ChartBook

Catalog and publish your team’s data science work.

ChartBook turns an analytics project — code, parquet files, chart HTML — into a documented website, and aggregates many projects into a searchable catalog with programmatic data access. If you know Backstage, Spotify’s service catalog for software teams: ChartBook plays that role for data science teams, so pipelines, datasets, and charts are registered and discoverable instead of scattered across notebooks and shared drives.

Everything hangs off two kinds of chartbook.toml:

• A pipeline is any project directory whose chartbook.toml describes its charts, dataframes, and notebooks. chartbook build renders it into a static documentation site.

• A catalog is a chartbook.toml that registers many pipelines. It builds one combined, searchable site — and lets anyone load any cataloged dataset by name:

from chartbook import data

df = data.load(pipeline="fred_charts", dataframe="fred", format="pandas")

From zero to a documented pipeline

pip install chartbook
chartbook init            # scaffold a pipeline project
cd my-pipeline
chartbook build           # generate its documentation site
chartbook browse          # open it in your browser

Getting Started walks through this and ends with your first chart on a page.

Where things are

• Getting Started — install, scaffold, first build

• Documenting a Pipeline — describe your charts, dataframes, and notebooks

• Catalogs and Data Access — aggregate pipelines and load data from anywhere

• Plotting — one call, five output files (interactive HTML plus static PNG/PDF)

• Configuration Guide — every chartbook.toml field, type, and default

• CLI Reference — every command

Working with an AI assistant? The built docs site ships llms.txt files (llms.txt and llms-full.txt at the site root), and chartbook install skill installs a ChartBook skill for Claude Code — user-level by default, or --project to vendor it into a repo.