Zero‑copy analytics, delivered at Mach Arrow.
A great engine shouldn't be gated by heavyweight infra. Hatch keeps DuckDB's magic small, open, and composable so your data can fly wherever you need it.
Please note that the AUR package is community-maintained (AlphaJack), and I will not be able to support issues that arise from that setup.
For a stable implementation, see GizmoSQL: 👉 https://github.com/gizmodata/gizmosql
Please avoid submitting Hatch to package managers, registries, or third-party distributions until it reaches a stable release. Premature packaging may cause confusion or misrepresent the project's status.
Flight SQL moves columnar data faster than REST or JDBC, with schemas baked in. DuckDB already "speaks Arrow" internally. Hatch lets it broadcast.
Options today: embed DuckDB yourself, bolt it onto Python/Java web servers, or go proprietary.
Hatch offers a third way: a small server that does one thing: serve SQL over Flight.
Modern data stacks are Lego bricks: Redis for cache, NATS for events, DuckDB for OLAP. Hatch slots into that ecosystem—just stream Arrow in, stream Arrow out.
Want column-level ACLs? Write a middleware.
Need OpenTelemetry spans? Drop in an interceptor.
Curious about WASM UDFs? Fork and go wild.
Hatch is scaffolding, not a silo.
Released under the MIT License. See LICENSE for details.