Inspiration
While writing Supplemental essays for our college applications, we realized that we needed to compare or fact-check the essays against credible sources, particularly school websites or college advising platforms. This inspired us to create SupplementSage, a tool that could help college applicants research schools and integrate their findings into supplemental essays. It was initially developed using LangChain tools for document processing and retrieval capabilities. However, we later recognized the efficiency and scalability offered by a search service like Cortex Search, so we integrated it into the app to improve the accuracy and speed of information retrieval for users.
What it does
SupplementSage is an AI-powered web application that aids college applicants in researching schools and incorporating their research into supplemental essays. Users can upload their personal PDFs containing essays and web links containing guides or details of a school's faculty. The platform also provides easy access to essay prompts from over 100 colleges, eliminating the need for repetitive copying and pasting. Some practical applications of SupplementSage include: •Gaining insights into an essay by providing the app with a personal essay draft and a website link containing writing advice. •Exploring the alignment between a faculty's values and specific essay prompts using details from a school's website.
How we built it
The app was developed entirely in Python, with the Streamlit framework for the user interface (frontend), Langchain for Document Processing and Cortex Search for Retrieval Augmented Generation. We utilized a Snowflake database containing two tables to store data: one for all supplemental essay prompts (scraped from the credible website "College Essay Advisors") and another for personal files and external content uploaded by users. The LangChain library was used to process the files, which were then uploaded to the Snowflake table using the Snowflake connector API.
Challenges we ran into
Initially, we faced difficulties navigating the Snowflake system, particularly with document operations through the Snowflake connector API. By thoroughly reading the documentation and watching online tutorials, we learned how to implement the necessary operations, including enabling users to automatically upload files directly from the application into their Snowflake table.
Accomplishments that we're proud of
We were thrilled to see our retrieval logic perform flawlessly and successfully deploy the application for practical use.
What we learned
We realized that we could learn a lot from reading other people's code. This genuinely helped us in solving our issues. For instance, we realized that we could use a multi-retrieval approach by separating the tables for essay prompts and external information. This improved the RAG process and allowed the app to perform more efficiently.
What's next for SupplementSage
We aim to onboard as many users as possible to gather valuable feedback. This will help us enhance the application with new features and make it even more user-friendly.
Log in or sign up for Devpost to join the conversation.