Combine GDS and Agentic AI for Automatic Report Generation
summary
Combines Neo4j GDS and LangGraph to turn Chicago Police datasets into automated investigative briefs: builds a co‑offending network, runs Louvain, PageRank and Betweenness to find communities and key players, then pipelines LangGraph agents for demographic, temporal and geographic analysis (Nominatim reverse‑geocoding and custody lookups) into HTML reports. Initial single‑prompt generation fails, parallel agents reduce latency but repeat content, chained prompts with reflection cut errors and boost quality ~10× while adding latency and infra complexity.
AI-written summary · about
chapters (16)
- 0:00Introduction to talk
- 2:08Problem with data overload
- 4:19Data sources and use case
- 6:47Building co-offending networks
- 9:19Community detection in networks
- 11:06Demo of network analysis
- 13:10Phase two overview
- 14:58First LLM report attempt
- 16:45Improving report generation
- 18:55Reflection and quality boost
- 20:42Prompt engineering techniques
- 23:11Evaluating prompt changes
- 24:18LangGraph workflow benefits
- 25:39Final report example
- 26:23Key takeaways summary
- 28:07Closing and Q&A
related talks
08mHarness Graph Power With Agentic AI: An Insights Discovery Use Case
01h 17mGraph Database and Retrieval Augmented Generation
01h 59mBuilding Agentic and Multi-Agent Systems with LangGraph
01h 40mKnowledge Graphs & GraphRAG: Techniques for Building Effective GenAI Applications
36mBuilding Agentic AI Assistants: Lessons from productionising agents in Preclinical R&D
30mNeurosymbolic Architecture in Production with Neo4j, GDS, Node2Vec & LLM Grounding
36mGraphRAG Explained 🔥 Building Smarter LLMs
30mPerformant GraphRAG Improving Neo4j to Drive Autonomous AI ...