A Visual Studio Code Plugin for Tracing Between Documentation and Source Code

Maintaining end-to-end traceability between natural-language requirements and source code is essential yet remains difficult. In practice links are created late, decay quickly, and live outside the IDE, so they are rarely updated. Existing automation yields either noisy high‑recall candidate lists (lexical / embedding retrieval) or costly Large Language Model (LLM) reasoning. Developers lack an in‑IDE, low‑cost aid that raises precision without heavy labeling. This thesis presents a Visual Studio Code extension that brings documentation‑to‑code traceability into the developer workflow and a retrieval‑augmented approach, TRAG, that combines vector search with LLM verification. We evaluate the TRAG against the ArDoCode baseline with gold‑standard links, reporting precision, recall, and F1. ArDoCode provides consistently high recall and the best F1 on JR, TM, and BBB, whereas TRAG improves precision and can exceed F1 on MS and slightly on TS (e.g., best MS 0.208 with mistral‑nemo‑cot; TS 0.316 with gemma3‑4b). Chain‑of‑thought shows mixed effects, helping when evidence is compact but reducing recall otherwise. We discuss design choices, threats to validity, and practical operating points. ... mehrOverall, retrieval‑augmented verification is a viable complement to recall‑oriented baselines: it raises precision when evidence is well‑separated.