Examples

The repository ships a complete, buildable demonstration in examples/: three pipelines aggregated by a catalog.

Project	What it demonstrates
fred_charts	A full pipeline: one dataframe pulled from FRED, six charts (including a dual-axis and a scatter), and two notebooks. Mixes docs_path write-up files with inline docs strings.
yield_curve	A larger pipeline: three dataframes, five charts, four notebooks and a model. rate_forecast depends on it.
rate_forecast	A pipeline that depends on another: one model, trained on yield_curve’s data. It declares the dependency, pins yield_curve’s snapshot with chartbook pin, and loads the pinned copy with data.load("yield_curve:fed_yield_curve").
catalog	A catalog registering all three pipelines under scoped IDs.

The _data, _output, and built docs/ artifacts are generated rather than committed (they’re gitignored), so each pipeline is built before the catalog aggregates it. examples/dodo.py orchestrates the whole sequence — build each pipeline to produce its parquet files and charts, build the catalog site, then copy it to a root docs/ directory for GitHub Pages:

git clone https://github.com/backofficedev/chartbook
cd chartbook/examples
doit

Individual pipelines have their own inputs (fred_charts and yield_curve pull from FRED), so the chartbook.toml files are also worth reading on their own as complete, real-world manifests.

rate_forecast is built after yield_curve, because it reads what that pipeline produces. It never reads yield_curve’s folder. Between the two builds, examples/dodo.py runs chartbook snapshot in yield_curve, which copies what it built into the store at ~/.chartbook/store/, labelled with the commit. rate_forecast’s first build step is chartbook pin, which pins that snapshot and writes chartbook.lock. Its README walks through what happens when yield_curve is rebuilt afterwards: nothing, until yield_curve makes a new snapshot and rate_forecast runs chartbook update.

A snapshot needs everything in the pipeline’s folder to be committed, so if you edit examples/yield_curve, commit before running doit in examples/.

Each example names its build command, [scripts] build = "doit". The same three steps by hand: chartbook run in examples/yield_curve builds that pipeline, chartbook snapshot there declares the build ready, and chartbook run in examples/rate_forecast pins it and builds rate_forecast. That last command does not build yield_curve: a folder you depend on is its owner’s to build.

The scaffold produced by chartbook init is itself a smaller worked example — see the cookiecutter template.