BoilerCompass

• 

  Front page of BoilerCompass

Inspiration

We created this to store all the important Purdue information within one place and be able to ask anything about what's going on at Purdue.

What it does

The user would ask it a questions in the chat box and the AI would go through it's database we've provided to give an optimal answer for the prompted queries.

How we built it

We used node.js with TypeScript to create the frontend of the application. Then we used Python web scraper that would pull all the data from websites regarding Purdue. We then linked a Deepseek AI model that would read all the data from the websites and put it into a response for you in the chatbot.

Challenges we ran into

Occasionally the system would break each time we run the web scraper for unknown reasons. Not only that but each web scraper would take 5-10min to finish its task so it was a really slow process.

Accomplishments that we're proud of

Was able to find a free AI model that is smart enough to format questions and not repeat the same thing over and over again.

What we learned

• We learned how to feed an AI model data and use its responses to create a conversation for the user and the model.
• Learned how to get frontend and backend to talk to each other through an API endpoint.
• Run a full stack application locally through two separate server processes in 2 different terminals.
• Made a LLM that would force its answers on a specific set of data.

What's next for BoilerCompass

• Adding an apartment searcher that takes into budget, location, roommates, and bathrooms.
• Add a GPS functionality to help direct you to your location
• Selection of transportation (Veo, bike, walking)

Built With

• css
• html
• node.js
• npm
• python
• react
• typescript
• uvicorn