Multi-Agent Operations War Room

A multi-agent AI system where four specialized agents collaborate to detect, diagnose, and solve operational problems in real-time.

What It Does

This system simulates an Operations War Room where AI agents work together like a team of specialists:

Agent	Role	What It Does
Coordinator	Session Leader	Orchestrates the session, prioritizes issues, synthesizes findings
Monitor	Early Warning	Scans KPIs, calculates health scores, detects anomalies
Analyst	Detective	Deep-dives into anomalies, examines patterns, finds root causes
Strategist	Advisor	Proposes solutions with cost/benefit estimates

Key Features

• Multi-agent orchestration: Coordinator routes tasks to specialized agents
• Statistical anomaly detection: Compares metrics against baseline using z-score approach
• Pattern-based root cause analysis: Matches data patterns to known problem categories
• Actionable recommendations: 3-tier solutions (immediate / short-term / long-term)
• Professional PDF report: Charts, findings, recommendations, and full agent conversation log
• Two modes: Rule-based (offline) and AI-powered (Google Gemini)

Quick Start

# Install dependencies
pip install -r requirements.txt

# Generate data with anomalies
python data_generator.py

# Run the war room
python main.py

Check results in:

• reports/war_room_report.pdf (full report)
• charts/ folder (4 PNG charts)

Anomaly Scenarios

The data generator injects realistic problems that the agents must detect:

# See all scenarios
python data_generator.py --list

# Default: mixed anomalies across different months
python data_generator.py --scenario mixed

# Staffing shortage across departments
python data_generator.py --scenario staffing-crisis

# Supply chain quality issues cascade
python data_generator.py --scenario quality-crisis

# Progressive IT system degradation
python data_generator.py --scenario system-failure

AI-Powered Mode

Get a free API key at https://aistudio.google.com/apikey, then:

export GEMINI_API_KEY=your-key-here
python main.py --with-ai

AI mode adds natural language insights on top of the data analysis. The system works fully without AI using rule-based logic.

How the War Room Session Works

1. COORDINATOR: "Monitor Agent, begin your scan."
2. MONITOR:     Scans 8,400 records, finds 3 anomalies
3. COORDINATOR: "Analyst, investigate the critical anomaly."
4. ANALYST:     Deep-dives into data, finds root cause
5. COORDINATOR: "Strategist, propose solutions."
6. STRATEGIST:  Recommends immediate, short-term, long-term actions
7. COORDINATOR: Synthesizes executive summary with action plan

Project Structure

multi-agent-operations/
├── main.py               # Entry point - runs the war room session
├── agents.py             # 4 agent definitions (Coordinator, Monitor, Analyst, Strategist)
├── kpi_engine.py         # KPI calculations and anomaly detection
├── data_generator.py     # Generate data with intentional anomalies
├── report_generator.py   # PDF report and chart generation
├── requirements.txt      # Python dependencies
├── data/                 # Generated data files
├── charts/               # Generated charts (4 PNG files)
└── reports/              # Generated PDF report

Technologies Used

• Python - Core language
• pandas / numpy - Data analysis and statistical calculations
• matplotlib / seaborn - Chart generation
• fpdf2 - PDF report generation
• scikit-learn - (available for future ML extensions)
• Google Gemini API - Optional AI-powered insights (free tier)

Author

Leo - AI Engineering Portfolio Project