[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-doca-sta":3,"mdc--hsamq2-key":34,"related-repo-nvidia-doca-sta":1455,"related-org-nvidia-doca-sta":1561},{"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-sta","configure NVMe-over-Fabrics storage targets with DOCA","Use this skill when the user is doing hands-on NVMe-over-Fabrics storage-target work on a BlueField DPU or ConnectX NIC with DOCA STA — standing up a doca_sta DOCA Core context that accelerates the target-side NVMe-oF data path over RDMA, defining doca_sta_subsystem targets (NQN + namespaces) backed by local NVMe-PCI backend disks (doca_sta_be), checking device support via doca_sta_cap_is_supported, sizing the per-connection I\u002FO queues, or debugging DOCA_ERROR_* from a STA call. Trigger even when the user does not say \"DOCA STA\" — typical implicit phrasings include \"my NVMe-oF Connect never completes\", \"Identify Controller times out over RoCE\", \"16 I\u002FO queues at depth 1024 — does this BlueField support that\", \"offload the nvmf target onto the DPU\", or \"DOCA_ERROR_IO_FAILED on an NVMe read\". Refuse and route elsewhere for DOCA install, raw RDMA data movement, raw packet I\u002FO, flow-rule programming, or initiator-side \u002F host NVMe stack work — those belong to other skills.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,17,20],{"name":13,"slug":14,"type":15},"Hardware","hardware","tag",{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},"Engineering","engineering",{"name":21,"slug":22,"type":15},"Storage","storage",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-30T05:28:17.499867","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-sta","---\nlicense: Apache-2.0\nname: doca-sta\ndescription: >\n  Use this skill when the user is doing hands-on NVMe-over-Fabrics\n  storage-target work on a BlueField DPU or ConnectX NIC with DOCA STA —\n  standing up a doca_sta DOCA Core context that accelerates the\n  target-side NVMe-oF data path over RDMA, defining\n  doca_sta_subsystem targets (NQN + namespaces) backed by local\n  NVMe-PCI backend disks (doca_sta_be), checking device support via\n  doca_sta_cap_is_supported, sizing the per-connection I\u002FO queues,\n  or debugging DOCA_ERROR_* from a STA call. Trigger even\n  when the user does not say \"DOCA STA\" — typical implicit phrasings\n  include \"my NVMe-oF Connect never completes\", \"Identify Controller\n  times out over RoCE\", \"16 I\u002FO queues at depth 1024 — does this\n  BlueField support that\", \"offload the nvmf target onto the DPU\", or\n  \"DOCA_ERROR_IO_FAILED on an NVMe read\". Refuse and route elsewhere\n  for DOCA install, raw RDMA data movement, raw packet I\u002FO,\n  flow-rule programming, or initiator-side \u002F host NVMe stack work\n  — those belong to other skills.\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 doca-sta`\n  (and `pkg-config doca-rdma` for the NVMe-over-RDMA transport) and\n  inspects \u002Fopt\u002Fmellanox\u002Fdoca\u002F{lib,include,samples,applications}.\n---\n\n# DOCA STA (Storage Target Acceleration)\n\n**Where to start:** This skill assumes DOCA is already installed and\nthe user is doing **hands-on NVMe-over-Fabrics storage-target work** on\na BlueField-class device with DOCA. Open [`TASKS.md`](TASKS.md) if\nthe user wants to *do* something (configure \u002F modify \u002F build \u002F run\n\u002F test \u002F debug); open [`CAPABILITIES.md`](CAPABILITIES.md) when the\nquestion is *what can DOCA STA express* on this version. If the\nuser has not installed DOCA yet, route to\n[`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md) first. If the user is\nasking *\"is this an NVMe-oF initiator\u002Fhost transport?\"*, the\nanswer is no — doca-sta accelerates the **target** side: it presents\nNVMe-oF `doca_sta_subsystem` targets backed by local NVMe-PCI\ndisks; the model lives in\n[`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes).\n\n## Example questions this skill answers well\n\nThe CLASSES of DOCA STA questions this skill is built to answer,\neach with one worked example. The agent should treat the *class*\nas the load-bearing piece — the worked example is a single\ninstance.\n\n- **\"How do I bring up an NVMe-oF target that uses the BlueField\n  to accelerate the storage data path?\"** — worked example: *\"define\n  a `doca_sta_subsystem` (NQN) with one namespace backed by a local\n  NVMe-PCI disk (`doca_sta_be`) and accept NVMe-over-RDMA\n  connections from a remote initiator\"*. Answered by the\n  target-model-and-lifecycle\n  workflow in [`TASKS.md ## configure`](TASKS.md#configure) +\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  target-object table.\n- **\"Can this BlueField accelerate an NVMe-oF target at all?\"** —\n  worked example: *\"my data center is RoCE end-to-end; does this\n  device support DOCA STA target acceleration?\"*. STA transport is\n  RDMA-only (there is no NVMe-over-TCP path). Answered by the\n  capability-query rule (`doca_sta_cap_is_supported` against a\n  `doca_devinfo`) in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  + the discovery step in\n  [`TASKS.md ## configure`](TASKS.md#configure).\n- **\"How deep can I size my I\u002FO queues, and how many I\u002FO queues\n  per connection?\"** — worked example: *\"I want 16 I\u002FO queues at\n  depth 1024 each — does this device support that?\"*. Answered by\n  the queue-sizing capability surface in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  + the queue-sizing step in\n  [`TASKS.md ## configure`](TASKS.md#configure) which gates on\n  the matching `doca_sta_get_max_*` query (e.g.\n  `doca_sta_get_max_qps`, `doca_sta_get_max_io_queue_size`).\n- **\"Which other DOCA libraries do I need alongside doca-sta?\"** —\n  worked example: *\"do I need doca-rdma directly, or does doca-sta\n  hide it from me?\"*. Answered by the substrate-library rule in\n  [`CAPABILITIES.md ## Safety policy`](CAPABILITIES.md#safety-policy)\n  + the env-prep checklist in\n  [`TASKS.md ## configure`](TASKS.md#configure) step 1, which\n  routes the steering side to\n  [`doca-flow`](..\u002Fdoca-flow\u002FSKILL.md) and the RDMA substrate\n  to [`doca-rdma`](..\u002Fdoca-rdma\u002FSKILL.md).\n- **\"Is this STA capability available on my installed DOCA?\"** —\n  worked example: *\"is STA target acceleration supported on this\n  BlueField + DOCA version?\"*. Answered by the version-and-device\n  overlay in\n  [`CAPABILITIES.md ## Version compatibility`](CAPABILITIES.md#version-compatibility),\n  which cross-links the canonical detection chain in\n  [`doca-version`](..\u002F..\u002Fdoca-version\u002FSKILL.md) and adds the\n  STA-specific cap-query rule (`pkg-config --modversion doca-sta`\n  is the build-time anchor; the runtime `doca_sta_cap_is_supported`\n  query is the truth).\n- **\"What does this `DOCA_ERROR_*` from a STA call mean and which\n  layer caused it?\"** — worked example: *\"`DOCA_ERROR_IO_FAILED`\n  on a submitted NVMe read I\u002FO against a target I can ping\"*.\n  Answered by the STA overlay on the cross-library taxonomy in\n  [`CAPABILITIES.md ## Error taxonomy`](CAPABILITIES.md#error-taxonomy)\n  + the layered ladder in\n  [`TASKS.md ## debug`](TASKS.md#debug) that escalates to\n  [`doca-debug`](..\u002F..\u002Fdoca-debug\u002FSKILL.md).\n\n## Audience\n\nThis skill serves **external developers building NVMe-over-Fabrics\nstorage targets that consume DOCA STA on BlueField** — i.e., users\nwhose code calls `doca_sta_*` (directly in C\u002FC++, or through\nFFI\u002Fbindings from another language) to accelerate the target-side\ndata path of an NVMe-oF target on the BlueField hardware: presenting\n`doca_sta_subsystem` targets (NQN + namespaces) backed by local\nNVMe-PCI disks (`doca_sta_be`) to remote initiators over RDMA. The\nskill is *not* for NVIDIA developers contributing to DOCA STA\nitself, and it is *not* for initiator\u002Fhost-side NVMe stacks.\n\n**Language scope.** DOCA STA ships as a C library with\n`pkg-config` module name `doca-sta`. DOCA STA ships **no public\nsamples** — it is absent from the DOCA libraries \u002F\nextension_libraries sample profiles — so the worked examples in\n`TASKS.md` build against the public headers directly rather than\nmodify a shipped sample. C and C++ consumers are the canonical\ncase. Other-language\nconsumers (Rust, Go, Python, …) consume the same `*.so` through\nFFI or language-specific bindings; the skill's contribution in\nthat case is to keep the target-model, lifecycle,\ncapability-discovery, queue-pair shape, substrate-dependency,\nand error-taxonomy guidance language-neutral, and to route the\nagent to the public C ABI as the authoritative surface that any\nwrapper will eventually call.\n\n## When to load this skill\n\nLoad this skill when the user is doing hands-on DOCA STA work,\nin any language. Concretely:\n\n- Initializing a `doca_sta` instance on a `doca_dev` opened\n  against a BlueField PF \u002F SF and configuring the NVMe-oF\n  target subsystems before `doca_ctx_start()`.\n- Defining target resources — `doca_sta_subsystem` (NQN +\n  namespaces) and `doca_sta_be` backend controllers (local\n  NVMe-PCI disks) — and accepting NVMe-oF connections (admin\n  queue plus N I\u002FO queues per connection) on the target side as\n  remote initiators connect over RDMA CM.\n- Reading or setting STA properties via the `doca_sta_set_*`\n  family, checking device support via `doca_sta_cap_is_supported`,\n  and querying sizing limits via the `doca_sta_get_max_*` family\n  (max I\u002FO queue depth, max number of queue pairs, max I\u002FO size,\n  max subsystems, max namespaces per subsystem, max backends).\n- Wiring the **NVMe-over-RDMA** transport — STA's only transport;\n  it lands on the `doca-rdma` substrate and uses RDMA CM for\n  connection establishment — for the target's I\u002FO queues.\n- Wiring DOCA Flow rules so that NVMe-oF traffic actually\n  reaches the STA-managed queues — the steering boundary is\n  `doca-flow`, not `doca-sta`.\n- Debugging a `DOCA_ERROR_*` returned from a STA call (lifecycle\n  vs. capability vs. transport-layer I\u002FO failure vs.\n  driver-below) and the per-queue events on the DOCA Core\n  progress engine.\n- Designing or extending non-C bindings (Rust, Go, Python, …)\n  that wrap the DOCA STA C ABI — for the lifecycle, queue-pair,\n  cap-query, and substrate-dependency rules the wrapper must\n  honor.\n\nDo **not** load this skill for general DOCA orientation, install\nof DOCA itself, raw RDMA data movement (use\n[`doca-rdma`](..\u002Fdoca-rdma\u002FSKILL.md)), raw packet I\u002FO on\nEthernet queues (use [`doca-eth`](..\u002Fdoca-eth\u002FSKILL.md)),\nflow-rule programming (use [`doca-flow`](..\u002Fdoca-flow\u002FSKILL.md)),\nor initiator\u002Fhost-side NVMe stack development\n(SPDK or kernel-nvme own that, not this skill). For DOCA\ndocumentation orientation, 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 STA-specific\nmaterial lives in two companion files:\n\n- `CAPABILITIES.md` — what DOCA STA can express on this\n  version: the target object model (`doca_sta_subsystem` \u002F\n  namespaces \u002F `doca_sta_be` backend NVMe-PCI disks),\n  the NVMe queue-pair shape (admin queue + I\u002FO queues over RDMA),\n  the RDMA-only transport,\n  the capability-query surface (`doca_sta_cap_is_supported` plus\n  the `doca_sta_get_max_*` sizing queries), the STA\n  error taxonomy (mapped onto the cross-library `DOCA_ERROR_*`\n  set), the observability surface (per-queue progress engine\n  events, capability snapshots), and the safety policy that\n  gates substrate-library, permission, and steering\n  preconditions.\n- `TASKS.md` — step-by-step workflows for the six in-scope\n  STA verbs: `configure`, `modify`, `build`, `run`, `test`,\n  `debug`. Plus a `Deferred task verbs` block that points\n  out-of-scope questions at the right next skill, and a\n  `Command appendix` of the recurring commands the agent\n  reaches for.\n\nThe skill assumes a BlueField (with DOCA installed at the\nstandard location) plus a remote NVMe-oF initiator reachable on\nthe fabric to connect into the accelerated target, and one or\nmore local NVMe-PCI disks to back the target's namespaces. It\ndoes not cover installing\nDOCA — that path goes through\n[`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md). It does not cover\ninitiator\u002Fhost-side NVMe stacks (SPDK `bdev_nvme`, kernel `nvme`\nhost) or NVMe protocol semantics above the accelerated target\ndata path — those are out of scope.\n\n## What this skill deliberately does not ship\n\nThis skill is **agent guidance**, not a samples or templates\nbundle. To keep the boundary clean, it deliberately does not\ncontain — and pull requests should not add:\n\n- **Pre-written DOCA STA application source code, in any\n  language.** DOCA STA ships **no public samples** — there is no\n  `\u002Fopt\u002Fmellanox\u002Fdoca\u002Fsamples\u002Fdoca_sta\u002F` directory, and STA is\n  absent from the libraries \u002F extension_libraries sample\n  profiles. The authoritative surface is the public headers under\n  $(pkg-config --variable=includedir doca-common) plus the public\n  DOCA STA guide; the agent builds against those directly rather\n  than modifying a shipped sample, per the\n  [`TASKS.md ## modify`](TASKS.md#modify) workflow.\n- **Initiator\u002Fhost-side NVMe stack glue.** SPDK `bdev_nvme`,\n  the kernel `nvme` host, and any initiator-side NVMe stack are\n  upstream projects out of scope for this skill — DOCA STA is\n  target-side acceleration, not an initiator transport provider.\n- **Standalone build manifests** (`meson.build`, `CMakeLists.txt`,\n  `Cargo.toml`, …) parked inside the skill. The agent\n  constructs the build manifest *in the user's project\n  directory* against the user's installed DOCA, where\n  `pkg-config --modversion doca-sta` is the source of truth.\n- **A `samples\u002F`, `bindings\u002F`, or `reference\u002F` subtree** of any\n  kind. A mock or incomplete artifact in this skill's tree,\n  even one labeled \"reference\", is misleading: users will\n  read it as buildable.\n\n## Loading order\n\n1. Read this `SKILL.md` first to confirm the user's question\n   is in scope.\n2. **For the STA capability matrix, the target object model,\n   queue-pair shape, RDMA-only transport, capability-query\n   rules, error taxonomy, observability, and safety policy,\n   see [CAPABILITIES.md](CAPABILITIES.md).**\n3. **For step-by-step workflows — configure, modify, build,\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,\n[`doca-rdma`](..\u002Fdoca-rdma\u002FSKILL.md) for the RDMA substrate\nthat NVMe-over-RDMA transport lands on,\n[`doca-flow`](..\u002Fdoca-flow\u002FSKILL.md) for the steering rules\nthat direct NVMe traffic to STA-managed queues, 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 \"STA-specific\nguidance\".\n\n## Related skills\n\n- [`doca-public-knowledge-map`](..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md) —\n  the routing table for every public DOCA documentation source\n  and the on-disk layout of an installed DOCA package. The\n  STA URL slug is `DOCA-STA`.\n- [`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md) — env preparation,\n  install verification, BlueField mode checks, and the\n  permission \u002F group-membership requirements for opening a\n  `doca_dev`. This skill assumes its preconditions are\n  satisfied.\n- [`doca-version`](..\u002F..\u002Fdoca-version\u002FSKILL.md) — canonical\n  DOCA version-handling rules. This skill's `## Version\n  compatibility` cross-links the four-way match rule and adds\n  only the STA-specific overlay (STA target-acceleration\n  availability windows, NVMe-oF feature-set device-conditional\n  support).\n- [`doca-structured-tools-contract`](..\u002F..