minto 2.0.0rc1

Experiment management and benchmark tools for mathematical optimization

MINTO: Jij Management and Insight tool for Optimization

minto is a Python library designed for developers working on research and development or proof-of-concept experiments using mathematical optimization. Positioned similarly to mlflow in the machine learning field, minto provides features such as saving optimization results, automatically computing benchmark metrics, and offering visualization tools for the results.

Primarily supporting Ising optimization problems, plans to extend its support to a wide range of optimization problems, such as MIP solvers, in the future.

Installation

minto can be easily installed using pip.

pip install minto

Documentation and Support

Documentation is hosted on Read the Docs: https://jij-inc-minto.readthedocs-hosted.com/latest/ The published site exposes both English and Japanese under /en/ and /ja/.

Tutorials will be provided in the future. Stay tuned!

Building Documentation Locally

The documentation is built with Jupyter Book 2 and MyST. The canonical build entrypoint is docs/build.py, which also regenerates API pages and converts Jupytext-backed sources before building each language.

Quick Start

The easiest way to build and view the documentation locally:

# Build and open the English and Japanese documentation
task book

# Build and open the Japanese documentation
task book_ja

# Build and open the English documentation
task book_en

# Stop local preview servers when you are done
task stop_book
task stop_book_ja
task stop_book_en

Editing Documentation

The documentation source is organized into separate English and Japanese trees:

docs/
├── en/                       # English documentation
│   ├── myst.yml
│   └── ...
├── ja/                       # Japanese documentation
│   ├── myst.yml
│   └── ...
├── index.html                # Read the Docs root entrypoint
├── build.py                  # Local/RTD build entrypoint
├── generate_api.py           # API doc generator
└── templates/                # API doc templates

Adding New Pages

1. Create your content file (.md, .ipynb, or Jupytext-backed .py) in the appropriate language directory
2. Add the file to project.toc in the corresponding language's myst.yml
3. Rebuild the documentation using the commands above

Read the Docs Configuration

The repository root .readthedocs.yaml builds both languages in one Read the Docs project and publishes them under /en/ and /ja/.

Common Issues

• Build errors: Make sure you have all dependencies installed with uv sync --dev
• Generated pages not updating: Re-run task build_book_ja, task build_book_en, task book, or uv run python docs/build.py --lang ...
• Preview server still running: Stop the local HTTP preview with task stop_book, task stop_book_ja, or task stop_book_en before reusing the same port
• API documentation: Generated automatically from source code by docs/generate_api.py using Griffe and Jinja2

How to Contribute

See CONTRIBUTING.md

---

Copyright (c) 2023 Jij Inc.