Kùzu Inc.

It's been a fruitful 2024 for Kùzu Inc., and we wish all our users and the larger community very warm season's greetings, and some well-earned time off with family and friends. Merry Christmas!

Our final release of 2024 is out! -> v0.7.1 🚀 In this minor release, we provide two new extensions: You can now interact with Apache Iceberg and Delta Lake tables in Kùzu! This is part of our larger interoperability goals with the data lake ecosystem for graph workloads. 👇🏽 ✅ Apache Iceberg is one of the most popular and widely adopted open source table formats for large-scale analytical datasets. You can now easily bring data from your Iceberg tables into Kùzu tables and query it in Cypher ✅ Delta Lake is an open-source storage framework that enables building a format agnostic Lakehouse architecture. Our Delta extension allows you to easily bring data from your Delta tables into Kùzu for your graph workloads ✅ You can use a combination of LOAD FROM, COPY FROM clauses in Kùzu to scan/copy from with Iceberg and Delta Lake tables directly into Kùzu, using just a single line of Cypher. In the Iceberg extension, you can also use the CALL function in Kùzu to inspect the metadata and snapshot information of your tables 🎉 ✅ We also included some bug fixes with this minor release, and we think that the new Iceberg and Delta extensions will be quite useful if you're already working within the data lake ecosystem in Python! Give the new extensions a try! it's just a couple additional lines in Kùzu - documentation links for both extensions are shown below. 👇🏽 Cheers to a great holiday week ahead, best wishes from the entire Kùzu Inc. team, and enjoy the new release! Release notes: https://lnkd.in/gUqeVV9c Iceberg extension docs: https://lnkd.in/g_rF93Xf Delta Lake extension docs: https://lnkd.in/gaBEdFt4 #datalake #dataengineering #kuzu #etl

Watch an excellent walkthrough of Kùzu on the Knowledge Graph Technology Showcase hosted by Ashleigh N. Faith. 👇🏽 ✅ Get started with Kùzu in one of three ways: The Python client, the CLI, and our interactive UI (Kùzu Explorer) ✅ Check out the performance and scalability aspects of Kùzu by running a 2-hop pathfinding query in a time-frame of milliseconds on a graph with 40 million edges ✅ We point to a recent release post on blog.kuzudb.com where we showcase benchmark results on *very* large graphs (17 billion edges), so indeed, Kùzu is designed for fast query performance on large graphs! ✅ We end with some takeaways on how embedded databases could make it simple for a broader subset of developers to use graphs in their applications, from data exploration & investigative analyses to large scale graph querying on specialized domains Many thanks to Ashleigh for having us, and please give the video a like and subscribe to her channel! Video URL: https://lnkd.in/dPV7A4Tz Kùzu docs: https://docs.kuzudb.com/

Our recent talk "Cypher vs. SQL: When do you need a graph database?" at Connected Data London was among the top talks at the event, as voted for by the attendees! The talk recordings are still on the way, but in the meantime, you can revisit this excellent video on the topic by Semih Salihoğlu: https://lnkd.in/empa6ysG As the year winds down, why not give yourself a primer on graph databases, fire up Kùzu and write some Cypher! 🚀 https://lnkd.in/g3QrMGgb

It's been a great week in London with Kùzu Inc. at Connected Data London 2024. It was awesome to be able to present the scope of the Connected Data Knowledge Graph Challenge, with George Anadiotis and others at the event, and a huge thank you to the organizers for managing this despite facing technical difficulties at various stages. The conference was filled with a lot of really interesting talks from other speakers - too many to attend (unless you could be in two places at once). I have some takeaways from this event from a graph technology perspective, so I'll do my bit to list them below. ✅ Graph RAG is still very relevant today, though there is a need for more educational material and tooling that can help make the landscape clearer and easier to work with for users. In particular, graph construction and downstream evaluation are essential to help users gain confidence to put their Graph RAG applications in production. Shoutout to Andreas Kollegger and his colleagues at Neo4j who are maintaining the educational material in https://graphrag.com/ to help with this. ✅ Graph visualization is FAR more than just pretty pictures - it's about what insights the visualization conveys, and this depends on the stakeholder who is interpreting the results, the data and what questions are being asked. Loads more interesting graph analytics and visualization work to be done in 2025! ✅ There are quite a few practitioners who are coming into the world of graphs who also get overwhelmed by the complexity of terminology, like RDF, ontologies/schemas and so on. It's up to developers like me and others (who understand the technology as well as the larger developments in the field) to create educational resources and material to help these folks who are coming from the worlds of relational, graph and document databases to get more comfortable thinking of their data as graphs in certain situations. I'm grateful to be living in an age where the tooling we have is so user-friendly and there is such a degree of open knowledge sharing at events like this. Looking forward to future Connected Data London conferences, and see you at more of these kinds of events in 2025! Time to head back to 🇨🇦 ✈️. #graph #database #rag #graphrag

