NeMo Data Designer Plugins

First-class NVIDIA-provided plugins for NeMo Data Designer. This repository is the NVIDIA-maintained curated first-party plugin catalog for Data Designer.

Quick Start

git clone git@github.com:NVIDIA-NeMo/DataDesignerPlugins.git
cd DataDesignerPlugins
make sync

Create a new plugin:

uv run ddp new my-plugin

This generates a column generator by default. Pass --type seed-reader or --type processor to scaffold those plugin types instead. Each scaffold creates a complete plugin skeleton under plugins/data-designer-my-plugin/ with config, implementation, entry point, docs, tests, and CODEOWNERS. See docs/authoring.md for the full authoring guide.

Repository Structure

DataDesignerPlugins/
|-- devtools/
|   `-- ddp/                          # Monorepo management tooling (ddp CLI, dev-only)
|-- catalog/                          # Machine-consumable plugin catalog data
|-- plugins/                          # One directory per plugin (auto-discovered by uv)
|   `-- data-designer-template/       # Reference implementation
`-- docs/                             # Zensical documentation source

Each plugin is an independent Python package with its own pyproject.toml, docs, tests, and CODEOWNERS. The root workspace auto-discovers plugins via plugins/*.

Plugin Catalog

The default NVIDIA plugin catalog URL is:

https://nvidia-nemo.github.io/DataDesignerPlugins/catalog/plugins.json

catalog/plugins.json is the checked-in JSON catalog registry for released Data Designer plugin packages. The published Pages site serves that catalog together with the static Python package index at https://nvidia-nemo.github.io/DataDesignerPlugins/simple/. See docs/catalogs.md for catalog discovery, catalog fields, install metadata, trust expectations, and external catalog setup guidance.

Development

Use the repo's Makefile targets for all development tasks:

make sync               # Install all packages (uv sync --all-packages)
make lint               # Lint and format check (ruff)
make format             # Auto-fix lint issues and reformat
make test               # Test each plugin in an isolated venv
make validate           # Run assert_valid_plugin on all entry points
make check              # Verify generated plugin docs, catalog validity, CODEOWNERS, and license headers
make plugin-docs        # Regenerate docs/plugins/ from per-plugin docs and metadata
make catalog PLUGIN=data-designer-my-plugin  # Register a package for its first release
make package-index      # Add catalog JSON and the static package index to site/
make qa-package-index   # Build/install a plugin through a scratch local package index
make docs               # Build the Zensical documentation site
make docs-server        # Serve docs locally at http://localhost:8000
make all                # lint + test + validate + check + docs (full local CI)

To test a single plugin in isolation:

make test-plugin PLUGIN=data-designer-my-plugin

If you change plugin docs or ownership, regenerate derived files:

make plugin-docs              # Regenerate plugin documentation site inputs
make codeowners               # Regenerate CODEOWNERS
make update-license-headers   # Fix SPDX headers

Register a package in the catalog only when preparing that package's first release:

make catalog PLUGIN=data-designer-my-plugin

The ddp CLI

The ddp command manages the monorepo. Run uv run ddp --help to see all subcommands:

Command	Description
ddp new <name>	Scaffold a new plugin
ddp catalog register <plugin>	Register one package for first release catalog discoverability
ddp catalog check	Validate the checked-in catalog JSON
ddp validate	Validate all installed plugins
ddp plugin-docs	Generate plugin docs site inputs
ddp package-index	Build, validate, merge, and QA the static package index
ddp codeowners	Aggregate CODEOWNERS to stdout
ddp license-headers	Add or check SPDX license headers
ddp bump <plugin> <part>	Bump a plugin's semantic version
ddp check-release <plugin> <version>	Validate plugin metadata for release

Releasing

make bump PLUGIN=data-designer-my-plugin PART=patch   # Bump version (major/minor/patch)
git add plugins/data-designer-my-plugin/pyproject.toml
git commit -m "chore(data-designer-my-plugin): bump version to 0.1.1"
make release PLUGIN=data-designer-my-plugin PUBLISH=1  # Build, tag, push, and publish the GitHub Release

Before a package's first release, register it once with make catalog PLUGIN=data-designer-my-plugin and commit catalog/plugins.json.

See docs/releasing.md for the full release guide.

License

Apache-2.0. See LICENSE.