AIDocs

Boost your assignment writing with our advanced notebook. It effortlessly blends AI-driven suggestions from your uploaded materials and ongoing work, taking your writing and research to the next level

Comment

Inspiration

  • With the recent artificial intelligence growth, our team is excited to explore the capabilities of large language models and how we can integrate them into our workflow. AIDocs is just one example of many ideas we had for large language model integrations!
  • The process of working with information, taking notes, and compiling thoughts together was out of date and could benefit significantly from the large language model revolution.

What it does

With AIDocs, you are able to create and edit a text document and chat with a GPT model simultaneously! Through the selection of documents, the GPT query automatically includes pre-filled context to aid in the generation of a more relevant response.

How we built it

Our tool utilises both a frontend component and a backend component. The frontend component is a React application hosted on Amazon Web Services and the backend component is a Flask application hosted on Google Cloud Platform. Along with that, we utilised various pre-existing libraries including React-Quill, Tesseract.js, Axios, etc.

Challenges we ran into

We wanted to use prompt engineering to use the LLM to determine whether the prompt it is receiving is asking it to edit the central document. However, we were unable to complete this as outputs from GPT3.5 were unstable. We plan on looking into different models as we continue our development.

Accomplishments that we're proud of

  • We built an entire working application which exceeded our initial MVP goals
  • Implementing the source_document logic in the frontend in order to properly query the backend api, which we built all by ourselves
  • Our first-time React developer was able to learn and contribute a significant amount to the frontend :)
  • Our other front-end developer is just happy overall

What we learned

  • Our first-time React developer learned... React
  • The backend team learned to write an API for the first time
  • Learned more about LLMs and the importance of context in queries
  • First time developing a ChatGPT integration!

What's next for AIDocs

  • Add the ability to web scrape an entire link automatically, adding context for queries
  • Add online collaboration feature for multiple users
  • Add file system and authentication, similar to Google Drive / OneDrive
  • Add more file manipulation for AI to manage central document
  • Add diagram manipulation from the AI chatbox
  • Gain a userbase :)
Share this project:

Updates