It's been a whirlwind week at Connected Data London for Kùzu Inc. Overall, we found a very vibrant and motivated #graph community here in London! We participated in two masterclasses on 11 December: ✅ Our AI Engineer Prashanth Rao participated in the Connected Data Knowledge Graph challenge, where the goal was to build a RAG application on top of Kùzu using the structured metadata and unstructured data from past events. We show how simple it is to get started using Kùzu for this kind of work. GitHub repo: https://lnkd.in/dzZBQcp4 ✅ Prof. Semih Salihoğlu and Prashanth Rao gave a workshop titled: "Cypher vs. SQL - when do you need a graph database?" This was all about the characteristics of Cypher, graph databases and a demonstration of the benefits of Cypher and a comparison with SQL for different queries. GitHub repo: https://lnkd.in/gxupYeyj From a Kùzu perspective, it's great for us to engage with a passionate crowd that wants to talk graphs all day. We look forward to the next phase to continue delivering a great solution! In case you couldn't make it to the event and want further information, don't hesitate to reach out. pip install kuzu

Happy Monday! We just released a tutorial showing how to build a simple Graph RAG pipeline using Kùzu's integration with LlamaIndex! The aim of this #graphrag demo is to answer questions from your unstructured text modelled as a graph. We'll use an interesting dataset of the scientist Marie Curie and her peers, and transform a block of unstructured text into a graph per a desired schema, persist the nodes/relationships extracted by an LLM in a Kùzu database, and then to answer questions through a text-to-Cypher pipeline. We'll cover the key steps involved in any RAG workflow, including: 1. Graph Construction (indexing): Use LlamaIndex and an LLM to extract nodes/relationships and store them in Kùzu. 2. Querying (also sometimes known as "serving"): Generate Cypher queries via a Text2Cypher pipeline to retrieve and respond to the user query in natural language. 3. Importantly, we highlight a downstream step that's important for constructing high quality knowledge graphs: data augmentation. You can leverage the Kùzu-LlamaIndex integration to fix inconsistencies by adding nodes/relationships and improving the factual quality of the triples stored in the graph. If you've used LlamaIndex for other tasks related to #rag, we hope this video will get you excited to turn your own data into graphs - give it a try! And do reach out to us on Discord for further tools and methodologies that can help you build your own Graph RAG workflows. 🚀 There's a lot more possible with this, so do reach out! https://kuzudb.com/chat https://lnkd.in/gX_WMXq4

We're happy to announce that we've started a newsletter, the first of which went out last week! The first issue talks about our recent release features as well as interesting links from various talks and demos that showcase how to work with graphs and Kùzu. If you're interested in keeping up with all the updates related to Kùzu, and if you haven't done so already, we highly recommend you sign up for our newsletter using the URL below! 👇🏽 https://lnkd.in/gXvYU-Rg

PyData Global 2024 is on, and there are two great talks on #graph and Kùzu lined up for tomorrow, 5th December, so please register and catch these live if you can! 1) A talk by the amazing Paco Nathan about constructing an investigative graph about money laundering, using spaCy, GliNER for entity extraction, and entity linking and entity resolution pipelines downstream. Should be interesting! Link: https://lnkd.in/gfVMqsm3 2) Another excellent talk by Alonso Silva Allende about building knowledge graph-based agents with structured text generation and open-weight LLMs. Uses llama.cpp + outlines + Kùzu. If you care about open source graph databases and open source tooling for agents/LLMs, this talk is for you! Link: https://lnkd.in/gdTJiGk2 It's great to see our awesome community continue to push their use of graphs and embedded graph databases. Please share your work and give us a shoutout if you use Kùzu for any interesting projects -- keep them coming! #graph #database #ai #pydata

Join us at PyData Global tomorrow at 11:30 AM EST, to hear about how to bring together the best of knowledge graphs and vector retrieval 😄. This talk will go over an application scenario that brings together the benefits of vector search with graph traversal. Knowledge graphs (or more generally, graphs), have long been used to model structured data that capture the connection between entities in the real world. Recently, there has been a lot of interest in the topic of Graph RAG, which aims to use graphs as part of the retrieval process in RAG, to enhance the outcomes. The talk will cover a practical example to showcase how Python developers can leverage the PyData ecosystem alongside two open source, embedded databases: Kùzu Inc. for the graph component, and LanceDB for the vector component of the retrieval. Register for this, and many other great talks through the "tickets" tab in the link below!👇🏽 https://lnkd.in/ghvKpr3c #graph #rag #database