Visual academic citation graph (ScholarGraph) built on Neo4j, FastAPI, React Flow, and LangChain
clearCites / ScholarGraph is an open-source visual discovery tool that maps academic research as an interactive knowledge graph. Papers are Neo4j nodes; citations, author overlaps, and funding relationships are edges. LangChain + OpenAI generates plain-English summaries and evaluates paper relationships (validates / builds_on / challenges). Users can navigate the full citation pedigree of any paper and identify publicly-funded research by funding-source tags.
Open-source academic knowledge graph for navigating the "lineage of ideas."
• Graph DB: Neo4j with Cypher DDL; nodes for Paper, Author, Keyword, Funder; edges for CITES, VALIDATES, BUILDS_ON, CHALLENGES, FUNDED_BY.
• Data Pipeline: Python ingestors for Semantic Scholar & CrossRef APIs; parser extracts keywords, authors, and funding sources; graph_pusher converts to Neo4j Nodes/Edges.
• Backend: FastAPI (async) with endpoints for citation chains, pedigree traversal, keyword search, author overlap, and AI relationship evaluation.
• AI: LangChain + OpenAI generates 3-sentence plain-English abstracts and returns structured relationship JSON: { relationship, correlation_value }.
• Frontend: React Flow interactive canvas (GraphCanvas) with drag, click, and zoom; PaperDetail sidebar; metric-based node scaling by impact_score.
• Infra: Docker Compose (Neo4j + FastAPI + React); fully containerised for one-command startup.