dbt Labs Open-Sources Declarative Charting Language 'dbt Charts' for AI Agents

dbt Labs has open-sourced dbt Charts, a declarative language that allows users to generate dashboards by conversing with AI agents while addressing the governance issues found in traditional BI tools. Previous AI data analysis approaches suffered from difficult auditing and increased token costs upon changes, as agents generated multiple files such as HTML/CSS/JS.

Declarative YAML-Based Architecture

dbt Charts decouples chart logic from BI tools by moving it into code (YAML). It declares entire interactive dashboards in a single auditable YAML file, defining data (WHAT) with SQL and visualization methods (HOW) with YAML. It currently supports 16 chart types and over 1,100 configuration options, and can render to static SVG, HTML, PNG, PDF, and more via the CLI (dct).

Integration with AI Agents and Development Workflows

The tool was designed with the premise that AI agents are proficient in code and SQL. It provides strict validation for YAML and SQL along with visualization checks (e.g., BAR-BAND-WIDTH-TOO-NARROW warnings) to strengthen the agent's feedback loop. Additionally, by placing a charts/ directory in the same Git repository as the dbt project's models/, model and chart changes can be managed within a single CI pipeline, preventing errors at the PR stage.

Platform and Licensing

The language itself is released under the Apache 2.0 license and can be used locally without an account. Platform features such as hosting, access control, and UI are provided as a separate service (dbtCharts.com), supporting data warehouse connections and permission-based sharing. Currently in pre-1.0 version, even if syntax changes occur, boards are migrated to maintain compatibility.