Performant GraphRAG Improving Neo4j to Drive Autonomous AI ...
summary
Neo4j is reshaped to power fast GraphRAG pipelines by building materialized views, search nodes and graph diffs that turn a 120 hour manual workflow into ~20 minutes of automated work; key wins are call statement optimization to stop query-plan branching, splitting MERGE into separate create/update passes to avoid all‑node scans, and converting UUIDs to 64‑bit longs plus label partitioning for 14x faster operations and much smaller indexes (example: long vs string IDs 1.1 GB vs 2.4 GB); result scales to hundreds of millions of nodes and feeds AI via the Model Context Protocol to auto-generate PRs and iterate safely.
AI-written summary · about
chapters (15)
- 0:00Introduction and background
- 1:11Code Logic's knowledge graph journey
- 3:57Moving from Postgres to Neo4j
- 5:48Search nodes and query performance
- 7:15Spring Data Repository issues
- 8:21Optimizing query plans with CALL
- 10:46Advanced CALL statement use cases
- 13:17Graph differentials and diff operations
- 14:20Performance issues with MERGE
- 16:27UUID to long conversion for performance
- 18:15Materialized views with labels and longs
- 20:27Why performance improvements matter
- 22:17Integrating AI with graph data
- 24:46Summary of performance optimizations
- 26:10Q&A session
related talks
16mEnhancing Retrieval-Augmented Generation with GraphRAG Patterns in Neo4j
24mGraph-Powered Job Matcher: From SQL Tables to Neo4j RAG in Production
39mEnhancing RAG with Neo4j Knowledge Graph
35mGraphRag in Action: Revolutionizing Fraud Detection Techniques
13mOptimizing Real-Time Job Matching: Integrating Neo4j With RAG
40mRAG Pipeline Evaluation
49mModern Graph Databases
27mAdaptive Knowledge Graphs for Dynamic RAG Systems