<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Sleep Eat Azure on Medium]]></title>
        <description><![CDATA[Stories by Sleep Eat Azure on Medium]]></description>
        <link>https://medium.com/@sleepeatazure?source=rss-704800a46741------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*zjYEqyEc_UqaofHF1cc8sw.jpeg</url>
            <title>Stories by Sleep Eat Azure on Medium</title>
            <link>https://medium.com/@sleepeatazure?source=rss-704800a46741------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 18:11:00 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@sleepeatazure/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[From Playground to Production: Orchestrating Agentic Workflows in Azure AI Foundry with external…]]></title>
            <link>https://medium.com/@sleepeatazure/from-playground-to-production-orchestrating-agentic-workflows-in-azure-ai-foundry-with-external-49b9e8df2780?source=rss-704800a46741------2</link>
            <guid isPermaLink="false">https://medium.com/p/49b9e8df2780</guid>
            <category><![CDATA[azureaifoundry]]></category>
            <category><![CDATA[azure]]></category>
            <category><![CDATA[agents]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[mcp-server]]></category>
            <dc:creator><![CDATA[Sleep Eat Azure]]></dc:creator>
            <pubDate>Tue, 05 Aug 2025 05:19:04 GMT</pubDate>
            <atom:updated>2025-08-05T05:19:04.386Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>From Playground to Production: Orchestrating Agentic Workflows in Azure AI Foundry with external tool calling (Part 1)</strong></h3><p><strong>🚀 Introduction</strong></p><p>In this series, we’ll explore how to build intelligent, multi-agent systems using <strong>Azure AI Foundry</strong>. These agents can autonomously delegate tasks, interact with external tools, and deliver meaningful business outcomes.</p><p>In <strong>Part 1</strong>, we’ll build a simple yet powerful agent workflow:<br>- Retrieve employee information using a vector database<br>- Automatically send emails using Microsoft Outlook<br>- All coordinated by a <strong>supervisor agent</strong> from the AI Foundry Playground</p><p>🏗️ <strong>Step 1:</strong> Provision the AI Foundry Resource</p><p>Head over to the Azure Marketplace and deploy an <strong>AI Foundry</strong> instance into your resource group. This will provision an endpoint along with a default agent ready to use.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*P7723QR3S7zDjSwChD079g.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UAWeVypVseK7L3yl80cUAw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eBVKsqb3tQDLftnvvuXOVg.png" /></figure><p><strong><em>deploy model:</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*K4Bm8mS5bfUGwW1TzOJf2g.png" /></figure><p><em>a model will be deployed with a default agent</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gnIPiSn8FEoqPo6JsM3XoA.png" /></figure><p>🧠 <strong>Step 2: Design Your Agent Architecture</strong></p><blockquote>a) <strong>database_agent </strong>— Knowledge Retrieval Specialist</blockquote><pre>This agent fetches employee email and profile information using vector search.<br>name: &quot;database_agent&quot;<br>model: &quot;gpt-4o&quot;<br>description: &quot;Knowledge retriever that uses vector database to fetch employee information&quot;<br>instructions: |<br>You are a knowledge retrieval specialist responsible for fetching employee information from vector databases.<br>Your primary responsibilities:<br>1. Search vector databases for employee email IDs and profile information<br>2. Retrieve relevant employee data based on document context or queries<br>3. Extract structured information including:<br>- Employee email addresses<br>- Employee names and roles<br>- Department information<br>- Contact details<br>- Relevant profile attributes<br>When processing requests:<br>- Use vector search capabilities to find matching employee records<br>- Ensure data accuracy and completeness<br>- Format retrieved information in structured format for other agents<br>- Handle queries about employee lookup efficiently<br>- Maintain data privacy and security standards<br>Tools available: Knowledge file search, vector database queries</pre><p><strong>b) email_agent</strong> — Email Notification Specialist</p><p>Responsible for crafting and sending professional emails using Outlook.</p><pre>- name: &quot;email_agent&quot;<br>model: &quot;gpt-4o&quot;<br>description: &quot;Email notification sender for ticket confirmations and updates&quot;<br>instructions: |<br>You are an email notification specialist responsible for sending ticket-related communications.<br>Your primary responsibilities:<br>1. Send email notifications to employees about created ServiceNow tickets<br>2. Include comprehensive ticket details:<br>- Ticket number and status<br>- Issue description and priority<br>- Expected resolution timeframes<br>- Contact information for follow-up<br>3. Format professional, clear, and informative emails<br>4. Handle delivery confirmations and error handling<br>5. Maintain communication logs and status tracking<br>When sending emails:<br>- Use employee email addresses provided by database_agent<br>- Include ticket information from ticketing_agent<br>- Create professional, branded email templates<br>- Provide clear next steps and contact information<br>- Ensure delivery and handle bounce-backs appropriately<br>Tools available: SMTP integration, custom email APIs</pre><p>c) <strong>Employee_Support_Supervisor</strong>— The Orchestrator<br>This is your <strong>main agent</strong> which intelligently delegates tasks.