Executable UDFs and a new dbt core beta lead the week

Executable UDFs arrive on ClickHouse Cloud

The week's most practical new capability is executable UDFs entering public beta on ClickHouse Cloud. You can now write a Python function, upload it as a zip to your cluster, and call it from SQL like any built-in. ClickHouse manages a pool of long-lived sandboxed processes and routes rows through them at query speed — meaning the function is callable anywhere SQL is: ad-hoc queries, joins, even materialized views that fire on every insert. This isn't a new idea (executable UDFs have been available in self-managed ClickHouse for a while), but the cloud beta removes the infrastructure barrier. For teams that need custom logic — say, a geocoding step or a regex-heavy transformation — without standing up a separate microservice, this is a direct path.

ClickHouse 26.5: performance at scale

Alongside the UDF beta, ClickHouse Release 26.5 landed with a record 51 performance optimizations. The headline numbers: ORDER BY … LIMIT pushdown through joins can deliver up to 20× faster queries, and a new GROUP BY … LIMIT shortcut avoids building unnecessary groups. There's also a new filesystem table function for running SQL directly against your local file system, which is useful for ad-hoc exploration without an ingest step. For observability practitioners, a companion post on ClickStack Event Deltas explains how to use ClickHouse's lesser-known feature for faster root-cause analysis in slow traces — a practical walkthrough worth reading if you're using HyperDX or building your own observability stack.

dbt-core: alpha meets beta

dbt-core had a bifurcated week. dbt-core v1.12.0b2 brings several quality-of-life features: the maximum seed size is now configurable, and versioned models can automatically create a latest_version pointer (a view aliasing the most recent version of a model) — opt-in via a flag or per-model config. The same day, dbt-core v2.0.0-alpha.1 shipped with a similar latest_version_view feature (reverted and re-added in the same alpha cycle, so expect it to stabilize). The alpha also includes an AI summary for the docs issue workflow. The takeaway: if you need the latest version pointer today, v1.12.0 beta is the stable path; v2.0.0 alpha is worth testing for forward compatibility.

Monitoring and community

Two ClickHouse blog posts round out the week with practical guidance. kubenetmon is a newly open-sourced tool for monitoring data transfer in ClickHouse Cloud — useful if you're running multi-cluster or egress-sensitive workloads. Separately, a guide on using clickhousectl and agents to monitor SLAs and scale walks through tagging queries, calculating SLAs, and automating scale decisions. On the Dagster side, Community Showcase Part 1 highlights creative projects built by the community — from public data exploration to infrastructure monitoring — with notes on why each team chose Dagster.