June 2023 Summaries

At Neo4j, two employees, AJ Kincaid and Joe Depeau, share their experiences of embracing their authentic selves in the workplace. They reflect on how Pride Month has evolved over time, from a struggle to be seen as just another family to a celebration of individuality and acceptance. The company values authenticity, inclusivity, and diversity, and strives to create an environment where every team member can bring their whole self to work. Through open dialogue, education, and support, Neo4j aims to foster a workplace that celebrates the richness of LGBTQ+ identities and experiences, while acknowledging the ongoing struggle for basic rights and freedoms in some parts of the world.

The authors of the article, who work on the @neo4j/graphql library, were facing performance issues with their Node.js GraphQL server. To identify and fix these problems, they used a combination of tools and techniques, including measuring the performance of their server, load testing with K6, and using Clinic.js to analyze their code's performance. They found that the main bottleneck was not in the database or the library itself, but rather in the JavaScript code processing the GraphQL queries. By optimizing this code and using a more efficient GraphQL parser, such as Yoga instead of Apollo, they were able to significantly improve the server's performance. The article provides a detailed example of how these improvements can be made and offers practical advice for developers who want to optimize their own Node.js GraphQL servers.

This week in Neo4j highlights various projects and news, starting with NaLLM, a natural language interface to a knowledge graph and creating a knowledge graph from unstructured data. Integration with Google Cloud's LLM platform, Vertex AI, is also announced, along with tips for presenting at NODES 2023. Another project, AuraDS, is showcased as a tool for creating graph features using new features in Vertex AI. Arrows.app, a web-based tool for drawing pictures of graphs, is mentioned, and Testcontainers has released their second monthly newsletter. Additionally, there are updates on the CovidGraph project, a HealthECCO presentation by Lea Gütebier, and various blog posts on fine-tuning Large Language Models and building graph-based machine learning models with Vertex AI. SSO for Neo4j Sandbox is also implemented, and ThreatMapper's architecture is described as a cloud-native application protection platform.

This post guides you through creating a Spring application with data stored in Neo4j and GraphQL support. You'll learn how to integrate Spring Data Neo4j with Spring for GraphQL, enabling you to create a unified GraphQL surface that presents the combined data from your Neo4j database and remote systems/microservices. The example project demonstrates how to use QueryByExampleExecutor and QuerydslPredicateExecutor to improve query performance and add support for scrolling/pagination. You'll also explore the use of Testcontainers to run the application in a test environment, making it easier to develop and test your GraphQL API. By following this tutorial, you'll gain hands-on experience with integrating Spring Data Neo44j and Spring for GraphQL, allowing you to build more efficient and scalable data-driven applications.

This week's newsletter from Neo4j covers various topics such as integrating graph databases into large language model applications, using Neo4j with Java Spring Boot and React, and community detection with Graph Data Science. The newsletter also highlights upcoming events like NODES 2023 and features interviews with experts in the field, including Payel Bhunia, Tomaz Bratanic, and Surya Kunju, who share their experiences and insights on using Neo4j for various use cases. Additionally, a new course is introduced, which teaches how to use Neo4j as the backend storage for a real-world application created with Java and Spring Boot. The newsletter concludes by highlighting the advocacy programs run by Yolande Poirier, including the Ninja program.

Predictive modeling is a fundamental task of data scientists that involves using statistical models to make predictions about the future from past data. It has numerous everyday uses in industry, including identifying content-violating posts for social media sites, predicting stock values, estimating insurance claims, and evaluating advertising campaign effectiveness. To build a predictive model, one must first collect and organize the dataset, clean the data by handling missing values and inconsistent formatting, and then choose a suitable methodology or algorithm. The chosen model is then built and fine-tuned using techniques such as cross-validation and hyperparameter tuning. Python's NumPy, pandas, and scikit-learn packages provide an efficient way to build predictive models, and tools like GridSearchCV can help optimize the model's performance. Additionally, Neo4j Graph Data Science offers a native Python client and intuitive API for querying and configuring data, making it easier to create predictive models and integrate them into enterprise data ecosystems.

With the increasing amount of generated data, organizations are turning to advanced analytics, data science, machine learning, and AI to drive better forecasting, more accurate predictions, and innovative solutions. However, many businesses fail to reap these benefits due to a lack of clear goals, poorly fitted teams, or insufficient access to high-quality data. To overcome these challenges, powerful graph technology can help by providing comprehensive domain knowledge, improved analytics, and new sources for machine learning feature engineering. Predictive modeling is the process of analyzing patterns in data to predict future outcomes, and its benefits include incorporating external variables, reducing time and cost, and enabling more informed business decisions. Common predictive tasks include regression, classification, anomaly detection, clustering, and similarity matching, each with various algorithms such as linear regression, logistic regression, decision trees, boosting methods, neural networks, KNN, and K-Means. Despite its potential, predictive modeling faces challenges like choosing the best model or algorithm, finding hidden insights for better decisions, and ensuring data security and privacy. Graph data science engines like Neo4j can help overcome these obstacles by providing a seamless integration with existing data ecosystems and enabling data scientists to make better predictions, build proof-of-concept models, and go to production quickly.

The integration of game engines like Unreal Engine with tools for natural interaction and social processes, such as Neo4j, has the potential to bring benefits to both industries. Researchers can focus on interaction management models while avoiding technical limitations set by the entertainment industry or the need to integrate new control devices. A framework called FANTASIA supports the development of RTI3D applications based on natural interaction and social processes in Unreal Engine, providing access to Neo4j from inside the engine. This integration enables the use of graph databases and Bayesian Networks for argumentation research, such as Artificial Neural and Graphical Models for Argumentation Research (ANGMAR), which can manage tasks like Automatic Speech Recognition and Intent Recognition using neural approaches and graphical models respectively. The framework also supports task prioritization with behavior trees, allowing the system to evaluate communication issues and generate clarification requests based on detected problems.

NeoDash is a low-code tool that allows users to easily explore their Neo4j database using Cypher queries, with features such as line graphs, tables, maps, and input fields for manipulating results. The author of the blog post uses NeoDash to improve the search experience for 197M chemical names in their graph database, leveraging the APOC package to extend Cypher functionality. By integrating NeoDash into their Python backend, they simplify string manipulation and reduce code duplication. With NeoDash, the author achieves a fuzzy full-text search with graph enhancements without additional code, making it an attractive tool for building prototypes and future projects.

Developers and data scientists are invited to present their graph-powered projects or techniques at NODES 2023, a free online conference that brings together the community to learn about the latest in graph-driven innovation. The perfect opportunity to share knowledge, find new collaboration opportunities, and establish oneself as an expert. Proposals can be submitted by June 30th for one of three talk tracks: Intelligent Applications, ML/AI, or Visualization. For inspiration, check out standout talks from NODES 2022, including presentations on using social networks, exploring graphs visually with Jupyter Notebooks, and making sense of geospatial data with knowledge graphs. The deadline to submit proposals is June 30th, and the full NODES 2022 playlist is available for reference.

Neo4j, a leading graph database, has recently released an update to its Workspace feature, which includes new features such as Saved Cypher, plan view in text format, file import pre-filter, and import cancel. Additionally, the company is hosting various events including GraphMeetup, APAC Training Series, and GraphSummit Series. The company also introduced a new tool called deps.dev, which analyzes software dependencies for graph databases. Other topics covered include Neo4j DevTools, creating synthetic data with the Mock Data Generator, and harnessing large language models with Neo4j.

The LangChain Cypher Search: Tips & Tricks blog post explores how to optimize prompts for better Cypher statement generation in Neo4j for use in Large Language Models (LLMs) applications. The authors highlight the importance of using few-shot capabilities of LLMs by providing Cypher statement examples, which can help improve accuracy and relevance of generated statements. They also demonstrate how to integrate graph algorithms from the Neo4j Graph Data Science library into LangChain applications to provide personalized recommendations. The post showcases various use cases for integrating knowledge graphs into LLM applications and provides code examples for improving Cypher search and using graph algorithms for recommendation generation.

This blog post explains how to implement Single Sign-On (SSO) for a Neo4j instance using Auth0, an external authentication system. SSO allows users to access multiple applications with their existing credentials and centralize user management. The authors of the post describe how they use OAuth 2.0, OpenID Connect, and JWT (JSON Web Tokens) to implement SSO for Neo4j Sandbox. They also provide a step-by-step guide on configuring the Neo4j instance with Auth0 using Docker. The process involves creating an environment file, setting up authentication providers, customizing login button text, and allowing access to the web application on port 7474 from the IdP's settings. With this setup, users can log in to their Neo4j instance by selecting SSO in their browser application.

In a significant update, Neo4j has introduced Single Sign-On (SSO) feature in its Sandbox environment, allowing users to access multiple applications with a single login credential. This new feature is now available alongside version 5.x of the platform, which includes several enhancements and new features. The SSO feature is implemented using Neo4j Central IdP (Identity Provider) Auth0, enabling seamless user experience. Users can log in to the database browser and Bloom applications through the Sandbox Login flow, and also use standard username/password credentials. To login with SSO from sandbox.neo4j.com, users need to create a sandbox instance, open Neo4j Browser or Bloom, press the "Sandbox Login" button, log in with their sandbox credentials, and verify their user email address. The introduction of this feature aims to improve workflow and overall experience for Neo4j users.