---

name: test description: | Runtime verification tests for bazzite-ai installation. Tests GPU detection, CUDA, PyTorch, service health, network connectivity, and pod lifecycles. Use when users need to verify their bazzite-ai installation works correctly.

Test - Runtime Verification

Overview

The test command provides comprehensive runtime verification for bazzite-ai installations. It tests GPU detection, CUDA/PyTorch functionality, service health, network connectivity, and pod container lifecycles.

Key Concept: Tests run on the LOCAL system to verify actual functionality, not just syntax.

Quick Reference

Parameters

ACTION Parameter

Copyujust test ACTION=""

Without ACTION, shows interactive menu (requires TTY).

Test Options

Copyujust test [ACTION] [--instance=N] [--image=IMAGE] [--cpus=N] [--ram=MB] [--ssh-port=PORT]

Available Tests

Quick Tests

Copyujust test quick      # GPU + service status (~30s)
ujust test status     # Test status summary

GPU Tests

Copyujust test gpu        # GPU detection and CDI check
ujust test cuda       # CUDA tests in nvidia container
ujust test pytorch    # PyTorch GPU access test

Service Tests

Copyujust test ollama     # Ollama health + quick inference
ujust test jupyter    # Jupyter service health
ujust test comfyui    # ComfyUI service health
ujust test openwebui  # Open WebUI service health
ujust test services   # All installed services status

Infrastructure Tests

Copyujust test config     # Configuration dispatcher test
ujust test network    # Registry connectivity test
ujust test apptainer  # Apptainer GPU detection

VM Tests

Copyujust test bootc                    # Ephemeral bootc VM (auto-cleanup)
ujust test bootc --image=stable     # Test specific image

Pod Lifecycle Tests

Copyujust test pods config --instance=91   # Configure test pods
ujust test pods start --instance=91    # Start test pods
ujust test pods status --instance=91   # Check test pods status
ujust test pods stop --instance=91     # Stop test pods
ujust test pods delete --instance=91   # Delete test pod configs
ujust test pods all --instance=91      # Full lifecycle test

Common Workflows

Verify New Installation

Copy# Quick verification
ujust test quick

# If issues found, run full suite
ujust test all

Verify GPU Support

Copy# Check GPU detection
ujust test gpu

# Test CUDA if NVIDIA
ujust test cuda

# Test PyTorch GPU access
ujust test pytorch

Verify Services

Copy# Test all services
ujust test services

# Or individual services
ujust test ollama
ujust test jupyter

Test Pod Lifecycle

Copy# Full lifecycle test with isolated instance
ujust test pods all --instance=91

Non-Interactive Usage

All tests work without TTY:

Copy# CI/automation-friendly
ujust test quick
ujust test gpu
ujust test all

Troubleshooting

GPU Not Detected

Symptom: ujust test gpu shows no GPU

Cause: GPU drivers not loaded or CDI not configured

Fix:

Copy# Check NVIDIA driver
nvidia-smi

# Check CDI
ls /etc/cdi/

# Setup GPU container support
ujust config gpu setup

CUDA Test Fails

Symptom: ujust test cuda fails

Cause: NVIDIA container toolkit not configured

Fix:

Copyujust config gpu setup
# May require reboot

Service Test Fails

Symptom: Service test shows unhealthy

Cause: Service not running or misconfigured

Fix:

Copy# Check specific service
ujust <service> status

# View logs
ujust <service> logs

# Restart service
ujust <service> restart

Cross-References

• GPU Setup: config skill (ujust config gpu setup)
• Service Management: Individual service skills (ollama, jupyter, etc.)
• Pod Management: pods skill for aggregate operations

When to Use This Skill

Use when the user asks about:

• "verify installation", "test bazzite-ai", "check if working"
• "GPU test", "CUDA test", "PyTorch test"
• "service health", "check services"
• "network connectivity", "registry access"
• "test pods", "lifecycle test"