Multi-Agent Fashion Stylist with Connected Agents

[aprilgittens]

Introduction

Last week at Microsoft Build, we announced Connected Agents! Connected Agents is within the Azure AI Foundry Agent Service and enables you to connect multiple agents together with no custom orchestration or handed-coded logic.

You start with a main agent and then connect subagents. The main agent works behind the scenes to determine which subagent to activate based on the prompt you provide and your defined activation steps.

In my project, I have a main Stylist agent that coordinates with a Weather subagent and a Closet subagent to determine whether an item in my closet is appropriate for the current weather forecast.

Using Connected Agents has literally saved me SO much time getting the agents to work together - so much so that I created a video to show you all just how simple it is to set this up! 🙆🏾‍♀️

Watch here: Create Multi-Agent Apps with Connected Agents

Features and Screenshots

Technical Details

Here's how it works:

1. Pass the location and an item of clothing in the closet to the agent
2. Weather agent uses a Langchain parser to extract the location and pass the city into the query string for the API call to the Weather API
3. Closet agent retrieves the item of clothing in question based on the information provided
4. Stylist agent makes the final decision based on what's returned from the subagents

Note: The Closet agent is grounded in Markdown files for each item of clothing in the closet. Information included is the brand/designer, description, color, fabric & materials, care instructions. Whether an item of clothing is "appropriate" for the weather is determined on the fabrication/material of the item of clothing.

Challenges and Solutions

For the demo in the video, I hardcoded the current location and temperature. However, I share a preview of the actual code that replaces the hardcoded values so that the API call could occur.

Other than that, this was a pretty simple configuration!

If anyone's interested in a code walkthrough of setting this up, feel free to let me know.

[techworld1510]

Hi @aprilgittens , Great Video of Implementation of Multi Agent using the Agent Service of Foundry.

I am looking to start with Code-first approach for building a Multi Agent architecture, is there any sample project which can be referred or it is recommended to create agent as using this Agent service and convert it to code?

[aprilgittens]

@techworld1510 thanks for watching!

Here's the doc that provides the portal approach as well as a code-first approach for C# and Python: https://learn.microsoft.com/azure/ai-foundry/agents/how-to/connected-agents?pivots=portal

[aroosachoudhary44-wq]

Hey everyone,

I’ve recently been diving into some open-source projects on GitHub related to speakers, audio playback, and sound processing, and I was surprised by how many creative repositories are out there! From ESP32 Bluetooth streaming setups to Python-based audio analysis tools — there’s something for every skill level.

While testing a few of these, I noticed many devs also face distorted or muffled sound during long runtimes, especially in DIY speaker builds. That’s when I found this simple online tool called Speaker Cleaner. It uses audio frequencies to remove dust or moisture stuck inside the speaker — and it actually works great for small devices and prototypes.

If you’re working on an embedded or IoT audio project, I’d highly recommend checking out some of these GitHub repos and pairing them with maintenance tools like this one. Clean sound = better testing results. 🎧