• Problem Statement
• Context & Motivation
• Objective
• System Workflow
  • 1. Document Ingestion
  • 2. Embedding & Indexing
  • 3. Query Execution (RAG Pipeline)
  • 4. Conversational Memory
• Key Components
• Repository Structure
• System Architecture
• Use Cases
• Installation
• License
• Contributing

• Organizations frequently manage large repositories of unstructured PDF documents such as technical manuals, compliance reports, internal policies, and product documentation. Manually retrieving relevant information from these files is time-consuming and inefficient.
• Traditional search mechanisms often lack semantic understanding and fail to provide context-aware answers, resulting in increased workload, delayed decision-making, and reduced productivity.

Leveraging machine learning, natural language processing, and Generative AI, this project automates traditionally manual processes like document review, insurance claim triage, or customer support through a PDF-based intelligent assistant.

• AI-enhanced understanding of unstructured content for faster analysis.
• Predictive insights to support proactive decision-making.
• Streamlined information flow across internal teams or external users.
• Based on real-world use cases like product queries, support FAQs, legal document analysis, etc.

This project aims to develop a Retrieval-Augmented Generation (RAG) based conversational agent that enables users to interact with a corpus of PDF documents through a natural language interface. The system is designed to:

• Automate information retrieval from enterprise documents.
• Provide real-time, contextually relevant responses to user queries.
• Maintain conversational memory for multi-turn interactions.
• Enhance user productivity and decision-making through intelligent document Q&A.

• All PDF files placed under the static/ directory are loaded using PyPDFLoader.
• Documents are segmented into semantically meaningful chunks using RecursiveCharacterTextSplitter.

• Each chunk is transformed into a vector representation using HuggingFaceEmbeddings (all-MiniLM-L6-v2).
• Vectors are stored in a FAISS index for fast and scalable similarity search.

• A user submits a natural language query via the chat UI.
• Retriever selects top relevant chunks from the FAISS index.
• These chunks are passed to a Groq-hosted large language model (mixtral-8x7b-32768) to generate an informed response.

• A session-based ChatMessageHistory maintains prior user inputs and bot replies.
• RunnableWithMessageHistory wraps the chain to handle contextual understanding across turns.

WebRAG_Chatbot/
├── pycache/ # Compiled Python cache
├── faiss/ # FAISS vector index
├── imgs/ # Image assets 
├── static/ # Source PDF documents (knowledge base)
├── templates/ # HTML templates for Flask frontend
├── .env # API keys and environment variables
├── README.md # Project documentation
├── app.py # Main application code (Flask + LangChain)
├── gitignore # Git ignore rules
├── prompts.py # Custom system prompts for RAG chain
├── requirements.txt                # Python dependencies
├── LICENSE

• Customer support based on internal knowledge base PDFs.
• Legal or compliance document Q&A.
• Employee onboarding or policy clarification assistant.
• Technical manuals or product documentation assistant.

Follow the steps below to set up and run the Multi-PDF RAG Chatbot locally:

git clone https://github.com/paht2005/WebRAG_Chatbot.git
cd WebRAG_Chatbot

python -m venv venv
source venv/bin/activate       # On Linux/macOS
venv\Scripts\activate          # On Windows

pip install -r requirements.txt

Create a file named .env in the root directory and add your API key:

GROQ_API_KEY=sk-xxxxxxxxxxxxxxxxxxxxxxxxxx

Do not share this file publicly. Use my .env for reference in the repo.

Place all your .pdf documents inside the static/ folder. These will be parsed and indexed automatically when the app starts.

python app.py

• Then open your browser and navigate to:

(http://127.0.0.1:5000)

• You can now start chatting with your documents in real time

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

I welcome contributions to improve this project! Feel free to fork, pull request, or open issues. Ideas welcome!