July 2024 Summaries

Neo4j Bloom is a low-code visualization tool that allows users to visualize and understand their graph database. It provides various features such as formatting, custom searches, actions, and expansions, enabling users to analyze and explore their data in a more interactive and intuitive way. With Neo4j Bloom, users can create visualizations of their graph data, filter and animate the information, run graph data science algorithms, and even add custom searches and actions to further customize their visualization experience. The tool is available on both Neo4j Desktop and Neo4j Aura platforms and offers a user-friendly interface for exploring and understanding complex graph data.

Neo4j recently hosted a gathering in San Francisco to discuss the latest developments and applications of graph databases, including GenAI. The platform is also exploring graph exploration from the MEAN stack, transforming unstructured data into structured data, and analyzing ransomware payments. Additionally, Neo4j has launched a user research panel where participants can share their experiences with the product and influence future development. Upcoming events include Neo4j Live: Graph-Powered Code Debugging with GenAI, conferences in Sydney and Wisconsin Dells, meetups in Melbourne and Sydney, and the NODES 2024 conference. Featured community members include Sharmistha Chatterjee, who will speak at NODES 2024 about the role of knowledge graphs in mental health diagnosis, and Andreas Kollegger, who summarized a recent gathering of GenAI experts discussing topics such as RAG to GraphRAG and graph agent with auto-gen. The platform is also highlighting its capabilities for transforming unstructured data into structured data using SemSpect and introducing Triplex, an innovative new model that exceeds the performance of GPT-4o at knowledge graph construction for less than one-tenth the cost.

The tutorial provides a step-by-step guide on loading data into Neo4j using pyneoinstance, a Python library that makes it easy to ingest data into Neo4j. The author uses the Scooby-Doo dataset as an example and guides readers through the process of creating Cypher queries, separating them from Python code, and executing the queries using pyneoinstance. The tutorial covers how to load series and episodes nodes, add main characters to the database, create relationships between monsters and main characters, and finally, run all the necessary queries to complete the ScoobyGraph. Throughout the process, the author emphasizes the importance of separating Python code from Cypher queries for better maintainability and scalability. The tutorial concludes by highlighting the benefits of using pyneoinstance and providing a starting point for readers to create their own graph databases from dataframes.

This article guides readers through creating a graph database using Cypher from their tabular data. The author, Consulting Engineer at Neo4j, shares their experience with graphs and highlights the benefits of using them in data science. They use the Scooby-Doo dataset to demonstrate how to turn tabular data into a graph, including structuring the data model, creating Cypher queries to load the data, and adding nodes and relationships. The author provides step-by-step instructions and code examples to help readers create their own graphs using Neo4j and Cypher. By following this tutorial, readers can explore the ScoobyGraph and gain hands-on experience with graph databases and Cypher.

The Neo4j Query API is a new interface for querying Neo4j data through HTTP, designed to provide human readability, machine readability, streamability, and flexibility. The API supports two formats: `application/json` and `application/vnd.neo4j.query`, with the latter providing more complex interactions with results, including support for Neo4j types such as nodes, dates, and points. The new format adds a `type/value` pairing object to JSON, allowing for machine readability and streamability, while also enabling flexibility in handling new types. However, this format does not allow for pushing data into Neo4j using the same type system, requiring additional steps or parameterization of queries.

A knowledge graph is an organized representation of real-world entities and their relationships, typically stored in a graph database. It captures the context and meaning of how these entities are connected, allowing for deeper data insights and applications such as generative AI, fraud detection, and master data management. Knowledge graphs can be built with a small scope to solve specific problems, and they offer flexibility, performance, and developer-friendly code compared to other approaches like triple stores or relational databases. They provide a powerful tool for storing and organizing data to enable more sophisticated understanding of that data, and various use cases include generative AI, fraud detection, master data management, supply chain management, investigative journalism, and drug discovery in healthcare research. To get started with knowledge graphs, one can download a free guide or explore learning resources such as Semantics workshops on GitHub, Ontology-Based Reasoning 101, and GraphAcademy for knowledge graph fundamentals using a property graph model in the Neo4j Graph Database.

A knowledge graph is an organized representation of real-world entities and their relationships, typically stored in a graph database. It organizes data and relationships alongside frameworks known as organizing principles, which capture the context and meaning of how entities are connected. Knowledge graphs can be used to solve various problems such as generative AI for enterprise search applications, fraud detection and analytics in financial services, master data management, supply chain management, investigative journalism, and drug discovery in healthcare research. To get started with knowledge graphs, it's recommended to download a free guide or take online courses to learn the concepts and techniques behind knowledge graphs and gain hands-on experience building, managing, querying, analyzing, and visualizing knowledge graphs.

