ChartBook

Turn a data pipeline into a documented, searchable site, and let other pipelines build on what it made.

A pipeline is a folder of code in git: a manifest, scripts, write-ups, and the data, charts and models the scripts produce. ChartBook reads that folder, records what is in it, and writes a site. Many pipelines make a catalog: one site over all of them, where any dataset can be loaded by name, and where one pipeline can pin exactly what another one built.

Three steps, each a transformation of the one before. The index in the middle is the point of the design: one JSON file that says what the site shows, with an ID and a content hash for everything, so that pages can be written for any site tool, two builds can be compared, and another program can read what you published without scraping it. Vocabulary defines these words, and every page here uses them in those senses.

Start

Install and build

pip install chartbook, scaffold a project with chartbook init, and have a site in a minute.

Getting Started

Document a pipeline

Describe charts, dataframes, notebooks, notes and models in chartbook.toml; each gets a page.

Documenting a Pipeline

Depend on another pipeline

Snapshots, pins and the lockfile: read exactly what another pipeline built, and nothing else.

Depending on Another Pipeline

The four documents

Four files, one record format, and none is another in disguise. A manifest asks, a lockfile records what was got, a snapshot records what was made, and an index records what a site shows.

From zero to a documented pipeline

pip install chartbook
chartbook init            # scaffold a pipeline project
cd <your-project>
chartbook docs generate   # write the site, and index.json beside it
chartbook docs open       # open it

from chartbook import data

df = data.load(pipeline="fred_charts", dataframe="fred", format="pandas")   # any cataloged dataset
rates = data.load("fed_yield_curve:treasury_rates")   # a dependency's dataframe: the pinned copy, by name

Where things are

User guide

• Getting Started: install, scaffold, first build

• Vocabulary: the words, defined once

• Documenting a Pipeline: the manifest and the pages it makes

• Depending on Another Pipeline: snapshots, pins, the lockfile

• Keeping Costly Outputs: keep = true, chartbook restore, and the stop before a build remakes it

• Catalogs and Data Access: one site over many pipelines; data.load()

• Plotting: one call, five chart files

• How the Build Works: the index, the writer, Sphinx

Reference

• Configuration Guide: every chartbook.toml field

• CLI Reference: every command

• API Reference: the Python modules

• Stability: what “experimental” means here

• Design: start here: the design, decision by decision

Working with an AI assistant? The built site ships llms.txt files at its root, and chartbook install claude-skill installs a ChartBook skill for Claude Code, which says when it is for another ChartBook.