Ovrtx

Sensor Processing Graphs: run your own GPU code as a pass over RTX render outputs. Covers the three-file node contract in CUDA and Slang, wiring a graph in USD, built-in nodes, cross-frame state, sensor composites, ray generation, scene queries and diagnosis. Use when the user asks to add SPG to a scene, write or debug an SPG node or launch script, work with `.cu.lua` or `.slang.lua`, author `info:spg:sourceAsset`, `omni:rtx:aov`, `orderedVars` or a `spg:` built-in node, post-process an AOV on the GPU, read a lidar or radar point cloud in a kernel, trace the scene from a node, or work out why an SPG output is empty or black.

Warming up the renderer before capturing output. Use when user asks about warmup frames, image quality, texture streaming, path tracing convergence, or why renders look noisy/incomplete.

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.

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 output routing.

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.

Discovering prims on the runtime stage and inspecting their attribute schemas. Use when user asks to find prims by type, filter by attribute, list all prims, or look up attribute types before reading or writing them.

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

Querying progress for asynchronous ovrtx operations. Use when user asks how to monitor long-running operations, show loading progress, inspect OperationStatus counters, use Operation.query_status(), or call ovrtx_query_op_status()/ovrtx_release_op_status().

Upgrade skill to migrate an existing ovrtx codebase from 0.2.0 to 0.3.0. Use when the user asks to "upgrade from 0.2 to 0.3", migrate ovrtx API usage, update 0.2 projects to 0.3, or fix code after moving from ovrtx 0.2.x to 0.3.x.

Setting up a new CMake C/C++ project that uses ovrtx. Use when user asks to create a new C project, set up CMake with ovrtx, scaffold a C++ app, or configure build dependencies.

Writing render settings to control rendering behavior. Use when user asks to change render settings, set max bounces, configure path tracing, change render quality, or modify RTX settings on a RenderProduct.

Assigning USD SemanticsAPI class and label metadata for semantic segmentation and ground-truth annotation outputs. Use when user asks to label objects, assign semantic class/label metadata, configure semantic segmentation labels, or author SemanticsAPI overrides.

Working with ovx_string_t in C and C++. Use when user asks about printing, comparing, or converting ovx strings.

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.

Render one ovrtx RenderProduct as sequential dataWindowNDC crops and stitch the mapped tiles into one image. Use for sliced or chunked image rendering, or when the user asks how to render a very large image; do not use for multi-camera tiled RenderProducts.

Upgrade skill: migrate user C/C++ application code from ovrtx 0.3.x standalone stage APIs to ovrtx 0.4.x attached to ovstage 0.1.x. Use when the user asks to "upgrade from 0.3 to 0.4 with ovstage", "ovrtx 0.4 ovstage 0.1 migration", "migrate my ovrtx app", or "port my application to ovstage". Reference implementations: `examples/c/minimal` and `examples/c/status-queries` (Recipe A).

Binding materials to prims at runtime. Use when user asks to assign a material, change a material, set material binding, or swap materials on a prim.

Viewport picking, marquee selection, pick-hit decoding, pickability, and selection outline drawing. Use when implementing click picking, drag selection, printing picked prim names, or highlighting selected prims.

Creating and configuring an ovrtx renderer instance. Use when user asks to create a renderer, initialize ovrtx, configure renderer options, or set up ovrtx.

Upgrade an existing Python or C/C++ application from ovrtx 0.4.x to 0.5.x. Use when porting application code, authored sensor USD, RenderVar output handling, DLPack interoperability, Windows renderer configuration, or an attached ovstage 0.1 workflow to the ovrtx 0.5 release train.

Replacing persistent ovrtx attribute bindings with reusable ovstage queries for repeated writes or maps. Use when user asks about persistent bindings, repeated writes, efficient animation loops, bind_attribute, or updating transforms every frame with caller-owned tensors.

Available camera render outputs for Real-Time Path-Tracing (RT2) mode. Use when user asks what AOVs/render vars are available, what format or dtype an output has, or how to read a specific output like depth, normals, albedo, or distance.

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 copy is acceptable.

Reading scalar or array attributes from prims into CPU or GPU tensors. Use when user asks to read an attribute value, fetch mesh data (points, faceVertexCounts, etc.), inspect a render setting, or sample transforms.

Writing scalar and array attribute data to prims. Use when user asks to write an attribute, set a property, change a material, set a color, or modify mesh data.

Writing 4x4 transform matrices to move, rotate, or scale prims. Use when user asks to move an object, set a transform, update position, animate a transform, or set a camera transform.

Authoring non-visual material labels for sensor simulation. Use when user asks to assign lidar/radar/acoustic material semantics, choose nonvisual base materials, coatings, or attributes, debug material IDs, or bind sensor-facing USD materials.

Setting up a new Python project that uses ovrtx. Use when user asks to create a new Python project, set up ovrtx in Python, create a pyproject.toml, or scaffold a Python app.

Mapping rendered output to access pixel data on CPU or GPU. Use when user asks to get pixels, read an image, save a PNG, display rendered output, or access render var data.

Upgrade skill to migrate an existing Python ovrtx codebase from 0.3.x to 0.4.x by moving scene population, cloning, queries, attribute reads, writes, mappings, and persistent binding workflows to ovstage while keeping rendering in ovrtx. Use when the user asks to upgrade from ovrtx 0.3 to 0.4, remove ovrtx 0.4 deprecation warnings, or adopt ovstage interop in an existing ovrtx project.

Running a simulation step to produce rendered frames. Use when user asks to render a frame, step the renderer, simulate a sensor, or get an image from ovrtx.

Authoring and configuring OmniLidar sensor prims and lidar PointCloud render outputs. Use when user asks to create a lidar scene, configure an OmniLidar prim, choose lidar output frame/coordinate behavior, or request lidar PointCloud channels.

Authoring and configuring OmniRadar sensor prims and radar PointCloud render outputs. Use when user asks to create a radar scene, configure an OmniRadar prim, choose radar output frame/coordinate behavior, configure radar scan outputs, or request radar PointCloud channels.

Mapping and reading lidar or radar PointCloud composite render-var tensors. Use when user asks how to access PointCloud output data, map Coordinates, Counts, Intensity, RCS, RadialVelocityMs, or TimeOffsetNs tensors, slice valid entries, or move sensor point clouds through CPU/CUDA memory. For channel meanings and units, use the interpreting-lidar-pointclouds or interpreting-radar-pointclouds skills.

Restricting RenderProducts to specific CUDA-visible device indices with the USD deviceIds attribute. Use when a scene, test, or example must constrain a RenderProduct to CUDA-visible GPU 0 or another explicit set of CUDA-visible GPUs, especially for multi-GPU CI or viewport picking.

Interpreting already-read lidar PointCloud tensors: channel meanings, units, valid point ranges, coordinate-frame implications, flags, IDs, normals, velocity, and visualization values. Use reading-sensor-pointclouds when the user needs to map or fetch the tensors first.

Interpreting 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 tensors first.