jogayush272/Sales_AI_Agent_ND

🙏 Naman Darshan AI — Chatbot & Booking Assistant

An advanced, production-grade AI sales and customer assistant built for Naman Darshan (operated by Traininglobe Consultancy Pvt Ltd). The assistant helps devotees plan temple visits (Darshan), book pujas, discover yatra packages, buy prasadam/chadhava, seek astrological services, and handle payment queries with zero-risk guarantees.

The application integrates an LLM-driven agentic chat workflow with a multi-layered RAG (Retrieval-Augmented Generation) pipeline, a local FAISS vector store, MongoDB database storage, a beautiful custom-styled Streamlit UI, and a Flask REST API complete with Razorpay payment webhook integration.

🏗️ Architecture Overview

The system runs in a co-existence model where the Streamlit UI (for interactive testing/direct user interface) and the Flask API (for backend integration and payment webhooks) share the same underlying core logic, database, and vector store.

graph TD
    %% Client interfaces
    A[Streamlit UI - Port 8501] -->|Query / Chat| C(RAG Pipeline / Agent Engine)
    B[Mobile App / Web Frontends] -->|POST /chat| D[Flask API Server - Port 5000]
    
    %% API Routing
    D -->|Query / Chat| C
    
    %% Core pipeline logic
    C -->|Search Context| E[(FAISS Vector Store)]
    C -->|Fetch History & Rules| F[(MongoDB Atlas)]
    C -->|Compile Dynamic Prompt| G[LangChain / Groq LLM]
    G -->|Generate Response| C
    
    %% Webhooks and payments
    H[Razorpay Webhook Callback] -->|POST /razorpay/webhook| D
    D -->|Update Payment Status| F

🌟 Key Features

1. 🛕 Devotional RAG Engine

• —FAISS Vector Store: Indices rich details on seasonal travel guidelines (optimal months for all major Indian temples), state-wise temple categories, Char Dham pathways, and Jyotirlinga circuits.
• —Intelligent QA & Retrieval: Pulls contextual data to resolve specific devotee queries instantly.

2. 🛡️ Sales & Objection Handling

• —Trust Building: Automatically addresses legal vs. brand name differences (e.g., brand Naman Darshan vs. legal entity Traininglobe Consultancy Pvt Ltd) using realistic company analogies like Zomato/Swiggy.
• —Payment Objections: Proactively offers secure bank transfers (NEFT/IMPS) to clear customer hesitation around QR code payments.
• —Tax Compliance: Explains GST exemption on religious coordination services under Indian tax laws.

3. 💳 Razorpay Payment Webhooks

• —Secure Validation: Flask server processes incoming Razorpay events (payment_link.paid) using HMAC SHA256 signature verification.
• —Real-time Database Sync: Automatically updates customer booking and payment status in MongoDB upon payment confirmation.

4. 🎨 Premium Streamlit Interface

• —Themed Ambiance: Custom CSS styling incorporating deep temple colors (marigold orange, dark woods, and gold gradients).
• —Devotional Ticker: Interactive scrolling ticker header displaying service offerings.
• —Custom Chat Bubbles: Clean user and assistant messaging avatars and cards.

📁 Repository Structure

├── app.py                      # Main Streamlit user interface & chat screen
├── flask_app.py                # Flask REST API server (Endpoints: /chat, webhooks, status)
├── rag_pipeline.py             # Main response generation logic combining FAISS & MongoDB
├── booking_flow.py             # User intent parser and interactive booking state-machine
├── build_knowledge_base.py     # Script to generate FAISS vectors from manual & DB data
├── update_mongodb.py           # Seeding script for objection handlers, company info, and rules
├── razorpay_handler.py         # Handles Razorpay signature verification and payment updates
├── memory.py / session_store.py# User chat history managers syncing to MongoDB
├── data_loader.py              # Loads initial temple and package data from MongoDB
├── retriever.py / embedder.py   # Embedding and VectorDB helper classes
├── requirements.txt            # Python dependencies list
└── run.ps1 / run.bat           # Scripts to run Streamlit & Flask concurrently

🚀 Installation & Setup

1. Clone & Set Up Virtual Environment

# Clone the repository
git clone <your-repository-url>
cd Sales-ai-backend

# Create a virtual environment
python -m venv myenv

# Activate the virtual environment
# On Windows (CMD):
myenv\Scripts\activate
# On Windows (PowerShell):
myenv\Scripts\activate.ps1
# On macOS/Linux:
source myenv/bin/activate

2. Install Dependencies

pip install -r requirements.txt

3. Configure Environment Variables

Create a .env file in the root directory and add the following keys:

GROQ_API_KEY=your_groq_api_key_here
MONGO_URI=your_mongodb_connection_uri_here
RAZORPAY_KEY_ID=your_razorpay_key_id_here
RAZORPAY_KEY_SECRET=your_razorpay_key_secret_here
RAZORPAY_WEBHOOK_SECRET=your_razorpay_webhook_secret_here

4. Seed Data & Rebuild Vector Store

Before running the applications, seed the company details/objections to MongoDB and build the local FAISS index:

# Seed MongoDB collections (agent rules, company info, objections)
python update_mongodb.py

# Extract MongoDB data and build local FAISS vector store
python build_knowledge_base.py

🏃 Running the Application

To run both the Streamlit Web UI and the Flask API Server concurrently:

• —On Windows (PowerShell):

    ./run.ps1

• —On Windows (Command Prompt / batch):

    run.bat

• —Manual launch:

    # Terminal 1: Launch Streamlit UI
    streamlit run app.py
    
    # Terminal 2: Launch Flask API Server
    python flask_app.py

📡 API Reference (Flask Server)

1. Send Message to Assistant

• —Endpoint: POST /chat
• —Payload:

    {
      "user_id": "unique-user-session-id",
      "query": "Kashi Vishwanath Darshan options tell me"
    }

• —Response:

    {
      "response": "🙏 Jai Shri Ram! For Kashi Vishwanath, we offer Assisted Darshan...",
      "history": [ ... ]
    }

2. Razorpay Webhook Receiver

• —Endpoint: POST /razorpay/webhook
• —Headers: Requires signature verification header X-Razorpay-Signature.
• —Event Handled: payment_link.paid
• —Response: {"status": "ok"} on successful capture.

3. Check Payment Status

• —Endpoint: GET /payment/status/<user_id>
• —Parameters: booking_id (Optional query parameter, e.g., ?booking_id=ND-12345)
• —Response: Returns current booking status and payment details. ======= title: Sales AI Agent ND emoji: 🚀 colorFrom: red colorTo: red sdk: docker app_port: 8501 tags:
• —streamlit pinned: false short_description: Chatbot for Naman Darshan Spiritual Services license: mit ---

Welcome to Streamlit!

Edit /src/streamlit_app.py to customize this app to your heart's desire. :heart:

If you have any questions, checkout our documentation and community forums.

>>>>>> d55abc73bb7782746828c8c638901ac01ab9844d