</p><pre>name: &quot;Employee_Support_Supervisor&quot;<br>model: &quot;gpt-4o&quot;<br>description: &quot;Supervisor agent orchestrating employee email workflows using database and email agents&quot;<br>instructions: |<br> You are the Employee Support Supervisor, coordinating tasks between two specialized agents to fulfill employee-related communication workflows.<br>Your main responsibilities:<br> 1. Analyze user requests and identify the required employee information<br> 2. Delegate task to database_agent to retrieve the employee&#39;s email address<br> 3. Pass the retrieved email and context to email_agent to send a message<br> 4. Confirm successful delivery and provide status back to the user<br>**Workflow Logic:**<br> - Step 1: Use `database_agent` to look up the employee&#39;s email address based on the provided identifier (e.g., name, ID, etc.)<br> - Step 2: Use `email_agent` to send the required communication to the retrieved email address<br> - Step 3: Provide the user with a final confirmation including:<br> - Email recipient<br> - Email content summary<br> - Delivery status<br>**Coordination Guidelines:**<br> - Maintain a clear, helpful, and professional tone<br> - Ensure successful handoff between agents<br> - Validate that an email address is retrieved before attempting to send<br> - Report any errors clearly with suggested actions<br> - Document key steps taken and outcomes delivered<br>**Connected Agents:**<br> - `database_agent`: Retrieves employee email based on provided details<br> - `email_agent`: Sends emails to employees using given subject and message body<br>Always provide the user with:<br> - Confirmation that the email was sent<br> - The email address used<br> - A brief summary of the email content<br> - Any follow-up recommendations if applicable</pre><p>once you create the the agents it will look like below</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Xo9wq8JAlV7ZeWbJjHLcXA.png" /></figure><p><strong><em>Step 3: Configure tool calling to the agents</em></strong></p><p>Now comes the secret sauce — <strong>tool calling</strong>.</p><p>For the <strong>database agent</strong>, we’ll integrate a vector database populated with a single PDF file that contains employee information. Since most vector stores don’t currently support CSV formats directly, we can simply convert the CSV into a PDF or Word document. This format difference doesn’t impact functionality because the system vectorizes the file content regardless of its original structure.</p><p>To demonstrate, I’ve converted a CSV file into a PDF that can now be used for embedding into the vector store.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*B8j67i21iP14jpyvD5ihMA.png" /></figure><p>let me know add this employee_mock.pdf file as vector store into the databse agent.In the agent configuration go to “Add Knowledge” and select “Files”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ru8zM-gTYVlAzumM986mKA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YTz9cHKf0ux01OpC0uLQAA.png" /></figure><p>Test the agent in playground</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*erMdXzpZTGFKf6BDj9j6pA.png" /></figure><p>Now configure the send email agent . So now we need to invoke an external tool which will send email.There are several ways of doing that using either logic app of azure functions. lets take the logic app route which is by far the easiet.</p><p>In email_agent configuration go to add actions and select Azure Logic Apps</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*adccSmQ5leeBH0miojZFbA.png" /></figure><p>You can use the templete given by Microsoft alreday to send Email using Outlook ( in most eneterprse cases you will be uinsg outlook 365 anyways)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pCH7j-PKptdrA65VPxRPMw.png" /></figure><p>Remeber to use the same resource group and subcription where your ai foundry resource is created</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zShFBAa6F4_43T7WK5RgSw.png" /></figure><p>Once you uthenticate with your outlook email a new new logic app connection will be created with the schema as below</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5UQIl1mUCR1InNptQzHhUA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*b5vloUJj9osmfHClHcNCyA.png" /></figure><p>Once you hit create a office_365 api connection and a logic app will be deployed in your resource group and you are good to go</p><p><strong><em>Step 4: test end to end in playground</em></strong></p><p>Now the final step we need to connect the sub agnets with supervisor.</p><p>go to the “connected agents” in the supervisor agent and add both your sub agnets</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iqXsrDipnoczIm1Mw_wBkA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/638/1*og4CQIHCNNw18LcVzGPNEg.png" /></figure><p>You can add some details how and when these agents will be called but that is optional .I have not added any description how these agents will be invoked rather rather deatils intruction to the supervisor how to delegate the task and expects the gpt-4o model to be intelligent enough to do the autonomous tool calling</p><p>Now is the time to test the system in action. Lets open the supervisor agent in playground and delegate him a task</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ay9EssiYUwE7w6y1W73q_g.png" /></figure><p>and this is the email I got</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yYXi-YnaKEnHtXZC2ExkWw.png" /></figure><p>🧩 <strong>What’s Next</strong></p><p>In <strong>Part 2</strong>, we’ll connect AI Foundry agents with a <strong>custom MCP server </strong>, allowing them to invoke your own tools and APIs — making your agents truly <strong>autonomous and extensible</strong>.</p><p>Stay tuned for more powerful patterns, integrations, and best practices.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=49b9e8df2780" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>