Fast-RAG Chatbot Pipeline

An intelligent Sales Bot powered by a Retrieval-Augmented Generation (RAG) pipeline. The project provides a scalable document ingestion framework, dynamic chunking, and an interactive chat interface to answer questions based on the uploaded company or product documents.

🚀 Key Features

• Customizable Chunking Strategies: Choose from Recursive, Sentence, Token, Semantic, and Fixed chunking strategies, and tweak the chunk size and overlap to fit your needs.
• Asynchronous Processing: Documents are processed in the background with real-time progress tracking, allowing you to use the interface uninterrupted.
• Modern React Interface: A sleek, high-end React frontend (Vite) utilizing a modern "Soft UI" design and real-time response streaming via Server-Sent Events (SSE).
• Agentic Capabilities: Leverages Langchain and agent frameworks to route queries effectively, with full Opik observability.
• Advanced Hybrid Retrieval: Uses Qdrant's built-in hybrid search with Reciprocal Rank Fusion (RRF) for dense and sparse vectors, followed by FlashRank reranking for maximum accuracy.

🛠️ Technology Stack

• Frontend: React + Vite (with custom Neumorphic/Glassmorphism CSS)
• Backend: FastAPI (with SSE streaming)
• LLM & RAG Framework: LangChain, LangChain-Groq, Opik (Observability)
• Embeddings & Search: sentence-transformers, Qdrant (RRF Hybrid Search), FastEmbed, FlashRank
• Document Loading: PyMuPDF

📁 Project Structure

Rag_Chatbot\
├── frontend/              # Modern React/Vite Frontend application
├── app.py                 # Legacy Streamlit Frontend application
├── main.py                # FastAPI Backend entry point (QA_Bot)
├── requirements.txt       # Project dependencies
├── .env                   # Environment variables (OpenAI/Groq keys, etc.)
├── backend/               # FastAPI routers (upload, query, metrics)
├── llm/                   # LLM clients and core agent interaction logic 
├── rag_pipeline/          # Core RAG components (ingestion, embedding, retrieval)
├── vectordb/              # Vector database storage and operations
├── uploads/               # Temporary storage for uploaded user files
├── utils/                 # Helper scripts and utilities
└── logs/                  # System and agent execution logs

⚙️ Setup and Installation

1. Clone the repository

Ensure you are in the project root directory.

2. Set up virtual environment (Recommended)

python -m venv venv
# Windows
venv\Scripts\activate
# Unix/MacOS
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Configure Environment Variables

Create a .env file in the root directory (if not exists) and add your API keys:

OPENAI_API_KEY=your_openai_api_key
GROQ_API_KEY=your_groq_api_key
# Add other necessary keys...

🏃 Running the Application

To run the application locally, you will need to start both the FastAPI backend and the React frontend. Note: Ensure Qdrant is running locally (e.g., via Docker on port 6334).

Start the FastAPI Backend

Open a terminal, navigate to the project root, and run:

uvicorn main:app --reload --port 8000

The API will be available at http://localhost:8000. You can test endpoints via http://localhost:8000/docs.

Start the React Frontend

In a separate terminal, navigate to the frontend directory and run:

npm install
npm run dev

Your browser will automatically open the Sales Bot interface.

💡 Usage

1. Open the React web application in your browser.
2. In the Upload Documents section, securely upload your data files.
3. Select your preferred Chunking Strategy and adjust dimensions if necessary.
4. Click Upload and watch the background process smoothly digest the data.
5. Head to the Chat with Sales Bot section to start extracting insights and asking domain-specific questions!