[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-doca-flow-grpc-server":3,"mdc-4jdha0-key":31,"related-repo-nvidia-doca-flow-grpc-server":1654,"related-org-nvidia-doca-flow-grpc-server":1758},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":20,"repoUrl":21,"updatedAt":22,"license":23,"forks":24,"topics":25,"repo":26,"sourceUrl":29,"mdContent":30},"doca-flow-grpc-server","configure DOCA Flow gRPC server security","PLAINTEXT-ONLY: the shipped `doca_flow_grpc` server uses `grpc::InsecureServerCredentials()` with NO TLS \u002F mTLS \u002F token-auth knob on the binary — transport security must come from external infrastructure (e.g. an mTLS proxy \u002F sidecar) on a trusted segment. Use this skill when bringing up, configuring, hardening, or debugging `doca_flow_grpc` — the DOCA-shipped gRPC remote-control surface in front of `doca-flow` that lets non-C++ clients (Python, Go, Rust, Java) program Flow pipes and entries over RPC instead of linking `libdoca_flow.so` directly. Trigger even when the user doesn't say 'doca-flow-grpc-server' or 'gRPC' — e.g. 'program Flow rules from Python on another host', 'remotely configure pipes on the BlueField', 'client times out connecting to the Flow server', 'where is the .proto for Flow', 'UNAUTHENTICATED \u002F FAILED_PRECONDITION on a Flow RPC'. Route elsewhere for the underlying doca-flow API, generic gRPC tooling (protoc, language bindings), or DOCA install \u002F BFB bring-up.\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],{"name":13,"slug":14,"type":15},"Security","security","tag",{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},"Infrastructure","infrastructure",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-30T05:28:45.423163","Apache-2.0",281,[],{"repoUrl":21,"stars":20,"forks":24,"topics":27,"description":28},[],"AI agent skills published by NVIDIA","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills\u002Ftree\u002FHEAD\u002Fskills\u002Fdoca-flow-grpc-server","---\nlicense: Apache-2.0\nname: doca-flow-grpc-server\ndescription: >\n  PLAINTEXT-ONLY: the shipped `doca_flow_grpc` server uses\n  `grpc::InsecureServerCredentials()` with NO TLS \u002F mTLS \u002F token-auth\n  knob on the binary — transport security must come from external\n  infrastructure (e.g. an mTLS proxy \u002F sidecar) on a trusted segment.\n  Use this skill when bringing up, configuring, hardening, or\n  debugging `doca_flow_grpc` — the DOCA-shipped gRPC remote-control\n  surface in front of `doca-flow` that lets non-C++ clients (Python,\n  Go, Rust, Java) program Flow pipes and entries over RPC instead of\n  linking `libdoca_flow.so` directly. Trigger even when the user\n  doesn't say 'doca-flow-grpc-server' or 'gRPC' — e.g. 'program Flow\n  rules from Python on another host', 'remotely configure pipes on the\n  BlueField', 'client times out connecting to the Flow server', 'where\n  is the .proto for Flow', 'UNAUTHENTICATED \u002F FAILED_PRECONDITION on a\n  Flow RPC'. Route elsewhere for the underlying doca-flow API, generic\n  gRPC tooling (protoc, language bindings), or DOCA install \u002F BFB\n  bring-up.\nmetadata:\n  kind: tool\ncompatibility: >\n  Requires DOCA on Linux (Ubuntu 22.04\u002F24.04 or RHEL\u002FSLES) with a\n  BlueField DPU or ConnectX NIC. The `doca_flow_grpc` binary is a build\n  artifact (install: false in tools\u002Fflow_grpc_server\u002Fmeson.build, gated by\n  flag_enable_grpc_support + flag_enable_grpc_flow_library) — NOT\n  installed under a default DOCA path; build it from the DOCA source tree\n  with gRPC enabled. Its `.proto` lives under libs\u002Fdoca_flow\u002Fgrpc\u002F.\n  Confirm Flow via `pkg-config doca-flow`.\n\n---\n\n# DOCA Flow gRPC Server (`doca_flow_grpc`)\n\n> **CRITICAL transport-security correction (Run-12 + R13).** The\n> shipped `doca_flow_grpc` \u002F `doca_flow_grpc_client`\n> binaries hard-code the gRPC plaintext credentials surface:\n> the **server** uses **`grpc::InsecureServerCredentials()`** (the\n> C++ gRPC server-side API in `tools\u002Fflow_grpc_server\u002Fserver\u002F`);\n> the **C++ client** uses\n> **`grpc::InsecureChannelCredentials()`** (the C++ gRPC\n> client-side API; the client lives in\n> `libs\u002Fdoca_flow\u002Fgrpc\u002Fclient\u002F`, compiled into the\n> `doca_flow` library, NOT under `tools\u002Fflow_grpc_client\u002F`);\n> the **Python\n> client** uses `grpc.aio.insecure_channel(...)`. Do NOT cite the\n> server-side string as `grpc::InsecureChannelCredentials()` —\n> that is the **client-side** API name and a Grep-against-source\n> verification will fail. There\n> is **no TLS, no mTLS, and no token-auth** knob on the shipped\n> control plane today. Any prose below (or in `CAPABILITIES.md`\n> \u002F `TASKS.md`) that frames \"mTLS \u002F token auth \u002F TLS posture\"\n> as a configurable knob on **this** server is the bundle's\n> previous aspirational framing and is wrong against the shipped\n> source. Treat the server as **plaintext-on-a-trusted-segment\n> only**: it MUST be bound on a control-plane-only network\n> segment behind an external proxy, sidecar, or VPN\n> that itself enforces TLS + identity. Any \"TLS \u002F mTLS \u002F token-\n> auth\" discussion below is about the operator's external\n> hardening layer, NOT a knob on this binary. Routing for an\n> TLS \u002F identity design discussion must stay on the selected\n> external proxy, sidecar, or VPN; never route it to a\n> shipped-today binary knob.\n\n**Where to start:** This is a tool skill for standing up and\noperating `doca_flow_grpc`, the DOCA-shipped gRPC remote-\ncontrol surface for `doca-flow`. Open [`TASKS.md`](TASKS.md) and\nstart at [`## configure`](TASKS.md#configure) to decide whether a\nremote control plane is the right answer at all (vs talking to\n`libdoca_flow.so` directly), then [`## run`](TASKS.md#run) for\nthe start → bind → one-client-smoke sequence, then\n[`## test`](TASKS.md#test) for the smoke-before-bulk loop that\ngates any RPC that mutates Flow \u002F dataplane state. Open\n[`CAPABILITIES.md`](CAPABILITIES.md) when the question is *what\nthe gRPC contract surface looks like* (the `.proto` files shipped\nunder the tool's source tree on the user's install), *which\nexternal proxy \u002F sidecar \u002F VPN protects the plaintext server*, *which language bindings\nthe gRPC ecosystem covers*, or *how to interpret the server's\nown logs alongside the live Flow application's logs*. If DOCA is\nnot installed, route to\n[`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md) first; if the user has\nnot stood up `doca-flow` yet, route to\n[`doca-flow`](..\u002F..\u002Flibs\u002Fdoca-flow\u002FSKILL.md) FIRST — the gRPC\nserver is a remote control plane on top of the Flow library, not\na replacement for it.\n\n## Example questions this skill answers well\n\nThe CLASSES of `doca_flow_grpc` questions this skill is\nbuilt to answer, each with one worked example. The class is the\nload-bearing piece; the worked example is one instance.\n\n- **\"Do I actually need a remote control plane for my Flow\n  pipeline, or should my client just link `libdoca_flow.so`\n  directly?\"** — worked example: *\"my client is a Python\n  service on a different host; can it program Flow rules\n  remotely?\"*. Answered by the *when-to-use-gRPC* decision in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  + the routing into\n  [`doca-flow`](..\u002F..\u002Flibs\u002Fdoca-flow\u002FSKILL.md) when a direct\n  library link is the better answer.\n- **\"Where is the gRPC contract surface actually defined on my\n  install?\"** — worked example: *\"I want to generate a Python\n  client; where do I get the `.proto` file?\"*. Answered by the\n  *the-`.proto`-file-is-the-source-of-truth* rule in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  + the language-bindings discussion of standard gRPC tooling\n  (`protoc` + the language-specific gRPC plugin per the\n  [official gRPC docs](https:\u002F\u002Fgrpc.io\u002Fdocs\u002F) on `grpc.io`).\n- **\"How do I harden the gRPC endpoint so it isn't an open\n  door into my dataplane?\"** — worked example: *\"the server is\n  bound on `0.0.0.0`; what should I do before exposing it?\"*.\n  Answered by the *admin attack surface* posture in\n  [`CAPABILITIES.md ## Safety policy`](CAPABILITIES.md#safety-policy)\n  + the external protection \u002F network-segment decision in\n  [`TASKS.md ## configure`](TASKS.md#configure).\n- **\"How do I smoke ONE client end-to-end before opening the\n  server to the fleet?\"** — worked example: *\"my Python client\n  can dial the endpoint; what is the first RPC I run to prove\n  it talks to the live Flow application?\"*. Answered by the\n  smoke-before-bulk loop in\n  [`TASKS.md ## test`](TASKS.md#test) +\n  [`CAPABILITIES.md ## Safety policy`](CAPABILITIES.md#safety-policy)\n  smoke-before-bulk rule.