chDB v3.1.2 has arrived with improved hardware compatibility. Additionally, we're excited to welcome our new contributor, Michael Eastham. 😎🎉

chDB has 13 contributors now!

Just discovered an impressive Python Notebook project called marimo , which has 12,000 stars on GitHub. using chdb v3.0

We're excited to announce that chdb-ruby is now available for testing. run! chDB on Rails🛞 github.com/chdb-io/chdb-r…

🏎️ What’s better than a fast database? A lazy one. ClickHouse’s new lazy materialization feature delays work until it’s absolutely needed, slashing runtimes by over 1,200×. Curious yet? Read the deep dive here → buff.ly/9CiK99o

I developed a compact RAG system last week. I realized that columnar databases could be well-suited for RAG vector search because: 1. Inserts are typically batched. 2. Datasets are usually massive, with 1,000-2,000 dimensions and numerous rows. 3. Query latency isn't a major

We are hiring! grnh.se/78b7f4fc4us

chdb-node v1.3.0 is out now with query bind support! 🎉 Big thanks to the amazing chdb community for their efforts! Check it out: github.com/chdb-io/chdb-n… Shoutout to github.com/rajdude0 for the contributions! 🚀

Potential of MCP in Database Applications is still underestimated auxten.com/potential-of-m…

chDB v3.3.0 is here, with support of JSON columns! 🚀🚀 github.com/chdb-io/chdb/r…

chDB v3.4.0 released with ClickHouse v25.5 inside🚀🚀🚀🚀 github.com/chdb-io/chdb/r…

chdb-go v1.10.0 release with steaming query supported

🎾 Data > Intuition: That nail-biting 5-set Alcaraz match in the 1st round at Wimbledon? The numbers tell a different story. Using ClickHouse/chDB to analyze point-by-point data, Mark Needham discovered Alcaraz was never truly in danger despite the drama. He had to build

Mark Needham Blog: dub.sh/5zbmNep App: buff.ly/d4gqcqk

chDB v3.4.1 released with better support for Delta Lake and Iceberg

In the end, every byte goes ocean 🌊 ⁦ClickHouse⁩

chDB v3.5.0 released with better support of PyArrow RecordBatchReader github.com/chdb-io/chdb/r…

Running #apachespark in pure driver mode (basically acting as a single-node compute) with 2 cores is too slow, but with 4 cores it’s ok-ish. Anything beyond 16 CPUs is a waste of compute since you’re just waiting on object store. spark in single node is not too bad actually !!