MoveWiser - A RAG AI powered chatbot for your website

Tech Stack

Frontend:

• Vue & Nuxt

Backend:

• Flask
• FAISS vector DB storage
• custom script to scrape URLs on a website

Details

MoveWiser is an AI chatbot that leverages a Retrieval-Augmented Generation (RAG) model to answer questions specifically tailored to your website!

This is how it works:

1. Scraping

The first task is to scrap all data available on a website. We took an approach of looking at sitemap.xml and downloading all pages followed by cleaning up and storing the images.

2. Generating Embeddings

We are using all-mpnet-base-v2 to generate embeddings and storing it in FAISS Vector storage database.

3. Flask Backend

When a query comes we use it to generate vector embedding and keyword based search and figure out the relevant parts.

4. LLM

We use Mistral AI's latest LLM that came just a few days back. We tried 13B-Llama2, but this was faster on our machine.

We used this repository to easily run, look at performance and create an API for the model.

5. Frontend

We did not have too much time for making it look good, and adding the features that we wanted to, but it is a functioning reponsive chatbot app.

Some interesting features

• Everything runs locally (from generating embeddings, to storing, and also running LLM)
• We have performed OCR on images and added also added it into generating embeddings.
• Alongside using similarity search, we wrote a simple token based keyword search to answer direct questions, which otherwise could be missed.

Running locally

• Run an API inference of any LLM you wish (we used oobabooga's text-generation-webui with Mistral-AI-7B)
• embeddings/data_get.py has the availibility to change the website URL and download all pages on that website. It also performs OCR on images and stores that.
• Create_Embeddings/generate_embeddings.py would generate embeddings for the downloaded website pages. Followed by merge_dbs.py which merges various faiss databases which were created, while preserving true source of data.
• On starting Create_Embeddings/flask_endpoint.py it creates an API endpoint where on querying a question it returns relevant parts.
• Go to /frontend and install npm packages yarn install followed by yarn dev.

Team: Ctrl+Alt+Defeat

• Divyansh Singh
• Param Gandhi
• Ritwik Sharma
• Vimarsh Shah