Skip to content

Enhance README for device mounts and validation in DLStreamer skill#789

Open
wdunia wants to merge 6 commits into
open-edge-platform:mainfrom
wdunia:coding-agent-improvements
Open

Enhance README for device mounts and validation in DLStreamer skill#789
wdunia wants to merge 6 commits into
open-edge-platform:mainfrom
wdunia:coding-agent-improvements

Conversation

@wdunia

@wdunia wdunia commented Apr 17, 2026

Copy link
Copy Markdown

Description

Fix ambiguous device mounting instructions in DL Streamer Coding Agent SKILL.md that caused the agent to generate incorrect docker run commands — either hard-coding --device /dev/accel on hosts without NPU (causing runtime errors) or omitting it entirely on hosts with NPU (missing hardware acceleration).

Changes (Step 5 — Run, Debug and Validate):

  • The docker run template in the README always includes both /dev/dri (GPU) and /dev/accel (NPU) mounts, plus a user-facing note: "Remove --device and --group-add lines for devices not present on your system."
  • Before validation, the agent probes host devices (ls /dev/dri/render* /dev/accel/accel*), removes --device/--group-add flags for missing devices, and selects the best inference device using priority NPU → GPU → CPU.
  • The README stays generic and is not rewritten to match the validation host.

This ensures:

  • The generated README works on any hardware (NPU+GPU, GPU-only, CPU-only) without manual editing beyond removing inapplicable lines.
  • The agent's validation run never fails due to mounting a non-existent device.
  • Inference always targets the best available XPU on the host.

Fixes # (issue)

Any Newly Introduced Dependencies

None.

How Has This Been Tested?

Validated on this branch by running the DLStreamer Coding Agent with the updated SKILL.md

Checklist:

  • I agree to use the MIT license for my code changes.
  • I have not introduced any 3rd party components incompatible with MIT.
  • I have not included any company confidential information, trade secret, password or security token.
  • I have performed a self-review of my code.

wdunia added 3 commits April 14, 2026 13:19
Strengthen container run guidance in the DLStreamer skill README: always mount both /dev/dri and /dev/accel, and ensure generated README docker run examples include both device mounts when a pipeline may use an NPU. Clarifies that a GPU-only fallback should be presented only as a secondary note for systems without NPU.
Add a pre-run verification in Step 7 to check for required host devices. Includes an example `ls /dev/dri/render* /dev/accel/accel*` command and guidance that if `/dev/accel` is missing the host has no NPU — switch to GPU for validation only and do not change the README's `docker run` command.
Refine instructions around mounting /dev/dri and /dev/accel and add guidance for systems missing XPU devices. Adds a README note to remove the corresponding --device and --group-add lines when a device is not present, and changes the host-check step to explicitly list XPU devices and instruct building the docker run command by removing missing device/group-add entries. Emphasizes choosing inference device priority NPU → GPU → CPU and keeping the generated README generic with all devices listed.
wdunia added 2 commits April 24, 2026 11:04
Update SKILL.md to instruct users to verify available XPU devices on the host before running the container (ls /dev/dri/render* /dev/accel/accel*). Advise removing --device and --group-add lines for missing devices when building the docker run command, and recommend selecting inference device priority NPU → GPU → CPU. Preserve README wording (keep all devices listed with a note to remove absent device lines).
Comment thread .github/skills/dlstreamer-coding-agent/SKILL.md Outdated
Update .github/skills/dlstreamer-coding-agent/SKILL.md to change the inference device selection priority from NPU → GPU → CPU to GPU → NPU → CPU. This makes GPUs the preferred device when available; the README wording remains generic and still instructs to remove device/group-add lines for devices not present on the host.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants