NVIDIA-Omniverse

Use when setting up a fresh clone of this repo to run OVRTX Dev Variant Presenter (also called the "variant studio" in the walkthrough video and earlier releases) — no .venv yet, ovrtx/ovstage/ovstream not installed, ModuleNotFoundError on…

Use when asked to start, run, launch, restart, or bring up the OVRTX Dev Variant Presenter server, stream, or viewport in this repo — also called the "variant studio" in the walkthrough video and earlier releases — or after the stream / GPU / server was lost…

Build a browser-based, RTX-streamed app for exploring and rendering USD VARIANT PERMUTATIONS — a live ovstream/WebRTC viewport plus instant variant switching, cameras, a turntable rig, grid/batch stills, a multi-track timeline, projects, post, and…

The visual design + layout for the Dev Variant Presenter frontend — the GitHub-dark + NVIDIA-green theme, the header / viewport+timeline / right-side tabbed control panel layout, and the component styles (pill chips with swatches, segmented buttons, blocks,…

Batch-render combinatorial USD variant permutations to disk with an ovrtx renderer: one-at-a-time and full-Cartesian matrix modes, a combinatorial-explosion guard, a {set}-{variant} naming convention, optional animation-range frame sequences, progress events,…

Post-process rendered variant permutations: browse a results folder of stills/videos, burn variant-label plates onto renders (originals untouched), assemble frame sequences into MP4, and compose a labeled contact sheet ("cut sheet"). Pure PIL/OpenCV/ffmpeg,…

Persist a Variant Presenter workspace: named PROJECTS (bundle a stage with its base variant selection, display/optics, selected camera, per-camera looks/framing, and saved timeline views) and an automatic crash-recovery SESSION checkpoint that restores the…

Build a multi-track, non-linear timeline (NLE) for USD variant + camera changes over time: a pure state-resolution engine (tracks/clips, state_at(t), frame sampling, mixer/slideshow presets), a client-side JS mirror for instant scrubbing, and a render-to-MP4…

High-level overview of a typical ovrtx application lifecycle. Use when user asks how to structure an ovrtx program, what the main steps are, or how the pieces fit together.

Asynchronous operation patterns including polling, timeouts, and non-blocking workflows. Use when user asks about async rendering, non-blocking operations, polling, timeouts, or parallel rendering.

Cloning USD subtrees to create copies at new paths. Use when user asks to clone, duplicate, copy a prim, or create instances of existing geometry.

GPU interop patterns for CUDA arrays, timeline semaphores, Vulkan shared memory, and MIG-safe CUDA-to-Vulkan device matching. Use when user asks about CUDA interop, GPU rendering pipelines, Vulkan interop, shared memory, timeline semaphores, MIG, or multi-GPU…

Error checking patterns for both C and Python. Use when user asks about error handling, checking errors, debugging ovrtx failures, or troubleshooting.

Interpret already-read radar PointCloud tensors: channel meanings, units, valid detection ranges, RCS, signed radial velocity, time offsets, flags, counts, and visualization values. Use reading-sensor-pointclouds when the user needs to map or fetch the…

Loading USD scenes into ovstage for rendering from files, URLs, or inline USDA strings. Use when user asks to load a USD scene, compose USD content, add cameras or RenderProducts to an existing USD layer, add referenced content, or create runtime geometry.

Zero-copy attribute map/unmap for direct memory access to ovstage buffers. Use when user asks about zero-copy writes, map attribute, direct memory access, or repeated mapped updates. Use attribute-bindings for repeated writes with caller-owned tensors when a…

Default material-authoring workflow for supported, unqualified USD material tasks. Use with usd-cli when a long-running coding agent assigns visual or non-visual materials, uses libraries, performs iterative visual review, and produces canonical artifacts and…

Own agent-driven physics planning, policy, component decisions, guarded repair, authoring, optional attested VoMP mass properties, simulation-backed validation, refinement, acceptance, and durable artifacts for a USD asset, using usd-cli only for low-level…

Use when the user explicitly requests a fixed-pipeline Content Agent CLI, service client, deployment, benchmark, or maintenance workflow under apps/: material-agent-cli, material-agent-client, physics-agent-cli, physics-agent-client, joint-agent-cli,…

Set up and route NVIDIA Content Agents through the default Agentic Content Workflow from the repository root. Use when the user asks for /quickstart, first-run setup, an unqualified supported content task, or prerequisite checks. Route an explicitly requested…

Author only approved articulation candidates, then validate saved-stage identity and joint-graph readback. Use after agent and required human review are durably bound.

Inspect bounded USD hierarchy, existing articulation, and usd-cli evidence before proposing joints. Use for exact Articulation candidate preparation and targeted reinspection.

