The Weekly Edge: Infinigraph, GQL in KQL, Youtu-GraphRAG & More [12 September 2025]

If you’re new to the world of graph technology – or just need a refresher – this week’s watch is the perfect place to start. In just under 2 hours, this GraphGeeks mini-course surveys the graph technology landscape from top to bottom, from beginner to advanced. Moderated by the Mon Motha of the Graph Republic Amy Hodler and taught by the Gandalf of Graphs Paco Nathan himself, this is well worth the binge. 

Part 1 of the series starts with the fundamentals, core regions of the landscape, common applications, frequent pitfalls, and what graph technology won’t solve. Part 2 of Graph Tech Demystified gets into more advanced topics, like motifs, probabilistic subgraphs, graph neural networks (GNNs), GraphRAG, and more. Grab Paco’s slides for Part 1 and Part 2 here.

Launch: Neo4j Announces Infinigraph, a Scalable, Distributed Graph Architecture

In case you’ve been living under an LPG rock, the big news from last week is that Neo4j has launched a breakthrough new graph architecture to unify transactional and operational workloads. According to Neo4j’s Dan McGrath, the new scalable, distributed graph architecture – called Infinigraph – allows you to run 100TB+ operational and analytical graph workloads in a single system. 

If you’re curious for more technical details, German San Agustin wrote a companion piece about property sharding in Infinigraph that dives deeper into the workings of the new Neo4j architecture. Kudos to the Neo4j team on such a big achievement!

Update: GQL on KQL Graph Semantics for Azure Data Explorer & Fabric Eventhouse

For those working in the Microsoft ecosystem, more big graph news: The Azure team is adding Graph Query Language (GQL) support to KQL graph semantics in public preview. This update allows you to run GQL queries on any Fabric Eventhouse or Azure Data Explorer, making it easier to work with graph data using the ISO industry-standard language. Check out the documentation to get started with some examples.

Microsoft PM Henning Rauch shared the news earlier this week and encouraged users to submit feedback while GQL support is still in public preview. This announcement represents another big step toward industry standardization around GQL as the future of the graph space, much as SQL is to relational databases.

Release: ArcadeDB v25.8.1 Offers New Schema Features, Enhanced Kubernetes Support & More

Completing the trifecta of release news this week is the GA release of ArcadeDB v25.8.1. The latest version of ArcadeDB multi-model database introduces a powerful new schema feature, enhances Kubernetes support, and resolves several important bugs to improve stability and performance.

Spearheaded by long-time graph veteran Luca Garulli, ArcadeDB supports key-value, document, search engine, vector, time-series, and graph data models all in one. 

Paper: Vertically Unified Agents for Graph Retrieval-Augmented Complex Reasoning

If this week’s news feels too hands-on for you, then brace yourself for some intellectual whiplash as we go highly theoretical: A team of Chinese researchers out of the Tencent YouTu Lab recently dropped a paper on vertically unified agents for graph retrieval-augmented complex reasoning, dubbed YouTu-GraphRAG. The proposed framework aims to bring together both graph construction and graph retrieval during the GraphRAG process. Check out the code on GitHub or grab the data on Hugging Face.

In case you haven’t heard of them, Tencent YouTu Lab (or 腾讯优图 meaning “excellent picture”) is an AI research department of Tencent that focuses on computer vision, including optical character recognition (OCR), image understanding, and image generation.  

P.S. Support for GQL on Ultipa Graph is almost ready for release on G.V(). Check out our dev preview on LinkedIn and let us know what you think.

That’s it for this week’s edition. Got something you want to nominate for inclusion in a future edition of the Weekly Edge? Ping us on on X | Bluesky | LinkedIn or email weeklyedge@gdotv.com.