Google Cloud Rapid Agent Hackathon 2026 — Dynatrace Track

SRE-GPT is an autonomous AI agent that monitors a production API, detects performance anomalies in real-time, self-heals through a cascaded remediation strategy, and generates AI-powered post-mortems — with zero human intervention.

Todo API (FastAPI + PostgreSQL on Render)
        ↓ real metrics every 60s
Dynatrace MCP Server (anomaly enrichment + event ingestion)
        ↓
SRE-GPT Agent (Python) + Gemini AI
        ↓ autonomous decision
Auto-Repair → Rollback (Render API) → Incident Report
        ↓
Glassmorphism Dashboard (live status.json polling)

• Real-time Monitoring — Measures actual latency, error rate, and availability from a live production API
• MCP Dynatrace Integration — Sends incident events to Dynatrace and queries observability data via the official MCP server
• Cascaded Remediation — Auto-repair first, rollback only if repair fails
• Autonomous Rollback — Triggers real redeployments via Render REST API
• Gemini AI Post-Mortems — Generates structured root cause analysis reports after every incident
• Live Dashboard — Glassmorphism UI with incident timeline, terminal log, sparkline charts, and cooldown progress bar
• Multi-language — EN/FR toggle on the dashboard

• Python 3.12+
• Node.js 18+
• A Render account (free tier)
• A Dynatrace trial account
• A Gemini API key (Google AI Studio)

git clone https://github.com/YOUR_USERNAME/sre-gpt.git
cd sre-gpt
copy .env.example .env
notepad .env

Fill in your .env:

DT_ENVIRONMENT_URL=https://YOUR_ENV.apps.dynatrace.com
DT_API_TOKEN=dt0c01.xxxxx
GEMINI_API_KEY=AIzaxxxxx
API_BASE_URL=https://your-api.onrender.com
RENDER_API_KEY=rnd_xxxxx
RENDER_SERVICE_ID=srv-xxxxx
ALERT_LATENCY_MS=1000
ALERT_ERROR_RATE=0.1

Push the api/ folder to a Render Web Service with:

• Build Command: pip install -r requirements.txt
• Start Command: uvicorn main:app --host 0.0.0.0 --port $PORT
• Environment Variable: DATABASE_URL = your Render PostgreSQL URL

$env:DT_ENVIRONMENT="https://YOUR_ENV.apps.dynatrace.com"
npx -y @dynatrace-oss/dynatrace-mcp-server@latest --http --port 8888

pip install requests python-dotenv google-genai psycopg2-binary
cd agent
python agent.py

HTML Dashboard (live agent data):

cd dashboard
python cors_server.py

Open http://localhost:4000 in your browser.

Note: cors_server.py is required instead of python -m http.server to allow cross-origin requests from the React dashboard on port 3000.

React Dashboard (advanced UI with PDF export):

npm install
npm run dev

Open http://localhost:3000 in your browser.

Note: Always launch from the correct directory. Use Ctrl+Shift+R to hard refresh if the browser shows a blank page.

# Inject latency fault
curl -X POST https://sre-gpt-api.onrender.com/simulate/latency?seconds=3

# Inject error fault  
curl -X POST https://sre-gpt-api.onrender.com/simulate/errors?rate=0.3

# Reset to normal
curl -X POST https://sre-gpt-api.onrender.com/simulate/reset

Or click "Simulate Incident" directly on the dashboard.

ANOMALY DETECTED
      ↓
Step 1 — Gemini analyzes root cause
      ↓
Step 2 — Auto-repair (reset simulation, wake-up ping)
      ↓ if metrics normalize → AUTO_REPAIRED ✅
      ↓ if still anomalous after 30s...
Step 3 — Real rollback via Render API
      ↓ ROLLBACK ✅
Step 4 — Gemini generates post-mortem report
      ↓ Dashboard updated + Dynatrace event sent

Dashboard url - Open https://sre-gpt.onrender.com/

The agent runs automatically every 5 minutes via GitHub Actions. Add these secrets to your GitHub repo (Settings → Secrets):

• DT_ENVIRONMENT_URL
• DT_API_TOKEN
• GEMINI_API_KEY
• API_BASE_URL
• RENDER_API_KEY
• RENDER_SERVICE_ID

MIT — See LICENSE