Release v1.3.0 · NVIDIA/fleet-intelligence-agent

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

NVIDIA Fleet Intelligence Agent v1.3.0

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

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

All release artifacts are checksummed. Verify downloads using:

sha256sum -c checksums.txt

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

Full Changelog: v1.2.1...v1.3.0