🤖 Agentic MLOps: Self-Healing Infrastructure Agent

An autonomous DevOps agent built to monitor, diagnose, and repair containerized applications. This project demonstrates Agentic Operations—using code to perform "Reason + Act" (ReAct) loops to maintain system uptime without human intervention.

🌟 The Problem & Solution

• Problem: Manual debugging of container configuration errors (like missing secrets) increases MTTR (Mean Time To Recovery) and requires 24/7 human monitoring.
• Solution: A "Doctor Agent" that utilizes the Docker SDK to perform real-time log analysis and autonomous remediation.

🛠️ Tech Stack

• Backend: FastAPI, Uvicorn (The "Patient" App)
• Containerization: Docker & Docker Compose
• Automation: Python Docker SDK (The "Doctor" Agent)
• Environment: Windows PowerShell / CMD

🏗️ Project Structure

self-healing-mcp-agent/
├── patient-app/           # The FastAPI application (The Patient)
│   ├── app.py             # Logic with intentional failure points
│   └── Dockerfile         # Container definition
├── doctor-agent/          # The Autonomous Agent (The Doctor)
│   └── agent.py           # The "Reason + Act" healing logic
└── docker-compose.yml     # Infrastructure orchestration

🚀 Step-by-Step Execution Guide

1. Build and Deploy the "Patient" From the root directory, build and launch the application. We intentionally "forget" the DB_SECRET to trigger a failure. PowerShell docker compose up --build -d

2. Trigger the Failure (The "Try" Phase) Access the health endpoint to generate a 500 error in the logs. PowerShell Invoke-RestMethod -Uri "http://localhost:8000/health"

3. Run the Autonomous "Doctor" (The "Heal" Phase) In a new terminal window, start the agent. It will detect the error and fix the container. PowerShell python doctor-agent/agent.py

4. Verify the Recovery Once the agent reports SUCCESS, verify the app is healthy. PowerShell Invoke-RestMethod -Uri "http://localhost:8000/health" Expected Output: {"status":"running","message":"Connected with TylerTexas_Verified_2026"}

🎓 Skills Acquired & Demonstrated

• Agentic Operations: Implementation of "Reason + Act" (ReAct) patterns for autonomous system management.
• Observability & Logging: Configuring Python applications to emit actionable telemetry for automated monitors.
• Infrastructure as Code (IaC): Managing container lifecycles and environment configurations via Docker Compose.
• Automated Remediation: Reducing MTTR (Mean Time To Recovery) through programmatic infrastructure state changes.
• Troubleshooting: Identifying and resolving logging bottlenecks (buffering) in containerized Windows environments.