Charts built for Chat
Short summary
dbt Labs open-sourced dbt Charts, a YAML-based language for building dashboards as code instead of through UI clicks. It's designed to work seamlessly with AI agents and integrates directly with dbt projects in Git.
- Charts declared in YAML can be validated, version-controlled, and changed alongside data models in a single Git branch, catching breaking changes before production.
- AI agents struggle with BI tool UIs but work fluently with code and SQL, so moving charts into YAML gives them a tighter feedback loop with strict validation and visualization warnings.
- The open-source language runs locally and in CI/CD pipelines, while dbtCharts.com adds hosting, access control, and conversational analytics on top without locking users into the platform.
Full summary
dbt Labs just open-sourced dbtCharts, a YAML-based language for declaring dashboards that sits between the messy freedom of writing custom code and the rigid constraints of traditional BI tools. The core insight is that AI agents work fluently in code and Git but stumble in UI-driven tools, so charts need to live where agents operate—in version-controlled files. A single YAML file declares an entire interactive dashboard: SQL queries define *what* data to show, while YAML and Jinja handle *how* to display it. The language supports 16 chart types with over 1,100 config options, but most users rarely touch them directly because styles cascade from themes and boards can inherit from each other, keeping files short and readable.
The real power emerges when you use dbtCharts inside a dbt project. Charts live in the same Git repo as your data models, so a model change and its corresponding chart update ship together in one branch and one CI run. If a renamed model or missing column breaks a chart, the validation fails before hitting production. This tight integration means validation happens at parse time—`dbt parse && dct validate charts/`—catching problems early. The platform also plans support for the dbt Semantic Layer, letting charts reference metrics as defined centrally rather than restating their SQL.
Alongside the open-source language, dbt Labs launched dbtCharts.com, a hosted platform in public beta that handles hosting, access control, and conversational analytics on top of the open standard. The platform includes a visual editor for tweaking charts, version history, permissions-based sharing, and a Claude-like chat interface with read-only warehouse access. Because everything syncs to the same YAML in your Git repo, changes from chat, the editor, or direct code all land in the same source of truth—no hidden data stacks, no lock-in. The language renders to SVG, HTML, PNG, PDF, and even the terminal, and you can install it locally with one command to start building without an account.