The astonishing dimensions and complexity of power systems render them impossible to be managed without the help of cutting-edge software. Due to a lack of scalable, reliable and well documented free and open-source solutions, system operators, regulators, and government agencies often rely on proprietary software to provide them information that ultimately will be used to serve energy to millions of people and shape the economy. However, proprietary software is frequently closed source, meaning that its users cannot generally see its source code. Consequently, the crucial step of computational-aided analysis can be secretive to the general public, allowing mistrust and potentially preventing valuable information from being incorporated into the decision-making process. In particular, a problem that has been dominated by proprietary software is the unit commitment (UC). Worldwide, system operators use proprietary software to daily solve their UCs and define the best scheduling of the generating units under their supervision to satisfy the forecast load. So far, attempts to solve UC instances with open-source solutions have been mainly restricted to small cases. Here, we propose a flexible decomposition framework based on dual dynamic integer programming that allows us to solve realistic instances of the UC in reasonable time. We are particularly interested in hydrothermal systems, and, for one such system with 7,475 buses, we show through exhaustive numerical experiments that our approach is a promising alternative to proprietary software. For this system, we can solve the UC with a completely free and open-source package in less than one hour, both with an asynchronous and a synchronous strategy.