dbt Charts Open Source Release: Declarative YAML Dashboards for AI‑Powered Analytics

TL;DR

dbt Charts 0.x introduces a declarative YAML language for full‑stack dashboards, enabling AI agents to create, validate, and version‑control charts as code while retaining the governance and auditability of traditional BI tools.

Why a New Dashboard Language?

Answer: Existing AI‑generated dashboards scatter output across HTML, CSS, JavaScript, and framework files, making audits slow and token‑heavy; BI tools keep AI on a narrow UI surface. dbt Charts moves the chart definition into a single, auditable YAML file that agents can read, write, and validate.

Agents excel at SQL, Git, and YAML, but struggle with UI‑centric tools.
A single YAML board captures variables, queries, chart specs, and layout in a self‑contained artifact.
Validation runs locally (dct validate) and flags visual problems before any UI is rendered.

Core Design: Unbundling the BI Stack

Answer: The modern data stack has already split ingestion, transformation, and compute from the BI layer; dbt Charts completes the unbundling by extracting the visualization layer into code.

Historical context – Traditional BI tools (MicroStrategy, Cognos) bundled ingestion → transformation → visualization. Since ~2015 the stack has been split, leaving BI tools with only visualization, UI, semantics, and hosting.
Shift to chat – As conversational agents become the primary front‑end, the UI‑first approach becomes a bottleneck. Moving charts to code aligns with agents’ strengths.

The YAML Specification

Answer: A dbt Charts board is a single YAML file that declares variables, SQL queries, chart types, and layout rows.

variables:
  status:
    column: main.documents.status

queries:
  doc_growth: |
    SELECT DATE_TRUNC('month', created_at) AS month,
           SUM(COUNT(*)) OVER (ORDER BY month) AS num_docs
    FROM main.documents
    WHERE {{ filter('status', status) }}
    GROUP BY 1

charts:
  growth:
    title: Documents created, all time
    type: area
    query: doc_growth
    x: month
    y: num_docs

rows:
  - growth

SQL for data – The queries block holds raw SQL, preserving dbt’s expressive power.
Markdown & Jinja – Prose can be added via Markdown; Jinja enables macros and variable interpolation, mirroring dbt’s templating.
Extensibility – Over 1,100 configuration options across 16 chart types and composable charts are available, yet most defaults are inherited via themes and board extensions.

Rendering & CI Integration

Answer: The dct CLI renders any board to SVG, HTML, PNG, PDF, or even terminal output, and can serve a static site for sharing.

# Render to SVG (or html, png, pdf, terminal)
dct render charts/documents.yml --format svg
# Serve a folder of rendered charts
 dct serve

Boards live alongside dbt models (charts/ next to models/) so a single PR can modify both transformation and visualization.
dbt parse && dct validate charts/ catches broken ref() calls or visual warnings before deployment.
Future support for the dbt Semantic Layer will allow charts to reference metrics instead of raw SQL.

Governance, Auditing, and AI Feedback Loops

Answer: dbt Charts provides strict validation of both YAML syntax and SQL semantics, plus visual warnings that surface before a chart is published.

$ dct render charts/revenue.yml
WARN-BAR-BAND-WIDTH-TOO-NARROW
182 bands x 2 series across 640px
Fix: roll up to a coarser grain.

WARN-TABLE-COLUMNS-OVERFLOW
Table needs 980px but only 640px is available.
Fix: drop columns or widen the slot.

The deterministic output (same YAML → same chart) enables reproducible CI pipelines.
Because the spec is version‑controlled, any change is traceable to a commit, satisfying audit requirements.

Community and Competitive Landscape

Answer: The project has sparked discussion about alternatives and positioning.

Malloy – An alternative DSL (Malloyyo/Publisher) also targets AI‑generated charts, but remains a separate ecosystem. Commenter @mrtimo notes Malloy’s free‑to‑use hosting versus dbt Charts’ optional SaaS layer.
Observable Framework – @bbkane asks how dbt Charts differs; the key distinction is a YAML‑first, dbt‑integrated spec versus Observable’s Markdown+JS approach.
Vega‑Lite – @engrav3er40 wonders about overlap; Vega‑Lite is a visual grammar, while dbt Charts couples that grammar with data‑model governance and CI integration.
Evidence, Superset, and other headless BI – Several commenters (e.g., @theodorewiles, @hackandthink) point out similar YAML‑based dashboard definitions, but dbt Charts uniquely nests within the dbt transformation layer and offers first‑class chat integration.
Design focus – The project enlisted data‑graphic designer RJ Andrews to create a cohesive visual language, addressing criticism that many code‑driven dashboards suffer from “visual noise”.

Hosted Platform (dbtCharts.com)

Answer: While the language is open source, dbtCharts.com provides a public‑beta SaaS layer for hosting, access control, and conversational analytics.

Connects to a warehouse, adds a chat interface (Claude/ChatGPT‑style) with read‑only permissions, and offers a visual editor for fine‑tuning.
All edits—whether from chat, UI, or code—commit back to the same YAML file in Git, preserving the single source of truth.
The platform handles the only remaining BI responsibilities that are hard to unbundle: hosting, authentication, and collaborative sharing.

Reception and Open Questions

Answer: Early community feedback highlights both enthusiasm and skepticism.

Positive reception – Users appreciate the deterministic, auditable artifacts (@mollifier14, @marcy_74) and see practical use cases for daily KPI updates in Slack/Teams.
Critiques – Some argue the innovation is incremental (@dgudkov) and question the choice of YAML as a “language” (@carterschonwald, @pgt). Others note that visual interactivity and semantic layers remain open problems (@RobGoretsky, @chupchap).
Future work – Planned support for the dbt Semantic Layer, richer interaction primitives, and broader UI integrations aim to address these concerns.

Getting Started

Answer: To try dbt Charts locally, install the CLI and render a sample board.

# Install via uv (or pip)
uv tool install dbt-charts
# Render a board to SVG
dct render charts/documents.yml --format svg

Repository: https://github.com/dbt-labs/dbt-charts (Apache 2.0)
Documentation: https://docs.dbtcharts.com/
Community Slack: #dbt-charts on the dbt Community Slack

Bottom Line

Answer: dbt Charts fills a gap between free‑form code generation and locked‑down BI UIs by providing a version‑controlled, YAML‑based dashboard language that integrates tightly with dbt, supports AI‑driven authoring, and enforces governance through CI‑ready validation.