\u002Fdoca-structured-tools-contract\u002FSKILL.md) —\n  the bundle's structured-tools precedence rule (detect \u002F\n  prefer \u002F fall back \u002F report). The Command appendix in\n  [TASKS.md](TASKS.md) honors this contract.\n- [`doca-programming-guide`](..\u002F..\u002Fdoca-programming-guide\u002FSKILL.md) —\n  general DOCA programming patterns shared by every library:\n  the canonical `pkg-config` + meson build pattern, the\n  universal modify-a-shipped-sample first-app workflow, the\n  universal lifecycle, the cross-library `DOCA_ERROR_*`\n  taxonomy, and the program-side debug order. This skill\n  layers STA specifics on top.\n- [`doca-rdma`](..\u002Fdoca-rdma\u002FSKILL.md) — the RDMA substrate\n  that NVMe-over-RDMA transport lands on. STA hides most of\n  the RDMA queue-pair details from the consumer, but the user\n  still needs `doca-rdma` linked in and the device's RDMA\n  capabilities discoverable for the NVMe-over-RDMA path to\n  work.\n- [`doca-eth`](..\u002Fdoca-eth\u002FSKILL.md) — the queue-pair\n  shape that STA's per-connection queue model echoes. Reach\n  here if the user is asking general questions about how\n  DOCA exposes queue-pairs that don't have an STA-specific\n  answer.\n- [`doca-flow`](..\u002Fdoca-flow\u002FSKILL.md) — the steering\n  surface that decides which NVMe-oF packets land on which\n  STA-managed queue. DOCA STA does *not* program steering\n  itself; an NVMe-oF target whose connections never come up\n  is often a missing or wrong Flow rule, not a STA bug.\n- [`doca-debug`](..\u002F..\u002Fdoca-debug\u002FSKILL.md) — the\n  cross-cutting debug ladder (install \u002F version \u002F build \u002F\n  link \u002F runtime \u002F program \u002F driver). STA-specific debug\n  (transport-type mismatches, queue-depth oversize,\n  IO-failed transport errors) overlays on top of that\n  ladder.\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-sta` (and `pkg-config doca-rdma` for the NVMe-over-RDMA transport) and inspects \u002Fopt\u002Fmellanox\u002Fdoca\u002F{lib,include,samples,applications}.\n",{"type":40,"children":41},"root",[42,51,155,162,174,552,558,605,652,658,663,794,848,854,866,984,1015,1021,1033,1176,1182,1225,1270,1276],{"type":43,"tag":44,"props":45,"children":47},"element","h1",{"id":46},"doca-sta-storage-target-acceleration",[48],{"type":49,"value":50},"text","DOCA STA (Storage Target Acceleration)",{"type":43,"tag":52,"props":53,"children":54},"p",{},[55,61,63,68,70,81,83,89,91,100,102,107,109,119,121,126,128,133,135,141,143,153],{"type":43,"tag":56,"props":57,"children":58},"strong",{},[59],{"type":49,"value":60},"Where to start:",{"type":49,"value":62}," This skill assumes DOCA is already installed and\nthe user is doing ",{"type":43,"tag":56,"props":64,"children":65},{},[66],{"type":49,"value":67},"hands-on NVMe-over-Fabrics storage-target work",{"type":49,"value":69}," on\na BlueField-class device with DOCA. Open ",{"type":43,"tag":71,"props":72,"children":74},"a",{"href":73},"TASKS.md",[75],{"type":43,"tag":76,"props":77,"children":79},"code",{"className":78},[],[80],{"type":49,"value":73},{"type":49,"value":82}," if\nthe user wants to ",{"type":43,"tag":84,"props":85,"children":86},"em",{},[87],{"type":49,"value":88},"do",{"type":49,"value":90}," something (configure \u002F modify \u002F build \u002F run\n\u002F test \u002F debug); open ",{"type":43,"tag":71,"props":92,"children":94},{"href":93},"CAPABILITIES.md",[95],{"type":43,"tag":76,"props":96,"children":98},{"className":97},[],[99],{"type":49,"value":93},{"type":49,"value":101}," when the\nquestion is ",{"type":43,"tag":84,"props":103,"children":104},{},[105],{"type":49,"value":106},"what can DOCA STA express",{"type":49,"value":108}," on this version. If the\nuser has not installed DOCA yet, route to\n",{"type":43,"tag":71,"props":110,"children":112},{"href":111},"..\u002F..\u002Fdoca-setup\u002FSKILL.md",[113],{"type":43,"tag":76,"props":114,"children":116},{"className":115},[],[117],{"type":49,"value":118},"doca-setup",{"type":49,"value":120}," first. If the user is\nasking ",{"type":43,"tag":84,"props":122,"children":123},{},[124],{"type":49,"value":125},"\"is this an NVMe-oF initiator\u002Fhost transport?\"",{"type":49,"value":127},", the\nanswer is no — doca-sta accelerates the ",{"type":43,"tag":56,"props":129,"children":130},{},[131],{"type":49,"value":132},"target",{"type":49,"value":134}," side: it presents\nNVMe-oF ",{"type":43,"tag":76,"props":136,"children":138},{"className":137},[],[139],{"type":49,"value":140},"doca_sta_subsystem",{"type":49,"value":142}," targets backed by local NVMe-PCI\ndisks; the model lives in\n",{"type":43,"tag":71,"props":144,"children":146},{"href":145},"CAPABILITIES.md#capabilities-and-modes",[147],{"type":43,"tag":76,"props":148,"children":150},{"className":149},[],[151],{"type":49,"value":152},"CAPABILITIES.md ## Capabilities and modes",{"type":49,"value":154},".",{"type":43,"tag":156,"props":157,"children":159},"h2",{"id":158},"example-questions-this-skill-answers-well",[160],{"type":49,"value":161},"Example questions this skill answers well",{"type":43,"tag":52,"props":163,"children":164},{},[165,167,172],{"type":49,"value":166},"The CLASSES of DOCA STA questions this skill is built to answer,\neach with one worked example. The agent should treat the ",{"type":43,"tag":84,"props":168,"children":169},{},[170],{"type":49,"value":171},"class",{"type":49,"value":173},"\nas the load-bearing piece — the worked example is a single\ninstance.",{"type":43,"tag":175,"props":176,"children":177},"ul",{},[178,233,291,357,424,479],{"type":43,"tag":179,"props":180,"children":181},"li",{},[182,187,189,209,211,221,223,231],{"type":43,"tag":56,"props":183,"children":184},{},[185],{"type":49,"value":186},"\"How do I bring up an NVMe-oF target that uses the BlueField\nto accelerate the storage data path?\"",{"type":49,"value":188}," — worked example: ",{"type":43,"tag":84,"props":190,"children":191},{},[192,194,199,201,207],{"type":49,"value":193},"\"define\na ",{"type":43,"tag":76,"props":195,"children":197},{"className":196},[],[198],{"type":49,"value":140},{"type":49,"value":200}," (NQN) with one namespace backed by a local\nNVMe-PCI disk (",{"type":43,"tag":76,"props":202,"children":204},{"className":203},[],[205],{"type":49,"value":206},"doca_sta_be",{"type":49,"value":208},") and accept NVMe-over-RDMA\nconnections from a remote initiator\"",{"type":49,"value":210},". Answered by the\ntarget-model-and-lifecycle\nworkflow in ",{"type":43,"tag":71,"props":212,"children":214},{"href":213},"TASKS.md#configure",[215],{"type":43,"tag":76,"props":216,"children":218},{"className":217},[],[219],{"type":49,"value":220},"TASKS.md ## configure",{"type":49,"value":222}," +\n",{"type":43,"tag":71,"props":224,"children":225},{"href":145},[226],{"type":43,"tag":76,"props":227,"children":229},{"className":228},[],[230],{"type":49,"value":152},{"type":49,"value":232},"\ntarget-object table.",{"type":43,"tag":179,"props":234,"children":235},{},[236,241,243,248,250,256,258,264,266,274],{"type":43,"tag":56,"props":237,"children":238},{},[239],{"type":49,"value":240},"\"Can this BlueField accelerate an NVMe-oF target at all?\"",{"type":49,"value":242}," —\nworked example: ",{"type":43,"tag":84,"props":244,"children":245},{},[246],{"type":49,"value":247},"\"my data center is RoCE end-to-end; does this\ndevice support DOCA STA target acceleration?\"",{"type":49,"value":249},". STA transport is\nRDMA-only (there is no NVMe-over-TCP path). Answered by the\ncapability-query rule (",{"type":43,"tag":76,"props":251,"children":253},{"className":252},[],[254],{"type":49,"value":255},"doca_sta_cap_is_supported",{"type":49,"value":257}," against a\n",{"type":43,"tag":76,"props":259,"children":261},{"className":260},[],[262],{"type":49,"value":263},"doca_devinfo",{"type":49,"value":265},") in\n",{"type":43,"tag":71,"props":267,"children":268},{"href":145},[269],{"type":43,"tag":76,"props":270,"children":272},{"className":271},[],[273],{"type":49,"value":152},{"type":43,"tag":175,"props":275,"children":276},{},[277],{"type":43,"tag":179,"props":278,"children":279},{},[280,282,290],{"type":49,"value":281},"the discovery step in\n",{"type":43,"tag":71,"props":283,"children":284},{"href":213},[285],{"type":43,"tag":76,"props":286,"children":288},{"className":287},[],[289],{"type":49,"value":220},{"type":49,"value":154},{"type":43,"tag":179,"props":292,"children":293},{},[294,299,300,305,307,315],{"type":43,"tag":56,"props":295,"children":296},{},[297],{"type":49,"value":298},"\"How deep can I size my I\u002FO queues, and how many I\u002FO queues\nper connection?