Eventropy

Our avatar :)

Inspiration

As avid hosts, we know organizing events can be tedious and often repetitive. Eventropy uses orchestrated AI agents to facilitate the creation and organization of events. Eventropy can be used by large companies for corporate events and individuals for personal get-togethers. We sought to automate as many of the repetitive tasks of event creation with agentic systems. We believed this app could specifically hit all categories for this hackathon, as it is placed in a unique position to unite communities through event creation. Eventropy can elevate the black community and organizations such as NSBE, as well as other organizations that want to unite people.

Moreover, AI agents are one of the fastest-growing technologies not just today, but of all time. The implementation of AI agents in fields other than software engineering is an avenue for significant profit and can generate a substantial amount of investment incentive. Not only that, the tech market for "Even Management as a Service" is predicted to go from 22B today to 56B by 2032, over doubling in the next 6 years.

What it does

Currently, Entropy allows you to create and plan events. Following the creation of an event, AI is used to split large events into smaller tasks based on the information provided. These tasks get split into four subsections: guests, venues & catering, entertainment & logistics and miscellaneous. Here, each subsection has its own agent who's personally designed to help you with the tasks in its domain. You can interact with these agents via chat, and thanks to the tools at their disposal, they can directly modify event details and task notes.

How we built it

We built our project as a desktop app on top of Electron. We used React (TypeScript) to render components and Supabase for persistent event and task storage. We also used Node.js and Claude for creating and orchestrating AI agents.

The AI orchestration works like this: When a conversation is initialized, the main process instantiates a fresh Claude conversation with a system prompt tailored to that agent's specialty. So each agent has an isolated thread. When a user sends a message in the category chat view, the Electron main process receives the request and initiates a Claude tool-use loop. Claude is given access to six TASK_TOOLS: list_tasks, read_task, write_task_body, update_task_status, read_event_details, and update_event_details. These tools allow Claude to directly query and mutate data in Supabase.

Challenges we ran into

Connecting to other tools like Gmail, the ability to search and Google Calendar proved to be too difficult, and we ended up leaving it out. Moreover, getting independent AI agents proved to be difficult as well, but we managed to succeed in this regard.

Accomplishments that we're proud of

We're really proud of this project because we believe this application is something that we would truly benefit from using when trying to organize events. It's also the first time we've created agentic systems to help in real-world tasks.

What we learned

We learned how to handle and orchestrate independent AI chats and utilize loops to create functioning and intelligent agents.

What's next for Eventropy

We're excited to continue working on Entropy. We have a lot of potential features in mind, but the immediate next step would be to give the agents more freedom. This would mean: allowing them to search, book appointments, message people and much more.

We would also like to tailor this project to organize community events to promote unity among marginalized communities.

Another way we can improve is by utilizing multiple LLMs in our project.