AIOps - Anomaly Analysis with Qwen3:8b

CPU and memory anomaly detection for local workloads using the Qwen3:8b LLM.

Requirements

• Python 3.8+
• Ollama (installed and running)
• Qwen3:8b model available in Ollama

Installation

1. Install Ollama

curl https://ollama.ai/install.sh | sh

2. Download the Qwen3:8b model

ollama pull qwen3:8b

3. Run the setup script

The start.sh script automatically:

• Verifies prerequisites
• Creates a virtual environment (venv)
• Installs Python dependencies
• Launches the app

./start.sh

Running

Option 1: Quick start (recommended)

./start.sh

The script will:

• Validate requirements
• Create the venv if needed
• Install dependencies
• Show the launcher menu

Option 2: Manual run

# Activate the venv
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Launch
python quick_analysis.py

Option 3: One-off quick analysis

./start.sh
# Choose option 1

Option 4: Continuous monitoring (60 seconds)

./start.sh
# Choose option 2

How It Works

1. Metric collection – Gather CPU, memory, and process info via psutil.
2. Anomaly detection – Flag readings when CPU > 80% or memory > 85%.
3. LLM analysis – Send metrics to the Qwen3:8b model for context.
4. Explanation – The model describes causes and recommended actions.

Customization

Adjust anomaly thresholds

analyzer.monitor_continuous(duration_seconds=120, interval_seconds=5)
# and
analyzer.is_anomaly(metrics, cpu_threshold=90, mem_threshold=90)

Switch models

analyzer = AnomalyAnalyzer(model="mistral")  # or another Ollama model

Project Structure

aiops/
├── anomaly_analyzer.py    # Continuous monitoring script
├── quick_analysis.py      # Single-measurement analysis
├── requirements.txt       # Python dependencies
└── README.md              # This file

Troubleshooting

"Cannot connect to Ollama"

ollama run qwen3:8b

Model not found

ollama pull qwen3:8b
ollama list

Slow performance

• Lower duration_seconds or increase interval_seconds.
• Qwen3:8b should run smoothly on most modern machines.

Using the Ollama API Directly

curl http://localhost:11434/api/generate -d '{
  "model": "qwen3:8b",
  "prompt": "What is a CPU anomaly?"
}'

Notes

• Qwen3:8b offers improved language understanding and reasoning.
• Everything runs locally; nothing is sent outside your environment.
• Useful for AIOps and DevOps teams needing quick diagnostics.