embodiment-card validates and renders structured robot embodiment metadata for
dataset and vision-language-action release review.
| Job | Make robot hardware, sensing, action, frame, control, environment, and limitation assumptions reviewable. |
| Built for | Robotics dataset curators, VLA researchers, release reviewers, and documentation owners. |
| Differentiator | Dependency-free schema validation plus Markdown and Hugging Face README rendering. |
| Produces | A validated JSON card or a human-readable embodiment card for release documentation. |
python -m pip install "embodiment-card==0.1.2"Run from a source checkout:
embodiment-card validate examples/mock_so101_card.json
embodiment-card render examples/mock_so101_card.json \
--hf-readme \
--out /tmp/SO101-EMBODIMENT.mdThe bundled SO-101 fixture validates and renders Hugging Face-compatible YAML tags followed by an embodiment card.
The v1 card requires:
- name and robot morphology;
- at least one sensor;
- action-space metadata;
- at least one coordinate frame;
- positive control rate;
- teleoperation method;
- environment assumptions;
- safety boundaries and known limitations.
Validation checks required fields and a small set of structural constraints. It does not verify hardware, calibration, units, coordinate transforms, or the truth of a disclosure.
- The CLI reads one local JSON file and writes Markdown to a file or stdout.
- It does not inspect robot media, connect to hardware, contact Hugging Face, or make network requests.
--hf-readmeformats local text only; it does not create or update a Hub repository.- No redaction is applied. Review cards for private paths, operators, sites, or hardware identifiers before sharing them.
- Validation covers required fields and simple structure only. It does not check calibration, coordinate transforms, units, or hardware truth.
- Rendered cards remain self-reported documentation, not a safety review, hardware certification, or deployment approval.
robostudio-engine exposes a direct robostudio card integration that imports
embodiment-card when the package is installed or available in the AuraOne
monorepo. The standalone card format can also be embedded in any dataset or
model README.
Verified on 2026-07-13:
- PyPI:
embodiment-card==0.1.2 - GitHub release:
v0.1.2 - Bundled cards describe mock tutorial datasets and do not establish hardware validation or safety assurance.
Fill a card from one real dataset's reviewed metadata, have the hardware owner verify frames and control rate, then publish the rendered disclosure beside the dataset.