Comprendia: AI-Augmented Code Comprehension
Comprendia is an Eclipse plugin that aims to improve the way developers understand Java code. It combines visualizing structural dependencies with LLM-generated code explanations on an interactive graph. The tool features four main elements: (1) a dynamic multi-edge dependency graph that includes live search and multiple layouts; (2) LLM explanations utilizing Graph-Aware Callee Pruning (GACP), which highlights relevant callees from the same graph; (3) an overlay for detecting code clones and suggesting refactor options; and (4) a CVE risk overlay from OSV.dev. GACP uses factors like graph distance and edge-type weighting to create consistent prompts traceable to graph nodes. We demonstrate Comprendia with a Java project showcasing known code clones and vulnerabilities.
Key facts
- Comprendia is an Eclipse plugin for Java program comprehension.
- It integrates structural dependency visualization with LLM-powered code explanation.
- The tool has four pillars: dependency graph, LLM explanations, clone detection, and CVE risk overlay.
- GACP (Graph-Aware Callee Pruning) selects relevant callees using the graph.
- GACP uses graph distance, inheritance collapse, and edge-type weighting.
- The tool is demonstrated on a Java project with known clones and vulnerabilities.
- The paper is available on arXiv with ID 2608.10290.
- The announcement type is cross.
Entities
Institutions
- Eclipse
- OSV.dev
- arXiv