Author(s) :

Umesh Joge, Utkarsh Sahare, Vipul Navghare, Vrushbh Agalawe, Yash Umak.

Volume/Issue :

Abstract :

Understanding unfamiliar source code is a significant challenge for developers, particularly in the absence of proper documentation and reliable resources. While Large Language Models (LLMs) have shown strong capabilities in code generation, their application in code comprehension remains limited and often dependent on internet connectivity. This paper presents an offline LLM-based assistant integrated within an Integrated Development Environment (IDE) to provide context-aware explanations, API details, and example usage directly from highlighted code. The proposed system eliminates the need for manual prompt engineering and reduces context switching by offering real-time assistance inside the development workflow. The system was evaluated through a user study comparing traditional web search and online AI tools. Experimental results demonstrate that the proposed approach reduces task completion time by up to 50% and improves code understanding accuracy to 88%. Additionally, user satisfaction was significantly higher due to offline availability and improved workflow continuity. The findings indicate that integrating offline LLMs within IDEs can enhance developer productivity, ensure data privacy, and provide efficient code comprehension support in low- connectivity environments.

No. of Downloads :

0