April 2026 Summaries

Neo4j's latest release introduces asynchronous I/O capabilities using io_uring on Linux, marking a significant evolution in its I/O handling approach. This update allows Neo4j to perform multiple I/O operations concurrently without blocking execution, which is particularly beneficial in cloud environments with high I/O latencies. Initially, asynchronous I/O support is implemented in the background page evictor and checkpointer, offering potential performance improvements in these components. While not enabled by default, asynchronous I/O can be activated by meeting specific prerequisites such as using Neo4j Enterprise Edition on a Linux system with the liburing library and JDK 25 or later. The new feature aims to enhance database performance by improving concurrency, throughput, and system resource utilization. As asynchronous I/O capabilities expand, Neo4j is poised for greater performance and scalability in future iterations.

This week's update from the world of Neo4j highlights the recent NODES AI event, where recordings are now available, including an extended keynote exploring the burgeoning role of context graphs in AI systems. The event also marked the launch of an Aura Agent Hackathon, encouraging innovative uses of Aura Agents. A detailed discussion unfolds on the replacement of APOC with native Cypher for improved memory safety and error handling in batch processing, while another post uses the metaphor of a sheep in a wolf-filled forest to explain how context graphs enable AI agents to reason beyond mere recall. The newsletter encourages participation in Neo4j’s User Research panel and previews upcoming conferences and workshops, emphasizing the transformative potential of graph technology and GenAI.

The blog post discusses the development of stateful AI through the integration of Neo4j Aura Agents with a memory layer to create a self-improving, closed-loop system. It highlights the concept of agentic memory, where agents not only execute tasks but also learn and evolve by storing experiences in a persistent memory, enabling them to optimize subagents over time. The approach involves using a memory model inspired by Andrej Karpathy's LLM Knowledge Bases, where learnings are stored in a structured format within a Neo4j database, allowing agents to build, test, and refine their strategies continuously. The article emphasizes the importance of a robust memory system, which allows the agent to retain insights and improve its performance across sessions, ultimately transforming it from a task executor to a problem-solving entity. The implementation involves an MCP server to manage the lifecycle of agents and a memory module to facilitate the retention and retrieval of knowledge, ensuring the agent can adapt and improve based on previous experiences.

In the Neo4j Developer Blog, Benjamin Squire explores modeling the Grand Sumo wrestling tournaments using graph databases like Neo4j, highlighting how such models can offer insights into the sport's complex structure. The Grand Sumo tournaments, held six times a year across Japan, feature divisions such as Juryo and Makuuchi, with a unique ranking system and no weight classes, leading to diverse match outcomes influenced by weight, speed, and agility. The blog delves into modeling this structure as a graph, utilizing Neo4j's Bloom and Graph Data Science Algorithms to visualize and analyze data, revealing that higher rank does not always equate to being the most central figure among wrestlers. The article emphasizes that innovative graph algorithms, like Eigenvector Centrality, can identify the most influential wrestlers, as seen with the top rikishi Hoshoryu, Onosato, and Aonishiki, and discusses the potential insights gained from such analyses. The post concludes by inviting readers to engage with the content and suggests future explorations of scaling this model using Neo4j Graph Data Science Native Snowflake App.

Neo4j and Google Cloud have expanded their collaboration to enhance the adoption of graph-powered AI applications, introducing new integrations and tools that support enterprises transitioning from AI experimentation to production. This includes the availability of Neo4j AI Agent within the Gemini Enterprise console, allowing seamless integration without a standalone marketplace, and the introduction of a Neo4j Gemini CLI extension to facilitate command-line deployment and query generation. Enhancements such as a complete agent memory API and Pub/Sub integration with Google Cloud bolster real-time data workflows and analytics. The Neo4j Aura agents, leveraging Gemini 3 as the reasoning engine, can now incorporate graph-grounded retrieval, while the Neo4j Terraform Provider aids in designing data flows. Furthermore, the Neo4j Community and Enterprise Editions are now more accessible through Google Cloud Marketplace, offering ease of deployment and tighter control over infrastructure, with a particular focus on sectors like healthcare that rely on connected data. These innovations reflect a commitment to making graph technology more practical and accessible for enterprises leveraging Google Cloud.