\n- **\"My client cannot reach the server — is the server down,\n  the wrong endpoint, an external-proxy mismatch, or a version\n  mismatch?\"** — worked example: *\"the client times out\n  connecting\"*. Answered by the layered error taxonomy in\n  [`CAPABILITIES.md ## Error taxonomy`](CAPABILITIES.md#error-taxonomy)\n  + the layered ladder in\n  [`TASKS.md ## debug`](TASKS.md#debug).\n- **\"Is my non-C++ client (Python \u002F Go \u002F Rust) actually the\n  right shape for the gRPC contract, or is there a cleaner\n  path?\"** — worked example: *\"I want a Rust client; what\n  does the `.proto`-generated API look like?\"*. Answered by the\n  language-bindings discussion in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  + the routing through standard gRPC tooling.\n\n## Audience\n\nThis skill serves **external operators, control-plane developers,\nand AI agents who need to program a running DOCA Flow pipeline\nfrom a non-C++ process across a network boundary** instead of\nlinking `libdoca_flow.so` directly into the controlling process.\nConcretely:\n\n- A control-plane engineer writing a Python \u002F Go \u002F Rust client\n  that programs Flow rules on a BlueField from outside the\n  BlueField's address space.\n- A platform operator running a Flow-using service on\n  BlueField who wants to expose a remote-control surface to a\n  centralized control plane.\n- An AI agent driving the *\"can I program these Flow rules\n  from this client \u002F this network position\"* triage step\n  before recommending a code change to the surrounding\n  doca-flow application.\n\nIt is **not** for users debugging the gRPC server's source code,\n**not** a substitute for the live public DOCA Flow gRPC Server\nguide on `docs.nvidia.com`, and **not** the place to learn the\n`doca-flow` API — that audience belongs in\n[`doca-flow`](..\u002F..\u002Flibs\u002Fdoca-flow\u002FSKILL.md).\n\n`doca_flow_grpc` is a **single CLI binary built from the DOCA\nsource tree** (`executable('doca_flow_grpc', ..., install: false)`\nin `tools\u002Fflow_grpc_server\u002Fmeson.build`, gated by\n`flag_enable_grpc_support`), plus its companion `.proto`\ncontract files under `libs\u002Fdoca_flow\u002Fgrpc\u002F`; per the tool's\nsource tree (`server\u002F`, `dpa_device\u002F`, `packet_buffering\u002F`) the\ntool can also be paired with a packet-buffering \u002F DPA-side\nhelper on configurations that need them. The skill uses the\nsame `kind: tool` three-file shape (`SKILL.md` + `CAPABILITIES.md` + `TASKS.md`) the rest of the bundle's tool slot uses — front matter at the top of this file already says `kind: tool`. (Prior bundle revisions said \"library three-file shape\" here; that wording was internally inconsistent with the front matter and is corrected.)\n\n## Language scope\n\nThis skill governs deployment, configuration, hardening, and\nclient-side bring-up across the languages standard gRPC tooling\ncovers — Python, Go, C++, Rust, Java, Node.js, C#, Kotlin, Ruby,\nPHP, Dart — via the language-specific gRPC plugin generated\nfrom the shipped `.proto` files (see the\n[gRPC language support](https:\u002F\u002Fgrpc.io\u002Fdocs\u002Flanguages\u002F) index\non `grpc.io`). The server itself is C++ + DOCA; the client\nlanguages are open, gated only by the standard `protoc` plugin\nset. For the `doca-flow` API the server programs, see\n[`doca-flow`](..\u002F..\u002Flibs\u002Fdoca-flow\u002FSKILL.md) — that surface is\nC-language.\n\n## When to load this skill\n\nLoad this skill when the user is — or the agent needs to — bring\nup `doca_flow_grpc` against a running `doca-flow`\napplication (or its preconditions) and connect a non-C++ client\nto it. Concretely:\n\n- Deciding whether a remote gRPC control plane is the right\n  surface (vs a direct `libdoca_flow.so` link in the client\n  process).\n- Locating the `.proto` files on the user's install so a\n  language-binding client can generate the appropriate\n  stubs.\n- Deciding the deployment's transport-security posture and\n  network segment. NOTE: the shipped server is plaintext-only\n  (`grpc::InsecureServerCredentials()`); TLS \u002F mTLS \u002F token-auth\n  are NOT binary configuration knobs — they are external\n  infrastructure concerns handled by a capable proxy, sidecar,\n  or VPN, and the\n  plaintext endpoint must stay on a trusted, isolated segment.\n- Standing up the server alongside a known-good Flow setup\n  and smoke-testing one client end-to-end before exposing\n  the endpoint to the fleet.\n- Diagnosing a connect \u002F version \u002F RPC failure through the\n  layered taxonomy.\n\nDo **not** load this skill for general DOCA orientation,\n`doca-flow` API work, DOCA install, or general gRPC tooling\n(use the [grpc.io](https:\u002F\u002Fgrpc.io\u002F) docs directly for those).\n\n## What this skill provides\n\nThis is a **thin loader**. Substantive material lives in two\ncompanion files:\n\n- `CAPABILITIES.md` — what `doca_flow_grpc` exposes:\n  the gRPC remote-control surface in front of `doca-flow`,\n  the `.proto`-files-as-authoritative-contract rule (the\n  shipped `.proto` files under the tool's source on the\n  user's install are the source of truth), the *when-to-use-\n  gRPC vs direct-library-link* decision, the language-\n  bindings story (any language standard gRPC tooling covers),\n  the external proxy \u002F sidecar \u002F VPN and network-segment decision, the\n  packet-buffering \u002F DPA-side option per the shipped\n  `packet_buffering\u002F` and `dpa_device\u002F` subtrees, the\n  version overlay (server rides the `doca-flow` library\n  version it links against), the layered error taxonomy\n  (server-not-started \u002F server-binding-failed \u002F external-layer-\n  rejected \u002F RPC-call-error \u002F Flow-precondition-failed \u002F\n  version \u002F cross-cutting), the observability surface (the\n  server's own logs + the live Flow application's logs +\n  the RPC client's status codes), and the safety policy\n  that treats the endpoint as an admin attack surface.\n- `TASKS.md` — step-by-step workflows for the in-scope task\n  verbs: `install` (route to setup; binary is built from\n  source with gRPC support enabled),\n  `configure` (decide remote-vs-direct, pick the external\n  proxy \u002F sidecar \u002F VPN and network segment), `build` (route to install),\n  `modify` (refuse — modify the deployment, not the binary),\n  `run` (start → bind → smoke), `test` (the\n  smoke-before-bulk loop with the client-side stub\n  generation step), `debug` (the layered diagnosis ladder),\n  `use` (the agent-side workflow for consuming a captured\n  gRPC server session), plus a `Deferred task verbs` block\n  and a `Command appendix`.\n\nThe skill assumes a host where DOCA is already installed (or\nthe NGC DOCA container is running) with the Flow library\npresent, a working `doca-flow` application to program against,\nand the operator's awareness that exposing a gRPC control plane\nis a high-stakes posture.\n\n## What this skill deliberately does not ship\n\nThis skill is **agent guidance**, not a samples or scripts\nbundle. To keep the boundary clean, it deliberately does not\ncontain — and pull requests should not add:\n\n- **Verbatim RPC method names, message field inventories, or\n  default endpoint paths.** The `.proto` files shipped under\n  the tool's source tree on the user's install are the\n  authoritative contract; copying them here pins the skill\n  to one release and silently rots when the contract\n  evolves.\n- **Pre-baked client code in any language.** The\n  language-specific gRPC plugin + the shipped `.proto` files\n  are the contract; client code generated from them on the\n  user's installed version is the right answer, not a stub\n  pinned to a snapshot.\n- **A pre-baked external security-layer configuration.** CA,\n  token, and mTLS configuration belong to the selected proxy,\n  sidecar, or VPN and its security review, never to\n  `doca_flow_grpc`.\n- **Wrappers, parsers, or scripts** that proxy the gRPC\n  endpoint into another protocol. The endpoint is the\n  endpoint; if a user wants HTTP\u002FJSON instead, that is a\n  separate concern outside this skill's scope.\n- **A `samples\u002F`, `bindings\u002F`, or `reference\u002F` subtree.**\n  Even one labeled *\"reference\"* is misleading: operators\n  will 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 (the user actually wants a remote gRPC control\n   plane on top of `doca-flow`, not a direct library link or\n   a different DOCA library).\n2. **For what the server exposes, the `.proto`-as-contract\n   rule, the language-bindings story, the external proxy \u002F\n   sidecar \u002F VPN and network-segment decision, version availability, the\n   layered error surface, observability, and safety posture,\n   see [CAPABILITIES.md](CAPABILITIES.md).