Example of optimized schedule produced from user prompt

Inspiration

People plan their days in lists but execute them on maps. Coordinating multiple locations with fixed arrival times (school drop-offs, work start times) is mentally taxing and error-prone. When I first start college, I tried to balance school and life with a rigid study schedule, but I realized that it was very difficult to plan and stick to a rigid schedule for such a spontenous life.

What it does

AI Day Planner is an application that accepts tasks with locations and time constraints through a chat bot in a conversational tone. It will then optimize the tasks based on locations, time constraints, and relation to other tasks. The chat bot will then generate a schedule and show the route of your tasks on the map.

How we built it

AI Day Planner was built with React and JavaScript on the frontend and Gemini and featherless.ai on the backend. It also utilizes Google Maps API to map the routes for the tasks on the screen.

The user inputs their daily tasks in a chat bot in conversational tone, and it is then sent into Gemini to organize the conversation text into a list of JSON objects representing each task and its time constraints, locations, and addresses.

Then, these JSON objects are passed in to featherless.ai to determine the most optimal order of tasks based on the user's location, time constraints, and location in relation to the other tasks. There is a fallback algorithm that orders the tasks based on flexible tasks and fixed tasks in case the featherless.ai call fails.

Challenges we ran into

The biggest challenge I ran into was figuring out how to get the Gemini prompt to correctly extract the address and location from the conversational input. Some tasks such as "stop by the grocery store", don't have a clearly visible location.

The solutions I came up with were:

Including examples of JSON objects in the Gemini prompt
Having a set of mappings in the Gemini prompt (Ex: grocery store -> HEB, Memorial Student Center -> MSC)
Passing in the user's location and current time into the Gemini prompt ## Accomplishments that we're proud of I am proud that I was able to turn my idea into a functioning MVP in 24 hours. This is my first hackathon, and, while I wasn't able to create everything I wanted to, and I had to make some sacrifices to meet the deadline, I am pround that I was able to get an general MVP made by myself for my first hackathon. ## What we learned Spending more time on a plan and a step by step blueprint could be more beneficial that diving right in and coding. I originally throught I had to get started as soon as possible to meet the deadline, but I feel I would have been more successful if I planned out more in the beginning.

I also learned that it is typically better to build larger projects with a team, and maybe could be beneficial to simplify your project idea if going solo.

Technically speaking, I learned alot about React, APIs, and utilizing AI tools in coding projects. While I had some experience with React and APIs, utilizing AI tools in coding projects was completely new to me and I definitely learned a lot from this project.

What's next for AI Day Planner

In the future, I want to build out this project more. There were a lot of compromises I made to meet this deadline, and in the future I want to built out the optimization algorithm better. I also want to add more features such as, exporting to calendar, and uploading an ics file to build your day around.