Skip to content
AI|Course

Grounding models with data · Intermediate

Implement GraphRAG with Knowledge Graphs

Learn why plain vector RAG breaks on multi-hop and corpus-wide questions, and how knowledge graphs fix it. Build a graph from unstructured text, retrieve over it, and evaluate the result.

7 lessons · 14 hr 15 min · Free

You will learn to

  • ▸Explain when GraphRAG beats vector RAG, and when it does not
  • ▸Model a domain as a property graph and query it with Cypher
  • ▸Build a knowledge graph from documents with an LLM extraction pipeline
  • ▸Pick the right retrieval pattern: vector, graph traversal, text-to-Cypher or community summaries
  • ▸Evaluate a GraphRAG system and reason about its cost

Before you start

  • ·Comfortable with Python
  • ·You have built or at least understand a basic RAG pipeline
  • ·An LLM API key (a free tier or a local model works)

Syllabus

  1. 01Why GraphRAG?Where vector RAG fails, and what a graph adds.45 min
  2. 02Knowledge graph foundationsProperty graphs, data modeling and your first Cypher queries.2 hr
  3. 03Building a graph from unstructured textChunking, schema-guided extraction and entity resolution.2 hr
  4. 04GraphRAG retrieval patternsVector plus traversal, hybrid search and text-to-Cypher.2 hr
  5. 05Communities: local vs global searchThe Microsoft GraphRAG approach to corpus-wide questions.1 hr 30 min
  6. 06Agents, evaluation and productionPut retrieval behind an agent, measure it, control the cost.2 hr
  7. 07Capstone projectBuild and ship a GraphRAG app on a corpus you care about.4 hr