\"",{"type":49,"value":188},{"type":43,"tag":84,"props":301,"children":302},{},[303],{"type":49,"value":304},"\"I want 16 I\u002FO queues at\ndepth 1024 each — does this device support that?\"",{"type":49,"value":306},". Answered by\nthe queue-sizing capability surface in\n",{"type":43,"tag":71,"props":308,"children":309},{"href":145},[310],{"type":43,"tag":76,"props":311,"children":313},{"className":312},[],[314],{"type":49,"value":152},{"type":43,"tag":175,"props":316,"children":317},{},[318],{"type":43,"tag":179,"props":319,"children":320},{},[321,323,331,333,339,341,347,349,355],{"type":49,"value":322},"the queue-sizing step in\n",{"type":43,"tag":71,"props":324,"children":325},{"href":213},[326],{"type":43,"tag":76,"props":327,"children":329},{"className":328},[],[330],{"type":49,"value":220},{"type":49,"value":332}," which gates on\nthe matching ",{"type":43,"tag":76,"props":334,"children":336},{"className":335},[],[337],{"type":49,"value":338},"doca_sta_get_max_*",{"type":49,"value":340}," query (e.g.\n",{"type":43,"tag":76,"props":342,"children":344},{"className":343},[],[345],{"type":49,"value":346},"doca_sta_get_max_qps",{"type":49,"value":348},", ",{"type":43,"tag":76,"props":350,"children":352},{"className":351},[],[353],{"type":49,"value":354},"doca_sta_get_max_io_queue_size",{"type":49,"value":356},").",{"type":43,"tag":179,"props":358,"children":359},{},[360,365,366,371,373,383],{"type":43,"tag":56,"props":361,"children":362},{},[363],{"type":49,"value":364},"\"Which other DOCA libraries do I need alongside doca-sta?\"",{"type":49,"value":242},{"type":43,"tag":84,"props":367,"children":368},{},[369],{"type":49,"value":370},"\"do I need doca-rdma directly, or does doca-sta\nhide it from me?\"",{"type":49,"value":372},". Answered by the substrate-library rule in\n",{"type":43,"tag":71,"props":374,"children":376},{"href":375},"CAPABILITIES.md#safety-policy",[377],{"type":43,"tag":76,"props":378,"children":380},{"className":379},[],[381],{"type":49,"value":382},"CAPABILITIES.md ## Safety policy",{"type":43,"tag":175,"props":384,"children":385},{},[386],{"type":43,"tag":179,"props":387,"children":388},{},[389,391,399,401,411,413,423],{"type":49,"value":390},"the env-prep checklist in\n",{"type":43,"tag":71,"props":392,"children":393},{"href":213},[394],{"type":43,"tag":76,"props":395,"children":397},{"className":396},[],[398],{"type":49,"value":220},{"type":49,"value":400}," step 1, which\nroutes the steering side to\n",{"type":43,"tag":71,"props":402,"children":404},{"href":403},"..\u002Fdoca-flow\u002FSKILL.md",[405],{"type":43,"tag":76,"props":406,"children":408},{"className":407},[],[409],{"type":49,"value":410},"doca-flow",{"type":49,"value":412}," and the RDMA substrate\nto ",{"type":43,"tag":71,"props":414,"children":416},{"href":415},"..\u002Fdoca-rdma\u002FSKILL.md",[417],{"type":43,"tag":76,"props":418,"children":420},{"className":419},[],[421],{"type":49,"value":422},"doca-rdma",{"type":49,"value":154},{"type":43,"tag":179,"props":425,"children":426},{},[427,432,433,438,440,450,452,462,464,470,472,477],{"type":43,"tag":56,"props":428,"children":429},{},[430],{"type":49,"value":431},"\"Is this STA capability available on my installed DOCA?\"",{"type":49,"value":242},{"type":43,"tag":84,"props":434,"children":435},{},[436],{"type":49,"value":437},"\"is STA target acceleration supported on this\nBlueField + DOCA version?\"",{"type":49,"value":439},". Answered by the version-and-device\noverlay in\n",{"type":43,"tag":71,"props":441,"children":443},{"href":442},"CAPABILITIES.md#version-compatibility",[444],{"type":43,"tag":76,"props":445,"children":447},{"className":446},[],[448],{"type":49,"value":449},"CAPABILITIES.md ## Version compatibility",{"type":49,"value":451},",\nwhich cross-links the canonical detection chain in\n",{"type":43,"tag":71,"props":453,"children":455},{"href":454},"..\u002F..\u002Fdoca-version\u002FSKILL.md",[456],{"type":43,"tag":76,"props":457,"children":459},{"className":458},[],[460],{"type":49,"value":461},"doca-version",{"type":49,"value":463}," and adds the\nSTA-specific cap-query rule (",{"type":43,"tag":76,"props":465,"children":467},{"className":466},[],[468],{"type":49,"value":469},"pkg-config --modversion doca-sta",{"type":49,"value":471},"\nis the build-time anchor; the runtime ",{"type":43,"tag":76,"props":473,"children":475},{"className":474},[],[476],{"type":49,"value":255},{"type":49,"value":478},"\nquery is the truth).",{"type":43,"tag":179,"props":480,"children":481},{},[482,495,496,509,511,521],{"type":43,"tag":56,"props":483,"children":484},{},[485,487,493],{"type":49,"value":486},"\"What does this ",{"type":43,"tag":76,"props":488,"children":490},{"className":489},[],[491],{"type":49,"value":492},"DOCA_ERROR_*",{"type":49,"value":494}," from a STA call mean and which\nlayer caused it?\"",{"type":49,"value":188},{"type":43,"tag":84,"props":497,"children":498},{},[499,501,507],{"type":49,"value":500},"\"",{"type":43,"tag":76,"props":502,"children":504},{"className":503},[],[505],{"type":49,"value":506},"DOCA_ERROR_IO_FAILED",{"type":49,"value":508},"\non a submitted NVMe read I\u002FO against a target I can ping\"",{"type":49,"value":510},".\nAnswered by the STA overlay on the cross-library taxonomy in\n",{"type":43,"tag":71,"props":512,"children":514},{"href":513},"CAPABILITIES.md#error-taxonomy",[515],{"type":43,"tag":76,"props":516,"children":518},{"className":517},[],[519],{"type":49,"value":520},"CAPABILITIES.md ## Error taxonomy",{"type":43,"tag":175,"props":522,"children":523},{},[524],{"type":43,"tag":179,"props":525,"children":526},{},[527,529,539,541,551],{"type":49,"value":528},"the layered ladder in\n",{"type":43,"tag":71,"props":530,"children":532},{"href":531},"TASKS.md#debug",[533],{"type":43,"tag":76,"props":534,"children":536},{"className":535},[],[537],{"type":49,"value":538},"TASKS.md ## debug",{"type":49,"value":540}," that escalates to\n",{"type":43,"tag":71,"props":542,"children":544},{"href":543},"..\u002F..\u002Fdoca-debug\u002FSKILL.md",[545],{"type":43,"tag":76,"props":546,"children":548},{"className":547},[],[549],{"type":49,"value":550},"doca-debug",{"type":49,"value":154},{"type":43,"tag":156,"props":553,"children":555},{"id":554},"audience",[556],{"type":49,"value":557},"Audience",{"type":43,"tag":52,"props":559,"children":560},{},[561,563,568,570,576,578,583,585,590,592,597,599,603],{"type":49,"value":562},"This skill serves ",{"type":43,"tag":56,"props":564,"children":565},{},[566],{"type":49,"value":567},"external developers building NVMe-over-Fabrics\nstorage targets that consume DOCA STA on BlueField",{"type":49,"value":569}," — i.e., users\nwhose code calls ",{"type":43,"tag":76,"props":571,"children":573},{"className":572},[],[574],{"type":49,"value":575},"doca_sta_*",{"type":49,"value":577}," (directly in C\u002FC++, or through\nFFI\u002Fbindings from another language) to accelerate the target-side\ndata path of an NVMe-oF target on the BlueField hardware: presenting\n",{"type":43,"tag":76,"props":579,"children":581},{"className":580},[],[582],{"type":49,"value":140},{"type":49,"value":584}," targets (NQN + namespaces) backed by local\nNVMe-PCI disks (",{"type":43,"tag":76,"props":586,"children":588},{"className":587},[],[589],{"type":49,"value":206},{"type":49,"value":591},") to remote initiators