In a blog post by Martin O'Hanlon, a technical curriculum developer at Neo4j, the potential to enhance coding agents like those in VS Code with Neo4j’s GraphAcademy content is explored, aiming to improve the educational capabilities of these agents. The initiative involves creating a Neo4j graph database on Aura containing GraphAcademy's open-source lessons, enabling an Aura agent to search, retrieve, and format this content. This agent is then exposed as an MCP server, allowing it to be integrated into developer environments, providing structured lessons, examples, and progression paths. The agent uses OpenAI’s text-embedding model for semantic searches, ensuring relevant lesson retrieval with a focus on challenging misconceptions and offering learning objectives. This setup not only facilitates better teaching interactions but also supports developers in learning through practical examples and structured guidance, enhancing their skills with Neo4j technologies.

The comprehensive guide by Giuseppe Villani explores the integration of Neo4j with Google's Genkit, highlighting the benefits of combining graph-based vector storage with semantic search and knowledge graph applications. As AI systems grow in complexity, traditional databases face challenges with multi-hop reasoning and structured relationships, which Graph Retrieval-Augmented Generation (GraphRAG) addresses by merging semantic understanding with knowledge graphs. Google Genkit, an open-source framework, facilitates AI application development through plugins, including a Neo4j plugin that enables native vector search and document storage as nodes, preserving their relational context. The guide details the integration process, from basic semantic search to advanced GraphRAG topologies and persistent chat memory, and provides insights into installation, configuration, indexing, retrieval, and the use of advanced strategies like metadata filtering and hybrid search. It also covers the potential of using custom retrieval queries and graph traversals to enhance data retrieval accuracy, emphasizing the role of GraphRAG capabilities in expanding context for AI systems. The integration offers a robust foundation for developing sophisticated AI applications by leveraging the structural context of knowledge graphs.

Paolo Delano, a Solutions Engineer at Neo4j, discusses the transformative impact of knowledge graphs on macroeconomic risk management and supply chain logistics in a world marked by unpredictability. The narrative follows Roberto, a beverage distributor in San Salvador, who uses a knowledge graph to navigate disruptions caused by protests at a crucial border crossing. Instead of relying on guesswork, Roberto employs an interconnected system that visualizes relationships between commodities, suppliers, and border crossings, transforming data into a predictive tool for managing logistical challenges. The article emphasizes the potential of agentic AI, with specialized agents monitoring various nodes like border crossings, commodities, and commercial hubs, to provide strategic insights and risk assessments. This approach allows for the identification of hidden dependencies and the simulation of economic scenarios, enabling policymakers and businesses to anticipate and mitigate risks proactively. By shifting from reactive to predictive strategies, the use of knowledge graphs offers a precise framework for economic policy and supply chain resilience, highlighting the importance of understanding interconnected economic entities to maintain stability.

Enterprises are increasingly focused on deploying AI agents capable of reasoning through complex data, with Neo4j's graph technology playing a pivotal role in overcoming challenges associated with disconnected data points. As demonstrated in Q1 2026, collaborations with major platforms like Google Cloud, AWS, Azure, Databricks, and Snowflake are enabling companies to integrate graph-powered intelligence directly into their existing data environments. This integration facilitates the transition from experimental AI to production-ready systems by providing structured, auditable, and context-rich data frameworks. Key successes include startups rapidly achieving ROI with Neo4j on Google Cloud, the U.S. Army's logistics modernization using AWS, and enhanced AI agent memory on Azure. Furthermore, innovative solutions like the Neo4j Graph Agent for Snowflake and the Neo4j Connector for Databricks streamline data processing and enable real-time insights. These advancements highlight the growing importance of connected intelligence, as AI agents become more capable of understanding relationships and solving complex problems. Neo4j's ongoing digital series invites users to explore these integrations and the transformative potential of graph databases in AI development.

The blog post explores the creation and challenges of using GraphRAG agents to simulate conversations with Stoic philosophers like Seneca, focusing on retrieving authentic philosophical responses from a knowledge graph. The author details the issue of these agents fabricating quotes instead of querying their database, a problem resolved by employing a situational prompt that casts the AI as a version of Seneca with unreliable memory, compelling it to verify quotes through tool-based searches. This approach reduced fabrication rates to zero on initial queries by framing the AI's limitations as intrinsic rather than instructive, enhancing the interaction's authenticity. The post also touches on the broader implications for AI systems, emphasizing the need for accurate self-perception in models with substantial domain knowledge to prevent hallucinations, and concludes with a humorous nod to the anthropomorphized nature of AI agents.

