Retrieval-Augmented Generation with LLAMA2

This Streamlit app called "Point of View Hub" allows users to either scrape text from a URL or upload a PDF. It then provides a chatbot interface where users can ask questions related to the scraped content, and the chatbot uses a language model to generate responses based on the provided text. It uses the RAG (Retrieval-Augmented Generation) model which blends retrieval-based and generation-based approaches in natural language processing. It employs a retrieval component to identify relevant documents from a corpus based on the input query, then utilizes a generation component to produce the final answer. This integration allows RAG models to leverage pre-existing knowledge effectively, resulting in accurate and informative responses across a wide range of queries.

The app integrates several technical components:

1. User Interface (UI):

   • It's built using Streamlit, a Python library for creating interactive web applications.
   • The UI includes options for scraping text from a URL or uploading a PDF file.
   • A chatbot interface allows users to interact with the system by asking questions.

2. Data Scraping:

   • For scraping text from a URL, the app uses the requests library to fetch the HTML content, and BeautifulSoup for parsing the content and extracting text.
   • For PDF uploads, it uses PyPDF2 to extract text from the uploaded PDF.

3. Language Model:

   • The app employs a language model for generating responses to user queries. Specifically, it uses a model called "TheBloke/Llama-2-13B-chat-GPTQ" (LLAMA2). This model is based on GPT (Generative Pre-trained Transformer) architecture, specialized for conversational question-answering tasks.

4. Pipeline and Embeddings:

   • It uses langchain library to set up the pipeline for processing text and generating responses.
   • Embeddings from Hugging Face models are used for text representations.
   • A retrieval-based question-answering (QA) system is implemented for responding to user queries, possibly involving multiple documents.

Overall, the app provides a user-friendly interface for interacting with a sophisticated language model, capable of answering queries based on the scraped content and is branded for a Deloitte use-case.

How to Run in Google Collab

The notebook "streamlit_via_collab" allows for a Streamlit app running on port 8501 in the background, then use localtunnel to expose it to the internet. Upload the app.py file and Deloitte logo and run the command:

streamlit run app.py & npx localtunnel --port 8501

Teammates

Name	Email	Socials
Zabih Buda	zzabih@schulich.yorku.ca	LinkedIn
Sabrina Renna	srenna@schulich.yorku.ca	LinkedIn
Vivek Periera	vivekp@schulich.yorku.ca	LinkedIn