StudyMuse

Architecture Diagram of StudyMuse

At StudyMuse, we understand that students are often overwhelmed by the sheer volume of text in their notes—as a result, finding specific information can be time-consuming. To address this, we've created a platform where students can elevate their notes to a new level of "smart" interactivity through retrieval augmented generation (RAG). Our website invites students to upload their study materials and engage with an interactive study companion, making learning more dynamic, efficient, and enjoyable.

Upon logging into StudyMuse, the student will be guided to a page where they can upload notes of all varieties. They can then pose questions to which StudyMuse responds with relevant information extracted directly from all their uploaded notes via RAG, creating a tailored and responsive study session.

We use MongoDB to house the user’s notes and various texts that they upload, ensuring secure and scalable data management. When a user asks a question, LlamaIndex is used if a relevant response can be generated from any previously uploaded text, otherwise a Claude response is returned, prefaced with a statement that the information does not come directly from the notes, and an option to add it to the notes.(We have attached an architecture diagram in the project media section)

We weren’t able to implement a few functionalities yet, leading to logical next steps: Allowing users to upload different forms of data (images, pdfs, slides, etc.) Creating user-specified workspaces for better organization Generating flashcards and practice problems based on notes

Once these steps are complete, a 1-1 personal learning ChatBot can be trained with the same RAG mechanism to help students learn more effectively.

The Tech Stack we used was: MongoDB LlamaIndex Anthropic’s Claude GPT Flask JavaScript HTML & CSS Bootstrap Python