Propose bounded revolute or prismatic articulation candidates from exact inspection evidence. Use to reason about endpoints, axes, limits, confidence, and native readiness before review.

Select, reject, or safely edit every articulation candidate through a digest-bound decision patch. Use before human review or accepted-only authoring.

Orchestrate an end-to-end NVIDIA self-managed streaming stack across AWS, Azure, or generic Kubernetes. Use when planning, generating, sequencing, deploying, or validating the full path from CSP or generic Kubernetes base infrastructure through NVCF…

Orchestrate in-place upgrades for an existing NVIDIA self-managed streaming stack across supported cloud providers. Use when the user asks for an upgrade plan, upgrade sequencing, component upgrade coordination, manifest-driven refresh, or an end-to-end…

Deploy NVIDIA Omniverse Content Cache (OVCC) alongside a MEGA simulation stack. Covers evaluation, sizing (NGINX queue-depth model), configuration for GCS/S3/Azure backends, Helm install from NGC (chart ovcontentcache v4.0.4), TLS setup, in-place upgrade from…

Deploy Omniverse Derived Cache (OVDC) 6.0.1 for Kit/write-render-worker shader caching on GKE, AKS, EKS or on-prem Kubernetes. Prepare a Helm deployment script with chart sizing presets and drivercachewrapper/UJITSO client values; optionally verify cold/warm…

Orchestrate the AWS KAS/NVCF EKS foundation by delegating to the component AWS infrastructure skills: aws-tf-conventions, aws-iam-vpc, eks-cluster, eks-nodes, eks-storage, ecr, eks-gpu-operator, eks-validate, and optionally ecr-mirror. Use when setting up,…

Project-specific Terraform rules for the AWS KAS/NVCF EKS foundation. Use when writing or reviewing AWS Terraform for VPC, EKS, node groups, IAM, ECR, state, tags, provider configuration, or before terraform apply.

Mirror every KAS/NVCF manifest image and optional Helm OCI chart from NGC into AWS ECR using generated regsync config.

Generate Terraform for AWS ECR repositories and EKS pull access needed by KAS/NVCF. Use after eks-cluster or eks-nodes and before ecr-mirror; this skill creates repositories only and does not mirror images.

Deploy and develop against NVIDIA Omniverse Storage APIs on any Kubernetes cluster (MicroK8s, EKS, AKS, GKE, bare metal). Two storage adapters: (1) Example Storage Adapter — Python filesystem reference implementation + Discovery for learning and custom…

Use when adding or reviewing a storage backend plugin for ovstorage.

Use when changing or reviewing the built-in authentication and authorization Layer or its TOML policy engine.

Use when Rust changes affect the ovstorage plugin ABI header and the checked-in generated copies must be refreshed.

Use when starting ovstorage-broker locally for development or integration testing with the current Stack and listener-auth configuration.

Run, troubleshoot, or explain ovstorage's app-level integration suite locally. Use when choosing between hermetic and live-cloud targets, exercising the Rust/C++ or Python legs, configuring a credential mapper or pre-exported credentials, or checking that…

Use when checking the Storage API wire contract that ovstorage-plugin-services-client depends on.

Use before opening or merging an ovstorage PR to run the project verification gate and inspect risky diffs.

Use for internal SMLib SmBrep model questions, validity and AssertValid audits, topology code and tests, or SmBrepData/BrepArray translation. Do not use for ordinary SM_API or Python operation usage.

Use when SMLib BRep healing (heal_brep, HealBrep, or healing during USD import) crashes, reports success but leaves gaps, seams or poles unrepaired, changes a model unexpectedly, or works on stale data after topology edits, and when changing the healer's…

Verify a suspected USD BRep or SMLib bug, distinguish API misuse and known limitations, and prepare a minimal reproduction and issue draft using this repository alone. Use when reporting a bug, checking whether unexpected behavior is a bug, or reducing a…

Use for questions, asset compliance checks, validator drift audits, and code or tests involving the proposed OmniSolid USD BrepArray schema.

Choose validation for SMLib changes or CI failures; map changed files to focused build/test commands. Do not use for implementation design.

Debug solid BRep booleans in SMLib, including entrypoints, dispatch, consumption, degeneracies, and `IntersectInsertRelate()`; not mesh booleans.

