ACM - Agent based Car Management

Inspiration

Listening to many of the amazing agent software options that were available to us, me and one of our partners decided that, because we are students who commute a long time for school through our cars, we decided to build an application that assists drivers with advanced intelligence that takes inputs on real-life data to help us get better gauged towards our driving experience, car maintenance, and overall savings.

What it does

The application takes in real-life data from the web when it comes down to fuel costs around gas stations in a local area based on where you commute. It also takes car sensor data that relies on sensors that the phone captures based on your current drive. Based on that, it gives better maintenance estimates and gauges a better understanding of car maintenance, when repairs need to happen and how much spending is actually done towards your car.

How we built it

We built much of the app infrastructure through SwiftUI. That's where all the data and information is displayed. Express is used for the backend, and when it comes down to agents, we mostly use them for the fuel maintenance estimates that give a better understanding of the things that the app presents.

For external data that we use, we use Tiny Fish for driving data for fuel insight around the local area. We also used an OpenAI key in order to get better, more unique, direct, and personalized responses based on the data we gathered from Tiny Fish. We structured all the actions to be displayed as best we could on the app itself.

Challenges we ran into

One of the main challenges that we ran into was how we could integrate WAPI with Tinyfish. The data that is publicly available on websites like Gas Prices, which show the gas prices, was too hard because Tinyfish wasn't able to retrieve all websites' data and scrape it. We had to redirect it to only a specific URL which accepts scraping by Tinyfish, and that is what we did.

Just to mention, the website that we had issues with was Gas Buddy, because whenever we tried to access the gas prices, they were dashed out because they had many blockers that we ran into that couldn't be accessed by Tinyfish.

Accomplishments that we're proud of

We started off with a consumer-based application that relied on simulated data from our car sensors, but we still do use that for maintenance. This time we were able to use the include agent for the fuel inside that actually applies the agent Tiny Fish, along with the express backend for both agents to successfully produce the output to be expected.