The LED Chatbot project draws inspiration from the current climate of AI use in the classroom. Embracing the challenge of building a chatbot that utilizes only textbook information, we aimed to create a visually striking and innovative user experience by incorporating bright neon lights and colors into the design, due to the Hackathon theme.
The chatbot is designed to extract information exclusively from a given textbook and provide relevant responses. We took textbook PDFs, transformed them into queryable embeddings, and stored them in a vector index that our OpenAI calls can extract data from.
Our journey was made possible by combining the power of React for the frontend, Flask for the backend API, and UnstructuredIO, Llama-Index, and OpenAI for the data extraction and query.
Using Docker containers to separate the two sides of the project – frontend and backend – gave us difficulties with communication. This technology was new to most of our group members, and we spent a lot of time debugging.
We're immensely proud of successfully blending technologies such as React, Flask, and OpenAI into a cohesive and visually appealing chatbot. The project represents a seamless fusion of functionality and aesthetics, achieving a balance between information retrieval and an engaging user experience.
We delved into the world of chatbot development, React, Flask, Docker, and OpenAI API, gaining valuable insights into the interplay of frontend design and backend logic.
We are looking into allowing user input textbooks which our current chatbot does not allow. This is due to the longer computation time and resources it takes to extract data by user request. We would also want image recognition during the data extraction phase.
- React
- Flask
- OpenAI API
- Llama Index
- Docker