Siyuan Xue

LeDA Agent

2025–2026

My role: Project team member; co-developed with my roommates

In July and August 2025, my roommates and I joined BUPT’s Undergraduate Innovation and Entrepreneurship Training Program and began building LeDA. We organized developer resources from the Awesome repository into a Neo4j knowledge graph and built a graph-based RAG agent with LangGraph. It retrieves resources in response to a user’s questions and returns relevant links. By persisting each user’s question history and intent, it also provides personalized learning-resource recommendations over time. LeDA went on to complete the program as a national-level project. Building something meaningful with friends was a thrill.

Your browser cannot play this video. Use the download link below.
A demonstration of resource search, persistent question history and intent, and personalized learning-resource recommendations based on that information.Download video
Your browser cannot play this video. Use the download link below.
Browsing the Neo4j knowledge graph. The full database contains more than 80,000 developer-resource nodes, which the agent uses for retrieval and queries.Download video
BUPT completion certificate for the national-level LeDA student innovation project, listing Siyuan Xue as a team member
Completion certificate for the national-level LeDA project, an educational multimodal recommendation system based on cognitive modeling and guidance. It passed its completion review in July 2026.

Related material

  • BUPT project page
  • Awesome GitHub repository