FinanQ

Unlock Financial Wisdom Instantly

Comment

Inspiration

Our inspiration behind this project stemmed from the desire to democratize access to financial information and empower individuals to make informed decisions in real-time. We recognized the overwhelming complexity and often daunting nature of the finance world, where individuals may struggle to navigate through vast amounts of data to find relevant insights. By leveraging the wealth of data available on platforms like Kiplinger and harnessing the power of Convex for backend processing, we aimed to create a solution that simplifies this process. Our goal is to provide users with instant access to accurate and actionable financial insights, regardless of their level of expertise or background in finance. Ultimately, we aspire to level the playing field and empower users to take control of their financial futures with confidence

What it does

Our project leverages data from Kiplinger to provide instant answers to financial queries. By integrating with Convex for backend processing and Clerk for authentication, our platform delivers accurate and personalized insights on stocks, investments, and more, empowering users to make informed decisions with ease.

How we built it

Tech stacks: Python, Flask, Typescript, React, Convex, Large Language Models, Google-Gemini, Clerk

  1. We first take an input query from the user which is then passed to the convex function. This function then calls our backend server which has the business logic. The backend is built using Python and Flask.
  2. In backend, we call the Kiplinger search API to fetch articles relevant to the user query. The articles are then scraped using beautifulSoup to get relevant texts.
  3. The user query along with the relevant text is then passed to Google's Large Language Model: Gemini to fetch the appropriate answer to the user query. We use Gemini's API to make this call.
  4. In between, we have also integrated Clerk's authentication to prompt users to sign up/log in before submitting the query.

Challenges we ran into

  • The scraping of articles from Kiplinger website is time taking and often makes our website slow.
  • The Large Language Model: Google Gemini often hallucinates and give wrong answers.

Accomplishments that we're proud of

  • We are proud of improving the UI from a simple HTML page to user-friendly React application.
  • Initially, we had used GPT API but in order to keeps things economical we decided to use Google-Gemini.
  • Understanding the working of Convex and Clerk was a challenging task too and we are proud of overcoming it. We understood the importance of going over documentations.

What we learned

  • We learned about the functioning of Convex and Clerk. It was inspiring to see how simple various tasks can be made with the help of these tools.
  • Importance of reading and following documentations.
  • The limitations of Large Language Models like GPT and Gemini.

What's next for FinanQ

There are few things that we would like to improve in our application:

  • The web scraping takes a lot of time and makes our application slow. We would like to work on that.
  • Upon testing we realised that Kiplinger has limited articles and the LLM does not always gives the best result. Hence, we would like to try a different method to get latest trends in finance.

Built With

Share this project:

Updates