Reproduce “CODENS: Transforming Code Changes into Living, Accessible, and Queryable Documentation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Abdelhak Kelious, Chyrine Tahri, Eliot Bardet

CODENS is a system transforming pull requests into a dynamically updated, queryable software knowledge graph for production codebases, addressing documentation obsolescence by continuously integrating code changes. An evaluation on a Ruby on Rails project showed high answer relevance (0.94) and completeness (4.45/5) using an agent-guided retrieval mechanism.

Key insights

  • Continuously processing pull requests allows for building a 'living' software knowledge graph that remains current with codebase evolution, overcoming documentation obsolescence.
  • Leveraging a typed software knowledge graph combined with vector embeddings enables both relational traversal and semantic search, enhancing the accuracy and context of retrieved information.
  • An LLM-powered ReAct agent, equipped with specific graph interaction tools, can perform guided, multi-hop exploration to generate more complete and precise answers for complex documentation queries.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.