#kuzudb — Public Fediverse posts

Another day, another "blazing fast" graph database bites the dust 🚀💥. Apparently, Kuzu DB devs have better things to do than support their own project—maybe they were too busy trying to enable #JavaScript on their own site? 😂🔧
https://kuzudb.com #blazingfast #graphdatabase #KuzuDB #fail #technews #HackerNews #ngated

Another day, another "blazing fast" graph database bites the dust 🚀💥. Apparently, Kuzu DB devs have better things to do than support their own project—maybe they were too busy trying to enable #JavaScript on their own site? 😂🔧
https://kuzudb.com #blazingfast #graphdatabase #KuzuDB #fail #technews #HackerNews #ngated

Kuzu DB devs no longer supporting the project

https://kuzudb.com

#HackerNews #KuzuDB #KuzuDBDevs #ProjectSupport #TechNews #DatabaseNews

Kuzu DB devs no longer supporting the project

https://kuzudb.com

#HackerNews #KuzuDB #KuzuDBDevs #ProjectSupport #TechNews #DatabaseNews

Kuzu, an in-process (embedded) graph database is fast becoming my favourite technology for 2025. In-process, fast, Cypher graph query language, vector indexes (for vector embeddings), full-text search... and, more.

Check it out, here: https://kuzudb.com/

#kuzudb #graph #database #cypher #vector #genai #contextualise

Kuzu, an in-process (embedded) graph database is fast becoming my favourite technology for 2025. In-process, fast, Cypher graph query language, vector indexes (for vector embeddings), full-text search... and, more.

Check it out, here: https://kuzudb.com/

#kuzudb #graph #database #cypher #vector #genai #contextualise

Pushing my Logseq graph into a #KuzuDB database on disk takes 5.0 seconds on average. The same logic pushing to a database in memory takes 4.8 seconds on average. (M1 MBP w/32GB RAM)

I strongly suspect this says more about my schema complexity than it says about KuzuDB.

Getting funky with the graph connections, using #KuzuDB to understand co-citations – how often two pages in my imported Logseq graph are linked by the same page, suggesting a contextual relationship.

I think I got the cypher query right. I IS LEARNING.

#KuzuDB 0.8.0 includes WASM support, for in-browser graph database shenanigans.

I have the worst idea for a Logseq plugin.

https://blog.kuzudb.com/post/kuzu-0.8.0-release/

And a selection of pages linking to my Python note. Alas, no rand() for ordering in #KuzuDB yet.

Doing entertainingly useless things with Cypher in #KuzuDB

The pages I've linked to most in my Logseq graph. Some inaccuracies, since I haven't included tag links, and "note" should be a property link.

New #KuzuDB release! Key items of interest for me: list lambdas for tighter description of operations on lists, you can write user-defined functions — stored procedures basically — in Python, and there's better support for Polars, SQLite, and JSON as data sources.

Plus it's started on Multi-Version Concurrency Control (MVCC) and there's already performance improvement when doing loads of small write transactions.

(The post is currently dated July 31 but the actual release was about 20 minutes ago.)

Kùzu 0.5.0 Release

@randomgeek Thank you for introducing me to #KùzuDB! It has become really useful for data science workloads and has now found a snug place alongside #DuckDB.

Credit to @randomgeek for making me aware of #KuzuDB. Their relational database interop makes hacking together graph stuff on top of relational databases *much* easier.

Got enough Logseq structure in #KuzuDB that now I'm on to a publishing workflow.

And because this is me we're talking about, I'm doing the script to generate site content from the database (which I built with Python) in JavaScript.

Steadily learning more about complex queries with Cypher and #KuzuDB

This, for example, is probably too much.

https://pastebin.com/H58LMcMX

Now #KuzuDB has branch / leaf relationships for branches. Still need to chew on some details, but this is starting to resemble my mental map for a #Logseq graph.

