Quick Start Guide - Azure AI Foundry¶

This notebook provides a hands-on introduction to Azure AI Foundry. You'll learn how to:

Initialize the AI Project client
List available models
Create a simple completion request
Handle basic error scenarios

Prerequisites¶

Completed environment setup from previous notebook
Azure credentials configured

Import Required Libraries and Setup¶

In the next cell, we'll:

Import the necessary Azure SDK libraries for authentication and AI Projects
Import standard Python libraries for environment variables and JSON handling
Initialize Azure credentials using DefaultAzureCredential
- This will automatically use your logged-in Azure CLI credentials
- Alternatively, it can use other authentication methods like environment variables or managed identity

In [6]:

  Copied!     
 
# Import required libraries
from azure.identity import DefaultAzureCredential
from azure.ai.projects import AIProjectClient
import os
import json

# Initialize credentials
credential = DefaultAzureCredential()
# Import required libraries from azure.identity import DefaultAzureCredential from azure.ai.projects import AIProjectClient import os import json # Initialize credentials credential = DefaultAzureCredential()

Initialize AI Project Client¶

Note: Before proceeding, ensure you:

Copy your .env.local file to .env

Update the project connection string in your .env file

Have a Hub and Project already provisioned in Azure AI Foundry

You can find your project connection string in Azure AI Foundry under your project's settings:

Creating the AI Project Client¶

In the next cell, we'll create an AI Project client using the connection string from our .env file.

Note: This example uses the synchronous client. For higher performance scenarios, you can also create an asynchronous client by importing asyncio and using the async methods from AIProjectClient.

The client will be used to:

Connect to your Azure AI Project using the connection string
Authenticate using Azure credentials
Enable making inference requests to your deployed models

In [ ]:

  Copied!     
 
from dotenv import load_dotenv

# Create AI Project client
# Load environment variables from .env file
load_dotenv()

try:
    client = AIProjectClient.from_connection_string(
        conn_str=os.getenv("PROJECT_CONNECTION_STRING"),
        credential=credential
    )
    print("✓ Successfully initialized AIProjectClient")
except Exception as e:
    print(f"× Error initializing client: {str(e)}")
from dotenv import load_dotenv # Create AI Project client # Load environment variables from .env file load_dotenv() try: client = AIProjectClient.from_connection_string( conn_str=os.getenv("PROJECT_CONNECTION_STRING"), credential=credential ) print("✓ Successfully initialized AIProjectClient") except Exception as e: print(f"× Error initializing client: {str(e)}")

Create a Simple Completion¶

Let's try a basic completion request:

Now that we have an authenticated client, let's use it to make a chat completion request. The code below demonstrates how to:

Get a ChatCompletionsClient from the azure-ai-inference package
Use it to make a simple completion request

We'll use the MODEL_DEPLOYMENT_NAME from our .env file, making it easy to switch between different deployed models without changing code. This could be an Azure OpenAI model, Microsoft model, or other providers that support chat completions.

Note: Make sure you have the azure-ai-inference package installed (from requirements.txt)

In [ ]:

  Copied!     
 
from azure.ai.inference.models import UserMessage

model_deployment_name = os.getenv("MODEL_DEPLOYMENT_NAME")

try:
    chat_client = client.inference.get_chat_completions_client()
    response = chat_client.complete(
        model=model_deployment_name, 
        messages=[UserMessage(content="How to be healthy in one sentence?")]
    )
    print(response.choices[0].message.content)
except Exception as e:
    print(f"An error occurred: {str(e)}")
from azure.ai.inference.models import UserMessage model_deployment_name = os.getenv("MODEL_DEPLOYMENT_NAME") try: chat_client = client.inference.get_chat_completions_client() response = chat_client.complete( model=model_deployment_name, messages=[UserMessage(content="How to be healthy in one sentence?")] ) print(response.choices[0].message.content) except Exception as e: print(f"An error occurred: {str(e)}")

Create a simple Agent¶

Using AI Agent Service, we can create a simple agent to answer health related questions.

Let's explore Azure AI Agent Service, a powerful tool for building intelligent agents.

