Inspiration
The first few months are a turbulent time for most new students, with thousands of new students scrambling to learn more about what to do and where to go. SamVsGoose.ai aims to provide a pair of guides that may help new arrivals better acclimate to the UMass community-- one by referencing official university sources and the other providing the voice of other students.
What it does
Allows users to have a text conversation with one of two chatbots at a time, toggled between by clicking on the title. Sam.ai is trained off official UMass websites, while Goose.ai is trained off reddit posts by UMass students. A FAQ section is provided on the left with a list of high-frequency topics that, when clicked, enter a sample question in the user input textbox. Details on Sam and Goose are provided in an info section to the right. Both of these can be shown or hidden depending on the user's needs.
How we built it
We used custom python scripts to scrape various separate groups of websites for Sam and Goose. The information was then stored in json files and sent to a mongoDB database. The front end was built with Svelte and the back end was built with Flask, with an OpenAI api as a middleman that dynamically transformed user input into function calls.
Challenges we ran into
When working with tons of data, we started getting rate limited by GitHub, OpenAI, and even by Reddit. We also faced integration problems with Nodejs and Flask, as well as formatting issues in the database.
Accomplishments that we're proud of
We are proud that we created a smooth, effective experience with many interactable features.
What we learned
We used many new languages and technologies we weren't familiar with, such as Svelte, Terraform, and MongoDB.
What's next for Sam Vs Geese
If we had more time, we would have included functional login and sign up pages, with robust security for users.
Built With
- amazon-web-services
- flask
- geese
- mongodb
- openai
- python
- svelte
- terraform
Log in or sign up for Devpost to join the conversation.