We are pleased to announce the release of our open source solver JuLS (Julia Local Search) on Amazon-Science GitHub. JuLS is a lightweight, open source and modular optimization solver that combines Constraint-Based Local Search (CBLS) and Constraint Programming (CP) to handle both generic and black-box constraints and objectives. Currently deployed in Amazon’s production environment, it successfully addresses complex logistics and scheduling challenges in middle-mile operations. Developed by scientists for scientists, this generic solver framework welcomes contributions and supports the modeling of a wide range of optimization problems. Its adaptable architecture and open design philosophy make it an ideal platform for both academic research and industrial applications. Backtests between JuLS and OR-Tools demonstrate comparable performance on standard optimization tasks. However, JuLS distinguishes itself through its seamless integration capabilities with external tools, particularly excelling in handling black-box constraints and objectives. This unique feature positions JuLS as a versatile solution for complex, real-world optimization scenarios that often require interfacing with proprietary or external systems.

Authors : - Axel Navarro : axelnav@amazon.fr - Arthur Dupuis : dupuisar@amazon.fr - Ilan Coulon : ilan.coulon@gmail.com - Ezra Reich : rezra@amazon.co.uk - Maxime Mulamba : mulmaxim@amazon.fr - Mehdi Oudaoud : mehdoud@amazon.fr

The software is released under the Apache 2.0 License.

Link : https://github.com/amazon-science/JuLS

We are glad for any comments and error reports (or even bug fixes) that you send us.

Looks interesting.

Would it be possible to use it as a MiniZinc backend?

Yes, that’s possible. JuLS’ CP layer can map FlatZinc constraints directly, so it could be used as a MiniZinc backend. We haven’t built that bridge yet, but it’s a natural extension and something we’d welcome contributions on.

Would be interesting to see in next years MiniZinc challenge as a way to compare with other solvers.