As enterprises integrate large language models (LLMs) with internal data sources using the Model Context Protocol (MCP), a crucial challenge arises in maintaining data security and access control, which is often overlooked. The article discusses the potential risks of data leakage, compliance violations, and trust erosion when identity-aware access control is not implemented, as MCP can allow unrestricted access to sensitive data. The solution proposed involves shifting from an identity vacuum to identity-driven access control by passing user-specific identity tokens through the MCP server to the data layer, using tools like Neo4j MCP Server and Keycloak for authentication. This approach ensures that, depending on their roles, users receive tailored access to data, thereby maintaining strict data governance and sovereignty. This method of secure access control is crucial for enterprises aiming to leverage AI while upholding data security and compliance standards.

Integrating Neo4j MCP with Laravel using the neo4j/laravel-boost package enhances the functionality of AI agents by providing them direct, secure access to graph databases, which improves their effectiveness and reduces hallucinations. By acting as a proxy, this package merges Neo4j tool manifests with Laravel Boost's registry, simplifying developer experience and maintaining a single, cohesive MCP endpoint. The integration supports both HTTP and STDIO transport modes, allowing flexible configuration through Laravel's .env file. This setup enables seamless execution of Cypher queries and schema retrieval, facilitating natural language-driven CRUD operations in development environments. Through the combination of Laravel and neo4j/laravel-boost, developers can leverage a robust, single-server architecture for AI applications, enhancing their development workflow by allowing agents to actively participate and generate relevant PHP code based on database interactions.

The blog post by Bryan Lee, a Solutions Engineer at Neo4j, discusses the security challenges of connecting large language models (LLMs) to enterprise databases and the implementation of the Model Context Protocol (MCP) as a solution. MCP is likened to a "USB-C for AI" because it allows easy integration of AI models with various tools and services, but it poses significant security risks due to the lack of identity-aware access control. Without proper controls, sensitive data can be exposed to unauthorized users, leading to data leaks, compliance violations, and a loss of trust. The article proposes using Identity Driven Access Control to pass users' identities through the MCP layer to enforce role-based access controls at the database level, specifically in Neo4j. This approach uses OIDC authentication to ensure that each request carries the user's identity, allowing the database to enforce fine-grained access controls based on roles and maintaining data governance while connecting LLMs to enterprise data. The blog highlights the importance of this method for maintaining security and integrity in AI-driven workflows and offers practical guidance for implementing such a system.

Neo4j has updated its listings on Google Cloud Marketplace to enhance user experience and ease of use, covering both its SaaS and self-managed offerings. The SaaS version, Neo4j Aura, is available in several tiers, including the Professional, Business Critical, and the Virtual Dedicated Cloud, with the latter being available through private offers. The Aura listings are pay-as-you-go, utilizing Google's real-time billing mechanism, and are designed for easy deployment. The self-managed versions, Neo4j Community Edition and Enterprise Edition, are also accessible on the marketplace, with the Community Edition being pay-as-you-go and the Enterprise Edition requiring users to bring their own license. These listings come with a user interface that wraps a Terraform module, providing flexibility for custom deployments. Neo4j aims to gather user feedback on these updates to continue improving its integration with Google Cloud.

The integration of Neo4j's Graph Intelligence Platform with Google Cloud enables the development of advanced AI agents that can navigate complex enterprise environments by leveraging structured graph-based memory systems. These agents surpass traditional ReAct loops by incorporating reasoning models, such as those from Gemini, and context engineering techniques like GraphRAG to enhance explainability, accuracy, and traceability. The Semantic Knowledge Layer acts as a navigational tool, helping agents identify and access the right data and APIs, and the graph-powered knowledge layer maintains the relationships, provenance, and structure essential for navigating enterprise data. These systems support task completion by storing long-term structured memory, allowing agents to learn from past interactions and providing a rich source of information for decision-making. The use of Neo4j's graph capabilities, coupled with Google Cloud's native services, forms a robust foundation for scalable and secure agentic systems capable of handling intricate organizational tasks across various domains such as supply chain, fraud detection, and healthcare.