Azure AI Agent Service is a fully managed service that helps developers build, deploy, and scale AI agents without managing infrastructure. It combines large language models with tools that allow agents to:

Answer questions using RAG (Retrieval Augmented Generation)
Perform actions through tool calling
Automate complex workflows

The code below demonstrates how to:

Create an agent with a code interpreter tool
Create a conversation thread
Send a message requesting BMI analysis
Process the request and get results
Save any generated visualizations

The agent will use the model specified in our .env file (MODEL_DEPLOYMENT_NAME) and will have access to a code interpreter tool for creating visualizations. This showcases how agents can combine natural language understanding with computational capabilities.

The visualization will be saved as a PNG file in the same folder as this notebook.

In [ ]:

  Copied!     
 
from azure.ai.projects.models import CodeInterpreterTool

try:
    # Create an agent with code interpreter
    code_interpreter = CodeInterpreterTool()
    agent = client.agents.create_agent(
        model=model_deployment_name,
        name="bmi-calculator",
        instructions="You are a health analyst who calculates BMI using US metrics (pounds, feet/inches). Use average US male measurements: 5'9\" (69 inches) and 198 pounds. Create a visualization showing where this BMI falls on the scale.",
        tools=code_interpreter.definitions,
        tool_resources=code_interpreter.resources,
    )
    
    thread = client.agents.create_thread()
    
    # Request BMI analysis
    message = client.agents.create_message(
        thread_id=thread.id,
        role="user",
        content="Calculate BMI for an average US male (5'9\", 198 lbs). Create a visualization showing where this BMI falls on the standard BMI scale from 15 to 35. Include the standard BMI categories (Underweight, Normal, Overweight, Obese) in the visualization."
    )
    
    # Process the request
    run = client.agents.create_and_process_run(thread_id=thread.id, assistant_id=agent.id)
    
    # Get and save any visualizations
    messages = client.agents.list_messages(thread_id=thread.id)
    for image_content in messages.image_contents:
        file_name = f"bmi_analysis_{image_content.image_file.file_id}.png"
        client.agents.save_file(file_id=image_content.image_file.file_id, file_name=file_name)
        print(f"Analysis saved as: {file_name}")
    
    # Print the analysis
    if last_msg := messages.get_last_text_message_by_sender("assistant"):
        print(f"Analysis: {last_msg.text.value}")
    
    # Cleanup
    client.agents.delete_agent(agent.id)
    
except Exception as e:
    print(f"An error occurred: {str(e)}")
from azure.ai.projects.models import CodeInterpreterTool try: # Create an agent with code interpreter code_interpreter = CodeInterpreterTool() agent = client.agents.create_agent( model=model_deployment_name, name="bmi-calculator", instructions="You are a health analyst who calculates BMI using US metrics (pounds, feet/inches). Use average US male measurements: 5'9\" (69 inches) and 198 pounds. Create a visualization showing where this BMI falls on the scale.", tools=code_interpreter.definitions, tool_resources=code_interpreter.resources, ) thread = client.agents.create_thread() # Request BMI analysis message = client.agents.create_message( thread_id=thread.id, role="user", content="Calculate BMI for an average US male (5'9\", 198 lbs). Create a visualization showing where this BMI falls on the standard BMI scale from 15 to 35. Include the standard BMI categories (Underweight, Normal, Overweight, Obese) in the visualization." ) # Process the request run = client.agents.create_and_process_run(thread_id=thread.id, assistant_id=agent.id) # Get and save any visualizations messages = client.agents.list_messages(thread_id=thread.id) for image_content in messages.image_contents: file_name = f"bmi_analysis_{image_content.image_file.file_id}.png" client.agents.save_file(file_id=image_content.image_file.file_id, file_name=file_name) print(f"Analysis saved as: {file_name}") # Print the analysis if last_msg := messages.get_last_text_message_by_sender("assistant"): print(f"Analysis: {last_msg.text.value}") # Cleanup client.agents.delete_agent(agent.id) except Exception as e: print(f"An error occurred: {str(e)}")