NVIDIA/fleet-intelligence-agent v1.3.0

v1.3.0

Repository: NVIDIA/fleet-intelligence-agent

Tag: v1.3.0

Published: 2026-06-04T18:56:12Z

Prerelease: no

Release notes:

NVIDIA Fleet Intelligence Agent v1.3.0

This release includes a host agent for GPU telemetry collection and attestation.

Available Packages

• DEB packages: amd64 and arm64 for Debian/Ubuntu systems
• RPM packages: x86_64 and aarch64 for RHEL/CentOS/Fedora systems
• Binary archives: Standalone binaries for both architectures

Installation

Download the appropriate package for your system from the assets below. For installation instructions, see the project README.

Verification

All release artifacts are checksummed. Verify downloads using:

sha256sum -c checksums.txt

Support

For issues and questions, please visit: https://github.com/NVIDIA/fleet-intelligence-agent/issues

What's Changed

• security: parameterize metrics component filter to prevent SQL injection by @rsampaio in https://github.com/NVIDIA/fleet-intelligence-agent/pull/178
• fix: change the helm default node selector to be gpu.deploy.dcgm by @jingxiang-z in https://github.com/NVIDIA/fleet-intelligence-agent/pull/181
• feat: add MALLOC_ARENA_MAX to limit the dcgm mem by @ambermingxin in https://github.com/NVIDIA/fleet-intelligence-agent/pull/182
• feat: add loop and retry for dcgm detection by @ambermingxin in https://github.com/NVIDIA/fleet-intelligence-agent/pull/180
• feat: add backend client and inventory attestation call by @jingxiang-z in https://github.com/NVIDIA/fleet-intelligence-agent/pull/176
• refactor: simplify inventory and attestation loops by @jingxiang-z in https://github.com/NVIDIA/fleet-intelligence-agent/pull/184
• fix: use canonical reflect pointer kind by @jingxiang-z in https://github.com/NVIDIA/fleet-intelligence-agent/pull/186
• feat: add agent enroll time in inventory loop by @ambermingxin in https://github.com/NVIDIA/fleet-intelligence-agent/pull/188
• Clarify the documentation on dependencies. by @ebalduf in https://github.com/NVIDIA/fleet-intelligence-agent/pull/190
• feat: create nvml component by @ambermingxin in https://github.com/NVIDIA/fleet-intelligence-agent/pull/194
• feat: add support for ada lovelace and ampere GPU architecture by @ambermingxin in https://github.com/NVIDIA/fleet-intelligence-agent/pull/197
• feat: add validation for agent values sending to backend request by @ambermingxin in https://github.com/NVIDIA/fleet-intelligence-agent/pull/196
• Remove unused SDK version detection by @rsampaio in https://github.com/NVIDIA/fleet-intelligence-agent/pull/192
• fix: restore gpu info gpus OTLP attribute by @jingxiang-z in https://github.com/NVIDIA/fleet-intelligence-agent/pull/201
• feat: set nodegroup and computezone on agent by @ambermingxin in https://github.com/NVIDIA/fleet-intelligence-agent/pull/200
• fix: context deadline exceeded shouldn't throw an error by @ambermingxin in https://github.com/NVIDIA/fleet-intelligence-agent/pull/203
• chore: disable accelerator-nvidia-processes by default (#209) by @ambermingxin in https://github.com/NVIDIA/fleet-intelligence-agent/pull/210
• fix: allow user to only set one field for ng or cz (#211) by @jingxiang-z in https://github.com/NVIDIA/fleet-intelligence-agent/pull/213

New Contributors

• @ebalduf made their first contribution in https://github.com/NVIDIA/fleet-intelligence-agent/pull/190

Full Changelog: https://github.com/NVIDIA/fleet-intelligence-agent/compare/v1.2.1...v1.3.0