Smart-Browse

Inspiration

When shopping for items, people end up using various resources like ChatGPT, Reddit, Amazon, BestBuy, Target, eBay, and other sites to figure out what is the most optimal product that fits your niche use-case. This process becomes tedious, mundane, and requires so many different tabs and spreadsheets to properly do your research before you make a purchase that really matters to you. We wanted to solve this issue by creating a dashboard that allows you to shop more smartly using AI.

What it does

The system first takes in a prompt for what kind of item you want to look for. Let us say you are shopping for a speaker.

Based on this, it searches through Reddit and its general knowledge base to figure out a set of dynamic questions to ask you about the specifics for that item. Continuing on the Speaker example, let us say an example question is "Where would you see yourself using this speaker?". The choices for such a question would be like so: "Inside the house", "While performing physical activity (eg. biking, running, basketball)", "Beach parties", etc.

Based on your choices to each of the questions, it will look through different websites like Amazon, BestBuy, Target, and eBay to get a list of potential items that fit the description of a Speaker. It then filters out the items that seem the most relevant to your needs and shows the choices it recommends. It creates a "deck of cards" for the items that you can actually use.

You can bookmark the items you like in your own lists, and you may make this list public for others to also look through and see what kinds of items people with similar needs might be able to buy.

How we built it

We used various different webscrapers, API calling functions, a Python Flask server, MongoDB for authentication and saving public/private lists, ElevenLabs's Text-to-Speech feature, and dynamic UI to allow the agent to control different parts of the UI to prompt the user in interesting means while maintaining simplicity.

Challenges we ran into

Some challenges we faced along the way were scraping data from websites like Amazon or eBay to be able to get images, prices, reviews, and other information that we can then use to display within our website. By using different technologies like Playwright, we were able to get the information we wanted and provide our users with the full experience.

Another challenge we faced was ...

Accomplishments that we're proud of

We are proud of building a user experience that chatbots like ChatGPT or Gemini cannot provide on their websites. By creating a more intuitive UI system that allows users to use sliders, checkboxes, and radio buttons while also displaying information similar to how "Spotify Wrapped" works and charts to show different specifications, we are able to give users a visual representation that other websites or chatbots cannot accomplish.

What we learned

We learned how to webscrape from websites, call different APIs, deploy our project, create a frontend-backend application from scratch, and create dynamic UI that agents can make to give a better user experience.