The blog post discusses the development and deployment of advanced AI agents using Neo4j's Graph Intelligence Platform on Google Cloud, highlighting the integration of graph-based memory and reasoning capabilities to enhance decision-making in complex enterprise environments. It explores how modern agentic systems have evolved to handle intricate tasks by utilizing Neo4j's graph-powered knowledge layers that maintain data relationships, provenance, and structure, thereby improving the explainability and traceability of agent decisions. The article outlines the use of GraphRAG for contextual retrieval and the semantic layer for navigating enterprise data landscapes, emphasizing the role of structured long-term memory in capturing organizational knowledge and decision flows. Neo4j's integration with Google's Vertex AI platform and tools like Gemini further supports the development of agents capable of performing reliable, context-aware operations by leveraging a blend of vector searches and graph traversals. Additionally, it introduces Aura Agents, a low-code architecture for democratizing access to knowledge graphs, and discusses the importance of reasoning memory and context graphs in making tacit decision processes explicit and accessible. The post also showcases how Neo4j's agent memory services and integration with Google's agent infrastructures, including the use of MCP (Model Context Protocol), facilitate the creation of stateful, explainable AI agents.

The text explores the challenges and solutions related to retrieving and retaining information in AI agent frameworks, specifically within the Microsoft Agent Framework using Neo4j's graph database technology. It highlights two primary issues: the retrieval of related but scattered data points and the lack of persistent memory in AI agents. The Neo4j context providers address these gaps by enabling graph-based retrieval of structured data and maintaining agent memory across sessions. The knowledge graph context provider enhances data retrieval by combining vector search with graph traversal, allowing agents to access interconnected information such as company products and risk factors, while the memory provider ensures continuity by storing conversation history and user preferences. Together, these providers improve the relevance and coherency of agent responses over time, transforming agents into more informed and personalized systems capable of leveraging both domain expertise and accumulated interaction history.

Neo4j has introduced a series of integrations designed to enhance agentic AI systems on Google Cloud by providing a persistent knowledge layer that improves contextual decision-making, accuracy, and traceability. These capabilities include deep integration with Google Gemini Enterprise and the Agent Development Kit (ADK), enabling Neo4j to work seamlessly with existing Google Cloud agent platforms. The neo4j-agent-memory package offers agents short-term and long-term memory, allowing them to store, learn from, and retrieve past actions, which aids in decision-making and reduces redundancy. The integration of Neo4j Agents with GraphRAG capabilities allows for advanced graph-grounded retrieval, enhancing agents' ability to perform complex reasoning tasks by navigating real relationships in the data rather than just performing vector similarity searches. These enhancements are particularly beneficial for sectors like finance, healthcare, and legal, where traceability and accurate decision-making are critical. With these tools, Google Cloud teams can implement an AI knowledge layer without needing significant infrastructure changes, thus optimizing their AI systems for better performance and reliability.

The guide provides a comprehensive overview of deploying a Neo4j Graph Database Agent into a production-ready enterprise environment using Google Gemini Enterprise, emphasizing the challenges of moving beyond local prototypes to ensure security, scalability, and cost-control. It details the use of the Model Context Protocol (MCP), the Google Agent Development Kit (ADK), and the Agent-to-Agent (A2A) protocol to create a decoupled, scalable system, with key architectural features including decoupled microservices, robust app-level security, granular cost control, and extensible logic for custom business requirements. The process involves deploying two separate services on Google Cloud Run: a standalone Neo4j MCP server and a custom Python ADK application, with stringent token management and authentication measures to maintain cost efficiency and security. The architecture supports customized queries and integrates with Gemini Enterprise through OAuth validation, facilitating natural language queries on complex graph structures. The guide concludes with deployment instructions, emphasizing the importance of observability, scalability, and extensibility in building secure, enterprise-grade AI agents.

In a blog post by Stu Moore and Olga Razvenskaia, the integration of graph algorithms with Graph Machine Learning (GraphML) is explored to enhance intrusion detection in Internet of Things (IoT) networks using Neo4j Graph Analytics for Snowflake. The process involves using the K-Nearest Neighbours (KNN) algorithm and GraphSAGE to identify and classify intrusions based on attribute-based graph construction, offering a more meaningful representation of node relationships compared to traditional methods. By leveraging Snowflake's secure data environment, the authors validate their approach using an academic paper and a dataset from the University of Queensland, demonstrating the effectiveness of using graph-based methods for network intrusion detection. The article highlights how the integration of these technologies allows for scalable, efficient, and accurate detection of network attacks, validated through experiments that showed the approach outperformed traditional machine learning methods. The piece concludes by encouraging users to explore Neo4j Graph Analytics for Snowflake, available in the Snowflake Marketplace, with a 30-day free trial and additional resources on Neo4j.com.