**\n3. **For the documented start sequence and the\n   smoke-before-bulk workflow — `install`, `configure`,\n   `build`, `modify`, `run`, `test`, `debug`, `use` — see\n   [TASKS.md](TASKS.md).**\n\n## Related skills\n\n- [`doca-flow`](..\u002F..\u002Flibs\u002Fdoca-flow\u002FSKILL.md) — the **base\n  library** the server's gRPC contract is a thin remote-\n  control wrapper over. Pipe \u002F entry \u002F rule semantics, the\n  validate-before-commit rule, the Flow counter \u002F inspector\n  surface all live there.\n- [`doca-flow-tune`](..\u002Fdoca-flow-tune\u002FSKILL.md) — the Flow\n  tuning tool. When a Flow-program change is recommended,\n  the change can be applied through the surrounding\n  application or — when the control plane is remote —\n  through this gRPC server's RPC surface.\n- [`doca-public-knowledge-map`](..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md)\n  — routing to the public DOCA Flow gRPC Server page on\n  `docs.nvidia.com` and the rest of the public DOCA\n  documentation set.\n- [`doca-version`](..\u002F..\u002Fdoca-version\u002FSKILL.md) — canonical\n  version-handling rules. The\n  [`## Version compatibility`](CAPABILITIES.md#version-compatibility)\n  section in this skill is a thin overlay on top.\n- [`doca-debug`](..\u002F..\u002Fdoca-debug\u002FSKILL.md) — the cross-cutting\n  debug ladder. gRPC server failures route into the ladder at\n  the runtime layer.\n- [`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md) — env preparation,\n  install verification, and the NGC DOCA container path.\n- [`doca-hardware-safety`](..\u002F..\u002Fdoca-hardware-safety\u002FSKILL.md) —\n  the cross-cutting hardware-safety meta-policy this skill's\n  `## Safety policy` overlays. Any state-changing RPC is a\n  potential dataplane-affecting change and must respect the\n  meta-policy.\n",{"data":32,"body":36},{"license":23,"name":4,"description":6,"metadata":33,"compatibility":35},{"kind":34},"tool","Requires DOCA on Linux (Ubuntu 22.04\u002F24.04 or RHEL\u002FSLES) with a BlueField DPU or ConnectX NIC. The `doca_flow_grpc` binary is a build artifact (install: false in tools\u002Fflow_grpc_server\u002Fmeson.build, gated by flag_enable_grpc_support + flag_enable_grpc_flow_library) — NOT installed under a default DOCA path; build it from the DOCA source tree with gRPC enabled. Its `.proto` lives under libs\u002Fdoca_flow\u002Fgrpc\u002F. Confirm Flow via `pkg-config doca-flow`.\n",{"type":37,"children":38},"root",[39,57,220,377,384,396,720,726,745,770,818,932,938,990,996,1015,1064,1090,1096,1108,1266,1278,1284,1296,1398,1404,1510,1516],{"type":40,"tag":41,"props":42,"children":44},"element","h1",{"id":43},"doca-flow-grpc-server-doca_flow_grpc",[45,48,55],{"type":46,"value":47},"text","DOCA Flow gRPC Server (",{"type":40,"tag":49,"props":50,"children":52},"code",{"className":51},[],[53],{"type":46,"value":54},"doca_flow_grpc",{"type":46,"value":56},")",{"type":40,"tag":58,"props":59,"children":60},"blockquote",{},[61],{"type":40,"tag":62,"props":63,"children":64},"p",{},[65,71,73,78,80,86,88,93,95,104,106,112,114,119,121,130,132,138,140,146,148,154,155,160,161,167,169,174,176,181,183,188,190,196,198,204,206,211,213,218],{"type":40,"tag":66,"props":67,"children":68},"strong",{},[69],{"type":46,"value":70},"CRITICAL transport-security correction (Run-12 + R13).",{"type":46,"value":72}," The\nshipped ",{"type":40,"tag":49,"props":74,"children":76},{"className":75},[],[77],{"type":46,"value":54},{"type":46,"value":79}," \u002F ",{"type":40,"tag":49,"props":81,"children":83},{"className":82},[],[84],{"type":46,"value":85},"doca_flow_grpc_client",{"type":46,"value":87},"\nbinaries hard-code the gRPC plaintext credentials surface:\nthe ",{"type":40,"tag":66,"props":89,"children":90},{},[91],{"type":46,"value":92},"server",{"type":46,"value":94}," uses ",{"type":40,"tag":66,"props":96,"children":97},{},[98],{"type":40,"tag":49,"props":99,"children":101},{"className":100},[],[102],{"type":46,"value":103},"grpc::InsecureServerCredentials()",{"type":46,"value":105}," (the\nC++ gRPC server-side API in ",{"type":40,"tag":49,"props":107,"children":109},{"className":108},[],[110],{"type":46,"value":111},"tools\u002Fflow_grpc_server\u002Fserver\u002F",{"type":46,"value":113},");\nthe ",{"type":40,"tag":66,"props":115,"children":116},{},[117],{"type":46,"value":118},"C++ client",{"type":46,"value":120}," uses\n",{"type":40,"tag":66,"props":122,"children":123},{},[124],{"type":40,"tag":49,"props":125,"children":127},{"className":126},[],[128],{"type":46,"value":129},"grpc::InsecureChannelCredentials()",{"type":46,"value":131}," (the C++ gRPC\nclient-side API; the client lives in\n",{"type":40,"tag":49,"props":133,"children":135},{"className":134},[],[136],{"type":46,"value":137},"libs\u002Fdoca_flow\u002Fgrpc\u002Fclient\u002F",{"type":46,"value":139},", compiled into the\n",{"type":40,"tag":49,"props":141,"children":143},{"className":142},[],[144],{"type":46,"value":145},"doca_flow",{"type":46,"value":147}," library, NOT under ",{"type":40,"tag":49,"props":149,"children":151},{"className":150},[],[152],{"type":46,"value":153},"tools\u002Fflow_grpc_client\u002F",{"type":46,"value":113},{"type":40,"tag":66,"props":156,"children":157},{},[158],{"type":46,"value":159},"Python\nclient",{"type":46,"value":94},{"type":40,"tag":49,"props":162,"children":164},{"className":163},[],[165],{"type":46,"value":166},"grpc.aio.insecure_channel(...)",{"type":46,"value":168},". Do NOT cite the\nserver-side string as ",{"type":40,"tag":49,"props":170,"children":172},{"className":171},[],[173],{"type":46,"value":129},{"type":46,"value":175}," —\nthat is the ",{"type":40,"tag":66,"props":177,"children":178},{},[179],{"type":46,"value":180},"client-side",{"type":46,"value":182}," API name and a Grep-against-source\nverification will fail. There\nis ",{"type":40,"tag":66,"props":184,"children":185},{},[186],{"type":46,"value":187},"no TLS, no mTLS, and no token-auth",{"type":46,"value":189}," knob on the shipped\ncontrol plane today. Any prose below (or in ",{"type":40,"tag":49,"props":191,"children":193},{"className":192},[],[194],{"type":46,"value":195},"CAPABILITIES.md",{"type":46,"value":197},"\n\u002F ",{"type":40,"tag":49,"props":199,"children":201},{"className":200},[],[202],{"type":46,"value":203},"TASKS.md",{"type":46,"value":205},") that frames \"mTLS \u002F token auth \u002F TLS posture\"\nas a configurable knob on ",{"type":40,"tag":66,"props":207,"children":208},{},[209],{"type":46,"value":210},"this",{"type":46,"value":212}," server is the bundle's\nprevious aspirational framing and is wrong against the shipped\nsource. Treat the server as ",{"type":40,"tag":66,"props":214,"children":215},{},[216],{"type":46,"value":217},"plaintext-on-a-trusted-segment\nonly",{"type":46,"value":219},": it MUST be bound on a control-plane-only network\nsegment behind an external proxy, sidecar, or VPN\nthat itself enforces TLS + identity. Any \"TLS \u002F mTLS \u002F token-\nauth\" discussion below is about the operator's external\nhardening layer, NOT a knob on this binary. Routing for an\nTLS \u002F identity design discussion must stay on the selected\nexternal proxy, sidecar, or VPN; never route it to a\nshipped-today binary knob.",{"type":40,"tag":62,"props":221,"children":222},{},[223,228,230,235,237,243,245,254,256,266,268,274,276,286,288,298,300,308,310,316,318,324,326,331,333,338,340,345,347,357,359,364,366,375],{"type":40,"tag":66,"props":224,"children":225},{},[226],{"type":46,"value":227},"Where to start:",{"type":46,"value":229}," This is a tool skill for standing up and\noperating ",{"type":40,"tag":49,"props":231,"children":233},{"className":232},[],[234],{"type":46,"value":54},{"type":46,"value":236},", the DOCA-shipped gRPC remote-\ncontrol surface for ",{"type":40,"tag":49,"props":238,"children":240},{"className":239},[],[241],{"type":46,"value":242},"doca-flow",{"type":46,"value":244},". Open ",{"type":40,"tag":246,"props":247,"children":248},"a",{"href":203},[249],{"type":40,"tag":49,"props":250,"children":252},{"className":251},[],[253],{"type":46,"value":203},{"type":46,"value":255}," and\nstart at ",{"type":40,"tag":246,"props":257,"children":259},{"href":258},"TASKS.md#configure",[260],{"type":40,"tag":49,"props":261,"children":263},{"className":262},[],[264],{"type":46,"value":265},"## configure",{"type":46,"value":267}," to decide whether a\nremote control plane is the right answer at all (vs talking to\n",{"type":40,"tag":49,"props":269,"children":271},{"className":270},[],[272],{"type":46,"value":273},"libdoca_flow.so",{"type":46,"value":275}," directly), then ",{"type":40,"tag":246,"props":277,"children":279},{"href":278},"TASKS.md#run",[280],{"type":40,"tag":49,"props":281,"children":283},{"className":282},[],[284],{"type":46,"value":285},"## run",{"type":46,"value":287}," for\nthe start → bind → one-client-smoke sequence, then\n",{"type":40,"tag":246,"props":289,"children":291},{"href":290},"TASKS.md#test",[292],{"type":40,"tag":49,"props":293,"children":295},{"className":294},[],[296],{"type":46,"value":297},"## test",{"type":46,"value":299}," for the smoke-before-bulk loop that\ngates any RPC that mutates Flow \u002F dataplane state. Open\n",{"type":40,"tag":246,"props":301,"children":302},{"href":195},[303],{"type":40,"tag":49,"props":304,"children":306},{"className":305},[],[307],{"type":46,"value":195},{"type":46,"value":309}," when the question is ",{"type":40,"tag":311,"props":312,"children":313},"em",{},[314],{"type":46,"value":315},"what\nthe gRPC contract surface looks like",{"type":46,"value":317}," (the ",{"type":40,"tag":49,"props":319,"children":321},{"className":320},[],[322],{"type":46,"value":323},".proto",{"type":46,"value":325}," files shipped\nunder the tool's source tree on the user's install), ",{"type":40,"tag":311,"props":327,"children":328},{},[329],{"type":46,"value":330},"which\nexternal proxy \u002F sidecar \u002F VPN protects the plaintext server",{"type":46,"value":332},", ",{"type":40,"tag":311,"props":334,"children":335},{},[336],{"type":46,"value":337},"which language bindings\nthe gRPC ecosystem covers",{"type":46,"value":339},", or ",{"type":40,"tag":311,"props":341,"children":342},{},[343],{"type":46,"value":344},"how to interpret the server's\nown logs alongside the live Flow application's logs",{"type":46,"value":346},". If DOCA is\nnot installed, route to\n",{"type":40,"tag":246,"props":348,"children":350},{"href":349},"..\u002F..\u002Fdoca-setup\u002FSKILL.md",[351],{"type":40,"tag":49,"props":352,"children":354},{"className":353},[],[355],{"type":46,"value":356},"doca-setup",{"type":46,"value":358}," first; if the user has\nnot stood up ",{"type":40,"tag":49,"props":360,"children":362},{"className":361},[],[363],{"type":46,"value":242},{"type":46,"value":365}," yet, route to\n",{"type":40,"tag":246,"props":367,"children":369},{"href":368},"..\u002F..\u002Flibs\u002Fdoca-flow\u002FSKILL.md",[370],{"type":40,"tag":49,"props":371,"children":373},{"className":372},[],[374],{"type":46,"value":242},{"type":46,"value":376}," FIRST — the gRPC\nserver is a remote control plane on top of the Flow library, not\na replacement for it.",{"type":40,"tag":378,"props":379,"children":381},"h2",{"id":380},"example-questions-this-skill-answers-well",[382],{"type":46,"value":383},"Example questions this skill answers well",{"type":40,"tag":62,"props":385,"children":386},{},[387,389,394],{"type":46,"value":388},"The CLASSES of ",{"type":40,"tag":49,"props":390,"children":392},{"className":391},[],[393],{"type":46,"value":54},{"type":46,"value":395}," questions this skill is\nbuilt to answer, each with one worked example. The class is the\nload-bearing piece; the worked example is one instance.",{"type":40,"tag":397,"props":398,"children":399},"ul",{},[400,460,539,599,636,681],{"type":40,"tag":401,"props":402,"children":403},"li",{},[404,416,418,423,425,430,432,442],{"type":40,"tag":66,"props":405,"children":406},{},[407,409,414],{"type":46,"value":408},"\"Do I actually need a remote control plane for my Flow\npipeline, or should my client just link ",{"type":40,"tag":49,"props":410,"children":412},{"className":411},[],[413],{"type":46,"value":273},{"type":46,"value":415},"\ndirectly?\"",{"type":46,"value":417}," — worked example: ",{"type":40,"tag":311,"props":419,"children":420},{},[421],{"type":46,"value":422},"\"my client is a Python\nservice on a different host; can it program Flow rules\nremotely?\"",{"type":46,"value":424},". Answered by the ",{"type":40,"tag":311,"props":426,"children":427},{},[428],{"type":46,"value":429},"when-to-use-gRPC",{"type":46,"value":431}," decision in\n",{"type":40,"tag":246,"props":433,"children":435},{"href":434},"CAPABILITIES.md#capabilities-and-modes",[436],{"type":40,"tag":49,"props":437,"children":439},{"className":438},[],[440],{"type":46,"value":441},"CAPABILITIES.md ## Capabilities and modes",{"type":40,"tag":397,"props":443,"children":444},{},[445],{"type":40,"tag":401,"props":446,"children":447},{},[448,450,458],{"type":46,"value":449},"the routing into\n",{"type":40,"tag":246,"props":451,"children":452},{"href":368},[453],{"type":40,"tag":49,"props":454,"children":456},{"className":455},[],[457],{"type":46,"value":242},{"type":46,"value":459}," when a direct\nlibrary link is the better answer.",{"type":40,"tag":401,"props":461,"children":462},{},[463,468,469,481,483,495,497,505],{"type":40,"tag":66,"props":464,"children":465},{},[466],{"type":46,"value":467},"\"Where is the gRPC contract surface actually defined on my\ninstall?\"",{"type":46,"value":417},{"type":40,"tag":311,"props":470,"children":471},{},[472,474,479],{"type":46,"value":473},"\"I want to generate a Python\nclient; where do I get the ",{"type":40,"tag":49,"props":475,"children":477},{"className":476},[],[478],{"type":46,"value":323},{"type":46,"value":480}," file?\"",{"type":46,"value":482},". Answered by the\n",{"type":40,"tag":311,"props":484,"children":485},{},[486,488,493],{"type":46,"value":487},"the-",{"type":40,"tag":49,"props":489,"children":491},{"className":490},[],[492],{"type":46,"value":323},{"type":46,"value":494},"-file-is-the-source-of-truth",{"type":46,"value":496}," rule in\n",{"type":40,"tag":246,"props":498,"children":499},{"href":434},[500],{"type":40,"tag":49,"props":501,"children":503},{"className":502},[],[504],{"type":46,"value":441},{"type":40,"tag":397,"props":506,"children":507},{},[508],{"type":40,"tag":401,"props":509,"children":510},{},[511,513,519,521,529,531,537],{"type":46,"value":512},"the language-bindings discussion of standard gRPC tooling\n(",{"type":40,"tag":49,"props":514,"children":516},{"className":515},[],[517],{"type":46,"value":518},"protoc",{"type":46,"value":520}," + the language-specific gRPC plugin per the\n",{"type":40,"tag":246,"props":522,"children":526},{"href":523,"rel":524},"https:\u002F\u002Fgrpc.io\u002Fdocs\u002F",[525],"nofollow",[527],{"type":46,"value":528},"official gRPC docs",{"type":46,"value":530}," on ",{"type":40,"tag":49,"props":532,"children":534},{"className":533},[],[535],{"type":46,"value":536},"grpc.io",{"type":46,"value":538},").",{"type":40,"tag":401,"props":540,"children":541},{},[542,547,548,561,563,568,570,580],{"type":40,"tag":66,"props":543,"children":544},{},[545],{"type":46,"value":546},"\"How do I harden the gRPC endpoint so it isn't an open\ndoor into my dataplane?\"",{"type":46,"value":417},{"type":40,"tag":311,"props":549,"children":550},{},[551,553,559],{"type":46,"value":552},"\"the server is\nbound on ",{"type":40,"tag":49,"props":554,"children":556},{"className":555},[],[557],{"type":46,"value":558},"0.0.0.0",{"type":46,"value":560},"; what should I do before exposing it?\"",{"type":46,"value":562},".\nAnswered by the ",{"type":40,"tag":311,"props":564,"children":565},{},[566],{"type":46,"value":567},"admin attack surface",{"type":46,"value":569}," posture in\n",{"type":40,"tag":246,"props":571,"children":573},{"href":572},"CAPABILITIES.md#safety-policy",[574],{"type":40,"tag":49,"props":575,"children":577},{"className":576},[],[578],{"type":46,"value":579},"CAPABILITIES.md ## Safety policy",{"type":40,"tag":397,"props":581,"children":582},{},[583],{"type":40,"tag":401,"props":584,"children":585},{},[586,588,597],{"type":46,"value":587},"the external protection \u002F network-segment decision in\n",{"type":40,"tag":246,"props":589,"children":590},{"href":258},[591],{"type":40,"tag":49,"props":592,"children":594},{"className":593},[],[595],{"type":46,"value":596},"TASKS.md ## configure",{"type":46,"value":598},".",{"type":40,"tag":401,"props":600,"children":601},{},[602,607,608,613,615,624,626,634],{"type":40,"tag":66,"props":603,"children":604},{},[605],{"type":46,"value":606},"\"How do I smoke ONE client end-to-end before opening the\nserver to the fleet?\"",{"type":46,"value":417},{"type":40,"tag":311,"props":609,"children":610},{},[611],{"type":46,"value":612},"\"my Python client\ncan dial the endpoint; what is the first RPC I run to prove\nit talks to the live Flow application?\"",{"type":46,"value":614},". Answered by the\nsmoke-before-bulk loop in\n",{"type":40,"tag":246,"props":616,"children":617},{"href":290},[618],{"type":40,"tag":49,"props":619,"children":621},{"className":620},[],[622],{"type":46,"value":623},"TASKS.md ## test",{"type":46,"value":625}," +\n",{"type":40,"tag":246,"props":627,"children":628},{"href":572},[629],{"type":40,"tag":49,"props":630,"children":632},{"className":631},[],[633],{"type":46,"value":579},{"type":46,"value":635},"\nsmoke-before-bulk rule.",{"type":40,"tag":401,"props":637,"children":638},{},[639,644,645,650,652,662],{"type":40,"tag":66,"props":640,"children":641},{},[642],{"type":46,"value":643},"\"My client cannot reach the server — is the server down,\nthe wrong endpoint, an external-proxy mismatch, or a version\nmismatch?