Use when debugging SMLib 3D curve-to-surface UV drop/projection failures, including drop_curve_to_surface, project_curve_to_surface, SmApiDropCurveToSrf, UV/3D segment pairing, seam/domain splits, and trim-boundary crossings. Do not use for standalone curve…

Use when diagnosing or fixing SMLib fillet, chamfer, blend, or fillet-removal failures involving selected edges/corners, radius feasibility, continuity/cross-section settings, solver or corner handling, and topology insertion. Do not use for generic booleans…

Creating and configuring an ovstream server instance. Use when user asks to create a server, configure ports, set up a streaming endpoint, or pick between server types.

Application-level messaging with the browser client via ovstream_utils.messaging — event types, correlation ids, automatic splitting and reassembly of large payloads, and per-event handlers on top of the raw app-message envelope. Use when user asks how to…

High-level overview of a typical ovstream application lifecycle. Use when user asks how to structure an ovstream program, what the main steps are, or how the pieces fit together.

Wiring connection / message / input callbacks for WebRTC, native, SHM, and CUDASHM streams. Use when user asks about receiving input events, handling messages, detecting client connect/disconnect, or sending messages back.

Composing ovstream with ovrtx, ovstage and ovui into one application — ownership of the scene, frames and CUDA streams, the per-tick order, one loop with two cadences, startup and shutdown order. Use when user asks how to build a streaming app from several OV…

CUDA buffer allocation, pitch alignment, BGRA channel order, and ctypes-bound cudart usage. Use when user asks about CUDA interop, pitched allocations, frame layout, or zero-copy GPU integration.

Writing a CUDASHM reader/consumer that attaches to an ovstream CUDASHM producer over CUDA IPC. Use when user asks about consuming GPU-resident frames, reading from a CUDA IPC ring, writing a simulation/post-processing kernel that ingests rendered output, or…

Error checking patterns for both C and Python plus logging. Use when user asks about error handling, debugging ovstream failures, registering a log callback, or troubleshooting.

Read saved validator artifacts and triage the issues that --fix could not auto-resolve, with per-rule prim lists and fix recommendations. Use after run-validators to decide what to do about remaining issues.

Diff two USD stages by prim/mesh/vertex/material count, file size, and validator summary. Use for before/after optimization comparisons.

Choose and run a ready-made preset operation stack from config_presets/. Use when you want a known-good starting point for a common optimization goal rather than assembling a custom config.

Create a USD proxy mesh sibling. Use to generate decimated, bbox, or LOD stand-ins, with optional render/proxy purpose variant set.

Triage a failing Usd Optimize operation. Use when an op errors, silently no-ops, or returns unexpected output.

Collapse duplicate prim hierarchies into instanceable internal references. Use when deduplicating subtrees or folding repeated prims into prototypes.

Quick USD stage inspection — reports stage info, prim counts, mesh statistics, materials, animation, and scene scale. Use before optimizing.

Add a new performance-validator rule that wraps a Usd Optimize analysis-mode operation. Use when creating a new validator rule class from scratch.

End-to-end ovstage application lifecycle: create an instance, identify prims/attributes via the path dictionary, write attribute columns at an ordinal, advance the write floor to seal them, query/read, and release. Use when user asks how to structure an…

Clone the subtree under a source prim to one or more new target paths in a single ordinal-keyed call — including the multi-environment pattern (stamp out N copies of a prototype). Use when the user asks to clone, duplicate, or copy a prim/subtree, or to spawn…

ovstage's asynchronous, ordinal-keyed submit/observe model: enqueue returns an op id before queued execution completes, and the CPU can run ahead. Covers the current synchronous submission, serialized backend-lane, and overlapping read/write/map limitations.…

How ovstage attribute data crosses the boundary as DLPack DLTensors — copy-in writes, copy-out reads, staging-backed map/unmap writes, CPU vs CUDA residency, cuda_sync GPU sync, and sparsity (index_map/mask). Use when user asks about DLTensor, tensor data…

Error checking patterns for the ovstage C API. Use when user asks about error handling, checking return codes, debugging ovstage failures, or troubleshooting a failed enqueue, wait, or fetch.

Export selected ovstage runtime state to USD from C or Python. Use for saved typed hierarchies, sparse overlays, reusable USD destinations, host-owned layer/stage identifiers, async export completion, or changed-property export. Prefer the metadata-driven…

Populate an ovstage instance from USD via the population API: open a USD file or inline USDA into the stage, choose data domains, drive the async populate op, and read populated prims back by usd-path. Use when the user asks to load/ingest USD into ovstage,…

Using the OVX path dictionary — the shared interning layer for tokens, prim paths, and prim-path lists across OV libraries. Use when user asks about interning strings/paths, ovx_token_t / ovx_primpath_t / ovx_primpath_list_t, or sharing paths with ovstage.

Integrate the OVXR lifecycle with an application-owned OVRTX renderer and OVStage. Use for a minimal XR render program or lifecycle design.

