chore: add a deployment example

Author: lpm0073

.gitignore

@@ -1,4 +1,4 @@
-keen_**
+automl-model
 
 # application data
 # -------------------------------------

conda.yaml

@@ -0,0 +1,9 @@
+# conda.yml
+name: automl-env
+dependencies:
+  - python=3.8
+  - pip
+  - pip:
+      - onnxruntime
+      - numpy
+      - pandas

docs/AZURE_ACCOUNT_SETUP.md

@@ -134,9 +134,20 @@ az provider register --namespace Microsoft.Compute
 az provider register --namespace Microsoft.Storage
 az provider register --namespace Microsoft.KeyVault
 az provider register --namespace Microsoft.ContainerInstance
+az provider register --namespace Microsoft.ContainerRegistry
+az provider register --namespace Microsoft.ContainerService
+
+# Additional providers for real-time endpoints
+az provider register --namespace Microsoft.Web
+az provider register --namespace Microsoft.Network
+az provider register --namespace Microsoft.Authorization
+az provider register --namespace Microsoft.Insights
+az provider register --namespace Microsoft.OperationalInsights
+az provider register --namespace Microsoft.Resources
+
 
 # Check registration status
-az provider list --query "[?namespace=='Microsoft.MachineLearningServices' || namespace=='Microsoft.Compute' || namespace=='Microsoft.Storage'].{Namespace:namespace, State:registrationState}" --output table
+az provider list --query "[?namespace=='Microsoft.MachineLearningServices' || namespace=='Microsoft.Compute' || namespace=='Microsoft.Storage' || namespace=='Microsoft.ContainerService' || namespace=='Microsoft.Web' || namespace=='Microsoft.Network' || namespace=='Microsoft.Authorization' || namespace=='Microsoft.Insights' || namespace=='Microsoft.OperationalInsights' || namespace=='Microsoft.Resources'].{Namespace:namespace, State:registrationState}" --output table
 ```
 
 **Note**: Registration can take 5-10 minutes. Wait for all to show "Registered" before proceeding.
@@ -188,12 +199,23 @@ az vm list-usage --location eastus2 --query "[?contains(name.value, 'Standard')]
 
 ### Common Quota Requirements for AutoML
 
-| VM Family                  | Recommended Quota | Use Case               |
-| -------------------------- | ----------------- | ---------------------- |
-| Standard DSv3 Family vCPUs | 2-10              | General AutoML jobs    |
-| Standard Dv3 Family vCPUs  | 16-32             | CPU-intensive training |
-| Standard FSv2 Family vCPUs | 16-32             | Fast compute jobs      |
-| Total Regional vCPUs       | 50-100            | Overall regional limit |
+| VM Family                       | Recommended Quota | Use Case                |
+| ------------------------------- | ----------------- | ----------------------- |
+| Standard DSv3 Family vCPUs      | 2-10              | General AutoML jobs     |
+| **Standard DASv4 Family vCPUs** | **4-10**          | **Real-time endpoints** |
+| Standard Dv3 Family vCPUs       | 16-32             | CPU-intensive training  |
+| Standard FSv2 Family vCPUs      | 16-32             | Fast compute jobs       |
+| Total Regional vCPUs            | 50-100            | Overall regional limit  |
+
+### Check Endpoint VM Quotas
+
+```console
+# Check quota for endpoint VMs (different from training compute)
+az vm list-usage --location eastus2 --query "[?contains(name.value, 'DASv4')]" --output table
+
+# Common endpoint VM families
+az vm list-usage --location eastus2 --query "[?contains(name.value, 'DASv4') || contains(name.value, 'DSv3') || contains(name.value, 'Dv3')]" --output table
+```
 
 ### Request Quota Increase in Azure Portal
 
@@ -325,6 +347,31 @@ az login
 az account set --subscription "your-subscription-name"
 ```
 
+### Issue 5: Troubleshooting Endpoint Creation
+
+If endpoint creation fails with "invalid request" errors, try these solutions:
+
+````console
+# 1. Use same VM family as training cluster
+az ml online-endpoint create \
+  --name titanic-test \
+  --auth-mode key \
+  --resource-group automl-resources \
+  --workspace-name automl-workspace
+
+# 2. Deploy with DSv3 family (same as training)
+az ml online-deployment create \
+  --name blue \
+  --endpoint-name titanic-test \
+  --model azureml:your-model-name@latest \
+  --instance-type Standard_D2s_v3 \
+  --instance-count 1 \
+  --resource-group automl-resources \
+  --workspace-name automl-workspace
+
+# 3. Alternative: Use Azure ML Studio interface
+# Go to ml.azure.com → Models → Deploy → Real-time endpoint
+
 ## 💰 Cost Management
 
 ### Estimated Costs for AutoML
@@ -349,7 +396,7 @@ az ml compute update \
   --idle-time-before-scale-down 300 \
   --resource-group automl-resources \
   --workspace-name automl-workspace
-```
+````
 
 ## ✅ Validation Checklist
 

docs/AZURE_DEPLOYMENT.md

@@ -0,0 +1,100 @@
+# Azure AutoML Real-Time Deployment
+
+Microsoft's recommended approach for deploying AutoML models to production for real-time endpoints is to download your AutoML model and deploy it using Azure Functions.
+
+## Prerequisites
+
+- Azure CLI installed and logged in
+- Azure Functions Core Tools installed
+- Python 3.8+ installed locally
+- Access to the Azure ML workspace containing your model
+
+```console
+# Install Azure Functions Core Tools (Mac)
+brew tap azure/functions
+brew install azure-functions-core-tools@4
+
+# Verify installation
+func --version
+```
+
+## Deployment
+
+### Step 1: Download Your AutoML Model
+
+```console
+# Download your trained AutoML model
+az ml model download \
+  --name titanic-survival \
+  --version 1 \
+  --resource-group ubc-cdl10 \
+  --workspace-name UBC-CDL10 \
+  --download-path ./titanic-survival-app/automl-model/
+```
+
+### Step 2: Create Azure Function App
+
+```console
+# Create a new storage account (name must be globally unique, 3-24 chars, lowercase/numbers only)
+az storage account create \
+  --name ubccdl10titanic \
+  --resource-group ubc-cdl10 \
+  --location eastus2 \
+  --sku Standard_LRS
+
+# Create Function App
+az functionapp create \
+  --resource-group ubc-cdl10 \
+  --consumption-plan-location eastus2 \
+  --runtime python \
+  --runtime-version 3.8 \
+  --functions-version 4 \
+  --name titanic-survival-app \
+  --storage-account ubccdl10titanic
+```
+
+### Step 3: Create Function Code
+
+- [**init**.py](../titanic-survival-app/__init__.py)
+- [requirements.txt](../titanic-survival-app/requirements.txt)
+- [function.json](../titanic-survival-app/function.json)
+
+### Step 4: Deploy Function
+
+```console
+cd titanic-survival-app
+func azure functionapp publish titanic-survival-app
+```
+
+### Step 5: Test Your Deployment
+
+```console
+# Test the function
+curl -X POST https://titanic-survival-app.azurewebsites.net/api/predict \
+  -H "Content-Type: application/json" \
+  -d '{"data": [[3, 1, 22, 1, 0, 7.25, 2]]}'
+```
+
+Expected response:
+
+```json
+{
+  "prediction": [0]
+}
+```
+
+## 🔧 Troubleshooting
+
+### Check Function Logs
+
+```console
+# Check function logs
+az functionapp log tail --name titanic-survival-app --resource-group ubc-cdl10
+```
+
+### Common Issues
+
+1. **Model Loading Errors**: Ensure model.pkl is in the same directory as **init**.py
+2. **Dependency Issues**: Pin all package versions in requirements.txt
+3. **Memory Issues**: Use Premium plan for large models
+4. **JSON Issues**: Ensure input data format matches training data

