May 2025 Summaries

Graph algorithms are powerful analytics tools that help explore and understand complex networks, revealing patterns and structures within data. They provide structured ways to navigate and analyze relationships between entities in various domains such as social networks, maps, organizational charts, or system components. Graph algorithms can uncover insights that traditional methods often miss, leading to improved decisions and better outcomes. By analyzing the connections within data, graph algorithms reveal hidden patterns and structures that can be used to answer questions like "How are these two entities connected?", "What's the shortest/cheapest/fastest route between A and B?", or "Which nodes are the most influential or critical?". These algorithms come in various types, including pathfinding, centrality, community detection, similarity, link prediction, and node embedding techniques. Graph algorithms are used across industries to tackle real-world problems such as detecting sophisticated fraud, powering smarter recommendation engines, optimizing supply chains and logistics, simplifying identity and access management, and enhancing machine learning with graph features. Neo4j Aura Graph Analytics provides an extensive library of 65+ pre-built graph algorithms that can be easily accessed and applied to data, making it a powerful tool for uncovering connected insights and improving decision-making.

Graph algorithms are powerful analytics tools that help explore and understand complex networks, revealing patterns and structures within them. They provide structured ways to navigate and analyze data, identifying connections, hubs, natural groups, and patterns in datasets. Graph algorithms are rooted in graph theory and are used across industries to tackle real-world problems, such as detecting sophisticated fraud, powering smarter recommendation engines, optimizing supply chains and logistics, simplifying identity and access management, and enhancing machine learning with graph features. The Neo4j Aura Graph Analytics provides a library of 65+ pre-built graph algorithms that can be easily accessed and applied to data, offering efficiency, scalability, and pay-as-you-go pricing.

Neo4j has been accredited by Singapore's Infocomm Media Development Authority (IMDA), a move that strengthens the company's ability to support innovation in graph technology, which is emerging as a powerful enabler for extracting meaningful insights from complex and connected data. The accreditation follows a rigorous vetting process and reflects Neo4j's transformative potential, scalability, and financial health. This recognition streamlines the Singaporean government's procurement process for software solutions and creates diverse opportunities for accredited organizations to work on large-scale public projects. With this achievement, Neo4j aims to deepen partnerships in Singapore, train future data professionals, and contribute meaningfully to the region's fast-growing digital ecosystem.

LightRAG, an advanced knowledge retrieval system, combines two parallel retrieval paths to provide comprehensive and accurate answers. The hybrid approach blends graph traversal with semantic vector similarity, allowing for explainable and contextually grounded responses. It uses dual-level keyword extraction to understand both the big picture and important details in a user's query, prioritizing well-connected and central entities first. LightRAG retrieves structured knowledge from a graph and surrounding documents using dual-level keyword semantic search and graph traversal over entities and relationships. In parallel, it runs a semantic vector search with conversation history augmentation, retrieving semantically similar chunks from a pre-indexed vector store. The system combines the strengths of both retrieval methods, leveraging graph structure to prioritize information and providing more comprehensive responses to user queries.

This edition of This Week in Neo4j explores powerful ways to use Neo4j across AI, analytics, and cybersecurity. Developers can build smarter chatbots using Vertex AI and Gemini, create knowledge graphs from unstructured data with GraphRAG, scale graph algorithms with Aura Analytics, and simplify Active Directory security with BloodHound-MCP. Additionally, the community is invited to share their code, models, and graph-powered insights at NODES 2025, a global conference spotlighting real-world apps, intelligent systems, and all things Neo4j.

Neo4j and AWS are helping organizations realize the benefits of generative AI (GenAI) by leveraging knowledge graphs, which provide a structured representation of an organization's domain knowledge. This grounding of GenAI enables the mitigation of risks such as lack of explainability, inaccurate answers, and hallucinations, allowing for trusted decision-making. Native vector search, native property graph databases like Neo4j, and GraphRAG are key technologies that improve GenAI quality and effectiveness by reducing hallucinations, improving insights from data, achieving greater accuracy, and enabling efficient retrieval of relevant information from large datasets. The integration of Neo4j with AWS provides a powerful combination for building GenAI applications that integrate contextually rich external data in real-time while being secure and compliant.

Graph analytics helps organizations understand complex relationships within large datasets, enabling them to make informed decisions and uncover hidden patterns. It's particularly useful for fraud detection, recommendation engines, supply chain optimization, cybersecurity threat detection, content networks, biomedical research, and IT infrastructure management. Neo4j Aura Graph Analytics provides a scalable platform that enables teams to build and expand graph solutions without specialized expertise, empowering them to gain actionable insights from their data.

The LightRAG system is a hybrid retrieval model that combines the strengths of both graph-based reasoning and semantic search. It uses a dual-level approach, where one path focuses on relationships (high-level keywords) and the other on entities (low-level keywords). Both paths leverage vector similarity to surface semantically relevant content and are enriched with graph traversal and metadata. The system retrieves structured knowledge from a graph and surrounding documents using keyword-based retrieval, while also running a semantic vector search in parallel. The final output is combined and merged into a single CSV-formatted string block for the LLM to generate an accurate and contextually grounded response. LightRAG offers practical benefits such as flexible design, easy addition of new information without rebuilding everything, making it particularly valuable for businesses with constantly evolving knowledge.

LightRAG is a new retrieval-augmented generation (RAG) technique that leverages knowledge graphs to enhance retrieval quality. It offers better answers for different question types, smarter focus on what matters, and easy updates with new information. The extraction process involves three stages: cleaning up the mess, breaking content into digestible chunks, and letting LLM identify the important entities and how they connect. LightRAG stores this information both as a knowledge graph and in a vector database, understanding meaning and explaining its reasoning. It makes use of Neo4j as a graph store and uses Cypher queries to insert or update entities and relationships. The technique provides a hybrid retrieval model similar to GraphRAG, unlocking powerful capabilities for semantic search and graph-powered reasoning.

