Agentic GraphRAG: Multi-Agent Knowledge Graph Construction for Research Teams
summary
A production-ready platform ingests PDFs and datasets and runs a four-agent pipeline — analyzer, extractor, merger and optimizer — built with Neo4j, LangChain, Cloudflare Workers and OpenAI chatkit; it chunks documents for parallel extraction, semantically merges 84 extracted entities down to 32, and writes 57 nodes and 179 relationships to Neo4j in ~2s using batch UNWIND+MERGE. Includes traceable eval with LangSmith and a chat interface. agentic pipeline semantic deduplication
AI-written summary · about
chapters (12)
- 0:00Introduction and motivation
- 2:59Current research challenges
- 4:25Multi-agent architecture overview
- 6:31Web app and user input
- 8:18Analyzer agent and schema definition
- 12:11Extractor agent and parallel processing
- 16:01Entity merger agent and deduplication
- 18:52Optimizer agent and graph quality
- 22:46Neo4j persistence and performance
- 24:55Chatbot integration and research assistant
- 26:00Future improvements and roadmap
- 28:08Q&A and closing remarks
related talks
12mAgentic GraphRAG: Autonomous Knowledge Graph Construction and Adaptive Retrieval
15mAgentic GraphRAG: Simplifying Retrieval Across Structured & Unstructured Data
21mAgentic GraphRAG: Autonomous Knowledge Graph Construction and Adaptive Retrieval
20mEnhancing RAG with Multi-Agent Integration
39mAgentic RAG Platform for AI-Assisted Workflows
01h 16mForget RAG Pipelines—Build Production Ready Agents in 15 Mins
36mGraphRAG Explained 🔥 Building Smarter LLMs
01h 26mThinking in Graphs: The Cognitive Foundations of Agentic AI