Official NVIDIA-authored guidance for NVIDIA cuDF GPU DataFrames, pandas acceleration, dask-cuDF, ETL, joins, groupby, CSV/Parquet I/O, nullable semantics, and multi-GPU DataFrame workloads.
日本語の概要は準備中です。原文の説明を表示しています。
Use this skill when choosing or running NeMo Relay installation for the CLI, Python, Node.js, Rust, OpenClaw, Hermes, or maintained framework integrations before runtime configuration or quick-start setup.
インストールする前に、エージェントに与えられる指示の中身を確認できます。
Choose the package or executable from the user's desired outcome. Stop after
installation and basic availability checks. Do not configure runtime behavior,
write plugins.toml, create scopes, register middleware, or build a first app
example from this skill.
Keep installation separate from first-use instrumentation.
If the user asks to install NeMo Relay but does not identify a target or desired outcome, ask one short clarifying question before giving commands:
Which install path do you want: CLI for coding-agent/local gateway use, language package for a Python/Node.js/Rust app, or framework integration for LangChain, LangGraph, Deep Agents, OpenClaw, or Hermes?
Do not ask when the user already names a CLI, language, framework, harness,
source checkout, or target project file such as pyproject.toml, package.json,
or Cargo.toml.
Use this order from least to most application-specific:
For "try Relay" requests, default to the CLI and temporary transparent run. Do not make persistent host-plugin installation the default.
Before persistent Codex installation, determine whether the user is operating
from Codex Desktop. Persistent nemo-relay install codex changes the active
Codex provider and can make the current or older Desktop threads appear missing
after restart because of an upstream provider-filtering bug. The threads are
not deleted.
For Codex Desktop users:
assets/codex-desktop-recovery.md as
NEMO_RELAY_CODEX_DESKTOP_RECOVERY.md in the user's workspace root before
running the persistent installer.Do not directly inspect, copy, delete, edit, or rewrite Codex session files,
private application configuration, or SQLite state to work around the
visibility bug. Supported nemo-relay install, uninstall, and doctor commands
may manage the Relay-generated provider and hook configuration.
The primary documented language paths are Rust, Python, and Node.js. Treat Go and raw FFI as source-first advanced surfaces, not normal package installs. Do not treat a first scope, subscriber, gateway, plugin config, or LLM call as installation verification.
Know about nemo-relay doctor, but use it in the right scope:
nemo-relay doctor when the user installed the CLI and reports config,
gateway, agent-readiness, plugin, exporter, or model-pricing issues.nemo-relay doctor --json when structured output will help an agent
inspect checks programmatically.nemo-relay doctor --plugin claude-code, nemo-relay doctor --plugin codex, or nemo-relay doctor --plugin all only for persistent host-plugin
installations.When doctor reports failures, summarize the failed checks and the specific remediation it suggests. Do not loop back to reinstalling every package unless the failed check points to a broken or missing install.
Choose the next workflow from the user's immediate outcome:
nemo-relay-get-started for a first working scope, tool call, LLM call,
or trial plugin setup.Avoid these installation-scope mistakes:
nemo-relay executable.plugins.toml, middleware registration, scopes, or quick-start
examples before the install step has been verified.Use these public entry points to confirm current installation guidance:
まだレビューはありません。使ってみた感想をお寄せください。
概要と使いどころ
Official NVIDIA-authored guidance for NVIDIA cuDF GPU DataFrames, pandas acceleration, dask-cuDF, ETL, joins, groupby, CSV/Parquet I/O, nullable semantics, and multi-GPU DataFrame workloads.
日本語の概要は準備中です。原文の説明を表示しています。
Customize NVIDIA Nemotron Voice Agent's Generic Pipecat example for healthcare appointment, five-field patient intake, or custom tool-calling workflows without a separate backend.
日本語の概要は準備中です。原文の説明を表示しています。
Calibrate a new dataset from live RTSP camera streams via the AutoMagicCalib REST API. Use when the user provides RTSP URLs or asks to calibrate live cameras; VIOS records clips, AMC ingests them, then runs calibration.
日本語の概要は準備中です。原文の説明を表示しています。
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'.
日本語の概要は準備中です。原文の説明を表示しています。
Calibrates pre-recorded `cam_*.mp4` datasets through the AutoMagicCalib REST API. Use for user-supplied local MP4s; route live RTSP streams to `amc-run-rtsp-calibration`.
日本語の概要は準備中です。原文の説明を表示しています。
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.
日本語の概要は準備中です。原文の説明を表示しています。