A practical implementation guide for integrating a knowledge graph alongside an existing vector RAG stack using the dual-index pattern. Covers schema-guided entity extraction with LangChain's LLMGraphTransformer and LlamaIndex's SchemaLLMPathExtractor, graph storage options (Neo4j, Neptune Analytics, Memgraph, FalkorDB), hybrid retrieval with entity linking and Reciprocal Rank Fusion, and complete LangGraph workflow code. Also addresses evaluation using Ragas multi-hop question generation, day-2 ops including incremental ingestion, graph drift monitoring, schema evolution, and access control. Includes a 30-60-90 rollout plan and eight common pitfalls such as entity explosion, LLM-hallucinated edges, Cypher timeouts, and graph query injection.
