FeynFox

Inspiration

The 2 most important things in the world are time and knowledge.Our mission is to implement the famous Feynman method with AI to increase productivity in learning new topics. With FeynFox, we aim to provide a way of using the Feynman method more effectively and independently.

What it does

FeynFox is built based on the idea of the Feynman method, where you reinforce your knowledge by explaining to others. First, the user needs to create an account with your Google credentials. After that, they will be directed to the profile page, where they can upload their learning materials. Then, FeynFox will generate talking points based on the given materials. The user will now select a talking point and start teaching FeynFox by talking into the microphone and the data will be converted into text by a built in speech to text system. That data will then be analyzed and a thorough report will be generated, giving the user the idea of where they did good and where they need to improve.

How we built it

We used React Vite as the main framework for the frontend, and Supabase as the database and authentication. We also used FastAPI to implement the embedding functions. We used the pgvector extension on Supabase's postgres database to store the embeddings in SQL column and run similarity search on it. For designing we also used Figma for collaboration.

Challenges we ran into

One of the biggest challenges we ran into was running some of the langchain’s libraries on react+Vite. We kept on running into Buffer problems and eventually, we gave up on TypeScript and used Python for running all the lanchain libraries.

Accomplishments that we're proud of

We are proud of the quality of the UI design of our app, how well the functionalities turned out, the integration of Supabase and Langchain, the accuracy of the text-to-speech and vice versa, and how well we were able to collaborate.

What we learned

We introduced new technologies in our tech stack: Flask, Supabase, and voice to text api. We had multiple problems come up in our database Supabase with inserting and querying the data from the database. To make it easier to embed text into Supabase, we used Flask since it is written in python. Flask was easy to pick up and Python made the backend easier for us compared to past technologies we have used. In addition, we used several apis that we have not used before.

What's next for FeynFox

For the future we plan to expand the media that we can accept on the platform to images and other document types. This could also expand to it having a huge pool of documents so one user's topics might help the context of some other user as well.

Built With

• claude
• langchain
• openai
• react
• supabase