dbt Labs open-sources dbt Charts to move BI dashboards into code

dbt Labs has open-sourced dbt Charts, a declarative language designed to move interactive dashboard definitions out of proprietary interfaces and into version-controlled code. The move aims to streamline how data teams and AI agents build and audit visualizations.

Released under the Apache 2.0 license, dbt Charts allows users to define dashboards using a combination of SQL for data retrieval, YAML for visualization configuration, Markdown for prose, and Jinja for variables and macros. The tool leverages Vega-Lite for chart assembly but adds a specialized layer for opinionated styling and repeatable SQL queries. Once defined, these charts can be rendered across multiple formats, including SVG, HTML, PNG, PDF, and terminal output.

The shift toward unbundled BI

For years, the "modern data stack" has unbundled the backend of data operations, separating ingestion and transformation into specialized tools. However, visualization has remained largely trapped within UI-first proprietary platforms. This creates a bottleneck for modern workflows, particularly as AI agents are increasingly tasked with generating reports.

Dave Fowler noted that while agents are fluent in SQL, Git, and code, they remain "clumsy in someone else’s UI," arguing that charts must move into the environments where agents actually work. By shifting the definition of a chart from a series of mouse clicks in a GUI to a YAML file in a repository, dbt Labs is attempting to apply the same software engineering rigor to BI that dbt previously applied to data transformation.

Why code-based charting matters

Treating dashboards as code enables the use of standard software development lifecycles, including version control and CI/CD validation. This approach ensures deterministic rendering, solving the problem of fragmented or "slop" artifacts often generated by AI agents using raw HTML or JavaScript.

Because dbt Charts integrates directly with dbt projects, visualizations can reside in the same Git repository as the underlying data models. By utilizing the `ref()` function, the tool ensures strict data lineage and validation, making it easier for human auditors to verify the SQL powering a visual. This positions the dashboard as a governed layer of the data stack rather than a disconnected presentation layer.

What's next for the ecosystem

Alongside the open-source release, dbt Labs is launching a hosted platform at dbtCharts.com in public beta. This platform will provide managed hosting, access control, and conversational analytics capabilities. As the tool moves into beta, the industry will be watching to see if a code-first approach can displace the dominance of traditional, UI-centric BI tools for enterprise reporting.