HN
Today

Charts built for Chat

dbt Labs introduces dbt Charts, an open-source declarative language (YAML + SQL) for building auditable, code-first dashboards. It aims to solve the messiness of AI-generated charts by providing a structured, version-controlled alternative to traditional BI tools. Hacker News appreciates the move towards unbundling BI, the emphasis on AI-agent friendliness, and the creation of maintainable data artifacts.

39
Score
15
Comments
#3
Highest Rank
10h
on Front Page
First Seen
Sep 14, 11:00 PM
Last Seen
Sep 15, 8:00 AM
Rank Over Time
5433344688

The Lowdown

dbt Charts is a new open-source project from dbt Labs, offering a declarative language for creating dashboards using YAML and SQL. The core idea is to enable AI agents to build analytical reports that are auditable, version-controlled, and easily maintainable, circumventing the issues of fragmented, hard-to-trace code typically generated by AI.

  • Addressing AI-Generated Mess: The tool tackles the problem of AI agents producing disparate HTML, CSS, JavaScript, and app files for dashboards, making them difficult to audit and expensive in terms of agent tokens and human time.
  • A Third Path: It positions itself as an alternative to the

The Gossip

Unbundling BI Bliss

Many commenters express excitement and agreement with the

Code Control vs. AI Chaos

A central debate revolves around the necessity of a structured DSL like dbt Charts when AI models are increasingly capable of generating raw code. Some argue that models will eventually be so good at single-shot generation from primitive libraries that DSLs become obsolete. The author (thingsilearned) counters that a domain-specific DSL maintains consistency, lowers maintenance, saves tokens for AI, and crucially, provides human readability and testability, which is vital for verifying data provenance and ensuring trust in AI-generated insights. The point is made that humans still need to verify the underlying SQL.

Visualizing Veracity

One intriguing comment questions the fundamental utility of data visualizations, suggesting they are often just