RAG Q&A Chatbot

Abstract

This repository implements a Retrieval-Augmented Generation (RAG) Q&A Chatbot using Streamlit for the user interface, FAISS as a vector store, and language models from OpenAI and HuggingFace. The system enables ingestion of documents or web pages, vector indexing, similarity-based retrieval, and context-aware answer generation.

Features

• Upload individual files (PDF, TXT, CSV, JSON, DOCX) or ingest entire webpages via URL.
• Automatic text splitting and chunking of documents for efficient embedding.
• Embedding vectors generated with sentence-transformers/huggingface and OpenAI embedding models.
• FAISS vector store for rapid nearest-neighbor search.
• Retrieval chain combining document chunks and LLM to produce concise, context-aware answers.
• Streamlit interface for seamless interaction.

Installation and Setup

1. Clone the repository:

   git clone https://github.com/your-org/your-repo.git
   cd your-repo

2. Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Create a .env file in the project root with the following variables:

   OPENAI_API_KEY=your_openai_api_key
   LANGCHAIN_API_KEY=your_langchain_api_key
   LANGCHAIN_PROJECT=your_langchain_project_name
   HF_TOKEN=your_huggingface_token

Usage

Launch the Streamlit application:

streamlit run rag_app.py

1. Navigate to the web interface opened by Streamlit.
2. Upload files or enter a URL to ingest content into the vector store.
3. Enter your question in the chat box.
4. View the generated answer, which is based on retrieved context.

Architecture Overview

Below is a high-level diagram illustrating the core workflow of the RAG Q&A Chatbot:

flowchart LR
    subgraph Ingestion
        A[User Upload / URL] --> B[Document Loader]
        B --> C[Text Splitter]
        C --> D[Embedding Model]
        D --> E[FAISS Vector Store]
    end

    subgraph QA["Query & Answer"]
        F[User Question] --> G[Retrieval Chain]
        G --> E
        E --> G
        G --> H[Language Model]
        H --> I[Answer]
        I --> J[Streamlit UI]
    end

    style Ingestion fill:#E3F2FD,stroke:#2196F3,stroke-width:2px
    style QA fill:#F1F8E9,stroke:#8BC34A,stroke-width:2px
    linkStyle default stroke:#999,stroke-width:1px

Loading

License

This project is licensed under the MIT License. See LICENSE for details.

Acknowledgements

This work leverages open-source libraries including:

• LangChain
• FAISS
• Streamlit
• OpenAI Python SDK
• HuggingFace sentence-transformers