From 86fd9121c4a76b9ab15795eee2655f87ce025790 Mon Sep 17 00:00:00 2001 From: davidmyriel Date: Wed, 5 Feb 2025 06:23:10 -0500 Subject: [PATCH] Update graphrag-qdrant-neo4j.md --- .../documentation/advanced-tutorials/graphrag-qdrant-neo4j.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/qdrant-landing/content/documentation/advanced-tutorials/graphrag-qdrant-neo4j.md b/qdrant-landing/content/documentation/advanced-tutorials/graphrag-qdrant-neo4j.md index 08ef463e7..25765e27e 100644 --- a/qdrant-landing/content/documentation/advanced-tutorials/graphrag-qdrant-neo4j.md +++ b/qdrant-landing/content/documentation/advanced-tutorials/graphrag-qdrant-neo4j.md @@ -14,6 +14,8 @@ social_preview_image: /documentation/examples/graphrag-qdrant-neo4j/social_previ | Time: 30 min | Level: Intermediate |Output: [GitHub](https://github.com/qdrant/examples/blob/master/graphrag_neo4j/graphrag.py)| | --- | ----------- | ----------- | +

+ To make Artificial Intelligence (AI) systems more intelligent and reliable, we face a paradox: Large Language Models (LLMs) possess remarkable reasoning capabilities, yet they struggle to connect information in ways humans find intuitive. While groundbreaking, Retrieval-Augmented Generation (RAG) approaches often fall short when tasked with complex information synthesis. When asked to connect disparate pieces of information or understand holistic concepts across large documents, these systems frequently miss crucial connections that would be obvious to human experts. To solve these problems, Microsoft introduced **GraphRAG,** which uses Knowledge Graphs (KGs) instead of vectors as a context for LLMs. GraphRAG depends mainly on LLMs for creating KGs and querying them. However, this reliance on LLMs can lead to many problems. We will address these challenges by combining vector databases with graph-based databases.