over RDMA. The\nskill is ",{"type":43,"tag":84,"props":593,"children":594},{},[595],{"type":49,"value":596},"not",{"type":49,"value":598}," for NVIDIA developers contributing to DOCA STA\nitself, and it is ",{"type":43,"tag":84,"props":600,"children":601},{},[602],{"type":49,"value":596},{"type":49,"value":604}," for initiator\u002Fhost-side NVMe stacks.",{"type":43,"tag":52,"props":606,"children":607},{},[608,613,615,621,623,628,630,635,637,642,644,650],{"type":43,"tag":56,"props":609,"children":610},{},[611],{"type":49,"value":612},"Language scope.",{"type":49,"value":614}," DOCA STA ships as a C library with\n",{"type":43,"tag":76,"props":616,"children":618},{"className":617},[],[619],{"type":49,"value":620},"pkg-config",{"type":49,"value":622}," module name ",{"type":43,"tag":76,"props":624,"children":626},{"className":625},[],[627],{"type":49,"value":4},{"type":49,"value":629},". DOCA STA ships ",{"type":43,"tag":56,"props":631,"children":632},{},[633],{"type":49,"value":634},"no public\nsamples",{"type":49,"value":636}," — it is absent from the DOCA libraries \u002F\nextension_libraries sample profiles — so the worked examples in\n",{"type":43,"tag":76,"props":638,"children":640},{"className":639},[],[641],{"type":49,"value":73},{"type":49,"value":643}," build against the public headers directly rather than\nmodify a shipped sample. C and C++ consumers are the canonical\ncase. Other-language\nconsumers (Rust, Go, Python, …) consume the same ",{"type":43,"tag":76,"props":645,"children":647},{"className":646},[],[648],{"type":49,"value":649},"*.so",{"type":49,"value":651}," through\nFFI or language-specific bindings; the skill's contribution in\nthat case is to keep the target-model, lifecycle,\ncapability-discovery, queue-pair shape, substrate-dependency,\nand error-taxonomy guidance language-neutral, and to route the\nagent to the public C ABI as the authoritative surface that any\nwrapper will eventually call.",{"type":43,"tag":156,"props":653,"children":655},{"id":654},"when-to-load-this-skill",[656],{"type":49,"value":657},"When to load this skill",{"type":43,"tag":52,"props":659,"children":660},{},[661],{"type":49,"value":662},"Load this skill when the user is doing hands-on DOCA STA work,\nin any language. Concretely:",{"type":43,"tag":175,"props":664,"children":665},{},[666,694,713,740,759,777,789],{"type":43,"tag":179,"props":667,"children":668},{},[669,671,677,679,685,687,693],{"type":49,"value":670},"Initializing a ",{"type":43,"tag":76,"props":672,"children":674},{"className":673},[],[675],{"type":49,"value":676},"doca_sta",{"type":49,"value":678}," instance on a ",{"type":43,"tag":76,"props":680,"children":682},{"className":681},[],[683],{"type":49,"value":684},"doca_dev",{"type":49,"value":686}," opened\nagainst a BlueField PF \u002F SF and configuring the NVMe-oF\ntarget subsystems before ",{"type":43,"tag":76,"props":688,"children":690},{"className":689},[],[691],{"type":49,"value":692},"doca_ctx_start()",{"type":49,"value":154},{"type":43,"tag":179,"props":695,"children":696},{},[697,699,704,706,711],{"type":49,"value":698},"Defining target resources — ",{"type":43,"tag":76,"props":700,"children":702},{"className":701},[],[703],{"type":49,"value":140},{"type":49,"value":705}," (NQN +\nnamespaces) and ",{"type":43,"tag":76,"props":707,"children":709},{"className":708},[],[710],{"type":49,"value":206},{"type":49,"value":712}," backend controllers (local\nNVMe-PCI disks) — and accepting NVMe-oF connections (admin\nqueue plus N I\u002FO queues per connection) on the target side as\nremote initiators connect over RDMA CM.",{"type":43,"tag":179,"props":714,"children":715},{},[716,718,724,726,731,733,738],{"type":49,"value":717},"Reading or setting STA properties via the ",{"type":43,"tag":76,"props":719,"children":721},{"className":720},[],[722],{"type":49,"value":723},"doca_sta_set_*",{"type":49,"value":725},"\nfamily, checking device support via ",{"type":43,"tag":76,"props":727,"children":729},{"className":728},[],[730],{"type":49,"value":255},{"type":49,"value":732},",\nand querying sizing limits via the ",{"type":43,"tag":76,"props":734,"children":736},{"className":735},[],[737],{"type":49,"value":338},{"type":49,"value":739}," family\n(max I\u002FO queue depth, max number of queue pairs, max I\u002FO size,\nmax subsystems, max namespaces per subsystem, max backends).",{"type":43,"tag":179,"props":741,"children":742},{},[743,745,750,752,757],{"type":49,"value":744},"Wiring the ",{"type":43,"tag":56,"props":746,"children":747},{},[748],{"type":49,"value":749},"NVMe-over-RDMA",{"type":49,"value":751}," transport — STA's only transport;\nit lands on the ",{"type":43,"tag":76,"props":753,"children":755},{"className":754},[],[756],{"type":49,"value":422},{"type":49,"value":758}," substrate and uses RDMA CM for\nconnection establishment — for the target's I\u002FO queues.",{"type":43,"tag":179,"props":760,"children":761},{},[762,764,769,771,776],{"type":49,"value":763},"Wiring DOCA Flow rules so that NVMe-oF traffic actually\nreaches the STA-managed queues — the steering boundary is\n",{"type":43,"tag":76,"props":765,"children":767},{"className":766},[],[768],{"type":49,"value":410},{"type":49,"value":770},", not ",{"type":43,"tag":76,"props":772,"children":774},{"className":773},[],[775],{"type":49,"value":4},{"type":49,"value":154},{"type":43,"tag":179,"props":778,"children":779},{},[780,782,787],{"type":49,"value":781},"Debugging a ",{"type":43,"tag":76,"props":783,"children":785},{"className":784},[],[786],{"type":49,"value":492},{"type":49,"value":788}," returned from a STA call (lifecycle\nvs. capability vs. transport-layer I\u002FO failure vs.\ndriver-below) and the per-queue events on the DOCA Core\nprogress engine.",{"type":43,"tag":179,"props":790,"children":791},{},[792],{"type":49,"value":793},"Designing or extending non-C bindings (Rust, Go, Python, …)\nthat wrap the DOCA STA C ABI — for the lifecycle, queue-pair,\ncap-query, and substrate-dependency rules the wrapper must\nhonor.",{"type":43,"tag":52,"props":795,"children":796},{},[797,799,803,805,813,815,825,827,835,837,847],{"type":49,"value":798},"Do ",{"type":43,"tag":56,"props":800,"children":801},{},[802],{"type":49,"value":596},{"type":49,"value":804}," load this skill for general DOCA orientation, install\nof DOCA itself, raw RDMA data movement (use\n",{"type":43,"tag":71,"props":806,"children":807},{"href":415},[808],{"type":43,"tag":76,"props":809,"children":811},{"className":810},[],[812],{"type":49,"value":422},{"type":49,"value":814},"), raw packet I\u002FO on\nEthernet queues (use ",{"type":43,"tag":71,"props":816,"children":818},{"href":817},"..\u002Fdoca-eth\u002FSKILL.md",[819],{"type":43,"tag":76,"props":820,"children":822},{"className":821},[],[823],{"type":49,"value":824},"doca-eth",{"type":49,"value":826},"),\nflow-rule programming (use ",{"type":43,"tag":71,"props":828,"children":829},{"href":403},[830],{"type":43,"tag":76,"props":831,"children":833},{"className":832},[],[834],{"type":49,"value":410},{"type":49,"value":836},"),\nor initiator\u002Fhost-side NVMe stack development\n(SPDK or kernel-nvme own that, not this skill). For DOCA\ndocumentation orientation, use\n",{"type":43,"tag":71,"props":838,"children":840},{"href":839},"..