AI-SRE

AI-SRE is an open-source platform that monitors your services 24/7, detects anomalies automatically, and uses Claude to tell you exactly what went wrong and how to fix it.

What takes an SRE 30 minutes of digging through dashboards, logs, and traces now takes about 10 seconds.

How it works

1. Continuous monitoring: metrics, logs, and traces collected across your stack
2. Anomaly detection: statistical analysis flags deviations from baseline
3. Automatic correlation: related logs, metrics, and traces grouped into a single incident view
4. Root cause analysis: Claude generates an RCA with what broke, why, how to fix it, and confidence scores

Stack

• API: FastAPI with Celery workers for background scanning
• Web UI: Next.js dashboard with @xyflow/react for flow visualization, glassmorphism design
• Observability: Grafana, Mimir, Loki, Tempo, Flower
• Data: PostgreSQL, Redis
• Integrations: Stripe billing, Slack notifications, Kubernetes client

Architecture

Next.js dashboard
      ↓
FastAPI + Celery workers
      ↓
Postgres · Redis · Grafana stack (Mimir, Loki, Tempo)
      ↓
Claude API (RCA generation)

All 11 services run via Docker Compose for local development. The platform includes incident management, service health views, integration settings, and an NLP query interface for asking questions about your infrastructure in plain English.

Why I built it

I got tired of waking up at 3 AM to debug production issues. Traditional monitoring drowns you in dashboards. AI-SRE correlates the signal and writes the diagnosis.