This article discusses the use of graph visualization tools like Bloom from Neo4j AuraDB to empower open-source cyber threat intelligence analysis. It highlights the importance of open-source cyber threat intelligence and its benefits, such as accessibility, cost-effectiveness, and collaboration among security professionals. The author demonstrates a low-code approach to combine Bloom with AlienVault's Open Threat Exchange (OTX) to enable more powerful visual threat investigations. The tutorial covers creating a database schema, launching Bloom, importing a perspective, exploring WannaCry ransomware intelligence, and discussing further discussion points. The article emphasizes the potential of graph visualization tools to simplify navigation through related information, provide on-demand intelligence retrieval, and facilitate scenario-specific analysis using Cypher.

This week's news from the world of graph databases includes the GraphRAG Manifesto, a visualisation tool with React, and an overview of CSV Import into Neo4j. The GraphGeeks podcast is also featured, discussing innovations in graph technology. Meanwhile, Hackerspace Mumbai shared their experience with leveraging LLMs for knowledge graph construction from unstructured data. Additionally, the community is invited to join the Neo4j User Research panel, which offers a chance to share experiences and influence product development. Upcoming events include conferences, meetups, and NODES 2024 registration. The featured community member, Mike Morley, showcased a Personal Knowledge Vault with Neo4j GraphRAG, transforming website URLs into structured graph documents.

The neo4j-graphrag Python package simplifies the integration of Neo4j into Retrieval-Augmented Generation (RAG) applications for developers. It equips users with tools to efficiently manage retrieval and generation processes, providing a variety of retrievers tailored for different retrieval strategies. The package facilitates graph Retrieval-Augmented Generation (GraphRAG), integrating graph databases with vector search. Users can connect to a pre-configured Neo4j demo database and visualize the data using Cypher commands. The package provides an official implementation for the OpenAI LLMs, but its interface is compatible with LangChain chat models and lets users write their own interface if needed. Users can use the VectorRetriever class from the neo4j-graphrag package to execute a straightforward retrieval query, and future posts will explore other retrieval strategies and how to customize different types of retrievers. The package code is open source, and users are invited to integrate it into their projects and share their insights via comments or on the GraphRAG Discord channel.

The text discusses integrating unstructured.io with Neo4j AuraDB to build a document knowledge graph. Unstructured.io is a tool for parsing and extracting data from PDF documents, while Neo4j AuraDB is a cloud-based graph database. The integration allows for the creation of a document knowledge graph that can be used to improve the efficiency, accuracy, and contextual relevance of Retrieval-Augmented Generation (RAG) systems. The text provides a step-by-step guide on how to use unstructured.io with Neo4j AuraDB, including installing the tool, setting up a Neo4j instance, and defining a document graph schema. It also discusses the challenges faced during the integration process, such as handling sentence breaks across pages and inconsistent recognition of page headers. The text concludes by highlighting the potential benefits of integrating unstructured.io with Neo4j AuraDB for building more effective RAG systems.

Welcome to This Week in Neo4j, your weekly fix for news from the world of graph databases! Our CTO, Philip Rathle, is interviewed about GraphRAG and GQL. We're creating Knowledge Graphs from Texts and Images using LangChain and Neo4j. You can turn a Relational Database into a Graph Database in Python. Anansi is an Enterprise Wrapper for Neo4j designed to streamline the experience for enterprises. Additionally, we have a User Research panel where you can share your experiences with a researcher and influence the future of Neo4j products, as well as upcoming events including conferences, meetups, and webinars.

Creating a graph of chemical reactions using tools like Neo4j and databases such as ORD can significantly enhance our understanding of complex reaction networks and their interdependencies. By representing chemical reactions as nodes and establishing relationships between input and output components, we can visualize intricate reaction processes and explore potential applications. The use of standardized systems for data representation, such as InChI, enables the development of applications that will greatly benefit computer-aided synthesis planning, reaction prediction, and other predictive chemistry tasks. This approach facilitates collaboration within the scientific community by providing a centralized repository for sharing data, ultimately promoting the advancement of computer-aided synthesis planning and related fields.

We're on the cusp of realizing that autoregressive LLMs alone won't suffice for making useful decisions with GenAI, and instead, we need to bring knowledge about things into the mix. This is where GraphRAG comes in - a technique that uses both vector-based RAG and fine-tuning, but also incorporates knowledge graphs to provide more accurate and complete answers. GraphRAG offers several advantages, including higher accuracy, easier development, and better explainability. It's particularly useful for applications where answer quality is essential, such as customer service or legal documents. With the Neo4j LLM Knowledge Graph Builder, users can create a knowledge graph in just a few clicks from unstructured text sources like PDFs, web pages, and YouTube videos. As GenAI progresses, knowledge graphs are becoming increasingly important for applications where answer quality is essential, explainability is needed, or fine-grained controls over access to data are required.

