Skip to content
LayerbaseLayerbase

DuckDB 1.5 Is Now on Layerbase

3 min readDuckDBAnalyticsDatabases

Create a DuckDB database on Layerbase today and you get DuckDB 1.5.5, the release the DuckDB team shipped on July 22. We moved managed DuckDB from the 1.4 line to 1.5.5 two days later.

That gap is the point. Managed database services usually run whatever version they qualified six months ago, and you find out how far behind you are the first time a query relies on something the docs say exists. We would rather track the engine closely, so when DuckDB cuts a release, updating the version you can provision is a short, boring process instead of a quarterly project.

What moving to 1.5 actually changes

Layerbase was on the 1.4 line, so this is a real step, not a patch bump. Across the 1.5 (Variegata) series, DuckDB picked up:

  • A friendlier CLI with better ergonomics for interactive work.
  • The VARIANT type: typed, binary semi-structured data in the spirit of Snowflake's VARIANT. Unlike a JSON column stored as text, each VARIANT value carries its own type information, which compresses better and queries faster.
  • A native GEOMETRY type built into the core engine.
  • A push into lakehouse formats, with DuckLake and Lance support and the experimental Quack remote protocol that turns DuckDB into a client-server database.

The full list is in the DuckDB 1.5.0 announcement. 1.5.5 itself is a stability release: a 128-bit DECIMAL statistics correctness fix, several concurrency crash fixes, and storage and Arrow bugfixes. That is exactly the kind of release you want sitting under a database you are not the one operating.

What managed DuckDB on Layerbase looks like

It is free. DuckDB is on the Layerbase free plan: no card, no usage meters, no per-query billing. You create one from the dashboard and have a connection string a couple of seconds later.

It speaks Postgres wire. DuckDB never shipped a network protocol of its own, so on Layerbase you connect over the Postgres wire protocol. Point psql, the pg npm package, psycopg, pgx, or a BI tool's Postgres data source at it and run SQL. No custom driver, no adapter.

There is a web SQL console. The database's dashboard page includes a query console, so checking a row count or eyeballing a table does not require a local client at all.

Free databases sleep when idle and wake on your next connection, so an analytics database you touch a few times a day is not quietly costing you anything between queries.

If you are already running DuckDB here

Nothing breaks. A database keeps the version it was created on: if you are on 1.4, your database stays on 1.4 and keeps running exactly as it did. We do not swap the engine out from under a live database. New databases default to 1.5.5, so when you want the 1.5 line, create a fresh database (or branch an existing one) and point at that.

Try it

DuckDB is embedded, so it is just as reasonable to prototype locally first. The Layerbase CLI runs the same engine on your machine with no account:

bash
npm i -g layerbase
lbase create analytics --engine duckdb --start --connect

When you are ready for a managed instance with a shareable connection string, create a DuckDB database on Layerbase. Our paid plans are flat monthly with no usage metering (pricing); DuckDB itself stays free.

New to the engine? Start with getting started with DuckDB. Weighing it against a columnar warehouse? ClickHouse vs DuckDB. Coming from MotherDuck? MotherDuck alternatives.