evento.ai

AI-powered integration that reads your GroupMe messages and automatically adds detected events to your Google Calendar.

Comment

Inspiration

The idea for evento.ai came from our own life as over-committed Lehigh students. Every semester, we joined 20+ clubs, genuinely wanting to attend their events. But event details get instantly drowned out by memes in GroupMe. We live by my Google Calendar; if an event isn't there, we forget it exists. We decided to build a system that would automatically organize my schedule, fixing my life and the lives of every student facing FOMO.

What it does

evento.ai is a serverless application that turns casual GroupMe chat messages into perfectly structured Google Calendar events, automatically.

It provides the necessary intelligence layer that simple automation tools like IFTTT lack. Our application monitors club chats, uses Generative AI to understand the conversation, extracts the event details, and pushes them to your primary calendar instantly.

How we built it

We set up a secure AWS Lambda function (Python) exposed via a single AWS API Gateway webhook endpoint. We securely stored all sensitive credentials (Gemini API Key, Google Refresh Token) in an Amazon S3 bucket and rigorously configured the Lambda's IAM Role to read them with the required least privilege. When a message hits, we engineered a Gemini Prompt to execute semantic extraction. The AI must detect the event and return the title, resolved date, time, and location as clean JSON. The code uses the long-lived Refresh Token to get a fresh Access Token and sends a direct HTTP POST request to the Google Calendar API, creating the event.

Challenges we ran into

(we ran into so many lol) We spent significant time resolving complex Git conflicts due to the unique "unrelated histories" error, forcing us to rewrite the branch history. Our code initially failed because the Python requests library was missing. This required us to successfully package and attach an external library using a Lambda Layer to our function. The Gemini AI kept wrapping its perfect JSON output in ugly Markdown code blocks (json ...), which crashed the Python parser. We fixed this by coding a robust Python cleanup function to strip these wrappers before loading the data.

Accomplishments that we're proud of

Successfully built an AI solution that outsmarts commercial tools (IFTTT) by handling unstructured data. Developed a scalable, secure Serverless Architecture on AWS Lambda, API Gateway, and S3, maximizing our use of the hackathon sponsor resources. Achieved 100% end-to-end functionality from a message in GroupMe to a created event in Google Calendar, solving a very difficult real-time integration problem.

What we learned

We learned that clean architecture is everything. We gained deep, practical experience in IAM security protocols, managing complex external Python dependencies in a serverless environment, and advanced Gemini prompt engineering to force precise JSON output.

What's next for evento.ai

Our next steps are to integrate the user-facing setup: fully coding the front-end webpage to automate the multi-bot creation process, which will allow a single user to monitor 20+ clubs automatically with a single click.

Built With

Share this project:

Updates