The Mahabharata chatbot has been upgraded with multi-LLM support, faster hosting via Cloud Run, and Hindi audio support for its responses. The update allows users to choose from different LLMs (GPT-4o, Gemini 2.5 Flash/Pro, Claude 3.7 Sonnet) and test how each model responds to the same query. This feature provides a more inclusive experience by making the chatbot's responses available in Hindi. The chatbot is now hosted on Cloud Run, which offers faster cold starts, auto-scaling, and easier API key management. Additionally, the chatbot has been enhanced with a new TTS (Text-to-Speech) system that converts English text to Hindi and plays it out loud. Future plans include expanding support for Tamil, Telugu, Bengali, and Kannada languages, as well as creating an epic digital sage with multilingual audio.

The Neo4j AuraDB with graph analytics workload is now available to all Microsoft Fabric customers, providing a seamless way to transform tabular data into a graph model and apply graph analytics. The workload uses GenAI assistance for creating the graph model and includes 65+ built-in algorithms. It offers exploration and interactivity using the Explore graph visualization interface, allowing users to visualize and interact with their graph data. The workload provides a free 14-day trial of AuraDB Professional database and enables writeback of graph insights into OneLake. Reviews from private preview customers have been positive, citing ease of use and the value of the generative AI assistance for creating the graph model.

To integrate Neo4j into a Java application, you can run a Cypher query and parse the results using the Neo4j Java driver. The process involves importing the driver, connecting to a server, verifying the connection, executing a Cypher query, parsing the results, and closing the connection. You can import the GraphDatabase and AuthTokens classes from the Neo4j driver package and use them to create an instance of the GraphDatabase.driver class, passing your credentials. The executableQuery method executes a Cypher query and returns the results, which can be parsed by iterating through the records and using get to retrieve return values. Once you finish with the driver, it's essential to call close to release any resources. Neo4j GraphAcademy offers various courses, including "Using Neo4j with Java," which covers topics such as passing parameters to queries, dealing with graph data types, and managing transactions. The course is available for free and provides a comprehensive introduction to developing Java applications with Neo4j.

This article demonstrates how to integrate LangChain4j with Neo4j, a high-performance graph database, to perform graph-based question answering. LangChain4j is a Java library that simplifies the integration of AI/LLM capabilities into Java applications. By integrating LangChain4j with Neo4j, developers can leverage advanced language models and graph data to answer complex queries based on the relationships and properties stored within the database. The article covers various aspects of the integration, including the creation of an Embedding Store instance, storing embeddings in a Neo4j database, executing searches, and implementing hybrid search capabilities. Additionally, it provides examples of how to use the LangChain4j framework with Neo4j for content retrieval and graph-based question answering. The article also discusses future updates and improvements to the LangChain4j library, including new features such as LLMGraphTransformer and GraphRAG concepts.

This week in Neo4j features the announcement of NODES 2025, a global graph developer conference taking place on November 6. The Call for Papers is open until June 15, allowing developers to share their code and models with the community. A Star Wars-themed Knowledge Graph Generator was also showcased, along with an introduction to Neo4j Aura Graph Analytics, a new serverless offering that allows users to run graph algorithms directly from any data source. Additionally, tutorials on using Cursor AI coding assistant with Neo4j MCP (Model Context Protocol) and various webinars and courses are available for learning graph technology.

Neo4j Aura Graph Analytics is a new serverless offering that delivers deeper insights from any data source and on any platform. It eliminates the need for custom queries, ETL pipelines, and specialized graph expertise, making graph analytics accessible to everyone. This new offering prioritizes speed, delivering insights 2x faster than open-source alternatives with parallelized in-memory processing of graph algorithms. It also offers up to 80% increase in model accuracy by transforming graph structures into ML-ready features with graph embeddings and 75% less code and zero ETL with prebuilt, optimized graph algorithms. Additionally, it provides no administrative overhead and lower TCO with pay-as-you-go pricing and independent scaling of compute and storage. Enterprises can use it to unlock the full value of organizational data, improve decision-making, and increase business value by uncovering critical patterns and insights in complex data.

Aura Graph Analytics is a powerful tool for Neo4j users, allowing them to run graph algorithms such as PageRank, node similarity, and community detection with ease. By projecting the relevant subgraph into memory, Aura Graph Analytics provides a fast and efficient way to analyze complex graphs without affecting the operational database. The tool requires minimal setup and can be used to solve a variety of use cases, including fraud detection, influence/leadership ranking, customer 360 identity resolution, supply chain risk management, product recommendations, IT asset risk mapping, internal talent mobility, recommendation embeddings, behavioral segmentation, and entity deduplication at scale. With Aura Graph Analytics, users can discover clusters, influences, risks, opportunities, and patterns hiding in their data that Cypher alone cannot easily surface, providing a new path for graph science and analysis.

GraphRAG is an agentic architecture designed to navigate complex domains like legal contracts by leveraging large language models (LLMs) and structured tools. By structuring legal information as a knowledge graph, users can increase answer accuracy using a LangGraph agent. The system involves constructing a knowledge graph in Neo4j, building a LangGraph agent that allows users to ask specific questions about the contracts, and implementing a contract retrieval tool with various attributes for filtering and aggregation. The LLM acts as the decision-maker, dynamically selecting which tools to invoke and executing multiple tools in sequence to fulfill complex requests. The system has been benchmarked using a dataset of 22 questions, showing promising results, but there is room for growth, including expanding clause coverage and refining tool design.