\"",{"type":46,"value":417},{"type":40,"tag":311,"props":646,"children":647},{},[648],{"type":46,"value":649},"\"the client times out\nconnecting\"",{"type":46,"value":651},". Answered by the layered error taxonomy in\n",{"type":40,"tag":246,"props":653,"children":655},{"href":654},"CAPABILITIES.md#error-taxonomy",[656],{"type":40,"tag":49,"props":657,"children":659},{"className":658},[],[660],{"type":46,"value":661},"CAPABILITIES.md ## Error taxonomy",{"type":40,"tag":397,"props":663,"children":664},{},[665],{"type":40,"tag":401,"props":666,"children":667},{},[668,670,680],{"type":46,"value":669},"the layered ladder in\n",{"type":40,"tag":246,"props":671,"children":673},{"href":672},"TASKS.md#debug",[674],{"type":40,"tag":49,"props":675,"children":677},{"className":676},[],[678],{"type":46,"value":679},"TASKS.md ## debug",{"type":46,"value":598},{"type":40,"tag":401,"props":682,"children":683},{},[684,689,690,702,704,712],{"type":40,"tag":66,"props":685,"children":686},{},[687],{"type":46,"value":688},"\"Is my non-C++ client (Python \u002F Go \u002F Rust) actually the\nright shape for the gRPC contract, or is there a cleaner\npath?\"",{"type":46,"value":417},{"type":40,"tag":311,"props":691,"children":692},{},[693,695,700],{"type":46,"value":694},"\"I want a Rust client; what\ndoes the ",{"type":40,"tag":49,"props":696,"children":698},{"className":697},[],[699],{"type":46,"value":323},{"type":46,"value":701},"-generated API look like?\"",{"type":46,"value":703},". Answered by the\nlanguage-bindings discussion in\n",{"type":40,"tag":246,"props":705,"children":706},{"href":434},[707],{"type":40,"tag":49,"props":708,"children":710},{"className":709},[],[711],{"type":46,"value":441},{"type":40,"tag":397,"props":713,"children":714},{},[715],{"type":40,"tag":401,"props":716,"children":717},{},[718],{"type":46,"value":719},"the routing through standard gRPC tooling.",{"type":40,"tag":378,"props":721,"children":723},{"id":722},"audience",[724],{"type":46,"value":725},"Audience",{"type":40,"tag":62,"props":727,"children":728},{},[729,731,736,738,743],{"type":46,"value":730},"This skill serves ",{"type":40,"tag":66,"props":732,"children":733},{},[734],{"type":46,"value":735},"external operators, control-plane developers,\nand AI agents who need to program a running DOCA Flow pipeline\nfrom a non-C++ process across a network boundary",{"type":46,"value":737}," instead of\nlinking ",{"type":40,"tag":49,"props":739,"children":741},{"className":740},[],[742],{"type":46,"value":273},{"type":46,"value":744}," directly into the controlling process.\nConcretely:",{"type":40,"tag":397,"props":746,"children":747},{},[748,753,758],{"type":40,"tag":401,"props":749,"children":750},{},[751],{"type":46,"value":752},"A control-plane engineer writing a Python \u002F Go \u002F Rust client\nthat programs Flow rules on a BlueField from outside the\nBlueField's address space.",{"type":40,"tag":401,"props":754,"children":755},{},[756],{"type":46,"value":757},"A platform operator running a Flow-using service on\nBlueField who wants to expose a remote-control surface to a\ncentralized control plane.",{"type":40,"tag":401,"props":759,"children":760},{},[761,763,768],{"type":46,"value":762},"An AI agent driving the ",{"type":40,"tag":311,"props":764,"children":765},{},[766],{"type":46,"value":767},"\"can I program these Flow rules\nfrom this client \u002F this network position\"",{"type":46,"value":769}," triage step\nbefore recommending a code change to the surrounding\ndoca-flow application.",{"type":40,"tag":62,"props":771,"children":772},{},[773,775,780,782,786,788,794,796,800,802,807,809,817],{"type":46,"value":774},"It is ",{"type":40,"tag":66,"props":776,"children":777},{},[778],{"type":46,"value":779},"not",{"type":46,"value":781}," for users debugging the gRPC server's source code,\n",{"type":40,"tag":66,"props":783,"children":784},{},[785],{"type":46,"value":779},{"type":46,"value":787}," a substitute for the live public DOCA Flow gRPC Server\nguide on ",{"type":40,"tag":49,"props":789,"children":791},{"className":790},[],[792],{"type":46,"value":793},"docs.nvidia.com",{"type":46,"value":795},", and ",{"type":40,"tag":66,"props":797,"children":798},{},[799],{"type":46,"value":779},{"type":46,"value":801}," the place to learn the\n",{"type":40,"tag":49,"props":803,"children":805},{"className":804},[],[806],{"type":46,"value":242},{"type":46,"value":808}," API — that audience belongs in\n",{"type":40,"tag":246,"props":810,"children":811},{"href":368},[812],{"type":40,"tag":49,"props":813,"children":815},{"className":814},[],[816],{"type":46,"value":242},{"type":46,"value":598},{"type":40,"tag":62,"props":819,"children":820},{},[821,826,828,833,835,841,843,849,851,857,859,864,866,872,874,880,881,887,888,894,896,902,904,910,912,917,918,923,925,930],{"type":40,"tag":49,"props":822,"children":824},{"className":823},[],[825],{"type":46,"value":54},{"type":46,"value":827}," is a ",{"type":40,"tag":66,"props":829,"children":830},{},[831],{"type":46,"value":832},"single CLI binary built from the DOCA\nsource tree",{"type":46,"value":834}," (",{"type":40,"tag":49,"props":836,"children":838},{"className":837},[],[839],{"type":46,"value":840},"executable('doca_flow_grpc', ..., install: false)",{"type":46,"value":842},"\nin ",{"type":40,"tag":49,"props":844,"children":846},{"className":845},[],[847],{"type":46,"value":848},"tools\u002Fflow_grpc_server\u002Fmeson.build",{"type":46,"value":850},", gated by\n",{"type":40,"tag":49,"props":852,"children":854},{"className":853},[],[855],{"type":46,"value":856},"flag_enable_grpc_support",{"type":46,"value":858},"), plus its companion ",{"type":40,"tag":49,"props":860,"children":862},{"className":861},[],[863],{"type":46,"value":323},{"type":46,"value":865},"\ncontract files under ",{"type":40,"tag":49,"props":867,"children":869},{"className":868},[],[870],{"type":46,"value":871},"libs\u002Fdoca_flow\u002Fgrpc\u002F",{"type":46,"value":873},"; per the tool's\nsource tree (",{"type":40,"tag":49,"props":875,"children":877},{"className":876},[],[878],{"type":46,"value":879},"server\u002F",{"type":46,"value":332},{"type":40,"tag":49,"props":882,"children":884},{"className":883},[],[885],{"type":46,"value":886},"dpa_device\u002F",{"type":46,"value":332},{"type":40,"tag":49,"props":889,"children":891},{"className":890},[],[892],{"type":46,"value":893},"packet_buffering\u002F",{"type":46,"value":895},") the\ntool can also be paired with a packet-buffering \u002F DPA-side\nhelper on configurations that need them. The skill uses the\nsame ",{"type":40,"tag":49,"props":897,"children":899},{"className":898},[],[900],{"type":46,"value":901},"kind: tool",{"type":46,"value":903}," three-file shape (",{"type":40,"tag":49,"props":905,"children":907},{"className":906},[],[908],{"type":46,"value":909},"SKILL.md",{"type":46,"value":911}," + ",{"type":40,"tag":49,"props":913,"children":915},{"className":914},[],[916],{"type":46,"value":195},{"type":46,"value":911},{"type":40,"tag":49,"props":919,"children":921},{"className":920},[],[922],{"type":46,"value":203},{"type":46,"value":924},") the rest of the bundle's tool slot uses — front matter at the top of this file already says ",{"type":40,"tag":49,"props":926,"children":928},{"className":927},[],[929],{"type":46,"value":901},{"type":46,"value":931},". (Prior bundle revisions said \"library three-file shape\" here; that wording was internally inconsistent with the front matter and is corrected.)",{"type":40,"tag":378,"props":933,"children":935},{"id":934},"language-scope",[936],{"type":46,"value":937},"Language scope",{"type":40,"tag":62,"props":939,"children":940},{},[941,943,948,950,957,959,964,966,971,973,978,980,988],{"type":46,"value":942},"This skill governs deployment, configuration, hardening, and\nclient-side bring-up across the languages standard gRPC tooling\ncovers — Python, Go, C++, Rust, Java, Node.js, C#, Kotlin, Ruby,\nPHP, Dart — via the language-specific gRPC plugin generated\nfrom the shipped ",{"type":40,"tag":49,"props":944,"children":946},{"className":945},[],[947],{"type":46,"value":323},{"type":46,"value":949}," files (see the\n",{"type":40,"tag":246,"props":951,"children":954},{"href":952,"rel":953},"https:\u002F\u002Fgrpc.io\u002Fdocs\u002Flanguages\u002F",[525],[955],{"type":46,"value":956},"gRPC language support",{"type":46,"value":958}," index\non ",{"type":40,"tag":49,"props":960,"children":962},{"className":961},[],[963],{"type":46,"value":536},{"type":46,"value":965},"). The server itself is C++ + DOCA; the client\nlanguages are open, gated only by the standard ",{"type":40,"tag":49,"props":967,"children":969},{"className":968},[],[970],{"type":46,"value":518},{"type":46,"value":972}," plugin\nset. For the ",{"type":40,"tag":49,"props":974,"children":976},{"className":975},[],[977],{"type":46,"value":242},{"type":46,"value":979}," API the server programs, see\n",{"type":40,"tag":246,"props":981,"children":982},{"href":368},[983],{"type":40,"tag":49,"props":984,"children":986},{"className":985},[],[987],{"type":46,"value":242},{"type":46,"value":989}," — that surface is\nC-language.",{"type":40,"tag":378,"props":991,"children":993},{"id":992},"when-to-load-this-skill",[994],{"type":46,"value":995},"When to load this skill",{"type":40,"tag":62,"props":997,"children":998},{},[999,1001,1006,1008,1013],{"type":46,"value":1000},"Load this skill when the user is — or the agent needs to — bring\nup ",{"type":40,"tag":49,"props":1002,"children":1004},{"className":1003},[],[1005],{"type":46,"value":54},{"type":46,"value":1007}," against a running ",{"type":40,"tag":49,"props":1009,"children":1011},{"className":1010},[],[1012],{"type":46,"value":242},{"type":46,"value":1014},"\napplication (or its preconditions) and connect a non-C++ client\nto it. Concretely:",{"type":40,"tag":397,"props":1016,"children":1017},{},[1018,1030,1042,1054,1059],{"type":40,"tag":401,"props":1019,"children":1020},{},[1021,1023,1028],{"type":46,"value":1022},"Deciding whether a remote gRPC control plane is the right\nsurface (vs a direct ",{"type":40,"tag":49,"props":1024,"children":1026},{"className":1025},[],[1027],{"type":46,"value":273},{"type":46,"value":1029}," link in the client\nprocess).",{"type":40,"tag":401,"props":1031,"children":1032},{},[1033,1035,1040],{"type":46,"value":1034},"Locating the ",{"type":40,"tag":49,"props":1036,"children":1038},{"className":1037},[],[1039],{"type":46,"value":323},{"type":46,"value":1041}," files on the user's install so a\nlanguage-binding client can generate the appropriate\nstubs.",{"type":40,"tag":401,"props":1043,"children":1044},{},[1045,1047,1052],{"type":46,"value":1046},"Deciding the deployment's transport-security posture and\nnetwork segment. NOTE: the shipped server is plaintext-only\n(",{"type":40,"tag":49,"props":1048,"children":1050},{"className":1049},[],[1051],{"type":46,"value":103},{"type":46,"value":1053},"); TLS \u002F mTLS \u002F token-auth\nare NOT binary configuration knobs — they are external\ninfrastructure concerns handled by a capable proxy, sidecar,\nor VPN, and the\nplaintext endpoint must stay on a trusted, isolated segment.",{"type":40,"tag":401,"props":1055,"children":1056},{},[1057],{"type":46,"value":1058},"Standing up the server alongside a known-good Flow setup\nand smoke-testing one client end-to-end before exposing\nthe endpoint to the fleet.",{"type":40,"tag":401,"props":1060,"children":1061},{},[1062],{"type":46,"value":1063},"Diagnosing a connect \u002F version \u002F RPC failure through the\nlayered taxonomy.",{"type":40,"tag":62,"props":1065,"children":1066},{},[1067,1069,1073,1075,1080,1082,1088],{"type":46,"value":1068},"Do ",{"type":40,"tag":66,"props":1070,"children":1071},{},[1072],{"type":46,"value":779},{"type":46,"value":1074}," load this skill for general DOCA orientation,\n",{"type":40,"tag":49,"props":1076,"children":1078},{"className":1077},[],[1079],{"type":46,"value":242},{"type":46,"value":1081}," API work, DOCA install, or general gRPC tooling\n(use the ",{"type":40,"tag":246,"props":1083,"children":1086},{"href":1084,"rel":1085},"https:\u002F\u002Fgrpc.io\u002F",[525],[1087],{"type":46,"value":536},{"type":46,"value":1089}," docs directly for those).",{"type":40,"tag":378,"props":1091,"children":1093},{"id":1092},"what-this-skill-provides",[1094],{"type":46,"value":1095},"What this skill provides",{"type":40,"tag":62,"props":1097,"children":1098},{},[1099,1101,1106],{"type":46,"value":1100},"This is a ",{"type":40,"tag":66,"props":1102,"children":1103},{},[1104],{"type":46,"value":1105},"thin loader",{"type":46,"value":1107},". Substantive material lives in two\ncompanion files:",{"type":40,"tag":397,"props":1109,"children":1110},{},[1111,1177],{"type":40,"tag":401,"props":1112,"children":1113},{},[1114,1119,1121,1126,1128,1133,1135,1140,1142,1147,1149,1154,1156,1161,1163,1168,1170,1175],{"type":40,"tag":49,"props":1115,"children":1117},{"className":1116},[],[1118],{"type":46,"value":195},{"type":46,"value":1120}," — what ",{"type":40,"tag":49,"props":1122,"children":1124},{"className":1123},[],[1125],{"type":46,"value":54},{"type":46,"value":1127}," exposes:\nthe gRPC remote-control surface in front of ",{"type":40,"tag":49,"props":1129,"children":1131},{"className":1130},[],[1132],{"type":46,"value":242},{"type":46,"value":1134},",\nthe ",{"type":40,"tag":49,"props":1136,"children":1138},{"className":1137},[],[1139],{"type":46,"value":323},{"type":46,"value":1141},"-files-as-authoritative-contract rule (the\nshipped ",{"type":40,"tag":49,"props":1143,"children":1145},{"className":1144},[],[1146],{"type":46,"value":323},{"type":46,"value":1148}," files under the tool's source on the\nuser's install are the source of truth), the ",{"type":40,"tag":311,"props":1150,"children":1151},{},[1152],{"type":46,"value":1153},"when-to-use-\ngRPC vs direct-library-link",{"type":46,"value":1155}," decision, the language-\nbindings story (any language standard gRPC tooling covers),\nthe external proxy \u002F sidecar \u002F VPN and network-segment decision, the\npacket-buffering \u002F DPA-side option per the shipped\n",{"type":40,"tag":49,"props":1157,"children":1159},{"className":1158},[],[1160],{"type":46,"value":893},{"type":46,"value":1162}," and ",{"type":40,"tag":49,"props":1164,"children":1166},{"className":1165},[],[1167],{"type":46,"value":886},{"type":46,"value":1169}," subtrees, the\nversion overlay (server rides the ",{"type":40,"tag":49,"props":1171,"children":1173},{"className":1172},[],[1174],{"type":46,"value":242},{"type":46,"value":1176}," library\nversion it links against), the layered error taxonomy\n(server-not-started \u002F server-binding-failed \u002F external-layer-\nrejected \u002F RPC-call-error \u002F Flow-precondition-failed \u002F\nversion \u002F cross-cutting), the observability surface (the\nserver's own logs + the live Flow application's logs +\nthe RPC client's status codes), and the safety policy\nthat treats the endpoint as an admin attack surface.",{"type":40,"tag":401,"props":1178,"children":1179},{},[1180,1185,1187,1193,1195,1201,1203,1209,1211,1217,1219,1225,1227,1233,1235,1241,1243,1249,1251,1257,1259,1265],{"type":40,"tag":49,"props":1181,"children":1183},{"className":1182},[],[1184],{"type":46,"value":203},{"type":46,"value":1186}," — step-by-step workflows for the in-scope task\nverbs: ",{"type":40,"tag":49,"props":1188,"children":1190},{"className":1189},[],[1191],{"type":46,"value":1192},"install",{"type":46,"value":1194}," (route to setup; binary is built from\nsource with gRPC support enabled),\n",{"type":40,"tag":49,"props":1196,"children":1198},{"className":1197},[],[1199],{"type":46,"value":1200},"configure",{"type":46,"value":1202}," (decide remote-vs-direct, pick the external\nproxy \u002F sidecar \u002F VPN and network segment), ",{"type":40,"tag":49,"props":1204,"children":1206},{"className":1205},[],[1207],{"type":46,"value":1208},"build",{"type":46,"value":1210}," (route to install),\n",{"type":40,"tag":49,"props":1212,"children":1214},{"className":1213},[],[1215],{"type":46,"value":1216},"modify",{"type":46,"value":1218}," (refuse — modify the deployment, not the binary),\n",{"type":40,"tag":49,"props":1220,"children":1222},{"className":1221},[],[1223],{"type":46,"value":1224},"run",{"type":46,"value":1226}," (start → bind → smoke), ",{"type":40,"tag":49,"props":1228,"children":1230},{"className":1229},[],[1231],{"type":46,"value":1232},"test",{"type":46,"value":1234}," (the\nsmoke-before-bulk loop with the client-side stub\ngeneration step), ",{"type":40,"tag":49,"props":1236,"children":1238},{"className":1237},[],[1239],{"type":46,"value":1240},"debug",{"type":46,"value":1242}," (the layered diagnosis ladder),\n",{"type":40,"tag":49,"props":1244,"children":1246},{"className":1245},[],[1247],{"type":46,"value":1248},"use",{"type":46,"value":1250}," (the agent-side workflow for consuming a captured\ngRPC server session), plus a ",{"type":40,"tag":49,"props":1252,"children":1254},{"className":1253},[],[1255],{"type":46,"value":1256},"Deferred task verbs",{"type":46,"value":1258}," block\nand a ",{"type":40,"tag":49,"props":1260,"children":1262},{"className":1261},[],[1263],{"type":46,"value":1264},"Command appendix",{"type":46,"value":598},{"type":40,"tag":62,"props":1267,"children":1268},{},[1269,1271,1276],{"type":46,"value":1270},"The skill assumes a host where DOCA is already installed (or\nthe NGC DOCA container is running) with the Flow library\npresent, a working ",{"type":40,"tag":49,"props":1272,"children":1274},{"className":1273},[],[1275],{"type":46,"value":242},{"type":46,"value":1277}," application to program against,\nand the operator's awareness that exposing a gRPC control plane\nis a high-stakes posture.",{"type":40,"tag":378,"props":1279,"children":1281},{"id":1280},"what-this-skill-deliberately-does-not-ship",[1282],{"type":46,"value":1283},"What this skill deliberately does not ship",{"type":40,"tag":62,"props":1285,"children":1286},{},[1287,1289,1294],{"type":46,"value":1288},"This skill is ",{"type":40,"tag":66,"props":1290,"children":1291},{},[1292],{"type":46,"value":1293},"agent guidance",{"type":46,"value":1295},", not a samples or scripts\nbundle. To keep the boundary clean, it deliberately does not\ncontain — and pull requests should not add:",{"type":40,"tag":397,"props":1297,"children":1298},{},[1299,1316,1333,1349,1359],{"type":40,"tag":401,"props":1300,"children":1301},{},[1302,1307,1309,1314],{"type":40,"tag":66,"props":1303,"children":1304},{},[1305],{"type":46,"value":1306},"Verbatim RPC method names, message field inventories, or\ndefault endpoint paths.",{"type":46,"value":1308}," The ",{"type":40,"tag":49,"props":1310,"children":1312},{"className":1311},[],[1313],{"type":46,"value":323},{"type":46,"value":1315}," files shipped under\nthe tool's source tree on the user's install are the\nauthoritative contract; copying them here pins the skill\nto one release and silently rots when the contract\nevolves.",{"type":40,"tag":401,"props":1317,"children":1318},{},[1319,1324,1326,1331],{"type":40,"tag":66,"props":1320,"children":1321},{},[1322],{"type":46,"value":1323},"Pre-baked client code in any language.",{"type":46,"value":1325}," The\nlanguage-specific gRPC plugin + the shipped ",{"type":40,"tag":49,"props":1327,"children":1329},{"className":1328},[],[1330],{"type":46,"value":323},{"type":46,"value":1332}," files\nare the contract; client code generated from them on the\nuser's installed version is the right answer, not a stub\npinned to a snapshot.",{"type":40,"tag":401,"props":1334,"children":1335},{},[1336,1341,1343,1348],{"type":40,"tag":66,"props":1337,"children":1338},{},[1339],{"type":46,"value":1340},"A pre-baked external security-layer configuration.",{"type":46,"value":1342}," CA,\ntoken, and mTLS configuration belong to the selected proxy,\nsidecar, or VPN and its security review, never to\n",{"type":40,"tag":49,"props":1344,"children":1346},{"className":1345},[],[1347],{"type":46,"value":54},{"type":46,"value":598},{"type":40,"tag":401,"props":1350,"children":1351},{},[1352,1357],{"type":40,"tag":66,"props":1353,"children":1354},{},[1355],{"type":46,"value":1356},"Wrappers, parsers, or scripts",{"type":46,"value":1358}," that proxy the gRPC\nendpoint into another protocol. The endpoint is the\nendpoint; if a user wants HTTP\u002FJSON instead, that is a\nseparate concern outside this skill's scope.",{"type":40,"tag":401,"props":1360,"children":1361},{},[1362,1389,1391,1396],{"type":40,"tag":66,"props":1363,"children":1364},{},[1365,1367,1373,1374,1380,1381,1387],{"type":46,"value":1366},"A ",{"type":40,"tag":49,"props":1368,"children":1370},{"className":1369},[],[1371],{"type":46,"value":1372},"samples\u002F",{"type":46,"value":332},{"type":40,"tag":49,"props":1375,"children":1377},{"className":1376},[],[1378],{"type":46,"value":1379},"bindings\u002F",{"type":46,"value":339},{"type":40,"tag":49,"props":1382,"children":1384},{"className":1383},[],[1385],{"type":46,"value":1386},"reference\u002F",{"type":46,"value":1388}," subtree.",{"type":46,"value":1390},"\nEven one labeled ",{"type":40,"tag":311,"props":1392,"children":1393},{},[1394],{"type":46,"value":1395},"\"reference\"",{"type":46,"value":1397}," is misleading: operators\nwill read it as buildable.",{"type":40,"tag":378,"props":1399,"children":1401},{"id":1400},"loading-order",[1402],{"type":46,"value":1403},"Loading order",{"type":40,"tag":1405,"props":1406,"children":1407},"ol",{},[1408,1427,1447],{"type":40,"tag":401,"props":1409,"children":1410},{},[1411,1413,1418,1420,1425],{"type":46,"value":1412},"Read this ",{"type":40,"tag":49,"props":1414,"children":1416},{"className":1415},[],[1417],{"type":46,"value":909},{"type":46,"value":1419}," first to confirm the user's question\nis in scope (the user actually wants a remote gRPC control\nplane on top of ",{"type":40,"tag":49,"props":1421,"children":1423},{"className":1422},[],[1424],{"type":46,"value":242},{"type":46,"value":1426},", not a direct library link or\na different DOCA library).",{"type":40,"tag":401,"props":1428,"children":1429},{},[1430],{"type":40,"tag":66,"props":1431,"children":1432},{},[1433,1435,1440,1442,1446],{"type":46,"value":1434},"For what the server exposes, the ",{"type":40,"tag":49,"props":1436,"children":1438},{"className":1437},[],[1439],{"type":46,"value":323},{"type":46,"value":1441},"-as-contract\nrule, the language-bindings story, the external proxy \u002F\nsidecar \u002F VPN and network-segment decision, version availability, the\nlayered error surface, observability, and safety posture,\nsee ",{"type":40,"tag":246,"props":1443,"children":1444},{"href":195},[1445],{"type":46,"value":195},{"type":46,"value":598},{"type":40,"tag":401,"props":1448,"children":1449},{},[1450],{"type":40,"tag":66,"props":1451,"children":1452},{},[1453,1455,1460,1461,1466,1468,1473,1474,1479,1480,1485,1486,1491,1492,1497,1498,1503,1505,1509],{"type":46,"value":1454},"For the documented start sequence and the\nsmoke-before-bulk workflow — ",{"type":40,"tag":49,"props":1456,"children":1458},{"className":1457},[],[1459],{"type":46,"value":1192},{"type":46,"value":332},{"type":40,"tag":49,"props":1462,"children":1464},{"className":1463},[],[1465],{"type":46,"value":1200},{"type":46,"value":1467},",\n",{"type":40,"tag":49,"props":1469,"children":1471},{"className":1470},[],[1472],{"type":46,"value":1208},{"type":46,"value":332},{"type":40,"tag":49,"props":1475,"children":1477},{"className":1476},[],[1478],{"type":46,"value":1216},{"type":46,"value":332},{"type":40,"tag":49,"props":1481,"children":1483},{"className":1482},[],[1484],{"type":46,"value":1224},{"type":46,"value":332},{"type":40,"tag":49,"props":1487,"children":1489},{"className":1488},[],[1490],{"type":46,"value":1232},{"type":46,"value":332},{"type":40,"tag":49,"props":1493,"children":1495},{"className":1494},[],[1496],{"type":46,"value":1240},{"type":46,"value":332},{"type":40,"tag":49,"props":1499,"children":1501},{"className":1500},[],[1502],{"type":46,"value":1248},{"type":46,"value":1504}," — see\n",{"type":40,"tag":246,"props":1506,"children":1507},{"href":203},[1508],{"type":46,"value":203},{"type":46,"value":598},{"type":40,"tag":378,"props":1511,"children":1513},{"id":1512},"related-skills",[1514],{"type":46,"value":1515},"Related skills",{"type":40,"tag":397,"props":1517,"children":1518},{},[1519,1539,1554,1576,1603,1618,1631],{"type":40,"tag":401,"props":1520,"children":1521},{},[1522,1530,1532,1537],{"type":40,"tag":246,"props":1523,"children":1524},{"href":368},[1525],{"type":40,"tag":49,"props":1526,"children":1528},{"className":1527},[],[1529],{"type":46,"value":242},{"type":46,"value":1531}," — the ",{"type":40,"tag":66,"props":1533,"children":1534},{},[1535],{"type":46,"value":1536},"base\nlibrary",{"type":46,"value":1538}," the server's gRPC contract is a thin remote-\ncontrol wrapper over. Pipe \u002F entry \u002F rule semantics, the\nvalidate-before-commit rule, the Flow counter \u002F inspector\nsurface all live there.",{"type":40,"tag":401,"props":1540,"children":1541},{},[1542,1552],{"type":40,"tag":246,"props":1543,"children":1545},{"href":1544},"..\u002Fdoca-flow-tune\u002FSKILL.md",[1546],{"type":40,"tag":49,"props":1547,"children":1549},{"className":1548},[],[1550],{"type":46,"value":1551},"doca-flow-tune",{"type":46,"value":1553}," — the Flow\ntuning tool. When a Flow-program change is recommended,\nthe change can be applied through the surrounding\napplication or — when the control plane is remote —\nthrough this gRPC server's RPC surface.",{"type":40,"tag":401,"props":1555,"children":1556},{},[1557,1567,1569,1574],{"type":40,"tag":246,"props":1558,"children":1560},{"href":1559},"..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md",[1561],{"type":40,"tag":49,"props":1562,"children":1564},{"className":1563},[],[1565],{"type":46,"value":1566},"doca-public-knowledge-map",{"type":46,"value":1568},"\n— routing to the public DOCA Flow gRPC Server page on\n",{"type":40,"tag":49,"props":1570,"children":1572},{"className":1571},[],[1573],{"type":46,"value":793},{"type":46,"value":1575}," and the rest of the public DOCA\ndocumentation set.",{"type":40,"tag":401,"props":1577,"children":1578},{},[1579,1589,1591,1601],{"type":40,"tag":246,"props":1580,"children":1582},{"href":1581},"..