Rocky: the typed graph between your code and your warehouse

What Rocky is

Rocky is a typed SQL compiler and run engine. You write SQL and TOML. Rocky compiles the project and reports the static problems it can find before anything runs. Then it sends generated SQL to your warehouse.

A clean compile does not prove general SQL validity or correct result values. See what Rocky checks.

Install and run

curl -fsSL https://raw.githubusercontent.com/rocky-data/rocky/main/engine/install.sh | bash
rocky playground my-first-project
cd my-first-project
rocky compile            # type-check
rocky run                # materialize the DAG on local DuckDB

You need no credentials. Other platforms: Installation. The full walk-through is the Quickstart.

See the project

Open the browser UI first. It needs no editor. Add the editor extension when you write models.

Browser UI: rocky serve --uiModels and the DAG, runs, plans waiting for review, and policy decisions. The release binaries include it.

VS Code extensionCompile errors, column types, and lineage as you type, through the rocky lsp language server.

What you get

Need	Command	Page
Typed compile errors with fixes	rocky compile	Compiler
Column-level lineage for PR review	rocky lineage-diff main	Introduction
Isolated branches and run replay	rocky branch create, rocky replay <run_id>	Shadow mode
Per-model cost, as an estimate and not a bill	rocky cost	Introduction
A way out, with no lock-in	rocky emit-sql	No lock-in

Who it is for

Rocky is built first for data platform engineers who run production-critical, multi-tenant pipelines on Databricks. That is where it has the most production mileage.

Snowflake and BigQuery adapters are Beta. See the Roadmap.

Next

IntroductionWhy Rocky exists, what it checks, and how far each part has shipped.

Import a dbt projectrocky import-dbt writes a runnable Rocky repo. You move over one piece at a time.

Using Dagster?dagster-rocky wraps the CLI as a ConfigurableResource.