One engine takes both workloads

Online DDL

Schema changes run against live writers

A schema epoch in the tuple slot lets ALTER TABLE add, drop, or retype a column without rewriting a row. Index builds publish, wait for in-flight writers, scan, and flip without blocking the table, and an incompatible type change runs as a shadow rewrite behind live traffic.

On a Raft group, DDL runs on the connection so the client sees the real command tag, and the applier picks it up if the client drops. Every statement in the grammar replicates the same way, covered on the consensus page. The catalog rows behind a schema migrate at upgrade time through a schema evolution registry, on the upgrades page.

Legibility

Every statement has help, every error has a code

One walk of the grammar writes both the reference pages and a dataset embedded in the binary, so the help a running server gives and the reference pages come from the same source.

Deprecations get the same treatment, with time-bounded windows, rate-limited warnings, and auto-generated migration guides, covered on the upgrades page.

SQL surface

psql connects unchanged

-- Read a table as it was at a point in time
SELECT * FROM orders AS OF TIMESTAMP '2026-05-06 09:00:00';

-- Branch the database, experiment in isolation, then merge or drop
CREATE BRANCH experiment FROM main;
USE BRANCH experiment;

-- Native vector search
CREATE VECTOR INDEX ON docs (embedding) WITH (metric = 'cosine');
SELECT id FROM docs ORDER BY embedding <=> $1 LIMIT 10;