Jetson Thor: Failed to Install Nemoclaw during OnBoard

[raymondlo84]

Description

Tried onboarding and got this error.

        throw new Error(`GPU proof failed: ${proof.label} (status ${statusText})${diagnosticSuffix}`);
              ^

Error: GPU proof failed: nvidia-smi (status 127): /bin/bash: line 1: nvidia-smi: command not found
    at verifyDirectSandboxGpu (/home/nvidia/.nemoclaw/source/dist/lib/onboard.js:1435:15)
    at createSandbox (/home/nvidia/.nemoclaw/source/dist/lib/onboard.js:4738:13)
    at process.processTicksAndRejections (node:internal/process/task_queues:103:5)
    at async Object.onboard [as runOnboard] (/home/nvidia/.nemoclaw/source/dist/lib/onboard.js:8823:27)
    at async runOnboardCommand (/home/nvidia/.nemoclaw/source/dist/lib/onboard/legacy-command.js:207:5)
    at async runOnboardAction (/home/nvidia/.nemoclaw/source/dist/lib/actions/onboard.js:26:5)
    at async runOnboardAction (/home/nvidia/.nemoclaw/source/dist/lib/actions/global.js:28:5)
    at async OnboardCliCommand.run (/home/nvidia/.nemoclaw/source/dist/lib/commands/onboard.js:18:9)
    at async OnboardCliCommand._run (/home/nvidia/.nemoclaw/source/node_modules/@oclif/core/lib/command.js:182:22)
    at async Config.runCommand (/home/nvidia/.nemoclaw/source/node_modules/@oclif/core/lib/config/config.js:445:25)

Try again and then it got stuck here.

:(

nvidia@nvidia-thor-raymond:~/llamacpp_serving$ nemoclaw onboard

NemoHermes Onboarding

[1/8] Preflight checks
──────────────────────────────────────────────────
✓ Docker is running
✓ Container DNS resolution works
✓ Container runtime: docker
✓ Container runtime resources: 14 vCPU / 122.8 GiB
✓ openshell CLI: openshell 0.0.39
✓ Port 8080 already owned by healthy NemoHermes runtime (OpenShell gateway)
✓ NVIDIA GPU detected (NVIDIA Thor, 125771 MB)
✓ Docker CDI GPU support detected (/etc/cdi/nvidia.yaml)
✓ Sandbox GPU: enabled (auto)
✓ Memory OK: 125771 MB RAM + 0 MB swap
NVIDIA GPU detected; enabling OpenShell GPU passthrough. Use --no-gpu to opt out.
Existing gateway was started without GPU passthrough.
To enable GPU, destroy the existing sandbox and gateway, then re-onboard:
nemoclaw destroy --yes && nemoclaw onboard --gpu

Reproduction Steps

Run nemohermes onboard, and it stopped

Environment

Jetson Thor

nvidia@nvidia-thor-raymond:~/llamacpp_serving$ uname -ar
Linux nvidia-thor-raymond 6.8.12-tegra #1 SMP PREEMPT Tue Dec 30 15:40:41 PST 2025 aarch64 aarch64 aarch64 GNU/Linux

Debug Output

Logs

Checklist

• I confirmed this bug is reproducible
• I searched existing issues and this is not a duplicate

[raymondlo84]

the workaround is use --no-gpu for onboarding for now.