An LLM Assistant for Software Project Onboarding
Knud Ronau Larsen, Magnus Edvall, Truong Ho-Quang, Felix Dobslaw, Rodi Jolak · 2025
New developers commonly join software teams for projects in the advanced stages of development. They frequently face barriers, especially in software comprehension, that impede their ability to make early contributions. To aid newcomers in understanding software projects, we have developed an LLM-based tool called SPAC-B that can answer software project-specific questions. In this study, a case study is conducted to investigate the accuracy of SPAC-B's answers to common developer questions when resolving issues on two open-source projects regarding relevance, completeness, and correctness. On a Likert scale (1–5), answers from SPAC-B scored a mean of 4.6 in relevance and 4.3 in completeness and correctness. An experiment with ten software developers is performed to further examine SPAC-B's ability to help newcomers formulate plans for resolving real-life open-source issues. Results show that the participants could make a better implementation plan with the use of SPAC-B and 8/10 participants found the tool to be helpful.