[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-doca-dms":3,"mdc--st8g7y-key":34,"related-repo-nvidia-doca-dms":1154,"related-org-nvidia-doca-dms":1258},{"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},"doca-dms","manage BlueField DPU device operations","Operate NVIDIA DOCA Management Service (`dmsd` + `dmspe`) on a BlueField, Arm\u002Fx86 host, or Kubernetes pod: choose deployment and authentication, configure `-allowed_users` and `dmsgroup`, use gNMI Get\u002FSet\u002FSubscribe, run supported gNOI workflows, and debug frontend\u002Fbackend failures. Trigger even without \"DMS\" for \"manage a remote BlueField over gRPC\", \"gNOI reboot from orchestrator\", or fleet-management requests. SAFETY: reboot, OS install, factory-reset, and managed-file deletion are destructive and require target-bound explicit confirmation; never invoke them speculatively. Route installation and library\u002FAPI build questions elsewhere, and route turnkey aggregation to the externally-productized DOCA Telemetry Service.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,19,20],{"name":13,"slug":14,"type":15},"Security","security","tag",{"name":17,"slug":18,"type":15},"Operations","operations",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"Engineering","engineering",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-30T05:28:36.40497","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\u002Fdoca-dms","---\nlicense: Apache-2.0\nname: doca-dms\ndescription: >\n  Operate NVIDIA DOCA Management Service (`dmsd` + `dmspe`) on a\n  BlueField, Arm\u002Fx86 host, or Kubernetes pod: choose deployment and\n  authentication, configure `-allowed_users` and `dmsgroup`, use gNMI\n  Get\u002FSet\u002FSubscribe, run supported gNOI workflows, and debug\n  frontend\u002Fbackend failures. Trigger even without \"DMS\" for \"manage a\n  remote BlueField over gRPC\", \"gNOI reboot from orchestrator\", or\n  fleet-management requests.\n  SAFETY: reboot, OS install, factory-reset, and managed-file deletion\n  are destructive and require target-bound explicit confirmation;\n  never invoke them speculatively. Route installation and library\u002FAPI\n  build questions elsewhere, and route turnkey aggregation to the\n  externally-productized DOCA Telemetry Service.\nmetadata:\n  kind: service\ncompatibility: >\n  DOCA service shipped with the DOCA install at \u002Fopt\u002Fmellanox\u002Fdoca\n  on the management endpoint (x86 host (non-DPU), BlueField Arm, or\n  Kubernetes pod) on Linux (Ubuntu 22.04\u002F24.04 or RHEL\u002FSLES); `dmsd`\n  + `dmspe` run there, and DMS is also pulled as an NGC container\n  image. Verify the public DMS guide version matches the installed\n  DOCA release before quoting flags or YANG paths.\n---\n\n# DOCA Management Service (DMS)\n\n> **⚠️ Destructive operations.** The gNOI `reboot`, `OS install`,\n> `factory-reset`, and managed-file deletion operations are\n> **irreversible** and **service-impacting**\n> — they can take a production BlueField or ConnectX offline or wipe its\n> configuration. Before issuing any of them the agent MUST: (1) verify the\n> target device identity, and (2) obtain explicit confirmation bound to\n> that target and action. In an interactive session this is an explicit\n> user reply naming\u002Faccepting both; in unattended execution it must be an\n> approved-system authorization artifact bound to both. Otherwise stop\n> with `confirmation_required`. Never invoke them speculatively or as\n> a side effect of another task. See the public DMS guide's safety\n> guidance for these operations.\n\n**Where to start:** This skill is for *operating* DMS, not for\n*linking against* a library. If the user wants to *deploy* or *run*\nthe daemon, open [`TASKS.md`](TASKS.md) and start at\n[`## configure`](TASKS.md#configure). If the question is *what shape\nof service is DMS and what protocols does it speak*, start at\n[`CAPABILITIES.md`](CAPABILITIES.md). If DOCA is not installed on the\nmanagement endpoint yet, route to\n[`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md) first.\n\n## Example questions this skill answers well\n\nThe CLASSES of DMS questions this skill is built to answer, each with\none worked example. The class is the load-bearing piece; the worked\nexample is one instance.\n\n- **\"Where should DMS run for my target topology?\"** — worked example:\n  *\"I have a host without a DPU; can I still manage a remote\n  ConnectX?\"*. Answered by the deployment-shape decision in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  + [`TASKS.md ## configure`](TASKS.md#configure).\n- **\"Which authentication mode should I pick for my security\n  posture?\"** — worked example: *\"a multi-tenant production env vs a\n  single-tenant lab\"*. Answered by the auth-mode trade-off table in\n  [`CAPABILITIES.md ## Safety policy`](CAPABILITIES.md#safety-policy)\n  + the auth-wiring step in\n  [`TASKS.md ## configure`](TASKS.md#configure).\n- **\"How do I issue a gNMI `Get` \u002F `Set` against a modeled path?\"** —\n  worked example: *\"set `\u002Finterfaces\u002Finterface\u002Fconfig\u002Fmtu` on a remote\n  interface\"*. Answered by the gNMI\u002FgNOI surface in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  + the request-shape workflow in\n  [`TASKS.md ## run`](TASKS.md#run).\n- **\"How do I run a gNOI operation (reboot, OS install, file\n  transfer)?\"** — worked example: *\"trigger a clean reboot of the\n  target via gNOI\"*. Answered by the gNOI catalog in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  + the run workflow in [`TASKS.md ## run`](TASKS.md#run).\n- **\"My DMS request returned an error — was it the frontend or the\n  backend?\"** — worked example: *\"`mlxconfig` failed under DMS but\n  works on the shell\"*. Answered by the frontend-vs-backend split in\n  [`CAPABILITIES.md ## Error taxonomy`](CAPABILITIES.md#error-taxonomy)\n  + the layered ladder in\n  [`TASKS.md ## debug`](TASKS.md#debug).\n- **\"Where do I read DMS logs and how do I rotate them?\"** — worked\n  example: *\"persistent log directory + journald + log-rotation\n  policy\"*. Answered by the logging surface in\n  [`CAPABILITIES.md ## Observability`](CAPABILITIES.md#observability)\n  + the log-rotation step in\n  [`TASKS.md ## configure`](TASKS.md#configure).\n\n## Audience\n\nThis skill serves **external operators and platform teams who deploy and\noperate DMS** to manage NVIDIA® BlueField® networking platforms or\nNVIDIA® ConnectX® SmartNICs from a centralized control plane. Concretely:\npeople running `dmsd`, integrating gNMI\u002FgNOI clients against it, choosing\nan authentication mode, or wiring DMS into a Kubernetes deployment.\n\nIt is **not** for NVIDIA developers contributing to DMS itself, and it\nis **not** a programming guide for *building applications on top of*\nDOCA libraries (that is `doca-programming-guide` plus the matching\nlibrary skill under `libs\u002F`). DMS is a **service**, not a library: the\nuser invokes it as a daemon and talks to it over gRPC; they do not link\nagainst `libdms.so` to write their own program.\n\n**Status note.** Per the public DMS guide, DMS is currently in **beta**,\nwith General Availability scoped to SPC-X use cases. The skill reflects\nthe public guide's posture: prescribe the documented launch \u002F auth \u002F\ndeployment paths, follow the documented security best practices, and\ndefer roadmap and GA-scope questions to the live public guide rather\nthan guessing.\n\n## When to load this skill\n\nLoad this skill when the user is doing **hands-on DMS operation work**\nagainst a BlueField or ConnectX target where DOCA is already installed\non the management endpoint (host, DPU, or pod). Concretely:\n\n- Deciding *where* DMS should run (host non-DPU \u002F BlueField Arm \u002F\n  Kubernetes pod) for a given target topology.\n- Bringing up the `dmsd` daemon — choosing SystemD vs manual launch,\n  selecting an authentication mode, wiring `-allowed_users` (the gRPC\n  client authorization boundary) and, if needed, `dmsgroup` (the\n  `dmspe` backend-helper Unix group).\n- Issuing `gNMI` `Get` \u002F `Set` requests against modeled paths\n  (e.g. `\u002Finterfaces\u002Finterface\u002Fconfig\u002Fmtu`).\n- Issuing `gNOI` operations: OS install, reboot, file transfer,\n  factory-reset, `mlxconfig`, containerz.\n- Choosing an authentication mode (localhost \u002F PAM \u002F credentials \u002F mTLS)\n  and understanding the security trade-offs the public guide calls out.\n- Reading or rotating DMS logs, configuring config persistency, or\n  recovering from a crashed daemon.\n- Debugging a DMS request that returned an error — separating\n  \"frontend rejected before reaching backend\" from \"backend executed and\n  the underlying tool (e.g. `mlxconfig`) failed\".\n\nDo **not** load this skill for general DOCA orientation, install of\nDOCA itself, or library-API questions. For those, route via\n[`doca-public-knowledge-map`](..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md),\n[`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md), or the matching\n`libs\u002F\u003Clibrary>` skill.\n\n## What this skill provides\n\nThis is a **thin loader**. Substantive material lives in two companion\nfiles:\n\n- `CAPABILITIES.md` — DMS architecture (frontend `dmsd` \u002F privileged\n  backend `dmspe`), management protocols (gNMI, gNOI), the YANG-based\n  unified configuration dictionary, deployment shapes, authentication\n  modes with their security trade-offs, the configuration-persistency\n  model, the logging surface, and DMS's documented security posture.\n- `TASKS.md` — step-by-step workflows for the in-scope DMS verbs:\n  `configure`, `build`, `modify`, `run`, `test`, `debug`, plus a\n  `Deferred task verbs` block routing out-of-scope questions.\n\nThe skill assumes a host where DOCA is already installed and the\noperator has root \u002F `sudo` access where the public guide says it is\nrequired. It does not cover installing DOCA — that path goes through\n[`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md).\n\n## What this skill deliberately does not ship\n\nThis skill is **agent guidance**, not a templates or sample-config\nbundle. To keep the boundary clean, it deliberately does not contain —\nand pull requests should not add:\n\n- **Pre-baked DMS configuration files** (YANG instance documents,\n  full-stack example configs, ready-to-run `dmsd` flag bundles)\n  intended to be copy-pasted into production. Configs are deployment-\n  specific and the safe answer for an external operator is to derive\n  them from the public guide against their own target. The agent's\n  job is to prescribe the *procedure* and quote the documented flags\n  and paths, not to ship a config the user might run unmodified.\n- **Pre-written gNMI \u002F gNOI client programs in any language.** The\n  client surface is standard gNMI \u002F gNOI (publicly documented); the\n  skill describes which paths and operations DMS supports, not how to\n  build a gRPC client in language X.\n- **TLS material, credentials, or PAM stanzas.** These are\n  user-environment artifacts; the skill points at the documented\n  configuration knobs and the documented security best practices and\n  stops there.\n- **A `samples\u002F`, `templates\u002F`, or `reference\u002F` subtree** of any\n  kind. A mock or incomplete artifact in this skill's tree, even one\n  labeled \"reference\", is misleading: operators will read it as\n  production-ready.\n\n## Loading order\n\n1. Read this `SKILL.md` first to confirm the user's question is in scope.\n2. **For the DMS architecture, deployment shapes, auth modes,\n   protocol\u002Fpath inventory, persistency, logging, and security\n   posture, see [CAPABILITIES.md](CAPABILITIES.md).**\n3. **For step-by-step workflows — configure, build, modify, run, test,\n   debug — see [TASKS.md](TASKS.md).**\n4. **Apply the destructive-operation gate in every phase.** Before any\n   gNOI OS install, reboot, factory reset, or managed-file deletion —\n   including a test or debug action — verify the exact target and obtain explicit\n   confirmation for that specific operation using the interactive or\n   approved-system mechanism defined in the warning above. No earlier\n   confirmation or workflow phase carries authorization forward.\n\n## Related skills\n\n- [`doca-public-knowledge-map`](..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md)\n  — the routing table to the public DMS guide and the rest of the\n  public DOCA documentation set.\n- [`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md) — env preparation and\n  install verification on the host where `dmsd` will run, including\n  the *I have no install yet* path via the public NGC DOCA container.\n  This skill assumes its preconditions are satisfied at the management\n  endpoint.\n- [`doca-programming-guide`](..\u002F..\u002Fdoca-programming-guide\u002FSKILL.md) —\n  general DOCA patterns. DMS is service-shaped not library-shaped, so\n  the build \u002F modify \u002F first-app pattern there does not apply directly,\n  but the cross-library `DOCA_ERROR_*` taxonomy and the\n  layered-debug order remain useful when DMS reports errors that\n  originated in a DOCA library it called.\n",{"data":35,"body":39},{"license":26,"name":4,"description":6,"metadata":36,"compatibility":38},{"kind":37},"service","DOCA service shipped with the DOCA install at \u002Fopt\u002Fmellanox\u002Fdoca on the management endpoint (x86 host (non-DPU), BlueField Arm, or Kubernetes pod) on Linux (Ubuntu 22.04\u002F24.04 or RHEL\u002FSLES); `dmsd` + `dmspe` run there, and DMS is also pulled as an NGC container image. Verify the public DMS guide version matches the installed DOCA release before quoting flags or YANG paths.\n",{"type":40,"children":41},"root",[42,51,114,207,214,219,518,524,544,599,616,622,634,759,799,805,817,904,926,932,944,1024,1030,1083,1089],{"type":43,"tag":44,"props":45,"children":47},"element","h1",{"id":46},"doca-management-service-dms",[48],{"type":49,"value":50},"text","DOCA Management Service (DMS)",{"type":43,"tag":52,"props":53,"children":54},"blockquote",{},[55],{"type":43,"tag":56,"props":57,"children":58},"p",{},[59,65,67,74,76,82,84,90,92,97,99,104,106,112],{"type":43,"tag":60,"props":61,"children":62},"strong",{},[63],{"type":49,"value":64},"⚠️ Destructive operations.",{"type":49,"value":66}," The gNOI ",{"type":43,"tag":68,"props":69,"children":71},"code",{"className":70},[],[72],{"type":49,"value":73},"reboot",{"type":49,"value":75},", ",{"type":43,"tag":68,"props":77,"children":79},{"className":78},[],[80],{"type":49,"value":81},"OS install",{"type":49,"value":83},",\n",{"type":43,"tag":68,"props":85,"children":87},{"className":86},[],[88],{"type":49,"value":89},"factory-reset",{"type":49,"value":91},", and managed-file deletion operations are\n",{"type":43,"tag":60,"props":93,"children":94},{},[95],{"type":49,"value":96},"irreversible",{"type":49,"value":98}," and ",{"type":43,"tag":60,"props":100,"children":101},{},[102],{"type":49,"value":103},"service-impacting",{"type":49,"value":105},"\n— they can take a production BlueField or ConnectX offline or wipe its\nconfiguration. Before issuing any of them the agent MUST: (1) verify the\ntarget device identity, and (2) obtain explicit confirmation bound to\nthat target and action. In an interactive session this is an explicit\nuser reply naming\u002Faccepting both; in unattended execution it must be an\napproved-system authorization artifact bound to both. Otherwise stop\nwith ",{"type":43,"tag":68,"props":107,"children":109},{"className":108},[],[110],{"type":49,"value":111},"confirmation_required",{"type":49,"value":113},". Never invoke them speculatively or as\na side effect of another task. See the public DMS guide's safety\nguidance for these operations.",{"type":43,"tag":56,"props":115,"children":116},{},[117,122,124,130,132,137,139,144,146,151,153,163,165,175,177,182,184,193,195,205],{"type":43,"tag":60,"props":118,"children":119},{},[120],{"type":49,"value":121},"Where to start:",{"type":49,"value":123}," This skill is for ",{"type":43,"tag":125,"props":126,"children":127},"em",{},[128],{"type":49,"value":129},"operating",{"type":49,"value":131}," DMS, not for\n",{"type":43,"tag":125,"props":133,"children":134},{},[135],{"type":49,"value":136},"linking against",{"type":49,"value":138}," a library. If the user wants to ",{"type":43,"tag":125,"props":140,"children":141},{},[142],{"type":49,"value":143},"deploy",{"type":49,"value":145}," or ",{"type":43,"tag":125,"props":147,"children":148},{},[149],{"type":49,"value":150},"run",{"type":49,"value":152},"\nthe daemon, open ",{"type":43,"tag":154,"props":155,"children":157},"a",{"href":156},"TASKS.md",[158],{"type":43,"tag":68,"props":159,"children":161},{"className":160},[],[162],{"type":49,"value":156},{"type":49,"value":164}," and start at\n",{"type":43,"tag":154,"props":166,"children":168},{"href":167},"TASKS.md#configure",[169],{"type":43,"tag":68,"props":170,"children":172},{"className":171},[],[173],{"type":49,"value":174},"## configure",{"type":49,"value":176},". If the question is ",{"type":43,"tag":125,"props":178,"children":179},{},[180],{"type":49,"value":181},"what shape\nof service is DMS and what protocols does it speak",{"type":49,"value":183},", start at\n",{"type":43,"tag":154,"props":185,"children":187},{"href":186},"CAPABILITIES.md",[188],{"type":43,"tag":68,"props":189,"children":191},{"className":190},[],[192],{"type":49,"value":186},{"type":49,"value":194},". If DOCA is not installed on the\nmanagement endpoint yet, route to\n",{"type":43,"tag":154,"props":196,"children":198},{"href":197},"..\u002F..\u002Fdoca-setup\u002FSKILL.md",[199],{"type":43,"tag":68,"props":200,"children":202},{"className":201},[],[203],{"type":49,"value":204},"doca-setup",{"type":49,"value":206}," first.",{"type":43,"tag":208,"props":209,"children":211},"h2",{"id":210},"example-questions-this-skill-answers-well",[212],{"type":49,"value":213},"Example questions this skill answers well",{"type":43,"tag":56,"props":215,"children":216},{},[217],{"type":49,"value":218},"The CLASSES of DMS questions this skill is built to answer, each with\none worked example. The class is the load-bearing piece; the worked\nexample is one instance.",{"type":43,"tag":220,"props":221,"children":222},"ul",{},[223,268,312,380,421,474],{"type":43,"tag":224,"props":225,"children":226},"li",{},[227,232,234,239,241,251],{"type":43,"tag":60,"props":228,"children":229},{},[230],{"type":49,"value":231},"\"Where should DMS run for my target topology?\"",{"type":49,"value":233}," — worked example:\n",{"type":43,"tag":125,"props":235,"children":236},{},[237],{"type":49,"value":238},"\"I have a host without a DPU; can I still manage a remote\nConnectX?\"",{"type":49,"value":240},". Answered by the deployment-shape decision in\n",{"type":43,"tag":154,"props":242,"children":244},{"href":243},"CAPABILITIES.md#capabilities-and-modes",[245],{"type":43,"tag":68,"props":246,"children":248},{"className":247},[],[249],{"type":49,"value":250},"CAPABILITIES.md ## Capabilities and modes",{"type":43,"tag":220,"props":252,"children":253},{},[254],{"type":43,"tag":224,"props":255,"children":256},{},[257,266],{"type":43,"tag":154,"props":258,"children":259},{"href":167},[260],{"type":43,"tag":68,"props":261,"children":263},{"className":262},[],[264],{"type":49,"value":265},"TASKS.md ## configure",{"type":49,"value":267},".",{"type":43,"tag":224,"props":269,"children":270},{},[271,276,278,283,285,295],{"type":43,"tag":60,"props":272,"children":273},{},[274],{"type":49,"value":275},"\"Which authentication mode should I pick for my security\nposture?\"",{"type":49,"value":277}," — worked example: ",{"type":43,"tag":125,"props":279,"children":280},{},[281],{"type":49,"value":282},"\"a multi-tenant production env vs a\nsingle-tenant lab\"",{"type":49,"value":284},". Answered by the auth-mode trade-off table in\n",{"type":43,"tag":154,"props":286,"children":288},{"href":287},"CAPABILITIES.md#safety-policy",[289],{"type":43,"tag":68,"props":290,"children":292},{"className":291},[],[293],{"type":49,"value":294},"CAPABILITIES.md ## Safety policy",{"type":43,"tag":220,"props":296,"children":297},{},[298],{"type":43,"tag":224,"props":299,"children":300},{},[301,303,311],{"type":49,"value":302},"the auth-wiring step in\n",{"type":43,"tag":154,"props":304,"children":305},{"href":167},[306],{"type":43,"tag":68,"props":307,"children":309},{"className":308},[],[310],{"type":49,"value":265},{"type":49,"value":267},{"type":43,"tag":224,"props":313,"children":314},{},[315,336,338,351,353,361],{"type":43,"tag":60,"props":316,"children":317},{},[318,320,326,328,334],{"type":49,"value":319},"\"How do I issue a gNMI ",{"type":43,"tag":68,"props":321,"children":323},{"className":322},[],[324],{"type":49,"value":325},"Get",{"type":49,"value":327}," \u002F ",{"type":43,"tag":68,"props":329,"children":331},{"className":330},[],[332],{"type":49,"value":333},"Set",{"type":49,"value":335}," against a modeled path?\"",{"type":49,"value":337}," —\nworked example: ",{"type":43,"tag":125,"props":339,"children":340},{},[341,343,349],{"type":49,"value":342},"\"set ",{"type":43,"tag":68,"props":344,"children":346},{"className":345},[],[347],{"type":49,"value":348},"\u002Finterfaces\u002Finterface\u002Fconfig\u002Fmtu",{"type":49,"value":350}," on a remote\ninterface\"",{"type":49,"value":352},". Answered by the gNMI\u002FgNOI surface in\n",{"type":43,"tag":154,"props":354,"children":355},{"href":243},[356],{"type":43,"tag":68,"props":357,"children":359},{"className":358},[],[360],{"type":49,"value":250},{"type":43,"tag":220,"props":362,"children":363},{},[364],{"type":43,"tag":224,"props":365,"children":366},{},[367,369,379],{"type":49,"value":368},"the request-shape workflow in\n",{"type":43,"tag":154,"props":370,"children":372},{"href":371},"TASKS.md#run",[373],{"type":43,"tag":68,"props":374,"children":376},{"className":375},[],[377],{"type":49,"value":378},"TASKS.md ## run",{"type":49,"value":267},{"type":43,"tag":224,"props":381,"children":382},{},[383,388,389,394,396,404],{"type":43,"tag":60,"props":384,"children":385},{},[386],{"type":49,"value":387},"\"How do I run a gNOI operation (reboot, OS install, file\ntransfer)?\"",{"type":49,"value":277},{"type":43,"tag":125,"props":390,"children":391},{},[392],{"type":49,"value":393},"\"trigger a clean reboot of the\ntarget via gNOI\"",{"type":49,"value":395},". Answered by the gNOI catalog in\n",{"type":43,"tag":154,"props":397,"children":398},{"href":243},[399],{"type":43,"tag":68,"props":400,"children":402},{"className":401},[],[403],{"type":49,"value":250},{"type":43,"tag":220,"props":405,"children":406},{},[407],{"type":43,"tag":224,"props":408,"children":409},{},[410,412,420],{"type":49,"value":411},"the run workflow in ",{"type":43,"tag":154,"props":413,"children":414},{"href":371},[415],{"type":43,"tag":68,"props":416,"children":418},{"className":417},[],[419],{"type":49,"value":378},{"type":49,"value":267},{"type":43,"tag":224,"props":422,"children":423},{},[424,429,430,443,445,455],{"type":43,"tag":60,"props":425,"children":426},{},[427],{"type":49,"value":428},"\"My DMS request returned an error — was it the frontend or the\nbackend?\"",{"type":49,"value":277},{"type":43,"tag":125,"props":431,"children":432},{},[433,435,441],{"type":49,"value":434},"\"",{"type":43,"tag":68,"props":436,"children":438},{"className":437},[],[439],{"type":49,"value":440},"mlxconfig",{"type":49,"value":442}," failed under DMS but\nworks on the shell\"",{"type":49,"value":444},". Answered by the frontend-vs-backend split in\n",{"type":43,"tag":154,"props":446,"children":448},{"href":447},"CAPABILITIES.md#error-taxonomy",[449],{"type":43,"tag":68,"props":450,"children":452},{"className":451},[],[453],{"type":49,"value":454},"CAPABILITIES.md ## Error taxonomy",{"type":43,"tag":220,"props":456,"children":457},{},[458],{"type":43,"tag":224,"props":459,"children":460},{},[461,463,473],{"type":49,"value":462},"the layered ladder in\n",{"type":43,"tag":154,"props":464,"children":466},{"href":465},"TASKS.md#debug",[467],{"type":43,"tag":68,"props":468,"children":470},{"className":469},[],[471],{"type":49,"value":472},"TASKS.md ## debug",{"type":49,"value":267},{"type":43,"tag":224,"props":475,"children":476},{},[477,482,484,489,491,501],{"type":43,"tag":60,"props":478,"children":479},{},[480],{"type":49,"value":481},"\"Where do I read DMS logs and how do I rotate them?\"",{"type":49,"value":483}," — worked\nexample: ",{"type":43,"tag":125,"props":485,"children":486},{},[487],{"type":49,"value":488},"\"persistent log directory + journald + log-rotation\npolicy\"",{"type":49,"value":490},". Answered by the logging surface in\n",{"type":43,"tag":154,"props":492,"children":494},{"href":493},"CAPABILITIES.md#observability",[495],{"type":43,"tag":68,"props":496,"children":498},{"className":497},[],[499],{"type":49,"value":500},"CAPABILITIES.md ## Observability",{"type":43,"tag":220,"props":502,"children":503},{},[504],{"type":43,"tag":224,"props":505,"children":506},{},[507,509,517],{"type":49,"value":508},"the log-rotation step in\n",{"type":43,"tag":154,"props":510,"children":511},{"href":167},[512],{"type":43,"tag":68,"props":513,"children":515},{"className":514},[],[516],{"type":49,"value":265},{"type":49,"value":267},{"type":43,"tag":208,"props":519,"children":521},{"id":520},"audience",[522],{"type":49,"value":523},"Audience",{"type":43,"tag":56,"props":525,"children":526},{},[527,529,534,536,542],{"type":49,"value":528},"This skill serves ",{"type":43,"tag":60,"props":530,"children":531},{},[532],{"type":49,"value":533},"external operators and platform teams who deploy and\noperate DMS",{"type":49,"value":535}," to manage NVIDIA® BlueField® networking platforms or\nNVIDIA® ConnectX® SmartNICs from a centralized control plane. Concretely:\npeople running ",{"type":43,"tag":68,"props":537,"children":539},{"className":538},[],[540],{"type":49,"value":541},"dmsd",{"type":49,"value":543},", integrating gNMI\u002FgNOI clients against it, choosing\nan authentication mode, or wiring DMS into a Kubernetes deployment.",{"type":43,"tag":56,"props":545,"children":546},{},[547,549,554,556,560,562,567,569,575,577,583,585,589,591,597],{"type":49,"value":548},"It is ",{"type":43,"tag":60,"props":550,"children":551},{},[552],{"type":49,"value":553},"not",{"type":49,"value":555}," for NVIDIA developers contributing to DMS itself, and it\nis ",{"type":43,"tag":60,"props":557,"children":558},{},[559],{"type":49,"value":553},{"type":49,"value":561}," a programming guide for ",{"type":43,"tag":125,"props":563,"children":564},{},[565],{"type":49,"value":566},"building applications on top of",{"type":49,"value":568},"\nDOCA libraries (that is ",{"type":43,"tag":68,"props":570,"children":572},{"className":571},[],[573],{"type":49,"value":574},"doca-programming-guide",{"type":49,"value":576}," plus the matching\nlibrary skill under ",{"type":43,"tag":68,"props":578,"children":580},{"className":579},[],[581],{"type":49,"value":582},"libs\u002F",{"type":49,"value":584},"). DMS is a ",{"type":43,"tag":60,"props":586,"children":587},{},[588],{"type":49,"value":37},{"type":49,"value":590},", not a library: the\nuser invokes it as a daemon and talks to it over gRPC; they do not link\nagainst ",{"type":43,"tag":68,"props":592,"children":594},{"className":593},[],[595],{"type":49,"value":596},"libdms.so",{"type":49,"value":598}," to write their own program.",{"type":43,"tag":56,"props":600,"children":601},{},[602,607,609,614],{"type":43,"tag":60,"props":603,"children":604},{},[605],{"type":49,"value":606},"Status note.",{"type":49,"value":608}," Per the public DMS guide, DMS is currently in ",{"type":43,"tag":60,"props":610,"children":611},{},[612],{"type":49,"value":613},"beta",{"type":49,"value":615},",\nwith General Availability scoped to SPC-X use cases. The skill reflects\nthe public guide's posture: prescribe the documented launch \u002F auth \u002F\ndeployment paths, follow the documented security best practices, and\ndefer roadmap and GA-scope questions to the live public guide rather\nthan guessing.",{"type":43,"tag":208,"props":617,"children":619},{"id":618},"when-to-load-this-skill",[620],{"type":49,"value":621},"When to load this skill",{"type":43,"tag":56,"props":623,"children":624},{},[625,627,632],{"type":49,"value":626},"Load this skill when the user is doing ",{"type":43,"tag":60,"props":628,"children":629},{},[630],{"type":49,"value":631},"hands-on DMS operation work",{"type":49,"value":633},"\nagainst a BlueField or ConnectX target where DOCA is already installed\non the management endpoint (host, DPU, or pod). Concretely:",{"type":43,"tag":220,"props":635,"children":636},{},[637,649,685,718,737,742,747],{"type":43,"tag":224,"props":638,"children":639},{},[640,642,647],{"type":49,"value":641},"Deciding ",{"type":43,"tag":125,"props":643,"children":644},{},[645],{"type":49,"value":646},"where",{"type":49,"value":648}," DMS should run (host non-DPU \u002F BlueField Arm \u002F\nKubernetes pod) for a given target topology.",{"type":43,"tag":224,"props":650,"children":651},{},[652,654,659,661,667,669,675,677,683],{"type":49,"value":653},"Bringing up the ",{"type":43,"tag":68,"props":655,"children":657},{"className":656},[],[658],{"type":49,"value":541},{"type":49,"value":660}," daemon — choosing SystemD vs manual launch,\nselecting an authentication mode, wiring ",{"type":43,"tag":68,"props":662,"children":664},{"className":663},[],[665],{"type":49,"value":666},"-allowed_users",{"type":49,"value":668}," (the gRPC\nclient authorization boundary) and, if needed, ",{"type":43,"tag":68,"props":670,"children":672},{"className":671},[],[673],{"type":49,"value":674},"dmsgroup",{"type":49,"value":676}," (the\n",{"type":43,"tag":68,"props":678,"children":680},{"className":679},[],[681],{"type":49,"value":682},"dmspe",{"type":49,"value":684}," backend-helper Unix group).",{"type":43,"tag":224,"props":686,"children":687},{},[688,690,696,698,703,704,709,711,716],{"type":49,"value":689},"Issuing ",{"type":43,"tag":68,"props":691,"children":693},{"className":692},[],[694],{"type":49,"value":695},"gNMI",{"type":49,"value":697}," ",{"type":43,"tag":68,"props":699,"children":701},{"className":700},[],[702],{"type":49,"value":325},{"type":49,"value":327},{"type":43,"tag":68,"props":705,"children":707},{"className":706},[],[708],{"type":49,"value":333},{"type":49,"value":710}," requests against modeled paths\n(e.g. ",{"type":43,"tag":68,"props":712,"children":714},{"className":713},[],[715],{"type":49,"value":348},{"type":49,"value":717},").",{"type":43,"tag":224,"props":719,"children":720},{},[721,722,728,730,735],{"type":49,"value":689},{"type":43,"tag":68,"props":723,"children":725},{"className":724},[],[726],{"type":49,"value":727},"gNOI",{"type":49,"value":729}," operations: OS install, reboot, file transfer,\nfactory-reset, ",{"type":43,"tag":68,"props":731,"children":733},{"className":732},[],[734],{"type":49,"value":440},{"type":49,"value":736},", containerz.",{"type":43,"tag":224,"props":738,"children":739},{},[740],{"type":49,"value":741},"Choosing an authentication mode (localhost \u002F PAM \u002F credentials \u002F mTLS)\nand understanding the security trade-offs the public guide calls out.",{"type":43,"tag":224,"props":743,"children":744},{},[745],{"type":49,"value":746},"Reading or rotating DMS logs, configuring config persistency, or\nrecovering from a crashed daemon.",{"type":43,"tag":224,"props":748,"children":749},{},[750,752,757],{"type":49,"value":751},"Debugging a DMS request that returned an error — separating\n\"frontend rejected before reaching backend\" from \"backend executed and\nthe underlying tool (e.g. ",{"type":43,"tag":68,"props":753,"children":755},{"className":754},[],[756],{"type":49,"value":440},{"type":49,"value":758},") failed\".",{"type":43,"tag":56,"props":760,"children":761},{},[762,764,768,770,780,781,789,791,797],{"type":49,"value":763},"Do ",{"type":43,"tag":60,"props":765,"children":766},{},[767],{"type":49,"value":553},{"type":49,"value":769}," load this skill for general DOCA orientation, install of\nDOCA itself, or library-API questions. For those, route via\n",{"type":43,"tag":154,"props":771,"children":773},{"href":772},"..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md",[774],{"type":43,"tag":68,"props":775,"children":777},{"className":776},[],[778],{"type":49,"value":779},"doca-public-knowledge-map",{"type":49,"value":83},{"type":43,"tag":154,"props":782,"children":783},{"href":197},[784],{"type":43,"tag":68,"props":785,"children":787},{"className":786},[],[788],{"type":49,"value":204},{"type":49,"value":790},", or the matching\n",{"type":43,"tag":68,"props":792,"children":794},{"className":793},[],[795],{"type":49,"value":796},"libs\u002F\u003Clibrary>",{"type":49,"value":798}," skill.",{"type":43,"tag":208,"props":800,"children":802},{"id":801},"what-this-skill-provides",[803],{"type":49,"value":804},"What this skill provides",{"type":43,"tag":56,"props":806,"children":807},{},[808,810,815],{"type":49,"value":809},"This is a ",{"type":43,"tag":60,"props":811,"children":812},{},[813],{"type":49,"value":814},"thin loader",{"type":49,"value":816},". Substantive material lives in two companion\nfiles:",{"type":43,"tag":220,"props":818,"children":819},{},[820,844],{"type":43,"tag":224,"props":821,"children":822},{},[823,828,830,835,837,842],{"type":43,"tag":68,"props":824,"children":826},{"className":825},[],[827],{"type":49,"value":186},{"type":49,"value":829}," — DMS architecture (frontend ",{"type":43,"tag":68,"props":831,"children":833},{"className":832},[],[834],{"type":49,"value":541},{"type":49,"value":836}," \u002F privileged\nbackend ",{"type":43,"tag":68,"props":838,"children":840},{"className":839},[],[841],{"type":49,"value":682},{"type":49,"value":843},"), management protocols (gNMI, gNOI), the YANG-based\nunified configuration dictionary, deployment shapes, authentication\nmodes with their security trade-offs, the configuration-persistency\nmodel, the logging surface, and DMS's documented security posture.",{"type":43,"tag":224,"props":845,"children":846},{},[847,852,854,860,861,867,868,874,875,880,881,887,888,894,896,902],{"type":43,"tag":68,"props":848,"children":850},{"className":849},[],[851],{"type":49,"value":156},{"type":49,"value":853}," — step-by-step workflows for the in-scope DMS verbs:\n",{"type":43,"tag":68,"props":855,"children":857},{"className":856},[],[858],{"type":49,"value":859},"configure",{"type":49,"value":75},{"type":43,"tag":68,"props":862,"children":864},{"className":863},[],[865],{"type":49,"value":866},"build",{"type":49,"value":75},{"type":43,"tag":68,"props":869,"children":871},{"className":870},[],[872],{"type":49,"value":873},"modify",{"type":49,"value":75},{"type":43,"tag":68,"props":876,"children":878},{"className":877},[],[879],{"type":49,"value":150},{"type":49,"value":75},{"type":43,"tag":68,"props":882,"children":884},{"className":883},[],[885],{"type":49,"value":886},"test",{"type":49,"value":75},{"type":43,"tag":68,"props":889,"children":891},{"className":890},[],[892],{"type":49,"value":893},"debug",{"type":49,"value":895},", plus a\n",{"type":43,"tag":68,"props":897,"children":899},{"className":898},[],[900],{"type":49,"value":901},"Deferred task verbs",{"type":49,"value":903}," block routing out-of-scope questions.",{"type":43,"tag":56,"props":905,"children":906},{},[907,909,915,917,925],{"type":49,"value":908},"The skill assumes a host where DOCA is already installed and the\noperator has root \u002F ",{"type":43,"tag":68,"props":910,"children":912},{"className":911},[],[913],{"type":49,"value":914},"sudo",{"type":49,"value":916}," access where the public guide says it is\nrequired. It does not cover installing DOCA — that path goes through\n",{"type":43,"tag":154,"props":918,"children":919},{"href":197},[920],{"type":43,"tag":68,"props":921,"children":923},{"className":922},[],[924],{"type":49,"value":204},{"type":49,"value":267},{"type":43,"tag":208,"props":927,"children":929},{"id":928},"what-this-skill-deliberately-does-not-ship",[930],{"type":49,"value":931},"What this skill deliberately does not ship",{"type":43,"tag":56,"props":933,"children":934},{},[935,937,942],{"type":49,"value":936},"This skill is ",{"type":43,"tag":60,"props":938,"children":939},{},[940],{"type":49,"value":941},"agent guidance",{"type":49,"value":943},", not a templates or sample-config\nbundle. To keep the boundary clean, it deliberately does not contain —\nand pull requests should not add:",{"type":43,"tag":220,"props":945,"children":946},{},[947,971,981,991],{"type":43,"tag":224,"props":948,"children":949},{},[950,955,957,962,964,969],{"type":43,"tag":60,"props":951,"children":952},{},[953],{"type":49,"value":954},"Pre-baked DMS configuration files",{"type":49,"value":956}," (YANG instance documents,\nfull-stack example configs, ready-to-run ",{"type":43,"tag":68,"props":958,"children":960},{"className":959},[],[961],{"type":49,"value":541},{"type":49,"value":963}," flag bundles)\nintended to be copy-pasted into production. Configs are deployment-\nspecific and the safe answer for an external operator is to derive\nthem from the public guide against their own target. The agent's\njob is to prescribe the ",{"type":43,"tag":125,"props":965,"children":966},{},[967],{"type":49,"value":968},"procedure",{"type":49,"value":970}," and quote the documented flags\nand paths, not to ship a config the user might run unmodified.",{"type":43,"tag":224,"props":972,"children":973},{},[974,979],{"type":43,"tag":60,"props":975,"children":976},{},[977],{"type":49,"value":978},"Pre-written gNMI \u002F gNOI client programs in any language.",{"type":49,"value":980}," The\nclient surface is standard gNMI \u002F gNOI (publicly documented); the\nskill describes which paths and operations DMS supports, not how to\nbuild a gRPC client in language X.",{"type":43,"tag":224,"props":982,"children":983},{},[984,989],{"type":43,"tag":60,"props":985,"children":986},{},[987],{"type":49,"value":988},"TLS material, credentials, or PAM stanzas.",{"type":49,"value":990}," These are\nuser-environment artifacts; the skill points at the documented\nconfiguration knobs and the documented security best practices and\nstops there.",{"type":43,"tag":224,"props":992,"children":993},{},[994,1022],{"type":43,"tag":60,"props":995,"children":996},{},[997,999,1005,1006,1012,1014,1020],{"type":49,"value":998},"A ",{"type":43,"tag":68,"props":1000,"children":1002},{"className":1001},[],[1003],{"type":49,"value":1004},"samples\u002F",{"type":49,"value":75},{"type":43,"tag":68,"props":1007,"children":1009},{"className":1008},[],[1010],{"type":49,"value":1011},"templates\u002F",{"type":49,"value":1013},", or ",{"type":43,"tag":68,"props":1015,"children":1017},{"className":1016},[],[1018],{"type":49,"value":1019},"reference\u002F",{"type":49,"value":1021}," subtree",{"type":49,"value":1023}," of any\nkind. A mock or incomplete artifact in this skill's tree, even one\nlabeled \"reference\", is misleading: operators will read it as\nproduction-ready.",{"type":43,"tag":208,"props":1025,"children":1027},{"id":1026},"loading-order",[1028],{"type":49,"value":1029},"Loading order",{"type":43,"tag":1031,"props":1032,"children":1033},"ol",{},[1034,1047,1060,1073],{"type":43,"tag":224,"props":1035,"children":1036},{},[1037,1039,1045],{"type":49,"value":1038},"Read this ",{"type":43,"tag":68,"props":1040,"children":1042},{"className":1041},[],[1043],{"type":49,"value":1044},"SKILL.md",{"type":49,"value":1046}," first to confirm the user's question is in scope.",{"type":43,"tag":224,"props":1048,"children":1049},{},[1050],{"type":43,"tag":60,"props":1051,"children":1052},{},[1053,1055,1059],{"type":49,"value":1054},"For the DMS architecture, deployment shapes, auth modes,\nprotocol\u002Fpath inventory, persistency, logging, and security\nposture, see ",{"type":43,"tag":154,"props":1056,"children":1057},{"href":186},[1058],{"type":49,"value":186},{"type":49,"value":267},{"type":43,"tag":224,"props":1061,"children":1062},{},[1063],{"type":43,"tag":60,"props":1064,"children":1065},{},[1066,1068,1072],{"type":49,"value":1067},"For step-by-step workflows — configure, build, modify, run, test,\ndebug — see ",{"type":43,"tag":154,"props":1069,"children":1070},{"href":156},[1071],{"type":49,"value":156},{"type":49,"value":267},{"type":43,"tag":224,"props":1074,"children":1075},{},[1076,1081],{"type":43,"tag":60,"props":1077,"children":1078},{},[1079],{"type":49,"value":1080},"Apply the destructive-operation gate in every phase.",{"type":49,"value":1082}," Before any\ngNOI OS install, reboot, factory reset, or managed-file deletion —\nincluding a test or debug action — verify the exact target and obtain explicit\nconfirmation for that specific operation using the interactive or\napproved-system mechanism defined in the warning above. No earlier\nconfirmation or workflow phase carries authorization forward.",{"type":43,"tag":208,"props":1084,"children":1086},{"id":1085},"related-skills",[1087],{"type":49,"value":1088},"Related skills",{"type":43,"tag":220,"props":1090,"children":1091},{},[1092,1105,1132],{"type":43,"tag":224,"props":1093,"children":1094},{},[1095,1103],{"type":43,"tag":154,"props":1096,"children":1097},{"href":772},[1098],{"type":43,"tag":68,"props":1099,"children":1101},{"className":1100},[],[1102],{"type":49,"value":779},{"type":49,"value":1104},"\n— the routing table to the public DMS guide and the rest of the\npublic DOCA documentation set.",{"type":43,"tag":224,"props":1106,"children":1107},{},[1108,1116,1118,1123,1125,1130],{"type":43,"tag":154,"props":1109,"children":1110},{"href":197},[1111],{"type":43,"tag":68,"props":1112,"children":1114},{"className":1113},[],[1115],{"type":49,"value":204},{"type":49,"value":1117}," — env preparation and\ninstall verification on the host where ",{"type":43,"tag":68,"props":1119,"children":1121},{"className":1120},[],[1122],{"type":49,"value":541},{"type":49,"value":1124}," will run, including\nthe ",{"type":43,"tag":125,"props":1126,"children":1127},{},[1128],{"type":49,"value":1129},"I have no install yet",{"type":49,"value":1131}," path via the public NGC DOCA container.\nThis skill assumes its preconditions are satisfied at the management\nendpoint.",{"type":43,"tag":224,"props":1133,"children":1134},{},[1135,1144,1146,1152],{"type":43,"tag":154,"props":1136,"children":1138},{"href":1137},"..\u002F..\u002Fdoca-programming-guide\u002FSKILL.md",[1139],{"type":43,"tag":68,"props":1140,"children":1142},{"className":1141},[],[1143],{"type":49,"value":574},{"type":49,"value":1145}," —\ngeneral DOCA patterns. DMS is service-shaped not library-shaped, so\nthe build \u002F modify \u002F first-app pattern there does not apply directly,\nbut the cross-library ",{"type":43,"tag":68,"props":1147,"children":1149},{"className":1148},[],[1150],{"type":49,"value":1151},"DOCA_ERROR_*",{"type":49,"value":1153}," taxonomy and the\nlayered-debug order remain useful when DMS reports errors that\noriginated in a DOCA library it called.",{"items":1155,"total":1257},[1156,1173,1187,1201,1213,1230,1243],{"slug":1157,"name":1157,"fn":1158,"description":1159,"org":1160,"tags":1161,"stars":23,"repoUrl":24,"updatedAt":1172},"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},[1162,1165,1168,1169],{"name":1163,"slug":1164,"type":15},"Data Analysis","data-analysis",{"name":1166,"slug":1167,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":1170,"slug":1171,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":1174,"name":1174,"fn":1175,"description":1176,"org":1177,"tags":1178,"stars":23,"repoUrl":24,"updatedAt":1186},"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},[1179,1182,1185],{"name":1180,"slug":1181,"type":15},"Deployment","deployment",{"name":1183,"slug":1184,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":1188,"name":1188,"fn":1189,"description":1190,"org":1191,"tags":1192,"stars":23,"repoUrl":24,"updatedAt":1200},"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},[1193,1196,1197],{"name":1194,"slug":1195,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":1198,"slug":1199,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":1202,"name":1202,"fn":1203,"description":1204,"org":1205,"tags":1206,"stars":23,"repoUrl":24,"updatedAt":1212},"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},[1207,1208,1209],{"name":1163,"slug":1164,"type":15},{"name":9,"slug":8,"type":15},{"name":1210,"slug":1211,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":1214,"name":1214,"fn":1215,"description":1216,"org":1217,"tags":1218,"stars":23,"repoUrl":24,"updatedAt":1229},"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},[1219,1222,1225,1226],{"name":1220,"slug":1221,"type":15},"Automation","automation",{"name":1223,"slug":1224,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":1227,"slug":1228,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":1231,"name":1231,"fn":1232,"description":1233,"org":1234,"tags":1235,"stars":23,"repoUrl":24,"updatedAt":1242},"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},[1236,1237,1240,1241],{"name":1180,"slug":1181,"type":15},{"name":1238,"slug":1239,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":17,"slug":18,"type":15},"2026-07-17T05:28:56.913999",{"slug":1244,"name":1244,"fn":1245,"description":1246,"org":1247,"tags":1248,"stars":23,"repoUrl":24,"updatedAt":1256},"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},[1249,1250,1253],{"name":9,"slug":8,"type":15},{"name":1251,"slug":1252,"type":15},"Quantum Computing","quantum-computing",{"name":1254,"slug":1255,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":1259,"total":1410},[1260,1278,1294,1305,1317,1331,1344,1358,1369,1378,1392,1401],{"slug":1261,"name":1261,"fn":1262,"description":1263,"org":1264,"tags":1265,"stars":1275,"repoUrl":1276,"updatedAt":1277},"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},[1266,1269,1272],{"name":1267,"slug":1268,"type":15},"Documentation","documentation",{"name":1270,"slug":1271,"type":15},"MCP","mcp",{"name":1273,"slug":1274,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":1279,"name":1279,"fn":1280,"description":1281,"org":1282,"tags":1283,"stars":1291,"repoUrl":1292,"updatedAt":1293},"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},[1284,1287,1288],{"name":1285,"slug":1286,"type":15},"Containers","containers",{"name":1180,"slug":1181,"type":15},{"name":1289,"slug":1290,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":1295,"name":1295,"fn":1296,"description":1297,"org":1298,"tags":1299,"stars":1291,"repoUrl":1292,"updatedAt":1304},"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},[1300,1303],{"name":1301,"slug":1302,"type":15},"CI\u002FCD","ci-cd",{"name":1180,"slug":1181,"type":15},"2026-07-14T05:25:59.97109",{"slug":1306,"name":1306,"fn":1307,"description":1308,"org":1309,"tags":1310,"stars":1291,"repoUrl":1292,"updatedAt":1316},"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},[1311,1312,1313],{"name":1301,"slug":1302,"type":15},{"name":1180,"slug":1181,"type":15},{"name":1314,"slug":1315,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":1318,"name":1318,"fn":1319,"description":1320,"org":1321,"tags":1322,"stars":1291,"repoUrl":1292,"updatedAt":1330},"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},[1323,1326,1327],{"name":1324,"slug":1325,"type":15},"Debugging","debugging",{"name":1314,"slug":1315,"type":15},{"name":1328,"slug":1329,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":1332,"name":1332,"fn":1333,"description":1334,"org":1335,"tags":1336,"stars":1291,"repoUrl":1292,"updatedAt":1343},"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},[1337,1340],{"name":1338,"slug":1339,"type":15},"Best Practices","best-practices",{"name":1341,"slug":1342,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":1345,"name":1345,"fn":1346,"description":1347,"org":1348,"tags":1349,"stars":1291,"repoUrl":1292,"updatedAt":1357},"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},[1350,1353,1356],{"name":1351,"slug":1352,"type":15},"Machine Learning","machine-learning",{"name":1354,"slug":1355,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":1359,"name":1359,"fn":1360,"description":1361,"org":1362,"tags":1363,"stars":1291,"repoUrl":1292,"updatedAt":1368},"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},[1364,1367],{"name":1365,"slug":1366,"type":15},"QA","qa",{"name":1210,"slug":1211,"type":15},"2026-07-14T05:25:53.673039",{"slug":1370,"name":1370,"fn":1371,"description":1372,"org":1373,"tags":1374,"stars":1291,"repoUrl":1292,"updatedAt":1377},"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},[1375,1376],{"name":1180,"slug":1181,"type":15},{"name":1183,"slug":1184,"type":15},"2026-07-14T05:25:49.362534",{"slug":1379,"name":1379,"fn":1380,"description":1381,"org":1382,"tags":1383,"stars":1291,"repoUrl":1292,"updatedAt":1391},"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},[1384,1387,1388],{"name":1385,"slug":1386,"type":15},"Code Review","code-review",{"name":1314,"slug":1315,"type":15},{"name":1389,"slug":1390,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":1393,"name":1393,"fn":1394,"description":1395,"org":1396,"tags":1397,"stars":1291,"repoUrl":1292,"updatedAt":1400},"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},[1398,1399],{"name":1365,"slug":1366,"type":15},{"name":1210,"slug":1211,"type":15},"2026-07-14T05:25:54.928983",{"slug":1402,"name":1402,"fn":1403,"description":1404,"org":1405,"tags":1406,"stars":1291,"repoUrl":1292,"updatedAt":1409},"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},[1407,1408],{"name":1220,"slug":1221,"type":15},{"name":1301,"slug":1302,"type":15},"2026-07-30T05:29:03.275638",496]