AI agents · SRE · Hackathon

Aegis, SRE Guardian

An incident response demo that monitors infrastructure, diagnoses common failures, and applies constrained remote fixes.

3 agentsmonitor, diagnose, remediate
Livemetrics streamed over WebSockets
AllowlistedSSH remediation commands

System goal

Demonstrate an agent workflow that can respond to infrastructure incidents while keeping remediation bounded and inspectable.

Response workflow

A Monitor agent detects conditions, a Diagnosis agent analyzes them, and a Remediation agent selects from a whitelisted set of SSH actions. Arbitrary shell input is not passed through to the remote server.

Operator view

The dashboard streams infrastructure metrics over WebSockets. A ChromaDB vector store supports retrieval from prior incidents, and the system can generate PDF post-mortems.

PythonFastAPIReactTypeScriptChromaDBWebSocketsParamiko

Scenarios and demo

The demo covers CPU surges, memory pressure, disk I/O, DDoS, and database bottlenecks. Built for the AMD Developer Hackathon 2025, AI Agents & Agentic Workflows track.

Try the demo ↗ · Source on GitHub ↗