pathos

Inspiration

PATHOS (noun): the emotional response or experience that connects to real-world struggles and challenges.

As a team of four software engineers, we have all faced the trials and tribulations of unreliable AI outputs. Whether it’s unrecognizable ChatGPT prompts or the Google AI dominating the top of search queries, sometimes all you need is a genuine human perspective. With one quick search, our model sifts forums like Reddit and StackOverflow to receive multiple raw solutions that come from real people–just like you!

What it does

Pathos takes in a search entry related to anything from the programming world and searches Reddit and StackOverflow for threads that have related queries. From there, it analyzes which of the comments is most related to what the user expects as an output. It selects multiple at once to read from and also provides a hyperlink to guide users to the original post.

How we built it

This search engine queries the Reddit and Stack Exchange APIs for questions and comments related to the user prompt. We used Gemini to parse the user entry into a searchable phrase, and used a FAISS index in order to rank the responses by similarity search. We used Python and Flash for the backend, and React and Tailwind for the frontend.

Challenges we ran into

Our Achilles’ heel was trying to match the information that our API returned to our search query. In stages of our initial testing, much of the API data would go down unrelated rabbit holes. We tried to optimize our search query through natural language processing with the Python NLTR library, but finding relevant computer science related filtering was outside of our time frame. Ultimately, through filtering mechanisms and keywords in order to streamline our data, we ended up utilizing a similarity search to handle a large number of comments and to find the most relevant ones.

Accomplishments that we're proud of

Our treatment of data had a long journey from beginning to end, and we are really proud of making our pipeline both smooth and timely. We also spent a lot of time on the UI design and are really proud of what we came up with.

What we learned

We learned a lot about the software development process, and how to utilize different languages together to create a product. Also, we got really good at terminal commands. Git. And debugging.

What's next for pathos

Looking into the future, the next step would definitely be to integrate more sources such as Github Discussions, API documentation, and Geeks4Geeks in order to create a more comprehensive search. Additionally, we would consider pioneering towards greater user customization to tailor results to users’ hardware and software needs.