Tutorials

These guides walk through retarget end to end: install, run, inspect results, tune configs, and scale up to batch jobs. They assume you have cloned the repository and can run commands from the repo root.

If you only need a one-liner to verify the install, start with Quickstart. For heterogeneous sensors, read Introduction first. It walks through the complete native-recording to observation to adaptation flow.

Suggested order

Step	Tutorial	You will
1	Your first retarget	Run the CLI, inspect a .npz result, evaluate quality
2	Python API	Compose and inspect RetargetingExperiment stages
3	Run configs	Drive repeatable experiments from TOML/YAML/JSON
4	Motion input	Load native capture or coherent motion files
5	Scenes and task kinds	Robot-only, object interaction, and climbing workflows
6	Batch and metrics	Process many motions and summarize evaluation reports
7	Export and view	Ship tracking NPZ files and preview results

Live code in the browser

Several tutorials include runnable Python blocks. To execute them in the docs UI:

1. Start MkDocs: uv sync --extra dev then uv run mkdocs serve
2. Start Jupyter: ./scripts/docs-jupyter.sh (second terminal)
3. Flip Live in the page header (left of the search bar)

The hosted site is read-only: no Live switch or Jupyter kernel.

See Interactive playground for troubleshooting.

Reference vs tutorials

Need	Go to
Mental model and engine flow	Architecture
Frames, units, resampling	Coordinate conventions
NPZ fields and provenance	Result schema
Batch manifests, resume, evaluation metrics	Batch and evaluation
Exporters, viewers, tracking NPZ layout	Export and visualization
Asset manifests and stores	Assets
Add robots, formats, objectives	Extending
API signatures	API