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md",[841],{"type":43,"tag":76,"props":842,"children":844},{"className":843},[],[845],{"type":49,"value":846},"doca-public-knowledge-map",{"type":49,"value":154},{"type":43,"tag":156,"props":849,"children":851},{"id":850},"what-this-skill-provides",[852],{"type":49,"value":853},"What this skill provides",{"type":43,"tag":52,"props":855,"children":856},{},[857,859,864],{"type":49,"value":858},"This is a ",{"type":43,"tag":56,"props":860,"children":861},{},[862],{"type":49,"value":863},"thin loader",{"type":49,"value":865},". The body keeps only the orientation\nneeded to pick the right next file. The substantive STA-specific\nmaterial lives in two companion files:",{"type":43,"tag":175,"props":867,"children":868},{},[869,914],{"type":43,"tag":179,"props":870,"children":871},{},[872,877,879,884,886,891,893,898,900,905,907,912],{"type":43,"tag":76,"props":873,"children":875},{"className":874},[],[876],{"type":49,"value":93},{"type":49,"value":878}," — what DOCA STA can express on this\nversion: the target object model (",{"type":43,"tag":76,"props":880,"children":882},{"className":881},[],[883],{"type":49,"value":140},{"type":49,"value":885}," \u002F\nnamespaces \u002F ",{"type":43,"tag":76,"props":887,"children":889},{"className":888},[],[890],{"type":49,"value":206},{"type":49,"value":892}," backend NVMe-PCI disks),\nthe NVMe queue-pair shape (admin queue + I\u002FO queues over RDMA),\nthe RDMA-only transport,\nthe capability-query surface (",{"type":43,"tag":76,"props":894,"children":896},{"className":895},[],[897],{"type":49,"value":255},{"type":49,"value":899}," plus\nthe ",{"type":43,"tag":76,"props":901,"children":903},{"className":902},[],[904],{"type":49,"value":338},{"type":49,"value":906}," sizing queries), the STA\nerror taxonomy (mapped onto the cross-library ",{"type":43,"tag":76,"props":908,"children":910},{"className":909},[],[911],{"type":49,"value":492},{"type":49,"value":913},"\nset), the observability surface (per-queue progress engine\nevents, capability snapshots), and the safety policy that\ngates substrate-library, permission, and steering\npreconditions.",{"type":43,"tag":179,"props":915,"children":916},{},[917,922,924,930,931,937,938,944,945,951,952,958,960,966,968,974,976,982],{"type":43,"tag":76,"props":918,"children":920},{"className":919},[],[921],{"type":49,"value":73},{"type":49,"value":923}," — step-by-step workflows for the six in-scope\nSTA verbs: ",{"type":43,"tag":76,"props":925,"children":927},{"className":926},[],[928],{"type":49,"value":929},"configure",{"type":49,"value":348},{"type":43,"tag":76,"props":932,"children":934},{"className":933},[],[935],{"type":49,"value":936},"modify",{"type":49,"value":348},{"type":43,"tag":76,"props":939,"children":941},{"className":940},[],[942],{"type":49,"value":943},"build",{"type":49,"value":348},{"type":43,"tag":76,"props":946,"children":948},{"className":947},[],[949],{"type":49,"value":950},"run",{"type":49,"value":348},{"type":43,"tag":76,"props":953,"children":955},{"className":954},[],[956],{"type":49,"value":957},"test",{"type":49,"value":959},",\n",{"type":43,"tag":76,"props":961,"children":963},{"className":962},[],[964],{"type":49,"value":965},"debug",{"type":49,"value":967},". Plus a ",{"type":43,"tag":76,"props":969,"children":971},{"className":970},[],[972],{"type":49,"value":973},"Deferred task verbs",{"type":49,"value":975}," block that points\nout-of-scope questions at the right next skill, and a\n",{"type":43,"tag":76,"props":977,"children":979},{"className":978},[],[980],{"type":49,"value":981},"Command appendix",{"type":49,"value":983}," of the recurring commands the agent\nreaches for.",{"type":43,"tag":52,"props":985,"children":986},{},[987,989,997,999,1005,1007,1013],{"type":49,"value":988},"The skill assumes a BlueField (with DOCA installed at the\nstandard location) plus a remote NVMe-oF initiator reachable on\nthe fabric to connect into the accelerated target, and one or\nmore local NVMe-PCI disks to back the target's namespaces. It\ndoes not cover installing\nDOCA — that path goes through\n",{"type":43,"tag":71,"props":990,"children":991},{"href":111},[992],{"type":43,"tag":76,"props":993,"children":995},{"className":994},[],[996],{"type":49,"value":118},{"type":49,"value":998},". It does not cover\ninitiator\u002Fhost-side NVMe stacks (SPDK ",{"type":43,"tag":76,"props":1000,"children":1002},{"className":1001},[],[1003],{"type":49,"value":1004},"bdev_nvme",{"type":49,"value":1006},", kernel ",{"type":43,"tag":76,"props":1008,"children":1010},{"className":1009},[],[1011],{"type":49,"value":1012},"nvme",{"type":49,"value":1014},"\nhost) or NVMe protocol semantics above the accelerated target\ndata path — those are out of scope.",{"type":43,"tag":156,"props":1016,"children":1018},{"id":1017},"what-this-skill-deliberately-does-not-ship",[1019],{"type":49,"value":1020},"What this skill deliberately does not ship",{"type":43,"tag":52,"props":1022,"children":1023},{},[1024,1026,1031],{"type":49,"value":1025},"This skill is ",{"type":43,"tag":56,"props":1027,"children":1028},{},[1029],{"type":49,"value":1030},"agent guidance",{"type":49,"value":1032},", not a samples or templates\nbundle. To keep the boundary clean, it deliberately does not\ncontain — and pull requests should not add:",{"type":43,"tag":175,"props":1034,"children":1035},{},[1036,1073,1097,1143],{"type":43,"tag":179,"props":1037,"children":1038},{},[1039,1044,1046,1051,1053,1059,1061,1071],{"type":43,"tag":56,"props":1040,"children":1041},{},[1042],{"type":49,"value":1043},"Pre-written DOCA STA application source code, in any\nlanguage.",{"type":49,"value":1045}," DOCA STA ships ",{"type":43,"tag":56,"props":1047,"children":1048},{},[1049],{"type":49,"value":1050},"no public samples",{"type":49,"value":1052}," — there is no\n",{"type":43,"tag":76,"props":1054,"children":1056},{"className":1055},[],[1057],{"type":49,"value":1058},"\u002Fopt\u002Fmellanox\u002Fdoca\u002Fsamples\u002Fdoca_sta\u002F",{"type":49,"value":1060}," directory, and STA is\nabsent from the libraries \u002F extension_libraries sample\nprofiles. The authoritative surface is the public headers under\n$(pkg-config --variable=includedir doca-common) plus the public\nDOCA STA guide; the agent builds against those directly rather\nthan modifying a shipped sample, per the\n",{"type":43,"tag":71,"props":1062,"children":1064},{"href":1063},"TASKS.md#modify",[1065],{"type":43,"tag":76,"props":1066,"children":1068},{"className":1067},[],[1069],{"type":49,"value":1070},"TASKS.md ## modify",{"type":49,"value":1072}," workflow.",{"type":43,"tag":179,"props":1074,"children":1075},{},[1076,1081,1083,1088,1090,1095],{"type":43,"tag":56,"props":1077,"children":1078},{},[1079],{"type":49,"value":1080},"Initiator\u002Fhost-side NVMe stack glue.",{"type":49,"value":1082}," SPDK ",{"type":43,"tag":76,"props":1084,"children":1086},{"className":1085},[],[1087],{"type":49,"value":1004},{"type":49,"value":1089},",\nthe kernel ",{"type":43,"tag":76,"props":1091,"children":1093},{"className":1092},[],[1094],{"type":49,"value":1012},{"type":49,"value":1096}," host, and any initiator-side NVMe stack are\nupstream projects out of scope for this skill — DOCA STA is\ntarget-side acceleration, not an initiator transport provider.",{"type":43,"tag":179,"props":1098,"children":1099},{},[1100,1105,1107,1113,1114,1120,1121,1127,1129,1134,1136,1141],{"type":43,"tag":56,"props":1101,"children":1102},{},[1103],{"type":49,"value":1104},"Standalone build manifests",{"type":49,"value":1106}," (",{"type":43,"tag":76,"props":1108,"children":1110},{"className":1109},[],[1111],{"type":49,"value":1112},"meson.build",{"type":49,"value":348},{"type":43,"tag":76,"props":1115,"children":1117},{"className":1116},[],[1118],{"type":49,"value":1119},"CMakeLists.txt",{"type":49,"value":959},{"type":43,"tag":76,"props":1122,"children":1124},{"className":1123},[],[1125],{"type":49,"value":1126},"Cargo.toml",{"type":49,"value":1128},", …) parked inside the skill. The agent\nconstructs the build manifest ",{"type":43,"tag":84,"props":1130,"children":1131},{},[1132],{"type":49,"value":1133},"in the user's project\ndirectory",{"type":49,"value":1135}," against the user's installed DOCA, where\n",{"type":43,"tag":76,"props":1137,"children":1139},{"className":1138},[],[1140],{"type":49,"value":469},{"type":49,"value":1142}," is the source of truth.",{"type":43,"tag":179,"props":1144,"children":1145},{},[1146,1174],{"type":43,"tag":56,"props":1147,"children":1148},{},[1149,1151,1157,1158,1164,1166,1172],{"type":49,"value":1150},"A ",{"type":43,"tag":76,"props":1152,"children":1154},{"className":1153},[],[1155],{"type":49,"value":1156},"samples\u002F",{"type":49,"value":348},{"type":43,"tag":76,"props":1159,"children":1161},{"className":1160},[],[1162],{"type":49,"value":1163},"bindings\u002F",{"type":49,"value":1165},", or ",{"type":43,"tag":76,"props":1167,"children":1169},{"className":1168},[],[1170],{"type":49,"value":1171},"reference\u002F",{"type":49,"value":1173}," subtree",{"type":49,"value":1175}," of any\nkind. A mock or incomplete artifact in this skill's tree,\neven one labeled \"reference\", is misleading: users will\nread it as buildable.",{"type":43,"tag":156,"props":1177,"children":1179},{"id":1178},"loading-order",[1180],{"type":49,"value":1181},"Loading order",{"type":43,"tag":1183,"props":1184,"children":1185},"ol",{},[1186,1199,1212],{"type":43,"tag":179,"props":1187,"children":1188},{},[1189,1191,1197],{"type":49,"value":1190},"Read this ",{"type":43,"tag":76,"props":1192,"children":1194},{"className":1193},[],[1195],{"type":49,"value":1196},"SKILL.md",{"type":49,"value":1198}," first to confirm the user's question\nis in scope.",{"type":43,"tag":179,"props":1200,"children":1201},{},[1202],{"type":43,"tag":56,"props":1203,"children":1204},{},[1205,1207,1211],{"type":49,"value":1206},"For the STA capability matrix, the target object model,\nqueue-pair shape, RDMA-only transport, capability-query\nrules, error taxonomy, observability, and safety policy,\nsee ",{"type":43,"tag":71,"props":1208,"children":1209},{"href":93},[1210],{"type":49,"value":93},{"type":49,"value":154},{"type":43,"tag":179,"props":1213,"children":1214},{},[1215],{"type":43,"tag":56,"props":1216,"children":1217},{},[1218,1220,1224],{"type":49,"value":1219},"For step-by-step workflows — configure, modify, build,\nrun, test, debug — see ",{"type":43,"tag":71,"props":1221,"children":1222},{"href":73},[1223],{"type":49,"value":73},{"type":49,"value":154},{"type":43,"tag":52,"props":1226,"children":1227},{},[1228,1230,1238,1240,1248,1250,1258,1260,1268],{"type":49,"value":1229},"Both companion files cross-link to each other,\n",{"type":43,"tag":71,"props":1231,"children":1232},{"href":454},[1233],{"type":43,"tag":76,"props":1234,"children":1236},{"className":1235},[],[1237],{"type":49,"value":461},{"type":49,"value":1239}," for the canonical\nversion-handling rules,\n",{"type":43,"tag":71,"props":1241,"children":1242},{"href":415},[1243],{"type":43,"tag":76,"props":1244,"children":1246},{"className":1245},[],[1247],{"type":49,"value":422},{"type":49,"value":1249}," for the RDMA substrate\nthat NVMe-over-RDMA transport lands on,\n",{"type":43,"tag":71,"props":1251,"children":1252},{"href":403},[1253],{"type":43,"tag":76,"props":1254,"children":1256},{"className":1255},[],[1257],{"type":49,"value":410},{"type":49,"value":1259}," for the steering rules\nthat direct NVMe traffic to STA-managed queues, and\n",{"type":43,"tag":71,"props":1261,"children":1262},{"href":839},[1263],{"type":43,"tag":76,"props":1264,"children":1266},{"className":1265},[],[1267],{"type":49,"value":846},{"type":49,"value":1269},"\nwhenever the right answer is \"look it up in the public docs or\nthe installed package layout\" rather than \"STA-specific\nguidance\".",{"type":43,"tag":156,"props":1271,"children":1273},{"id":1272},"related-skills",[1274],{"type":49,"value":1275},"Related skills",{"type":43,"tag":175,"props":1277,"children":1278},{},[1279,1299,1319,1340,1361,1390,1410,1423,1442],{"type":43,"tag":179,"props":1280,"children":1281},{},[1282,1290,1292,1298],{"type":43,"tag":71,"props":1283,"children":1284},{"href":839},[1285],{"type":43,"tag":76,"props":1286,"children":1288},{"className":1287},[],[1289],{"type":49,"value":846},{"type":49,"value":1291}," —\nthe routing table for every public DOCA documentation source\nand the on-disk layout of an installed DOCA package. The\nSTA URL slug is ",{"type":43,"tag":76,"props":1293,"children":1295},{"className":1294},[],[1296],{"type":49,"value":1297},"DOCA-STA",{"type":49,"value":154},{"type":43,"tag":179,"props":1300,"children":1301},{},[1302,1310,1312,1317],{"type":43,"tag":71,"props":1303,"children":1304},{"href":111},[1305],{"type":43,"tag":76,"props":1306,"children":1308},{"className":1307},[],[1309],{"type":49,"value":118},{"type":49,"value":1311}," — env preparation,\ninstall verification, BlueField mode checks, and the\npermission \u002F group-membership requirements for opening a\n",{"type":43,"tag":76,"props":1313,"children":1315},{"className":1314},[],[1316],{"type":49,"value":684},{"type":49,"value":1318},". This skill assumes its preconditions are\nsatisfied.",{"type":43,"tag":179,"props":1320,"children":1321},{},[1322,1330,1332,1338],{"type":43,"tag":71,"props":1323,"children":1324},{"href":454},[1325],{"type":43,"tag":76,"props":1326,"children":1328},{"className":1327},[],[1329],{"type":49,"value":461},{"type":49,"value":1331}," — canonical\nDOCA version-handling rules. This skill's ",{"type":43,"tag":76,"props":1333,"children":1335},{"className":1334},[],[1336],{"type":49,"value":1337},"## Version compatibility",{"type":49,"value":1339}," cross-links the four-way match rule and adds\nonly the STA-specific overlay (STA target-acceleration\navailability windows, NVMe-oF feature-set device-conditional\nsupport).",{"type":43,"tag":179,"props":1341,"children":1342},{},[1343,1353,1355,1359],{"type":43,"tag":71,"props":1344,"children":1346},{"href":1345},"..\u002F..\u002Fdoca-structured-tools-contract\u002FSKILL.md",[1347],{"type":43,"tag":76,"props":1348,"children":1350},{"className":1349},[],[1351],{"type":49,"value":1352},"doca-structured-tools-contract",{"type":49,"value":1354}," —\nthe bundle's structured-tools precedence rule (detect \u002F\nprefer \u002F fall back \u002F report). The Command appendix in\n",{"type":43,"tag":71,"props":1356,"children":1357},{"href":73},[1358],{"type":49,"value":73},{"type":49,"value":1360}," honors this contract.",{"type":43,"tag":179,"props":1362,"children":1363},{},[1364,1374,1376,1381,1383,1388],{"type":43,"tag":71,"props":1365,"children":1367},{"href":1366},"..\u002F..\u002Fdoca-programming-guide\u002FSKILL.md",[1368],{"type":43,"tag":76,"props":1369,"children":1371},{"className":1370},[],[1372],{"type":49,"value":1373},"doca-programming-guide",{"type":49,"value":1375}," —\ngeneral DOCA programming patterns shared by every library:\nthe canonical ",{"type":43,"tag":76,"props":1377,"children":1379},{"className":1378},[],[1380],{"type":49,"value":620},{"type":49,"value":1382}," + meson build pattern, the\nuniversal modify-a-shipped-sample first-app workflow, the\nuniversal lifecycle, the cross-library ",{"type":43,"tag":76,"props":1384,"children":1386},{"className":1385},[],[1387],{"type":49,"value":492},{"type":49,"value":1389},"\ntaxonomy, and the program-side debug order. This skill\nlayers STA specifics on top.",{"type":43,"tag":179,"props":1391,"children":1392},{},[1393,1401,1403,1408],{"type":43,"tag":71,"props":1394,"children":1395},{"href":415},[1396],{"type":43,"tag":76,"props":1397,"children":1399},{"className":1398},[],[1400],{"type":49,"value":422},{"type":49,"value":1402}," — the RDMA substrate\nthat NVMe-over-RDMA transport lands on. STA hides most of\nthe RDMA queue-pair details from the consumer, but the user\nstill needs ",{"type":43,"tag":76,"props":1404,"children":1406},{"className":1405},[],[1407],{"type":49,"value":422},{"type":49,"value":1409}," linked in and the device's RDMA\ncapabilities discoverable for the NVMe-over-RDMA path to\nwork.",{"type":43,"tag":179,"props":1411,"children":1412},{},[1413,1421],{"type":43,"tag":71,"props":1414,"children":1415},{"href":817},[1416],{"type":43,"tag":76,"props":1417,"children":1419},{"className":1418},[],[1420],{"type":49,"value":824},{"type":49,"value":1422}," — the queue-pair\nshape that STA's per-connection queue model echoes. Reach\nhere if the user is asking general questions