Finsight

Your AI finance bro

Comment

Finsight

Your AI finance bro, Finn 😎
Bridging the gap between new investors and financial markets through stock market briefs and “snapshots.”
Developed for YHack 2024.
Check out our GitHub!

About

Finsight is a RAG (retrieval-augmented generation) AI app that helps new investors understand financial markets through experiential learning. We created a pipeline for market data streamed from Yahoo Finance coupled with a state-of-the-art LLM (large language model) that processes data and delivers high-quality, unbiased market briefings to users through a friendly user interface.

Our Process

We were motivated by the difficulty of learning about the stock market, as it is taking a bigger role in our day-to-day wealth management lives. So, we determined a need for an educational platform to help new investors understand what the financial market looks like daily and provide critical insights on their portfolio. As a result, Finsight came to be. We spent the first few hours planning our idea, including developing our unique “RAG” pipeline from data sources like Yahoo Finance to the LLM to the user interface with LangChain wrappers. We also sketched our idea on a whiteboard while researching similar designs: Yahoo Finance, Robinhood, Motley Fools, etc.

This was the second time most of our team members had developed with ReactJS and Firebase, so working with these web technologies was a challenge, especially with passing data between components. Our team learned a lot about teamwork and communication, especially using GitHub as a centralized platform for code management (merge conflicts!). Furthermore, we learned a lot about the ideation process and how to plan for software with many moving parts (components, working with LLMs, API development, etc).

Current Features

  • Top company stock listings dashboard for users to explore and add to their favorite stocks list
  • In-depth RAG-generated stock briefings for day, month, quarter, and year generated by Finn, our SOTA LLM RAG pipeline
  • Easy-to-follow tutorial
  • Chat option to query for more information on financial reports
  • Definitions for each analysis to help users learn about financial terms

Future Work

  • Deploy to a permanent cloud location and custom domain
  • Experiment with different LLMs, Prompt Templates, and RAG pipelines to improve response quality

Technical

Finsense was developed with the following technologies:

  • Frontend: ReactJS
  • API: FastAPI, HuggingFace LLM (gemma:2b, published on Feb 21, 2024), Ollama, LangChain
  • Database: Firebase Firestore
  • Auth: Firebase Auth

Local Hosting

To run the project locally, you may clone the repository linked above. You must have Node version 18+ and pip/Python 3.11+ installed on your machine.

Setup

git clone https://github.com/charlestang06/finsight

Then, install the dependencies.

cd web-app
npm install
cd ..
cd finance-rag-server
pip install -r requirements.txt

Run the server. The server will be listening on port 8000.

cd finance-rag-server
uvicorn main:app --reload
cd ..

Open a new command line and run the web-app.

cd web-app
npm start

Go to localhost:3000 to view the project.

Versions

  • Version 1.0 (03/31/2024)
    • Initial Release
    • See [Commits history]

License

MIT

Built With

Share this project:

Updates