dbt's end-of-support warning sweep lands across every release line

The deprecation push

The defining story of the week is that dbt Labs began warning users on deprecated dbt versions, and it shipped that warning across an astonishing breadth of release lines. The same feature landed in v1.1.6, v1.2.7, v1.3.8, v1.5.12, v1.6.19, v1.7.20, v1.8.10, v1.9.11, v1.10.23, v1.11.13, and v1.12.x. This is a rare case of old-version maintenance being treated as first-class work. If you're still on an ancient line, expect a prompt nudging you forward — and take the hint, because the warning also arrived alongside real maintenance.

The most consequential piece is the removal of Python 3.8 support in dbt-core v1.4.10, which now treats the EOL runtime as unsupported. The same change rolled into v1.5.12, v1.6.19, and v1.12.x, so if you're on Python 3.8 you're effectively being cut off across the board. Meanwhile the maintenance train also fixed the semver comparison logic and JSON log formatting issues in the older releases, so those lines aren't just getting warnings — they're getting genuine polish while they remain in the support window.

The 2.0.0 preview

The second headline is dbt Core v2.0.0-beta.1, which is where the real feature work lives. The headline additions are selector-based linting and formatting — you can now run dbt lint --selector name or --select/--exclude to target a subset of your project instead of linting everything. That's a significant workflow improvement for large projects where full-project lint runs are slow.

The beta also advances dbt's state machinery with speculative node submission during prefetch, which promises faster state-aware runs. Databricks Azure service principal auth and opt-in Redshift grants round out the integration work. What's notable is the quality-of-life trajectory here: the new version isn't just adding features, it's fixing the painful edges. The related patch releases this week — v1.12.2's fix for spurious Databricks config warnings, v1.11.13's fix for versioned model unit test parsing, and v1.12.1's corrections to unit test node counts and OTel span instrumentation — all point at the same theme: dbt is tightening its correctness around the newer features as 2.0 approaches.

ClickHouse's observability expansion

On the ClickHouse side, the story is about extending the ClickStack observability platform. The Terraform provider now manages ClickStack dashboards, alerts, sources, and webhooks, which is a meaningful step toward infrastructure-as-code for observability configs. The Grafana plugin's 4.20 release brings compact query mode, click-to-filter log investigation, guided editors for variables and annotations, and OpenTelemetry dashboards.

The deeper play is the integration between ClickStack and Hud sharing trace IDs, which pairs service-level observability with function-level runtime forensics. The idea is that coding agents can assess risky changes before shipping and catch regressions with production context. And the bitdrift mobile observability integration brings mobile-native telemetry to the platform, rounding out ClickStack's coverage from backend to device.

Finally, the pg_clickhouse v0.10.0 release pushes 16 of 22 TPC-H queries down in full, which is a substantial performance milestone for Postgres users querying ClickHouse through the extension.