MATCH
    (block:Block)-[:InPage]->(page:Page {name: "My Logseq Workflow"}),
    (block)-[InBranch]->(b:Block)
RETURN *;

Now #KuzuDB has branch / leaf relationships for branches. Still need to chew on some details, but this is starting to resemble my mental map for a #Logseq graph.

MATCH
    (block:Block)-[:InPage]->(page:Page {name: "My Logseq Workflow"}),
    (block)-[InBranch]->(b:Block)
RETURN *;

Remember how I said importing to #KuzuDB from CSV took about 6s, and 18s using CREATE statements in transactions (90 without)? And how I wanted to see how loading from a DataFrame compared?

Well after a brief digression and bug report, I got it working.

Range of 1.5 to 1.92 seconds after several timed runs.

Mind you these times include wiping out existing data files, loading and parsing the 20,327 blocks in the 1,576 pages of my Logseq graph, and then copying into the database.

So yeah. Loading from DataFrames is good.

This is in Python. Probably faster in Rust, but they don't support Polars yet. And I'd need to learn Rust.

Remember how I said importing to #KuzuDB from CSV took about 6s, and 18s using CREATE statements in transactions (90 without)? And how I wanted to see how loading from a DataFrame compared?

Well after a brief digression and bug report, I got it working.

Range of 1.5 to 1.92 seconds after several timed runs.

Mind you these times include wiping out existing data files, loading and parsing the 20,327 blocks in the 1,576 pages of my Logseq graph, and then copying into the database.

So yeah. Loading from DataFrames is good.

This is in Python. Probably faster in Rust, but they don't support Polars yet. And I'd need to learn Rust.

Generating then importing CSV takes about 6s, but has required constant fiddling with text formatting for multiline blocks, quote characters, stuff like that.

A whole lot of CREATE statements takes 18s, but hey it was more like 90 before I wrapped those statements up in a couple transactions.

#KuzuDB also has functionality to load from a DataFrame, and I'll probably check that out too. See how it compares.

Generating then importing CSV takes about 6s, but has required constant fiddling with text formatting for multiline blocks, quote characters, stuff like that.

A whole lot of CREATE statements takes 18s, but hey it was more like 90 before I wrapped those statements up in a couple transactions.

#KuzuDB also has functionality to load from a DataFrame, and I'll probably check that out too. See how it compares.

Remembered that in a #Logseq style PKM graph, every node is a page. The details are in the edges. Updated code to reflect this in #KuzuDB.

Remembered that in a #Logseq style PKM graph, every node is a page. The details are in the edges. Updated code to reflect this in #KuzuDB.

Got enough data in the database that I can start learning Cypher and making queries.

So I think this is all direct links between public pages in my #Logseq graph, as interpreted by my code and #KuzuDB.

Got enough data in the database that I can start learning Cypher and making queries.

So I think this is all direct links between public pages in my #Logseq graph, as interpreted by my code and #KuzuDB.

Very rudimentary – I haven't added attributes or support for placeholder links – but now I've got my #Logseq graph connections loaded into #KuzuDB, with a pretty good graph view in the Kuzu Explorer interface.

Was hoping to get this much done before we were on the road for our move. Pleased to have met my deadline.

I'm 1 for 34,982! Go me! :party_parrot:​

Very rudimentary – I haven't added attributes or support for placeholder links – but now I've got my #Logseq graph connections loaded into #KuzuDB, with a pretty good graph view in the Kuzu Explorer interface.

Was hoping to get this much done before we were on the road for our move. Pleased to have met my deadline.

I'm 1 for 34,982! Go me! :party_parrot:​

Kùzu is another open source property graph database management system with support of Cypher query language just like #neo4j and #memgraph (and Oracle with PGQL): https://kuzudb.com/

#kuzudb #propertygraph

Kùzu is another open source property graph database management system with support of Cypher query language just like #neo4j and #memgraph (and Oracle with PGQL): https://kuzudb.com/

#kuzudb #propertygraph