Identify supported OVXR seams for application composition and GPU resources. Use for application composition layers, borrowed Vulkan bindings or custom projection UI.

Choose OVXR context descriptor, runtime, scene metadata and lifecycle settings. Use for context creation failures or live-versus-recreate configuration.

Select OVXR view configuration, foveation policy and capture domain. Use for warped images, eye resolution, feature negotiation or multi-GPU topology.

Drive OVXR convenience or begin, bind and submit frame scheduling. Use for frame-plan ownership, no-render states, timing or cancellation.

Consume OVXR CUDA frame captures without violating borrowed buffer lifetime. Use for acquire-frame, expired pointers, capture correlation or readback domains.

Integrate OpenXR actions, hands or extensions through OVXR borrowed handles. Use for extension enablement, native input timing or shared-session ownership.

Configure a C or C++ application against the OVXR native package. Use for CMake setup, linking, or platform runtime deployment.

Cloud deployment performance — caching architecture, pre-warming strategy, container cold starts, NVCF, OKAS, UCC, GXCache, DDCS. Triggers on "cloud performance", "container startup", "K8s profiling", "cloud caching", "cold start", "NVCF", "GXCache", "OKAS",…

Content preparation, validation, CAD import, materials pipeline, and physics — preventing performance issues at the source. Triggers on "import CAD", "CAD to USD", "validate USD", "asset validator", "usdchecker", "material pipeline", "primitive collider",…

Fabric Scene Delegate (FSD) configuration — the small set of settings partners actually need to consider, with the mergeMaterials default (off by default since Kit 108) context and the deprecated instanceCompactTransforms warning. Triggers on "FSD", "Fabric…

GPU and system memory, threading, async rendering, and texture sizing. Triggers on "GPU memory", "VRAM", "out of GPU memory", "RAM usage", "system memory", "memory leak", "thread count", "8K texture", "DDS", "texture compression", "viewport memory".

Entry point for any Omniverse Kit or OpenUSD performance question. Triggers on "my scene is slow", "low FPS", "scene takes forever to load", "GPU crash", "out of memory", "performance problem", "how do I speed up", "optimize my scene". Classifies the problem…

Profiling Omniverse Kit applications — F8/Profiler Window, Tracy, the activity timeline for load-time analysis, Chrome Tracing, Nsight Systems, and the RTX performance heatmap. Triggers on "how to profile", "where is the bottleneck", "Tracy", "F8", "GPU…

RTX rendering performance — RT 2.0 vs Path Tracing, DLSS, glass, bounces, multi-GPU. Triggers on "render settings", "RTX", "DLSS", "RT 2.0 vs path tracing", "bounces", "samples per pixel", "refraction", "glass performance", "multi-GPU", "improve FPS",…

Bottleneck-driven Scene Optimizer workflow — operations, performance validators, trade-offs, and when not to merge meshes. Triggers on "Scene Optimizer", "optimize scene", "merge meshes", "decimate", "deduplicate", "reduce triangles", "performance validator",…

Create an ovphysx instance, attach an ovstage scene, step simulation, and clean up. Use when starting a new ovphysx project, testing basic integration, or learning the minimal workflow.

Clone one USD environment subtree into many runtime-only PhysX environments for RL-style batched simulation. Use when a caller needs replicated physics environments, clone API ordering, or target transform guidance.

Read ovphysx simulation output with the ovstage-native output-read API when results must retain ovstage identity or write back without rebuilding paths. Covers Python `PhysX.query()` + `PhysX.read()` operations and `PhysX.read_sync()`, C `ovphysx_query()` /…

Push caller-owned bulk control inputs and state into an ovphysx simulation with the session write API — Python `PhysX.write_sync()` or the `PhysX.query()` + `PhysX.write()` operations, C `ovphysx_write()`. Covers writing rigid-body and articulation state and…

Author USD physics content for ovstage to populate and ovphysx to simulate (rigid bodies, colliders, mass, physics scene). Use when creating or editing .usda or USD scenes for ovstage and ovphysx, not when calling the runtime step or tensor API. Keywords USD…

Exchange CPU simulation state — poses, velocities, joint targets — as caller-owned NumPy arrays through tensor bindings. The tensor-binding CODE API is deprecated in ovphysx 0.6 in favor of the session read/write API (Python PhysX.read / PhysX.write, C…

Exchange GPU simulation state as caller-owned CUDA tensors (DLPack, GPU-to-GPU with no CPU staging) through tensor bindings. The tensor-binding CODE API is deprecated in ovphysx 0.6 in favor of the session read/write API (Python PhysX.read / PhysX.write, C…