Inspiration
We use LLMs for summarizing meetings, coaching after meetings, and providing real-time information from meetings. What if we just had them do the meeting itself?
What it does
It serves as a proxy for each team member to have the meeting on behalf of the team member. It uses notes and uploaded documents to determine what to say. There is a meeting creator who can invite members to "join" the meeting. The join by adding content and important things they thing will be valuable to the meeting. The creator can then run that meeting at anytime to get a log, summary, and task list.
How we built it
It is a flask app with autogen and llamaindex that also just goes straight to an inference endpoint via the llama-3-70b inference endpoint mindsdb set up. Also added AgentOps which turned out to be useful as I'm not super experienced with agents
Challenges we ran into
I was new to autogen so I struggled with finding the golden path working with it and hack around a lot
Accomplishments that we're proud of
Glad to have something to play with that ships
What we learned
I think I definitely start to understand the deeper challenges of the nondeterministic nature of agents specifically with LLMs.
What's next for Automeet
It is strange, but at first it was just kind of funny, but using it starts to actually feel a little bit useful. Like there are coordination and other things that are not just calendar invites etc. I think I'll probably develop it a little further
Log in or sign up for Devpost to join the conversation.