.12.0

The runtime exposes the installed version as openpls.__version__, which is handy for methodology sections and bug reports.

Use from an AI assistant (MCP)

Since 1.12.0 the engine ships a local Model Context Protocol server, so an assistant like Claude or Cursor can inspect a dataset and fit a model by calling the engine on your machine. Nothing is uploaded; the tools read datasets from local file paths.

pip install "openpls-engine[mcp]"

Point your MCP client at the openpls-mcp command, e.g. in Claude Desktop / Claude Code:

{
  "mcpServers": {
    "openpls": { "command": "openpls-mcp" }
  }
}

It exposes two tools: describe_dataset(path) (row count and per-column name/type, so the assistant can pick indicator columns) and fit_pls_sem(data_path, latent_variables, structural_paths, scheme, scaled, missing_codes) (R² per construct, path coefficients with estimate / p-value / f², reliability with Cronbach’s alpha, rho_A, rho_C, AVE, model fit SRMR / d_ULS, and the goodness-of-fit index).

From source

If you need the bleeding edge or want to contribute:

git clone https://github.com/jojacobsen/openpls-engine.git
cd openpls-engine
python3 -m pip install -e .

For development with the test suite and linter:

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements-dev.txt
pip install -e .

pytest          # run the test suite
ruff check .    # lint

Verify the install

python -c "import openpls; print(openpls.__version__)"

You should see the version number you installed (for example 1.0.2).

A note on the import name

The PyPI distribution is named openpls-engine, and the Python package you import is named openpls:

from openpls import Plspm
import openpls.config as c
from openpls.scheme import Scheme
from openpls.mode import Mode

Migrating from 0.7.x

1.0.0 renamed the import namespace from plspm to openpls. The PyPI distribution name (openpls-engine) is unchanged. To upgrade existing notebooks and scripts:

import plspm.config as c
from plspm.plspm import Plspm
from plspm.scheme import Scheme
from plspm.mode import Mode
import openpls.config as c
from openpls import Plspm
from openpls.scheme import Scheme
from openpls.mode import Mode

Plspm is now re-exported at the top level, so from openpls import Plspm is the recommended entry point.

1.0.0 also aligns indicator standardization with SmartPLS 4: each indicator is now standardized with its own (Bessel-corrected) sample standard deviation instead of a pooled block standard deviation. Loadings, R², HTMT, f² and Q² are scale-invariant and therefore unaffected; outer weights and composite scores now match the SmartPLS convention for mixed-scale models. See the changelog for the full list.