Rhetor: GenAI-Assisted MILP Modelling Environment

Rhetor is an Integrated Modelling Environment for researchers, students, and practitioners who model and solve mixed-integer linear programming (MILP) problems. It supports a subset of the Optimisation Programming Language (OPL) syntax and is built on PyOPL, a Python library for parsing and solving OPL-like mathematical programming models. Rhetor also provides optional GenAI assistance for drafting, checking, and refining mathematical models.

Install:

pip install rhetor

Start Rhetor:

pyopl

To enable GenAI features, set at least one of the following environment variables before starting Rhetor:

• OPENAI_API_KEY — for OpenAI models
• GEMINI_API_KEY — for Google Gemini

or install and run ollama locally.

Example (macOS / bash):

export OPENAI_API_KEY="sk-..."

Alternatively, you can use rhetor in Jupyter notebook

or you can use the Rhetor Skills and PyOPL MCP to integrate Rhetor with your favorite agentic workflow (e.g. Github Copilot).

• Write OPL-like mathematical programming models and data files.
• Parse, compile, validate, and solve supported MILP models.
• Use GenAI assistants to draft models, repair errors, and refine model/data pairs.
• Work from the desktop IDE, Python, Jupyter notebooks, or the PyOPL MCP integration.
• Explore practical examples including knapsack, scheduling, routing, production, and facility location.

Rhetor can solve supported models with Gurobi or the SciPy/HiGHS backend. Optional GenAI features support OpenAI and Google Gemini through API keys, or Ollama for local models.

Rhetor embeds GenAI features that support the modelling process: features can be explored within the system in a conversational fashion. A complete user guide and examples for PyOPL are available at:

• User guide
• Gallery of examples
• Sample PyOPL models

A selection of LinkedIn posts about Rhetor is available below.

Rhetor is released under the MIT License.

Rossi, R. & Prestwich, S. D. (2026).
Grammar-Aware Literate Generative Mathematical Programming with Compiler-in-the-Loop.
arXiv:2601.17670 [cs.PL].

@misc{rossi2026grammarawareliterategenerativemathematical,
            title={Grammar-Aware Literate Generative Mathematical Programming with Compiler-in-the-Loop},
            author={Roberto Rossi and Steven D. Prestwich},
            year={2026},
            eprint={2601.17670},
            archivePrefix={arXiv},
            primaryClass={cs.PL},
            url={https://arxiv.org/abs/2601.17670},
          }