Kit-CAE: Reference Application

Kit-CAE is a reference application demonstrating technologies for CAE and scientific data workflows. It shows how OpenUSD can serve as a composition and access layer over engineering data from solvers, AI surrogates, sensors, and more, while keeping source files in their native format. No conversion or copying is required, so data provenance and fidelity are preserved.

The key technologies integrated in Kit-CAE include:

• OpenUSD Schemas & File-Format Plugins: A USD-native foundation for composing and lazily accessing scientific datasets without conversion. Reference implementations are provided for CGNS, EnSight, VTK, OpenFOAM, and others, but the same approach extends to any format.
• NVIDIA Warp: GPU-accelerated data processing and visualization algorithms for interactive exploration of large datasets.
• RTX & IndeX Rendering: High-fidelity visualization of engineering data, including surfaces, volumes, and particles.
• Kit Application Framework: Extensible UI, pixel streaming, and integration with the broader set of Omniverse libraries.

Each technology can be used independently. The repository is structured to make that separation clear. Developers are encouraged to use Kit-CAE as a testbed for integrating their own tools with any of the enablement technologies provided.

Kit-CAE builds on two companion projects:

• CAE OpenUSD Plugins provides the OpenUSD schemas and file-format plugins used to compose and lazily access scientific datasets in their native formats.
• Warp SimData provides the common scientific-data model and Warp-accelerated processing algorithms used by the visualization operators.

Quick Start

Build

# On Linux
./repo.sh build -r

# On Windows
repo.bat build -r

USD schema source code is generated during fetch and compiled automatically during the build process.

See Build Instructions for details on prerequisites, optional dependencies, and Kit SDK version selection.

Note: Native VTK and other file-format support is included in the default build. Optional pip packages are needed only by legacy data delegates. See Optional Dependencies.

Launch

# Basic application
./repo.sh launch -n omni.cae.kit

# Legacy VTK Python compatibility environment
./repo.sh launch -n omni.cae_vtk.kit

See Launch Instructions for running sample scripts and detailed launch options.

Documentation

Getting Started

• Build Instructions - Prerequisites, building, and dependencies
• Launch Instructions - Launching the app and running samples
• User Guide - Step-by-step feature walkthroughs
• Reference Guide - Assortment of notes on some topics for power users and developers.

Architecture & Design

• Scientific Data Schemas - Active OmniSci data model and legacy CAE schema boundary
• CAE Viz Schemas - Visualization operator schemas (omni.cae.viz)
• Extensions Overview - Omniverse extensions architecture
• Legacy Data Delegate API - Compatibility reference for retired CaeFieldArray workflows
• Format Onboarding Tutorial - End-to-end custom format integration walkthrough

Advanced Topics

• Integration Guide - Combining with Kit Application Template apps
• Kit SDK Version Selection - Managing Kit SDK versions

License

Development using the Omniverse Kit SDK is subject to the Omniverse licensing terms.

This project uses several open-source libraries:

1. CGNS - CFD Notation System
2. h5py - Python interface for HDF5
3. HDF5 - High performance data software library
4. VTK - Visualization Toolkit
5. Zlib - General purpose data compression library
6. pyparsing - Python Parsing
7. nvtx - NVTX Code Annotation
8. lz4 - LZ4 compression library
9. trimesh - Trimesh library

Kit-CAE also uses CAE OpenUSD Plugins and Warp SimData. Refer to those repositories for the licenses and notices covering their additional third-party dependencies.

Review the license terms of these open source projects and companion-project dependencies before use.

Agent Skills

Kit-CAE ships with agent skills in skills/ that let agents (Codex, Claude Code, OpenClaw, Cursor, etc.) drive Kit-CAE programmatically: importing data, setting up visualizations, and capturing renders without manual intervention.

Onboarding

1. Point your agent’s skill/tool search path at /skills/.
2. The agent reads the relevant SKILL.md on demand — no further configuration needed.

Skills reference each other via relative paths (for example, cae-core/references/kit-cae-api.md), so the skills/ directory is self-contained and portable.

Onboarding Prompt

Paste this into Codex, Claude Code, OpenClaw, Cursor, or any skills-aware agent after pointing it at this repo:

This repository ships agent skills under `skills/` (cae-core,
cae-visualization, cae-capture, cae-streaming). Read each SKILL.md, run any
preflight or environment checks the skills prescribe, and confirm Kit-CAE is
ready to use on this system. Then give me a short summary of what each skill
enables so I know what to ask you to do next.

Contributing

We provide this source code as-is and are currently not accepting outside contributions.