Beta Release - CVXlab is currently in beta (v1.0.0b1). The API may change before the stable 1.0.0 release. Documentation is under active development.
CVXlab is an open-source Python laboratory for modeling and solving convex optimization problems. It extends cvxpy with user-friendly interfaces, integrated data management and support for multiple, interconnected optimization models.
From PyPI (once published):
pip install cvxlabFrom source (for development):
git clone https://github.com/cvxlab/cvxlab.git
cd cvxlab
pip install -e .[docs]See the Installation Guide for detailed instructions.
CVXlab allows you to define optimization problems using:
- General-purpose model generator: Model problems as you would mathematically, without restrictive solver forms.
- Almost no-code required: Build models using Excel or YAML—no coding required.
- Centralized data management: Centralized data input/output via SQLite database.
- Multi-Model Support: Generate and solve multiple integrated or decomposed optimization problems.
- Powerful engine embedded: Built on cvxpy package, leveraging its extensive solver support.
Typical workflow:
The figure below provides a synthetic and simplified overview of the CVXlab modeling process.
In generating and handling a CVXlab model, the user must follow the five fundamental activities summarized below:
- The user defines the model settings and the related structure: model scope, structure of variables, and list of mathematical expressions, including equalities, inequalities and (eventually) objective function. This activity requires almost no coding, as model definition can be performed via Excel files or YAML configuration files.
- The user proceeds by generating a CVXlab Model object, consisting in a Python class instance embedding all the model settings and the methods useful to manage the model. At the same time, other items are generated, including the SQLite database file (to store all model data), and the Excel files serving as blank templates for collecting exogenous data from the user.
- The user feeds input data to SQLite database through blank Excel template files. Specifically, user defines the data input required to characterize exogenous model variables.
- The numerical problem is generated, exogenous data fetched from the database, and the problem is solved through CVXPY engine.
- If problem is successfuly solved, results are finally exported to the database. Due to the structure of the relational database, it can be easily linked and inspected via Excel or SQL queries, or imported into Business Intelligence tools (such as PowerBI or Tableau) for more elaborated data visualization and analysis.
Full documentation is available at cvxlab.readthedocs.io. You can also browse the source documentation in the docs/source directory.
See CHANGELOG.md for a detailed history of changes.
We welcome contributions from the community! Please see CONTRIBUTING.md for guidelines.
Submit issues and ideas for improvements in GitHub GitHub Issues
Licensed under the Apache License 2.0. See LICENSE for details.
If you use CVXlab in academic work, please cite our papers. For industry use, we'd love to hear your feedback—reach out via email (matteovincenzo.rocco@polimi.it).