fleet-intelligence-client

Fleet Intelligence Client

github.com/NVIDIA/fleet-intelligence-client

Go SDK and nvfleetctl CLI for the NVIDIA Fleet Intelligence customer API. Use it to inspect GPU fleet health, inventory, alerts, and reports from a terminal or Go program.

What is included

• nvfleetctl: CLI for authentication, fleet inspection, alerts, and reports.
• pkg/fleetintelligence: public Go SDK over the generated OpenAPI client.
• api/openapi: public customer API contract used by the SDK and CLI.

Requirements

• Go 1.23+ for source builds and go install.
• An NGC service key from https://org.ngc.nvidia.com/identity-access/service-keys. See the Fleet Intelligence API reference for key-generation instructions.
• Linux, macOS, or Windows on a Go-supported amd64/arm64 platform.

Install

Use a prebuilt binary

Download the nvfleetctl archive for your platform from the release artifacts.

Linux and macOS

Choose the matching OS and architecture:

• Linux amd64: nvfleetctl_<version>_linux_amd64.tar.gz
• Linux arm64: nvfleetctl_<version>_linux_arm64.tar.gz
• macOS Intel: nvfleetctl_<version>_darwin_amd64.tar.gz
• macOS Apple Silicon: nvfleetctl_<version>_darwin_arm64.tar.gz

Then extract and install:

tar -xzf nvfleetctl_<version>_<os>_<arch>.tar.gz
chmod +x nvfleetctl
sudo mv nvfleetctl /usr/local/bin/nvfleetctl
nvfleetctl version

On macOS, Gatekeeper quarantines binaries downloaded outside the App Store. If macOS blocks the binary, clear the quarantine attribute (adjust the path to wherever you installed it):

xattr -d com.apple.quarantine /usr/local/bin/nvfleetctl

If you do not have permission to write to /usr/local/bin, install to a user-local directory:

mkdir -p "$HOME/.local/bin"
mv nvfleetctl "$HOME/.local/bin/nvfleetctl"

Make sure $HOME/.local/bin is on your PATH, then verify the install:

nvfleetctl version

Windows

Download the matching Windows zip:

• Windows amd64: nvfleetctl_<version>_windows_amd64.zip
• Windows arm64: nvfleetctl_<version>_windows_arm64.zip

Unzip the archive, then add the directory containing nvfleetctl.exe to your PATH. Verify the install:

nvfleetctl version

Install with Go

Use Go installer:

# While this repository is private, set GOPRIVATE first so `go install`
# fetches via git instead of the public proxy:
export GOPRIVATE=github.com/NVIDIA/fleet-intelligence-client
go install github.com/NVIDIA/fleet-intelligence-client/cmd/nvfleetctl@latest

Build from source

git clone https://github.com/NVIDIA/fleet-intelligence-client.git
cd fleet-intelligence-client
make build
./bin/nvfleetctl --help

Quick Start

nvfleetctl auth login --key <your-ngc-service-key>
nvfleetctl node list
nvfleetctl alert list
nvfleetctl report inventory

Common output and pagination flags:

nvfleetctl node list --all --output json
nvfleetctl node list --page 0 --page-size 25
nvfleetctl node list --health Degraded,Unhealthy

Run nvfleetctl <command> --help for command-specific flags.

Use with AI agents (Claude Code and Codex)

This repo ships an Agent Skill that teaches a coding agent how to drive nvfleetctl to answer fleet questions in plain language ("how many nodes are unhealthy?", "any critical alerts?"). The skill is a single portable SKILL.md file, so it works in any agent that supports the Agent Skills format.

The agent still calls the nvfleetctl binary, so install and authenticate it first.

Download skills/nvfleetctl/SKILL.md from this repo and place it in a nvfleetctl/ folder under your agent's skills directory:

• Claude Code: ~/.claude/skills/nvfleetctl/SKILL.md (or .claude/skills/nvfleetctl/SKILL.md to scope it to a single project)
• Codex: ~/.codex/skills/nvfleetctl/SKILL.md

The agent discovers the skill automatically — just ask a fleet question and it invokes nvfleetctl.

Documentation

• Examples
• Architecture
• OpenAPI contract
• Changelog
• Contributing
• Security policy

Development

make build
make test
make lint
make check

Commit subjects follow conventional commits: <type>(<scope>): <subject>. Contributions require DCO sign-off; see CONTRIBUTING.md.

Support

This client is experimental and under active development. Open a GitHub issue for bugs, feature requests, or questions. Do not file public issues for security vulnerabilities; follow SECURITY.md.

License

Apache License 2.0. See LICENSE.