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. 🎧

[swertresresultshistory]

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. 🎧

[swertresresultshistory]

Open-source audio projects really are a goldmine, especially ESP32 stuff. Distortion issues are super common in DIY builds, so that Speaker Cleaner tip is actually pretty useful. Clean hardware definitely makes debugging software way easier.

[maryamayesha8800-lang]

You’re absolutely right — GitHub has become a goldmine for audio, speaker, and DSP-related open-source projects, especially in the ESP32 + IoT space. Issues like distortion or muffled output in long-run DIY builds are quite common, so combining proper signal processing techniques with maintenance tools can really improve testing accuracy and overall audio performance.

Thanks for sharing the Speaker Cleaner tool as well — interesting approach for handling physical sound issues in small devices and prototypes. Definitely worth exploring alongside software-based audio enhancement and filtering solutions.

Overall, this is a helpful direction for anyone working in embedded audio systems or experimental speaker projects.

[maryamayesha8800-lang]

Thank you for the detailed insights — you explained it really well! 👏

You’re absolutely right that distortion and muffled output are common challenges in ESP32 and IoT-based audio projects. Combining proper DSP techniques with hardware-level optimization can significantly improve long-term performance.

I also found that alongside testing tools, having a clear understanding of digital optimization and performance strategies can make a big difference. For those interested in exploring more about digital solutions and smart tech approaches, this resource is quite useful:
🌐 https://mjunaidtariq.com/

[zohanalibabar2028-droid]

Hi everyone,

When working with AI applications or multimedia projects, audio clarity can make a huge difference in testing, output quality, and user experience. Many developers focus on software optimization but overlook the hardware side — like speakers, earphones, or microphones — which can affect results.

I found SpeakersCleaner
to be a really useful resource for maintaining and troubleshooting device audio. It provides practical guides for cleaning speakers, troubleshooting sound issues, and improving audio performance on various devices.

Even small adjustments in hardware setup can complement software improvements and help achieve more accurate AI audio processing or multimedia results.

Would love to hear if others have tips for balancing device maintenance with AI and audio workflows!

[swertresresultshistory]

Totally agree with this. Hardware is often the silent bottleneck in audio and AI projects. You can optimize models all day, but if the speaker or mic is compromised, results will still feel “off.”

Regular maintenance plus simple tools like that can really help isolate whether an issue is software or hardware-related. Curious to see what workflows others use too — especially for long-running tests or prototypes.

[charlesedward23400-ui]

Thanks for sharing this wonderful post! If you start feeling a bit hungry during or after work, grab something fresh and tasty. Eating healthy keeps your energy up and helps you perform at your best. E