[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-nurec-index":3,"mdc--wcqi6a-key":31,"related-org-nvidia-nurec-index":2595,"related-repo-nvidia-nurec-index":2756},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":20,"repoUrl":21,"updatedAt":22,"license":23,"forks":24,"topics":25,"repo":26,"sourceUrl":29,"mdContent":30},"nurec-index","route NVIDIA NuRec and asset tasks","Router for NVIDIA NuRec \u002F NRE \u002F 3DGUT \u002F USDZ \u002F NCore V4 \u002F asset harvest \u002F frame cleanup tasks — picks the right sibling (nre, ncore, asset-harvester, nurec-fixer, physical-ai-datasets). Use when the sub-skill is unclear or a multi-stage pipeline is needed; do NOT use for non-NuRec tasks or to run any pipeline itself.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,17],{"name":13,"slug":14,"type":15},"3D","3d","tag",{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},"Simulation","simulation",16,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills","2026-07-14T05:32:31.902229","CC-BY-4.0 AND Apache-2.0",4,[],{"repoUrl":21,"stars":20,"forks":24,"topics":27,"description":28},[],"Agent skills for Neural Reconstruction Engine","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills\u002Ftree\u002FHEAD\u002Fskills\u002Fnurec-index","---\nname: nurec-index\ndescription: >-\n Router for NVIDIA NuRec \u002F NRE \u002F 3DGUT \u002F USDZ \u002F NCore V4 \u002F asset\n harvest \u002F frame cleanup tasks — picks the right sibling (nre,\n ncore, asset-harvester, nurec-fixer, physical-ai-datasets). Use\n when the sub-skill is unclear or a multi-stage pipeline is needed;\n do NOT use for non-NuRec tasks or to run any pipeline itself.\nversion: \"0.2.4\"\ntools:\n - Read\nlicense: CC-BY-4.0 AND Apache-2.0\nmetadata:\n author: NVIDIA NRS \u003Cnurec-skills@nvidia.com>\n tags:\n - nurec\n - index\n - router\n - table-of-contents\n canonical_repo: https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills\n canonical_skills_dir: .agents\u002Fskills\u002F\n agentskills_io_compatible: true\n trigger_keywords:\n - nurec\n - nurec index\n - nurec router\n - neural reconstruction engine\n - NRE\n - 3DGUT\n - 3DGRT\n - USDZ\n - sensorsim\n - novel view synthesis\n - PhysicalAI-Autonomous-Vehicles-NuRec\n - NuRec pipeline\n - NuRec workflow\n - warm serve-grpc\n - nre thin client\n - batch_render_rgb\n - nurec teardown\n---\n\n# NuRec Skills Index\n\nA routing skill. It decides *which* sibling skill to read next for a\nNuRec \u002F Neural Reconstruction task. Five siblings cover the full\npipeline:\n\n- `physical-ai-datasets` — find an existing NVIDIA dataset.\n- `ncore` — convert raw sensor data into NCore V4.\n- `nre` — train a 3DGUT reconstruction and render novel views from\n NCore V4 (or a pre-trained USDZ).\n- `asset-harvester` — extract per-object 3D Gaussian Splat assets\n from sparse AV-clip views.\n- `nurec-fixer` — clean up artifacts in already-rendered frames.\n\n**Use this index when** the user mentions NuRec \u002F NRE \u002F 3DGUT \u002F\nUSDZ \u002F \"render this clip\" \u002F \"convert this bag\" \u002F \"extract objects\"\nbut the right sub-skill is not yet obvious. Always read this first.\n\n**Do NOT use this index when:**\n\n- The right sub-skill is already obvious (open it directly).\n- The task is not a NuRec task (this skill will not help).\n- Hands-on implementation steps are needed — defer to the sub-skill\n this index points at.\n\n## Purpose\n\nThis skill exists so an agent never has to *guess* which NuRec-family\nskill to read next. It is a hand-curated router for the\nfive-skill NuRec family and nothing else.\n\n**Use cases this skill is built for:**\n\n- **Disambiguate a NuRec request.** The user says \"render this\n clip\", \"convert this bag\", \"fix these frames\", or \"extract this\n car\" and you need to pick the one sibling skill that owns that\n verb.\n- **Bootstrap a multi-stage pipeline.** The user's goal needs two or\n three siblings in a fixed order (convert → train → render, dataset\n → render, harvest → insert, render → harmonize). This index points\n you at the right starting sibling and the matching workflow A–G in\n [`references\u002Fworkflows.md`](references\u002Fworkflows.md).\n- **Translate NuRec jargon for a beginner.** NuRec vs NRE, USDZ vs\n NCore V4, 3DGUT vs 3DGRT, NRE's built-in Fixer vs the standalone\n DiffusionHarmonizer — see [Easy mix-ups](#easy-mix-ups).\n- **Locate a sibling skill that is not on disk.** Defer to\n [`references\u002Fdiscovery.md`](references\u002Fdiscovery.md) instead of\n guessing a path.\n- **Plan disk cleanup across the family.** A full NuRec workflow\n can leave 150 GB+ behind; defer to\n [`references\u002Fteardown.md`](references\u002Fteardown.md) for the\n documented order.\n\n**Use cases this skill is explicitly NOT built for:**\n\n- Running any container, training job, rendering job, conversion,\n dataset download, server, or teardown — always handed off to the\n sibling.\n- Routing tasks outside the NuRec family (Omniverse, Isaac Sim,\n CARLA, Cosmos-* training, generic Hugging Face downloads).\n- Discovering newly-added sibling skills automatically — the\n catalogue is hand-curated and must be edited by hand (see\n [Keeping this index up to date](#keeping-this-index-up-to-date)).\n\n## Instructions\n\nFollow these steps when answering a NuRec-shaped question:\n\n1. **Classify the request.** Read the user's goal and match it to a\n row in [Pick a skill](#pick-a-skill). Multi-stage tasks\n (convert + train, dataset + render) usually start with `ncore`\n or `physical-ai-datasets` and then hand off to `nre`.\n2. **Open exactly one sibling skill next.** Refer to it by `name:`\n (e.g. `nre`), not by file path — names are portable across\n runtimes. If the sibling is not on disk locally, follow\n [`references\u002Fdiscovery.md`](references\u002Fdiscovery.md).\n3. **Defer execution to the sibling.** This index is read-only and\n describes routing only; it never runs containers, training,\n rendering, conversion, or downloads itself.\n4. **For multi-step pipelines, walk a workflow in order.** Pick a\n workflow ID (A–G) from\n [`references\u002Fworkflows.md`](references\u002Fworkflows.md) and open\n the named siblings one at a time, in the listed order. Do not\n collapse steps from that file into this index.\n5. **On disk cleanup**, follow\n [`references\u002Fteardown.md`](references\u002Fteardown.md). A complete\n NuRec workflow can leave 150 GB+ on disk; each sibling owns its\n own teardown.\n6. **Never echo secrets** (`NGC_API_KEY`, `HF_TOKEN`). Use each\n sibling's `scripts\u002Fvalidate_setup.py` when present, or\n `hf auth whoami`. See the \"Secrets\" block in\n [`references\u002Fteardown.md`](references\u002Fteardown.md).\n\n## What is NuRec?\n\n**NuRec** (NVIDIA Omniverse Neural Reconstruction) takes a\nrecording from cameras and LiDAR — usually from a self-driving car\nor a robot — and turns it into a 3D scene that can be re-rendered\nfrom any angle.\n\nNames that appear often:\n\n- **NRE** — \"Neural Reconstruction Engine\", the program that does\n the actual training and rendering. NuRec is the product name;\n NRE is the engine inside it.\n- **USDZ** — the file format the trained scene is saved in. A zip\n archive that Omniverse, Isaac Sim, and CARLA know how to open.\n- **NCore V4** — the input format. Raw recordings must be\n converted into NCore V4 before NRE can consume them.\n- **3DGUT \u002F 3DGRT** — two flavours of 3D Gaussian Splatting that\n NRE uses internally. Most users never pick between them; the\n default Hydra recipe handles it.\n\nA typical NuRec project has three stages:\n\n1. **Get the input** — convert a recording into NCore V4, or\n download a pre-converted dataset from Hugging Face.\n2. **Train the reconstruction** — feed NCore V4 to NRE; out comes\n a USDZ file.\n3. **Render new views** — point NRE at the USDZ to render images,\n videos, or LiDAR sweeps from any camera angle.\n\nSome projects skip step 2 entirely by downloading a USDZ that\nNVIDIA has already trained.\n\n## Pick a skill\n\nMatch the user's goal in the left column, then open the skill on\nthe right. Arrows mean \"do these in order\".\n\n| Goal | Skill to read |\n|------|---------------|\n| Find or download a NuRec dataset NVIDIA has published | `physical-ai-datasets` |\n| Convert camera \u002F LiDAR \u002F radar \u002F depth \u002F stereo into NCore V4 | `ncore` |\n| Write a new converter for a sensor setup not yet supported (drone, RGB-D, ROS 2 bag, COLMAP, …) | `ncore` |\n| Train a 3D reconstruction from an NCore clip | `ncore` → `nre` |\n| Generate the extra inputs NRE needs (segmentation, depth, ego mask) | `nre` (via `nre-tools` container) |\n| Render a USDZ along the **original** camera positions | `nre` |\n| Render at full resolution \u002F highest quality | `nre` (\"Quality presets\" inside that skill) |\n| Render along a **shifted** trajectory | `nre` |\n| Render through a server so a simulator can ask for frames | `nre` (`serve-grpc`) |\n| Render the same USDZ many times back-to-back from Python with minimal per-call latency | `nre` (warm `serve-grpc` + thin Python client \u002F `batch_render_rgb`) |\n| Render LiDAR sweeps (point clouds) from a USDZ | `nre` (`render-grpc --lidar`) |\n| Skip training and just render an NVIDIA-built driving scene | `physical-ai-datasets` → `nre` |\n| Skip training and use a pre-built indoor robotics scene | `physical-ai-datasets` → `nre` (then Isaac Sim 5.1) |\n| Extract individual 3D objects (cars, pedestrians) from a driving clip | `asset-harvester` |\n| Add, remove, or replace cars \u002F pedestrians in a NuRec scene | `asset-harvester` → `nre` |\n| Clean up rendered frames (ghosting, floaters, flickering, inserted-object lighting) | `nurec-fixer`, **or** `--enable-difix` inside `nre` |\n| Export the scene as PLY \u002F mesh \u002F depth maps \u002F ego mask | `nre` |\n| Upgrade an old USDZ so newer NRE versions load it faster | `nre` (`upgrade-artifact`) |\n| Open a USDZ or PLY in a browser viewer | `nre` (`viewer` \u002F `ply_viewer`) |\n| Measure rendering quality (PSNR \u002F SSIM \u002F LPIPS) | `nre` (`eval-rendering-metrics`) |\n| Benchmark different reconstruction methods on the same scenes | `physical-ai-datasets` (`PhysicalAI-NuRec-PPISP`) → `nre` |\n| Train on multiple GPUs or on SLURM | `nre` (Workflow D) |\n\nFor multi-step pipelines, see\n[`references\u002Fworkflows.md`](references\u002Fworkflows.md) (workflows\nA–G).\n\n## The skills in this folder\n\nOpen siblings by their **Name** — that is the canonical\nidentifier. The **Folder** column is just where the skill lives in\nthis repo if it has been cloned locally.\n\n| Name | Folder | What it does |\n|------|--------|--------------|\n| `physical-ai-datasets` | `physical-ai-datasets\u002F` | Catalog and download recipes for every NVIDIA Physical AI dataset on Hugging Face (driving, robotics, manipulation, NuRec scenes, benchmarks). |\n| `ncore` | `ncore\u002F` | Converts any sensor recording into NCore V4. Also covers writing a new converter. |\n| `nre` | `nre\u002F` | The Neural Reconstruction Engine itself. Trains reconstructions, renders frames, exports meshes \u002F point clouds \u002F depth, edits actors, runs the gRPC server, browses results, evaluates quality. |\n| `asset-harvester` | `asset-harvester\u002F` | Apache-2.0 pipeline that extracts individual 3D objects from sparse driving-clip views as `.ply` Gaussian splats plus metadata. |\n| `nurec-fixer` | `nurec-fixer\u002F` | Standalone DiffusionHarmonizer workflow that cleans up rendered frames, harmonizes inserted actors, evaluates PSNR\u002FLPIPS, and optionally fine-tunes the model. |\n\n## Easy mix-ups\n\nThese pairs sound similar but are different things. When in doubt,\ncome back here.\n\n- **NuRec vs NRE.** NuRec is the product name; NRE is the engine\n inside it. Both map to the same skill: `nre`.\n- **NRE's built-in Fixer vs standalone DiffusionHarmonizer.**\n `--enable-difix` inside `nre` is an inline NRE rendering\n feature. The `nurec-fixer` skill covers the standalone public\n DiffusionHarmonizer release (code at `NVIDIA\u002Fharmonizer`, model\n at `nvidia\u002FDiffusionHarmonizer`) for frames already on disk,\n paired evaluation, and fine-tuning. Do not assume these two\n paths share cache layout or weights unless the NRE tag's own\n docs say so.\n- **`ncore` vs `nre`.** They run **in order**, never as\n alternatives. `ncore` produces the input format; `nre` reads it.\n- **`asset-harvester` vs `nre`'s `export-external-assets`.** Asset\n Harvester **produces** the per-object `.ply` files. `nre`'s\n `export-external-assets` **packages** them into a USDZ. Always\n Asset Harvester first.\n- **Cosmos-Drive-Dreams vs PhysicalAI-Autonomous-Vehicles-NuRec.**\n Both are AV datasets on Hugging Face; both are managed by\n `physical-ai-datasets`. Cosmos-Drive-Dreams is **synthetic**\n weather-augmented video (CC-BY-4.0). The NuRec dataset is\n **real** driving scenes turned into renderable USDZs (gated AV\n License).\n\n## Prerequisites\n\nThis index is read-only and needs no tooling. The *real*\nprerequisites live in each sibling skill:\n\n| Sibling | Hard requirements |\n|---------|-------------------|\n| `ncore` | Python 3.10+, `pip install nvidia-ncore`; source data on disk |\n| `nre` | Linux x86_64, NVIDIA GPU (Ampere+, ≥24 GB VRAM), Docker 23+, NVIDIA Container Toolkit, `NGC_API_KEY` |\n| `asset-harvester` | Linux + conda, NVIDIA driver ≥570, ~16 GB VRAM, `HF_TOKEN` |\n| `nurec-fixer` | Linux, NVIDIA GPU (Ampere+), Docker, NVIDIA Container Toolkit, `HF_TOKEN`; `NGC_API_KEY` may be needed for `nvcr.io` pulls |\n| `physical-ai-datasets` | Python + `huggingface_hub`, `HF_TOKEN` (gated datasets need license acceptance) |\n\nEach sibling skill ships `scripts\u002Fvalidate_setup.py` (where\napplicable) — run it before invoking the workflow. Never echo\nsecret env vars; see\n[`references\u002Fteardown.md`](references\u002Fteardown.md).\n\n## Examples\n\nConcrete routing examples. The user prompt is on the left; the\ncorrect action this index should take is on the right.\n\n### Example 1 — single-skill routing\n\n> **User:** \"I have a Waymo Open recording. How do I get it into a\n> format NRE accepts?\"\n\n- Match the row \"Convert camera \u002F LiDAR \u002F radar \u002F depth \u002F stereo\n into NCore V4\" in [Pick a skill](#pick-a-skill).\n- Open the `ncore` skill next; do not run any commands from this\n index.\n\n### Example 2 — multi-stage pipeline\n\n> **User:** \"I have a driving clip. I want to train NuRec and then\n> render a new camera trajectory through it.\"\n\n- Match \"Train a 3D reconstruction from an NCore clip\" → `ncore`\n → `nre`.\n- Cross-check workflow A in\n [`references\u002Fworkflows.md`](references\u002Fworkflows.md).\n- Open `ncore` first, then `nre`, in that order.\n\n### Example 3 — skip training, just render\n\n> **User:** \"Can I just see NuRec working on a scene NVIDIA already\n> built?\"\n\n- Match \"Skip training and just render an NVIDIA-built driving\n scene\" → `physical-ai-datasets` → `nre`.\n- Cross-check workflow B.\n- Open `physical-ai-datasets` first to download **one** scene\n (~1.5–2 GB), then `nre` to render.\n\n### Example 4 — actor editing\n\n> **User:** \"I want to add a pedestrian to this NuRec scene.\"\n\n- Match \"Add, remove, or replace cars \u002F pedestrians\" →\n `asset-harvester` → `nre`.\n- Cross-check workflow D.\n- Confirm the original NCore clip is on disk first (Asset\n Harvester needs it), then open `asset-harvester`, then `nre`\n with `serve-grpc --enable-editing-actors`.\n\n### Example 5 — ambiguous \"fix it\" request\n\n> **User:** \"My rendered frames look fuzzy with weird floaters.\n> Can you clean them up?\"\n\n- Match the \"Clean up rendered frames\" row. Two valid paths:\n - **Quick path:** `--enable-difix` inside `nre` if the user is\n already rendering through NRE.\n - **Standalone path:** `nurec-fixer` for already-rendered\n frames on disk or for paired evaluation \u002F fine-tuning.\n- Ask which interface the user wants, then open the matching\n skill. See workflow E for details.\n\n### Example 6 — sibling skill not on disk\n\n> **User:** \"Where do I find the `nre` skill? It is not in my\n> repo.\"\n\n- Do not guess a path. Follow\n [`references\u002Fdiscovery.md`](references\u002Fdiscovery.md): look under\n `.agents\u002Fskills\u002Fnre\u002FSKILL.md`, then `.claude\u002Fskills\u002Fnre\u002F`, then\n `.cursor\u002Fskills\u002Fnre\u002F`, then `~\u002F.cursor\u002Fskills\u002Fnre\u002F`, and as a\n last resort clone\n `https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills`.\n\n## Limitations\n\n- **Routes only**; never runs pipelines. For hands-on steps, open\n the sibling skill this index points to.\n- **Hand-curated catalogue.** A newly-added sibling skill is not\n discoverable here until someone updates\n [Pick a skill](#pick-a-skill).\n- **Names are portable, paths are not.** Cross-skill links assume\n the canonical layout (`.agents\u002Fskills\u002F\u003Cname>\u002FSKILL.md`). In a\n different runtime layout, prefer name-based skill resolution\n over file paths.\n- **No Omniverse \u002F Isaac Sim integration steps** — those live in\n upstream Omniverse \u002F Isaac docs, not in the NuRec skill family.\n\n## Troubleshooting\n\n| Symptom | Likely cause | Resolution |\n|---------|--------------|------------|\n| Agent picked the wrong sibling skill | The user's task spans multiple stages (e.g. convert + train) | Re-read [Pick a skill](#pick-a-skill) and follow the arrows; multi-stage tasks usually start with `ncore` or `physical-ai-datasets`, then hand off to `nre`. |\n| Sibling skill not found on disk | The host repo only has the index | Follow [`references\u002Fdiscovery.md`](references\u002Fdiscovery.md). |\n| Stale link to `ncore-data-conversion` | Older snapshots used that name; the skill is now `ncore` | Update the link to `ncore`. |\n| User wants to delete disk artifacts | NuRec workflow caches grow large | Walk [`references\u002Fteardown.md`](references\u002Fteardown.md) in the documented order. |\n| User asks \"should I retrain or just clean up frames?\" | Conflating reconstruction vs post-processing | Retrain → `nre`; clean already-rendered frames → `nurec-fixer`. |\n\n## References\n\nDetailed material that this index intentionally keeps out of the\nhot path. Read only when the matching section above points there.\n\n- [`references\u002Fworkflows.md`](references\u002Fworkflows.md) — full\n step-by-step multi-skill workflows A–G.\n- [`references\u002Fteardown.md`](references\u002Fteardown.md) — disk\n cleanup order across all five siblings plus secrets-handling\n policy.\n- [`references\u002Fdiscovery.md`](references\u002Fdiscovery.md) — how to\n locate or fetch a sibling skill that is not already on disk;\n thin-local-skill policy and upstream links.\n\n## Keeping this index up to date\n\nThis index is **hand-curated** — it groups skills by what users\nwant to do, not alphabetically. There is no generator script; edit\nthis `SKILL.md` by hand whenever the sibling set changes.\n\nWhen a new sibling skill is added or a use case shifts:\n\n1. Add a row to [Pick a skill](#pick-a-skill) for the new use\n case.\n2. Add a row to\n [The skills in this folder](#the-skills-in-this-folder).\n3. If the new skill changes a multi-step pipeline, update\n [`references\u002Fworkflows.md`](references\u002Fworkflows.md).\n4. Confirm the sibling's `metadata.upstream` field still points at\n the canonical upstream repo or container.\n\nOtherwise the index will quietly drift and beginners will end up\nreading the wrong skill.\n",{"data":32,"body":62},{"name":4,"description":6,"version":33,"tools":34,"license":23,"metadata":36},"0.2.4",[35],"Read",{"author":37,"tags":38,"canonical_repo":21,"canonical_skills_dir":43,"agentskills_io_compatible":44,"trigger_keywords":45},"NVIDIA NRS \u003Cnurec-skills@nvidia.com>",[39,40,41,42],"nurec","index","router","table-of-contents",".agents\u002Fskills\u002F",true,[39,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61],"nurec index","nurec router","neural reconstruction engine","NRE","3DGUT","3DGRT","USDZ","sensorsim","novel view synthesis","PhysicalAI-Autonomous-Vehicles-NuRec","NuRec pipeline","NuRec workflow","warm serve-grpc","nre thin client","batch_render_rgb","nurec teardown",{"type":63,"children":64},"root",[65,74,88,149,160,168,186,193,205,213,307,315,341,347,352,529,535,545,550,591,596,629,634,639,644,1178,1193,1199,1218,1376,1381,1386,1584,1590,1601,1757,1778,1784,1789,1796,1810,1835,1841,1853,1907,1913,1925,1975,1981,1993,2044,2050,2062,2116,2122,2141,2195,2201,2257,2263,2455,2461,2466,2508,2513,2533,2538,2590],{"type":66,"tag":67,"props":68,"children":70},"element","h1",{"id":69},"nurec-skills-index",[71],{"type":72,"value":73},"text","NuRec Skills Index",{"type":66,"tag":75,"props":76,"children":77},"p",{},[78,80,86],{"type":72,"value":79},"A routing skill. It decides ",{"type":66,"tag":81,"props":82,"children":83},"em",{},[84],{"type":72,"value":85},"which",{"type":72,"value":87}," sibling skill to read next for a\nNuRec \u002F Neural Reconstruction task. Five siblings cover the full\npipeline:",{"type":66,"tag":89,"props":90,"children":91},"ul",{},[92,105,116,127,138],{"type":66,"tag":93,"props":94,"children":95},"li",{},[96,103],{"type":66,"tag":97,"props":98,"children":100},"code",{"className":99},[],[101],{"type":72,"value":102},"physical-ai-datasets",{"type":72,"value":104}," — find an existing NVIDIA dataset.",{"type":66,"tag":93,"props":106,"children":107},{},[108,114],{"type":66,"tag":97,"props":109,"children":111},{"className":110},[],[112],{"type":72,"value":113},"ncore",{"type":72,"value":115}," — convert raw sensor data into NCore V4.",{"type":66,"tag":93,"props":117,"children":118},{},[119,125],{"type":66,"tag":97,"props":120,"children":122},{"className":121},[],[123],{"type":72,"value":124},"nre",{"type":72,"value":126}," — train a 3DGUT reconstruction and render novel views from\nNCore V4 (or a pre-trained USDZ).",{"type":66,"tag":93,"props":128,"children":129},{},[130,136],{"type":66,"tag":97,"props":131,"children":133},{"className":132},[],[134],{"type":72,"value":135},"asset-harvester",{"type":72,"value":137}," — extract per-object 3D Gaussian Splat assets\nfrom sparse AV-clip views.",{"type":66,"tag":93,"props":139,"children":140},{},[141,147],{"type":66,"tag":97,"props":142,"children":144},{"className":143},[],[145],{"type":72,"value":146},"nurec-fixer",{"type":72,"value":148}," — clean up artifacts in already-rendered frames.",{"type":66,"tag":75,"props":150,"children":151},{},[152,158],{"type":66,"tag":153,"props":154,"children":155},"strong",{},[156],{"type":72,"value":157},"Use this index when",{"type":72,"value":159}," the user mentions NuRec \u002F NRE \u002F 3DGUT \u002F\nUSDZ \u002F \"render this clip\" \u002F \"convert this bag\" \u002F \"extract objects\"\nbut the right sub-skill is not yet obvious. Always read this first.",{"type":66,"tag":75,"props":161,"children":162},{},[163],{"type":66,"tag":153,"props":164,"children":165},{},[166],{"type":72,"value":167},"Do NOT use this index when:",{"type":66,"tag":89,"props":169,"children":170},{},[171,176,181],{"type":66,"tag":93,"props":172,"children":173},{},[174],{"type":72,"value":175},"The right sub-skill is already obvious (open it directly).",{"type":66,"tag":93,"props":177,"children":178},{},[179],{"type":72,"value":180},"The task is not a NuRec task (this skill will not help).",{"type":66,"tag":93,"props":182,"children":183},{},[184],{"type":72,"value":185},"Hands-on implementation steps are needed — defer to the sub-skill\nthis index points at.",{"type":66,"tag":187,"props":188,"children":190},"h2",{"id":189},"purpose",[191],{"type":72,"value":192},"Purpose",{"type":66,"tag":75,"props":194,"children":195},{},[196,198,203],{"type":72,"value":197},"This skill exists so an agent never has to ",{"type":66,"tag":81,"props":199,"children":200},{},[201],{"type":72,"value":202},"guess",{"type":72,"value":204}," which NuRec-family\nskill to read next. It is a hand-curated router for the\nfive-skill NuRec family and nothing else.",{"type":66,"tag":75,"props":206,"children":207},{},[208],{"type":66,"tag":153,"props":209,"children":210},{},[211],{"type":72,"value":212},"Use cases this skill is built for:",{"type":66,"tag":89,"props":214,"children":215},{},[216,226,248,265,286],{"type":66,"tag":93,"props":217,"children":218},{},[219,224],{"type":66,"tag":153,"props":220,"children":221},{},[222],{"type":72,"value":223},"Disambiguate a NuRec request.",{"type":72,"value":225}," The user says \"render this\nclip\", \"convert this bag\", \"fix these frames\", or \"extract this\ncar\" and you need to pick the one sibling skill that owns that\nverb.",{"type":66,"tag":93,"props":227,"children":228},{},[229,234,236,246],{"type":66,"tag":153,"props":230,"children":231},{},[232],{"type":72,"value":233},"Bootstrap a multi-stage pipeline.",{"type":72,"value":235}," The user's goal needs two or\nthree siblings in a fixed order (convert → train → render, dataset\n→ render, harvest → insert, render → harmonize). This index points\nyou at the right starting sibling and the matching workflow A–G in\n",{"type":66,"tag":237,"props":238,"children":240},"a",{"href":239},"references\u002Fworkflows.md",[241],{"type":66,"tag":97,"props":242,"children":244},{"className":243},[],[245],{"type":72,"value":239},{"type":72,"value":247},".",{"type":66,"tag":93,"props":249,"children":250},{},[251,256,258,264],{"type":66,"tag":153,"props":252,"children":253},{},[254],{"type":72,"value":255},"Translate NuRec jargon for a beginner.",{"type":72,"value":257}," NuRec vs NRE, USDZ vs\nNCore V4, 3DGUT vs 3DGRT, NRE's built-in Fixer vs the standalone\nDiffusionHarmonizer — see ",{"type":66,"tag":237,"props":259,"children":261},{"href":260},"#easy-mix-ups",[262],{"type":72,"value":263},"Easy mix-ups",{"type":72,"value":247},{"type":66,"tag":93,"props":266,"children":267},{},[268,273,275,284],{"type":66,"tag":153,"props":269,"children":270},{},[271],{"type":72,"value":272},"Locate a sibling skill that is not on disk.",{"type":72,"value":274}," Defer to\n",{"type":66,"tag":237,"props":276,"children":278},{"href":277},"references\u002Fdiscovery.md",[279],{"type":66,"tag":97,"props":280,"children":282},{"className":281},[],[283],{"type":72,"value":277},{"type":72,"value":285}," instead of\nguessing a path.",{"type":66,"tag":93,"props":287,"children":288},{},[289,294,296,305],{"type":66,"tag":153,"props":290,"children":291},{},[292],{"type":72,"value":293},"Plan disk cleanup across the family.",{"type":72,"value":295}," A full NuRec workflow\ncan leave 150 GB+ behind; defer to\n",{"type":66,"tag":237,"props":297,"children":299},{"href":298},"references\u002Fteardown.md",[300],{"type":66,"tag":97,"props":301,"children":303},{"className":302},[],[304],{"type":72,"value":298},{"type":72,"value":306}," for the\ndocumented order.",{"type":66,"tag":75,"props":308,"children":309},{},[310],{"type":66,"tag":153,"props":311,"children":312},{},[313],{"type":72,"value":314},"Use cases this skill is explicitly NOT built for:",{"type":66,"tag":89,"props":316,"children":317},{},[318,323,328],{"type":66,"tag":93,"props":319,"children":320},{},[321],{"type":72,"value":322},"Running any container, training job, rendering job, conversion,\ndataset download, server, or teardown — always handed off to the\nsibling.",{"type":66,"tag":93,"props":324,"children":325},{},[326],{"type":72,"value":327},"Routing tasks outside the NuRec family (Omniverse, Isaac Sim,\nCARLA, Cosmos-* training, generic Hugging Face downloads).",{"type":66,"tag":93,"props":329,"children":330},{},[331,333,339],{"type":72,"value":332},"Discovering newly-added sibling skills automatically — the\ncatalogue is hand-curated and must be edited by hand (see\n",{"type":66,"tag":237,"props":334,"children":336},{"href":335},"#keeping-this-index-up-to-date",[337],{"type":72,"value":338},"Keeping this index up to date",{"type":72,"value":340},").",{"type":66,"tag":187,"props":342,"children":344},{"id":343},"instructions",[345],{"type":72,"value":346},"Instructions",{"type":66,"tag":75,"props":348,"children":349},{},[350],{"type":72,"value":351},"Follow these steps when answering a NuRec-shaped question:",{"type":66,"tag":353,"props":354,"children":355},"ol",{},[356,394,428,438,458,478],{"type":66,"tag":93,"props":357,"children":358},{},[359,364,366,372,374,379,381,386,388,393],{"type":66,"tag":153,"props":360,"children":361},{},[362],{"type":72,"value":363},"Classify the request.",{"type":72,"value":365}," Read the user's goal and match it to a\nrow in ",{"type":66,"tag":237,"props":367,"children":369},{"href":368},"#pick-a-skill",[370],{"type":72,"value":371},"Pick a skill",{"type":72,"value":373},". Multi-stage tasks\n(convert + train, dataset + render) usually start with ",{"type":66,"tag":97,"props":375,"children":377},{"className":376},[],[378],{"type":72,"value":113},{"type":72,"value":380},"\nor ",{"type":66,"tag":97,"props":382,"children":384},{"className":383},[],[385],{"type":72,"value":102},{"type":72,"value":387}," and then hand off to ",{"type":66,"tag":97,"props":389,"children":391},{"className":390},[],[392],{"type":72,"value":124},{"type":72,"value":247},{"type":66,"tag":93,"props":395,"children":396},{},[397,402,404,410,412,417,419,427],{"type":66,"tag":153,"props":398,"children":399},{},[400],{"type":72,"value":401},"Open exactly one sibling skill next.",{"type":72,"value":403}," Refer to it by ",{"type":66,"tag":97,"props":405,"children":407},{"className":406},[],[408],{"type":72,"value":409},"name:",{"type":72,"value":411},"\n(e.g. ",{"type":66,"tag":97,"props":413,"children":415},{"className":414},[],[416],{"type":72,"value":124},{"type":72,"value":418},"), not by file path — names are portable across\nruntimes. If the sibling is not on disk locally, follow\n",{"type":66,"tag":237,"props":420,"children":421},{"href":277},[422],{"type":66,"tag":97,"props":423,"children":425},{"className":424},[],[426],{"type":72,"value":277},{"type":72,"value":247},{"type":66,"tag":93,"props":429,"children":430},{},[431,436],{"type":66,"tag":153,"props":432,"children":433},{},[434],{"type":72,"value":435},"Defer execution to the sibling.",{"type":72,"value":437}," This index is read-only and\ndescribes routing only; it never runs containers, training,\nrendering, conversion, or downloads itself.",{"type":66,"tag":93,"props":439,"children":440},{},[441,446,448,456],{"type":66,"tag":153,"props":442,"children":443},{},[444],{"type":72,"value":445},"For multi-step pipelines, walk a workflow in order.",{"type":72,"value":447}," Pick a\nworkflow ID (A–G) from\n",{"type":66,"tag":237,"props":449,"children":450},{"href":239},[451],{"type":66,"tag":97,"props":452,"children":454},{"className":453},[],[455],{"type":72,"value":239},{"type":72,"value":457}," and open\nthe named siblings one at a time, in the listed order. Do not\ncollapse steps from that file into this index.",{"type":66,"tag":93,"props":459,"children":460},{},[461,466,468,476],{"type":66,"tag":153,"props":462,"children":463},{},[464],{"type":72,"value":465},"On disk cleanup",{"type":72,"value":467},", follow\n",{"type":66,"tag":237,"props":469,"children":470},{"href":298},[471],{"type":66,"tag":97,"props":472,"children":474},{"className":473},[],[475],{"type":72,"value":298},{"type":72,"value":477},". A complete\nNuRec workflow can leave 150 GB+ on disk; each sibling owns its\nown teardown.",{"type":66,"tag":93,"props":479,"children":480},{},[481,486,488,494,496,502,504,510,512,518,520,528],{"type":66,"tag":153,"props":482,"children":483},{},[484],{"type":72,"value":485},"Never echo secrets",{"type":72,"value":487}," (",{"type":66,"tag":97,"props":489,"children":491},{"className":490},[],[492],{"type":72,"value":493},"NGC_API_KEY",{"type":72,"value":495},", ",{"type":66,"tag":97,"props":497,"children":499},{"className":498},[],[500],{"type":72,"value":501},"HF_TOKEN",{"type":72,"value":503},"). Use each\nsibling's ",{"type":66,"tag":97,"props":505,"children":507},{"className":506},[],[508],{"type":72,"value":509},"scripts\u002Fvalidate_setup.py",{"type":72,"value":511}," when present, or\n",{"type":66,"tag":97,"props":513,"children":515},{"className":514},[],[516],{"type":72,"value":517},"hf auth whoami",{"type":72,"value":519},". See the \"Secrets\" block in\n",{"type":66,"tag":237,"props":521,"children":522},{"href":298},[523],{"type":66,"tag":97,"props":524,"children":526},{"className":525},[],[527],{"type":72,"value":298},{"type":72,"value":247},{"type":66,"tag":187,"props":530,"children":532},{"id":531},"what-is-nurec",[533],{"type":72,"value":534},"What is NuRec?",{"type":66,"tag":75,"props":536,"children":537},{},[538,543],{"type":66,"tag":153,"props":539,"children":540},{},[541],{"type":72,"value":542},"NuRec",{"type":72,"value":544}," (NVIDIA Omniverse Neural Reconstruction) takes a\nrecording from cameras and LiDAR — usually from a self-driving car\nor a robot — and turns it into a 3D scene that can be re-rendered\nfrom any angle.",{"type":66,"tag":75,"props":546,"children":547},{},[548],{"type":72,"value":549},"Names that appear often:",{"type":66,"tag":89,"props":551,"children":552},{},[553,562,571,581],{"type":66,"tag":93,"props":554,"children":555},{},[556,560],{"type":66,"tag":153,"props":557,"children":558},{},[559],{"type":72,"value":49},{"type":72,"value":561}," — \"Neural Reconstruction Engine\", the program that does\nthe actual training and rendering. NuRec is the product name;\nNRE is the engine inside it.",{"type":66,"tag":93,"props":563,"children":564},{},[565,569],{"type":66,"tag":153,"props":566,"children":567},{},[568],{"type":72,"value":52},{"type":72,"value":570}," — the file format the trained scene is saved in. A zip\narchive that Omniverse, Isaac Sim, and CARLA know how to open.",{"type":66,"tag":93,"props":572,"children":573},{},[574,579],{"type":66,"tag":153,"props":575,"children":576},{},[577],{"type":72,"value":578},"NCore V4",{"type":72,"value":580}," — the input format. Raw recordings must be\nconverted into NCore V4 before NRE can consume them.",{"type":66,"tag":93,"props":582,"children":583},{},[584,589],{"type":66,"tag":153,"props":585,"children":586},{},[587],{"type":72,"value":588},"3DGUT \u002F 3DGRT",{"type":72,"value":590}," — two flavours of 3D Gaussian Splatting that\nNRE uses internally. Most users never pick between them; the\ndefault Hydra recipe handles it.",{"type":66,"tag":75,"props":592,"children":593},{},[594],{"type":72,"value":595},"A typical NuRec project has three stages:",{"type":66,"tag":353,"props":597,"children":598},{},[599,609,619],{"type":66,"tag":93,"props":600,"children":601},{},[602,607],{"type":66,"tag":153,"props":603,"children":604},{},[605],{"type":72,"value":606},"Get the input",{"type":72,"value":608}," — convert a recording into NCore V4, or\ndownload a pre-converted dataset from Hugging Face.",{"type":66,"tag":93,"props":610,"children":611},{},[612,617],{"type":66,"tag":153,"props":613,"children":614},{},[615],{"type":72,"value":616},"Train the reconstruction",{"type":72,"value":618}," — feed NCore V4 to NRE; out comes\na USDZ file.",{"type":66,"tag":93,"props":620,"children":621},{},[622,627],{"type":66,"tag":153,"props":623,"children":624},{},[625],{"type":72,"value":626},"Render new views",{"type":72,"value":628}," — point NRE at the USDZ to render images,\nvideos, or LiDAR sweeps from any camera angle.",{"type":66,"tag":75,"props":630,"children":631},{},[632],{"type":72,"value":633},"Some projects skip step 2 entirely by downloading a USDZ that\nNVIDIA has already trained.",{"type":66,"tag":187,"props":635,"children":637},{"id":636},"pick-a-skill",[638],{"type":72,"value":371},{"type":66,"tag":75,"props":640,"children":641},{},[642],{"type":72,"value":643},"Match the user's goal in the left column, then open the skill on\nthe right. Arrows mean \"do these in order\".",{"type":66,"tag":645,"props":646,"children":647},"table",{},[648,667],{"type":66,"tag":649,"props":650,"children":651},"thead",{},[652],{"type":66,"tag":653,"props":654,"children":655},"tr",{},[656,662],{"type":66,"tag":657,"props":658,"children":659},"th",{},[660],{"type":72,"value":661},"Goal",{"type":66,"tag":657,"props":663,"children":664},{},[665],{"type":72,"value":666},"Skill to read",{"type":66,"tag":668,"props":669,"children":670},"tbody",{},[671,688,704,720,743,769,792,810,833,858,889,913,935,959,975,997,1034,1050,1074,1106,1130,1160],{"type":66,"tag":653,"props":672,"children":673},{},[674,680],{"type":66,"tag":675,"props":676,"children":677},"td",{},[678],{"type":72,"value":679},"Find or download a NuRec dataset NVIDIA has published",{"type":66,"tag":675,"props":681,"children":682},{},[683],{"type":66,"tag":97,"props":684,"children":686},{"className":685},[],[687],{"type":72,"value":102},{"type":66,"tag":653,"props":689,"children":690},{},[691,696],{"type":66,"tag":675,"props":692,"children":693},{},[694],{"type":72,"value":695},"Convert camera \u002F LiDAR \u002F radar \u002F depth \u002F stereo into NCore V4",{"type":66,"tag":675,"props":697,"children":698},{},[699],{"type":66,"tag":97,"props":700,"children":702},{"className":701},[],[703],{"type":72,"value":113},{"type":66,"tag":653,"props":705,"children":706},{},[707,712],{"type":66,"tag":675,"props":708,"children":709},{},[710],{"type":72,"value":711},"Write a new converter for a sensor setup not yet supported (drone, RGB-D, ROS 2 bag, COLMAP, …)",{"type":66,"tag":675,"props":713,"children":714},{},[715],{"type":66,"tag":97,"props":716,"children":718},{"className":717},[],[719],{"type":72,"value":113},{"type":66,"tag":653,"props":721,"children":722},{},[723,728],{"type":66,"tag":675,"props":724,"children":725},{},[726],{"type":72,"value":727},"Train a 3D reconstruction from an NCore clip",{"type":66,"tag":675,"props":729,"children":730},{},[731,736,738],{"type":66,"tag":97,"props":732,"children":734},{"className":733},[],[735],{"type":72,"value":113},{"type":72,"value":737}," → ",{"type":66,"tag":97,"props":739,"children":741},{"className":740},[],[742],{"type":72,"value":124},{"type":66,"tag":653,"props":744,"children":745},{},[746,751],{"type":66,"tag":675,"props":747,"children":748},{},[749],{"type":72,"value":750},"Generate the extra inputs NRE needs (segmentation, depth, ego mask)",{"type":66,"tag":675,"props":752,"children":753},{},[754,759,761,767],{"type":66,"tag":97,"props":755,"children":757},{"className":756},[],[758],{"type":72,"value":124},{"type":72,"value":760}," (via ",{"type":66,"tag":97,"props":762,"children":764},{"className":763},[],[765],{"type":72,"value":766},"nre-tools",{"type":72,"value":768}," container)",{"type":66,"tag":653,"props":770,"children":771},{},[772,784],{"type":66,"tag":675,"props":773,"children":774},{},[775,777,782],{"type":72,"value":776},"Render a USDZ along the ",{"type":66,"tag":153,"props":778,"children":779},{},[780],{"type":72,"value":781},"original",{"type":72,"value":783}," camera positions",{"type":66,"tag":675,"props":785,"children":786},{},[787],{"type":66,"tag":97,"props":788,"children":790},{"className":789},[],[791],{"type":72,"value":124},{"type":66,"tag":653,"props":793,"children":794},{},[795,800],{"type":66,"tag":675,"props":796,"children":797},{},[798],{"type":72,"value":799},"Render at full resolution \u002F highest quality",{"type":66,"tag":675,"props":801,"children":802},{},[803,808],{"type":66,"tag":97,"props":804,"children":806},{"className":805},[],[807],{"type":72,"value":124},{"type":72,"value":809}," (\"Quality presets\" inside that skill)",{"type":66,"tag":653,"props":811,"children":812},{},[813,825],{"type":66,"tag":675,"props":814,"children":815},{},[816,818,823],{"type":72,"value":817},"Render along a ",{"type":66,"tag":153,"props":819,"children":820},{},[821],{"type":72,"value":822},"shifted",{"type":72,"value":824}," trajectory",{"type":66,"tag":675,"props":826,"children":827},{},[828],{"type":66,"tag":97,"props":829,"children":831},{"className":830},[],[832],{"type":72,"value":124},{"type":66,"tag":653,"props":834,"children":835},{},[836,841],{"type":66,"tag":675,"props":837,"children":838},{},[839],{"type":72,"value":840},"Render through a server so a simulator can ask for frames",{"type":66,"tag":675,"props":842,"children":843},{},[844,849,850,856],{"type":66,"tag":97,"props":845,"children":847},{"className":846},[],[848],{"type":72,"value":124},{"type":72,"value":487},{"type":66,"tag":97,"props":851,"children":853},{"className":852},[],[854],{"type":72,"value":855},"serve-grpc",{"type":72,"value":857},")",{"type":66,"tag":653,"props":859,"children":860},{},[861,866],{"type":66,"tag":675,"props":862,"children":863},{},[864],{"type":72,"value":865},"Render the same USDZ many times back-to-back from Python with minimal per-call latency",{"type":66,"tag":675,"props":867,"children":868},{},[869,874,876,881,883,888],{"type":66,"tag":97,"props":870,"children":872},{"className":871},[],[873],{"type":72,"value":124},{"type":72,"value":875}," (warm ",{"type":66,"tag":97,"props":877,"children":879},{"className":878},[],[880],{"type":72,"value":855},{"type":72,"value":882}," + thin Python client \u002F ",{"type":66,"tag":97,"props":884,"children":886},{"className":885},[],[887],{"type":72,"value":60},{"type":72,"value":857},{"type":66,"tag":653,"props":890,"children":891},{},[892,897],{"type":66,"tag":675,"props":893,"children":894},{},[895],{"type":72,"value":896},"Render LiDAR sweeps (point clouds) from a USDZ",{"type":66,"tag":675,"props":898,"children":899},{},[900,905,906,912],{"type":66,"tag":97,"props":901,"children":903},{"className":902},[],[904],{"type":72,"value":124},{"type":72,"value":487},{"type":66,"tag":97,"props":907,"children":909},{"className":908},[],[910],{"type":72,"value":911},"render-grpc --lidar",{"type":72,"value":857},{"type":66,"tag":653,"props":914,"children":915},{},[916,921],{"type":66,"tag":675,"props":917,"children":918},{},[919],{"type":72,"value":920},"Skip training and just render an NVIDIA-built driving scene",{"type":66,"tag":675,"props":922,"children":923},{},[924,929,930],{"type":66,"tag":97,"props":925,"children":927},{"className":926},[],[928],{"type":72,"value":102},{"type":72,"value":737},{"type":66,"tag":97,"props":931,"children":933},{"className":932},[],[934],{"type":72,"value":124},{"type":66,"tag":653,"props":936,"children":937},{},[938,943],{"type":66,"tag":675,"props":939,"children":940},{},[941],{"type":72,"value":942},"Skip training and use a pre-built indoor robotics scene",{"type":66,"tag":675,"props":944,"children":945},{},[946,951,952,957],{"type":66,"tag":97,"props":947,"children":949},{"className":948},[],[950],{"type":72,"value":102},{"type":72,"value":737},{"type":66,"tag":97,"props":953,"children":955},{"className":954},[],[956],{"type":72,"value":124},{"type":72,"value":958}," (then Isaac Sim 5.1)",{"type":66,"tag":653,"props":960,"children":961},{},[962,967],{"type":66,"tag":675,"props":963,"children":964},{},[965],{"type":72,"value":966},"Extract individual 3D objects (cars, pedestrians) from a driving clip",{"type":66,"tag":675,"props":968,"children":969},{},[970],{"type":66,"tag":97,"props":971,"children":973},{"className":972},[],[974],{"type":72,"value":135},{"type":66,"tag":653,"props":976,"children":977},{},[978,983],{"type":66,"tag":675,"props":979,"children":980},{},[981],{"type":72,"value":982},"Add, remove, or replace cars \u002F pedestrians in a NuRec scene",{"type":66,"tag":675,"props":984,"children":985},{},[986,991,992],{"type":66,"tag":97,"props":987,"children":989},{"className":988},[],[990],{"type":72,"value":135},{"type":72,"value":737},{"type":66,"tag":97,"props":993,"children":995},{"className":994},[],[996],{"type":72,"value":124},{"type":66,"tag":653,"props":998,"children":999},{},[1000,1005],{"type":66,"tag":675,"props":1001,"children":1002},{},[1003],{"type":72,"value":1004},"Clean up rendered frames (ghosting, floaters, flickering, inserted-object lighting)",{"type":66,"tag":675,"props":1006,"children":1007},{},[1008,1013,1014,1019,1021,1027,1029],{"type":66,"tag":97,"props":1009,"children":1011},{"className":1010},[],[1012],{"type":72,"value":146},{"type":72,"value":495},{"type":66,"tag":153,"props":1015,"children":1016},{},[1017],{"type":72,"value":1018},"or",{"type":72,"value":1020}," ",{"type":66,"tag":97,"props":1022,"children":1024},{"className":1023},[],[1025],{"type":72,"value":1026},"--enable-difix",{"type":72,"value":1028}," inside ",{"type":66,"tag":97,"props":1030,"children":1032},{"className":1031},[],[1033],{"type":72,"value":124},{"type":66,"tag":653,"props":1035,"children":1036},{},[1037,1042],{"type":66,"tag":675,"props":1038,"children":1039},{},[1040],{"type":72,"value":1041},"Export the scene as PLY \u002F mesh \u002F depth maps \u002F ego mask",{"type":66,"tag":675,"props":1043,"children":1044},{},[1045],{"type":66,"tag":97,"props":1046,"children":1048},{"className":1047},[],[1049],{"type":72,"value":124},{"type":66,"tag":653,"props":1051,"children":1052},{},[1053,1058],{"type":66,"tag":675,"props":1054,"children":1055},{},[1056],{"type":72,"value":1057},"Upgrade an old USDZ so newer NRE versions load it faster",{"type":66,"tag":675,"props":1059,"children":1060},{},[1061,1066,1067,1073],{"type":66,"tag":97,"props":1062,"children":1064},{"className":1063},[],[1065],{"type":72,"value":124},{"type":72,"value":487},{"type":66,"tag":97,"props":1068,"children":1070},{"className":1069},[],[1071],{"type":72,"value":1072},"upgrade-artifact",{"type":72,"value":857},{"type":66,"tag":653,"props":1075,"children":1076},{},[1077,1082],{"type":66,"tag":675,"props":1078,"children":1079},{},[1080],{"type":72,"value":1081},"Open a USDZ or PLY in a browser viewer",{"type":66,"tag":675,"props":1083,"children":1084},{},[1085,1090,1091,1097,1099,1105],{"type":66,"tag":97,"props":1086,"children":1088},{"className":1087},[],[1089],{"type":72,"value":124},{"type":72,"value":487},{"type":66,"tag":97,"props":1092,"children":1094},{"className":1093},[],[1095],{"type":72,"value":1096},"viewer",{"type":72,"value":1098}," \u002F ",{"type":66,"tag":97,"props":1100,"children":1102},{"className":1101},[],[1103],{"type":72,"value":1104},"ply_viewer",{"type":72,"value":857},{"type":66,"tag":653,"props":1107,"children":1108},{},[1109,1114],{"type":66,"tag":675,"props":1110,"children":1111},{},[1112],{"type":72,"value":1113},"Measure rendering quality (PSNR \u002F SSIM \u002F LPIPS)",{"type":66,"tag":675,"props":1115,"children":1116},{},[1117,1122,1123,1129],{"type":66,"tag":97,"props":1118,"children":1120},{"className":1119},[],[1121],{"type":72,"value":124},{"type":72,"value":487},{"type":66,"tag":97,"props":1124,"children":1126},{"className":1125},[],[1127],{"type":72,"value":1128},"eval-rendering-metrics",{"type":72,"value":857},{"type":66,"tag":653,"props":1131,"children":1132},{},[1133,1138],{"type":66,"tag":675,"props":1134,"children":1135},{},[1136],{"type":72,"value":1137},"Benchmark different reconstruction methods on the same scenes",{"type":66,"tag":675,"props":1139,"children":1140},{},[1141,1146,1147,1153,1155],{"type":66,"tag":97,"props":1142,"children":1144},{"className":1143},[],[1145],{"type":72,"value":102},{"type":72,"value":487},{"type":66,"tag":97,"props":1148,"children":1150},{"className":1149},[],[1151],{"type":72,"value":1152},"PhysicalAI-NuRec-PPISP",{"type":72,"value":1154},") → ",{"type":66,"tag":97,"props":1156,"children":1158},{"className":1157},[],[1159],{"type":72,"value":124},{"type":66,"tag":653,"props":1161,"children":1162},{},[1163,1168],{"type":66,"tag":675,"props":1164,"children":1165},{},[1166],{"type":72,"value":1167},"Train on multiple GPUs or on SLURM",{"type":66,"tag":675,"props":1169,"children":1170},{},[1171,1176],{"type":66,"tag":97,"props":1172,"children":1174},{"className":1173},[],[1175],{"type":72,"value":124},{"type":72,"value":1177}," (Workflow D)",{"type":66,"tag":75,"props":1179,"children":1180},{},[1181,1183,1191],{"type":72,"value":1182},"For multi-step pipelines, see\n",{"type":66,"tag":237,"props":1184,"children":1185},{"href":239},[1186],{"type":66,"tag":97,"props":1187,"children":1189},{"className":1188},[],[1190],{"type":72,"value":239},{"type":72,"value":1192}," (workflows\nA–G).",{"type":66,"tag":187,"props":1194,"children":1196},{"id":1195},"the-skills-in-this-folder",[1197],{"type":72,"value":1198},"The skills in this folder",{"type":66,"tag":75,"props":1200,"children":1201},{},[1202,1204,1209,1211,1216],{"type":72,"value":1203},"Open siblings by their ",{"type":66,"tag":153,"props":1205,"children":1206},{},[1207],{"type":72,"value":1208},"Name",{"type":72,"value":1210}," — that is the canonical\nidentifier. The ",{"type":66,"tag":153,"props":1212,"children":1213},{},[1214],{"type":72,"value":1215},"Folder",{"type":72,"value":1217}," column is just where the skill lives in\nthis repo if it has been cloned locally.",{"type":66,"tag":645,"props":1219,"children":1220},{},[1221,1240],{"type":66,"tag":649,"props":1222,"children":1223},{},[1224],{"type":66,"tag":653,"props":1225,"children":1226},{},[1227,1231,1235],{"type":66,"tag":657,"props":1228,"children":1229},{},[1230],{"type":72,"value":1208},{"type":66,"tag":657,"props":1232,"children":1233},{},[1234],{"type":72,"value":1215},{"type":66,"tag":657,"props":1236,"children":1237},{},[1238],{"type":72,"value":1239},"What it does",{"type":66,"tag":668,"props":1241,"children":1242},{},[1243,1268,1293,1318,1351],{"type":66,"tag":653,"props":1244,"children":1245},{},[1246,1254,1263],{"type":66,"tag":675,"props":1247,"children":1248},{},[1249],{"type":66,"tag":97,"props":1250,"children":1252},{"className":1251},[],[1253],{"type":72,"value":102},{"type":66,"tag":675,"props":1255,"children":1256},{},[1257],{"type":66,"tag":97,"props":1258,"children":1260},{"className":1259},[],[1261],{"type":72,"value":1262},"physical-ai-datasets\u002F",{"type":66,"tag":675,"props":1264,"children":1265},{},[1266],{"type":72,"value":1267},"Catalog and download recipes for every NVIDIA Physical AI dataset on Hugging Face (driving, robotics, manipulation, NuRec scenes, benchmarks).",{"type":66,"tag":653,"props":1269,"children":1270},{},[1271,1279,1288],{"type":66,"tag":675,"props":1272,"children":1273},{},[1274],{"type":66,"tag":97,"props":1275,"children":1277},{"className":1276},[],[1278],{"type":72,"value":113},{"type":66,"tag":675,"props":1280,"children":1281},{},[1282],{"type":66,"tag":97,"props":1283,"children":1285},{"className":1284},[],[1286],{"type":72,"value":1287},"ncore\u002F",{"type":66,"tag":675,"props":1289,"children":1290},{},[1291],{"type":72,"value":1292},"Converts any sensor recording into NCore V4. Also covers writing a new converter.",{"type":66,"tag":653,"props":1294,"children":1295},{},[1296,1304,1313],{"type":66,"tag":675,"props":1297,"children":1298},{},[1299],{"type":66,"tag":97,"props":1300,"children":1302},{"className":1301},[],[1303],{"type":72,"value":124},{"type":66,"tag":675,"props":1305,"children":1306},{},[1307],{"type":66,"tag":97,"props":1308,"children":1310},{"className":1309},[],[1311],{"type":72,"value":1312},"nre\u002F",{"type":66,"tag":675,"props":1314,"children":1315},{},[1316],{"type":72,"value":1317},"The Neural Reconstruction Engine itself. Trains reconstructions, renders frames, exports meshes \u002F point clouds \u002F depth, edits actors, runs the gRPC server, browses results, evaluates quality.",{"type":66,"tag":653,"props":1319,"children":1320},{},[1321,1329,1338],{"type":66,"tag":675,"props":1322,"children":1323},{},[1324],{"type":66,"tag":97,"props":1325,"children":1327},{"className":1326},[],[1328],{"type":72,"value":135},{"type":66,"tag":675,"props":1330,"children":1331},{},[1332],{"type":66,"tag":97,"props":1333,"children":1335},{"className":1334},[],[1336],{"type":72,"value":1337},"asset-harvester\u002F",{"type":66,"tag":675,"props":1339,"children":1340},{},[1341,1343,1349],{"type":72,"value":1342},"Apache-2.0 pipeline that extracts individual 3D objects from sparse driving-clip views as ",{"type":66,"tag":97,"props":1344,"children":1346},{"className":1345},[],[1347],{"type":72,"value":1348},".ply",{"type":72,"value":1350}," Gaussian splats plus metadata.",{"type":66,"tag":653,"props":1352,"children":1353},{},[1354,1362,1371],{"type":66,"tag":675,"props":1355,"children":1356},{},[1357],{"type":66,"tag":97,"props":1358,"children":1360},{"className":1359},[],[1361],{"type":72,"value":146},{"type":66,"tag":675,"props":1363,"children":1364},{},[1365],{"type":66,"tag":97,"props":1366,"children":1368},{"className":1367},[],[1369],{"type":72,"value":1370},"nurec-fixer\u002F",{"type":66,"tag":675,"props":1372,"children":1373},{},[1374],{"type":72,"value":1375},"Standalone DiffusionHarmonizer workflow that cleans up rendered frames, harmonizes inserted actors, evaluates PSNR\u002FLPIPS, and optionally fine-tunes the model.",{"type":66,"tag":187,"props":1377,"children":1379},{"id":1378},"easy-mix-ups",[1380],{"type":72,"value":263},{"type":66,"tag":75,"props":1382,"children":1383},{},[1384],{"type":72,"value":1385},"These pairs sound similar but are different things. When in doubt,\ncome back here.",{"type":66,"tag":89,"props":1387,"children":1388},{},[1389,1405,1449,1491,1553],{"type":66,"tag":93,"props":1390,"children":1391},{},[1392,1397,1399,1404],{"type":66,"tag":153,"props":1393,"children":1394},{},[1395],{"type":72,"value":1396},"NuRec vs NRE.",{"type":72,"value":1398}," NuRec is the product name; NRE is the engine\ninside it. Both map to the same skill: ",{"type":66,"tag":97,"props":1400,"children":1402},{"className":1401},[],[1403],{"type":72,"value":124},{"type":72,"value":247},{"type":66,"tag":93,"props":1406,"children":1407},{},[1408,1413,1418,1419,1424,1426,1431,1433,1439,1441,1447],{"type":66,"tag":153,"props":1409,"children":1410},{},[1411],{"type":72,"value":1412},"NRE's built-in Fixer vs standalone DiffusionHarmonizer.",{"type":66,"tag":97,"props":1414,"children":1416},{"className":1415},[],[1417],{"type":72,"value":1026},{"type":72,"value":1028},{"type":66,"tag":97,"props":1420,"children":1422},{"className":1421},[],[1423],{"type":72,"value":124},{"type":72,"value":1425}," is an inline NRE rendering\nfeature. The ",{"type":66,"tag":97,"props":1427,"children":1429},{"className":1428},[],[1430],{"type":72,"value":146},{"type":72,"value":1432}," skill covers the standalone public\nDiffusionHarmonizer release (code at ",{"type":66,"tag":97,"props":1434,"children":1436},{"className":1435},[],[1437],{"type":72,"value":1438},"NVIDIA\u002Fharmonizer",{"type":72,"value":1440},", model\nat ",{"type":66,"tag":97,"props":1442,"children":1444},{"className":1443},[],[1445],{"type":72,"value":1446},"nvidia\u002FDiffusionHarmonizer",{"type":72,"value":1448},") for frames already on disk,\npaired evaluation, and fine-tuning. Do not assume these two\npaths share cache layout or weights unless the NRE tag's own\ndocs say so.",{"type":66,"tag":93,"props":1450,"children":1451},{},[1452,1468,1470,1475,1477,1482,1484,1489],{"type":66,"tag":153,"props":1453,"children":1454},{},[1455,1460,1462,1467],{"type":66,"tag":97,"props":1456,"children":1458},{"className":1457},[],[1459],{"type":72,"value":113},{"type":72,"value":1461}," vs ",{"type":66,"tag":97,"props":1463,"children":1465},{"className":1464},[],[1466],{"type":72,"value":124},{"type":72,"value":247},{"type":72,"value":1469}," They run ",{"type":66,"tag":153,"props":1471,"children":1472},{},[1473],{"type":72,"value":1474},"in order",{"type":72,"value":1476},", never as\nalternatives. ",{"type":66,"tag":97,"props":1478,"children":1480},{"className":1479},[],[1481],{"type":72,"value":113},{"type":72,"value":1483}," produces the input format; ",{"type":66,"tag":97,"props":1485,"children":1487},{"className":1486},[],[1488],{"type":72,"value":124},{"type":72,"value":1490}," reads it.",{"type":66,"tag":93,"props":1492,"children":1493},{},[1494,1517,1519,1524,1526,1531,1533,1538,1540,1545,1546,1551],{"type":66,"tag":153,"props":1495,"children":1496},{},[1497,1502,1503,1508,1510,1516],{"type":66,"tag":97,"props":1498,"children":1500},{"className":1499},[],[1501],{"type":72,"value":135},{"type":72,"value":1461},{"type":66,"tag":97,"props":1504,"children":1506},{"className":1505},[],[1507],{"type":72,"value":124},{"type":72,"value":1509},"'s ",{"type":66,"tag":97,"props":1511,"children":1513},{"className":1512},[],[1514],{"type":72,"value":1515},"export-external-assets",{"type":72,"value":247},{"type":72,"value":1518}," Asset\nHarvester ",{"type":66,"tag":153,"props":1520,"children":1521},{},[1522],{"type":72,"value":1523},"produces",{"type":72,"value":1525}," the per-object ",{"type":66,"tag":97,"props":1527,"children":1529},{"className":1528},[],[1530],{"type":72,"value":1348},{"type":72,"value":1532}," files. ",{"type":66,"tag":97,"props":1534,"children":1536},{"className":1535},[],[1537],{"type":72,"value":124},{"type":72,"value":1539},"'s\n",{"type":66,"tag":97,"props":1541,"children":1543},{"className":1542},[],[1544],{"type":72,"value":1515},{"type":72,"value":1020},{"type":66,"tag":153,"props":1547,"children":1548},{},[1549],{"type":72,"value":1550},"packages",{"type":72,"value":1552}," them into a USDZ. Always\nAsset Harvester first.",{"type":66,"tag":93,"props":1554,"children":1555},{},[1556,1561,1563,1568,1570,1575,1577,1582],{"type":66,"tag":153,"props":1557,"children":1558},{},[1559],{"type":72,"value":1560},"Cosmos-Drive-Dreams vs PhysicalAI-Autonomous-Vehicles-NuRec.",{"type":72,"value":1562},"\nBoth are AV datasets on Hugging Face; both are managed by\n",{"type":66,"tag":97,"props":1564,"children":1566},{"className":1565},[],[1567],{"type":72,"value":102},{"type":72,"value":1569},". Cosmos-Drive-Dreams is ",{"type":66,"tag":153,"props":1571,"children":1572},{},[1573],{"type":72,"value":1574},"synthetic",{"type":72,"value":1576},"\nweather-augmented video (CC-BY-4.0). The NuRec dataset is\n",{"type":66,"tag":153,"props":1578,"children":1579},{},[1580],{"type":72,"value":1581},"real",{"type":72,"value":1583}," driving scenes turned into renderable USDZs (gated AV\nLicense).",{"type":66,"tag":187,"props":1585,"children":1587},{"id":1586},"prerequisites",[1588],{"type":72,"value":1589},"Prerequisites",{"type":66,"tag":75,"props":1591,"children":1592},{},[1593,1595,1599],{"type":72,"value":1594},"This index is read-only and needs no tooling. The ",{"type":66,"tag":81,"props":1596,"children":1597},{},[1598],{"type":72,"value":1581},{"type":72,"value":1600},"\nprerequisites live in each sibling skill:",{"type":66,"tag":645,"props":1602,"children":1603},{},[1604,1620],{"type":66,"tag":649,"props":1605,"children":1606},{},[1607],{"type":66,"tag":653,"props":1608,"children":1609},{},[1610,1615],{"type":66,"tag":657,"props":1611,"children":1612},{},[1613],{"type":72,"value":1614},"Sibling",{"type":66,"tag":657,"props":1616,"children":1617},{},[1618],{"type":72,"value":1619},"Hard requirements",{"type":66,"tag":668,"props":1621,"children":1622},{},[1623,1647,1668,1689,1727],{"type":66,"tag":653,"props":1624,"children":1625},{},[1626,1634],{"type":66,"tag":675,"props":1627,"children":1628},{},[1629],{"type":66,"tag":97,"props":1630,"children":1632},{"className":1631},[],[1633],{"type":72,"value":113},{"type":66,"tag":675,"props":1635,"children":1636},{},[1637,1639,1645],{"type":72,"value":1638},"Python 3.10+, ",{"type":66,"tag":97,"props":1640,"children":1642},{"className":1641},[],[1643],{"type":72,"value":1644},"pip install nvidia-ncore",{"type":72,"value":1646},"; source data on disk",{"type":66,"tag":653,"props":1648,"children":1649},{},[1650,1658],{"type":66,"tag":675,"props":1651,"children":1652},{},[1653],{"type":66,"tag":97,"props":1654,"children":1656},{"className":1655},[],[1657],{"type":72,"value":124},{"type":66,"tag":675,"props":1659,"children":1660},{},[1661,1663],{"type":72,"value":1662},"Linux x86_64, NVIDIA GPU (Ampere+, ≥24 GB VRAM), Docker 23+, NVIDIA Container Toolkit, ",{"type":66,"tag":97,"props":1664,"children":1666},{"className":1665},[],[1667],{"type":72,"value":493},{"type":66,"tag":653,"props":1669,"children":1670},{},[1671,1679],{"type":66,"tag":675,"props":1672,"children":1673},{},[1674],{"type":66,"tag":97,"props":1675,"children":1677},{"className":1676},[],[1678],{"type":72,"value":135},{"type":66,"tag":675,"props":1680,"children":1681},{},[1682,1684],{"type":72,"value":1683},"Linux + conda, NVIDIA driver ≥570, ~16 GB VRAM, ",{"type":66,"tag":97,"props":1685,"children":1687},{"className":1686},[],[1688],{"type":72,"value":501},{"type":66,"tag":653,"props":1690,"children":1691},{},[1692,1700],{"type":66,"tag":675,"props":1693,"children":1694},{},[1695],{"type":66,"tag":97,"props":1696,"children":1698},{"className":1697},[],[1699],{"type":72,"value":146},{"type":66,"tag":675,"props":1701,"children":1702},{},[1703,1705,1710,1712,1717,1719,1725],{"type":72,"value":1704},"Linux, NVIDIA GPU (Ampere+), Docker, NVIDIA Container Toolkit, ",{"type":66,"tag":97,"props":1706,"children":1708},{"className":1707},[],[1709],{"type":72,"value":501},{"type":72,"value":1711},"; ",{"type":66,"tag":97,"props":1713,"children":1715},{"className":1714},[],[1716],{"type":72,"value":493},{"type":72,"value":1718}," may be needed for ",{"type":66,"tag":97,"props":1720,"children":1722},{"className":1721},[],[1723],{"type":72,"value":1724},"nvcr.io",{"type":72,"value":1726}," pulls",{"type":66,"tag":653,"props":1728,"children":1729},{},[1730,1738],{"type":66,"tag":675,"props":1731,"children":1732},{},[1733],{"type":66,"tag":97,"props":1734,"children":1736},{"className":1735},[],[1737],{"type":72,"value":102},{"type":66,"tag":675,"props":1739,"children":1740},{},[1741,1743,1749,1750,1755],{"type":72,"value":1742},"Python + ",{"type":66,"tag":97,"props":1744,"children":1746},{"className":1745},[],[1747],{"type":72,"value":1748},"huggingface_hub",{"type":72,"value":495},{"type":66,"tag":97,"props":1751,"children":1753},{"className":1752},[],[1754],{"type":72,"value":501},{"type":72,"value":1756}," (gated datasets need license acceptance)",{"type":66,"tag":75,"props":1758,"children":1759},{},[1760,1762,1767,1769,1777],{"type":72,"value":1761},"Each sibling skill ships ",{"type":66,"tag":97,"props":1763,"children":1765},{"className":1764},[],[1766],{"type":72,"value":509},{"type":72,"value":1768}," (where\napplicable) — run it before invoking the workflow. Never echo\nsecret env vars; see\n",{"type":66,"tag":237,"props":1770,"children":1771},{"href":298},[1772],{"type":66,"tag":97,"props":1773,"children":1775},{"className":1774},[],[1776],{"type":72,"value":298},{"type":72,"value":247},{"type":66,"tag":187,"props":1779,"children":1781},{"id":1780},"examples",[1782],{"type":72,"value":1783},"Examples",{"type":66,"tag":75,"props":1785,"children":1786},{},[1787],{"type":72,"value":1788},"Concrete routing examples. The user prompt is on the left; the\ncorrect action this index should take is on the right.",{"type":66,"tag":1790,"props":1791,"children":1793},"h3",{"id":1792},"example-1-single-skill-routing",[1794],{"type":72,"value":1795},"Example 1 — single-skill routing",{"type":66,"tag":1797,"props":1798,"children":1799},"blockquote",{},[1800],{"type":66,"tag":75,"props":1801,"children":1802},{},[1803,1808],{"type":66,"tag":153,"props":1804,"children":1805},{},[1806],{"type":72,"value":1807},"User:",{"type":72,"value":1809}," \"I have a Waymo Open recording. How do I get it into a\nformat NRE accepts?\"",{"type":66,"tag":89,"props":1811,"children":1812},{},[1813,1823],{"type":66,"tag":93,"props":1814,"children":1815},{},[1816,1818,1822],{"type":72,"value":1817},"Match the row \"Convert camera \u002F LiDAR \u002F radar \u002F depth \u002F stereo\ninto NCore V4\" in ",{"type":66,"tag":237,"props":1819,"children":1820},{"href":368},[1821],{"type":72,"value":371},{"type":72,"value":247},{"type":66,"tag":93,"props":1824,"children":1825},{},[1826,1828,1833],{"type":72,"value":1827},"Open the ",{"type":66,"tag":97,"props":1829,"children":1831},{"className":1830},[],[1832],{"type":72,"value":113},{"type":72,"value":1834}," skill next; do not run any commands from this\nindex.",{"type":66,"tag":1790,"props":1836,"children":1838},{"id":1837},"example-2-multi-stage-pipeline",[1839],{"type":72,"value":1840},"Example 2 — multi-stage pipeline",{"type":66,"tag":1797,"props":1842,"children":1843},{},[1844],{"type":66,"tag":75,"props":1845,"children":1846},{},[1847,1851],{"type":66,"tag":153,"props":1848,"children":1849},{},[1850],{"type":72,"value":1807},{"type":72,"value":1852}," \"I have a driving clip. I want to train NuRec and then\nrender a new camera trajectory through it.\"",{"type":66,"tag":89,"props":1854,"children":1855},{},[1856,1874,1888],{"type":66,"tag":93,"props":1857,"children":1858},{},[1859,1861,1866,1868,1873],{"type":72,"value":1860},"Match \"Train a 3D reconstruction from an NCore clip\" → ",{"type":66,"tag":97,"props":1862,"children":1864},{"className":1863},[],[1865],{"type":72,"value":113},{"type":72,"value":1867},"\n→ ",{"type":66,"tag":97,"props":1869,"children":1871},{"className":1870},[],[1872],{"type":72,"value":124},{"type":72,"value":247},{"type":66,"tag":93,"props":1875,"children":1876},{},[1877,1879,1887],{"type":72,"value":1878},"Cross-check workflow A in\n",{"type":66,"tag":237,"props":1880,"children":1881},{"href":239},[1882],{"type":66,"tag":97,"props":1883,"children":1885},{"className":1884},[],[1886],{"type":72,"value":239},{"type":72,"value":247},{"type":66,"tag":93,"props":1889,"children":1890},{},[1891,1893,1898,1900,1905],{"type":72,"value":1892},"Open ",{"type":66,"tag":97,"props":1894,"children":1896},{"className":1895},[],[1897],{"type":72,"value":113},{"type":72,"value":1899}," first, then ",{"type":66,"tag":97,"props":1901,"children":1903},{"className":1902},[],[1904],{"type":72,"value":124},{"type":72,"value":1906},", in that order.",{"type":66,"tag":1790,"props":1908,"children":1910},{"id":1909},"example-3-skip-training-just-render",[1911],{"type":72,"value":1912},"Example 3 — skip training, just render",{"type":66,"tag":1797,"props":1914,"children":1915},{},[1916],{"type":66,"tag":75,"props":1917,"children":1918},{},[1919,1923],{"type":66,"tag":153,"props":1920,"children":1921},{},[1922],{"type":72,"value":1807},{"type":72,"value":1924}," \"Can I just see NuRec working on a scene NVIDIA already\nbuilt?\"",{"type":66,"tag":89,"props":1926,"children":1927},{},[1928,1945,1950],{"type":66,"tag":93,"props":1929,"children":1930},{},[1931,1933,1938,1939,1944],{"type":72,"value":1932},"Match \"Skip training and just render an NVIDIA-built driving\nscene\" → ",{"type":66,"tag":97,"props":1934,"children":1936},{"className":1935},[],[1937],{"type":72,"value":102},{"type":72,"value":737},{"type":66,"tag":97,"props":1940,"children":1942},{"className":1941},[],[1943],{"type":72,"value":124},{"type":72,"value":247},{"type":66,"tag":93,"props":1946,"children":1947},{},[1948],{"type":72,"value":1949},"Cross-check workflow B.",{"type":66,"tag":93,"props":1951,"children":1952},{},[1953,1954,1959,1961,1966,1968,1973],{"type":72,"value":1892},{"type":66,"tag":97,"props":1955,"children":1957},{"className":1956},[],[1958],{"type":72,"value":102},{"type":72,"value":1960}," first to download ",{"type":66,"tag":153,"props":1962,"children":1963},{},[1964],{"type":72,"value":1965},"one",{"type":72,"value":1967}," scene\n(~1.5–2 GB), then ",{"type":66,"tag":97,"props":1969,"children":1971},{"className":1970},[],[1972],{"type":72,"value":124},{"type":72,"value":1974}," to render.",{"type":66,"tag":1790,"props":1976,"children":1978},{"id":1977},"example-4-actor-editing",[1979],{"type":72,"value":1980},"Example 4 — actor editing",{"type":66,"tag":1797,"props":1982,"children":1983},{},[1984],{"type":66,"tag":75,"props":1985,"children":1986},{},[1987,1991],{"type":66,"tag":153,"props":1988,"children":1989},{},[1990],{"type":72,"value":1807},{"type":72,"value":1992}," \"I want to add a pedestrian to this NuRec scene.\"",{"type":66,"tag":89,"props":1994,"children":1995},{},[1996,2013,2018],{"type":66,"tag":93,"props":1997,"children":1998},{},[1999,2001,2006,2007,2012],{"type":72,"value":2000},"Match \"Add, remove, or replace cars \u002F pedestrians\" →\n",{"type":66,"tag":97,"props":2002,"children":2004},{"className":2003},[],[2005],{"type":72,"value":135},{"type":72,"value":737},{"type":66,"tag":97,"props":2008,"children":2010},{"className":2009},[],[2011],{"type":72,"value":124},{"type":72,"value":247},{"type":66,"tag":93,"props":2014,"children":2015},{},[2016],{"type":72,"value":2017},"Cross-check workflow D.",{"type":66,"tag":93,"props":2019,"children":2020},{},[2021,2023,2028,2030,2035,2037,2043],{"type":72,"value":2022},"Confirm the original NCore clip is on disk first (Asset\nHarvester needs it), then open ",{"type":66,"tag":97,"props":2024,"children":2026},{"className":2025},[],[2027],{"type":72,"value":135},{"type":72,"value":2029},", then ",{"type":66,"tag":97,"props":2031,"children":2033},{"className":2032},[],[2034],{"type":72,"value":124},{"type":72,"value":2036},"\nwith ",{"type":66,"tag":97,"props":2038,"children":2040},{"className":2039},[],[2041],{"type":72,"value":2042},"serve-grpc --enable-editing-actors",{"type":72,"value":247},{"type":66,"tag":1790,"props":2045,"children":2047},{"id":2046},"example-5-ambiguous-fix-it-request",[2048],{"type":72,"value":2049},"Example 5 — ambiguous \"fix it\" request",{"type":66,"tag":1797,"props":2051,"children":2052},{},[2053],{"type":66,"tag":75,"props":2054,"children":2055},{},[2056,2060],{"type":66,"tag":153,"props":2057,"children":2058},{},[2059],{"type":72,"value":1807},{"type":72,"value":2061}," \"My rendered frames look fuzzy with weird floaters.\nCan you clean them up?\"",{"type":66,"tag":89,"props":2063,"children":2064},{},[2065,2111],{"type":66,"tag":93,"props":2066,"children":2067},{},[2068,2070],{"type":72,"value":2069},"Match the \"Clean up rendered frames\" row. Two valid paths:\n",{"type":66,"tag":89,"props":2071,"children":2072},{},[2073,2095],{"type":66,"tag":93,"props":2074,"children":2075},{},[2076,2081,2082,2087,2088,2093],{"type":66,"tag":153,"props":2077,"children":2078},{},[2079],{"type":72,"value":2080},"Quick path:",{"type":72,"value":1020},{"type":66,"tag":97,"props":2083,"children":2085},{"className":2084},[],[2086],{"type":72,"value":1026},{"type":72,"value":1028},{"type":66,"tag":97,"props":2089,"children":2091},{"className":2090},[],[2092],{"type":72,"value":124},{"type":72,"value":2094}," if the user is\nalready rendering through NRE.",{"type":66,"tag":93,"props":2096,"children":2097},{},[2098,2103,2104,2109],{"type":66,"tag":153,"props":2099,"children":2100},{},[2101],{"type":72,"value":2102},"Standalone path:",{"type":72,"value":1020},{"type":66,"tag":97,"props":2105,"children":2107},{"className":2106},[],[2108],{"type":72,"value":146},{"type":72,"value":2110}," for already-rendered\nframes on disk or for paired evaluation \u002F fine-tuning.",{"type":66,"tag":93,"props":2112,"children":2113},{},[2114],{"type":72,"value":2115},"Ask which interface the user wants, then open the matching\nskill. See workflow E for details.",{"type":66,"tag":1790,"props":2117,"children":2119},{"id":2118},"example-6-sibling-skill-not-on-disk",[2120],{"type":72,"value":2121},"Example 6 — sibling skill not on disk",{"type":66,"tag":1797,"props":2123,"children":2124},{},[2125],{"type":66,"tag":75,"props":2126,"children":2127},{},[2128,2132,2134,2139],{"type":66,"tag":153,"props":2129,"children":2130},{},[2131],{"type":72,"value":1807},{"type":72,"value":2133}," \"Where do I find the ",{"type":66,"tag":97,"props":2135,"children":2137},{"className":2136},[],[2138],{"type":72,"value":124},{"type":72,"value":2140}," skill? It is not in my\nrepo.\"",{"type":66,"tag":89,"props":2142,"children":2143},{},[2144],{"type":66,"tag":93,"props":2145,"children":2146},{},[2147,2149,2157,2159,2165,2166,2172,2174,2180,2181,2187,2189,2194],{"type":72,"value":2148},"Do not guess a path. Follow\n",{"type":66,"tag":237,"props":2150,"children":2151},{"href":277},[2152],{"type":66,"tag":97,"props":2153,"children":2155},{"className":2154},[],[2156],{"type":72,"value":277},{"type":72,"value":2158},": look under\n",{"type":66,"tag":97,"props":2160,"children":2162},{"className":2161},[],[2163],{"type":72,"value":2164},".agents\u002Fskills\u002Fnre\u002FSKILL.md",{"type":72,"value":2029},{"type":66,"tag":97,"props":2167,"children":2169},{"className":2168},[],[2170],{"type":72,"value":2171},".claude\u002Fskills\u002Fnre\u002F",{"type":72,"value":2173},", then\n",{"type":66,"tag":97,"props":2175,"children":2177},{"className":2176},[],[2178],{"type":72,"value":2179},".cursor\u002Fskills\u002Fnre\u002F",{"type":72,"value":2029},{"type":66,"tag":97,"props":2182,"children":2184},{"className":2183},[],[2185],{"type":72,"value":2186},"~\u002F.cursor\u002Fskills\u002Fnre\u002F",{"type":72,"value":2188},", and as a\nlast resort clone\n",{"type":66,"tag":97,"props":2190,"children":2192},{"className":2191},[],[2193],{"type":72,"value":21},{"type":72,"value":247},{"type":66,"tag":187,"props":2196,"children":2198},{"id":2197},"limitations",[2199],{"type":72,"value":2200},"Limitations",{"type":66,"tag":89,"props":2202,"children":2203},{},[2204,2214,2229,2247],{"type":66,"tag":93,"props":2205,"children":2206},{},[2207,2212],{"type":66,"tag":153,"props":2208,"children":2209},{},[2210],{"type":72,"value":2211},"Routes only",{"type":72,"value":2213},"; never runs pipelines. For hands-on steps, open\nthe sibling skill this index points to.",{"type":66,"tag":93,"props":2215,"children":2216},{},[2217,2222,2224,2228],{"type":66,"tag":153,"props":2218,"children":2219},{},[2220],{"type":72,"value":2221},"Hand-curated catalogue.",{"type":72,"value":2223}," A newly-added sibling skill is not\ndiscoverable here until someone updates\n",{"type":66,"tag":237,"props":2225,"children":2226},{"href":368},[2227],{"type":72,"value":371},{"type":72,"value":247},{"type":66,"tag":93,"props":2230,"children":2231},{},[2232,2237,2239,2245],{"type":66,"tag":153,"props":2233,"children":2234},{},[2235],{"type":72,"value":2236},"Names are portable, paths are not.",{"type":72,"value":2238}," Cross-skill links assume\nthe canonical layout (",{"type":66,"tag":97,"props":2240,"children":2242},{"className":2241},[],[2243],{"type":72,"value":2244},".agents\u002Fskills\u002F\u003Cname>\u002FSKILL.md",{"type":72,"value":2246},"). In a\ndifferent runtime layout, prefer name-based skill resolution\nover file paths.",{"type":66,"tag":93,"props":2248,"children":2249},{},[2250,2255],{"type":66,"tag":153,"props":2251,"children":2252},{},[2253],{"type":72,"value":2254},"No Omniverse \u002F Isaac Sim integration steps",{"type":72,"value":2256}," — those live in\nupstream Omniverse \u002F Isaac docs, not in the NuRec skill family.",{"type":66,"tag":187,"props":2258,"children":2260},{"id":2259},"troubleshooting",[2261],{"type":72,"value":2262},"Troubleshooting",{"type":66,"tag":645,"props":2264,"children":2265},{},[2266,2287],{"type":66,"tag":649,"props":2267,"children":2268},{},[2269],{"type":66,"tag":653,"props":2270,"children":2271},{},[2272,2277,2282],{"type":66,"tag":657,"props":2273,"children":2274},{},[2275],{"type":72,"value":2276},"Symptom",{"type":66,"tag":657,"props":2278,"children":2279},{},[2280],{"type":72,"value":2281},"Likely cause",{"type":66,"tag":657,"props":2283,"children":2284},{},[2285],{"type":72,"value":2286},"Resolution",{"type":66,"tag":668,"props":2288,"children":2289},{},[2290,2334,2361,2396,2424],{"type":66,"tag":653,"props":2291,"children":2292},{},[2293,2298,2303],{"type":66,"tag":675,"props":2294,"children":2295},{},[2296],{"type":72,"value":2297},"Agent picked the wrong sibling skill",{"type":66,"tag":675,"props":2299,"children":2300},{},[2301],{"type":72,"value":2302},"The user's task spans multiple stages (e.g. convert + train)",{"type":66,"tag":675,"props":2304,"children":2305},{},[2306,2308,2312,2314,2319,2321,2326,2328,2333],{"type":72,"value":2307},"Re-read ",{"type":66,"tag":237,"props":2309,"children":2310},{"href":368},[2311],{"type":72,"value":371},{"type":72,"value":2313}," and follow the arrows; multi-stage tasks usually start with ",{"type":66,"tag":97,"props":2315,"children":2317},{"className":2316},[],[2318],{"type":72,"value":113},{"type":72,"value":2320}," or ",{"type":66,"tag":97,"props":2322,"children":2324},{"className":2323},[],[2325],{"type":72,"value":102},{"type":72,"value":2327},", then hand off to ",{"type":66,"tag":97,"props":2329,"children":2331},{"className":2330},[],[2332],{"type":72,"value":124},{"type":72,"value":247},{"type":66,"tag":653,"props":2335,"children":2336},{},[2337,2342,2347],{"type":66,"tag":675,"props":2338,"children":2339},{},[2340],{"type":72,"value":2341},"Sibling skill not found on disk",{"type":66,"tag":675,"props":2343,"children":2344},{},[2345],{"type":72,"value":2346},"The host repo only has the index",{"type":66,"tag":675,"props":2348,"children":2349},{},[2350,2352,2360],{"type":72,"value":2351},"Follow ",{"type":66,"tag":237,"props":2353,"children":2354},{"href":277},[2355],{"type":66,"tag":97,"props":2356,"children":2358},{"className":2357},[],[2359],{"type":72,"value":277},{"type":72,"value":247},{"type":66,"tag":653,"props":2362,"children":2363},{},[2364,2375,2385],{"type":66,"tag":675,"props":2365,"children":2366},{},[2367,2369],{"type":72,"value":2368},"Stale link to ",{"type":66,"tag":97,"props":2370,"children":2372},{"className":2371},[],[2373],{"type":72,"value":2374},"ncore-data-conversion",{"type":66,"tag":675,"props":2376,"children":2377},{},[2378,2380],{"type":72,"value":2379},"Older snapshots used that name; the skill is now ",{"type":66,"tag":97,"props":2381,"children":2383},{"className":2382},[],[2384],{"type":72,"value":113},{"type":66,"tag":675,"props":2386,"children":2387},{},[2388,2390,2395],{"type":72,"value":2389},"Update the link to ",{"type":66,"tag":97,"props":2391,"children":2393},{"className":2392},[],[2394],{"type":72,"value":113},{"type":72,"value":247},{"type":66,"tag":653,"props":2397,"children":2398},{},[2399,2404,2409],{"type":66,"tag":675,"props":2400,"children":2401},{},[2402],{"type":72,"value":2403},"User wants to delete disk artifacts",{"type":66,"tag":675,"props":2405,"children":2406},{},[2407],{"type":72,"value":2408},"NuRec workflow caches grow large",{"type":66,"tag":675,"props":2410,"children":2411},{},[2412,2414,2422],{"type":72,"value":2413},"Walk ",{"type":66,"tag":237,"props":2415,"children":2416},{"href":298},[2417],{"type":66,"tag":97,"props":2418,"children":2420},{"className":2419},[],[2421],{"type":72,"value":298},{"type":72,"value":2423}," in the documented order.",{"type":66,"tag":653,"props":2425,"children":2426},{},[2427,2432,2437],{"type":66,"tag":675,"props":2428,"children":2429},{},[2430],{"type":72,"value":2431},"User asks \"should I retrain or just clean up frames?\"",{"type":66,"tag":675,"props":2433,"children":2434},{},[2435],{"type":72,"value":2436},"Conflating reconstruction vs post-processing",{"type":66,"tag":675,"props":2438,"children":2439},{},[2440,2442,2447,2449,2454],{"type":72,"value":2441},"Retrain → ",{"type":66,"tag":97,"props":2443,"children":2445},{"className":2444},[],[2446],{"type":72,"value":124},{"type":72,"value":2448},"; clean already-rendered frames → ",{"type":66,"tag":97,"props":2450,"children":2452},{"className":2451},[],[2453],{"type":72,"value":146},{"type":72,"value":247},{"type":66,"tag":187,"props":2456,"children":2458},{"id":2457},"references",[2459],{"type":72,"value":2460},"References",{"type":66,"tag":75,"props":2462,"children":2463},{},[2464],{"type":72,"value":2465},"Detailed material that this index intentionally keeps out of the\nhot path. Read only when the matching section above points there.",{"type":66,"tag":89,"props":2467,"children":2468},{},[2469,2482,2495],{"type":66,"tag":93,"props":2470,"children":2471},{},[2472,2480],{"type":66,"tag":237,"props":2473,"children":2474},{"href":239},[2475],{"type":66,"tag":97,"props":2476,"children":2478},{"className":2477},[],[2479],{"type":72,"value":239},{"type":72,"value":2481}," — full\nstep-by-step multi-skill workflows A–G.",{"type":66,"tag":93,"props":2483,"children":2484},{},[2485,2493],{"type":66,"tag":237,"props":2486,"children":2487},{"href":298},[2488],{"type":66,"tag":97,"props":2489,"children":2491},{"className":2490},[],[2492],{"type":72,"value":298},{"type":72,"value":2494}," — disk\ncleanup order across all five siblings plus secrets-handling\npolicy.",{"type":66,"tag":93,"props":2496,"children":2497},{},[2498,2506],{"type":66,"tag":237,"props":2499,"children":2500},{"href":277},[2501],{"type":66,"tag":97,"props":2502,"children":2504},{"className":2503},[],[2505],{"type":72,"value":277},{"type":72,"value":2507}," — how to\nlocate or fetch a sibling skill that is not already on disk;\nthin-local-skill policy and upstream links.",{"type":66,"tag":187,"props":2509,"children":2511},{"id":2510},"keeping-this-index-up-to-date",[2512],{"type":72,"value":338},{"type":66,"tag":75,"props":2514,"children":2515},{},[2516,2518,2523,2525,2531],{"type":72,"value":2517},"This index is ",{"type":66,"tag":153,"props":2519,"children":2520},{},[2521],{"type":72,"value":2522},"hand-curated",{"type":72,"value":2524}," — it groups skills by what users\nwant to do, not alphabetically. There is no generator script; edit\nthis ",{"type":66,"tag":97,"props":2526,"children":2528},{"className":2527},[],[2529],{"type":72,"value":2530},"SKILL.md",{"type":72,"value":2532}," by hand whenever the sibling set changes.",{"type":66,"tag":75,"props":2534,"children":2535},{},[2536],{"type":72,"value":2537},"When a new sibling skill is added or a use case shifts:",{"type":66,"tag":353,"props":2539,"children":2540},{},[2541,2552,2563,2577],{"type":66,"tag":93,"props":2542,"children":2543},{},[2544,2546,2550],{"type":72,"value":2545},"Add a row to ",{"type":66,"tag":237,"props":2547,"children":2548},{"href":368},[2549],{"type":72,"value":371},{"type":72,"value":2551}," for the new use\ncase.",{"type":66,"tag":93,"props":2553,"children":2554},{},[2555,2557,2562],{"type":72,"value":2556},"Add a row to\n",{"type":66,"tag":237,"props":2558,"children":2560},{"href":2559},"#the-skills-in-this-folder",[2561],{"type":72,"value":1198},{"type":72,"value":247},{"type":66,"tag":93,"props":2564,"children":2565},{},[2566,2568,2576],{"type":72,"value":2567},"If the new skill changes a multi-step pipeline, update\n",{"type":66,"tag":237,"props":2569,"children":2570},{"href":239},[2571],{"type":66,"tag":97,"props":2572,"children":2574},{"className":2573},[],[2575],{"type":72,"value":239},{"type":72,"value":247},{"type":66,"tag":93,"props":2578,"children":2579},{},[2580,2582,2588],{"type":72,"value":2581},"Confirm the sibling's ",{"type":66,"tag":97,"props":2583,"children":2585},{"className":2584},[],[2586],{"type":72,"value":2587},"metadata.upstream",{"type":72,"value":2589}," field still points at\nthe canonical upstream repo or container.",{"type":66,"tag":75,"props":2591,"children":2592},{},[2593],{"type":72,"value":2594},"Otherwise the index will quietly drift and beginners will end up\nreading the wrong skill.",{"items":2596,"total":2755},[2597,2615,2633,2644,2656,2670,2683,2697,2710,2721,2735,2744],{"slug":2598,"name":2598,"fn":2599,"description":2600,"org":2601,"tags":2602,"stars":2612,"repoUrl":2613,"updatedAt":2614},"nemoclaw-user-guide","retrieve NemoClaw documentation and configuration","Guides human users' AI agents to the NemoClaw docs MCP server and canonical Fern documentation in Markdown form. Use when users ask how to install, configure, operate, troubleshoot, secure, or learn NemoClaw with an AI coding assistant. Trigger keywords - nemoclaw docs, use nemoclaw with ai agent, nemoclaw mcp docs, nemoclaw install help, nemoclaw quickstart, nemoclaw markdown docs, llms.txt, agent skills.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2603,2606,2609],{"name":2604,"slug":2605,"type":15},"Documentation","documentation",{"name":2607,"slug":2608,"type":15},"MCP","mcp",{"name":2610,"slug":2611,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":2616,"name":2616,"fn":2617,"description":2618,"org":2619,"tags":2620,"stars":2630,"repoUrl":2631,"updatedAt":2632},"mcore-build-and-dependency","manage Megatron-LM development environments","Container-based dev environment setup and dependency management for Megatron-LM. Covers acquiring and launching the CI container, uv package management, and updating uv.lock.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2621,2624,2627],{"name":2622,"slug":2623,"type":15},"Containers","containers",{"name":2625,"slug":2626,"type":15},"Deployment","deployment",{"name":2628,"slug":2629,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":2634,"name":2634,"fn":2635,"description":2636,"org":2637,"tags":2638,"stars":2630,"repoUrl":2631,"updatedAt":2643},"mcore-bump-base-image","update NVIDIA PyTorch base images","Bump the NVIDIA PyTorch base image (`nvcr.io\u002Fnvidia\u002Fpytorch:YY.MM-py3`) used by Megatron-LM CI. Covers the two pin sites (GitHub CI in `docker\u002F.ngc_version.dev` and GitLab CI in `.gitlab\u002Fstages\u002F01.build.yml`), the post-bump CI loop (re-run functional tests, refresh golden values, mark broken tests), and the gotchas that bit PRs",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2639,2642],{"name":2640,"slug":2641,"type":15},"CI\u002FCD","ci-cd",{"name":2625,"slug":2626,"type":15},"2026-07-14T05:25:59.97109",{"slug":2645,"name":2645,"fn":2646,"description":2647,"org":2648,"tags":2649,"stars":2630,"repoUrl":2631,"updatedAt":2655},"mcore-cicd","manage CI\u002FCD pipelines for Megatron-LM","CI\u002FCD reference for Megatron-LM. Covers CI pipeline structure, PR scope labels, triggering internal GitLab CI (which force-pushes the current branch to a pull-request\u002FBRANCH ref — always dry-run and verify the destination first; never run against shared or protected branches), and CI failure investigation.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2650,2651,2652],{"name":2640,"slug":2641,"type":15},{"name":2625,"slug":2626,"type":15},{"name":2653,"slug":2654,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":2657,"name":2657,"fn":2658,"description":2659,"org":2660,"tags":2661,"stars":2630,"repoUrl":2631,"updatedAt":2669},"mcore-create-issue","investigate CI failures and create issues","Investigate a failing GitHub Actions run or job and create a GitHub issue for the failure.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2662,2665,2666],{"name":2663,"slug":2664,"type":15},"Debugging","debugging",{"name":2653,"slug":2654,"type":15},{"name":2667,"slug":2668,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":2671,"name":2671,"fn":2672,"description":2673,"org":2674,"tags":2675,"stars":2630,"repoUrl":2631,"updatedAt":2682},"mcore-linting-and-formatting","lint and format Megatron-LM code","Linting and formatting for Megatron-LM. Covers running autoformat.sh, tools (ruff, black, isort, pylint, mypy), and code style rules.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2676,2679],{"name":2677,"slug":2678,"type":15},"Best Practices","best-practices",{"name":2680,"slug":2681,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":2684,"name":2684,"fn":2685,"description":2686,"org":2687,"tags":2688,"stars":2630,"repoUrl":2631,"updatedAt":2696},"mcore-migrate-gpt-to-hybrid","migrate Megatron-LM models to HybridModel","Migration guide for moving Megatron Core GPTModel checkpoints, model providers, training commands, and layer mappings to HybridModel.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2689,2692,2695],{"name":2690,"slug":2691,"type":15},"Machine Learning","machine-learning",{"name":2693,"slug":2694,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":2698,"name":2698,"fn":2699,"description":2700,"org":2701,"tags":2702,"stars":2630,"repoUrl":2631,"updatedAt":2709},"mcore-onboard-gb200-1node-tests","onboard functional tests for GB200","Onboard 1-node GitHub MR functional tests for GB200 from existing mr-scoped 2-node tests.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2703,2706],{"name":2704,"slug":2705,"type":15},"QA","qa",{"name":2707,"slug":2708,"type":15},"Testing","testing","2026-07-14T05:25:53.673039",{"slug":2711,"name":2711,"fn":2712,"description":2713,"org":2714,"tags":2715,"stars":2630,"repoUrl":2631,"updatedAt":2720},"mcore-run-on-slurm","launch distributed training jobs on SLURM","How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-rank failure diagnosis.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2716,2717],{"name":2625,"slug":2626,"type":15},{"name":2718,"slug":2719,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":2722,"name":2722,"fn":2723,"description":2724,"org":2725,"tags":2726,"stars":2630,"repoUrl":2631,"updatedAt":2734},"mcore-split-pr","split pull requests to reduce review load","Split a PR into multiple PRs to reduce the number of required CODEOWNERS reviewer groups.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2727,2730,2731],{"name":2728,"slug":2729,"type":15},"Code Review","code-review",{"name":2653,"slug":2654,"type":15},{"name":2732,"slug":2733,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":2736,"name":2736,"fn":2737,"description":2738,"org":2739,"tags":2740,"stars":2630,"repoUrl":2631,"updatedAt":2743},"mcore-testing","run and manage Megatron-LM tests","Test system for Megatron-LM. Covers test layout, recipe YAML structure, adding and running unit and functional tests, golden values, marker filters, and CI parity.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2741,2742],{"name":2704,"slug":2705,"type":15},{"name":2707,"slug":2708,"type":15},"2026-07-14T05:25:54.928983",{"slug":2745,"name":2745,"fn":2746,"description":2747,"org":2748,"tags":2749,"stars":2630,"repoUrl":2631,"updatedAt":2754},"nightly-sync","manage nightly main-to-dev sync workflows","Domain knowledge for the nightly main-to-dev sync workflow. Covers merge strategy, CI architecture, failure investigation, and known issues.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2750,2753],{"name":2751,"slug":2752,"type":15},"Automation","automation",{"name":2640,"slug":2641,"type":15},"2026-07-30T05:29:03.275638",496,{"items":2757,"total":2825},[2758,2770,2784,2794,2804,2810],{"slug":135,"name":135,"fn":2759,"description":2760,"org":2761,"tags":2762,"stars":20,"repoUrl":21,"updatedAt":2769},"extract 3D Gaussian assets with Asset Harvester","Use to install and run NVIDIA Asset Harvester (Apache-2.0) to extract per-object 3D Gaussian Splat assets (`gaussians.ply`) from AV NCore V4 clips or masked single images via SparseViewDiT + TokenGS, optionally producing `metadata.yaml` for NuRec object insertion. Do NOT use for full-scene reconstruction (use `nre`) or for inputs without per-object masks.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2763,2764,2767,2768],{"name":13,"slug":14,"type":15},{"name":2765,"slug":2766,"type":15},"Computer Vision","computer-vision",{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},"2026-07-14T05:32:37.216976",{"slug":113,"name":113,"fn":2771,"description":2772,"org":2773,"tags":2774,"stars":20,"repoUrl":21,"updatedAt":2783},"convert sensor datasets to NVIDIA NCore format","Use when converting any sensor dataset into NVIDIA NCore V4 format (and feeding it to NuRec or a robotics-to-sim \"r2s\" pipeline). Covers ingesting raw cameras, LiDARs, radars, IMUs, depth or stereo into V4 sequences; authoring a new converter from the template; adapting PAI \u002F Waymo \u002F PandaSet \u002F NuScenes to V4; handling non-AV rigs (mono+depth, mono+lidar, stereo, multi-stereo, RGB-D, COLMAP \u002F SfM, ROS2 bag); and diagnosing a broken converter against `validate.py`. Do NOT use to train reconstructions (use `nre`) or to extract per-object 3D assets (use `asset-harvester`). Trigger keywords: ncore, ncore v4, convert, ingest, zarr, itar, nurec, waymo, pandaset, nuscenes, pai, hyperion, colmap, scannetpp, stereo, multi-stereo, mono+depth, mono+lidar, kitti, sfm, camera, lidar, radar, imu, cuboid, poses, intrinsics, ego mask, ros2, rosbag, mcap, realsense, zed, rgb-d, r2s, robotics, sam2.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2775,2778,2779,2782],{"name":2776,"slug":2777,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":2780,"slug":2781,"type":15},"Robotics","robotics",{"name":18,"slug":19,"type":15},"2026-07-14T05:32:35.853952",{"slug":124,"name":124,"fn":2785,"description":2786,"org":2787,"tags":2788,"stars":20,"repoUrl":21,"updatedAt":2793},"train 3D Gaussian Splat models with NRE","Use to drive NVIDIA Omniverse NuRec \u002F Neural Reconstruction Engine (NRE) via the public NGC containers nvcr.io\u002Fnvidia\u002Fnre\u002Fnre and nvcr.io\u002Fnvidia\u002Fnre\u002Fnre-tools (NGC_API_KEY required) — train 3DGUT Gaussian reconstructions from NCore clips, generate aux data, render frames or LiDAR sweeps (local or warm `serve-grpc`), export PLY\u002Fdepth\u002Fmesh\u002FUSDZ, edit actors, and evaluate metrics. Do NOT use for per-object asset capture (use `asset-harvester`) or sensor-to-NCore conversion (use `ncore`).",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2789,2790,2791,2792],{"name":13,"slug":14,"type":15},{"name":2690,"slug":2691,"type":15},{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},"2026-07-14T05:32:34.548655",{"slug":146,"name":146,"fn":2795,"description":2796,"org":2797,"tags":2798,"stars":20,"repoUrl":21,"updatedAt":2803},"harmonize and fine-tune 3D reconstruction frames","Use to run NVIDIA DiffusionHarmonizer (public successor to the older Fixer recipes) to enhance, harmonize, evaluate, or fine-tune novel-view frames from NRE \u002F NuRec \u002F 3DGS \u002F NeRF reconstructions. Do NOT use for training the 3D reconstruction itself (use `nre`) or for sensor-to-NCore conversion (use `ncore`).",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2799,2800,2801,2802],{"name":13,"slug":14,"type":15},{"name":2690,"slug":2691,"type":15},{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},"2026-07-14T05:32:38.465728",{"slug":4,"name":4,"fn":5,"description":6,"org":2805,"tags":2806,"stars":20,"repoUrl":21,"updatedAt":22},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2807,2808,2809],{"name":13,"slug":14,"type":15},{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},{"slug":102,"name":102,"fn":2811,"description":2812,"org":2813,"tags":2814,"stars":20,"repoUrl":21,"updatedAt":2824},"find and download NVIDIA Physical AI datasets","Use when the user wants to find, download, or pick a NVIDIA Physical AI dataset on Hugging Face for autonomous-vehicle, robotics, spatial intelligence, manipulation, or neural-reconstruction workflows. Catalog of every dataset under huggingface.co\u002Fnvidia with the `PhysicalAI-` prefix, organised by domain (AV, Robotics-Manipulation, Robotics-GR00T, Robotics-mindmap, Robotics-NuRec, Spatial Intelligence, Grasping, Healthcare, Sim-Ready, Material properties), with per-dataset size, format, gating, license, and the downstream sibling skill (`ncore`, `nre`, `asset-harvester`, `nurec-fixer`) or upstream tool (Isaac Sim, CARLA, Isaac-GR00T, Cosmos-*) that consumes it. Do NOT use as a runtime — it routes you elsewhere. Trigger keywords: nvidia physical ai dataset, PhysicalAI- dataset, hf nvidia dataset, NCore dataset, NuRec dataset, GR00T dataset, GraspGen, SimReady, Cosmos-Drive-Dreams, Lyra SDG, Open-H-Embodiment, huggingface-cli download, physical_ai_av, dataset gated, RDS-HQ.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2815,2818,2821,2822,2823],{"name":2816,"slug":2817,"type":15},"Datasets","datasets",{"name":2819,"slug":2820,"type":15},"Hugging Face","hugging-face",{"name":2690,"slug":2691,"type":15},{"name":9,"slug":8,"type":15},{"name":2780,"slug":2781,"type":15},"2026-07-14T05:32:33.182445",6]