GraphReader Agentic RAG: Rethinking Long-Context Retrieval Systems
summary
Reproduces the GraphReader paper with open-source tools, building a Neo4j knowledge graph from section-wise chunks, extracting atomic facts and key elements with LLM prompts, indexing fact, section and full‑text embeddings, and running a LangGraph/LangChain agent that fuses vector hits then hops neighbors. Demo ingests a finance PDF, surfaces nodes like Tesla and $25.7B, answers a Q4 revenue query with traversal limit 3. Notes compute cost and LLM nondeterminism.
AI-written summary · about
chapters (10)
- 0:00Introduction and setup
- 1:54Traditional RAG challenges
- 4:02Introducing GraphReader
- 8:28Demo setup and tools
- 10:18Document ingestion process
- 13:31Graph visualization and data structure
- 15:13Retrieval and agentic workflow
- 20:37Query execution and answer generation
- 23:11Limitations and future work
- 25:18Resources and closing
related talks
36mGraphRAG Explained 🔥 Building Smarter LLMs
53mPractical GraphRAG - Making LLMs smarter with Knowledge Graphs
39mEnhancing RAG with Neo4j Knowledge Graph
20mPractical GraphRAG: Making LLMs smarter with Knowledge Graphs
26mPractical GraphRAG Making LLMs smarter with Knowledge Graphs
32mGraphRAG: Powering Up LLMs with Knowledge Graphs
23mConnecting the Dots With Context Graphs
25mGraphRAG: Using LLMs to Build and Interrogate a Regulatory Knowledge Graph