NeedACar | Devpost

Original prompt
First response
Follow up questions

Inspiration

We have always thought the process of purchasing a car can be overwhelming, especially when it involves talking to salespeople or customer service agents. Sometimes, you just want a simple, hassle-free experience. That’s what inspired us to create NeedACar. We wanted to build a solution that would make finding the right car as easy and stress-free as possible, without the need for long conversations or complicated steps. We aimed to provide a personalized experience where users can simply chat and get the car recommendations they need, tailored to their specific requirements. It’s all about convenience and simplicity, just the way it should be.

What it does

NeedACar is a chat application where users can ask for car recommendations based on specific requirements. NeedACar will then provide the user with different options with pictures and it even asks clarifying questions. The user can then ask follow up questions to further refine the search. Don't feel like typing? NeedACar also supports speech-to-text input.

How we built it

With our computers, our hearts, sweat and tears haha just kidding. First we started off by designing the system in a way to leverage SQL querying and semantic search with embeddings. We then designed the front end using React components.

Challenges we ran into

Our biggest issue was how to handle user queries, we needed some attributes to be passed as a SQL queries and some not to, so we needed to figure out a way to refine the SQL query.
The speech-to-text feauture was difficult to implement as it was our first time working on something like this.
Deciding how to format the cards for the carousel was also a difficult decision because we were not sure what information was relevant.
The accuracy of the search was also challenging as we needed to refine the LLM to a point where it was returning factual information consistently.

Accomplishments that we're proud of

We're proud of everything we've accomplished! We put a lot of effort into this project, and we believe the results really reflect that. We created a sleek, user-friendly UI, and we're particularly happy with the carousel of images and the speech-to-text feature. We're also really proud of the accuracy of the search functionality and how well it handles follow-up questions!

What we learned

We learned how to effectively combine both frontend and backend components to create a creative solution to our problem. We found a good balance between working individually to maintain progress, while also collaborating as a team. Along the way, we handled several challenging merge conflicts, which taught us a lot about using Git in collaborative projects.