The new Cypher co-pilot is a feature available in Neo4j Aura that assists users with generating Cypher queries. It's designed for both beginners and experienced users, aiming to save time by providing a good initial statement to work from. The co-pilot uses a base LLM from OpenAI, which keeps the model primed with the context of the current database schema. Users input their natural language question, and the LLM generates a Cypher candidate that can be edited and adjusted before execution. While the co-pilot is useful for generating queries, it's not perfect and may produce incorrect results, so users need to validate the output carefully. The feature has limitations, including hallucinating model shortcuts, using deprecated syntax, and variable naming clashes. To improve its quality, Neo4j plans to fine-tune a custom model in the coming months.

Graph databases like Neo4j are being explored as a solution to combat disinformation campaigns on social media, particularly in the context of the 2024 U.S. presidential election. These databases can identify malicious accounts by uncovering connections between them that would otherwise go undetected. By representing social media advertising accounts and credentials as nodes in a graph database, researchers can analyze patterns and anomalies to distinguish between legitimate and malicious connected accounts. Graph databases also maintain historical data and relationships associated with deleted accounts, enabling analysis of their legitimacy. Furthermore, they can identify connections between accounts that don't share common credentials, shedding light on previously hidden influence and malicious activity. By leveraging relationship analysis, graph databases can help ensure the integrity of social media advertising during elections.

The implementation of the "From Local to Global" GraphRAG approach combines text extraction, network analysis, and LLM prompting and summarization for improved RAG accuracy. The pipeline begins with input text from documents, which are processed to generate a graph. The graph is then converted back into natural language text, where the generated text contains condensed information about specific entities or graph communities previously spread across multiple documents. A knowledge graph data representation allows quickly combining information from multiple documents or data sources about particular entities. After constructing the knowledge graph, a combination of graph algorithms and LLM prompting generates natural language summaries of communities of entities found in the knowledge graph. These summaries contain condensed information spreading across multiple data sources and documents for particular entities and communities. The approach involves using Neo4j as the underlying graph store and LangChain for implementing the GraphRAG pipeline. The code repository is available, and the project page can be accessed. Entity resolution is crucial when constructing a knowledge graph to ensure that each entity is uniquely represented. An LLM-based entity resolution process is implemented to identify potential duplicates and decide which entities should be merged. Element summarization involves generating natural language summaries of nodes and relationships in the graph, while community summarization generates summaries for communities of entities found in the knowledge graph. The final step involves storing the community summaries back to the database. Overall, this implementation demonstrates a new approach to GraphRAG that combines text extraction, network analysis, and LLM prompting and summarization for improved RAG accuracy.

Neo4j Desktop 1.6.0 introduces restrictions on third-party graph apps due to limited usage and increasing maintenance costs. The restricted apps include Semspect, Graphlytic, yFiles Neo4j Data explorer, GraphXR, and Neo4j Commander 3. These apps will no longer be signed and verified as trusted by Neo4j, and installation via certain methods will not work. However, Neo4j Labs Graph Apps will still be available pre-installed in Neo4j Desktop 1.5.9, and users can access them through the Gallery App or file protocol. Users who need to use third-party Graph Apps are advised to consider alternatives such as using Neo4j Desktop 1.5.9, SemSpect, Graphlytic, yFiles Neo4j Data explorer, or reaching out to developers for custom solutions.

Neo4j, the graph database company, has released its monthly podcast featuring Oleg Šelajev, a Neo4j Ninja and speaker at NODES. The podcast discusses topics such as Testcontainers and AI in the Java ecosystem. Additionally, Neo4j is promoting an open-source knowledge graph schema library called WhyHow.AI, which offers various schemas for different domains. The company has also released tutorials on how to convert text to Cypher using a language model and building a chatbot with LangChain. Furthermore, Neo4j encourages users to join its user research panel, where they can share their experiences and influence the future of Neo4j products. Gary Lilienfield, a Neo4j Ninja, is featured as a community member, while Oleg Šelajev is also discussed in more detail in the podcast episode. The company has upcoming events, including a livestream on July 9, conferences in Berlin and Bangaluru, and meetups in Berlin and Bangalore. Finally, Purring Lynx shares their positive experience with Neo4j, highlighting its versatility as a graph database.

GraphRAG is a method for using vector search to find relevant documents in a knowledge graph, with the goal of providing more accurate and reliable answers to questions. The author demonstrates this by building a graph of movies and TV series, and then uses GraphRAG to ask ChatGPT a question about Robb Stark's family, which it correctly answers based on the synopsis of a relevant document. However, when asked a question about a movie that was released after ChatGPT's training data, it hallucinates an answer instead of saying "I don't know". The author suggests using GraphRAG to improve the accuracy and reliability of AI-powered chatbots like ChatGPT by leveraging the power of knowledge graphs.