Validating an AIF Asset#

This guide shows how to validate an NVIDIA AI Factory (AIF) data center asset (CDU, CRAH, UPS, or Compute Rack) against the AIF-Entity profile.

If you only want to confirm your setup works, the minimal Foo sample in the top-level README is the quickest check. This page is the AIF path: use it once you have a real data center asset to validate.

Prerequisites#

Python 3.10 to 3.12 (3.12 recommended).
The simready-validate toolchain installed in an active virtual environment. If you already completed Environment setup in the top-level README, you have this; just activate that virtual environment and continue. If not, create and activate a virtual environment (see the README) and, from the repository root, run:
```
pip install -r nv_core/validator_sample/requirements.txt
```
This pulls simready-validate, omniverse-usd-profiles (which ships the omni.capabilities module the AIF validators import), and omniverse-asset-validator. No local build/codegen step is required to run validation.
An AIF asset to validate. See the AIF-Entity authoring guide for the required stage composition and the per-equipment-class attribute sets.

The asset class drives validation#

A single AIF-Entity profile covers all four equipment classes. The aif:core:assetClass property on the asset selects which expectations apply:

`aif:core:assetClass`	Thermal	Electrical
`CDU`	Yes	Yes
`CRAH`	Yes	Yes
`UPS`	No	Yes
`Rack`	No	No

You do not pick a different profile per class; you set aif:core:assetClass on the asset and validate against AIF-Entity.

Running validation#

From the repository root:

simready-validate \
  --rules-path    nv_core/sr_specs/docs/capabilities \
  --features-path nv_core/sr_specs/docs/features \
  --profiles-path nv_core/sr_specs/docs/profiles/profiles.toml \
  --profile AIF-Entity --version 0.1.0 \
  -v path/to/your_asset.usd

The aif:core:assetClass property on the asset selects which expectations apply (CDU, CRAH, UPS, or Rack), so the same AIF-Entity invocation works for every class. Write results to a file for programmatic inspection with --output results.json.

The asset path can point anywhere on disk. It does not need to live inside this repository. Pass an absolute or relative path to any .usd/.usda and the validator will resolve and check it. Note that the asset must be self-contained: any references it makes to external files (textures, sublayers, payloads) must resolve, or those references will be reported as failures.

Stamping the result into the asset#

Add --stamp-asset-validation to record the outcome in the asset’s customLayerData under SimReady_Metadata (see the README for the stamp schema). This gives downstream tools a portable record of what passed, against which profile, and when.

Expected results per class#

All four pass the AIF-Entity checks with zero failing requirements.

`assetClass`	Expected	Notes
`CDU`	Pass	`Generic_CDU` fixture
`CRAH`	Pass	`CW375` fixture
`UPS`	Pass	`EXLS1` fixture
`Rack`	Pass	`gb300` fixture. Large geometry; full validation run takes ~20 min

Where this fits#

This repository is the spec and validation rules. To author an AIF asset from CAD (ingestion, optimization, metadata), use the AIF Pipeline Samples, then validate the output here against AIF-Entity.