AutoGen/AutoGen Studio Community Resources

[victordibia]

Using this discussion thread to curate a list of AutoGen / AutoGen studio learning resources created by the community in addition to the official documentation available

Title	Link	Author
Introduction to AutoGen Studio v0.5 | Build AI Agents 🤖 from Scratch	Video	@mayank953
SEO-friendly content generation chatbot with AutoGen	Video	@M-H-Amini
Arxiv paper hunter and summarizer chatbot with AutoGen	Video	@M-H-Amini
Chat with YouTube videos using AutoGen	Video	@M-H-Amini
Chat with datasets using AutoGen	Video	@M-H-Amini
Team of agents for code development in AutoGen	Video	@M-H-Amini
Agents lively debate in AutoGen	Video	@M-H-Amini

If you have created AutoGen/AutoGen studio resources and would like them listed here, comment below.

[M-H-Amini]

Hi @victordibia

Thanks for creating this conversation. I have a playlist in my YouTube channel dedicated for Agentic projects with AutoGen. Here's the link of the playlist: https://www.youtube.com/playlist?list=PLEe-UC96P-yaP7-jz3HLrieQdNNFuy_xN

And here's the list of projects:

• AutoGen Crash Course with 6 Projects: https://www.youtube.com/watch?v=YeJG_H-sw84&list=PLEe-UC96P-yaP7-jz3HLrieQdNNFuy_xN&index=1&t=1s
• SEO-friendly content generation chatbot with AutoGen: https://www.youtube.com/watch?v=Cvrf9MGAAP4&list=PLEe-UC96P-yaP7-jz3HLrieQdNNFuy_xN&index=2
• Arxiv paper hunter and summarizer chatbot with AutoGen: https://www.youtube.com/watch?v=rd1NFIgr9M0&list=PLEe-UC96P-yaP7-jz3HLrieQdNNFuy_xN&index=3
• Chat with YouTube videos using AutoGen: https://www.youtube.com/watch?v=ozd85AtvgiQ&list=PLEe-UC96P-yaP7-jz3HLrieQdNNFuy_xN&index=5
• Chat with datasets using AutoGen: https://www.youtube.com/watch?v=ss6lown4_ks&list=PLEe-UC96P-yaP7-jz3HLrieQdNNFuy_xN&index=7
• Team of agents for code development in AutoGen: https://www.youtube.com/watch?v=8INMwnI7pJg&list=PLEe-UC96P-yaP7-jz3HLrieQdNNFuy_xN&index=9
• Agents lively debate in AutoGen: https://www.youtube.com/watch?v=c-WAkBEla48&list=PLEe-UC96P-yaP7-jz3HLrieQdNNFuy_xN&index=11

[victordibia]

+1. Updated

[dheeraj-sachdeva]

Hi Victor, Autogen team, Autogen enthusiasts,

I'm using Autogen v0.4.x in a web app where the /ws/chat route defines a _user_input() function tied to a UserProxyAgent. This input function works well during a RoundRobin GroupChat, allowing the user to indirectly interact with multiple assistant agents via the team chat.

However, after the RoundRobin chat completes, I want to let the user continue chatting directly with individual assistant agents (e.g., to follow up with one agent's response), without re-entering the GroupChat flow. But since the input function is still connected to the UserProxyAgent, the next user input continues to route messages to the GroupChat.

My constraints and goals:
I don't want to recreate the team or agents — just transition cleanly.

Questions:
What is the recommended way to switch from a completed GroupChat to direct chats with individual agents?
Is there an officially supported way to reset the UserProxyAgent state or toggle its routing behavior?
or Should I be using human-in-the-loop feature here? If so, how?

Any best practices or pointers would be greatly appreciated!

Thanks for the awesome framework — it’s been great to prototype multi-agent systems with it.

[victordibia]

So your setup has custom/niche behavior.

Here's what I would recommend: build a custom orchestrator/Python script that manages your agents and the behaviors you care about.

Here are things your orchestrator should do:

• Keep track of some type of response state (e.g., team or agent). The UI/user or some logic can define whether it's in team or agent state (and which agent should be responding)
• If in team state, you can manually write your own loop: keep a list of messages and direct which agents respond using either logic or an LLM. You can also create a GroupChat here.
• If in agent state, route to the agent by calling its run method directly or by adding it to a new GroupChat with a new proxy

Another option is to:

• Use GroupChat as usual
• In the UI, select which agent you want to chat with directly, then recreate a new version of that agent with the [same state](https://microsoft.github.io/autogen/stable//user-guide/agentchat-user-guide/tutorial/state.html) (call agent.save_state(), then load the state on the new version)
• Chat with the recreated version of the agent either by calling its run method directly or by adding it to a new GroupChat with a new proxy

Either way, you'll need to implement custom logic!

Let us know how it goes!

[dheeraj-sachdeva]

Thank you, Victor! I missed the obvious that you reminded: Use GroupChat with 1 agent. Cloning of original agent w/ its state preloaded worked out perfectly.

Can you point out to me how to best use Memory feature between Agents and Teams? I'm thinking that at the end of the group chat, I somehow generate some basic facts and add them to the Memory, so that next individual agent group-chat can inherit this memory to allow the user to seamlessly continue the conversation without repeating what the core topic was or some of the basic facts (I intend to reload the team state into the individual agent group chat as well but that's sometimes too much context, where as I need some basic facts to be available to the agent groups.

[M-H-Amini]

Hey @victordibia

Hope you're doing well. I was just wondering if there's any plan to put the community resources in the examples page in AutoGen's documentation as it would be way easier to get discovered.

I would be more than happy to help with that if it is decided.

Kind regards and thanks for your kind consideration :)

[victordibia]

None for now.
GH discussions are quite discoverable.

[M-H-Amini]

Thanks for the swift reply :)