🍕 Open Hack

Appearance

Create Your First Agent ​

In this quickstart, we’ll walk through the process of creating your very first AI agent using the Microsoft Foundry Agent Service.

By the end, you’ll have a simple agent running locally that you can interact with in real time.

Login to Azure ​

Before you can use the Microsoft Foundry Agent Service, you need to sign in to your Azure subscription.

Run the following command and follow the on-screen instructions. Use credentials that have access to your Microsoft Foundry resource:

shell
az login --use-device-code

Install Required Packages ​

Next, install the Python packages needed to work with Microsoft Foundry and manage environment variables:

shell
pip install openai 
pip install azure-identity 
pip install python-dotenv
pip install azure-ai-projects==2.0.0b1

Create a Basic Agent ​

We’ll now create a basic Python script that defines and runs an agent.

  • Start by creating a new file called: agent.py in the workshop folder

Add Imports to agent.py ​

These imports bring in the Foundry Projects SDK, environment handling, and helper classes:

python
import os
from dotenv import load_dotenv
from azure.identity import DefaultAzureCredential
from azure.ai.projects import AIProjectClient
from azure.ai.projects.models import PromptAgentDefinition

Create a .env File ​

We'll store secrets (such as your project connection string and model deployment name) in an environment file for security and flexibility.

Create a file named .env with the following content:

txt
PROJECT_ENDPOINT="<enter it here>"
MODEL_DEPLOYMENT_NAME="<enter your model deployment name here>"

You can find the connection string in the Microsoft Foundry portal, on the homepage of your project. To go to the homepage click on Overview in the Microsoft Foundry portal. The connection string has the following format: https://[name of Foundry resource].services.ai.azure.com/projects/[project name]

Load the .env File ​

Load environment variables into your script by adding this line to agent.py:

python
load_dotenv()

Create an AIProjectClient Instance ​

This client connects your script to the Microsoft Foundry service using the connection string and your Azure credentials.

python
project_client = AIProjectClient(
    endpoint=os.environ["PROJECT_ENDPOINT"],
    credential=DefaultAzureCredential(),
)

Get the OpenAI Client ​

Get the OpenAI client from the project client to interact with conversations:

python
openai_client = project_client.get_openai_client()

Create the Agent ​

Now, let's create the agent itself using the create_version method with a PromptAgentDefinition:

python
agent = project_client.agents.create_version(
    agent_name="hello-world-agent",
    definition=PromptAgentDefinition(
        model=os.environ["MODEL_DEPLOYMENT_NAME"],
        instructions="You are a helpful assistant that can answer questions about pizza ordering.",
    ),
)

Create a Conversation ​

Agents interact within conversations. A conversation is like a container that stores all messages exchanged between the user and the agent.

python
conversation = openai_client.conversations.create()
print(f"Created conversation (id: {conversation.id})")

Chat with the Agent ​

This loop lets you send messages to the agent and receive responses. Type into the terminal, and the agent will respond.

python
while True:
    # Get the user input
    user_input = input("You: ")

    if user_input.lower() in ["exit", "quit"]:
        print("Exiting the chat.")
        break

    # Get the agent response
    response = openai_client.responses.create(
        conversation=conversation.id,
        input=user_input,
        extra_body={"agent": {"name": agent.name, "type": "agent_reference"}},
    )

    # Print the agent response
    print(f"Assistant: {response.output_text}")

Run the Agent ​

Finally, run the Python script:

shell
python agent.py

You can now chat with your agent directly in the terminal. Type exit or quit to stop the conversation.

Recap ​

In this quickstart, you have:

  • Logged in to Azure
  • Retrieved a connection string
  • Separated secrets from code using .env
  • Created a basic agent with the Microsoft Foundry Agent Service
  • Started a conversation with the agent
  • Interacted with the agent using the OpenAI client