FlashGPT

Challenge yourself, review your notes using your flashcards
Challenge yourself, review your notes using your flashcards
Flashcard storage for you to access anytime, anywhere
Generate flashcards in an instant
Track your progress and frequency of reviews
Login Screen

Inspiration

Studying with flashcards can be one of the most efficient ways to memorise content. However, as busy students, we're too busy (and lazy) to create flashcards, so why not ask SOTA LLMs like GPT to do it for us? The time-consuming nature of flashcard creation inspired us to create a platform that is able to generate flashcards directly from notes, slides, etc, within seconds. Through FlashGPT, we hope students can go straight to revision, cutting out the hassle of manually creating flashcards.

What it does

FlashGPT takes pdf files from the user and turns them into a deck of flash cards for users to instantly use for their revision. FlashGPT parses the user's pdf files in the backend and sends the extracted text data to OpenAI's GPT model, where it is prompted to return data for the flashcards. The flashcards are stored on sanity.io, which users can access through FlashGPT.

How we built it

We used Figma to rapidly prototype the UI of FlashGPT. As for the tech stack, we used Nextjs for server-side rendering, Nodejs, sanity.io CMS to manage our data and lastly prompt engineered GPT through OpenAI's completion API.

Challenges we ran into

Given that Sanity.io is a CMS, we struggled to build the backend infrastructure with the lack of database api support (huge mistake on our part). However, we learned how to get past the lack of infrastructure through documentation of Sanity's groq queries, where we where able to set up CRUD operations between the application and the database. We also struggled with NextJS, serverSideRendering, where logic had to be separated by client-side and server-side.

What we learned

Learning how various frameworks like NextJs's server-side rendering, the Sanity CMS and even prompt engineering GPT were extremely fulfilling. We're happy to be able to piece together a functional application that we think others will enjoy using as well.

What's next for FlashGPT

We hope to add flexibility to FlashGPT, allowing users to not only generate flashcards but also edit GPT's generations and even create flashcards from scratch. Additionally, we hope to migrate to a db management system like Pocketbase or Firebasem, where maintaining FlashGPT's API calls and db management would be much easier and cheaper as well. We also hope to push more features to help the busy student cope with their revisions, through schedules, recently viewed and view frequency.