about how\nDOCA exposes queue-pairs that don't have an STA-specific\nanswer.",{"type":43,"tag":179,"props":1424,"children":1425},{},[1426,1434,1436,1440],{"type":43,"tag":71,"props":1427,"children":1428},{"href":403},[1429],{"type":43,"tag":76,"props":1430,"children":1432},{"className":1431},[],[1433],{"type":49,"value":410},{"type":49,"value":1435}," — the steering\nsurface that decides which NVMe-oF packets land on which\nSTA-managed queue. DOCA STA does ",{"type":43,"tag":84,"props":1437,"children":1438},{},[1439],{"type":49,"value":596},{"type":49,"value":1441}," program steering\nitself; an NVMe-oF target whose connections never come up\nis often a missing or wrong Flow rule, not a STA bug.",{"type":43,"tag":179,"props":1443,"children":1444},{},[1445,1453],{"type":43,"tag":71,"props":1446,"children":1447},{"href":543},[1448],{"type":43,"tag":76,"props":1449,"children":1451},{"className":1450},[],[1452],{"type":49,"value":550},{"type":49,"value":1454}," — the\ncross-cutting debug ladder (install \u002F version \u002F build \u002F\nlink \u002F runtime \u002F program \u002F driver). STA-specific debug\n(transport-type mismatches, queue-depth oversize,\nIO-failed transport errors) overlays on top of that\nladder.",{"items":1456,"total":1560},[1457,1474,1488,1502,1514,1531,1546],{"slug":1458,"name":1458,"fn":1459,"description":1460,"org":1461,"tags":1462,"stars":23,"repoUrl":24,"updatedAt":1473},"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},[1463,1466,1469,1470],{"name":1464,"slug":1465,"type":15},"Data Analysis","data-analysis",{"name":1467,"slug":1468,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":1471,"slug":1472,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":1475,"name":1475,"fn":1476,"description":1477,"org":1478,"tags":1479,"stars":23,"repoUrl":24,"updatedAt":1487},"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},[1480,1483,1486],{"name":1481,"slug":1482,"type":15},"Deployment","deployment",{"name":1484,"slug":1485,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":1489,"name":1489,"fn":1490,"description":1491,"org":1492,"tags":1493,"stars":23,"repoUrl":24,"updatedAt":1501},"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},[1494,1497,1498],{"name":1495,"slug":1496,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":1499,"slug":1500,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":1503,"name":1503,"fn":1504,"description":1505,"org":1506,"tags":1507,"stars":23,"repoUrl":24,"updatedAt":1513},"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},[1508,1509,1510],{"name":1464,"slug":1465,"type":15},{"name":9,"slug":8,"type":15},{"name":1511,"slug":1512,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":1515,"name":1515,"fn":1516,"description":1517,"org":1518,"tags":1519,"stars":23,"repoUrl":24,"updatedAt":1530},"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},[1520,1523,1526,1527],{"name":1521,"slug":1522,"type":15},"Automation","automation",{"name":1524,"slug":1525,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":1528,"slug":1529,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":1532,"name":1532,"fn":1533,"description":1534,"org":1535,"tags":1536,"stars":23,"repoUrl":24,"updatedAt":1545},"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},[1537,1538,1541,1542],{"name":1481,"slug":1482,"type":15},{"name":1539,"slug":1540,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":1543,"slug":1544,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":1547,"name":1547,"fn":1548,"description":1549,"org":1550,"tags":1551,"stars":23,"repoUrl":24,"updatedAt":1559},"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},[1552,1553,1556],{"name":9,"slug":8,"type":15},{"name":1554,"slug":1555,"type":15},"Quantum Computing","quantum-computing",{"name":1557,"slug":1558,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":1562,"total":1713},[1563,1581,1597,1608,1620,1634,1647,1661,1672,1681,1695,1704],{"slug":1564,"name":1564,"fn":1565,"description":1566,"org":1567,"tags":1568,"stars":1578,"repoUrl":1579,"updatedAt":1580},"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},[1569,1572,1575],{"name":1570,"slug":1571,"type":15},"Documentation","documentation",{"name":1573,"slug":1574,"type":15},"MCP","mcp",{"name":1576,"slug":1577,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":1582,"name":1582,"fn":1583,"description":1584,"org":1585,"tags":1586,"stars":1594,"repoUrl":1595,"updatedAt":1596},"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},[1587,1590,1591],{"name":1588,"slug":1589,"type":15},"Containers","containers",{"name":1481,"slug":1482,"type":15},{"name":1592,"slug":1593,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":1598,"name":1598,"fn":1599,"description":1600,"org":1601,"tags":1602,"stars":1594,"repoUrl":1595,"updatedAt":1607},"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},[1603,1606],{"name":1604,"slug":1605,"type":15},"CI\u002FCD","ci-cd",{"name":1481,"slug":1482,"type":15},"2026-07-14T05:25:59.97109",{"slug":1609,"name":1609,"fn":1610,"description":1611,"org":1612,"tags":1613,"stars":1594,"repoUrl":1595,"updatedAt":1619},"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},[1614,1615,1616],{"name":1604,"slug":1605,"type":15},{"name":1481,"slug":1482,"type":15},{"name":1617,"slug":1618,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":1621,"name":1621,"fn":1622,"description":1623,"org":1624,"tags":1625,"stars":1594,"repoUrl":1595,"updatedAt":1633},"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},[1626,1629,1630],{"name":1627,"slug":1628,"type":15},"Debugging","debugging",{"name":1617,"slug":1618,"type":15},{"name":1631,"slug":1632,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":1635,"name":1635,"fn":1636,"description":1637,"org":1638,"tags":1639,"stars":1594,"repoUrl":1595,"updatedAt":1646},"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},[1640,1643],{"name":1641,"slug":1642,"type":15},"Best Practices","best-practices",{"name":1644,"slug":1645,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":1648,"name":1648,"fn":1649,"description":1650,"org":1651,"tags":1652,"stars":1594,"repoUrl":1595,"updatedAt":1660},"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},[1653,1656,1659],{"name":1654,"slug":1655,"type":15},"Machine Learning","machine-learning",{"name":1657,"slug":1658,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":1662,"name":1662,"fn":1663,"description":1664,"org":1665,"tags":1666,"stars":1594,"repoUrl":1595,"updatedAt":1671},"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},[1667,1670],{"name":1668,"slug":1669,"type":15},"QA","qa",{"name":1511,"slug":1512,"type":15},"2026-07-14T05:25:53.673039",{"slug":1673,"name":1673,"fn":1674,"description":1675,"org":1676,"tags":1677,"stars":1594,"repoUrl":1595,"updatedAt":1680},"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},[1678,1679],{"name":1481,"slug":1482,"type":15},{"name":1484,"slug":1485,"type":15},"2026-07-14T05:25:49.362534",{"slug":1682,"name":1682,"fn":1683,"description":1684,"org":1685,"tags":1686,"stars":1594,"repoUrl":1595,"updatedAt":1694},"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},[1687,1690,1691],{"name":1688,"slug":1689,"type":15},"Code Review","code-review",{"name":1617,"slug":1618,"type":15},{"name":1692,"slug":1693,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":1696,"name":1696,"fn":1697,"description":1698,"org":1699,"tags":1700,"stars":1594,"repoUrl":1595,"updatedAt":1703},"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},[1701,1702],{"name":1668,"slug":1669,"type":15},{"name":1511,"slug":1512,"type":15},"2026-07-14T05:25:54.928983",{"slug":1705,"name":1705,"fn":1706,"description":1707,"org":1708,"tags":1709,"stars":1594,"repoUrl":1595,"updatedAt":1712},"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},[1710,1711],{"name":1521,"slug":1522,"type":15},{"name":1604,"slug":1605,"type":15},"2026-07-30T05:29:03.275638",496]