🔍 AI-Powered Root Cause Analysis System

Automatically detects application errors and generates AI-driven root cause analysis — no manual log reading, no API keys, no data leaving your machine.

💡 Why I built this

On-call engineers waste hours manually grepping logs during incidents. I wanted to explore whether a local LLM could replace that entire workflow — ingest live logs, detect anomalies, and return a human-readable root cause + fix in seconds. This project is that proof of concept, built entirely with open-source tooling and zero cloud dependencies.

📸 Demo

🎬 Live Dashboard

🤖 AI Root Cause Analysis API

✅ Health Check

🔔 Prometheus Alerts

🧠 Sample RCA output

{
  "root_cause": "ZeroDivisionError in /cause-error endpoint",
  "explanation": "The Flask route triggers division by zero, causing an unhandled 500 response.",
  "fix": "Add input validation — if divisor == 0, return a 400 before performing the division.",
  "confidence": "high"
}

🏗️ Architecture

Flask App ──► Promtail ──► Loki ──► RCA Service ──► Phi-4-mini (Ollama)
          ──► Prometheus ──► Grafana Dashboard        └──► REST API (/rca)

Why this stack?

• Loki over ELK — lightweight, Docker-friendly, no JVM overhead
• Phi-4-mini via Ollama — runs on a laptop GPU, zero data leakage, no API costs
• Prometheus + Grafana — industry-standard observability used by every major SRE team

🛠️ Tech Stack

Layer	Technology
Application	Flask (Python)
Log Collection	Promtail → Loki
Metrics	Prometheus
Visualization	Grafana
AI/LLM	Microsoft Phi-4-mini via Ollama
Infrastructure	Docker Compose

✨ Key Features

• Automated RCA — AI analyzes logs and returns root cause + fix in seconds
• Local LLM — Phi-4-mini runs on-device, no API keys, no data leakage
• Real-time dashboards — Live log streaming and error highlighting in Grafana
• REST API — Trigger analysis via GET /rca, integrate into any workflow

🚀 Quick Start

Prerequisites: Docker Desktop, Ollama with Phi-4-mini

# Pull the model
ollama pull phi4-mini

# Clone and run
git clone https://github.com/vivek1251/ai-rca-system.git
cd ai-rca-system

# Configure environment
cp .env.example .env

# Start all services
docker compose up -d

🧪 Testing

# Simulate errors
curl http://localhost:5000/cause-error
curl http://localhost:5000/stress

# Get AI root cause analysis
curl http://localhost:8000/rca

📡 Endpoints

URL	Purpose
localhost:5000	Flask demo app
localhost:8000/rca	AI root cause analysis
localhost:8000/health	Service health check
localhost:9090	Prometheus metrics
localhost:3000	Grafana (see .env for credentials)

🧬 Running tests

pip install pytest pytest-cov
python -m pytest tests/ -v

🤝 Contributing

See CONTRIBUTING.md for ideas, extension points, and dev setup instructions.

📄 License

MIT