rigyd

SDK + CLI for Rigyd: compose an asset from intent and optional reference images, or convert an existing 3D model, into a physics-enabled SimReady simulation asset — USD for NVIDIA Isaac Sim and MJCF for MuJoCo.

Zero dependencies. pip install rigyd and go.

CLI

pip install rigyd
rigyd login                                   # stores your rgyd_live_... key

rigyd compose "compact desktop stapler" \
  --task "press the upper arm to staple paper" \
  --asset-class articulated --export all

rigyd convert chair.glb --tris 50000 --export all

rigyd compositions list
rigyd compositions get <composition_id>
rigyd compositions submit <composition_id> --export all
rigyd jobs list
rigyd jobs get <job_id>
rigyd download <job_id> --export mujoco       # re-download any job, 0 credits
rigyd simulate <job_id> --scene drop          # physics demo video, 0 credits
rigyd whoami                                  # user + credit balance

• --export takes a format (usd, mjcf, all) or a simulator alias (isaac → USD, mujoco → MJCF). Default: usd.
• Progress goes to stderr, the result path to stdout, so it composes: blender $(rigyd convert scan.obj --export usd). Add --json for a machine-readable manifest (agents/CI).

Input	Command	Cost
Idea + optional 1 or 4 images	rigyd compose PROMPT --task TASK	prototype price from rigyd pricing
3D file (.glb/.gltf/.fbx/.obj/.stl/.ply/.usd*)	rigyd convert FILE	free

rigyd compose submits to SimReady automatically and downloads the result. Add --review to stop when the prototype is ready; later run rigyd compositions submit <composition_id>.

Python SDK

import rigyd
rigyd.configure()                                  # key from login/env

job = rigyd.convert(file="chair.glb")
job.wait(on_progress=lambda j: print(j.status, j.stage, j.progress))
xml_path = job.download(fmt="mjcf")                # or "usd" / "all"
print(rigyd.account())                             # user + credit balance

composition = rigyd.compose(
    prompt="compact desktop stapler",
    robot_task="press the upper arm to staple paper",
    asset_class="articulated",
    auto_submit=True,
    physical_context={
        "dimensions_m": {"length": 0.15, "width": 0.04, "height": 0.08},
        "materials": ["ABS plastic", "steel"],
        "fixed_base": False,
    },
)
composition.wait()
job = composition.conversion_job()
usd_path = job.download(fmt="usd")

MuJoCo extra

pip install "rigyd[mujoco]"

model = rigyd.load_model(file="chair.glb")  # -> mujoco.MjModel, ready to mj_step

Loading into a live simulator

• NVIDIA Isaac Sim — use the Rigyd SimReady Importer extension (Omniverse Community Registry: rigyd.simready) to generate and load assets onto the stage without leaving the app. This CLI produces the same USD files for offline/scripted pipelines.
• MuJoCo — rigyd.load_model(...) above, or mujoco.MjModel.from_xml_path(<path from rigyd download --export mujoco>).

Configuration

Key resolution order: --api-key flag → RIGYD_API_KEY env → ~/.config/rigyd/config.json (written by rigyd login, mode 600).

License

MIT — see LICENSE.