SimpleNotes

Inspiration

Many students don't like taking notes for their classes, and they often just plug their online texts into chatGPT before a test to summarize them. This method of "studying" does not lead to retention of information, meaning that most of what you "studied" will be forgotten the next day. Busy students may not have time to write all of their notes, so we aim to take a review-based notetaking method (Cornell Notes) and meet them halfway.

What it does

The program prompts the user for a segment of text. Then, it asks the user to skim through the text again and ask some questions (this requires actively thinking about the text). It answers these questions using the llama3 AI model (running locally). When the user is done asking questions, it formats the data into an easy to read/review document.

How we built it

We used the Llama3 AI, running locally, and langchain, to parse through user-submitted text and answer questions about it. We then used fpdf to format the output into a pdf that the user can view, save, and edit later.

Challenges we ran into

A major setback was downloading and installing a large language model locally. After hours of watching tutorials and researching different chatbot options, we decided on Llama 3. Another challenge we encountered was that Llama 3 has a slow running time, which meant that testing took longer than it could if we had access to a free and faster chatbot.

Accomplishments that we're proud of

We are proud of our ability to utilize a large language model in Python and the fact that we were able to turn the data into a pdf.

What we learned

We learned a lot about LLMS, various python libraries, and how machine capability affects LLM performance when running locally

What's next for SimpleNotes

We'd like to port the output to LaTex code so that it can be styled better. We might also consider purchasing and API key so that this can run faster.

Built With

• llama
• python