SUMM

The Knowledge Graph Mullet: Trimming GraphRAG Complexity - William Lyon

Introduction to Knowledge Graph Mullet 00:00

  • William Lyon introduces the concept of the knowledge graph mullet, likening it to a mullet hairstyle that is low-maintenance yet versatile.
  • The knowledge graph mullet combines property graphs in the front and RDF triples in the back, creating a hybrid approach for knowledge graphs.

Property Graph vs. RDF 01:14

  • Property graphs focus on nodes, relationships, and key-value properties, while RDF emphasizes ontologies and triples.
  • The talk aims to demonstrate how to leverage both paradigms for enhanced graph data modeling and querying.

Understanding Knowledge Graphs 03:00

  • A knowledge graph is defined as an instance of a property graph, consisting of nodes with labels, relationships, and properties.
  • The importance of having a canonical representation of entities in a knowledge graph is emphasized.

Dgraph Overview 04:32

  • Dgraph is an open-source project designed for large-scale graph data, inspired by Google's Spanner graph.
  • It uses a hybrid model, employing property graphs for data modeling and RDF for data interchange.

Querying with Dgraph 07:47

  • DQL, the query language for Dgraph, is influenced by GraphQL and starts with a well-defined starting point for traversals.
  • The structure of DQL queries resembles JSON and includes specifications for properties and traversals.

Building a News Knowledge Graph 09:09

  • An example is provided to build a knowledge graph of news articles, focusing on entities like organizations, topics, and authors.
  • Chunking unstructured data into nodes is discussed, where each paragraph of an article may represent a node.

Graph RAG Approaches 11:00

  • The concept of graph RAG (Retrieval-Augmented Generation) is introduced, emphasizing different entry points for data retrieval.
  • Vector similarity search is mentioned as a method to find relevant chunks of unstructured data.

Hands-On Dgraph Examples 12:55

  • A demonstration of executing DQL queries using Dgraph’s Rattell workbench is provided, including simple and complex queries.
  • Geographic distance searches and vector similarity searches are showcased as methods to traverse and explore the knowledge graph.

Dgraph MCP Server Features 15:48

  • The latest Dgraph release includes enterprise features, and the Model Context Protocol (MCP) server for model interaction is introduced.
  • MCP enables database interaction and use cases for exploratory data analysis and agentic coding environments are discussed.

Building AI Agents with Modus Framework 24:15

  • The Modus agent orchestration framework is introduced as an open-source tool for creating AI agents that interact with knowledge graphs.
  • It allows for the creation of domain-specific agents using prompts and exposes tools via the MCP server.

Conclusion and Resources 32:31

  • The talk concludes with a summary of how to utilize the knowledge graph mullet for effective graph RAG workflows and AI agent development.
  • Viewers are directed to additional resources and slides related to the presentation.
HomeSubmitSaved