Ask Muze

Inspiration

As students, we are faced with countless, overwhelming decisions at every moment of the day. We're stressed, overworked, and always short on time. Having all these decisions will inevitably lead to decision paralysis and overthinking. What if there was a tool to help take the pressure off making these decisions and instill confidence in our choices? This is what inspired us to make Ask Muze, a multi-purpose chatbot designed to bring us all, student or other, some peace of mind.

What it does

Given a specific dilemma from the user, Ask Muze uses conversational AI to engage in a dynamic, guided question-and-answer flow. It collects relevant user inputs through a series of targeted yes/no questions, then analyzes the responses to deliver a personalized, comprehensive, and context-aware recommendation designed to optimize the decision-making process we struggle with every day.

How we built it

We first explored OpenAI's API "assistants" feature to create a proof of concept. After determining that our idea was doable, we decided to create our backend using Flask to support the API in Python. For the frontend, we used React (create-react-app) and CSS for styling, while the UI was designed and visualized in Figma Dev Mode. The frontend and backend were connected via Axios. Additionally, we explored deployment options using GoDaddy (for the domain www.askmuze.co) and Google Cloud.

Challenges we ran into

To begin, we spent lots of time in the brainstorming phase of the project. We considered a large variety of topics but eventually decided on a simpler idea that we could implement efficiently, rather than a more creative idea that would be harder to execute. On the technical side, a general difficulty was initially linking the frontend and backend cohesively with Axios. We also attempted to use Terraform in our attempts at deployment, but the process quickly became convoluted. The entire process was also done without a docker file which proved to complicate things.

Accomplishments that we're proud of

We are proud of the learning curve we navigated to teach ourselves much of the software and tools used for Ask Muze, especially given the short time frame of the hackathon. Ultimately, we are proud of being able to implement something that is of tangible value and serves a wide range of diverse user demographics, addressing real-world challenges and enhancing decision-making processes for a broad audience.

What we learned

For some of us, this was our first hackathon and first time using API's. We learned the general structure of backend/frontend communication along with the interaction with the API. None of us were familiar with Flask but we managed to pick it up and integrate it along the way. Additionally, but of equal importance, we developed our transferable skills of collaboration, individual and team problem-solving, persistence, and assessing individual skills to contribute most effectively to the project.

What's next for Ask Muze

There were a few features we weren't able to implement to the fullest extent that we'd like to expand upon in the future. We'd like to refine the decision-making algorithm to a more optimized and efficient version, as well as keep a record of user data by retaining past conversations in a database to help with future conversations. There are more specific cases that Ask Muze could be useful for, such as cost-benefit analysis, weighted scoring models, risk-benefit analysis, and other technical applications. Additionally, with prompt injection being a growing limitation of AI models, we'd want to work towards keeping Ask Muze safe from this.