\u002F..\u002Fdoca-version\u002FSKILL.md",[1583],{"type":40,"tag":49,"props":1584,"children":1586},{"className":1585},[],[1587],{"type":46,"value":1588},"doca-version",{"type":46,"value":1590}," — canonical\nversion-handling rules. The\n",{"type":40,"tag":246,"props":1592,"children":1594},{"href":1593},"CAPABILITIES.md#version-compatibility",[1595],{"type":40,"tag":49,"props":1596,"children":1598},{"className":1597},[],[1599],{"type":46,"value":1600},"## Version compatibility",{"type":46,"value":1602},"\nsection in this skill is a thin overlay on top.",{"type":40,"tag":401,"props":1604,"children":1605},{},[1606,1616],{"type":40,"tag":246,"props":1607,"children":1609},{"href":1608},"..\u002F..\u002Fdoca-debug\u002FSKILL.md",[1610],{"type":40,"tag":49,"props":1611,"children":1613},{"className":1612},[],[1614],{"type":46,"value":1615},"doca-debug",{"type":46,"value":1617}," — the cross-cutting\ndebug ladder. gRPC server failures route into the ladder at\nthe runtime layer.",{"type":40,"tag":401,"props":1619,"children":1620},{},[1621,1629],{"type":40,"tag":246,"props":1622,"children":1623},{"href":349},[1624],{"type":40,"tag":49,"props":1625,"children":1627},{"className":1626},[],[1628],{"type":46,"value":356},{"type":46,"value":1630}," — env preparation,\ninstall verification, and the NGC DOCA container path.",{"type":40,"tag":401,"props":1632,"children":1633},{},[1634,1644,1646,1652],{"type":40,"tag":246,"props":1635,"children":1637},{"href":1636},"..\u002F..\u002Fdoca-hardware-safety\u002FSKILL.md",[1638],{"type":40,"tag":49,"props":1639,"children":1641},{"className":1640},[],[1642],{"type":46,"value":1643},"doca-hardware-safety",{"type":46,"value":1645}," —\nthe cross-cutting hardware-safety meta-policy this skill's\n",{"type":40,"tag":49,"props":1647,"children":1649},{"className":1648},[],[1650],{"type":46,"value":1651},"## Safety policy",{"type":46,"value":1653}," overlays. Any state-changing RPC is a\npotential dataplane-affecting change and must respect the\nmeta-policy.",{"items":1655,"total":1757},[1656,1673,1685,1699,1711,1728,1743],{"slug":1657,"name":1657,"fn":1658,"description":1659,"org":1660,"tags":1661,"stars":20,"repoUrl":21,"updatedAt":1672},"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},[1662,1665,1668,1669],{"name":1663,"slug":1664,"type":15},"Data Analysis","data-analysis",{"name":1666,"slug":1667,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":1670,"slug":1671,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":1674,"name":1674,"fn":1675,"description":1676,"org":1677,"tags":1678,"stars":20,"repoUrl":21,"updatedAt":1684},"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},[1679,1682,1683],{"name":1680,"slug":1681,"type":15},"Deployment","deployment",{"name":18,"slug":19,"type":15},{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":1686,"name":1686,"fn":1687,"description":1688,"org":1689,"tags":1690,"stars":20,"repoUrl":21,"updatedAt":1698},"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},[1691,1694,1695],{"name":1692,"slug":1693,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":1696,"slug":1697,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":1700,"name":1700,"fn":1701,"description":1702,"org":1703,"tags":1704,"stars":20,"repoUrl":21,"updatedAt":1710},"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},[1705,1706,1707],{"name":1663,"slug":1664,"type":15},{"name":9,"slug":8,"type":15},{"name":1708,"slug":1709,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":1712,"name":1712,"fn":1713,"description":1714,"org":1715,"tags":1716,"stars":20,"repoUrl":21,"updatedAt":1727},"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},[1717,1720,1723,1724],{"name":1718,"slug":1719,"type":15},"Automation","automation",{"name":1721,"slug":1722,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":1725,"slug":1726,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":1729,"name":1729,"fn":1730,"description":1731,"org":1732,"tags":1733,"stars":20,"repoUrl":21,"updatedAt":1742},"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},[1734,1735,1738,1739],{"name":1680,"slug":1681,"type":15},{"name":1736,"slug":1737,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":1740,"slug":1741,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":1744,"name":1744,"fn":1745,"description":1746,"org":1747,"tags":1748,"stars":20,"repoUrl":21,"updatedAt":1756},"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},[1749,1750,1753],{"name":9,"slug":8,"type":15},{"name":1751,"slug":1752,"type":15},"Quantum Computing","quantum-computing",{"name":1754,"slug":1755,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":1759,"total":1910},[1760,1778,1794,1805,1817,1831,1844,1858,1869,1878,1892,1901],{"slug":1761,"name":1761,"fn":1762,"description":1763,"org":1764,"tags":1765,"stars":1775,"repoUrl":1776,"updatedAt":1777},"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},[1766,1769,1772],{"name":1767,"slug":1768,"type":15},"Documentation","documentation",{"name":1770,"slug":1771,"type":15},"MCP","mcp",{"name":1773,"slug":1774,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":1779,"name":1779,"fn":1780,"description":1781,"org":1782,"tags":1783,"stars":1791,"repoUrl":1792,"updatedAt":1793},"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},[1784,1787,1788],{"name":1785,"slug":1786,"type":15},"Containers","containers",{"name":1680,"slug":1681,"type":15},{"name":1789,"slug":1790,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":1795,"name":1795,"fn":1796,"description":1797,"org":1798,"tags":1799,"stars":1791,"repoUrl":1792,"updatedAt":1804},"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},[1800,1803],{"name":1801,"slug":1802,"type":15},"CI\u002FCD","ci-cd",{"name":1680,"slug":1681,"type":15},"2026-07-14T05:25:59.97109",{"slug":1806,"name":1806,"fn":1807,"description":1808,"org":1809,"tags":1810,"stars":1791,"repoUrl":1792,"updatedAt":1816},"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},[1811,1812,1813],{"name":1801,"slug":1802,"type":15},{"name":1680,"slug":1681,"type":15},{"name":1814,"slug":1815,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":1818,"name":1818,"fn":1819,"description":1820,"org":1821,"tags":1822,"stars":1791,"repoUrl":1792,"updatedAt":1830},"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},[1823,1826,1827],{"name":1824,"slug":1825,"type":15},"Debugging","debugging",{"name":1814,"slug":1815,"type":15},{"name":1828,"slug":1829,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":1832,"name":1832,"fn":1833,"description":1834,"org":1835,"tags":1836,"stars":1791,"repoUrl":1792,"updatedAt":1843},"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},[1837,1840],{"name":1838,"slug":1839,"type":15},"Best Practices","best-practices",{"name":1841,"slug":1842,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":1845,"name":1845,"fn":1846,"description":1847,"org":1848,"tags":1849,"stars":1791,"repoUrl":1792,"updatedAt":1857},"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},[1850,1853,1856],{"name":1851,"slug":1852,"type":15},"Machine Learning","machine-learning",{"name":1854,"slug":1855,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":1859,"name":1859,"fn":1860,"description":1861,"org":1862,"tags":1863,"stars":1791,"repoUrl":1792,"updatedAt":1868},"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},[1864,1867],{"name":1865,"slug":1866,"type":15},"QA","qa",{"name":1708,"slug":1709,"type":15},"2026-07-14T05:25:53.673039",{"slug":1870,"name":1870,"fn":1871,"description":1872,"org":1873,"tags":1874,"stars":1791,"repoUrl":1792,"updatedAt":1877},"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},[1875,1876],{"name":1680,"slug":1681,"type":15},{"name":18,"slug":19,"type":15},"2026-07-14T05:25:49.362534",{"slug":1879,"name":1879,"fn":1880,"description":1881,"org":1882,"tags":1883,"stars":1791,"repoUrl":1792,"updatedAt":1891},"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},[1884,1887,1888],{"name":1885,"slug":1886,"type":15},"Code Review","code-review",{"name":1814,"slug":1815,"type":15},{"name":1889,"slug":1890,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":1893,"name":1893,"fn":1894,"description":1895,"org":1896,"tags":1897,"stars":1791,"repoUrl":1792,"updatedAt":1900},"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},[1898,1899],{"name":1865,"slug":1866,"type":15},{"name":1708,"slug":1709,"type":15},"2026-07-14T05:25:54.928983",{"slug":1902,"name":1902,"fn":1903,"description":1904,"org":1905,"tags":1906,"stars":1791,"repoUrl":1792,"updatedAt":1909},"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},[1907,1908],{"name":1718,"slug":1719,"type":15},{"name":1801,"slug":1802,"type":15},"2026-07-30T05:29:03.275638",496]