Frigate 0.18 (Full Access Beta) - OpenVINO GPU Context Error on Intel N5030 Mini PC

Configuration
1

Hi everyone,

I am trying to configure face recognition in Frigate v0.18-0 on Home Assistant OS (HAOS), running on a NiPoGi AK1PLUS Mini PC with an Intel Pentium Silver N5030 processor (Gemini Lake Refresh / Intel UHD Graphics 605 - Gen9).

I want to offload the OpenVINO detector to the integrated GPU (iGPU) instead of using the CPU, but Frigate fails to start when using device: GPU.

Here is the exact error message I am getting in the Frigate logs:

RuntimeError: Exception from src/inference/src/cpp/core.cpp:134:
Check 'contexts.count(device_id)' failed at src/plugins/intel_gpu/src/plugin/plugin.cpp:206:
[GPU] Context was not initialized for 0 device

frigate.util.services: Unable to collect Intel GPU stats: no Intel GPU (i915/xe) found in /sys/class/drm. Check that the driver is loaded on the host

My Environment & Setup:

  • Hardware: Mini PC Intel N5030 (Gen9 iGPU)

  • OS: Home Assistant OS (HAOS)

  • Add-on version: Frigate (Full Access) Beta v0.18.0-beta3

  • Protection Mode: Disabled (Turned off in the add-on settings)

  • Hardware Access: Enabled (The toggle is active on the Full Access add-on info tab)

Since I am using the Full Access Beta add-on, the traditional “Configuration” tab with the visual Devicessection is hidden/not accessible. The add-on should theoretically map /dev/dri automatically with Protection Mode off, but it seems OpenVINO cannot detect or initialize the GPU context inside the container.

My frigate.yml relevant snippet:

yaml

environment_vars:
  LIBVA_DRIVER_NAME: iHD
  MESA_VA_DRIVER_NAME: iris

detectors:
  ov:
    type: openvino
    device: GPU

face_recognition:
  enabled: true
  model_size: small

How can I properly expose the Intel iGPU /dev/dri device to the Frigate Full Access add-on in HAOS so OpenVINO can initialize the GPU context?

Any help or insights would be greatly appreciated! Thank you.