Skip to main content

Lab 13: GraphRAG Mini

Difficulty: Advanced · Estimated time: ~4–6 hours

Objective​

Build a minimal GraphRAG pipeline:

  • extract entities + relations from a small corpus
  • store a graph (NetworkX / Neo4j / similar)
  • use graph expansion to guide retrieval

Requirements​

  • 1-hop and 2-hop expansion supported
  • a comparison to hybrid search on multi-hop questions

Deliverables​

  • extract_graph.py
  • graph_store.py
  • query.py
  • REPORT.md with:
    • example multi-hop questions
    • qualitative + quantitative comparison