BMSCE-MCP-Server-Client-

KedarJevargi/BMSCE-MCP-Server-Client-

3.2

If you are the rightful owner of BMSCE-MCP-Server-Client- and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.

BMSCE Assistant is a friendly AI-powered chatbot designed for BMS College of Engineering students, utilizing the Model Context Protocol (MCP) and Mistral 7B to provide real-time updates and academic assistance.

Tools
3
Resources
0
Prompts
0

🎓 BMSCE Assistant

A friendly AI-powered chatbot for BMS College of Engineering students, built using the Model Context Protocol (MCP) and Mistral 7B. Get real-time updates on college events, notifications, and query academic documents—all through a natural conversational interface.

✨ Features

  • 📰 Latest News & Events: Stay updated with college events, workshops, festivals, and happenings
  • 📢 College Notifications: Access official notices, circulars, announcements, and deadlines
  • 🔍 Knowledge Base Query: Search through uploaded PDF documents (syllabus, resumes, academic content)
  • 💬 Natural Conversation: Friendly, student-oriented chat interface powered by Mistral 7B
  • 🚀 MCP Architecture: Built on Model Context Protocol for extensible tool integration

🏗️ Architecture

┌─────────────────┐
│   User Input    │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  client.py      │  ◄── Mistral 7B LLM (via Ollama)
│  (MCP Client)   │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│   main.py       │
│  (MCP Server)   │
│                 │
│  Tools:         │
│  • get_latest_news
│  • get_college_notifications
│  • query_knowledge_base
└────────┬────────┘
         │
    ┌────┴────┬──────────────┐
    ▼         ▼              ▼
┌────────┐ ┌──────┐ ┌────────────────┐
│Web     │ │Web   │ │ChromaDB        │
│Scraper │ │Scraper│ │Vector Store    │
│(News)  │ │(Notif)│ │(PDF Documents) │
└────────┘ └──────┘ └────────────────┘

📋 Prerequisites

  • Python 3.8+
  • Ollama with the following models:
    • mistral:7b (LLM)
    • nomic-embed-text:v1.5 (Embeddings)

Install Ollama

# macOS/Linux
curl -fsSL https://ollama.com/install.sh | sh

# Windows
# Download from https://ollama.com/download

Pull Required Models

ollama pull mistral:7b
ollama pull nomic-embed-text:v1.5

🚀 Installation

  1. Clone the repository
git clone <your-repo-url>
cd bmsce-assistant
  1. Install Python dependencies
pip install -r requirements.txt
  1. Set up the vector database (optional, for PDF querying)

If you want to add PDF documents to the knowledge base:

# Edit vector_db.py and add your PDF files to the pdf_files list
python vector_db.py

This will create a chroma_storage directory with indexed documents.

💻 Usage

Running the Assistant

Start the chatbot with:

python client.py

You'll see a welcome screen:

🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓

   Welcome to BMSCE Assistant! 🤖
   Your friendly AI helper for all things BMS College

🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓 🎓

✅ Connected to MCP Server

💬 Hey there! Ask me about college events, notifications, or anything else!
   Type 'quit' or 'exit' when you're done.

────────────────────────────────────────────────────────────────

You:

Example Queries

News & Events:

You: What events are happening this month?
You: Tell me about upcoming workshops
You: Any festivals coming up?

Notifications:

You: What are the latest notifications?
You: Show me recent announcements
You: Any important deadlines?

Knowledge Base:

You: Search for syllabus information
You: Find details about [specific topic]
You: What does the document say about [query]?

General Chat:

You: Hi!
You: Thank you
You: How are you?

Exiting

Type any of the following to exit:

  • quit
  • exit
  • bye
  • goodbye
  • Or press Ctrl+C

📁 Project Structure

bmsce-assistant/
│
├── client.py              # MCP client with Mistral integration
├── main.py                # MCP server with tool definitions
├── web_scrap.py           # Web scrapers for BMSCE website
├── vector_db.py           # ChromaDB setup and PDF indexing
├── requirements.txt       # Python dependencies
├── .gitignore            # Git ignore file
│
├── chroma_storage/        # ChromaDB persistent storage (auto-created)
└── *.pdf                  # Your PDF documents to index

🛠️ Components

1. MCP Server (main.py)

Defines three tools:

  • get_latest_news(): Scrapes news and events from BMSCE website
  • get_college_notifications(): Scrapes college notifications
  • query_knowledge_base(query_text, n_results=3): Queries ChromaDB vector store

2. MCP Client (client.py)

  • Connects to the MCP server
  • Uses Mistral 7B to:
    • Interpret user queries
    • Select appropriate tools
    • Generate natural, friendly responses
  • Provides a conversational interface

3. Web Scraper (web_scrap.py)

  • Scrapes the BMSCE website (https://bmsce.ac.in)
  • Extracts structured data from:
    • News & Events section
    • College Notifications section
  • Returns data in JSON format

4. Vector Database (vector_db.py)

  • Uses ChromaDB for semantic search
  • Indexes PDF documents with Nomic embeddings
  • Supports chunked document retrieval

🔧 Configuration

Adding PDF Documents

Edit vector_db.py:

if __name__ == "__main__":
    pdf_files = [
        "syllabus.pdf",
        "handbook.pdf",
        "your_document.pdf"
    ]
    for pdf in pdf_files:
        add_pdf_to_vectordb(pdf)

Then run:

python vector_db.py

Customizing LLM Behavior

In client.py, you can modify:

  • Temperature: Controls randomness (0.1 = focused, 0.9 = creative)
  • Model: Change mistral:7b to other Ollama models
  • Prompts: Edit system prompts for different personalities

Adjusting Chunk Size

In vector_db.py:

chunks = split_text(pdf_text, chunk_size=1000, overlap=100)

🤝 How It Works

  1. User asks a question → Sent to client.py
  2. Mistral analyzes the query and decides which tool to use
  3. Client calls the tool on the MCP server (main.py)
  4. Server executes the tool:
    • Web scraping for news/notifications
    • Vector search for knowledge base queries
  5. Raw data returned to client
  6. Mistral formats the data into a natural response
  7. User sees friendly answer 🎉

🐛 Troubleshooting

"Not connected to server" error

  • Make sure both client.py and main.py are in the same directory
  • Check that Python can execute main.py as a subprocess

Ollama model not found

# Pull the models
ollama pull mistral:7b
ollama pull nomic-embed-text:v1.5

# Verify installation
ollama list

ChromaDB errors

  • Delete chroma_storage folder and re-run vector_db.py
  • Ensure PDF files exist in the project directory

Web scraping fails

  • Check your internet connection
  • Verify the BMSCE website is accessible
  • Website structure may have changed (update selectors in web_scrap.py)

📝 Dependencies

Key libraries:

  • fastmcp - MCP server framework
  • mcp - MCP client SDK
  • ollama - Ollama Python client
  • chromadb - Vector database
  • beautifulsoup4 - Web scraping
  • PyPDF2 - PDF text extraction

See requirements.txt for complete list.

🚧 Future Enhancements

  • Add caching for web scraping results
  • Support for more document formats (DOCX, TXT)
  • Database for persistent conversation history
  • Voice input/output
  • Web interface (Gradio/Streamlit)
  • Multi-language support
  • Integration with college calendar
  • Student authentication system

Made with ❤️ for BMSCE students