requirements/local.txt

@@ -7,6 +7,7 @@
 
 -r base.txt
 
+azure.functions==1.23.0
 
 # Code linters, formatters, and security scanners
 # ------------

titanic-survival-app/README.md

@@ -0,0 +1,46 @@
+# Azure function deployment resources
+
+This directory contains the deployment resources for a machine learning model created using Azure AutoML (Automated Machine Learning).
+See [Azure AutoML Real-Time Deployment](../docs/AZURE_DEPLOYMENT.md)
+
+## Overview
+
+These files enable deployment of an AutoML-trained model as a serverless Azure Function for real-time inference. The model was trained using Azure's AutoML service and is now packaged for production deployment.
+
+## Files
+
+- **`__init__.py`** - Main Azure Function handler that loads the AutoML model and processes prediction requests
+- **`requirements.txt`** - Python dependencies required for the function runtime
+- **`function.json`** - Azure Functions configuration defining the HTTP trigger and bindings
+- **`automl-model/`** - Directory containing the downloaded AutoML model artifacts
+
+## Usage
+
+This function is designed to be deployed to Azure Functions and provides a REST API endpoint for real-time model predictions.
+
+### API Endpoint
+
+```bash
+curl -X POST https://titanic-survival-app.azurewebsites.net/api/predict \
+  -H "Content-Type: application/json" \
+  -d '{"data": [[3, 1, 22, 1, 0, 7.25, 2]]}'
+```
+
+### Response
+
+```json
+{
+  "prediction": [0]
+}
+```
+
+## Deployment
+
+For complete deployment instructions, see the [Azure Deployment Guide](../docs/AZURE_DEPLOYMENT.md).
+
+## Model Information
+
+- **Source**: Azure AutoML
+- **Task**: Binary Classification (Titanic Survival Prediction)
+- **Input Features**: Passenger class, sex, age, siblings/spouses, parents/children, fare, embarked port
+- **Output**: Survival prediction (0 = did not survive, 1 = survived)

titanic-survival-app/__init__.py

@@ -0,0 +1,70 @@
+"""
+https://github.com/FullStackWithLawrence/azureml-example
+./titanic-survival-app/__init__.py
+
+Azure Function to serve predictions from a pre-trained AutoML model.
+
+EXECUTION ENVIRONMENT:
+- This code runs on Azure Functions cloud servers (NOT on your local computer)
+- Deployed via 'func azure functionapp publish' command from your local machine
+- Accessible globally via HTTPS endpoint: https://titanic-survival-app.azurewebsites.net
+
+DEPLOYMENT FLOW:
+1. Model downloaded locally: 'az ml model download' → ./automl-model/
+    az ml model download \
+        --name titanic-survival \
+        --version 1 \
+        --resource-group ubc-cdl10 \
+        --workspace-name UBC-CDL10 \
+        --download-path path/to/this/repo/titanic-survival-app/automl-model/
+
+2. Function code created locally in ./titanic-survival-app/__init__.py (this file)
+
+3. Entire titanic-survival-app/ folder uploaded to Azure via 'func azure functionapp publish'
+    cd titanic-survival-app
+    func azure functionapp publish titanic-survival-app
+
+4. Azure Functions runtime loads and executes this code on their servers
+
+MODEL LOADING:
+- The model.pkl file is loaded ONCE when the Azure Function starts up (cold start)
+- Model file exists on Azure servers because it was packaged and uploaded with this code
+- Path './automl-model/model.pkl' is relative to this file's location on Azure servers
+
+RUNTIME BEHAVIOR:
+- When HTTP requests arrive at the Azure endpoint, Azure executes main() function
+- Input data comes from HTTP POST requests with JSON payload
+- Predictions are computed using the pre-loaded model on Azure servers
+- Results are returned as JSON HTTP responses
+
+This function expects input data in JSON format and returns predictions in JSON format.
+"""
+
+import json
+import os
+import pickle
+
+import azure.functions as func
+import pandas as pd
+
+
+# Load model once at startup
+model_path = os.path.join(os.path.dirname(__file__), "./automl-model/model.pkl")
+with open(model_path, "rb") as f:
+    model = pickle.load(f)  # nosec B301
+
+
+def main(req: func.HttpRequest) -> func.HttpResponse:
+    try:
+        # Get input data
+        req_body = req.get_json()
+        input_data = pd.DataFrame(req_body["data"])
+
+        # Make prediction
+        prediction = model.predict(input_data)
+
+        # Return result
+        return func.HttpResponse(json.dumps({"prediction": prediction.tolist()}), mimetype="application/json")
+    # pylint: disable=W0718
+    except Exception as e:
+        return func.HttpResponse(json.dumps({"error": str(e)}), status_code=400)

titanic-survival-app/function.json

@@ -0,0 +1,17 @@
+{
+  "scriptFile": "__init__.py",
+  "bindings": [
+    {
+      "authLevel": "function",
+      "type": "httpTrigger",
+      "direction": "in",
+      "name": "req",
+      "methods": ["post"]
+    },
+    {
+      "type": "http",
+      "direction": "out",
+      "name": "$return"
+    }
+  ]
+}

titanic-survival-app/requirements.txt

@@ -0,0 +1,9 @@
+"""
+# titanic-survival-app/requirements.txt
+# This file contains the Python package dependencies for the Azure Function App
+# that serves the Titanic dataset model.
+"""
+azure-functions
+pandas
+numpy
+scikit-learn