From AI Workflows to Graph-Native Reasoning: A Dify Plugin for Querying Neo4j

Session Track: Knowledge Graphs & GraphRAG

Session Time: 01:30 pm - 02:00 pm April 15

Session description

AI workflows orchestrate large language models, tools, and data sources into repeatable pipelines that can retrieve information, transform it, and act on results with minimal human intervention. Dify is an open, agentic workflow platform that lets users visually compose these pipelines, connect to a wide range of models and tools, and observe how AI applications behave in production. Within this ecosystem, many real-world use cases demand not only unstructured text retrieval but also robust access to structured, interconnected domain knowledge. Knowledge graphs, and Neo4j in particular, provide a graph-native representation of entities and relationships that is essential for transparent, explainable reasoning in complex domains.

This work introduces a Neo4j query plugin for Dify that bridges agentic workflows with graph databases, allowing AI agents to issue Cypher queries to Neo4j directly from within Dify workflows and integrate structured graph results into multi-step reasoning. The plugin enables AI workflows to ground their outputs in curated graph data, support multi-hop questions over rich relationship structures, and combine RAG-style document retrieval with graph-native context.

The talk will cover how the plugin works, how to configure it from the Dify Marketplace, and how to embed it into Dify workflows. The plugin will be showcased with several life science examples, illustrating how graph-grounded AI can support complex queries over biological entities, diseases, and therapeutic relationships.