RAG For Papers with Code

RAG For Papers with Code is a easier method to scan through tons of ML research by just hitting a button. A RAG system that mines large amount of information easily. You’ll ask it your questions in natural language and it’ll answer according to relevant papers it finds on Papers With Code.

On the backend side, the system is powered with a Retrieval Augmented Generation (RAG) framework that relies on a scalable serverless vector database called Upstash, for embeddings we are using BGE models on the HuggingFace, and Mixtral-8x7B-Instruct-v0.1 as LLM from HuggingFace.

On the front-end side, this assistant will be integrated into an interactive and easily deployable web application built with Streamlit.

Requirements

• HuggingFace API TOKEN
• Upstash Account
• Upstash URL and TOKEN after creating the index

Steps to Replicate

1. Clone the repo

     git clone https://github.com/wittyicon29/PWC-RAG.git

2. Move to the workspace directory

     cd PWC-RAG

3. Indexing To index data into the vector DB, you first need to create an index on Upstash and fill in the credentials in the .env file:

    UPSTASH_URL=...
    UPSTASH_TOKEN=...
   

   Run the following command:

     python -m src.index_papers --query "Mistral" --limit 200

   Result of indexing 200 chunks matching the "Mistral" query.

   

   

4. Running the streamlit applications locally

   Before running the streamlit app, you have to set the Huggingface API token in the '.env' file:

     HUGGINGFACE_API_TOKEN=...

   Now you can the streamlit app

     python -m streamlit run  src/app.py

   

Notes

• You can use any Embedding model supported by Langchain
• You can try different LLMs on Langchain to evaluate the RAG system
• The app is ready to deploy on Google Cloud Run using docker or Streamlit cloud.

Reference

How To Build an LLM-Powered App To Chat with PapersWithCode