[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-doca-telemetry":3,"mdc-lt3g2e-key":34,"related-org-nvidia-doca-telemetry":1086,"related-repo-nvidia-doca-telemetry":1247},{"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-telemetry","read DOCA hardware telemetry events","Use this skill to read DOCA hardware-counter events from a `doca_dev` through the per-domain Telemetry reader libraries: `doca_telemetry_pcc`, `_dpa`, `_diag`, `_adp_retx`, `_phy`, and `_pci`. It covers capability checks, context creation, startup, and per-domain reads or samples. Trigger for implicit requests such as \"read PCC counters from my BlueField app\", \"sample DPA counter exports\", or \"expose PHY, PCI, or DIAG counters from this doca_dev\". This is the counter-reader surface, not a NetFlow, IPFIX, or local-socket collector. Route publishing and export to `doca-telemetry-exporter`; route deployed DOCA Telemetry Service (DTS), collectors, and plain stdout logging elsewhere.\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},"Observability","observability","tag",{"name":17,"slug":18,"type":15},"Monitoring","monitoring",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"Engineering","engineering",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-30T05:28:47.42631","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-telemetry","---\nlicense: Apache-2.0\nname: doca-telemetry\ndescription: >\n  Use this skill to read DOCA hardware-counter events from a\n  `doca_dev` through the per-domain Telemetry reader libraries:\n  `doca_telemetry_pcc`, `_dpa`, `_diag`, `_adp_retx`, `_phy`,\n  and `_pci`. It covers capability checks, context creation,\n  startup, and per-domain reads or samples. Trigger for implicit\n  requests such as \"read PCC counters from my BlueField app\",\n  \"sample DPA counter exports\", or \"expose PHY, PCI, or DIAG\n  counters from this doca_dev\". This is the counter-reader\n  surface, not a NetFlow, IPFIX, or local-socket collector.\n  Route publishing and export to `doca-telemetry-exporter`;\n  route deployed DOCA Telemetry Service (DTS), collectors, and\n  plain stdout logging elsewhere.\nmetadata:\n  kind: library\ncompatibility: >\n  Requires DOCA SDK installed at \u002Fopt\u002Fmellanox\u002Fdoca on Linux (Ubuntu\n  22.04\u002F24.04 or RHEL\u002FSLES) with a BlueField DPU or ConnectX NIC\n  attached. Reads the user's local install via `pkg-config\n  doca-telemetry` and inspects\n  \u002Fopt\u002Fmellanox\u002Fdoca\u002F{lib,include,samples,applications}.\n---\n\n# DOCA Telemetry\n\n**Where to start:** This skill assumes DOCA is already installed\nand the user is doing **hands-on hardware-counter-reader work** —\nopening a per-domain `doca_telemetry_\u003Cdomain>` context against a\n`doca_dev` and reading the latest hardware-counter snapshot for\nthat domain. The library is the **counter-READER** half of DOCA\ntelemetry; it is NOT a NetFlow \u002F IPFIX collector and NOT a\ngeneric schema-event consumer (the bundle previously framed it\nthat way and that framing was wrong — there is no NetFlow \u002F\nIPFIX \u002F local-socket transport surface in the public header).\nOpen [`TASKS.md`](TASKS.md) if the user wants to *do* something\n(configure \u002F build \u002F modify \u002F run \u002F test \u002F debug); open\n[`CAPABILITIES.md`](CAPABILITIES.md) when the question is *which\nhardware-counter domains can this device read* on this install\n(PCC, DPA, DIAG, ADP_RETX, PHY, PCI). If the user has not\ninstalled DOCA yet, route to\n[`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md) first. If the user is\nconfused about whether they want this library (HW-counter\nreader on a `doca_dev`) or\n[`doca-telemetry-exporter`](..\u002Fdoca-telemetry-exporter\u002FSKILL.md)\n(the publisher \u002F export side, which is a separate library and\npublishes structured telemetry \u002F labeled metrics \u002F OTLP logs),\nread the reader-vs-exporter role split in\n[`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\nBEFORE configuring anything; mixing the two is the load-bearing\nfirst-app failure for this skill. If the user is asking about\n**DOCA Telemetry Service (DTS) as deployed**, route to\n[`doca-public-knowledge-map`](..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md)\nnon-goals — DTS is out of scope for this bundle.\n\n## Audience\n\nThis skill serves **external developers building applications\nthat READ DOCA hardware counters from a `doca_dev` through one\nor more of the six per-domain DOCA Telemetry reader libraries**\n(`doca_telemetry_pcc` \u002F `_dpa` \u002F `_diag` \u002F `_adp_retx` \u002F `_phy` \u002F\n`_pci`) — i.e., users whose application code calls\n`doca_telemetry_\u003Cdomain>_*` (directly in C\u002FC++, or through FFI \u002F\nbindings from another language) to open a per-domain context on\na `doca_dev`, configure the per-domain sample window, and read\nthe hardware-counter snapshot for that domain. It is *not* for\nNVIDIA developers contributing to DOCA Telemetry itself, and it\nis *not* for users writing the **publishing \u002F export** side —\nthat is\n[`doca-telemetry-exporter`](..\u002Fdoca-telemetry-exporter\u002FSKILL.md),\na separate library and a separate skill.\n\n**Language scope.** DOCA Telemetry's per-domain reader libraries\nship as a C surface with `pkg-config` module name\n`doca-telemetry`. The shipped samples are written in C. C and\nC++ readers are the canonical case; the worked examples in\n`TASKS.md` assume that path. Other-language readers (Rust, Go,\nPython, …) consume the same `*.so` through FFI or\nlanguage-specific bindings; the skill's contribution in that\ncase is to keep the reader-vs-exporter distinction, the per-\ndomain cap-query-first discipline, the per-domain DOCA Core\nlifecycle on the `doca_dev`, the sample-window discipline, and\nthe error-taxonomy guidance language-neutral, and to route the\nagent to the public per-domain C ABI as the authoritative\nsurface that any wrapper will eventually call.\n\n## When to load this skill\n\nLoad this skill when the user is doing hands-on DOCA Telemetry\n**hardware-counter-reader** work, in any language. Concretely:\n\n- Picking the right per-domain header for the counters the user\n  wants (`doca_telemetry_pcc.h` for Programmable Congestion\n  Control counters, `_dpa.h` for DPA counters, `_diag.h` for\n  generic device diagnostic counters, `_adp_retx.h` for ADP\n  retransmit counters, `_phy.h` for physical-layer counters,\n  `_pci.h` for PCI \u002F PCIe counters) and confirming the device\n  supports it via the per-domain `_cap_is_supported(devinfo)`\n  query — **except `_pci`, which has no single `_cap_is_supported`\n  and instead exposes per-feature caps like\n  `doca_telemetry_pci_cap_management_info_is_supported` \u002F\n  `_cap_perf_counters_1_is_supported`.**\n- Opening a per-domain `doca_telemetry_\u003Cdomain>` context on a\n  `doca_dev`, walking the per-domain lifecycle\n  (`doca_telemetry_\u003Cdomain>_create(dev)` → per-domain setters →\n  `doca_telemetry_\u003Cdomain>_start`), configuring the per-domain\n  sample window, and reading the hardware-counter snapshot for\n  that domain. Note this is a per-domain `_create`\u002F`_start`\n  surface, not the generic `doca_ctx_*` progress-engine\n  lifecycle.\n- Reading the device + library capability surface before assuming\n  a counter family is available: use\n  `doca_telemetry_\u003Cdomain>_cap_is_supported` only for `pcc`,\n  `dpa`, `diag`, `adp_retx`, and `phy`; use the matching\n  per-feature `doca_telemetry_pci_cap_*_is_supported` query for\n  PCI.\n- Handling per-domain `DOCA_ERROR_*` returns from a counter\n  read (lifecycle vs. device-doesn't-support-this-domain vs.\n  per-domain `AGAIN`-means-snapshot-not-ready vs. permission \u002F\n  driver) and the per-read status reported back to the\n  application.\n- Choosing between DOCA Telemetry (hardware-counter READER) and\n  an adjacent option:\n  [`doca-telemetry-exporter`](..\u002Fdoca-telemetry-exporter\u002FSKILL.md)\n  when the user actually wants to PUBLISH \u002F EXPORT the counter\n  values (OTLP \u002F Prometheus \u002F labeled metrics);\n  `doca-log` when plain structured stdout logging is enough; a\n  generic Prometheus \u002F OpenTelemetry client library when the\n  counter source is a non-DOCA program; the externally-\n  productized DOCA Telemetry Service (DTS, out of scope) when\n  the user wants a turnkey aggregator.\n- Designing or extending non-C bindings (Rust, Go, Python, …)\n  that wrap the per-domain reader C ABI — for the reader-vs-\n  exporter distinction, the per-domain cap-query-first rule,\n  the per-domain `doca_dev` lifecycle, the sample-window\n  discipline, and the error rules the wrapper must honor.\n\nDo **not** load this skill for general DOCA orientation, install\nof DOCA itself, the **publishing \u002F export** side\n([`doca-telemetry-exporter`](..\u002Fdoca-telemetry-exporter\u002FSKILL.md)\nhas its own skill), the externally-productized DOCA Telemetry\nService (DTS — out of scope), or non-reader library questions.\nFor those, use\n[`doca-public-knowledge-map`](..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md).\n\n## What this skill provides\n\nThis is a **thin loader**. The body keeps only the orientation\nneeded to pick the right next file. The substantive\nhardware-counter-reader material lives in two companion files:\n\n- `CAPABILITIES.md` — what the per-domain readers can express on\n  this install: the reader-vs-exporter role-split rule, the\n  six shipped sub-libraries (`doca_telemetry_pcc` \u002F `_dpa` \u002F\n  `_diag` \u002F `_adp_retx` \u002F `_phy` \u002F `_pci`) and which counter\n  family each one exposes, the per-domain DOCA Core lifecycle\n  on a `doca_dev`, the domain-level capability query for PCC \u002F\n  DPA \u002F DIAG \u002F ADP_RETX \u002F PHY and the per-feature capability\n  queries for PCI, the\n  reader error taxonomy (mapped onto the cross-library\n  `DOCA_ERROR_*` set, with the `NOT_SUPPORTED`-means-domain-\n  not-exposed-on-this-device rule and the `AGAIN`-means-\n  snapshot-not-ready rule called out explicitly), the\n  observability surface (per-read status + per-domain cap-\n  query snapshot at configure time), the safety policy that\n  gates per-domain reads behind the cap-query result, and the\n  path-selection rule against\n  [`doca-telemetry-exporter`](..\u002Fdoca-telemetry-exporter\u002FSKILL.md),\n  `doca-log`, and standalone Prometheus \u002F OpenTelemetry \u002F DTS.\n- `TASKS.md` — step-by-step workflows for the six in-scope\n  reader verbs: `configure`, `build`, `modify`, `run`, `test`,\n  `debug`. Plus a `Deferred task verbs` block that points\n  out-of-scope questions at the right next skill.\n\nThe skill assumes a host where DOCA is already installed at\nthe standard location and a target BlueField DPU or ConnectX NIC\nis available. The [`TASKS.md ## run`](TASKS.md#run) workflow\nopens the corresponding `doca_dev` and requires the per-domain\ncap-query to return `DOCA_SUCCESS`. It does not cover installing\nDOCA — that path goes through\n[`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md) — and it does not\ncover writing the publishing \u002F export side, which is\n[`doca-telemetry-exporter`](..\u002Fdoca-telemetry-exporter\u002FSKILL.md).\n\n## Loading order\n\n1. Read this `SKILL.md` first to confirm the user's question\n   is in scope (specifically, that the user wants to READ a\n   per-domain hardware counter via the per-domain reader API\n   on a `doca_dev` — not PUBLISH counters, which is\n   [`doca-telemetry-exporter`](..\u002Fdoca-telemetry-exporter\u002FSKILL.md);\n   not deploy DTS, which is out of scope; and not stand up a\n   NetFlow \u002F IPFIX collector, which this library does not\n   expose a surface for).\n2. **For the reader-vs-exporter rule, the six per-domain\n   sub-libraries, the per-domain DOCA Core lifecycle on a\n   `doca_dev`, the per-domain capability query, the error\n   taxonomy (including the `NOT_SUPPORTED`-means-domain-\n   not-exposed-on-this-device rule and the `AGAIN`-means-\n   snapshot-not-ready rule), observability, the safety\n   policy, and the path-selection rule, see\n   [CAPABILITIES.md](CAPABILITIES.md).**\n3. **For step-by-step workflows — configure, build, modify,\n   run, test, debug — see [TASKS.md](TASKS.md).**\n\nBoth companion files cross-link to each other,\n[`doca-version`](..\u002F..\u002Fdoca-version\u002FSKILL.md) for the canonical\nversion-handling rules, and\n[`doca-public-knowledge-map`](..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md)\nwhenever the right answer is \"look it up in the public docs or\nthe installed package layout\" rather than \"reader-specific\nguidance\".\n\n## Example questions this skill answers well\n\nSee [`references\u002Fdetails.md`](references\u002Fdetails.md#example-questions-this-skill-answers-well).\n## What this skill deliberately does not ship\n\nSee [`references\u002Fdetails.md`](references\u002Fdetails.md#what-this-skill-deliberately-does-not-ship).\n## Related skills\n\nSee [`references\u002Fdetails.md`](references\u002Fdetails.md#related-skills).\n",{"data":35,"body":39},{"license":26,"name":4,"description":6,"metadata":36,"compatibility":38},{"kind":37},"library","Requires DOCA SDK installed at \u002Fopt\u002Fmellanox\u002Fdoca on Linux (Ubuntu 22.04\u002F24.04 or RHEL\u002FSLES) with a BlueField DPU or ConnectX NIC attached. Reads the user's local install via `pkg-config doca-telemetry` and inspects \u002Fopt\u002Fmellanox\u002Fdoca\u002F{lib,include,samples,applications}.\n",{"type":40,"children":41},"root",[42,50,193,200,309,356,362,374,650,686,692,704,860,911,917,997,1024,1030,1046,1052,1066,1072],{"type":43,"tag":44,"props":45,"children":46},"element","h1",{"id":4},[47],{"type":48,"value":49},"text","DOCA Telemetry",{"type":43,"tag":51,"props":52,"children":53},"p",{},[54,60,62,67,69,76,78,84,86,91,93,103,105,111,113,122,124,129,131,141,143,148,150,160,162,172,174,179,181,191],{"type":43,"tag":55,"props":56,"children":57},"strong",{},[58],{"type":48,"value":59},"Where to start:",{"type":48,"value":61}," This skill assumes DOCA is already installed\nand the user is doing ",{"type":43,"tag":55,"props":63,"children":64},{},[65],{"type":48,"value":66},"hands-on hardware-counter-reader work",{"type":48,"value":68}," —\nopening a per-domain ",{"type":43,"tag":70,"props":71,"children":73},"code",{"className":72},[],[74],{"type":48,"value":75},"doca_telemetry_\u003Cdomain>",{"type":48,"value":77}," context against a\n",{"type":43,"tag":70,"props":79,"children":81},{"className":80},[],[82],{"type":48,"value":83},"doca_dev",{"type":48,"value":85}," and reading the latest hardware-counter snapshot for\nthat domain. The library is the ",{"type":43,"tag":55,"props":87,"children":88},{},[89],{"type":48,"value":90},"counter-READER",{"type":48,"value":92}," half of DOCA\ntelemetry; it is NOT a NetFlow \u002F IPFIX collector and NOT a\ngeneric schema-event consumer (the bundle previously framed it\nthat way and that framing was wrong — there is no NetFlow \u002F\nIPFIX \u002F local-socket transport surface in the public header).\nOpen ",{"type":43,"tag":94,"props":95,"children":97},"a",{"href":96},"TASKS.md",[98],{"type":43,"tag":70,"props":99,"children":101},{"className":100},[],[102],{"type":48,"value":96},{"type":48,"value":104}," if the user wants to ",{"type":43,"tag":106,"props":107,"children":108},"em",{},[109],{"type":48,"value":110},"do",{"type":48,"value":112}," something\n(configure \u002F build \u002F modify \u002F run \u002F test \u002F debug); open\n",{"type":43,"tag":94,"props":114,"children":116},{"href":115},"CAPABILITIES.md",[117],{"type":43,"tag":70,"props":118,"children":120},{"className":119},[],[121],{"type":48,"value":115},{"type":48,"value":123}," when the question is ",{"type":43,"tag":106,"props":125,"children":126},{},[127],{"type":48,"value":128},"which\nhardware-counter domains can this device read",{"type":48,"value":130}," on this install\n(PCC, DPA, DIAG, ADP_RETX, PHY, PCI). If the user has not\ninstalled DOCA yet, route to\n",{"type":43,"tag":94,"props":132,"children":134},{"href":133},"..\u002F..\u002Fdoca-setup\u002FSKILL.md",[135],{"type":43,"tag":70,"props":136,"children":138},{"className":137},[],[139],{"type":48,"value":140},"doca-setup",{"type":48,"value":142}," first. If the user is\nconfused about whether they want this library (HW-counter\nreader on a ",{"type":43,"tag":70,"props":144,"children":146},{"className":145},[],[147],{"type":48,"value":83},{"type":48,"value":149},") or\n",{"type":43,"tag":94,"props":151,"children":153},{"href":152},"..\u002Fdoca-telemetry-exporter\u002FSKILL.md",[154],{"type":43,"tag":70,"props":155,"children":157},{"className":156},[],[158],{"type":48,"value":159},"doca-telemetry-exporter",{"type":48,"value":161},"\n(the publisher \u002F export side, which is a separate library and\npublishes structured telemetry \u002F labeled metrics \u002F OTLP logs),\nread the reader-vs-exporter role split in\n",{"type":43,"tag":94,"props":163,"children":165},{"href":164},"CAPABILITIES.md#capabilities-and-modes",[166],{"type":43,"tag":70,"props":167,"children":169},{"className":168},[],[170],{"type":48,"value":171},"CAPABILITIES.md ## Capabilities and modes",{"type":48,"value":173},"\nBEFORE configuring anything; mixing the two is the load-bearing\nfirst-app failure for this skill. If the user is asking about\n",{"type":43,"tag":55,"props":175,"children":176},{},[177],{"type":48,"value":178},"DOCA Telemetry Service (DTS) as deployed",{"type":48,"value":180},", route to\n",{"type":43,"tag":94,"props":182,"children":184},{"href":183},"..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md",[185],{"type":43,"tag":70,"props":186,"children":188},{"className":187},[],[189],{"type":48,"value":190},"doca-public-knowledge-map",{"type":48,"value":192},"\nnon-goals — DTS is out of scope for this bundle.",{"type":43,"tag":194,"props":195,"children":197},"h2",{"id":196},"audience",[198],{"type":48,"value":199},"Audience",{"type":43,"tag":51,"props":201,"children":202},{},[203,205,217,219,225,227,233,234,240,241,247,248,254,256,262,264,270,272,277,279,284,286,290,292,297,299,307],{"type":48,"value":204},"This skill serves ",{"type":43,"tag":55,"props":206,"children":207},{},[208,210,215],{"type":48,"value":209},"external developers building applications\nthat READ DOCA hardware counters from a ",{"type":43,"tag":70,"props":211,"children":213},{"className":212},[],[214],{"type":48,"value":83},{"type":48,"value":216}," through one\nor more of the six per-domain DOCA Telemetry reader libraries",{"type":48,"value":218},"\n(",{"type":43,"tag":70,"props":220,"children":222},{"className":221},[],[223],{"type":48,"value":224},"doca_telemetry_pcc",{"type":48,"value":226}," \u002F ",{"type":43,"tag":70,"props":228,"children":230},{"className":229},[],[231],{"type":48,"value":232},"_dpa",{"type":48,"value":226},{"type":43,"tag":70,"props":235,"children":237},{"className":236},[],[238],{"type":48,"value":239},"_diag",{"type":48,"value":226},{"type":43,"tag":70,"props":242,"children":244},{"className":243},[],[245],{"type":48,"value":246},"_adp_retx",{"type":48,"value":226},{"type":43,"tag":70,"props":249,"children":251},{"className":250},[],[252],{"type":48,"value":253},"_phy",{"type":48,"value":255}," \u002F\n",{"type":43,"tag":70,"props":257,"children":259},{"className":258},[],[260],{"type":48,"value":261},"_pci",{"type":48,"value":263},") — i.e., users whose application code calls\n",{"type":43,"tag":70,"props":265,"children":267},{"className":266},[],[268],{"type":48,"value":269},"doca_telemetry_\u003Cdomain>_*",{"type":48,"value":271}," (directly in C\u002FC++, or through FFI \u002F\nbindings from another language) to open a per-domain context on\na ",{"type":43,"tag":70,"props":273,"children":275},{"className":274},[],[276],{"type":48,"value":83},{"type":48,"value":278},", configure the per-domain sample window, and read\nthe hardware-counter snapshot for that domain. It is ",{"type":43,"tag":106,"props":280,"children":281},{},[282],{"type":48,"value":283},"not",{"type":48,"value":285}," for\nNVIDIA developers contributing to DOCA Telemetry itself, and it\nis ",{"type":43,"tag":106,"props":287,"children":288},{},[289],{"type":48,"value":283},{"type":48,"value":291}," for users writing the ",{"type":43,"tag":55,"props":293,"children":294},{},[295],{"type":48,"value":296},"publishing \u002F export",{"type":48,"value":298}," side —\nthat is\n",{"type":43,"tag":94,"props":300,"children":301},{"href":152},[302],{"type":43,"tag":70,"props":303,"children":305},{"className":304},[],[306],{"type":48,"value":159},{"type":48,"value":308},",\na separate library and a separate skill.",{"type":43,"tag":51,"props":310,"children":311},{},[312,317,319,325,327,332,334,339,341,347,349,354],{"type":43,"tag":55,"props":313,"children":314},{},[315],{"type":48,"value":316},"Language scope.",{"type":48,"value":318}," DOCA Telemetry's per-domain reader libraries\nship as a C surface with ",{"type":43,"tag":70,"props":320,"children":322},{"className":321},[],[323],{"type":48,"value":324},"pkg-config",{"type":48,"value":326}," module name\n",{"type":43,"tag":70,"props":328,"children":330},{"className":329},[],[331],{"type":48,"value":4},{"type":48,"value":333},". The shipped samples are written in C. C and\nC++ readers are the canonical case; the worked examples in\n",{"type":43,"tag":70,"props":335,"children":337},{"className":336},[],[338],{"type":48,"value":96},{"type":48,"value":340}," assume that path. Other-language readers (Rust, Go,\nPython, …) consume the same ",{"type":43,"tag":70,"props":342,"children":344},{"className":343},[],[345],{"type":48,"value":346},"*.so",{"type":48,"value":348}," through FFI or\nlanguage-specific bindings; the skill's contribution in that\ncase is to keep the reader-vs-exporter distinction, the per-\ndomain cap-query-first discipline, the per-domain DOCA Core\nlifecycle on the ",{"type":43,"tag":70,"props":350,"children":352},{"className":351},[],[353],{"type":48,"value":83},{"type":48,"value":355},", the sample-window discipline, and\nthe error-taxonomy guidance language-neutral, and to route the\nagent to the public per-domain C ABI as the authoritative\nsurface that any wrapper will eventually call.",{"type":43,"tag":194,"props":357,"children":359},{"id":358},"when-to-load-this-skill",[360],{"type":48,"value":361},"When to load this skill",{"type":43,"tag":51,"props":363,"children":364},{},[365,367,372],{"type":48,"value":366},"Load this skill when the user is doing hands-on DOCA Telemetry\n",{"type":43,"tag":55,"props":368,"children":369},{},[370],{"type":48,"value":371},"hardware-counter-reader",{"type":48,"value":373}," work, in any language. Concretely:",{"type":43,"tag":375,"props":376,"children":377},"ul",{},[378,475,534,594,615,638],{"type":43,"tag":379,"props":380,"children":381},"li",{},[382,384,390,392,398,400,406,408,414,416,422,424,430,432,438,440],{"type":48,"value":383},"Picking the right per-domain header for the counters the user\nwants (",{"type":43,"tag":70,"props":385,"children":387},{"className":386},[],[388],{"type":48,"value":389},"doca_telemetry_pcc.h",{"type":48,"value":391}," for Programmable Congestion\nControl counters, ",{"type":43,"tag":70,"props":393,"children":395},{"className":394},[],[396],{"type":48,"value":397},"_dpa.h",{"type":48,"value":399}," for DPA counters, ",{"type":43,"tag":70,"props":401,"children":403},{"className":402},[],[404],{"type":48,"value":405},"_diag.h",{"type":48,"value":407}," for\ngeneric device diagnostic counters, ",{"type":43,"tag":70,"props":409,"children":411},{"className":410},[],[412],{"type":48,"value":413},"_adp_retx.h",{"type":48,"value":415}," for ADP\nretransmit counters, ",{"type":43,"tag":70,"props":417,"children":419},{"className":418},[],[420],{"type":48,"value":421},"_phy.h",{"type":48,"value":423}," for physical-layer counters,\n",{"type":43,"tag":70,"props":425,"children":427},{"className":426},[],[428],{"type":48,"value":429},"_pci.h",{"type":48,"value":431}," for PCI \u002F PCIe counters) and confirming the device\nsupports it via the per-domain ",{"type":43,"tag":70,"props":433,"children":435},{"className":434},[],[436],{"type":48,"value":437},"_cap_is_supported(devinfo)",{"type":48,"value":439},"\nquery — ",{"type":43,"tag":55,"props":441,"children":442},{},[443,445,450,452,458,460,466,467,473],{"type":48,"value":444},"except ",{"type":43,"tag":70,"props":446,"children":448},{"className":447},[],[449],{"type":48,"value":261},{"type":48,"value":451},", which has no single ",{"type":43,"tag":70,"props":453,"children":455},{"className":454},[],[456],{"type":48,"value":457},"_cap_is_supported",{"type":48,"value":459},"\nand instead exposes per-feature caps like\n",{"type":43,"tag":70,"props":461,"children":463},{"className":462},[],[464],{"type":48,"value":465},"doca_telemetry_pci_cap_management_info_is_supported",{"type":48,"value":255},{"type":43,"tag":70,"props":468,"children":470},{"className":469},[],[471],{"type":48,"value":472},"_cap_perf_counters_1_is_supported",{"type":48,"value":474},".",{"type":43,"tag":379,"props":476,"children":477},{},[478,480,485,487,492,494,500,502,508,510,516,518,524,526,532],{"type":48,"value":479},"Opening a per-domain ",{"type":43,"tag":70,"props":481,"children":483},{"className":482},[],[484],{"type":48,"value":75},{"type":48,"value":486}," context on a\n",{"type":43,"tag":70,"props":488,"children":490},{"className":489},[],[491],{"type":48,"value":83},{"type":48,"value":493},", walking the per-domain lifecycle\n(",{"type":43,"tag":70,"props":495,"children":497},{"className":496},[],[498],{"type":48,"value":499},"doca_telemetry_\u003Cdomain>_create(dev)",{"type":48,"value":501}," → per-domain setters →\n",{"type":43,"tag":70,"props":503,"children":505},{"className":504},[],[506],{"type":48,"value":507},"doca_telemetry_\u003Cdomain>_start",{"type":48,"value":509},"), configuring the per-domain\nsample window, and reading the hardware-counter snapshot for\nthat domain. Note this is a per-domain ",{"type":43,"tag":70,"props":511,"children":513},{"className":512},[],[514],{"type":48,"value":515},"_create",{"type":48,"value":517},"\u002F",{"type":43,"tag":70,"props":519,"children":521},{"className":520},[],[522],{"type":48,"value":523},"_start",{"type":48,"value":525},"\nsurface, not the generic ",{"type":43,"tag":70,"props":527,"children":529},{"className":528},[],[530],{"type":48,"value":531},"doca_ctx_*",{"type":48,"value":533}," progress-engine\nlifecycle.",{"type":43,"tag":379,"props":535,"children":536},{},[537,539,545,547,553,555,561,563,569,570,576,578,584,586,592],{"type":48,"value":538},"Reading the device + library capability surface before assuming\na counter family is available: use\n",{"type":43,"tag":70,"props":540,"children":542},{"className":541},[],[543],{"type":48,"value":544},"doca_telemetry_\u003Cdomain>_cap_is_supported",{"type":48,"value":546}," only for ",{"type":43,"tag":70,"props":548,"children":550},{"className":549},[],[551],{"type":48,"value":552},"pcc",{"type":48,"value":554},",\n",{"type":43,"tag":70,"props":556,"children":558},{"className":557},[],[559],{"type":48,"value":560},"dpa",{"type":48,"value":562},", ",{"type":43,"tag":70,"props":564,"children":566},{"className":565},[],[567],{"type":48,"value":568},"diag",{"type":48,"value":562},{"type":43,"tag":70,"props":571,"children":573},{"className":572},[],[574],{"type":48,"value":575},"adp_retx",{"type":48,"value":577},", and ",{"type":43,"tag":70,"props":579,"children":581},{"className":580},[],[582],{"type":48,"value":583},"phy",{"type":48,"value":585},"; use the matching\nper-feature ",{"type":43,"tag":70,"props":587,"children":589},{"className":588},[],[590],{"type":48,"value":591},"doca_telemetry_pci_cap_*_is_supported",{"type":48,"value":593}," query for\nPCI.",{"type":43,"tag":379,"props":595,"children":596},{},[597,599,605,607,613],{"type":48,"value":598},"Handling per-domain ",{"type":43,"tag":70,"props":600,"children":602},{"className":601},[],[603],{"type":48,"value":604},"DOCA_ERROR_*",{"type":48,"value":606}," returns from a counter\nread (lifecycle vs. device-doesn't-support-this-domain vs.\nper-domain ",{"type":43,"tag":70,"props":608,"children":610},{"className":609},[],[611],{"type":48,"value":612},"AGAIN",{"type":48,"value":614},"-means-snapshot-not-ready vs. permission \u002F\ndriver) and the per-read status reported back to the\napplication.",{"type":43,"tag":379,"props":616,"children":617},{},[618,620,628,630,636],{"type":48,"value":619},"Choosing between DOCA Telemetry (hardware-counter READER) and\nan adjacent option:\n",{"type":43,"tag":94,"props":621,"children":622},{"href":152},[623],{"type":43,"tag":70,"props":624,"children":626},{"className":625},[],[627],{"type":48,"value":159},{"type":48,"value":629},"\nwhen the user actually wants to PUBLISH \u002F EXPORT the counter\nvalues (OTLP \u002F Prometheus \u002F labeled metrics);\n",{"type":43,"tag":70,"props":631,"children":633},{"className":632},[],[634],{"type":48,"value":635},"doca-log",{"type":48,"value":637}," when plain structured stdout logging is enough; a\ngeneric Prometheus \u002F OpenTelemetry client library when the\ncounter source is a non-DOCA program; the externally-\nproductized DOCA Telemetry Service (DTS, out of scope) when\nthe user wants a turnkey aggregator.",{"type":43,"tag":379,"props":639,"children":640},{},[641,643,648],{"type":48,"value":642},"Designing or extending non-C bindings (Rust, Go, Python, …)\nthat wrap the per-domain reader C ABI — for the reader-vs-\nexporter distinction, the per-domain cap-query-first rule,\nthe per-domain ",{"type":43,"tag":70,"props":644,"children":646},{"className":645},[],[647],{"type":48,"value":83},{"type":48,"value":649}," lifecycle, the sample-window\ndiscipline, and the error rules the wrapper must honor.",{"type":43,"tag":51,"props":651,"children":652},{},[653,655,659,661,665,667,675,677,685],{"type":48,"value":654},"Do ",{"type":43,"tag":55,"props":656,"children":657},{},[658],{"type":48,"value":283},{"type":48,"value":660}," load this skill for general DOCA orientation, install\nof DOCA itself, the ",{"type":43,"tag":55,"props":662,"children":663},{},[664],{"type":48,"value":296},{"type":48,"value":666}," side\n(",{"type":43,"tag":94,"props":668,"children":669},{"href":152},[670],{"type":43,"tag":70,"props":671,"children":673},{"className":672},[],[674],{"type":48,"value":159},{"type":48,"value":676},"\nhas its own skill), the externally-productized DOCA Telemetry\nService (DTS — out of scope), or non-reader library questions.\nFor those, use\n",{"type":43,"tag":94,"props":678,"children":679},{"href":183},[680],{"type":43,"tag":70,"props":681,"children":683},{"className":682},[],[684],{"type":48,"value":190},{"type":48,"value":474},{"type":43,"tag":194,"props":687,"children":689},{"id":688},"what-this-skill-provides",[690],{"type":48,"value":691},"What this skill provides",{"type":43,"tag":51,"props":693,"children":694},{},[695,697,702],{"type":48,"value":696},"This is a ",{"type":43,"tag":55,"props":698,"children":699},{},[700],{"type":48,"value":701},"thin loader",{"type":48,"value":703},". The body keeps only the orientation\nneeded to pick the right next file. The substantive\nhardware-counter-reader material lives in two companion files:",{"type":43,"tag":375,"props":705,"children":706},{},[707,799],{"type":43,"tag":379,"props":708,"children":709},{},[710,715,717,722,723,728,729,734,735,740,741,746,747,752,754,759,761,766,768,774,776,781,783,791,792,797],{"type":43,"tag":70,"props":711,"children":713},{"className":712},[],[714],{"type":48,"value":115},{"type":48,"value":716}," — what the per-domain readers can express on\nthis install: the reader-vs-exporter role-split rule, the\nsix shipped sub-libraries (",{"type":43,"tag":70,"props":718,"children":720},{"className":719},[],[721],{"type":48,"value":224},{"type":48,"value":226},{"type":43,"tag":70,"props":724,"children":726},{"className":725},[],[727],{"type":48,"value":232},{"type":48,"value":255},{"type":43,"tag":70,"props":730,"children":732},{"className":731},[],[733],{"type":48,"value":239},{"type":48,"value":226},{"type":43,"tag":70,"props":736,"children":738},{"className":737},[],[739],{"type":48,"value":246},{"type":48,"value":226},{"type":43,"tag":70,"props":742,"children":744},{"className":743},[],[745],{"type":48,"value":253},{"type":48,"value":226},{"type":43,"tag":70,"props":748,"children":750},{"className":749},[],[751],{"type":48,"value":261},{"type":48,"value":753},") and which counter\nfamily each one exposes, the per-domain DOCA Core lifecycle\non a ",{"type":43,"tag":70,"props":755,"children":757},{"className":756},[],[758],{"type":48,"value":83},{"type":48,"value":760},", the domain-level capability query for PCC \u002F\nDPA \u002F DIAG \u002F ADP_RETX \u002F PHY and the per-feature capability\nqueries for PCI, the\nreader error taxonomy (mapped onto the cross-library\n",{"type":43,"tag":70,"props":762,"children":764},{"className":763},[],[765],{"type":48,"value":604},{"type":48,"value":767}," set, with the ",{"type":43,"tag":70,"props":769,"children":771},{"className":770},[],[772],{"type":48,"value":773},"NOT_SUPPORTED",{"type":48,"value":775},"-means-domain-\nnot-exposed-on-this-device rule and the ",{"type":43,"tag":70,"props":777,"children":779},{"className":778},[],[780],{"type":48,"value":612},{"type":48,"value":782},"-means-\nsnapshot-not-ready rule called out explicitly), the\nobservability surface (per-read status + per-domain cap-\nquery snapshot at configure time), the safety policy that\ngates per-domain reads behind the cap-query result, and the\npath-selection rule against\n",{"type":43,"tag":94,"props":784,"children":785},{"href":152},[786],{"type":43,"tag":70,"props":787,"children":789},{"className":788},[],[790],{"type":48,"value":159},{"type":48,"value":554},{"type":43,"tag":70,"props":793,"children":795},{"className":794},[],[796],{"type":48,"value":635},{"type":48,"value":798},", and standalone Prometheus \u002F OpenTelemetry \u002F DTS.",{"type":43,"tag":379,"props":800,"children":801},{},[802,807,809,815,816,822,823,829,830,836,837,843,844,850,852,858],{"type":43,"tag":70,"props":803,"children":805},{"className":804},[],[806],{"type":48,"value":96},{"type":48,"value":808}," — step-by-step workflows for the six in-scope\nreader verbs: ",{"type":43,"tag":70,"props":810,"children":812},{"className":811},[],[813],{"type":48,"value":814},"configure",{"type":48,"value":562},{"type":43,"tag":70,"props":817,"children":819},{"className":818},[],[820],{"type":48,"value":821},"build",{"type":48,"value":562},{"type":43,"tag":70,"props":824,"children":826},{"className":825},[],[827],{"type":48,"value":828},"modify",{"type":48,"value":562},{"type":43,"tag":70,"props":831,"children":833},{"className":832},[],[834],{"type":48,"value":835},"run",{"type":48,"value":562},{"type":43,"tag":70,"props":838,"children":840},{"className":839},[],[841],{"type":48,"value":842},"test",{"type":48,"value":554},{"type":43,"tag":70,"props":845,"children":847},{"className":846},[],[848],{"type":48,"value":849},"debug",{"type":48,"value":851},". Plus a ",{"type":43,"tag":70,"props":853,"children":855},{"className":854},[],[856],{"type":48,"value":857},"Deferred task verbs",{"type":48,"value":859}," block that points\nout-of-scope questions at the right next skill.",{"type":43,"tag":51,"props":861,"children":862},{},[863,865,875,877,882,884,890,892,900,902,910],{"type":48,"value":864},"The skill assumes a host where DOCA is already installed at\nthe standard location and a target BlueField DPU or ConnectX NIC\nis available. The ",{"type":43,"tag":94,"props":866,"children":868},{"href":867},"TASKS.md#run",[869],{"type":43,"tag":70,"props":870,"children":872},{"className":871},[],[873],{"type":48,"value":874},"TASKS.md ## run",{"type":48,"value":876}," workflow\nopens the corresponding ",{"type":43,"tag":70,"props":878,"children":880},{"className":879},[],[881],{"type":48,"value":83},{"type":48,"value":883}," and requires the per-domain\ncap-query to return ",{"type":43,"tag":70,"props":885,"children":887},{"className":886},[],[888],{"type":48,"value":889},"DOCA_SUCCESS",{"type":48,"value":891},". It does not cover installing\nDOCA — that path goes through\n",{"type":43,"tag":94,"props":893,"children":894},{"href":133},[895],{"type":43,"tag":70,"props":896,"children":898},{"className":897},[],[899],{"type":48,"value":140},{"type":48,"value":901}," — and it does not\ncover writing the publishing \u002F export side, which is\n",{"type":43,"tag":94,"props":903,"children":904},{"href":152},[905],{"type":43,"tag":70,"props":906,"children":908},{"className":907},[],[909],{"type":48,"value":159},{"type":48,"value":474},{"type":43,"tag":194,"props":912,"children":914},{"id":913},"loading-order",[915],{"type":48,"value":916},"Loading order",{"type":43,"tag":918,"props":919,"children":920},"ol",{},[921,951,984],{"type":43,"tag":379,"props":922,"children":923},{},[924,926,932,934,939,941,949],{"type":48,"value":925},"Read this ",{"type":43,"tag":70,"props":927,"children":929},{"className":928},[],[930],{"type":48,"value":931},"SKILL.md",{"type":48,"value":933}," first to confirm the user's question\nis in scope (specifically, that the user wants to READ a\nper-domain hardware counter via the per-domain reader API\non a ",{"type":43,"tag":70,"props":935,"children":937},{"className":936},[],[938],{"type":48,"value":83},{"type":48,"value":940}," — not PUBLISH counters, which is\n",{"type":43,"tag":94,"props":942,"children":943},{"href":152},[944],{"type":43,"tag":70,"props":945,"children":947},{"className":946},[],[948],{"type":48,"value":159},{"type":48,"value":950},";\nnot deploy DTS, which is out of scope; and not stand up a\nNetFlow \u002F IPFIX collector, which this library does not\nexpose a surface for).",{"type":43,"tag":379,"props":952,"children":953},{},[954],{"type":43,"tag":55,"props":955,"children":956},{},[957,959,964,966,971,972,977,979,983],{"type":48,"value":958},"For the reader-vs-exporter rule, the six per-domain\nsub-libraries, the per-domain DOCA Core lifecycle on a\n",{"type":43,"tag":70,"props":960,"children":962},{"className":961},[],[963],{"type":48,"value":83},{"type":48,"value":965},", the per-domain capability query, the error\ntaxonomy (including the ",{"type":43,"tag":70,"props":967,"children":969},{"className":968},[],[970],{"type":48,"value":773},{"type":48,"value":775},{"type":43,"tag":70,"props":973,"children":975},{"className":974},[],[976],{"type":48,"value":612},{"type":48,"value":978},"-means-\nsnapshot-not-ready rule), observability, the safety\npolicy, and the path-selection rule, see\n",{"type":43,"tag":94,"props":980,"children":981},{"href":115},[982],{"type":48,"value":115},{"type":48,"value":474},{"type":43,"tag":379,"props":985,"children":986},{},[987],{"type":43,"tag":55,"props":988,"children":989},{},[990,992,996],{"type":48,"value":991},"For step-by-step workflows — configure, build, modify,\nrun, test, debug — see ",{"type":43,"tag":94,"props":993,"children":994},{"href":96},[995],{"type":48,"value":96},{"type":48,"value":474},{"type":43,"tag":51,"props":998,"children":999},{},[1000,1002,1012,1014,1022],{"type":48,"value":1001},"Both companion files cross-link to each other,\n",{"type":43,"tag":94,"props":1003,"children":1005},{"href":1004},"..\u002F..\u002Fdoca-version\u002FSKILL.md",[1006],{"type":43,"tag":70,"props":1007,"children":1009},{"className":1008},[],[1010],{"type":48,"value":1011},"doca-version",{"type":48,"value":1013}," for the canonical\nversion-handling rules, and\n",{"type":43,"tag":94,"props":1015,"children":1016},{"href":183},[1017],{"type":43,"tag":70,"props":1018,"children":1020},{"className":1019},[],[1021],{"type":48,"value":190},{"type":48,"value":1023},"\nwhenever the right answer is \"look it up in the public docs or\nthe installed package layout\" rather than \"reader-specific\nguidance\".",{"type":43,"tag":194,"props":1025,"children":1027},{"id":1026},"example-questions-this-skill-answers-well",[1028],{"type":48,"value":1029},"Example questions this skill answers well",{"type":43,"tag":51,"props":1031,"children":1032},{},[1033,1035,1045],{"type":48,"value":1034},"See ",{"type":43,"tag":94,"props":1036,"children":1038},{"href":1037},"references\u002Fdetails.md#example-questions-this-skill-answers-well",[1039],{"type":43,"tag":70,"props":1040,"children":1042},{"className":1041},[],[1043],{"type":48,"value":1044},"references\u002Fdetails.md",{"type":48,"value":474},{"type":43,"tag":194,"props":1047,"children":1049},{"id":1048},"what-this-skill-deliberately-does-not-ship",[1050],{"type":48,"value":1051},"What this skill deliberately does not ship",{"type":43,"tag":51,"props":1053,"children":1054},{},[1055,1056,1065],{"type":48,"value":1034},{"type":43,"tag":94,"props":1057,"children":1059},{"href":1058},"references\u002Fdetails.md#what-this-skill-deliberately-does-not-ship",[1060],{"type":43,"tag":70,"props":1061,"children":1063},{"className":1062},[],[1064],{"type":48,"value":1044},{"type":48,"value":474},{"type":43,"tag":194,"props":1067,"children":1069},{"id":1068},"related-skills",[1070],{"type":48,"value":1071},"Related skills",{"type":43,"tag":51,"props":1073,"children":1074},{},[1075,1076,1085],{"type":48,"value":1034},{"type":43,"tag":94,"props":1077,"children":1079},{"href":1078},"references\u002Fdetails.md#related-skills",[1080],{"type":43,"tag":70,"props":1081,"children":1083},{"className":1082},[],[1084],{"type":48,"value":1044},{"type":48,"value":474},{"items":1087,"total":1246},[1088,1106,1124,1135,1147,1161,1174,1188,1201,1212,1226,1235],{"slug":1089,"name":1089,"fn":1090,"description":1091,"org":1092,"tags":1093,"stars":1103,"repoUrl":1104,"updatedAt":1105},"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},[1094,1097,1100],{"name":1095,"slug":1096,"type":15},"Documentation","documentation",{"name":1098,"slug":1099,"type":15},"MCP","mcp",{"name":1101,"slug":1102,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":1107,"name":1107,"fn":1108,"description":1109,"org":1110,"tags":1111,"stars":1121,"repoUrl":1122,"updatedAt":1123},"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},[1112,1115,1118],{"name":1113,"slug":1114,"type":15},"Containers","containers",{"name":1116,"slug":1117,"type":15},"Deployment","deployment",{"name":1119,"slug":1120,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":1125,"name":1125,"fn":1126,"description":1127,"org":1128,"tags":1129,"stars":1121,"repoUrl":1122,"updatedAt":1134},"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},[1130,1133],{"name":1131,"slug":1132,"type":15},"CI\u002FCD","ci-cd",{"name":1116,"slug":1117,"type":15},"2026-07-14T05:25:59.97109",{"slug":1136,"name":1136,"fn":1137,"description":1138,"org":1139,"tags":1140,"stars":1121,"repoUrl":1122,"updatedAt":1146},"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},[1141,1142,1143],{"name":1131,"slug":1132,"type":15},{"name":1116,"slug":1117,"type":15},{"name":1144,"slug":1145,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":1148,"name":1148,"fn":1149,"description":1150,"org":1151,"tags":1152,"stars":1121,"repoUrl":1122,"updatedAt":1160},"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},[1153,1156,1157],{"name":1154,"slug":1155,"type":15},"Debugging","debugging",{"name":1144,"slug":1145,"type":15},{"name":1158,"slug":1159,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":1162,"name":1162,"fn":1163,"description":1164,"org":1165,"tags":1166,"stars":1121,"repoUrl":1122,"updatedAt":1173},"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},[1167,1170],{"name":1168,"slug":1169,"type":15},"Best Practices","best-practices",{"name":1171,"slug":1172,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":1175,"name":1175,"fn":1176,"description":1177,"org":1178,"tags":1179,"stars":1121,"repoUrl":1122,"updatedAt":1187},"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},[1180,1183,1186],{"name":1181,"slug":1182,"type":15},"Machine Learning","machine-learning",{"name":1184,"slug":1185,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":1189,"name":1189,"fn":1190,"description":1191,"org":1192,"tags":1193,"stars":1121,"repoUrl":1122,"updatedAt":1200},"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},[1194,1197],{"name":1195,"slug":1196,"type":15},"QA","qa",{"name":1198,"slug":1199,"type":15},"Testing","testing","2026-07-14T05:25:53.673039",{"slug":1202,"name":1202,"fn":1203,"description":1204,"org":1205,"tags":1206,"stars":1121,"repoUrl":1122,"updatedAt":1211},"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},[1207,1208],{"name":1116,"slug":1117,"type":15},{"name":1209,"slug":1210,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":1213,"name":1213,"fn":1214,"description":1215,"org":1216,"tags":1217,"stars":1121,"repoUrl":1122,"updatedAt":1225},"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},[1218,1221,1222],{"name":1219,"slug":1220,"type":15},"Code Review","code-review",{"name":1144,"slug":1145,"type":15},{"name":1223,"slug":1224,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":1227,"name":1227,"fn":1228,"description":1229,"org":1230,"tags":1231,"stars":1121,"repoUrl":1122,"updatedAt":1234},"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},[1232,1233],{"name":1195,"slug":1196,"type":15},{"name":1198,"slug":1199,"type":15},"2026-07-14T05:25:54.928983",{"slug":1236,"name":1236,"fn":1237,"description":1238,"org":1239,"tags":1240,"stars":1121,"repoUrl":1122,"updatedAt":1245},"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},[1241,1244],{"name":1242,"slug":1243,"type":15},"Automation","automation",{"name":1131,"slug":1132,"type":15},"2026-07-30T05:29:03.275638",496,{"items":1248,"total":1344},[1249,1266,1276,1290,1300,1315,1330],{"slug":1250,"name":1250,"fn":1251,"description":1252,"org":1253,"tags":1254,"stars":23,"repoUrl":24,"updatedAt":1265},"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},[1255,1258,1261,1262],{"name":1256,"slug":1257,"type":15},"Data Analysis","data-analysis",{"name":1259,"slug":1260,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":1263,"slug":1264,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":1267,"name":1267,"fn":1268,"description":1269,"org":1270,"tags":1271,"stars":23,"repoUrl":24,"updatedAt":1275},"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},[1272,1273,1274],{"name":1116,"slug":1117,"type":15},{"name":1209,"slug":1210,"type":15},{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":1277,"name":1277,"fn":1278,"description":1279,"org":1280,"tags":1281,"stars":23,"repoUrl":24,"updatedAt":1289},"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},[1282,1285,1286],{"name":1283,"slug":1284,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":1287,"slug":1288,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":1291,"name":1291,"fn":1292,"description":1293,"org":1294,"tags":1295,"stars":23,"repoUrl":24,"updatedAt":1299},"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},[1296,1297,1298],{"name":1256,"slug":1257,"type":15},{"name":9,"slug":8,"type":15},{"name":1198,"slug":1199,"type":15},"2026-07-17T05:29:03.913266",{"slug":1301,"name":1301,"fn":1302,"description":1303,"org":1304,"tags":1305,"stars":23,"repoUrl":24,"updatedAt":1314},"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},[1306,1307,1310,1311],{"name":1242,"slug":1243,"type":15},{"name":1308,"slug":1309,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":1312,"slug":1313,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":1316,"name":1316,"fn":1317,"description":1318,"org":1319,"tags":1320,"stars":23,"repoUrl":24,"updatedAt":1329},"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},[1321,1322,1325,1326],{"name":1116,"slug":1117,"type":15},{"name":1323,"slug":1324,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":1327,"slug":1328,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":1331,"name":1331,"fn":1332,"description":1333,"org":1334,"tags":1335,"stars":23,"repoUrl":24,"updatedAt":1343},"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},[1336,1337,1340],{"name":9,"slug":8,"type":15},{"name":1338,"slug":1339,"type":15},"Quantum Computing","quantum-computing",{"name":1341,"slug":1342,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305]