Skip to content

first-deployment.md

Latest commit

 

History

History
190 lines (134 loc) · 4.04 KB

first-deployment.md

File metadata and controls

190 lines (134 loc) · 4.04 KB

3️⃣ Deploy to Azure

Now let's deploy your agent to Azure for production use.

🎯 What You'll Deploy

A production-ready setup with:

  • Azure OpenAI for the LLM
  • Azure Container Apps for hosting
  • Managed Identity for secure auth
  • Application Insights for monitoring

🚀 Option 1: Quick Deploy with azd (Recommended)

The fastest way to get to production:

# Clone the production accelerator
git clone --recurse-submodules \
  https://github.com/microsoft/Deploy-Your-AI-Application-In-Production.git

cd Deploy-Your-AI-Application-In-Production

# Login and deploy (~45 minutes)
azd auth login
azd up

This deploys:

  • ✅ Azure AI Foundry
  • ✅ Azure AI Search
  • ✅ Microsoft Fabric (optional)
  • ✅ Microsoft Purview (governance)
  • ✅ Private networking
  • ✅ Managed identities

📖 Full Deployment Guide

🏗️ Option 2: Deploy Infrastructure Only

If you want just the core AI infrastructure from this repo:

cd azure-agentic-engineering/infra/bicep

# Get your principal ID
PRINCIPAL_ID=$(az ad signed-in-user show --query id -o tsv)

# Deploy
az deployment group create \
  --resource-group rg-my-agents \
  --template-file main.bicep \
  --parameters principalId=$PRINCIPAL_ID environmentName=dev

This creates:

  • Azure OpenAI with GPT-4o
  • Azure AI Search
  • Key Vault
  • Storage Account
  • Log Analytics

🐳 Option 3: Containerize Your Agent

Package your agent as a container for Azure Container Apps:

Dockerfile

FROM python:3.12-slim

WORKDIR /app

COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

COPY . .

EXPOSE 8000

CMD ["python", "-m", "uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]

requirements.txt

agent-framework>=1.0.0b1
azure-identity>=1.15.0
fastapi>=0.109.0
uvicorn>=0.27.0

main.py (API wrapper)

from fastapi import FastAPI
from agent_framework.azure import AzureOpenAIResponsesClient
from azure.identity import DefaultAzureCredential
import os

app = FastAPI()

credential = DefaultAzureCredential()
agent = AzureOpenAIResponsesClient(
    endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],
    credential=credential,
).as_agent(
    name="ProductionAgent",
    instructions="You are a helpful assistant."
)

@app.post("/chat")
async def chat(message: str):
    response = await agent.run(message)
    return {"response": response}

@app.get("/health")
async def health():
    return {"status": "healthy"}

Deploy to Container Apps

# Build and push
az acr build --registry myregistry --image my-agent:v1 .

# Deploy
az containerapp create \
  --name my-agent \
  --resource-group rg-my-agents \
  --image myregistry.azurecr.io/my-agent:v1 \
  --target-port 8000 \
  --ingress external \
  --env-vars AZURE_OPENAI_ENDPOINT=https://... \
  --user-assigned /subscriptions/.../my-identity

✅ Post-Deployment Checklist

After deploying, verify everything works:

  • Health check: curl https://your-app.azurecontainerapps.io/health
  • Test chat: curl -X POST https://your-app.../chat?message=Hello
  • Check logs: Azure Portal → Container Apps → Logs
  • Monitor: Azure Portal → Application Insights

🔒 Security Checklist

Before going live:

  • Using Managed Identity (no API keys)
  • Private endpoints enabled (for production)
  • HTTPS only
  • RBAC configured with least privilege
  • Diagnostic logging enabled
  • Content Safety enabled (if user-facing)

💰 Cost Considerations

Resource Approximate Cost
Azure OpenAI (GPT-4o) ~$0.005/1K tokens
Container Apps ~$0.000024/vCPU-second
AI Search (Basic) ~$75/month
Key Vault ~$0.03/10K operations

💡 Use the Azure Pricing Calculator for estimates.

➡️ Next Step

Your agent is deployed! Let's explore what's next.

Continue to: Next Steps →