LoanGuard AI is a graph-based agentic AI system designed for compliance monitoring and financial crime investigation in Australian financial services, leveraging a knowledge graph to ensure explainability and traceability of loan approval decisions. By structuring data into a three-layer graph, LoanGuard AI connects borrowers, loan applications, and regulatory standards, allowing for seamless and auditable compliance assessments. This architecture supports faster investigations and audit readiness by embedding reasoning chains and cited evidence within the system, distinguishing it from typical models where explainability is retrofitted. The system utilizes Neo4j to traverse interconnected data nodes, ensuring that every decision is backed by a traceable chain of reasoning, which is crucial in regulated industries. Through separation of retrieval and reasoning processes, LoanGuard AI maintains clear fault boundaries, enhancing system reliability and trustworthiness. This approach not only addresses compliance but also highlights the importance of designing systems where reasoning is structurally integrated rather than appended after development.

The blog post discusses the advantages of migrating from APOC's `apoc.periodic.iterate` procedure to Cypher's `CALL {…} IN TRANSACTIONS` (CIT) for handling bulk data operations in Neo4j. APOC has traditionally been used to manage transactional data operations while avoiding out-of-memory (OOM) errors by allowing control over batch sizes. However, CIT has evolved to offer similar functionalities with additional benefits such as improved memory tracking, error handling, and enhanced readability and error detection capabilities through Cypher's tooling. CIT provides a more modern and efficient approach by simplifying query planning and execution, offering built-in query statistics, and handling entity transactions more effectively, thus reducing the risk of OOM errors and improving performance. The post highlights the ongoing improvements and support for CIT, contrasting it with APOC's maintenance mode status, and suggests that switching to CIT provides a cleaner, more reliable, and future-proof solution for managing complex queries in Neo4j.

This guide provides a detailed walkthrough on creating and integrating an Okta OIDC service account with Neo4j, emphasizing the benefits of using service accounts for secure, non-human identity management. By leveraging Okta as an identity provider and using OAuth 2.0 Client Credentials, users can avoid storing long-lived Neo4j credentials, instead opting for short-lived access tokens that improve secret management and allow fine-grained control over access. The guide covers essential steps such as creating a service application in Okta, configuring API scopes and claims, setting up Single Sign-On (SSO) in Neo4j, and authenticating with a JWT via the Python driver. The integration ensures that Neo4j trusts and validates tokens issued by Okta and maps claims to appropriate roles, enhancing security and auditability for automated workloads.

In the Neo4j Developer Blog, William Lyon discusses building a multi-agent memory system using a single Neo4j graph to enable AI agents to share short-term, long-term, and reasoning memory, crucial for collaborative problem-solving in fields like financial services. Traditional agent systems isolate memory, leading to inefficiencies and blind spots, especially in regulated industries where shared memory can prevent duplicated efforts, ensure awareness of flagged activities, and maintain an audit trail for compliance. By integrating Neo4j's graph database with AWS Strands, agents like a KYC analyst and a credit assessment agent can access and contribute to a unified, structured memory, enhancing decision-making and traceability. The shared memory model eliminates the need for message queues by allowing agents to query a common graph, ensuring immediate visibility of any agent's findings and creating a deterministic retrieval process that supports compliance inquiries. The system leverages Neo4j's graph architecture for precise memory management and incorporates reasoning traces as first-class nodes, offering a transparent and auditable AI memory framework.

AI agents, despite their advancements, often hit a ceiling in their ability to reason due to unstructured memories that lack logical connections, leading them to repeat mistakes and struggle in new environments. The concept of a context graph is introduced as a solution to this problem, transforming an agent's memory from a collection of isolated facts to a structured web of knowledge that maps relationships between decisions, outcomes, and the environment. This is illustrated through the story of an AI sheep navigating a digital forest, evolving from a reactive agent with short-term memory to one with long-term recall and eventually contextual reasoning, enabling it to learn and adapt strategically. The context graph allows agents to leverage structured experiences, providing a scaffold for language models to deduce and apply rules efficiently, enhancing an agent's ability to reason and adapt to changes. The article suggests that implementing context graphs can significantly improve AI reasoning by embedding the "why" alongside the "what," and provides guidance on building such systems using frameworks like Neo4j.

Enterprise AI projects often fail not because of flawed models, but due to a lack of understanding of the intricate relationships within a business, a problem termed as the "Context Gap." While AI models excel in generating predictions, they struggle with reasoning because they lack the structured relational context needed to truly understand business dynamics. This disconnect stems from traditional data storage methods that focus on retrieval rather than relationship modeling, leading to inaccurate and unreliable AI outputs. To address this, businesses are urged to adopt graph-based approaches that emphasize modeling relationships, which can significantly enhance AI reasoning, decision-making, and trustworthiness. By incorporating a knowledge layer that makes these relationships explicit, companies can transform AI from merely sounding intelligent to genuinely understanding and navigating complex business environments, thus gaining a competitive edge.

This edition of "This Week in Neo4j" highlights the latest developments in graph databases, focusing on the Neo4j Community Edition's major AI upgrade, which now includes native vector search, visual graph exploration, and streamlined cloud deployment options. The release aims to facilitate the transition from prototype to production-ready applications, particularly with GraphRAG and agentic AI systems, supported by tools like Neo4j Bloom and the create-context-graph CLI. It also announces an upcoming global graph-and-AI event, NODES AI, featuring sessions on Context Graph, GraphRAG, and AI in production. Additionally, GraphAcademy introduces a new course on building AI agents in Neo4j Aura, emphasizing no/low-code solutions and celebrating a milestone of 200,000 learners. The newsletter further discusses the architectural significance of the three-layer memory model for AI agents and provides insights into integrating Neo4j with Java using Spring Data.

Many enterprise AI initiatives struggle with traditional data architectures that aren't designed for AI's relational and contextual needs, making AI projects expensive experiments rather than valuable tools. A knowledge layer, built on a knowledge graph, resolves this by providing a centralized platform where AI can access relationships and context, enhancing decision-making accuracy and explainability. AI systems using graph-based grounding show significantly improved accuracy in tasks, as demonstrated by a study indicating a threefold improvement in large language model Q&A accuracy. This approach allows organizations to maintain their existing data infrastructure while enabling AI to reason effectively over data. Companies that have implemented a knowledge layer report substantial benefits, such as improved fraud detection, reduced ad costs, and accelerated compliance monitoring, positioning AI as a tool for reliable business outcomes.

The blog provides insights into the application of digital twins and knowledge graphs across various industries, highlighting their potential to optimize operations and uncover hidden inefficiencies. It features case studies and interviews, such as the use of Aura Graph Analytics on Databricks to identify bottlenecks in flight networks, Neanex's application of Neo4j in the construction industry, and London's traffic operations digital twin. Additionally, the blog includes perspectives from industry leaders like David Brian Ward of Telegraph Hill Software, emphasizing the transformative impact of digital twins in innovation and operational efficiency across sectors.

The text delves into the complexities of transforming graph structures into machine-readable features through embeddings, emphasizing the limitations of traditional adjacency matrices as graphs scale. It discusses the FastRP algorithm, which compresses high-dimensional data into lower-dimensional embeddings using random projection matrices, thereby preserving the relative distances between data points. The process involves converting nodes into one-hot vectors, aggregating neighbor embeddings with adjacency matrices, and refining these through reprojection and weighted sums to maintain distinctiveness. This method allows for the compact representation of graph data, making it compatible with standard machine learning models, thereby enhancing the efficiency of analyzing complex graph structures. The text is part of a broader series on graph algorithms, highlighting the practical applications of these techniques in analytics.

The article explains how to create a conversational AI agent with persistent memory using the Microsoft Agent Framework (MAF) and Neo4j, illustrated through the development of the TfL Explorer, a London transport assistant. This AI agent can remember user preferences and previous interactions, enabling it to provide more personalized and context-aware responses. The framework integrates the neo4j-agent-memory library, allowing the agent to access a shared Neo4j instance that stores both transport data and the agent's memory. This integration utilizes a pluggable architecture that combines various model providers and memory backends, such as Microsoft Foundry, Azure OpenAI, and Neo4j, to manage short-term, long-term, and reasoning memories. The article highlights the architectural benefits of using a single database for domain data and memory, which simplifies operations and enhances the agent's reasoning capabilities by linking memory entities to domain nodes. It also details how to set up the memory client, the types of memories managed, and the implementation of custom tools using the transport graph to execute Cypher queries directly within Neo4j. The agent's memory operations are non-blocking, and its tool calls return structured data, facilitating visualization in the frontend through Mapbox GL JS and Neo4j NVL.

