Skip to content

Vector search-based chatbot to help retail investors analyze company 10K and 10Q filings.

License

Notifications You must be signed in to change notification settings

Chubbyman2/Mango10K

Repository files navigation

Mango10K

Introduction

Your friendly neighbourhood vector search-based chatbot, here to help retail investors navigate the complex and tedious task of analyzing company 10K and 10Q filings!

Getting Started

To get started, you'll need a MongoDB Atlas cluster, OpenAI API key, Langchain API key, and a Nomic API key. See .env.example for more information.

Vector Search Index

Once you have your MongoDB Atlas cluster, database, and collection, you'll need to apply a vector search index to it. See these instructions for the exact steps. We used the following configuration:

{
  "fields": [
    {
      "numDimensions": 768,
      "path": "embedding",
      "similarity": "cosine",
      "type": "vector"
    }
  ]
}

Prerequisites

gradio==4.25.0
langchain==0.1.14
langchain-community==0.0.31
langchain-nomic==0.0.2
openai==1.16.2
pydantic==2.6.4
pymongo==4.6.3
python-dotenv==1.0.1
sec_cik_mapper==2.1.0
unstructured==0.13.2

Embeddings and Vector Search

To try an example, run seed.py to generate and store the vector embeddings from the most recent 10K's and 10Q's from AAPL and TSLA. Make sure you have your vector search index configured! Then, you can run frontend.py to launch the Gradio web app, and ask some questions regarding the financials of Tesla and Apple!

Technologies

MongoDB Atlas

MongoDB Atlas was used to store the vector embeddings generated, as well as return results from vector search.

OpenAI

OpenAI's gpt-3.5-turbo was used to generate coherent responses based on the retrieved embeddings from MongoDB Atlas, in order to answer the user's question.

Nomic

Nomic's model was used for the actual vector embeddings. Specifically, we used the nomic-embed-text-v1.5 model.

LangChain

LangChain was used to load the HTML documents, perform the vector store and vector search, as well as generate the actual response. The backend model was OpenAI's, but the module used was LangChain.

Gradio

Gradio was used to create the web app which users can interact with in order to use our application.

Team

Special thanks to Daniel, Shah, Somyaranjan, Calvin, and Charles of team MangoDB!

License

This project is licensed under the MIT License - see the LICENSE file for details.

About

Vector search-based chatbot to help retail investors analyze company 10K and 10Q filings.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages