Endless loop during startup

[mirek967]

version 4.10.0

Environment, CPU architecture, OS, and Version:
linux, amd64, gentoo.
Linux host 6.18.52-gentoo #1 SMP PREEMPT Thu Sep 17 13:05:37 CEST 2026 x86_64 Intel(R) Core(TM) i9-14900HX GenuineIntel GNU/Linux

Describe the bug
when try to start local-ai, it blocks with endless loop at call of intel_gpu_top.
intel_gpu_top is called with -s 1 parameter, which means endless checking every 1ms.
Which I think is not intended.

To Reproduce
get new installation, and just start.

Expected behavior
web interface would start

Logs
Sep 22 15:14:30 DEBUG GPU vendor detected via ghw vendor="nvidia" caller={caller.file="/run/portage/portage/sci-ml/local-ai-4.10.0-r1/work/LocalAI-4.10.0/pkg/xsysinfo/gpu.go" caller.L=240 }
Sep 22 15:14:30 DEBUG GPU vendor gpuVendor="nvidia" caller={caller.file="/run/portage/portage/sci-ml/local-ai-4.10.0-r1/work/LocalAI-4.10.0/pkg/system/state.go" caller.L=87 }

Additional context

[mirek967]

problem occur in pkg/xsysinfo/gpu.go
at line 831
cmd := exec.Command("intel_gpu_top", "-J", "-s", "1")

[mirek967]

I have one more observation.
Removing of intel_gpu_top from the system solved problem. Local-ai starts normally in such a case.

[leilei3167]

Hi — I can reproduce the hang from intel_gpu_top -J -s 1 blocking startup on machines with Intel GPU tools installed. I’m looking into a one-shot sample (or timeout) in pkg/xsysinfo/gpu.go and will open a PR shortly if that checks out.