Scripts¶
Every script the repo tracks lives under scripts/. None of them ships in the wheel. Run each one from the repo root.
Docs¶
| Script | What it does | Run |
|---|---|---|
scripts/check_heading_case.py |
Fails on a docs heading that isn't sentence case. The heading-case pre-commit hook runs it on each changed page. |
python scripts/check_heading_case.py docs/**/*.md |
scripts/logo/generate_wordmark.py |
Draws the nexus word mark in the Michroma font, docs/assets/logo.svg and logo-light.svg. |
uv run --with fonttools python scripts/logo/generate_wordmark.py |
scripts/logo/generate_subtitle.py |
Draws the tagline under the word mark, docs/assets/subtitle.svg and subtitle-light.svg. |
uv run --with fonttools python scripts/logo/generate_subtitle.py |
scripts/logo/generate_lockup.py |
Stacks the word mark over the tagline, docs/assets/lockup.svg and lockup-light.svg, and sets the dark one on the home page photo, lockup-banner.svg. Run it after the word mark and the tagline. |
uv run python scripts/logo/generate_lockup.py |
scripts/logo/generate_favicon.py |
Draws the browser-tab icon, docs/assets/favicon.ico. |
uv run --with pillow python scripts/logo/generate_favicon.py |
scripts/logo/README.md says how to change the logo. The social card layout, scripts/logo/nexus-card.yml, is the folder mkdocs.yml names in cards_layout_dir.
Assets¶
Converting scenes covers obj_to_usd.py, site_scan_splat.py, spawn_site.py, author_scene_origin.py and prepare_asset_upload.py. The scripts below build the assets the catalog ships. Each writes a file that prepare_asset_upload.py then hashes for upload, see Bringing your own assets.
| Script | What it does | Run |
|---|---|---|
scripts/assets/author_powerline_scene.py |
Builds the powerline scene from Poly Haven assets. |
uv run python scripts/assets/author_powerline_scene.py /tmp/powerline.usdz |
scripts/assets/author_slalom_scene.py |
Builds the slalom scene: three pillars the sampling Model Predictive Control (MPC) example steers around. |
uv run python scripts/assets/author_slalom_scene.py /tmp/slalom.usdz |
scripts/assets/author_cesium_scene.py |
Builds the cesium scene: the Cesium globe with no token. It runs in the Kit image, which it starts itself. |
uv run python scripts/assets/author_cesium_scene.py --out assets/local/cesium.usd |
scripts/assets/convert_urdf.py |
Converts the Astro Max Unified Robot Description Format (URDF) file to the Universal Scene Description (USD) file the RL example spawns. It needs the full Isaac Sim, so it runs inside an Isaac Lab image. | See below. |
To run convert_urdf.py, set URDF to the input file and OUT to the output folder. USD_NAME sets the filename and defaults to the input's stem with .usda. The public Isaac Lab image needs no login:
docker run --rm --gpus all -v "$PWD":/work -w /work \
-e ACCEPT_EULA=Y -e PRIVACY_CONSENT=Y \
nvcr.io/nvidia/isaac-lab:3.0.0-beta2 \
bash -lc 'URDF=/work/astro_max.urdf OUT=/work/out \
/isaac-sim/python.sh /work/scripts/assets/convert_urdf.py'
CI¶
The workflows under .github/workflows/ run these. Each also runs by hand with the command shown.
| Script | What it does | Run |
|---|---|---|
scripts/ci/evaluate_examples.py |
Flies the examples and gates each against scripts/ci/examples_baselines.json, see Benchmarking. |
uv run --group ci --extra examples python scripts/ci/evaluate_examples.py --only pid |
scripts/ci/merge_eval_parts.py |
Joins the results the gpu-examples boxes upload, one example each, into one folder. |
python3 scripts/ci/merge_eval_parts.py <parts dir> <out dir> |
scripts/ci/peer_legs.py |
Plans the peer legs of gpu-peers: the peers a pull request's diff touches, read from the peers tree, and what each one's box runs. |
python3 scripts/ci/peer_legs.py --all |
scripts/ci/vale_added.py |
Runs Vale on the lines the branch adds against origin/main, the check vale-changed makes, and fails on any alert among them. The vale-added pre-commit hook runs it. |
python3 scripts/ci/vale_added.py |
scripts/ci/benchmark_cell.py |
Flies one cell of the benchmark matrix: one vehicle, scene, and device. | uv run python scripts/ci/benchmark_cell.py --vehicle astro_max_base --scene empty --device cpu |
scripts/ci/benchmark_matrix.py |
Flies every cell of the benchmark matrix and writes docs/data/rtf_matrix.json. In CI, --list, --cell and --merge split the run across boxes. |
uv run python scripts/ci/benchmark_matrix.py --only physics |
scripts/ci/run_rl_example.sh |
Trains, records and gates the Astro Max RL example, see Astro Max RL. | bash scripts/ci/run_rl_example.sh |
scripts/ci/check_rl_stats.py |
Gates a training run's statistics against nexus-rl/stats_baseline.json. |
python scripts/ci/check_rl_stats.py --train-stats train.json |
scripts/ci/px4_image_cache.sh |
Loads the PX4 Software In The Loop (SITL) image from a GPU box's cache before a run, and saves it after. | bash scripts/ci/px4_image_cache.sh load |
scripts/ci/discover_aws_runner_config.py |
Lists the Amazon Web Services (AWS) zones, images and subnets where a GPU box can start, for the start step to try in order. Every GPU run calls it through gpu-runner.yml. By hand, it helps to debug a box that won't start, and it needs AWS credentials. |
AWS_REGION_CANDIDATES="us-west-2 us-east-1" AWS_INSTANCE_TYPE=g5.2xlarge python3 scripts/ci/discover_aws_runner_config.py |