The blog post by Irina Loghin discusses the vulnerabilities in satellite supply chain systems, emphasizing the potential risks posed by cybersecurity threats. It highlights how interconnected components from various suppliers create numerous attack surfaces, which can lead to significant disruptions if a critical satellite is compromised. Loghin presents a project using Neo4j's graph database to model the relationships between satellites, components, vulnerabilities, suppliers, and other entities, offering a visual representation of potential risks. The project aims to improve security by allowing analysts to quickly identify vulnerabilities and compromised roles using graph algorithms. Additionally, a Neo4j Aura Agent was developed to facilitate interaction with the database using plain language, eliminating the need for specialized knowledge in Cypher, thus aiding security and space analysts in their mission to protect satellite infrastructure. The blog underscores the importance of proactive measures in securing satellite systems and encourages further discussion and exploration of these issues.

The guide provides a detailed solution for securely exposing a Neo4j Enterprise Edition cluster across Virtual Private Clouds (VPCs) using AWS PrivateLink, without public exposure and with full Transport Layer Security (TLS) and clean private DNS configurations. It addresses challenges commonly faced in enterprise environments, such as the need for secure database connectivity and the risks associated with exposing network connectivity fully. The solution involves using AWS services like Network Load Balancers and PrivateLink to manage cross-VPC connectivity while maintaining cluster communication strictly internal. By configuring Neo4j nodes with specific advertised addresses for Bolt and HTTPS, and implementing security measures like least privilege access and private DNS, the guide ensures high availability, secure client connections, and minimal attack surfaces. This proven approach, based on real-world implementation, offers a scalable, secure, and efficient architecture for enterprises needing cross-region, cross-VPC connectivity for Neo4j.

Neo4j's Aura Graph Analytics offers a powerful method for extracting insights from data stored in Databricks by using graph algorithms without the need to set up new infrastructure or move data. This approach allows businesses to uncover patterns related to recommendations, fraud detection, customer behavior analysis, and more by modeling data as interconnected nodes and relationships. Various algorithms, such as similarity, pathfinding, community detection, centrality, and embedding algorithms, provide targeted solutions for industries like retail, financial services, and logistics, enhancing tasks like product recommendations, delivery route optimization, and fraud detection. Graph projections, created within serverless sessions, enable efficient in-memory processing of data to generate valuable insights, which can be integrated back into Databricks for further analysis or machine learning applications. This method facilitates the identification of complex patterns in data that are not easily visible in traditional table formats, thereby enabling organizations to make data-driven decisions with greater precision.

Corydon Baylor, a Senior Manager in Technical Product Marketing at Neo4j, discusses using Aura Graph Analytics on Databricks to detect bottlenecks in flight networks. The approach involves transforming flight route data into a graph model and applying various algorithms to identify potential disruptions. Using the Weakly Connected Components algorithm confirms the connectivity of the network, while Betweenness Centrality identifies critical hubs like Ted Stevens Anchorage International Airport, which plays a significant role in cargo operations. Simulating the closure of Anchorage due to weather and rerouting through Dijkstra's Shortest Path algorithm demonstrates the importance of alternative routes to maintain operational flow. This method highlights the benefits of using graph analytics to treat data as a dynamic and connected structure, allowing for real-time stress testing and rerouting.

Aura Graph Analytics (AGA) offers a solution to the inefficiencies of persistent server models in graph analytics by providing an on-demand, ephemeral compute model that aligns infrastructure costs with actual usage. Traditional persistent analytics environments, like AuraDS, often incur unnecessary expenses due to idle time and require complex infrastructure management, including data replication and server orchestration. AGA addresses these issues by allowing users to pay only for the compute jobs they execute, thus eliminating "hidden taxes" such as peak capacity and idle state costs. This approach simplifies workflows by reducing the need for manual data movement and infrastructure management, enabling engineers to focus on valuable data insights rather than operational overhead. The shift to on-demand compute is particularly advantageous for episodic workloads, ensuring that analytics are always run on the freshest data with the latest algorithms without the need for constant server maintenance.