Azure OpenAI - Complete Guide

PROBLEM

You need to create an application that leverages LLM (Large Language Models) to generate text, code, or images, or to calculate embeddings or transcribe audio. How can you do this using Azure OpenAI?

SOLUTION

Use the Azure OpenAI service to access and manage models like gpt-4, gpt-35-turbo, text-embedding-ada, dall-e, whisper, etc.

REQUIRED COMPONENTS

Component	Description
Azure OpenAI Resource	Main resource to create in the portal or via CLI
Azure CLI / Azure Portal	To create and manage resources and model deployments
API Key and Endpoint	To authenticate and make REST calls or via SDK
Model to deploy	E.g.: gpt-35-turbo, text-embedding-ada, dall-e, whisper
Deployment Name	Model alias for the application
SDK (Python: openai)	To integrate models into your code (chat, completions, embeddings, etc.)

OPERATIONAL PROCEDURE

1. Create Azure OpenAI Resource (CLI)

az cognitiveservices account create \
-n MyOpenAIResource \
-g OAIResourceGroup \
-l eastus \
--kind OpenAI \
--sku s0 \
--subscription subscriptionID

2. Deploy a Model (e.g., GPT-35 Turbo)

az cognitiveservices account deployment create \
   -g OAIResourceGroup \
   -n MyOpenAIResource \
   --deployment-name MyModel \
   --model-name gpt-35-turbo \
   --model-version "0125"  \
   --model-format OpenAI \
   --sku-name "Standard" \
   --sku-capacity 1

3. Usage via API (REST – chat completions example)

curl https://<ENDPOINT>.openai.azure.com/openai/deployments/<DEPLOYMENT_NAME>/chat/completions?api-version=2023-03-15-preview \
  -H "Content-Type: application/json" \
  -H "api-key: <API_KEY>" \
  -d '{
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "What is Azure OpenAI?"}
    ]
}'

4. Usage via Python SDK

from openai import AzureOpenAI

client = AzureOpenAI(
    azure_endpoint="<YOUR_ENDPOINT_NAME>",
    api_key="<YOUR_API_KEY>",
    api_version="2024-02-15-preview"
)

response = client.chat.completions.create(
    model="<YOUR_DEPLOYMENT_NAME>",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "What is Azure OpenAI?"}
    ]
)
print(response.choices[0].message.content)

AVAILABLE MODEL TYPES

Category	Model (example)	Description
LLM Chat	gpt-4, gpt-35-turbo	Text generation/conversations/chatbots
Code	gpt-35-turbo	Code generation
Embedding	text-embedding-ada:002	Transforms text into numerical vectors
Images	dall-e	Generates images from prompts (preview)
Speech-to-text	whisper	Transcribes audio to text
Text-to-speech	text-to-speech	Generates voice from text

REQUIRED PERMISSIONS

Azure Role	Permissions
Cognitive Services OpenAI User	Resource viewing, playground
Cognitive Services OpenAI Contributor	Deployment creation

COMMON USE CASES

Objective	Recommended Model	Endpoint
Chatbot / Assistant	gpt-35-turbo	chat/completions
Semantic search engine	text-embedding-ada:002	embeddings
Code generation	gpt-4 or gpt-35-turbo	completions
Text similarity analysis	text-similarity-*	embeddings
Image generation	dall-e	images/generations
Voice transcription	whisper	audio/transcriptions

SIMULATED EXAM QUESTIONS

1. You need an enterprise chatbot for a website. Which Azure OpenAI model would you use and how would you configure it?

2. You need to implement semantic search to compare product descriptions. Which model do you use?

3. You want to automatically generate Python code from natural language prompts. Which model and endpoint?

4. You want to integrate AI into your Python code: what parameters are needed to connect?

5. What RBAC roles are required to create a GPT model deployment in an enterprise?