Task 04: Introduction to Azure AI Agent Service

From the project1 page, select Models + endpoints from the left menu.

Select + Deploy model, then select Deploy base model.

Select the gpt-4o model, then select Confirm.

Set the Deployment type to Standard, set the Tokens per Minute Rate Limit to 250K, then select Deploy.

Switch to the tab with Azure portal.

In the search bar at the top, search for bing and then select Bing Resources.

From the Bing Resources page, select + Add, then select + Grounding with Bing Search.

On the Create a Grounding with Bing Search resource page, select your resource group and pricing tier. Give it a name of bingsrch and select Review + Create, then select Create.

Return to the tab with the Azure AI model deployment and select Agents from the left menu.

Under Select an Azure OpenAI Service resource, select your hub and select Let’s go.

Under Select or deploy a model, select gpt-4o then select Next.

This will create a new agent using the specified model.

Select the agent to open the Setup pane.

In the Instructions field of the Setup pane, enter the following:

 Understand User Query:
 Analyze the user's query to identify if it requires real-time information (e.g., weather, date, news).
    
 Use Bing Search Tool for Real-Time Data:
 If the query involves up-to-date information, use the Bing Search tool to retrieve relevant data.
    
 Craft a Clear, Concise Response:
 Extract the relevant information (e.g., temperature, news) and provide the answer in a simple and direct way.
    
 Ask for Clarification if Needed:
 If the query is vague or missing details (e.g., location for weather), ask the user for more information.

Under Knowledge in the Setup pane, select + Add, then select Grounding with Bing Search.

Select the + Create connectioh connection, then select Add connection next to the bingsrch resource.

The Knowledge section allows you to specify the source of information for the agent. In this case, we’re using the Grounding with Bing Search service to retrieve up-to-date information from external sources, beyond our local dataset. You can also set the source to a pre-existing search index or local data. If you want the agent to be able to retrieve both local and web-based results, you can add a connection for each.

Under Actions in the Setup pane, select + Add, then select Code interpreter.

On the Add code interpreter action page, select Select local files and then select the products.xlsx file created earlier.

Select Upload and Save.

The Actions section allows you to specify additional tasks for the agent beyond simple data retrieval. The Code interpreter tool can be used for tasks like performing calculations or creating visualizations from your data.

From the upper right of the Setup pane, select Try in playground.

In the Agents playground chat, enter What is the weather like in New York?

The gpt-4o model doesn’t have direct access to the current date. However, by using the Bing Search tool, the agent can retrieve up-to-date information for time-sensitive queries.

In the chat, enter What is the average price of the products in the xlsx file?

The code interpreter tool allows for more complex queries about your data. In this case, we used it to retrieve the average price of the products in our data set.