[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-physical-ai-defect-image-generation":3,"mdc-ibanaq-key":34,"related-org-nvidia-physical-ai-defect-image-generation":3219,"related-repo-nvidia-physical-ai-defect-image-generation":3378},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":23,"repoUrl":24,"updatedAt":25,"license":26,"forks":27,"topics":28,"repo":29,"sourceUrl":32,"mdContent":33},"physical-ai-defect-image-generation","generate defect inspection images with NVIDIA Cosmos","Use when the user wants to orchestrate defect image generation with NVIDIA Cosmos AnomalyGen (Cosmos-Predict2-derived) on OSMO for PCBA, metal surface, and glass inspection. The Day 0 path handles cold-start with USD-to-ROI, image-edit augmentation, and AnomalyGen to create initial PCBA datasets. The Day 1 path performs inference and labeling on real images. This skill helps with first-time asset setup, creation of finetuning checkpoints, and configuring deployment.\nTrigger keywords: defect image generation, dig workflow, dig pipeline, defect image detection workflow, aoi pipeline, aoi anomalygen, usd2roi anomalygen, day 0 pcba, day 1 pcba, day 1 real-photo alignment, day 1 manual roi, metal surface anomaly, glass defect, anomalygen finetune, setup_pcb, setup_metal, setup_glass, setup_pretrained, dig setup, dig datasets, dig pretrained checkpoint, dig image-edit endpoint, cosmos defect generation, cosmos-predict2 defect, cosmos-anomalygen, cosmos predict2 finetune.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,17,20],{"name":13,"slug":14,"type":15},"Automation","automation","tag",{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},"Computer Vision","computer-vision",{"name":21,"slug":22,"type":15},"Image Generation","image-generation",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-14T05:26:15.105577","CC-BY-4.0 AND Apache-2.0",281,[],{"repoUrl":24,"stars":23,"forks":27,"topics":30,"description":31},[],"AI agent skills published by NVIDIA","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills\u002Ftree\u002FHEAD\u002Fskills\u002Fphysical-ai-defect-image-generation","---\nname: physical-ai-defect-image-generation\ndescription: >-\n  Use when the user wants to orchestrate defect image generation with NVIDIA Cosmos AnomalyGen (Cosmos-Predict2-derived) on OSMO for PCBA, metal surface, and glass inspection. The Day 0 path handles cold-start with USD-to-ROI, image-edit augmentation, and AnomalyGen to create initial PCBA datasets. The Day 1 path performs inference and labeling on real images. This skill helps with first-time asset setup, creation of finetuning checkpoints, and configuring deployment.\n\n  Trigger keywords: defect image generation, dig workflow, dig pipeline, defect image detection workflow, aoi pipeline, aoi anomalygen, usd2roi anomalygen, day 0 pcba, day 1 pcba, day 1 real-photo alignment, day 1 manual roi, metal surface anomaly, glass defect, anomalygen finetune, setup_pcb, setup_metal, setup_glass, setup_pretrained, dig setup, dig datasets, dig pretrained checkpoint, dig image-edit endpoint, cosmos defect generation, cosmos-predict2 defect, cosmos-anomalygen, cosmos predict2 finetune.\nversion: \"1.0.1\"\nlicense: CC-BY-4.0 AND Apache-2.0\ntools:\n  - Read\n  - Shell\nmetadata:\n  owner: NVIDIA\n  service: physical-ai-data-factory\n  version: 1.0.1\n  reviewed: 2026-06-23\n  author: NVIDIA\n  tags:\n    - physical-ai\n    - defect-image-generation\n    - aoi\n    - anomalygen\n    - usd2roi\n    - cosmos\n    - cosmos-predict2\n    - cosmos-anomalygen\n---\n\n# Physical AI Defect Image Generation\n\n\n## Table of Contents\n\n- [Supported Flows](#supported-flows)\n- [Disambiguation](#disambiguation-handle-vague-requests-before-committing) (full table in `references\u002Fdisambiguation.md`)\n- [Step 0: Select Flow, Cookbook, and Gather Inputs](#step-0-select-flow-cookbook-and-gather-inputs)\n- [Common Preconditions](#common-preconditions-all-flows) (long-form in `references\u002Fpreconditions.md`)\n- [Flow walkthroughs](#flow-walkthroughs) (one entry per flow; details in `references\u002Fflows\u002F`)\n- [OSMO Monitoring](#osmo-monitoring)\n- [Supporting files](#supporting-files)\n\nEnd-to-end orchestration of defect image generation, augmentation, and labeling pipelines for AOI (Automated Optical Inspection) datasets. **AnomalyGen = Cosmos-Predict2-2B finetuned per use case** (Cosmos-AnomalyGen-PCB-2B, -Metal-2B, -Glass-2B). Every flow has a canonical OSMO workflow YAML in `assets\u002Fconfigs\u002F` that chains all steps non-interactively. Use-case cookbooks in `assets\u002Fcookbooks\u002F` provide PCBA usd2roi\u002Fimage-edit configs and AnomalyGen training configs for PCBA, metal surface, and glass inspection. This skill governs flow selection, data handoffs, and submit commands; component internals live in each component's `SKILL.md`.\n\n## Supported Flows\n\n| Flow | Entry point | OSMO YAML | Steps | Use cases |\n|------|-------------|-----------|-------|-----------|\n| **Day 0 — Texture Defects** | CAD scene USD (`pcba_target.yaml` ships in the cookbook) | `texture_defect_generation_day0.yaml` | usd2roi (scan_grid + per-cell ROI crops) → image-edit augmentation (`nvidia\u002FQwen-Image-Edit-NVPCB-OVSL2SL`) → finetune-or-passthrough → infer (anomalygen labels inline, **including missing-component**) | PCBA |\n| **Day 0 — Good Image** *(usd2roi + Image-Edit)* | CAD scene USD + per-board `pcba_target.yaml` \u002F `day0_image.yaml` \u002F `day0_crop.yaml` | `good_image_generation.yaml` | usd2roi-render (scan_grid + per-cell ROI crop) → Qwen Image-Edit (OVSL2SL appearance transfer) | PCBA clean-image set (ChangeNet golden halves, finetune positives, real-photo pairing) |\n| **Day 0 — Structural Defects** | CAD scene USD + per-board `pcba_target.yaml` | `structural_defect_generation.yaml` | isaac-render (pose defects: shift \u002F tombstone \u002F sideflip) + per-component crop (single pod) → Qwen Image-Edit (OVSL2SL lighting transfer; pose geometry preserved) | PCBA pose-defect set; ChangeNet defect halves |\n| **Day 1 — Infer + Label (real-photo alignment, DEFAULT)** | CAD-derived USD + real PCBA photo (both ship in `datasets\u002Fpcb\u002Fassets`) | `texture_defect_generation_day1_real_alignment.yaml` | usd2roi day-1 render → MI register → per-ROI crop → yq-render config → finetune-or-passthrough → infer (anomalygen labels inline) | **Default PCBA Day 1.** Raw AOI screenshot of any usd2roi-supported board |\n| **Day 1 — Infer + Label (manual ROI)** | Pre-captured clean images + ROI masks (NGC artifact or user upload) | `texture_defect_generation_day1_manual_roi.yaml` | yq-render config → finetune-or-passthrough → infer (anomalygen labels inline) | Metal surface, glass (no USD\u002Freal-photo flow); PCBA **only when user explicitly asks** for pre-captured ROI experimentation |\n| **Finetune Only** | Labeled anomaly URL artifact | `finetune.yaml` | yq-render config → finetune (validate_dataset → prep_testcase → torchrun) | Any use case; produces checkpoint for Day 0 or Day 1. Requires raw training data under `\u003Cdig_url_root>\u002Fdatasets\u002F\u003Cusecase>\u002Fraw` (see `assets\u002Fconfigs\u002Fsetup\u002Fsetup_\u003Cusecase>.yaml`). |\n\nAll flows run on OSMO. Day 0 flows require `image_edit_endpoint` (Qwen Image-Edit OVSL2SL — existing URL or local deploy from `references\u002Fnim\u002F`); Finetune Only has no external endpoints.\n\n### Pick the right workflow for the user's defect class\n\n| Defect class | Workflow | Mechanism |\n|---|---|---|\n| Clean \u002F good \u002F scan-grid \u002F `normal_img + cad_mask` pairs | `good_image_generation.yaml` | usd2roi-render + Qwen Image-Edit |\n| Texture defects (solder bridge, scratch, discoloration) **AND missing-component** (handled natively by AnomalyGen, NOT structural) | `texture_defect_generation_day0.yaml` | Qwen Image-Edit + AnomalyGen AMP\u002FSDG |\n| Structural \u002F pose defects (tombstone, shift, sideflip) | `structural_defect_generation.yaml` | IsaacSim pose perturbation |\n| Day 1 inference + labeling on a real image | `texture_defect_generation_day1_real_alignment.yaml` (PCBA default) or `texture_defect_generation_day1_manual_roi.yaml` (metal\u002Fglass; PCBA only when user explicitly asks for pre-captured ROI \u002F skip-alignment) | usd2roi day-1 registration (real-alignment) or direct inference (manual-ROI) |\n\nChangeNet golden\u002Fdefect pairs: submit `good_image_generation.yaml` + `structural_defect_generation.yaml` with the same `--set name=` (two-submission pairing convention).\n\n> **Day 0 and Day 1 share the same downstream shape**: a Jinja-gated `finetune-job` (omitted when `use_pretrained_checkpoint=true`) feeding `anomaly-infer`. Day 0 prepends `usd2roi-render` + `augment-image-edit`; Day 1 starts from `\u003Cdig_url_root>\u002Fdatasets\u002F\u003Cusecase>\u002Fraw`. Per-stage detail: each flow's walkthrough.\n\n### User intent → knob mapping\n\n**Every OV flow is two-stage**: `crop_max_emit=N` caps the *final* per-cell crops (stage 2); `render_patches=N` caps *raw* scan-grid patches (stage 1, each yielding multiple crops). **DO NOT auto-map \"generate N images\" → `render_patches=N`** (wrong stage). `crop_max_emit` does not exist on `structural_defect_generation.yaml` (one crop per component — use `render_patches`) or `texture_defect_generation_day1_real_alignment.yaml` (narrow via the cookbook's `crop.classes` whitelist). Full knob table, smoke-test recipes, defaults, caveats: `references\u002Fknob_mapping.md`.\n\n### Structural-defect sizing (no `crop_max_emit` knob exists)\n\nStructural output is **non-linear in `render_patches`** — doubling frames adds ~1.6–1.7× crops, not 2×. Don't use `crop_max_emit` (no effect) or `render_patches=0` (fails). Validated yield table + target-size formula: `references\u002Fflows\u002Fstructural_defect_generation.md` §\"Sizing the output\". For ambiguous \"generate N images\", surface the calibration table via `AskUserQuestion`.\n\n---\n\n## Disambiguation: handle vague requests before committing\n\nUnderspecified prompts (\"generate me some images\", \"run the PCBA flow\", \"give me defects\") **must not** be resolved by silently assuming a flow \u002F usecase \u002F knob mapping. When intent is ambiguous, pause and present candidate interpretations via `AskUserQuestion` (2–4 mutually exclusive options) before submitting. Disambiguate the load-bearing choices: **which flow, which use case, what stage a count refers to, finetune vs. passthrough**.\n\nSettled defaults you should NOT disambiguate: PCBA Day 1 → real-alignment; board → `0603_H100`; image-edit endpoint → local cluster service (`references\u002Fnim\u002F`); `use_pretrained_checkpoint=true`; Day 1 real-alignment `default_spatial_dependency=cad` (fall back to `free` only when CAD masks are unavailable, see `references\u002Fflows\u002Ftexture_defect_generation_day1_real_alignment.md`).\n\n**`dig_url_root` is the one exception — NO silent default.** First-time (no memory entry), MUST elicit via `AskUserQuestion` before any submit \u002F `osmo data upload` \u002F `preflight_urls.sh`. `s3:\u002F\u002Fosmo-workflows\u002Fdig` is a *suggestion to confirm*, never auto-picked (~80 GB+ lands there). Later runs may reuse the remembered value silently. See Step 0 + memory rules (§4).\n\n**Full trigger table, prompt construction, and when-NOT-to-ask exceptions: `references\u002Fdisambiguation.md`** — load before assembling `AskUserQuestion` options for any vague request.\n\n---\n\n## Step 0: Select Flow, Cookbook, and Gather Inputs\n\n**Before this step**, if the request is vague (e.g. \"generate me images\", \"run the PCBA flow\", \"give me defects\"), pause and run the disambiguation cheat sheet above — present candidate interpretations via `AskUserQuestion` and let the user pick. Don't auto-pick a load-bearing default the user didn't actually choose.\n\n### First-time gate\n\nIf memory has no entries for this user, ASK the up-front preference questions in ONE `AskUserQuestion` call BEFORE any preflight \u002F `osmo` \u002F `kubectl` \u002F `osmo data upload`, save to memory (§4), then proceed. Bundle:\n\n- **`dig_url_root`** — MUST be elicited, not auto-picked. Offer `s3:\u002F\u002Fosmo-workflows\u002Fdig` as a confirmable suggestion; else user provides their own OSMO-supported storage prefix. ~80 GB+ lands here. No escape hatch other than memory-recall of a previously confirmed value.\n- **Default OSMO `--pool`** — candidates from `osmo profile list` → `pool.accessible`.\n- **Pod-template confirmation** — only when `osmo config show POD_TEMPLATE` returns 403 (§2 has the exact question).\n- **Image-edit endpoint** — Day 0 only: Option A (existing URL) vs Option B (deploy local NIM).\n\nSubsequent conversations read these silently from memory. Per-flow choices (use case, checkpoint vs finetune, board, knobs) are asked each time — see below.\n\n### Preflight ordering (after the first-time gate)\n\nRun §1 `preflight_credentials.sh` → §2 `preflight_pod_template.sh` → §3 `preflight_urls.sh \u003Cflow> \u003Cusecase>` → §4 generate the run stamp. **Cadence**: §1 and §2 are once-per-conversation gates with cross-conversation memory caching (see §4a in `references\u002Fpreconditions.md`) — skip when memory records them as already verified \u002F user-confirmed. §3 runs before every submit (varies by flow). §4 is the agent's job — fresh `$STAMP` per submit.\n\nPod-template enforcement is two layers: the pre-submit `preflight_pod_template.sh` gate (§2) plus an in-pod runtime preflight on every OV + training task (fails fast on missing `\u002Fusr\u002Fshare\u002Fnvidia\u002Fnvoptix.bin` or `\u002Fdev\u002Fshm` \u003C 16 GiB). Runtime failure despite §2 passing → template was patched out → route to `physical-ai-infrastructure-setup-and-resilient-scaling`. Missing creds \u002F URL artifacts → offer to submit `setup\u002Fsetup_\u003Ccase>.yaml` + `setup\u002Fsetup_pretrained.yaml` first.\n\nThen ask the user in one message — per-flow choices only (the first-time gate above already covered `dig_url_root`, pool, pod-template, and endpoint preferences; pull those from memory):\n\n1. **Use case** — PCBA (use Day 0 + pcb cookbook), metal surface (Day 1 + metal_surface cookbook), glass (Day 1 + glass cookbook), or custom?\n2. **Checkpoint available?** — If yes (`use_pretrained_checkpoint=true`), use `\u003Cdig_url_root>\u002Fmodels\u002F\u003Cusecase>` and provide `checkpoint_step`. If no, finetune from `\u003Cdig_url_root>\u002Fdatasets\u002F\u003Cusecase>\u002Fraw`.\n3. **Local-NIM pool capacity check** (Day 0 Option B only) — before `kubectl apply`, check `Total Capacity` via `physical-ai-infrastructure-setup-and-resilient-scaling`. `Total Capacity \u003C 2` cannot host NIM + DIG concurrently → ask user to add GPUs or switch to Option A. `image_edit_model` is always `nvidia\u002FQwen-Image-Edit-NVPCB-OVSL2SL`, never generic `qwen-image-edit`.\n4. **Save user preferences to memory** — after the first-time gate (and after any submit diverging from a documented default), persist load-bearing choices (`dig_url_root`, OSMO pool, default board, image-edit endpoint, pod-template state, osmo-admin role). **Never save** `image_edit_model` (constant — saving invites drift) or ephemeral state (STAMP, one-off `anomaly_types_json`). Full table: **`references\u002Fpreconditions.md` §4a \"Memory rules\"**. Read relevant memories at the start of every new conversation and apply silently.\n\nReview the relevant flow reference before asking — most values have sensible defaults. Day 1 routing: PCBA defaults to `real_alignment`; metal\u002Fglass have no USD flow so always `manual_roi`; don't ask the user \"manual or real-alignment?\" for PCBA unless they explicitly ask to skip alignment.\n\n---\n\n## Common Preconditions (all flows)\n\nQuick reference. Long-form: `references\u002Fpreconditions.md`.\n\n1. **OSMO credentials + tokens** — once per conversation. **If a `.env` exists in the workspace, source it first** (`set -a; . .\u002F.env; set +a`) so `HF_TOKEN` is exported. Run `scripts\u002Fpreflight_credentials.sh`; authoritative check is the OSMO cred `hf-token` is provisioned (images are public on `nvcr.io\u002Fnvidia\u002F` — no registry cred needed). Pass `--no-probe` in restricted-egress shells. See `references\u002Fpreconditions.md` §1.\n2. **Pod template** — once per conversation, with cross-conversation memory caching (see Step 0 §6). Skip when memory records the cluster verified \u002F user-confirmed \u002F 409-skipped. Otherwise run `scripts\u002Fpreflight_pod_template.sh` and branch on exit code (0=verified \u002F 1=patch via infra skill \u002F 2=ask-user (HTTP 403) \u002F 3=skip (HTTP 409) \u002F 4=env-fix). Full branching prose and prompts in `references\u002Fpreconditions.md` §2.\n3. **Required URL artifacts** — before every submit. Run `DIG_URL_ROOT=\u003Cdig_url_root> scripts\u002Fpreflight_urls.sh \u003Cflow> \u003Cusecase> [variant]`. If anything is missing, **stop and submit the relevant `setup\u002Fsetup_\u003Ccase>.yaml` + `setup\u002Fsetup_pretrained.yaml` first** (the OSMO setup workflows) — see `references\u002Fsetup.md`. **Never download assets locally to work around a problem; if setup fails on credentials, ask the user to rectify them and re-submit on OSMO.** Per-flow checklist:\n\n   | Flow | Use case | Required URL artifacts under `\u003Cdig_url_root>` |\n   |---|---|---|\n   | Day 0 — Texture Defects | PCBA | `models\u002Fpretrained`, `models\u002Fpcb`, `datasets\u002Fpcb\u002Fraw`, `datasets\u002Fpcb\u002Fassets` |\n   | Day 0 — Good Image | PCBA | `datasets\u002Fpcb\u002Fassets` only |\n   | Day 0 — Structural Defects | PCBA | `datasets\u002Fpcb\u002Fassets` only |\n   | Day 1 | Metal surface | `models\u002Fpretrained`, `models\u002Fmetal_surface`, `datasets\u002Fmetal_surface\u002Fraw` |\n   | Day 1 | Glass | `models\u002Fpretrained`, `models\u002Fglass`, `datasets\u002Fglass\u002Fraw` |\n   | Day 1 real-photo alignment | PCBA | Day 1 PCBA plus `datasets\u002Fpcb\u002Fassets` |\n   | Finetune Only | Any | `models\u002Fpretrained`, `datasets\u002F\u003Cusecase>\u002Fraw` |\n\n   Built-in `usecase` values are `pcb`, `metal_surface`, `glass`. See `references\u002Fpreconditions.md` §3.\n\n4. **Name stamping** — regenerate `$STAMP=$(cat \u002Fproc\u002Fsys\u002Fkernel\u002Frandom\u002Fuuid | cut -c1-8)` before every submit and pass `--set name=\u003Cflow>-$STAMP`. Production YAMLs ship no `name` default. See `references\u002Fpreconditions.md` §4.\n5. **Glass case (UC3) — Roboflow zip** — only for `setup_glass.yaml`. Upload `mobile_screen.zip` to an OSMO URL prefix first; pass `--set uc3_zip_url_root=\u003Cprefix>`. Full procedure: `references\u002Fsetup.md` §\"Glass case (UC3)\".\n\n---\n\n## Flow walkthroughs\n\nEach flow's full walkthrough — group diagrams, prerequisites, submit-command variants, data handoffs, per-stage troubleshooting — lives under `references\u002Fflows\u002F`. The agent should read the matching file before submitting any flow it hasn't run in the current conversation.\n\n| Flow | Workflow YAML | Walkthrough |\n|---|---|---|\n| **Day 0 — Texture Defects (PCBA)** | `assets\u002Fconfigs\u002Ftexture_defect_generation_day0.yaml` | `references\u002Fflows\u002Ftexture_defect_generation_day0.md` |\n| **Day 0 — Good Image (PCBA)** | `assets\u002Fconfigs\u002Fgood_image_generation.yaml` | `references\u002Fflows\u002Fgood_image_generation.md` |\n| **Day 0 — Structural Defects (PCBA)** | `assets\u002Fconfigs\u002Fstructural_defect_generation.yaml` | `references\u002Fflows\u002Fstructural_defect_generation.md` |\n| **Day 1 — Infer + Label (real-photo alignment, default PCBA)** | `assets\u002Fconfigs\u002Ftexture_defect_generation_day1_real_alignment.yaml` | `references\u002Fflows\u002Ftexture_defect_generation_day1_real_alignment.md` |\n| **Day 1 — Infer + Label (manual ROI, metal\u002Fglass + PCBA experimentation)** | `assets\u002Fconfigs\u002Ftexture_defect_generation_day1_manual_roi.yaml` | `references\u002Fflows\u002Ftexture_defect_generation_day1_manual_roi.md` |\n| **Finetune Only** | `assets\u002Fconfigs\u002Ffinetune.yaml` | `references\u002Fflows\u002Ffinetune.md` |\n\n### Cross-flow invariants\n\n- `use_pretrained_checkpoint=true` (default) → passthrough against `models\u002F\u003Cusecase>`. Set to `false` to insert an in-pod `finetune-job` group (cookbook yq-patched in-pod, no pre-submit render step).\n- Day 0 emits per-cell `crop\u002F\u003CMATERIAL>\u002F\u003Ccell>\u002F...` trees; Day 1 emits per-ROI crops registered against the USD; structural emits flat per-component crops.\n- Shipped per-usecase `checkpoint_step` + `anomaly_types_json` defaults: see `references\u002Fpreconditions.md` §\"Shipped checkpoint and `anomaly_types_json` defaults\".\n\n---\n\n## OSMO Monitoring\n\n**Load `references\u002Fmonitoring.md` before any `osmo workflow submit`, `osmo workflow query`, or `osmo workflow logs` action in this skill.** It defines the polling cadence, task-status interpretation, log-pull escalation thresholds, failure-classification routing, and what to surface to the user vs. silently retry. Do not assemble a post-submit watch loop or status summary from memory — re-read it on the first such action of every conversation.\n\n```bash\nosmo workflow query \u003Cworkflow_id> --format-type json | jq '{status, tasks: [.groups[].tasks[] | {name, status, exit_code}]}'\nosmo workflow logs \u003Cworkflow_id> -t \u003Ctask_name> -n 200\nosmo data download \u003Cdig_url_root>\u002Fruns\u002F\u003Cname>\u002Fanomaly .\u002Foutput\u002Fanomaly-\u003Cname>\u002F\n```\n\nMonitoring discipline: `references\u002Fmonitoring.md`. Retrieval: `references\u002Foutput_retrieval.md`. Presentation: `references\u002Foutput_rendering.md`. Gotchas: `references\u002Ftroubleshooting.md`.\n\n---\n\n## Response Template\n\nFor \"show me the plan \u002F recipe\" requests, emit your final response with these labeled sections (so nothing truncates mid-recipe):\n\n**Workflow:** `\u003Cflow name>` → `assets\u002Fconfigs\u002F\u003Cyaml>`\n\n**Preflights:** `scripts\u002Fpreflight_credentials.sh`; `scripts\u002Fpreflight_urls.sh \u003C0|1|finetune> \u003Cusecase> [variant]`\n\n**Required URL Artifacts under `\u003Cdig_url_root>`:** enumerate per Common Preconditions §3 for the chosen flow.\n\n**Submit Command:**\n\n```bash\nSTAMP=$(cat \u002Fproc\u002Fsys\u002Fkernel\u002Frandom\u002Fuuid | cut -c1-8)\nosmo workflow submit assets\u002Fconfigs\u002F\u003Cyaml> --pool \u003Cpool> \\\n  --set name=\u003Cflow>-$STAMP dig_url_root=\u003Croot> usecase=\u003Cusecase> \\\n        image_edit_endpoint=\u003Cendpoint> image_edit_model=nvidia\u002FQwen-Image-Edit-NVPCB-OVSL2SL \\\n        checkpoint_step=\u003Cstep> 'anomaly_types_json=\u003Ctypes>'\n```\n\n**Monitoring:** load `references\u002Fmonitoring.md` before running the submit; apply its polling cadence + log-pull thresholds after `osmo workflow submit` returns a workflow id.\n\n**Output Location:** `\u003Cdig_url_root>\u002Fruns\u002F\u003Cflow>-$STAMP\u002Fanomaly\u002F` (per-flow override: see flow walkthrough).\n\n---\n\n## Supporting files\n\nFull inventory — workflow YAMLs, cookbooks, scripts table, references, evals, component skills — in **`references\u002Fcontents.md`**. Top-level dirs: `assets\u002Fconfigs\u002F`, `assets\u002Fcookbooks\u002F`, `scripts\u002F`, `references\u002F`, `evals\u002F`.\n",{"data":35,"body":52},{"name":4,"description":6,"version":36,"license":26,"tools":37,"metadata":40},"1.0.1",[38,39],"Read","Shell",{"owner":9,"service":41,"version":36,"reviewed":42,"author":9,"tags":43},"physical-ai-data-factory","2026-06-23",[44,45,46,47,48,49,50,51],"physical-ai","defect-image-generation","aoi","anomalygen","usd2roi","cosmos","cosmos-predict2","cosmos-anomalygen",{"type":53,"children":54},"root",[55,63,70,168,206,211,553,574,581,716,743,803,809,906,919,966,970,976,1001,1051,1104,1126,1129,1134,1151,1157,1190,1272,1277,1283,1334,1385,1397,1562,1583,1586,1592,1603,2119,2122,2127,2139,2336,2342,2423,2426,2431,2472,2719,2754,2757,2763,2768,2790,2812,2829,2837,3115,3139,3156,3159,3164,3213],{"type":56,"tag":57,"props":58,"children":59},"element","h1",{"id":4},[60],{"type":61,"value":62},"text","Physical AI Defect Image Generation",{"type":56,"tag":64,"props":65,"children":67},"h2",{"id":66},"table-of-contents",[68],{"type":61,"value":69},"Table of Contents",{"type":56,"tag":71,"props":72,"children":73},"ul",{},[74,85,105,114,132,150,159],{"type":56,"tag":75,"props":76,"children":77},"li",{},[78],{"type":56,"tag":79,"props":80,"children":82},"a",{"href":81},"#supported-flows",[83],{"type":61,"value":84},"Supported Flows",{"type":56,"tag":75,"props":86,"children":87},{},[88,94,96,103],{"type":56,"tag":79,"props":89,"children":91},{"href":90},"#disambiguation-handle-vague-requests-before-committing",[92],{"type":61,"value":93},"Disambiguation",{"type":61,"value":95}," (full table in ",{"type":56,"tag":97,"props":98,"children":100},"code",{"className":99},[],[101],{"type":61,"value":102},"references\u002Fdisambiguation.md",{"type":61,"value":104},")",{"type":56,"tag":75,"props":106,"children":107},{},[108],{"type":56,"tag":79,"props":109,"children":111},{"href":110},"#step-0-select-flow-cookbook-and-gather-inputs",[112],{"type":61,"value":113},"Step 0: Select Flow, Cookbook, and Gather Inputs",{"type":56,"tag":75,"props":115,"children":116},{},[117,123,125,131],{"type":56,"tag":79,"props":118,"children":120},{"href":119},"#common-preconditions-all-flows",[121],{"type":61,"value":122},"Common Preconditions",{"type":61,"value":124}," (long-form in ",{"type":56,"tag":97,"props":126,"children":128},{"className":127},[],[129],{"type":61,"value":130},"references\u002Fpreconditions.md",{"type":61,"value":104},{"type":56,"tag":75,"props":133,"children":134},{},[135,141,143,149],{"type":56,"tag":79,"props":136,"children":138},{"href":137},"#flow-walkthroughs",[139],{"type":61,"value":140},"Flow walkthroughs",{"type":61,"value":142}," (one entry per flow; details in ",{"type":56,"tag":97,"props":144,"children":146},{"className":145},[],[147],{"type":61,"value":148},"references\u002Fflows\u002F",{"type":61,"value":104},{"type":56,"tag":75,"props":151,"children":152},{},[153],{"type":56,"tag":79,"props":154,"children":156},{"href":155},"#osmo-monitoring",[157],{"type":61,"value":158},"OSMO Monitoring",{"type":56,"tag":75,"props":160,"children":161},{},[162],{"type":56,"tag":79,"props":163,"children":165},{"href":164},"#supporting-files",[166],{"type":61,"value":167},"Supporting files",{"type":56,"tag":169,"props":170,"children":171},"p",{},[172,174,180,182,188,190,196,198,204],{"type":61,"value":173},"End-to-end orchestration of defect image generation, augmentation, and labeling pipelines for AOI (Automated Optical Inspection) datasets. ",{"type":56,"tag":175,"props":176,"children":177},"strong",{},[178],{"type":61,"value":179},"AnomalyGen = Cosmos-Predict2-2B finetuned per use case",{"type":61,"value":181}," (Cosmos-AnomalyGen-PCB-2B, -Metal-2B, -Glass-2B). Every flow has a canonical OSMO workflow YAML in ",{"type":56,"tag":97,"props":183,"children":185},{"className":184},[],[186],{"type":61,"value":187},"assets\u002Fconfigs\u002F",{"type":61,"value":189}," that chains all steps non-interactively. Use-case cookbooks in ",{"type":56,"tag":97,"props":191,"children":193},{"className":192},[],[194],{"type":61,"value":195},"assets\u002Fcookbooks\u002F",{"type":61,"value":197}," provide PCBA usd2roi\u002Fimage-edit configs and AnomalyGen training configs for PCBA, metal surface, and glass inspection. This skill governs flow selection, data handoffs, and submit commands; component internals live in each component's ",{"type":56,"tag":97,"props":199,"children":201},{"className":200},[],[202],{"type":61,"value":203},"SKILL.md",{"type":61,"value":205},".",{"type":56,"tag":64,"props":207,"children":209},{"id":208},"supported-flows",[210],{"type":61,"value":84},{"type":56,"tag":212,"props":213,"children":214},"table",{},[215,249],{"type":56,"tag":216,"props":217,"children":218},"thead",{},[219],{"type":56,"tag":220,"props":221,"children":222},"tr",{},[223,229,234,239,244],{"type":56,"tag":224,"props":225,"children":226},"th",{},[227],{"type":61,"value":228},"Flow",{"type":56,"tag":224,"props":230,"children":231},{},[232],{"type":61,"value":233},"Entry point",{"type":56,"tag":224,"props":235,"children":236},{},[237],{"type":61,"value":238},"OSMO YAML",{"type":56,"tag":224,"props":240,"children":241},{},[242],{"type":61,"value":243},"Steps",{"type":56,"tag":224,"props":245,"children":246},{},[247],{"type":61,"value":248},"Use cases",{"type":56,"tag":250,"props":251,"children":252},"tbody",{},[253,311,374,413,460,502],{"type":56,"tag":220,"props":254,"children":255},{},[256,265,278,287,306],{"type":56,"tag":257,"props":258,"children":259},"td",{},[260],{"type":56,"tag":175,"props":261,"children":262},{},[263],{"type":61,"value":264},"Day 0 — Texture Defects",{"type":56,"tag":257,"props":266,"children":267},{},[268,270,276],{"type":61,"value":269},"CAD scene USD (",{"type":56,"tag":97,"props":271,"children":273},{"className":272},[],[274],{"type":61,"value":275},"pcba_target.yaml",{"type":61,"value":277}," ships in the cookbook)",{"type":56,"tag":257,"props":279,"children":280},{},[281],{"type":56,"tag":97,"props":282,"children":284},{"className":283},[],[285],{"type":61,"value":286},"texture_defect_generation_day0.yaml",{"type":56,"tag":257,"props":288,"children":289},{},[290,292,298,300,305],{"type":61,"value":291},"usd2roi (scan_grid + per-cell ROI crops) → image-edit augmentation (",{"type":56,"tag":97,"props":293,"children":295},{"className":294},[],[296],{"type":61,"value":297},"nvidia\u002FQwen-Image-Edit-NVPCB-OVSL2SL",{"type":61,"value":299},") → finetune-or-passthrough → infer (anomalygen labels inline, ",{"type":56,"tag":175,"props":301,"children":302},{},[303],{"type":61,"value":304},"including missing-component",{"type":61,"value":104},{"type":56,"tag":257,"props":307,"children":308},{},[309],{"type":61,"value":310},"PCBA",{"type":56,"tag":220,"props":312,"children":313},{},[314,330,355,364,369],{"type":56,"tag":257,"props":315,"children":316},{},[317,322,324],{"type":56,"tag":175,"props":318,"children":319},{},[320],{"type":61,"value":321},"Day 0 — Good Image",{"type":61,"value":323}," ",{"type":56,"tag":325,"props":326,"children":327},"em",{},[328],{"type":61,"value":329},"(usd2roi + Image-Edit)",{"type":56,"tag":257,"props":331,"children":332},{},[333,335,340,342,348,349],{"type":61,"value":334},"CAD scene USD + per-board ",{"type":56,"tag":97,"props":336,"children":338},{"className":337},[],[339],{"type":61,"value":275},{"type":61,"value":341}," \u002F ",{"type":56,"tag":97,"props":343,"children":345},{"className":344},[],[346],{"type":61,"value":347},"day0_image.yaml",{"type":61,"value":341},{"type":56,"tag":97,"props":350,"children":352},{"className":351},[],[353],{"type":61,"value":354},"day0_crop.yaml",{"type":56,"tag":257,"props":356,"children":357},{},[358],{"type":56,"tag":97,"props":359,"children":361},{"className":360},[],[362],{"type":61,"value":363},"good_image_generation.yaml",{"type":56,"tag":257,"props":365,"children":366},{},[367],{"type":61,"value":368},"usd2roi-render (scan_grid + per-cell ROI crop) → Qwen Image-Edit (OVSL2SL appearance transfer)",{"type":56,"tag":257,"props":370,"children":371},{},[372],{"type":61,"value":373},"PCBA clean-image set (ChangeNet golden halves, finetune positives, real-photo pairing)",{"type":56,"tag":220,"props":375,"children":376},{},[377,385,394,403,408],{"type":56,"tag":257,"props":378,"children":379},{},[380],{"type":56,"tag":175,"props":381,"children":382},{},[383],{"type":61,"value":384},"Day 0 — Structural Defects",{"type":56,"tag":257,"props":386,"children":387},{},[388,389],{"type":61,"value":334},{"type":56,"tag":97,"props":390,"children":392},{"className":391},[],[393],{"type":61,"value":275},{"type":56,"tag":257,"props":395,"children":396},{},[397],{"type":56,"tag":97,"props":398,"children":400},{"className":399},[],[401],{"type":61,"value":402},"structural_defect_generation.yaml",{"type":56,"tag":257,"props":404,"children":405},{},[406],{"type":61,"value":407},"isaac-render (pose defects: shift \u002F tombstone \u002F sideflip) + per-component crop (single pod) → Qwen Image-Edit (OVSL2SL lighting transfer; pose geometry preserved)",{"type":56,"tag":257,"props":409,"children":410},{},[411],{"type":61,"value":412},"PCBA pose-defect set; ChangeNet defect halves",{"type":56,"tag":220,"props":414,"children":415},{},[416,424,436,445,450],{"type":56,"tag":257,"props":417,"children":418},{},[419],{"type":56,"tag":175,"props":420,"children":421},{},[422],{"type":61,"value":423},"Day 1 — Infer + Label (real-photo alignment, DEFAULT)",{"type":56,"tag":257,"props":425,"children":426},{},[427,429,435],{"type":61,"value":428},"CAD-derived USD + real PCBA photo (both ship in ",{"type":56,"tag":97,"props":430,"children":432},{"className":431},[],[433],{"type":61,"value":434},"datasets\u002Fpcb\u002Fassets",{"type":61,"value":104},{"type":56,"tag":257,"props":437,"children":438},{},[439],{"type":56,"tag":97,"props":440,"children":442},{"className":441},[],[443],{"type":61,"value":444},"texture_defect_generation_day1_real_alignment.yaml",{"type":56,"tag":257,"props":446,"children":447},{},[448],{"type":61,"value":449},"usd2roi day-1 render → MI register → per-ROI crop → yq-render config → finetune-or-passthrough → infer (anomalygen labels inline)",{"type":56,"tag":257,"props":451,"children":452},{},[453,458],{"type":56,"tag":175,"props":454,"children":455},{},[456],{"type":61,"value":457},"Default PCBA Day 1.",{"type":61,"value":459}," Raw AOI screenshot of any usd2roi-supported board",{"type":56,"tag":220,"props":461,"children":462},{},[463,471,476,485,490],{"type":56,"tag":257,"props":464,"children":465},{},[466],{"type":56,"tag":175,"props":467,"children":468},{},[469],{"type":61,"value":470},"Day 1 — Infer + Label (manual ROI)",{"type":56,"tag":257,"props":472,"children":473},{},[474],{"type":61,"value":475},"Pre-captured clean images + ROI masks (NGC artifact or user upload)",{"type":56,"tag":257,"props":477,"children":478},{},[479],{"type":56,"tag":97,"props":480,"children":482},{"className":481},[],[483],{"type":61,"value":484},"texture_defect_generation_day1_manual_roi.yaml",{"type":56,"tag":257,"props":486,"children":487},{},[488],{"type":61,"value":489},"yq-render config → finetune-or-passthrough → infer (anomalygen labels inline)",{"type":56,"tag":257,"props":491,"children":492},{},[493,495,500],{"type":61,"value":494},"Metal surface, glass (no USD\u002Freal-photo flow); PCBA ",{"type":56,"tag":175,"props":496,"children":497},{},[498],{"type":61,"value":499},"only when user explicitly asks",{"type":61,"value":501}," for pre-captured ROI experimentation",{"type":56,"tag":220,"props":503,"children":504},{},[505,513,518,527,532],{"type":56,"tag":257,"props":506,"children":507},{},[508],{"type":56,"tag":175,"props":509,"children":510},{},[511],{"type":61,"value":512},"Finetune Only",{"type":56,"tag":257,"props":514,"children":515},{},[516],{"type":61,"value":517},"Labeled anomaly URL artifact",{"type":56,"tag":257,"props":519,"children":520},{},[521],{"type":56,"tag":97,"props":522,"children":524},{"className":523},[],[525],{"type":61,"value":526},"finetune.yaml",{"type":56,"tag":257,"props":528,"children":529},{},[530],{"type":61,"value":531},"yq-render config → finetune (validate_dataset → prep_testcase → torchrun)",{"type":56,"tag":257,"props":533,"children":534},{},[535,537,543,545,551],{"type":61,"value":536},"Any use case; produces checkpoint for Day 0 or Day 1. Requires raw training data under ",{"type":56,"tag":97,"props":538,"children":540},{"className":539},[],[541],{"type":61,"value":542},"\u003Cdig_url_root>\u002Fdatasets\u002F\u003Cusecase>\u002Fraw",{"type":61,"value":544}," (see ",{"type":56,"tag":97,"props":546,"children":548},{"className":547},[],[549],{"type":61,"value":550},"assets\u002Fconfigs\u002Fsetup\u002Fsetup_\u003Cusecase>.yaml",{"type":61,"value":552},").",{"type":56,"tag":169,"props":554,"children":555},{},[556,558,564,566,572],{"type":61,"value":557},"All flows run on OSMO. Day 0 flows require ",{"type":56,"tag":97,"props":559,"children":561},{"className":560},[],[562],{"type":61,"value":563},"image_edit_endpoint",{"type":61,"value":565}," (Qwen Image-Edit OVSL2SL — existing URL or local deploy from ",{"type":56,"tag":97,"props":567,"children":569},{"className":568},[],[570],{"type":61,"value":571},"references\u002Fnim\u002F",{"type":61,"value":573},"); Finetune Only has no external endpoints.",{"type":56,"tag":575,"props":576,"children":578},"h3",{"id":577},"pick-the-right-workflow-for-the-users-defect-class",[579],{"type":61,"value":580},"Pick the right workflow for the user's defect class",{"type":56,"tag":212,"props":582,"children":583},{},[584,605],{"type":56,"tag":216,"props":585,"children":586},{},[587],{"type":56,"tag":220,"props":588,"children":589},{},[590,595,600],{"type":56,"tag":224,"props":591,"children":592},{},[593],{"type":61,"value":594},"Defect class",{"type":56,"tag":224,"props":596,"children":597},{},[598],{"type":61,"value":599},"Workflow",{"type":56,"tag":224,"props":601,"children":602},{},[603],{"type":61,"value":604},"Mechanism",{"type":56,"tag":250,"props":606,"children":607},{},[608,637,665,686],{"type":56,"tag":220,"props":609,"children":610},{},[611,624,632],{"type":56,"tag":257,"props":612,"children":613},{},[614,616,622],{"type":61,"value":615},"Clean \u002F good \u002F scan-grid \u002F ",{"type":56,"tag":97,"props":617,"children":619},{"className":618},[],[620],{"type":61,"value":621},"normal_img + cad_mask",{"type":61,"value":623}," pairs",{"type":56,"tag":257,"props":625,"children":626},{},[627],{"type":56,"tag":97,"props":628,"children":630},{"className":629},[],[631],{"type":61,"value":363},{"type":56,"tag":257,"props":633,"children":634},{},[635],{"type":61,"value":636},"usd2roi-render + Qwen Image-Edit",{"type":56,"tag":220,"props":638,"children":639},{},[640,652,660],{"type":56,"tag":257,"props":641,"children":642},{},[643,645,650],{"type":61,"value":644},"Texture defects (solder bridge, scratch, discoloration) ",{"type":56,"tag":175,"props":646,"children":647},{},[648],{"type":61,"value":649},"AND missing-component",{"type":61,"value":651}," (handled natively by AnomalyGen, NOT structural)",{"type":56,"tag":257,"props":653,"children":654},{},[655],{"type":56,"tag":97,"props":656,"children":658},{"className":657},[],[659],{"type":61,"value":286},{"type":56,"tag":257,"props":661,"children":662},{},[663],{"type":61,"value":664},"Qwen Image-Edit + AnomalyGen AMP\u002FSDG",{"type":56,"tag":220,"props":666,"children":667},{},[668,673,681],{"type":56,"tag":257,"props":669,"children":670},{},[671],{"type":61,"value":672},"Structural \u002F pose defects (tombstone, shift, sideflip)",{"type":56,"tag":257,"props":674,"children":675},{},[676],{"type":56,"tag":97,"props":677,"children":679},{"className":678},[],[680],{"type":61,"value":402},{"type":56,"tag":257,"props":682,"children":683},{},[684],{"type":61,"value":685},"IsaacSim pose perturbation",{"type":56,"tag":220,"props":687,"children":688},{},[689,694,711],{"type":56,"tag":257,"props":690,"children":691},{},[692],{"type":61,"value":693},"Day 1 inference + labeling on a real image",{"type":56,"tag":257,"props":695,"children":696},{},[697,702,704,709],{"type":56,"tag":97,"props":698,"children":700},{"className":699},[],[701],{"type":61,"value":444},{"type":61,"value":703}," (PCBA default) or ",{"type":56,"tag":97,"props":705,"children":707},{"className":706},[],[708],{"type":61,"value":484},{"type":61,"value":710}," (metal\u002Fglass; PCBA only when user explicitly asks for pre-captured ROI \u002F skip-alignment)",{"type":56,"tag":257,"props":712,"children":713},{},[714],{"type":61,"value":715},"usd2roi day-1 registration (real-alignment) or direct inference (manual-ROI)",{"type":56,"tag":169,"props":717,"children":718},{},[719,721,726,728,733,735,741],{"type":61,"value":720},"ChangeNet golden\u002Fdefect pairs: submit ",{"type":56,"tag":97,"props":722,"children":724},{"className":723},[],[725],{"type":61,"value":363},{"type":61,"value":727}," + ",{"type":56,"tag":97,"props":729,"children":731},{"className":730},[],[732],{"type":61,"value":402},{"type":61,"value":734}," with the same ",{"type":56,"tag":97,"props":736,"children":738},{"className":737},[],[739],{"type":61,"value":740},"--set name=",{"type":61,"value":742}," (two-submission pairing convention).",{"type":56,"tag":744,"props":745,"children":746},"blockquote",{},[747],{"type":56,"tag":169,"props":748,"children":749},{},[750,755,757,763,765,771,773,779,781,787,788,794,796,801],{"type":56,"tag":175,"props":751,"children":752},{},[753],{"type":61,"value":754},"Day 0 and Day 1 share the same downstream shape",{"type":61,"value":756},": a Jinja-gated ",{"type":56,"tag":97,"props":758,"children":760},{"className":759},[],[761],{"type":61,"value":762},"finetune-job",{"type":61,"value":764}," (omitted when ",{"type":56,"tag":97,"props":766,"children":768},{"className":767},[],[769],{"type":61,"value":770},"use_pretrained_checkpoint=true",{"type":61,"value":772},") feeding ",{"type":56,"tag":97,"props":774,"children":776},{"className":775},[],[777],{"type":61,"value":778},"anomaly-infer",{"type":61,"value":780},". Day 0 prepends ",{"type":56,"tag":97,"props":782,"children":784},{"className":783},[],[785],{"type":61,"value":786},"usd2roi-render",{"type":61,"value":727},{"type":56,"tag":97,"props":789,"children":791},{"className":790},[],[792],{"type":61,"value":793},"augment-image-edit",{"type":61,"value":795},"; Day 1 starts from ",{"type":56,"tag":97,"props":797,"children":799},{"className":798},[],[800],{"type":61,"value":542},{"type":61,"value":802},". Per-stage detail: each flow's walkthrough.",{"type":56,"tag":575,"props":804,"children":806},{"id":805},"user-intent-knob-mapping",[807],{"type":61,"value":808},"User intent → knob mapping",{"type":56,"tag":169,"props":810,"children":811},{},[812,817,819,825,827,832,834,840,842,847,849,859,861,867,869,874,876,882,884,889,891,897,899,905],{"type":56,"tag":175,"props":813,"children":814},{},[815],{"type":61,"value":816},"Every OV flow is two-stage",{"type":61,"value":818},": ",{"type":56,"tag":97,"props":820,"children":822},{"className":821},[],[823],{"type":61,"value":824},"crop_max_emit=N",{"type":61,"value":826}," caps the ",{"type":56,"tag":325,"props":828,"children":829},{},[830],{"type":61,"value":831},"final",{"type":61,"value":833}," per-cell crops (stage 2); ",{"type":56,"tag":97,"props":835,"children":837},{"className":836},[],[838],{"type":61,"value":839},"render_patches=N",{"type":61,"value":841}," caps ",{"type":56,"tag":325,"props":843,"children":844},{},[845],{"type":61,"value":846},"raw",{"type":61,"value":848}," scan-grid patches (stage 1, each yielding multiple crops). ",{"type":56,"tag":175,"props":850,"children":851},{},[852,854],{"type":61,"value":853},"DO NOT auto-map \"generate N images\" → ",{"type":56,"tag":97,"props":855,"children":857},{"className":856},[],[858],{"type":61,"value":839},{"type":61,"value":860}," (wrong stage). ",{"type":56,"tag":97,"props":862,"children":864},{"className":863},[],[865],{"type":61,"value":866},"crop_max_emit",{"type":61,"value":868}," does not exist on ",{"type":56,"tag":97,"props":870,"children":872},{"className":871},[],[873],{"type":61,"value":402},{"type":61,"value":875}," (one crop per component — use ",{"type":56,"tag":97,"props":877,"children":879},{"className":878},[],[880],{"type":61,"value":881},"render_patches",{"type":61,"value":883},") or ",{"type":56,"tag":97,"props":885,"children":887},{"className":886},[],[888],{"type":61,"value":444},{"type":61,"value":890}," (narrow via the cookbook's ",{"type":56,"tag":97,"props":892,"children":894},{"className":893},[],[895],{"type":61,"value":896},"crop.classes",{"type":61,"value":898}," whitelist). Full knob table, smoke-test recipes, defaults, caveats: ",{"type":56,"tag":97,"props":900,"children":902},{"className":901},[],[903],{"type":61,"value":904},"references\u002Fknob_mapping.md",{"type":61,"value":205},{"type":56,"tag":575,"props":907,"children":909},{"id":908},"structural-defect-sizing-no-crop_max_emit-knob-exists",[910,912,917],{"type":61,"value":911},"Structural-defect sizing (no ",{"type":56,"tag":97,"props":913,"children":915},{"className":914},[],[916],{"type":61,"value":866},{"type":61,"value":918}," knob exists)",{"type":56,"tag":169,"props":920,"children":921},{},[922,924,934,936,941,943,949,951,957,959,965],{"type":61,"value":923},"Structural output is ",{"type":56,"tag":175,"props":925,"children":926},{},[927,929],{"type":61,"value":928},"non-linear in ",{"type":56,"tag":97,"props":930,"children":932},{"className":931},[],[933],{"type":61,"value":881},{"type":61,"value":935}," — doubling frames adds ~1.6–1.7× crops, not 2×. Don't use ",{"type":56,"tag":97,"props":937,"children":939},{"className":938},[],[940],{"type":61,"value":866},{"type":61,"value":942}," (no effect) or ",{"type":56,"tag":97,"props":944,"children":946},{"className":945},[],[947],{"type":61,"value":948},"render_patches=0",{"type":61,"value":950}," (fails). Validated yield table + target-size formula: ",{"type":56,"tag":97,"props":952,"children":954},{"className":953},[],[955],{"type":61,"value":956},"references\u002Fflows\u002Fstructural_defect_generation.md",{"type":61,"value":958}," §\"Sizing the output\". For ambiguous \"generate N images\", surface the calibration table via ",{"type":56,"tag":97,"props":960,"children":962},{"className":961},[],[963],{"type":61,"value":964},"AskUserQuestion",{"type":61,"value":205},{"type":56,"tag":967,"props":968,"children":969},"hr",{},[],{"type":56,"tag":64,"props":971,"children":973},{"id":972},"disambiguation-handle-vague-requests-before-committing",[974],{"type":61,"value":975},"Disambiguation: handle vague requests before committing",{"type":56,"tag":169,"props":977,"children":978},{},[979,981,986,988,993,995,1000],{"type":61,"value":980},"Underspecified prompts (\"generate me some images\", \"run the PCBA flow\", \"give me defects\") ",{"type":56,"tag":175,"props":982,"children":983},{},[984],{"type":61,"value":985},"must not",{"type":61,"value":987}," be resolved by silently assuming a flow \u002F usecase \u002F knob mapping. When intent is ambiguous, pause and present candidate interpretations via ",{"type":56,"tag":97,"props":989,"children":991},{"className":990},[],[992],{"type":61,"value":964},{"type":61,"value":994}," (2–4 mutually exclusive options) before submitting. Disambiguate the load-bearing choices: ",{"type":56,"tag":175,"props":996,"children":997},{},[998],{"type":61,"value":999},"which flow, which use case, what stage a count refers to, finetune vs. passthrough",{"type":61,"value":205},{"type":56,"tag":169,"props":1002,"children":1003},{},[1004,1006,1012,1014,1019,1021,1026,1028,1034,1036,1042,1044,1050],{"type":61,"value":1005},"Settled defaults you should NOT disambiguate: PCBA Day 1 → real-alignment; board → ",{"type":56,"tag":97,"props":1007,"children":1009},{"className":1008},[],[1010],{"type":61,"value":1011},"0603_H100",{"type":61,"value":1013},"; image-edit endpoint → local cluster service (",{"type":56,"tag":97,"props":1015,"children":1017},{"className":1016},[],[1018],{"type":61,"value":571},{"type":61,"value":1020},"); ",{"type":56,"tag":97,"props":1022,"children":1024},{"className":1023},[],[1025],{"type":61,"value":770},{"type":61,"value":1027},"; Day 1 real-alignment ",{"type":56,"tag":97,"props":1029,"children":1031},{"className":1030},[],[1032],{"type":61,"value":1033},"default_spatial_dependency=cad",{"type":61,"value":1035}," (fall back to ",{"type":56,"tag":97,"props":1037,"children":1039},{"className":1038},[],[1040],{"type":61,"value":1041},"free",{"type":61,"value":1043}," only when CAD masks are unavailable, see ",{"type":56,"tag":97,"props":1045,"children":1047},{"className":1046},[],[1048],{"type":61,"value":1049},"references\u002Fflows\u002Ftexture_defect_generation_day1_real_alignment.md",{"type":61,"value":552},{"type":56,"tag":169,"props":1052,"children":1053},{},[1054,1065,1067,1072,1074,1080,1081,1087,1089,1095,1097,1102],{"type":56,"tag":175,"props":1055,"children":1056},{},[1057,1063],{"type":56,"tag":97,"props":1058,"children":1060},{"className":1059},[],[1061],{"type":61,"value":1062},"dig_url_root",{"type":61,"value":1064}," is the one exception — NO silent default.",{"type":61,"value":1066}," First-time (no memory entry), MUST elicit via ",{"type":56,"tag":97,"props":1068,"children":1070},{"className":1069},[],[1071],{"type":61,"value":964},{"type":61,"value":1073}," before any submit \u002F ",{"type":56,"tag":97,"props":1075,"children":1077},{"className":1076},[],[1078],{"type":61,"value":1079},"osmo data upload",{"type":61,"value":341},{"type":56,"tag":97,"props":1082,"children":1084},{"className":1083},[],[1085],{"type":61,"value":1086},"preflight_urls.sh",{"type":61,"value":1088},". ",{"type":56,"tag":97,"props":1090,"children":1092},{"className":1091},[],[1093],{"type":61,"value":1094},"s3:\u002F\u002Fosmo-workflows\u002Fdig",{"type":61,"value":1096}," is a ",{"type":56,"tag":325,"props":1098,"children":1099},{},[1100],{"type":61,"value":1101},"suggestion to confirm",{"type":61,"value":1103},", never auto-picked (~80 GB+ lands there). Later runs may reuse the remembered value silently. See Step 0 + memory rules (§4).",{"type":56,"tag":169,"props":1105,"children":1106},{},[1107,1117,1119,1124],{"type":56,"tag":175,"props":1108,"children":1109},{},[1110,1112],{"type":61,"value":1111},"Full trigger table, prompt construction, and when-NOT-to-ask exceptions: ",{"type":56,"tag":97,"props":1113,"children":1115},{"className":1114},[],[1116],{"type":61,"value":102},{"type":61,"value":1118}," — load before assembling ",{"type":56,"tag":97,"props":1120,"children":1122},{"className":1121},[],[1123],{"type":61,"value":964},{"type":61,"value":1125}," options for any vague request.",{"type":56,"tag":967,"props":1127,"children":1128},{},[],{"type":56,"tag":64,"props":1130,"children":1132},{"id":1131},"step-0-select-flow-cookbook-and-gather-inputs",[1133],{"type":61,"value":113},{"type":56,"tag":169,"props":1135,"children":1136},{},[1137,1142,1144,1149],{"type":56,"tag":175,"props":1138,"children":1139},{},[1140],{"type":61,"value":1141},"Before this step",{"type":61,"value":1143},", if the request is vague (e.g. \"generate me images\", \"run the PCBA flow\", \"give me defects\"), pause and run the disambiguation cheat sheet above — present candidate interpretations via ",{"type":56,"tag":97,"props":1145,"children":1147},{"className":1146},[],[1148],{"type":61,"value":964},{"type":61,"value":1150}," and let the user pick. Don't auto-pick a load-bearing default the user didn't actually choose.",{"type":56,"tag":575,"props":1152,"children":1154},{"id":1153},"first-time-gate",[1155],{"type":61,"value":1156},"First-time gate",{"type":56,"tag":169,"props":1158,"children":1159},{},[1160,1162,1167,1169,1175,1176,1182,1183,1188],{"type":61,"value":1161},"If memory has no entries for this user, ASK the up-front preference questions in ONE ",{"type":56,"tag":97,"props":1163,"children":1165},{"className":1164},[],[1166],{"type":61,"value":964},{"type":61,"value":1168}," call BEFORE any preflight \u002F ",{"type":56,"tag":97,"props":1170,"children":1172},{"className":1171},[],[1173],{"type":61,"value":1174},"osmo",{"type":61,"value":341},{"type":56,"tag":97,"props":1177,"children":1179},{"className":1178},[],[1180],{"type":61,"value":1181},"kubectl",{"type":61,"value":341},{"type":56,"tag":97,"props":1184,"children":1186},{"className":1185},[],[1187],{"type":61,"value":1079},{"type":61,"value":1189},", save to memory (§4), then proceed. Bundle:",{"type":56,"tag":71,"props":1191,"children":1192},{},[1193,1213,1244,1262],{"type":56,"tag":75,"props":1194,"children":1195},{},[1196,1204,1206,1211],{"type":56,"tag":175,"props":1197,"children":1198},{},[1199],{"type":56,"tag":97,"props":1200,"children":1202},{"className":1201},[],[1203],{"type":61,"value":1062},{"type":61,"value":1205}," — MUST be elicited, not auto-picked. Offer ",{"type":56,"tag":97,"props":1207,"children":1209},{"className":1208},[],[1210],{"type":61,"value":1094},{"type":61,"value":1212}," as a confirmable suggestion; else user provides their own OSMO-supported storage prefix. ~80 GB+ lands here. No escape hatch other than memory-recall of a previously confirmed value.",{"type":56,"tag":75,"props":1214,"children":1215},{},[1216,1227,1229,1235,1237,1243],{"type":56,"tag":175,"props":1217,"children":1218},{},[1219,1221],{"type":61,"value":1220},"Default OSMO ",{"type":56,"tag":97,"props":1222,"children":1224},{"className":1223},[],[1225],{"type":61,"value":1226},"--pool",{"type":61,"value":1228}," — candidates from ",{"type":56,"tag":97,"props":1230,"children":1232},{"className":1231},[],[1233],{"type":61,"value":1234},"osmo profile list",{"type":61,"value":1236}," → ",{"type":56,"tag":97,"props":1238,"children":1240},{"className":1239},[],[1241],{"type":61,"value":1242},"pool.accessible",{"type":61,"value":205},{"type":56,"tag":75,"props":1245,"children":1246},{},[1247,1252,1254,1260],{"type":56,"tag":175,"props":1248,"children":1249},{},[1250],{"type":61,"value":1251},"Pod-template confirmation",{"type":61,"value":1253}," — only when ",{"type":56,"tag":97,"props":1255,"children":1257},{"className":1256},[],[1258],{"type":61,"value":1259},"osmo config show POD_TEMPLATE",{"type":61,"value":1261}," returns 403 (§2 has the exact question).",{"type":56,"tag":75,"props":1263,"children":1264},{},[1265,1270],{"type":56,"tag":175,"props":1266,"children":1267},{},[1268],{"type":61,"value":1269},"Image-edit endpoint",{"type":61,"value":1271}," — Day 0 only: Option A (existing URL) vs Option B (deploy local NIM).",{"type":56,"tag":169,"props":1273,"children":1274},{},[1275],{"type":61,"value":1276},"Subsequent conversations read these silently from memory. Per-flow choices (use case, checkpoint vs finetune, board, knobs) are asked each time — see below.",{"type":56,"tag":575,"props":1278,"children":1280},{"id":1279},"preflight-ordering-after-the-first-time-gate",[1281],{"type":61,"value":1282},"Preflight ordering (after the first-time gate)",{"type":56,"tag":169,"props":1284,"children":1285},{},[1286,1288,1294,1296,1302,1304,1310,1312,1317,1319,1324,1326,1332],{"type":61,"value":1287},"Run §1 ",{"type":56,"tag":97,"props":1289,"children":1291},{"className":1290},[],[1292],{"type":61,"value":1293},"preflight_credentials.sh",{"type":61,"value":1295}," → §2 ",{"type":56,"tag":97,"props":1297,"children":1299},{"className":1298},[],[1300],{"type":61,"value":1301},"preflight_pod_template.sh",{"type":61,"value":1303}," → §3 ",{"type":56,"tag":97,"props":1305,"children":1307},{"className":1306},[],[1308],{"type":61,"value":1309},"preflight_urls.sh \u003Cflow> \u003Cusecase>",{"type":61,"value":1311}," → §4 generate the run stamp. ",{"type":56,"tag":175,"props":1313,"children":1314},{},[1315],{"type":61,"value":1316},"Cadence",{"type":61,"value":1318},": §1 and §2 are once-per-conversation gates with cross-conversation memory caching (see §4a in ",{"type":56,"tag":97,"props":1320,"children":1322},{"className":1321},[],[1323],{"type":61,"value":130},{"type":61,"value":1325},") — skip when memory records them as already verified \u002F user-confirmed. §3 runs before every submit (varies by flow). §4 is the agent's job — fresh ",{"type":56,"tag":97,"props":1327,"children":1329},{"className":1328},[],[1330],{"type":61,"value":1331},"$STAMP",{"type":61,"value":1333}," per submit.",{"type":56,"tag":169,"props":1335,"children":1336},{},[1337,1339,1344,1346,1352,1354,1360,1362,1368,1370,1376,1377,1383],{"type":61,"value":1338},"Pod-template enforcement is two layers: the pre-submit ",{"type":56,"tag":97,"props":1340,"children":1342},{"className":1341},[],[1343],{"type":61,"value":1301},{"type":61,"value":1345}," gate (§2) plus an in-pod runtime preflight on every OV + training task (fails fast on missing ",{"type":56,"tag":97,"props":1347,"children":1349},{"className":1348},[],[1350],{"type":61,"value":1351},"\u002Fusr\u002Fshare\u002Fnvidia\u002Fnvoptix.bin",{"type":61,"value":1353}," or ",{"type":56,"tag":97,"props":1355,"children":1357},{"className":1356},[],[1358],{"type":61,"value":1359},"\u002Fdev\u002Fshm",{"type":61,"value":1361}," \u003C 16 GiB). Runtime failure despite §2 passing → template was patched out → route to ",{"type":56,"tag":97,"props":1363,"children":1365},{"className":1364},[],[1366],{"type":61,"value":1367},"physical-ai-infrastructure-setup-and-resilient-scaling",{"type":61,"value":1369},". Missing creds \u002F URL artifacts → offer to submit ",{"type":56,"tag":97,"props":1371,"children":1373},{"className":1372},[],[1374],{"type":61,"value":1375},"setup\u002Fsetup_\u003Ccase>.yaml",{"type":61,"value":727},{"type":56,"tag":97,"props":1378,"children":1380},{"className":1379},[],[1381],{"type":61,"value":1382},"setup\u002Fsetup_pretrained.yaml",{"type":61,"value":1384}," first.",{"type":56,"tag":169,"props":1386,"children":1387},{},[1388,1390,1395],{"type":61,"value":1389},"Then ask the user in one message — per-flow choices only (the first-time gate above already covered ",{"type":56,"tag":97,"props":1391,"children":1393},{"className":1392},[],[1394],{"type":61,"value":1062},{"type":61,"value":1396},", pool, pod-template, and endpoint preferences; pull those from memory):",{"type":56,"tag":1398,"props":1399,"children":1400},"ol",{},[1401,1411,1450,1512],{"type":56,"tag":75,"props":1402,"children":1403},{},[1404,1409],{"type":56,"tag":175,"props":1405,"children":1406},{},[1407],{"type":61,"value":1408},"Use case",{"type":61,"value":1410}," — PCBA (use Day 0 + pcb cookbook), metal surface (Day 1 + metal_surface cookbook), glass (Day 1 + glass cookbook), or custom?",{"type":56,"tag":75,"props":1412,"children":1413},{},[1414,1419,1421,1426,1428,1434,1436,1442,1444,1449],{"type":56,"tag":175,"props":1415,"children":1416},{},[1417],{"type":61,"value":1418},"Checkpoint available?",{"type":61,"value":1420}," — If yes (",{"type":56,"tag":97,"props":1422,"children":1424},{"className":1423},[],[1425],{"type":61,"value":770},{"type":61,"value":1427},"), use ",{"type":56,"tag":97,"props":1429,"children":1431},{"className":1430},[],[1432],{"type":61,"value":1433},"\u003Cdig_url_root>\u002Fmodels\u002F\u003Cusecase>",{"type":61,"value":1435}," and provide ",{"type":56,"tag":97,"props":1437,"children":1439},{"className":1438},[],[1440],{"type":61,"value":1441},"checkpoint_step",{"type":61,"value":1443},". If no, finetune from ",{"type":56,"tag":97,"props":1445,"children":1447},{"className":1446},[],[1448],{"type":61,"value":542},{"type":61,"value":205},{"type":56,"tag":75,"props":1451,"children":1452},{},[1453,1458,1460,1466,1468,1474,1476,1481,1482,1488,1490,1496,1498,1503,1505,1511],{"type":56,"tag":175,"props":1454,"children":1455},{},[1456],{"type":61,"value":1457},"Local-NIM pool capacity check",{"type":61,"value":1459}," (Day 0 Option B only) — before ",{"type":56,"tag":97,"props":1461,"children":1463},{"className":1462},[],[1464],{"type":61,"value":1465},"kubectl apply",{"type":61,"value":1467},", check ",{"type":56,"tag":97,"props":1469,"children":1471},{"className":1470},[],[1472],{"type":61,"value":1473},"Total Capacity",{"type":61,"value":1475}," via ",{"type":56,"tag":97,"props":1477,"children":1479},{"className":1478},[],[1480],{"type":61,"value":1367},{"type":61,"value":1088},{"type":56,"tag":97,"props":1483,"children":1485},{"className":1484},[],[1486],{"type":61,"value":1487},"Total Capacity \u003C 2",{"type":61,"value":1489}," cannot host NIM + DIG concurrently → ask user to add GPUs or switch to Option A. ",{"type":56,"tag":97,"props":1491,"children":1493},{"className":1492},[],[1494],{"type":61,"value":1495},"image_edit_model",{"type":61,"value":1497}," is always ",{"type":56,"tag":97,"props":1499,"children":1501},{"className":1500},[],[1502],{"type":61,"value":297},{"type":61,"value":1504},", never generic ",{"type":56,"tag":97,"props":1506,"children":1508},{"className":1507},[],[1509],{"type":61,"value":1510},"qwen-image-edit",{"type":61,"value":205},{"type":56,"tag":75,"props":1513,"children":1514},{},[1515,1520,1522,1527,1529,1534,1535,1540,1542,1548,1550,1560],{"type":56,"tag":175,"props":1516,"children":1517},{},[1518],{"type":61,"value":1519},"Save user preferences to memory",{"type":61,"value":1521}," — after the first-time gate (and after any submit diverging from a documented default), persist load-bearing choices (",{"type":56,"tag":97,"props":1523,"children":1525},{"className":1524},[],[1526],{"type":61,"value":1062},{"type":61,"value":1528},", OSMO pool, default board, image-edit endpoint, pod-template state, osmo-admin role). ",{"type":56,"tag":175,"props":1530,"children":1531},{},[1532],{"type":61,"value":1533},"Never save",{"type":61,"value":323},{"type":56,"tag":97,"props":1536,"children":1538},{"className":1537},[],[1539],{"type":61,"value":1495},{"type":61,"value":1541}," (constant — saving invites drift) or ephemeral state (STAMP, one-off ",{"type":56,"tag":97,"props":1543,"children":1545},{"className":1544},[],[1546],{"type":61,"value":1547},"anomaly_types_json",{"type":61,"value":1549},"). Full table: ",{"type":56,"tag":175,"props":1551,"children":1552},{},[1553,1558],{"type":56,"tag":97,"props":1554,"children":1556},{"className":1555},[],[1557],{"type":61,"value":130},{"type":61,"value":1559}," §4a \"Memory rules\"",{"type":61,"value":1561},". Read relevant memories at the start of every new conversation and apply silently.",{"type":56,"tag":169,"props":1563,"children":1564},{},[1565,1567,1573,1575,1581],{"type":61,"value":1566},"Review the relevant flow reference before asking — most values have sensible defaults. Day 1 routing: PCBA defaults to ",{"type":56,"tag":97,"props":1568,"children":1570},{"className":1569},[],[1571],{"type":61,"value":1572},"real_alignment",{"type":61,"value":1574},"; metal\u002Fglass have no USD flow so always ",{"type":56,"tag":97,"props":1576,"children":1578},{"className":1577},[],[1579],{"type":61,"value":1580},"manual_roi",{"type":61,"value":1582},"; don't ask the user \"manual or real-alignment?\" for PCBA unless they explicitly ask to skip alignment.",{"type":56,"tag":967,"props":1584,"children":1585},{},[],{"type":56,"tag":64,"props":1587,"children":1589},{"id":1588},"common-preconditions-all-flows",[1590],{"type":61,"value":1591},"Common Preconditions (all flows)",{"type":56,"tag":169,"props":1593,"children":1594},{},[1595,1597,1602],{"type":61,"value":1596},"Quick reference. Long-form: ",{"type":56,"tag":97,"props":1598,"children":1600},{"className":1599},[],[1601],{"type":61,"value":130},{"type":61,"value":205},{"type":56,"tag":1398,"props":1604,"children":1605},{},[1606,1686,1711,2037,2078],{"type":56,"tag":75,"props":1607,"children":1608},{},[1609,1614,1616,1629,1631,1637,1639,1645,1647,1653,1655,1661,1663,1669,1671,1677,1679,1684],{"type":56,"tag":175,"props":1610,"children":1611},{},[1612],{"type":61,"value":1613},"OSMO credentials + tokens",{"type":61,"value":1615}," — once per conversation. ",{"type":56,"tag":175,"props":1617,"children":1618},{},[1619,1621,1627],{"type":61,"value":1620},"If a ",{"type":56,"tag":97,"props":1622,"children":1624},{"className":1623},[],[1625],{"type":61,"value":1626},".env",{"type":61,"value":1628}," exists in the workspace, source it first",{"type":61,"value":1630}," (",{"type":56,"tag":97,"props":1632,"children":1634},{"className":1633},[],[1635],{"type":61,"value":1636},"set -a; . .\u002F.env; set +a",{"type":61,"value":1638},") so ",{"type":56,"tag":97,"props":1640,"children":1642},{"className":1641},[],[1643],{"type":61,"value":1644},"HF_TOKEN",{"type":61,"value":1646}," is exported. Run ",{"type":56,"tag":97,"props":1648,"children":1650},{"className":1649},[],[1651],{"type":61,"value":1652},"scripts\u002Fpreflight_credentials.sh",{"type":61,"value":1654},"; authoritative check is the OSMO cred ",{"type":56,"tag":97,"props":1656,"children":1658},{"className":1657},[],[1659],{"type":61,"value":1660},"hf-token",{"type":61,"value":1662}," is provisioned (images are public on ",{"type":56,"tag":97,"props":1664,"children":1666},{"className":1665},[],[1667],{"type":61,"value":1668},"nvcr.io\u002Fnvidia\u002F",{"type":61,"value":1670}," — no registry cred needed). Pass ",{"type":56,"tag":97,"props":1672,"children":1674},{"className":1673},[],[1675],{"type":61,"value":1676},"--no-probe",{"type":61,"value":1678}," in restricted-egress shells. See ",{"type":56,"tag":97,"props":1680,"children":1682},{"className":1681},[],[1683],{"type":61,"value":130},{"type":61,"value":1685}," §1.",{"type":56,"tag":75,"props":1687,"children":1688},{},[1689,1694,1696,1702,1704,1709],{"type":56,"tag":175,"props":1690,"children":1691},{},[1692],{"type":61,"value":1693},"Pod template",{"type":61,"value":1695}," — once per conversation, with cross-conversation memory caching (see Step 0 §6). Skip when memory records the cluster verified \u002F user-confirmed \u002F 409-skipped. Otherwise run ",{"type":56,"tag":97,"props":1697,"children":1699},{"className":1698},[],[1700],{"type":61,"value":1701},"scripts\u002Fpreflight_pod_template.sh",{"type":61,"value":1703}," and branch on exit code (0=verified \u002F 1=patch via infra skill \u002F 2=ask-user (HTTP 403) \u002F 3=skip (HTTP 409) \u002F 4=env-fix). Full branching prose and prompts in ",{"type":56,"tag":97,"props":1705,"children":1707},{"className":1706},[],[1708],{"type":61,"value":130},{"type":61,"value":1710}," §2.",{"type":56,"tag":75,"props":1712,"children":1713},{},[1714,1719,1721,1727,1729,1747,1749,1755,1756,1761,1763,1994,1998,2000,2006,2008,2014,2015,2021,2022,2028,2030,2035],{"type":56,"tag":175,"props":1715,"children":1716},{},[1717],{"type":61,"value":1718},"Required URL artifacts",{"type":61,"value":1720}," — before every submit. Run ",{"type":56,"tag":97,"props":1722,"children":1724},{"className":1723},[],[1725],{"type":61,"value":1726},"DIG_URL_ROOT=\u003Cdig_url_root> scripts\u002Fpreflight_urls.sh \u003Cflow> \u003Cusecase> [variant]",{"type":61,"value":1728},". If anything is missing, ",{"type":56,"tag":175,"props":1730,"children":1731},{},[1732,1734,1739,1740,1745],{"type":61,"value":1733},"stop and submit the relevant ",{"type":56,"tag":97,"props":1735,"children":1737},{"className":1736},[],[1738],{"type":61,"value":1375},{"type":61,"value":727},{"type":56,"tag":97,"props":1741,"children":1743},{"className":1742},[],[1744],{"type":61,"value":1382},{"type":61,"value":1746}," first",{"type":61,"value":1748}," (the OSMO setup workflows) — see ",{"type":56,"tag":97,"props":1750,"children":1752},{"className":1751},[],[1753],{"type":61,"value":1754},"references\u002Fsetup.md",{"type":61,"value":1088},{"type":56,"tag":175,"props":1757,"children":1758},{},[1759],{"type":61,"value":1760},"Never download assets locally to work around a problem; if setup fails on credentials, ask the user to rectify them and re-submit on OSMO.",{"type":61,"value":1762}," Per-flow checklist:",{"type":56,"tag":212,"props":1764,"children":1765},{},[1766,1791],{"type":56,"tag":216,"props":1767,"children":1768},{},[1769],{"type":56,"tag":220,"props":1770,"children":1771},{},[1772,1776,1780],{"type":56,"tag":224,"props":1773,"children":1774},{},[1775],{"type":61,"value":228},{"type":56,"tag":224,"props":1777,"children":1778},{},[1779],{"type":61,"value":1408},{"type":56,"tag":224,"props":1781,"children":1782},{},[1783,1785],{"type":61,"value":1784},"Required URL artifacts under ",{"type":56,"tag":97,"props":1786,"children":1788},{"className":1787},[],[1789],{"type":61,"value":1790},"\u003Cdig_url_root>",{"type":56,"tag":250,"props":1792,"children":1793},{},[1794,1835,1856,1876,1911,1945,1967],{"type":56,"tag":220,"props":1795,"children":1796},{},[1797,1801,1805],{"type":56,"tag":257,"props":1798,"children":1799},{},[1800],{"type":61,"value":264},{"type":56,"tag":257,"props":1802,"children":1803},{},[1804],{"type":61,"value":310},{"type":56,"tag":257,"props":1806,"children":1807},{},[1808,1814,1816,1822,1823,1829,1830],{"type":56,"tag":97,"props":1809,"children":1811},{"className":1810},[],[1812],{"type":61,"value":1813},"models\u002Fpretrained",{"type":61,"value":1815},", ",{"type":56,"tag":97,"props":1817,"children":1819},{"className":1818},[],[1820],{"type":61,"value":1821},"models\u002Fpcb",{"type":61,"value":1815},{"type":56,"tag":97,"props":1824,"children":1826},{"className":1825},[],[1827],{"type":61,"value":1828},"datasets\u002Fpcb\u002Fraw",{"type":61,"value":1815},{"type":56,"tag":97,"props":1831,"children":1833},{"className":1832},[],[1834],{"type":61,"value":434},{"type":56,"tag":220,"props":1836,"children":1837},{},[1838,1842,1846],{"type":56,"tag":257,"props":1839,"children":1840},{},[1841],{"type":61,"value":321},{"type":56,"tag":257,"props":1843,"children":1844},{},[1845],{"type":61,"value":310},{"type":56,"tag":257,"props":1847,"children":1848},{},[1849,1854],{"type":56,"tag":97,"props":1850,"children":1852},{"className":1851},[],[1853],{"type":61,"value":434},{"type":61,"value":1855}," only",{"type":56,"tag":220,"props":1857,"children":1858},{},[1859,1863,1867],{"type":56,"tag":257,"props":1860,"children":1861},{},[1862],{"type":61,"value":384},{"type":56,"tag":257,"props":1864,"children":1865},{},[1866],{"type":61,"value":310},{"type":56,"tag":257,"props":1868,"children":1869},{},[1870,1875],{"type":56,"tag":97,"props":1871,"children":1873},{"className":1872},[],[1874],{"type":61,"value":434},{"type":61,"value":1855},{"type":56,"tag":220,"props":1877,"children":1878},{},[1879,1884,1889],{"type":56,"tag":257,"props":1880,"children":1881},{},[1882],{"type":61,"value":1883},"Day 1",{"type":56,"tag":257,"props":1885,"children":1886},{},[1887],{"type":61,"value":1888},"Metal surface",{"type":56,"tag":257,"props":1890,"children":1891},{},[1892,1897,1898,1904,1905],{"type":56,"tag":97,"props":1893,"children":1895},{"className":1894},[],[1896],{"type":61,"value":1813},{"type":61,"value":1815},{"type":56,"tag":97,"props":1899,"children":1901},{"className":1900},[],[1902],{"type":61,"value":1903},"models\u002Fmetal_surface",{"type":61,"value":1815},{"type":56,"tag":97,"props":1906,"children":1908},{"className":1907},[],[1909],{"type":61,"value":1910},"datasets\u002Fmetal_surface\u002Fraw",{"type":56,"tag":220,"props":1912,"children":1913},{},[1914,1918,1923],{"type":56,"tag":257,"props":1915,"children":1916},{},[1917],{"type":61,"value":1883},{"type":56,"tag":257,"props":1919,"children":1920},{},[1921],{"type":61,"value":1922},"Glass",{"type":56,"tag":257,"props":1924,"children":1925},{},[1926,1931,1932,1938,1939],{"type":56,"tag":97,"props":1927,"children":1929},{"className":1928},[],[1930],{"type":61,"value":1813},{"type":61,"value":1815},{"type":56,"tag":97,"props":1933,"children":1935},{"className":1934},[],[1936],{"type":61,"value":1937},"models\u002Fglass",{"type":61,"value":1815},{"type":56,"tag":97,"props":1940,"children":1942},{"className":1941},[],[1943],{"type":61,"value":1944},"datasets\u002Fglass\u002Fraw",{"type":56,"tag":220,"props":1946,"children":1947},{},[1948,1953,1957],{"type":56,"tag":257,"props":1949,"children":1950},{},[1951],{"type":61,"value":1952},"Day 1 real-photo alignment",{"type":56,"tag":257,"props":1954,"children":1955},{},[1956],{"type":61,"value":310},{"type":56,"tag":257,"props":1958,"children":1959},{},[1960,1962],{"type":61,"value":1961},"Day 1 PCBA plus ",{"type":56,"tag":97,"props":1963,"children":1965},{"className":1964},[],[1966],{"type":61,"value":434},{"type":56,"tag":220,"props":1968,"children":1969},{},[1970,1974,1979],{"type":56,"tag":257,"props":1971,"children":1972},{},[1973],{"type":61,"value":512},{"type":56,"tag":257,"props":1975,"children":1976},{},[1977],{"type":61,"value":1978},"Any",{"type":56,"tag":257,"props":1980,"children":1981},{},[1982,1987,1988],{"type":56,"tag":97,"props":1983,"children":1985},{"className":1984},[],[1986],{"type":61,"value":1813},{"type":61,"value":1815},{"type":56,"tag":97,"props":1989,"children":1991},{"className":1990},[],[1992],{"type":61,"value":1993},"datasets\u002F\u003Cusecase>\u002Fraw",{"type":56,"tag":1995,"props":1996,"children":1997},"br",{},[],{"type":61,"value":1999},"Built-in ",{"type":56,"tag":97,"props":2001,"children":2003},{"className":2002},[],[2004],{"type":61,"value":2005},"usecase",{"type":61,"value":2007}," values are ",{"type":56,"tag":97,"props":2009,"children":2011},{"className":2010},[],[2012],{"type":61,"value":2013},"pcb",{"type":61,"value":1815},{"type":56,"tag":97,"props":2016,"children":2018},{"className":2017},[],[2019],{"type":61,"value":2020},"metal_surface",{"type":61,"value":1815},{"type":56,"tag":97,"props":2023,"children":2025},{"className":2024},[],[2026],{"type":61,"value":2027},"glass",{"type":61,"value":2029},". See ",{"type":56,"tag":97,"props":2031,"children":2033},{"className":2032},[],[2034],{"type":61,"value":130},{"type":61,"value":2036}," §3.",{"type":56,"tag":75,"props":2038,"children":2039},{},[2040,2045,2047,2053,2055,2061,2063,2069,2071,2076],{"type":56,"tag":175,"props":2041,"children":2042},{},[2043],{"type":61,"value":2044},"Name stamping",{"type":61,"value":2046}," — regenerate ",{"type":56,"tag":97,"props":2048,"children":2050},{"className":2049},[],[2051],{"type":61,"value":2052},"$STAMP=$(cat \u002Fproc\u002Fsys\u002Fkernel\u002Frandom\u002Fuuid | cut -c1-8)",{"type":61,"value":2054}," before every submit and pass ",{"type":56,"tag":97,"props":2056,"children":2058},{"className":2057},[],[2059],{"type":61,"value":2060},"--set name=\u003Cflow>-$STAMP",{"type":61,"value":2062},". Production YAMLs ship no ",{"type":56,"tag":97,"props":2064,"children":2066},{"className":2065},[],[2067],{"type":61,"value":2068},"name",{"type":61,"value":2070}," default. See ",{"type":56,"tag":97,"props":2072,"children":2074},{"className":2073},[],[2075],{"type":61,"value":130},{"type":61,"value":2077}," §4.",{"type":56,"tag":75,"props":2079,"children":2080},{},[2081,2086,2088,2094,2096,2102,2104,2110,2112,2117],{"type":56,"tag":175,"props":2082,"children":2083},{},[2084],{"type":61,"value":2085},"Glass case (UC3) — Roboflow zip",{"type":61,"value":2087}," — only for ",{"type":56,"tag":97,"props":2089,"children":2091},{"className":2090},[],[2092],{"type":61,"value":2093},"setup_glass.yaml",{"type":61,"value":2095},". Upload ",{"type":56,"tag":97,"props":2097,"children":2099},{"className":2098},[],[2100],{"type":61,"value":2101},"mobile_screen.zip",{"type":61,"value":2103}," to an OSMO URL prefix first; pass ",{"type":56,"tag":97,"props":2105,"children":2107},{"className":2106},[],[2108],{"type":61,"value":2109},"--set uc3_zip_url_root=\u003Cprefix>",{"type":61,"value":2111},". Full procedure: ",{"type":56,"tag":97,"props":2113,"children":2115},{"className":2114},[],[2116],{"type":61,"value":1754},{"type":61,"value":2118}," §\"Glass case (UC3)\".",{"type":56,"tag":967,"props":2120,"children":2121},{},[],{"type":56,"tag":64,"props":2123,"children":2125},{"id":2124},"flow-walkthroughs",[2126],{"type":61,"value":140},{"type":56,"tag":169,"props":2128,"children":2129},{},[2130,2132,2137],{"type":61,"value":2131},"Each flow's full walkthrough — group diagrams, prerequisites, submit-command variants, data handoffs, per-stage troubleshooting — lives under ",{"type":56,"tag":97,"props":2133,"children":2135},{"className":2134},[],[2136],{"type":61,"value":148},{"type":61,"value":2138},". The agent should read the matching file before submitting any flow it hasn't run in the current conversation.",{"type":56,"tag":212,"props":2140,"children":2141},{},[2142,2162],{"type":56,"tag":216,"props":2143,"children":2144},{},[2145],{"type":56,"tag":220,"props":2146,"children":2147},{},[2148,2152,2157],{"type":56,"tag":224,"props":2149,"children":2150},{},[2151],{"type":61,"value":228},{"type":56,"tag":224,"props":2153,"children":2154},{},[2155],{"type":61,"value":2156},"Workflow YAML",{"type":56,"tag":224,"props":2158,"children":2159},{},[2160],{"type":61,"value":2161},"Walkthrough",{"type":56,"tag":250,"props":2163,"children":2164},{},[2165,2194,2223,2251,2279,2308],{"type":56,"tag":220,"props":2166,"children":2167},{},[2168,2176,2185],{"type":56,"tag":257,"props":2169,"children":2170},{},[2171],{"type":56,"tag":175,"props":2172,"children":2173},{},[2174],{"type":61,"value":2175},"Day 0 — Texture Defects (PCBA)",{"type":56,"tag":257,"props":2177,"children":2178},{},[2179],{"type":56,"tag":97,"props":2180,"children":2182},{"className":2181},[],[2183],{"type":61,"value":2184},"assets\u002Fconfigs\u002Ftexture_defect_generation_day0.yaml",{"type":56,"tag":257,"props":2186,"children":2187},{},[2188],{"type":56,"tag":97,"props":2189,"children":2191},{"className":2190},[],[2192],{"type":61,"value":2193},"references\u002Fflows\u002Ftexture_defect_generation_day0.md",{"type":56,"tag":220,"props":2195,"children":2196},{},[2197,2205,2214],{"type":56,"tag":257,"props":2198,"children":2199},{},[2200],{"type":56,"tag":175,"props":2201,"children":2202},{},[2203],{"type":61,"value":2204},"Day 0 — Good Image (PCBA)",{"type":56,"tag":257,"props":2206,"children":2207},{},[2208],{"type":56,"tag":97,"props":2209,"children":2211},{"className":2210},[],[2212],{"type":61,"value":2213},"assets\u002Fconfigs\u002Fgood_image_generation.yaml",{"type":56,"tag":257,"props":2215,"children":2216},{},[2217],{"type":56,"tag":97,"props":2218,"children":2220},{"className":2219},[],[2221],{"type":61,"value":2222},"references\u002Fflows\u002Fgood_image_generation.md",{"type":56,"tag":220,"props":2224,"children":2225},{},[2226,2234,2243],{"type":56,"tag":257,"props":2227,"children":2228},{},[2229],{"type":56,"tag":175,"props":2230,"children":2231},{},[2232],{"type":61,"value":2233},"Day 0 — Structural Defects (PCBA)",{"type":56,"tag":257,"props":2235,"children":2236},{},[2237],{"type":56,"tag":97,"props":2238,"children":2240},{"className":2239},[],[2241],{"type":61,"value":2242},"assets\u002Fconfigs\u002Fstructural_defect_generation.yaml",{"type":56,"tag":257,"props":2244,"children":2245},{},[2246],{"type":56,"tag":97,"props":2247,"children":2249},{"className":2248},[],[2250],{"type":61,"value":956},{"type":56,"tag":220,"props":2252,"children":2253},{},[2254,2262,2271],{"type":56,"tag":257,"props":2255,"children":2256},{},[2257],{"type":56,"tag":175,"props":2258,"children":2259},{},[2260],{"type":61,"value":2261},"Day 1 — Infer + Label (real-photo alignment, default PCBA)",{"type":56,"tag":257,"props":2263,"children":2264},{},[2265],{"type":56,"tag":97,"props":2266,"children":2268},{"className":2267},[],[2269],{"type":61,"value":2270},"assets\u002Fconfigs\u002Ftexture_defect_generation_day1_real_alignment.yaml",{"type":56,"tag":257,"props":2272,"children":2273},{},[2274],{"type":56,"tag":97,"props":2275,"children":2277},{"className":2276},[],[2278],{"type":61,"value":1049},{"type":56,"tag":220,"props":2280,"children":2281},{},[2282,2290,2299],{"type":56,"tag":257,"props":2283,"children":2284},{},[2285],{"type":56,"tag":175,"props":2286,"children":2287},{},[2288],{"type":61,"value":2289},"Day 1 — Infer + Label (manual ROI, metal\u002Fglass + PCBA experimentation)",{"type":56,"tag":257,"props":2291,"children":2292},{},[2293],{"type":56,"tag":97,"props":2294,"children":2296},{"className":2295},[],[2297],{"type":61,"value":2298},"assets\u002Fconfigs\u002Ftexture_defect_generation_day1_manual_roi.yaml",{"type":56,"tag":257,"props":2300,"children":2301},{},[2302],{"type":56,"tag":97,"props":2303,"children":2305},{"className":2304},[],[2306],{"type":61,"value":2307},"references\u002Fflows\u002Ftexture_defect_generation_day1_manual_roi.md",{"type":56,"tag":220,"props":2309,"children":2310},{},[2311,2318,2327],{"type":56,"tag":257,"props":2312,"children":2313},{},[2314],{"type":56,"tag":175,"props":2315,"children":2316},{},[2317],{"type":61,"value":512},{"type":56,"tag":257,"props":2319,"children":2320},{},[2321],{"type":56,"tag":97,"props":2322,"children":2324},{"className":2323},[],[2325],{"type":61,"value":2326},"assets\u002Fconfigs\u002Ffinetune.yaml",{"type":56,"tag":257,"props":2328,"children":2329},{},[2330],{"type":56,"tag":97,"props":2331,"children":2333},{"className":2332},[],[2334],{"type":61,"value":2335},"references\u002Fflows\u002Ffinetune.md",{"type":56,"tag":575,"props":2337,"children":2339},{"id":2338},"cross-flow-invariants",[2340],{"type":61,"value":2341},"Cross-flow invariants",{"type":56,"tag":71,"props":2343,"children":2344},{},[2345,2378,2391],{"type":56,"tag":75,"props":2346,"children":2347},{},[2348,2353,2355,2361,2363,2369,2371,2376],{"type":56,"tag":97,"props":2349,"children":2351},{"className":2350},[],[2352],{"type":61,"value":770},{"type":61,"value":2354}," (default) → passthrough against ",{"type":56,"tag":97,"props":2356,"children":2358},{"className":2357},[],[2359],{"type":61,"value":2360},"models\u002F\u003Cusecase>",{"type":61,"value":2362},". Set to ",{"type":56,"tag":97,"props":2364,"children":2366},{"className":2365},[],[2367],{"type":61,"value":2368},"false",{"type":61,"value":2370}," to insert an in-pod ",{"type":56,"tag":97,"props":2372,"children":2374},{"className":2373},[],[2375],{"type":61,"value":762},{"type":61,"value":2377}," group (cookbook yq-patched in-pod, no pre-submit render step).",{"type":56,"tag":75,"props":2379,"children":2380},{},[2381,2383,2389],{"type":61,"value":2382},"Day 0 emits per-cell ",{"type":56,"tag":97,"props":2384,"children":2386},{"className":2385},[],[2387],{"type":61,"value":2388},"crop\u002F\u003CMATERIAL>\u002F\u003Ccell>\u002F...",{"type":61,"value":2390}," trees; Day 1 emits per-ROI crops registered against the USD; structural emits flat per-component crops.",{"type":56,"tag":75,"props":2392,"children":2393},{},[2394,2396,2401,2402,2407,2409,2414,2416,2421],{"type":61,"value":2395},"Shipped per-usecase ",{"type":56,"tag":97,"props":2397,"children":2399},{"className":2398},[],[2400],{"type":61,"value":1441},{"type":61,"value":727},{"type":56,"tag":97,"props":2403,"children":2405},{"className":2404},[],[2406],{"type":61,"value":1547},{"type":61,"value":2408}," defaults: see ",{"type":56,"tag":97,"props":2410,"children":2412},{"className":2411},[],[2413],{"type":61,"value":130},{"type":61,"value":2415}," §\"Shipped checkpoint and ",{"type":56,"tag":97,"props":2417,"children":2419},{"className":2418},[],[2420],{"type":61,"value":1547},{"type":61,"value":2422}," defaults\".",{"type":56,"tag":967,"props":2424,"children":2425},{},[],{"type":56,"tag":64,"props":2427,"children":2429},{"id":2428},"osmo-monitoring",[2430],{"type":61,"value":158},{"type":56,"tag":169,"props":2432,"children":2433},{},[2434,2470],{"type":56,"tag":175,"props":2435,"children":2436},{},[2437,2439,2445,2447,2453,2454,2460,2462,2468],{"type":61,"value":2438},"Load ",{"type":56,"tag":97,"props":2440,"children":2442},{"className":2441},[],[2443],{"type":61,"value":2444},"references\u002Fmonitoring.md",{"type":61,"value":2446}," before any ",{"type":56,"tag":97,"props":2448,"children":2450},{"className":2449},[],[2451],{"type":61,"value":2452},"osmo workflow submit",{"type":61,"value":1815},{"type":56,"tag":97,"props":2455,"children":2457},{"className":2456},[],[2458],{"type":61,"value":2459},"osmo workflow query",{"type":61,"value":2461},", or ",{"type":56,"tag":97,"props":2463,"children":2465},{"className":2464},[],[2466],{"type":61,"value":2467},"osmo workflow logs",{"type":61,"value":2469}," action in this skill.",{"type":61,"value":2471}," It defines the polling cadence, task-status interpretation, log-pull escalation thresholds, failure-classification routing, and what to surface to the user vs. silently retry. Do not assemble a post-submit watch loop or status summary from memory — re-read it on the first such action of every conversation.",{"type":56,"tag":2473,"props":2474,"children":2479},"pre",{"className":2475,"code":2476,"language":2477,"meta":2478,"style":2478},"language-bash shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","osmo workflow query \u003Cworkflow_id> --format-type json | jq '{status, tasks: [.groups[].tasks[] | {name, status, exit_code}]}'\nosmo workflow logs \u003Cworkflow_id> -t \u003Ctask_name> -n 200\nosmo data download \u003Cdig_url_root>\u002Fruns\u002F\u003Cname>\u002Fanomaly .\u002Foutput\u002Fanomaly-\u003Cname>\u002F\n","bash","",[2480],{"type":56,"tag":97,"props":2481,"children":2482},{"__ignoreMap":2478},[2483,2562,2629],{"type":56,"tag":2484,"props":2485,"children":2488},"span",{"class":2486,"line":2487},"line",1,[2489,2494,2500,2505,2511,2516,2522,2527,2532,2537,2542,2547,2552,2557],{"type":56,"tag":2484,"props":2490,"children":2492},{"style":2491},"--shiki-light:#E2931D;--shiki-default:#FFCB6B;--shiki-dark:#FFCB6B",[2493],{"type":61,"value":1174},{"type":56,"tag":2484,"props":2495,"children":2497},{"style":2496},"--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D",[2498],{"type":61,"value":2499}," workflow",{"type":56,"tag":2484,"props":2501,"children":2502},{"style":2496},[2503],{"type":61,"value":2504}," query",{"type":56,"tag":2484,"props":2506,"children":2508},{"style":2507},"--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF",[2509],{"type":61,"value":2510}," \u003C",{"type":56,"tag":2484,"props":2512,"children":2513},{"style":2496},[2514],{"type":61,"value":2515},"workflow_i",{"type":56,"tag":2484,"props":2517,"children":2519},{"style":2518},"--shiki-light:#90A4AE;--shiki-default:#EEFFFF;--shiki-dark:#BABED8",[2520],{"type":61,"value":2521},"d",{"type":56,"tag":2484,"props":2523,"children":2524},{"style":2507},[2525],{"type":61,"value":2526},">",{"type":56,"tag":2484,"props":2528,"children":2529},{"style":2496},[2530],{"type":61,"value":2531}," --format-type",{"type":56,"tag":2484,"props":2533,"children":2534},{"style":2496},[2535],{"type":61,"value":2536}," json",{"type":56,"tag":2484,"props":2538,"children":2539},{"style":2507},[2540],{"type":61,"value":2541}," |",{"type":56,"tag":2484,"props":2543,"children":2544},{"style":2491},[2545],{"type":61,"value":2546}," jq",{"type":56,"tag":2484,"props":2548,"children":2549},{"style":2507},[2550],{"type":61,"value":2551}," '",{"type":56,"tag":2484,"props":2553,"children":2554},{"style":2496},[2555],{"type":61,"value":2556},"{status, tasks: [.groups[].tasks[] | {name, status, exit_code}]}",{"type":56,"tag":2484,"props":2558,"children":2559},{"style":2507},[2560],{"type":61,"value":2561},"'\n",{"type":56,"tag":2484,"props":2563,"children":2565},{"class":2486,"line":2564},2,[2566,2570,2574,2579,2583,2587,2591,2595,2600,2604,2609,2614,2618,2623],{"type":56,"tag":2484,"props":2567,"children":2568},{"style":2491},[2569],{"type":61,"value":1174},{"type":56,"tag":2484,"props":2571,"children":2572},{"style":2496},[2573],{"type":61,"value":2499},{"type":56,"tag":2484,"props":2575,"children":2576},{"style":2496},[2577],{"type":61,"value":2578}," logs",{"type":56,"tag":2484,"props":2580,"children":2581},{"style":2507},[2582],{"type":61,"value":2510},{"type":56,"tag":2484,"props":2584,"children":2585},{"style":2496},[2586],{"type":61,"value":2515},{"type":56,"tag":2484,"props":2588,"children":2589},{"style":2518},[2590],{"type":61,"value":2521},{"type":56,"tag":2484,"props":2592,"children":2593},{"style":2507},[2594],{"type":61,"value":2526},{"type":56,"tag":2484,"props":2596,"children":2597},{"style":2496},[2598],{"type":61,"value":2599}," -t",{"type":56,"tag":2484,"props":2601,"children":2602},{"style":2507},[2603],{"type":61,"value":2510},{"type":56,"tag":2484,"props":2605,"children":2606},{"style":2496},[2607],{"type":61,"value":2608},"task_nam",{"type":56,"tag":2484,"props":2610,"children":2611},{"style":2518},[2612],{"type":61,"value":2613},"e",{"type":56,"tag":2484,"props":2615,"children":2616},{"style":2507},[2617],{"type":61,"value":2526},{"type":56,"tag":2484,"props":2619,"children":2620},{"style":2496},[2621],{"type":61,"value":2622}," -n",{"type":56,"tag":2484,"props":2624,"children":2626},{"style":2625},"--shiki-light:#F76D47;--shiki-default:#F78C6C;--shiki-dark:#F78C6C",[2627],{"type":61,"value":2628}," 200\n",{"type":56,"tag":2484,"props":2630,"children":2632},{"class":2486,"line":2631},3,[2633,2637,2642,2647,2651,2656,2661,2665,2670,2675,2680,2684,2688,2693,2698,2702,2706,2710,2714],{"type":56,"tag":2484,"props":2634,"children":2635},{"style":2491},[2636],{"type":61,"value":1174},{"type":56,"tag":2484,"props":2638,"children":2639},{"style":2496},[2640],{"type":61,"value":2641}," data",{"type":56,"tag":2484,"props":2643,"children":2644},{"style":2496},[2645],{"type":61,"value":2646}," download",{"type":56,"tag":2484,"props":2648,"children":2649},{"style":2507},[2650],{"type":61,"value":2510},{"type":56,"tag":2484,"props":2652,"children":2653},{"style":2496},[2654],{"type":61,"value":2655},"dig_url_roo",{"type":56,"tag":2484,"props":2657,"children":2658},{"style":2518},[2659],{"type":61,"value":2660},"t",{"type":56,"tag":2484,"props":2662,"children":2663},{"style":2507},[2664],{"type":61,"value":2526},{"type":56,"tag":2484,"props":2666,"children":2667},{"style":2496},[2668],{"type":61,"value":2669},"\u002Fruns\u002F",{"type":56,"tag":2484,"props":2671,"children":2672},{"style":2507},[2673],{"type":61,"value":2674},"\u003C",{"type":56,"tag":2484,"props":2676,"children":2677},{"style":2496},[2678],{"type":61,"value":2679},"nam",{"type":56,"tag":2484,"props":2681,"children":2682},{"style":2518},[2683],{"type":61,"value":2613},{"type":56,"tag":2484,"props":2685,"children":2686},{"style":2507},[2687],{"type":61,"value":2526},{"type":56,"tag":2484,"props":2689,"children":2690},{"style":2496},[2691],{"type":61,"value":2692},"\u002Fanomaly",{"type":56,"tag":2484,"props":2694,"children":2695},{"style":2496},[2696],{"type":61,"value":2697}," .\u002Foutput\u002Fanomaly-",{"type":56,"tag":2484,"props":2699,"children":2700},{"style":2507},[2701],{"type":61,"value":2674},{"type":56,"tag":2484,"props":2703,"children":2704},{"style":2496},[2705],{"type":61,"value":2679},{"type":56,"tag":2484,"props":2707,"children":2708},{"style":2518},[2709],{"type":61,"value":2613},{"type":56,"tag":2484,"props":2711,"children":2712},{"style":2507},[2713],{"type":61,"value":2526},{"type":56,"tag":2484,"props":2715,"children":2716},{"style":2496},[2717],{"type":61,"value":2718},"\u002F\n",{"type":56,"tag":169,"props":2720,"children":2721},{},[2722,2724,2729,2731,2737,2739,2745,2747,2753],{"type":61,"value":2723},"Monitoring discipline: ",{"type":56,"tag":97,"props":2725,"children":2727},{"className":2726},[],[2728],{"type":61,"value":2444},{"type":61,"value":2730},". Retrieval: ",{"type":56,"tag":97,"props":2732,"children":2734},{"className":2733},[],[2735],{"type":61,"value":2736},"references\u002Foutput_retrieval.md",{"type":61,"value":2738},". Presentation: ",{"type":56,"tag":97,"props":2740,"children":2742},{"className":2741},[],[2743],{"type":61,"value":2744},"references\u002Foutput_rendering.md",{"type":61,"value":2746},". Gotchas: ",{"type":56,"tag":97,"props":2748,"children":2750},{"className":2749},[],[2751],{"type":61,"value":2752},"references\u002Ftroubleshooting.md",{"type":61,"value":205},{"type":56,"tag":967,"props":2755,"children":2756},{},[],{"type":56,"tag":64,"props":2758,"children":2760},{"id":2759},"response-template",[2761],{"type":61,"value":2762},"Response Template",{"type":56,"tag":169,"props":2764,"children":2765},{},[2766],{"type":61,"value":2767},"For \"show me the plan \u002F recipe\" requests, emit your final response with these labeled sections (so nothing truncates mid-recipe):",{"type":56,"tag":169,"props":2769,"children":2770},{},[2771,2776,2777,2783,2784],{"type":56,"tag":175,"props":2772,"children":2773},{},[2774],{"type":61,"value":2775},"Workflow:",{"type":61,"value":323},{"type":56,"tag":97,"props":2778,"children":2780},{"className":2779},[],[2781],{"type":61,"value":2782},"\u003Cflow name>",{"type":61,"value":1236},{"type":56,"tag":97,"props":2785,"children":2787},{"className":2786},[],[2788],{"type":61,"value":2789},"assets\u002Fconfigs\u002F\u003Cyaml>",{"type":56,"tag":169,"props":2791,"children":2792},{},[2793,2798,2799,2804,2806],{"type":56,"tag":175,"props":2794,"children":2795},{},[2796],{"type":61,"value":2797},"Preflights:",{"type":61,"value":323},{"type":56,"tag":97,"props":2800,"children":2802},{"className":2801},[],[2803],{"type":61,"value":1652},{"type":61,"value":2805},"; ",{"type":56,"tag":97,"props":2807,"children":2809},{"className":2808},[],[2810],{"type":61,"value":2811},"scripts\u002Fpreflight_urls.sh \u003C0|1|finetune> \u003Cusecase> [variant]",{"type":56,"tag":169,"props":2813,"children":2814},{},[2815,2827],{"type":56,"tag":175,"props":2816,"children":2817},{},[2818,2820,2825],{"type":61,"value":2819},"Required URL Artifacts under ",{"type":56,"tag":97,"props":2821,"children":2823},{"className":2822},[],[2824],{"type":61,"value":1790},{"type":61,"value":2826},":",{"type":61,"value":2828}," enumerate per Common Preconditions §3 for the chosen flow.",{"type":56,"tag":169,"props":2830,"children":2831},{},[2832],{"type":56,"tag":175,"props":2833,"children":2834},{},[2835],{"type":61,"value":2836},"Submit Command:",{"type":56,"tag":2473,"props":2838,"children":2840},{"className":2475,"code":2839,"language":2477,"meta":2478,"style":2478},"STAMP=$(cat \u002Fproc\u002Fsys\u002Fkernel\u002Frandom\u002Fuuid | cut -c1-8)\nosmo workflow submit assets\u002Fconfigs\u002F\u003Cyaml> --pool \u003Cpool> \\\n  --set name=\u003Cflow>-$STAMP dig_url_root=\u003Croot> usecase=\u003Cusecase> \\\n        image_edit_endpoint=\u003Cendpoint> image_edit_model=nvidia\u002FQwen-Image-Edit-NVPCB-OVSL2SL \\\n        checkpoint_step=\u003Cstep> 'anomaly_types_json=\u003Ctypes>'\n",[2841],{"type":56,"tag":97,"props":2842,"children":2843},{"__ignoreMap":2478},[2844,2886,2952,3041,3076],{"type":56,"tag":2484,"props":2845,"children":2846},{"class":2486,"line":2487},[2847,2852,2857,2862,2867,2871,2876,2881],{"type":56,"tag":2484,"props":2848,"children":2849},{"style":2518},[2850],{"type":61,"value":2851},"STAMP",{"type":56,"tag":2484,"props":2853,"children":2854},{"style":2507},[2855],{"type":61,"value":2856},"=$(",{"type":56,"tag":2484,"props":2858,"children":2859},{"style":2491},[2860],{"type":61,"value":2861},"cat",{"type":56,"tag":2484,"props":2863,"children":2864},{"style":2496},[2865],{"type":61,"value":2866}," \u002Fproc\u002Fsys\u002Fkernel\u002Frandom\u002Fuuid",{"type":56,"tag":2484,"props":2868,"children":2869},{"style":2507},[2870],{"type":61,"value":2541},{"type":56,"tag":2484,"props":2872,"children":2873},{"style":2491},[2874],{"type":61,"value":2875}," cut",{"type":56,"tag":2484,"props":2877,"children":2878},{"style":2496},[2879],{"type":61,"value":2880}," -c1-8",{"type":56,"tag":2484,"props":2882,"children":2883},{"style":2507},[2884],{"type":61,"value":2885},")\n",{"type":56,"tag":2484,"props":2887,"children":2888},{"class":2486,"line":2564},[2889,2893,2897,2902,2907,2911,2916,2921,2925,2930,2934,2939,2943,2947],{"type":56,"tag":2484,"props":2890,"children":2891},{"style":2491},[2892],{"type":61,"value":1174},{"type":56,"tag":2484,"props":2894,"children":2895},{"style":2496},[2896],{"type":61,"value":2499},{"type":56,"tag":2484,"props":2898,"children":2899},{"style":2496},[2900],{"type":61,"value":2901}," submit",{"type":56,"tag":2484,"props":2903,"children":2904},{"style":2496},[2905],{"type":61,"value":2906}," assets\u002Fconfigs\u002F",{"type":56,"tag":2484,"props":2908,"children":2909},{"style":2507},[2910],{"type":61,"value":2674},{"type":56,"tag":2484,"props":2912,"children":2913},{"style":2496},[2914],{"type":61,"value":2915},"yam",{"type":56,"tag":2484,"props":2917,"children":2918},{"style":2518},[2919],{"type":61,"value":2920},"l",{"type":56,"tag":2484,"props":2922,"children":2923},{"style":2507},[2924],{"type":61,"value":2526},{"type":56,"tag":2484,"props":2926,"children":2927},{"style":2496},[2928],{"type":61,"value":2929}," --pool",{"type":56,"tag":2484,"props":2931,"children":2932},{"style":2507},[2933],{"type":61,"value":2510},{"type":56,"tag":2484,"props":2935,"children":2936},{"style":2496},[2937],{"type":61,"value":2938},"poo",{"type":56,"tag":2484,"props":2940,"children":2941},{"style":2518},[2942],{"type":61,"value":2920},{"type":56,"tag":2484,"props":2944,"children":2945},{"style":2507},[2946],{"type":61,"value":2526},{"type":56,"tag":2484,"props":2948,"children":2949},{"style":2518},[2950],{"type":61,"value":2951}," \\\n",{"type":56,"tag":2484,"props":2953,"children":2954},{"class":2486,"line":2631},[2955,2960,2965,2969,2974,2979,2983,2988,2993,2998,3002,3007,3011,3015,3020,3024,3029,3033,3037],{"type":56,"tag":2484,"props":2956,"children":2957},{"style":2496},[2958],{"type":61,"value":2959},"  --set",{"type":56,"tag":2484,"props":2961,"children":2962},{"style":2496},[2963],{"type":61,"value":2964}," name=",{"type":56,"tag":2484,"props":2966,"children":2967},{"style":2507},[2968],{"type":61,"value":2674},{"type":56,"tag":2484,"props":2970,"children":2971},{"style":2496},[2972],{"type":61,"value":2973},"flo",{"type":56,"tag":2484,"props":2975,"children":2976},{"style":2518},[2977],{"type":61,"value":2978},"w",{"type":56,"tag":2484,"props":2980,"children":2981},{"style":2507},[2982],{"type":61,"value":2526},{"type":56,"tag":2484,"props":2984,"children":2985},{"style":2496},[2986],{"type":61,"value":2987},"-",{"type":56,"tag":2484,"props":2989,"children":2990},{"style":2518},[2991],{"type":61,"value":2992},"$STAMP ",{"type":56,"tag":2484,"props":2994,"children":2995},{"style":2496},[2996],{"type":61,"value":2997},"dig_url_root=",{"type":56,"tag":2484,"props":2999,"children":3000},{"style":2507},[3001],{"type":61,"value":2674},{"type":56,"tag":2484,"props":3003,"children":3004},{"style":2496},[3005],{"type":61,"value":3006},"roo",{"type":56,"tag":2484,"props":3008,"children":3009},{"style":2518},[3010],{"type":61,"value":2660},{"type":56,"tag":2484,"props":3012,"children":3013},{"style":2507},[3014],{"type":61,"value":2526},{"type":56,"tag":2484,"props":3016,"children":3017},{"style":2496},[3018],{"type":61,"value":3019}," usecase=",{"type":56,"tag":2484,"props":3021,"children":3022},{"style":2507},[3023],{"type":61,"value":2674},{"type":56,"tag":2484,"props":3025,"children":3026},{"style":2496},[3027],{"type":61,"value":3028},"usecas",{"type":56,"tag":2484,"props":3030,"children":3031},{"style":2518},[3032],{"type":61,"value":2613},{"type":56,"tag":2484,"props":3034,"children":3035},{"style":2507},[3036],{"type":61,"value":2526},{"type":56,"tag":2484,"props":3038,"children":3039},{"style":2518},[3040],{"type":61,"value":2951},{"type":56,"tag":2484,"props":3042,"children":3044},{"class":2486,"line":3043},4,[3045,3050,3054,3059,3063,3067,3072],{"type":56,"tag":2484,"props":3046,"children":3047},{"style":2496},[3048],{"type":61,"value":3049},"        image_edit_endpoint=",{"type":56,"tag":2484,"props":3051,"children":3052},{"style":2507},[3053],{"type":61,"value":2674},{"type":56,"tag":2484,"props":3055,"children":3056},{"style":2496},[3057],{"type":61,"value":3058},"endpoin",{"type":56,"tag":2484,"props":3060,"children":3061},{"style":2518},[3062],{"type":61,"value":2660},{"type":56,"tag":2484,"props":3064,"children":3065},{"style":2507},[3066],{"type":61,"value":2526},{"type":56,"tag":2484,"props":3068,"children":3069},{"style":2496},[3070],{"type":61,"value":3071}," image_edit_model=nvidia\u002FQwen-Image-Edit-NVPCB-OVSL2SL",{"type":56,"tag":2484,"props":3073,"children":3074},{"style":2518},[3075],{"type":61,"value":2951},{"type":56,"tag":2484,"props":3077,"children":3079},{"class":2486,"line":3078},5,[3080,3085,3089,3094,3098,3102,3106,3111],{"type":56,"tag":2484,"props":3081,"children":3082},{"style":2496},[3083],{"type":61,"value":3084},"        checkpoint_step=",{"type":56,"tag":2484,"props":3086,"children":3087},{"style":2507},[3088],{"type":61,"value":2674},{"type":56,"tag":2484,"props":3090,"children":3091},{"style":2496},[3092],{"type":61,"value":3093},"ste",{"type":56,"tag":2484,"props":3095,"children":3096},{"style":2518},[3097],{"type":61,"value":169},{"type":56,"tag":2484,"props":3099,"children":3100},{"style":2507},[3101],{"type":61,"value":2526},{"type":56,"tag":2484,"props":3103,"children":3104},{"style":2507},[3105],{"type":61,"value":2551},{"type":56,"tag":2484,"props":3107,"children":3108},{"style":2496},[3109],{"type":61,"value":3110},"anomaly_types_json=\u003Ctypes>",{"type":56,"tag":2484,"props":3112,"children":3113},{"style":2507},[3114],{"type":61,"value":2561},{"type":56,"tag":169,"props":3116,"children":3117},{},[3118,3123,3125,3130,3132,3137],{"type":56,"tag":175,"props":3119,"children":3120},{},[3121],{"type":61,"value":3122},"Monitoring:",{"type":61,"value":3124}," load ",{"type":56,"tag":97,"props":3126,"children":3128},{"className":3127},[],[3129],{"type":61,"value":2444},{"type":61,"value":3131}," before running the submit; apply its polling cadence + log-pull thresholds after ",{"type":56,"tag":97,"props":3133,"children":3135},{"className":3134},[],[3136],{"type":61,"value":2452},{"type":61,"value":3138}," returns a workflow id.",{"type":56,"tag":169,"props":3140,"children":3141},{},[3142,3147,3148,3154],{"type":56,"tag":175,"props":3143,"children":3144},{},[3145],{"type":61,"value":3146},"Output Location:",{"type":61,"value":323},{"type":56,"tag":97,"props":3149,"children":3151},{"className":3150},[],[3152],{"type":61,"value":3153},"\u003Cdig_url_root>\u002Fruns\u002F\u003Cflow>-$STAMP\u002Fanomaly\u002F",{"type":61,"value":3155}," (per-flow override: see flow walkthrough).",{"type":56,"tag":967,"props":3157,"children":3158},{},[],{"type":56,"tag":64,"props":3160,"children":3162},{"id":3161},"supporting-files",[3163],{"type":61,"value":167},{"type":56,"tag":169,"props":3165,"children":3166},{},[3167,3169,3178,3180,3185,3186,3191,3192,3198,3199,3205,3206,3212],{"type":61,"value":3168},"Full inventory — workflow YAMLs, cookbooks, scripts table, references, evals, component skills — in ",{"type":56,"tag":175,"props":3170,"children":3171},{},[3172],{"type":56,"tag":97,"props":3173,"children":3175},{"className":3174},[],[3176],{"type":61,"value":3177},"references\u002Fcontents.md",{"type":61,"value":3179},". Top-level dirs: ",{"type":56,"tag":97,"props":3181,"children":3183},{"className":3182},[],[3184],{"type":61,"value":187},{"type":61,"value":1815},{"type":56,"tag":97,"props":3187,"children":3189},{"className":3188},[],[3190],{"type":61,"value":195},{"type":61,"value":1815},{"type":56,"tag":97,"props":3193,"children":3195},{"className":3194},[],[3196],{"type":61,"value":3197},"scripts\u002F",{"type":61,"value":1815},{"type":56,"tag":97,"props":3200,"children":3202},{"className":3201},[],[3203],{"type":61,"value":3204},"references\u002F",{"type":61,"value":1815},{"type":56,"tag":97,"props":3207,"children":3209},{"className":3208},[],[3210],{"type":61,"value":3211},"evals\u002F",{"type":61,"value":205},{"type":56,"tag":3214,"props":3215,"children":3216},"style",{},[3217],{"type":61,"value":3218},"html .light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html.light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}",{"items":3220,"total":3377},[3221,3239,3257,3268,3280,3294,3307,3321,3334,3345,3359,3368],{"slug":3222,"name":3222,"fn":3223,"description":3224,"org":3225,"tags":3226,"stars":3236,"repoUrl":3237,"updatedAt":3238},"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},[3227,3230,3233],{"name":3228,"slug":3229,"type":15},"Documentation","documentation",{"name":3231,"slug":3232,"type":15},"MCP","mcp",{"name":3234,"slug":3235,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":3240,"name":3240,"fn":3241,"description":3242,"org":3243,"tags":3244,"stars":3254,"repoUrl":3255,"updatedAt":3256},"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},[3245,3248,3251],{"name":3246,"slug":3247,"type":15},"Containers","containers",{"name":3249,"slug":3250,"type":15},"Deployment","deployment",{"name":3252,"slug":3253,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":3258,"name":3258,"fn":3259,"description":3260,"org":3261,"tags":3262,"stars":3254,"repoUrl":3255,"updatedAt":3267},"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},[3263,3266],{"name":3264,"slug":3265,"type":15},"CI\u002FCD","ci-cd",{"name":3249,"slug":3250,"type":15},"2026-07-14T05:25:59.97109",{"slug":3269,"name":3269,"fn":3270,"description":3271,"org":3272,"tags":3273,"stars":3254,"repoUrl":3255,"updatedAt":3279},"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},[3274,3275,3276],{"name":3264,"slug":3265,"type":15},{"name":3249,"slug":3250,"type":15},{"name":3277,"slug":3278,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":3281,"name":3281,"fn":3282,"description":3283,"org":3284,"tags":3285,"stars":3254,"repoUrl":3255,"updatedAt":3293},"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},[3286,3289,3290],{"name":3287,"slug":3288,"type":15},"Debugging","debugging",{"name":3277,"slug":3278,"type":15},{"name":3291,"slug":3292,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":3295,"name":3295,"fn":3296,"description":3297,"org":3298,"tags":3299,"stars":3254,"repoUrl":3255,"updatedAt":3306},"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},[3300,3303],{"name":3301,"slug":3302,"type":15},"Best Practices","best-practices",{"name":3304,"slug":3305,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":3308,"name":3308,"fn":3309,"description":3310,"org":3311,"tags":3312,"stars":3254,"repoUrl":3255,"updatedAt":3320},"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},[3313,3316,3319],{"name":3314,"slug":3315,"type":15},"Machine Learning","machine-learning",{"name":3317,"slug":3318,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":3322,"name":3322,"fn":3323,"description":3324,"org":3325,"tags":3326,"stars":3254,"repoUrl":3255,"updatedAt":3333},"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},[3327,3330],{"name":3328,"slug":3329,"type":15},"QA","qa",{"name":3331,"slug":3332,"type":15},"Testing","testing","2026-07-14T05:25:53.673039",{"slug":3335,"name":3335,"fn":3336,"description":3337,"org":3338,"tags":3339,"stars":3254,"repoUrl":3255,"updatedAt":3344},"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},[3340,3341],{"name":3249,"slug":3250,"type":15},{"name":3342,"slug":3343,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":3346,"name":3346,"fn":3347,"description":3348,"org":3349,"tags":3350,"stars":3254,"repoUrl":3255,"updatedAt":3358},"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},[3351,3354,3355],{"name":3352,"slug":3353,"type":15},"Code Review","code-review",{"name":3277,"slug":3278,"type":15},{"name":3356,"slug":3357,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":3360,"name":3360,"fn":3361,"description":3362,"org":3363,"tags":3364,"stars":3254,"repoUrl":3255,"updatedAt":3367},"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},[3365,3366],{"name":3328,"slug":3329,"type":15},{"name":3331,"slug":3332,"type":15},"2026-07-14T05:25:54.928983",{"slug":3369,"name":3369,"fn":3370,"description":3371,"org":3372,"tags":3373,"stars":3254,"repoUrl":3255,"updatedAt":3376},"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},[3374,3375],{"name":13,"slug":14,"type":15},{"name":3264,"slug":3265,"type":15},"2026-07-30T05:29:03.275638",496,{"items":3379,"total":3475},[3380,3397,3407,3421,3431,3446,3461],{"slug":3381,"name":3381,"fn":3382,"description":3383,"org":3384,"tags":3385,"stars":23,"repoUrl":24,"updatedAt":3396},"accelerated-computing-cudf","accelerate data processing with cuDF","Official NVIDIA-authored guidance for NVIDIA cuDF GPU DataFrames, pandas acceleration, dask-cuDF, ETL, joins, groupby, CSV\u002FParquet I\u002FO, nullable semantics, and multi-GPU DataFrame workloads.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[3386,3389,3392,3393],{"name":3387,"slug":3388,"type":15},"Data Analysis","data-analysis",{"name":3390,"slug":3391,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":3394,"slug":3395,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":3398,"name":3398,"fn":3399,"description":3400,"org":3401,"tags":3402,"stars":23,"repoUrl":24,"updatedAt":3406},"aiq-deploy","deploy and manage NVIDIA AI-Q infrastructure","Use when asked to install, deploy, run, validate, troubleshoot, or stop NVIDIA AI-Q Blueprint infrastructure.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[3403,3404,3405],{"name":3249,"slug":3250,"type":15},{"name":3342,"slug":3343,"type":15},{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":3408,"name":3408,"fn":3409,"description":3410,"org":3411,"tags":3412,"stars":23,"repoUrl":24,"updatedAt":3420},"aiq-research","conduct deep research with AI-Q","Use when asked to run deep research or AI-Q research through a reachable NVIDIA AI-Q Blueprint backend.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[3413,3416,3417],{"name":3414,"slug":3415,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":3418,"slug":3419,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":3422,"name":3422,"fn":3423,"description":3424,"org":3425,"tags":3426,"stars":23,"repoUrl":24,"updatedAt":3430},"amc-run-sample-calibration","run AMC sample dataset calibration","Run end-to-end calibration on the shipped sample dataset (sdg_08_2_sample_data_010926.zip) against a running AMC microservice. Use when user says 'test sample dataset', 'run sample calibration', 'verify AMC install', or 'launch and test'.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[3427,3428,3429],{"name":3387,"slug":3388,"type":15},{"name":9,"slug":8,"type":15},{"name":3331,"slug":3332,"type":15},"2026-07-17T05:29:03.913266",{"slug":3432,"name":3432,"fn":3433,"description":3434,"org":3435,"tags":3436,"stars":23,"repoUrl":24,"updatedAt":3445},"amc-run-video-calibration","calibrate video datasets with AutoMagicCalib","Calibrate a new dataset from pre-recorded video files via the AutoMagicCalib REST API. Use when user has local MP4s and says 'calibrate my videos', 'run AMC on these videos', or similar. For RTSP\u002Flive streams, use amc-run-rtsp-calibration instead.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[3437,3438,3441,3442],{"name":13,"slug":14,"type":15},{"name":3439,"slug":3440,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":3443,"slug":3444,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":3447,"name":3447,"fn":3448,"description":3449,"org":3450,"tags":3451,"stars":23,"repoUrl":24,"updatedAt":3460},"amc-setup-calibration-stack","deploy AutoMagicCalib microservice with Docker","Launch AutoMagicCalib microservice and web UI from NGC release images via Docker Compose. Use when user says 'deploy auto calibration', 'launch auto calibration', 'launch AMC', 'start MS+UI', or 'set up auto-magic-calib'. Requires NGC API key.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[3452,3453,3456,3457],{"name":3249,"slug":3250,"type":15},{"name":3454,"slug":3455,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":3458,"slug":3459,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":3462,"name":3462,"fn":3463,"description":3464,"org":3465,"tags":3466,"stars":23,"repoUrl":24,"updatedAt":3474},"cudaq-guide","develop quantum applications with CUDA-Q","CUDA-Q onboarding guide for installation, test programs, GPU simulation, QPU hardware, and quantum applications.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[3467,3468,3471],{"name":9,"slug":8,"type":15},{"name":3469,"slug":3470,"type":15},"Quantum Computing","quantum-computing",{"name":3472,"slug":3473,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305]