CARS: A Context-Aware Recommender System for Sustainable Software Package Selection

CARS: the Context-Aware Recommender System. In the software development world, there is a constant push towards innovation. New technologies are continuously replacing older ones. New software packages are released, while old ones are deprecated. The software world is a space with many problems and even more solutions. There is now a sea of packages to sift through in order to find the one that is the right fit for any specific project. The CARS project aims to alleviate some of this manual labour by sifting through all of these packages and their context for the user; our pipelines gather data from package indices such as PyPI and CRAN to find which packages exist, we scour GitHub repositories to find how these packages are used and we comb through stack overflow posts to understand what the community thinks of these packages. Through the use of big data, Large Language Models and other modern technologies, we can understand the context and use cases of software packages, interpret the user’s query and recommend them software packages that fit their need. The aim is that the user can specify their requirements in natural language in order to get relevant recommendations.