READMEmain

🧠 MCP + Ollama Streamlit Chatbot

This repository contains a fully functional multi-agent chatbot powered by the Model Context Protocol (MCP), Ollama with the qwen3:1.7b model, and a Streamlit-based frontend. The chatbot supports tool calling and integrates domain-specific utilities like weather APIs, math evaluation, and CSV dataset analysis.

📁 Project Structure

.
├── app.py             # Streamlit frontend interface
├── client.py          # Async MCP client with Ollama integration and tool handling
├── server.py          # MCP-compatible server with weather, math, and dataset tools
├── data/
│   └── dataset.csv    # Sample CSV file for dataset analysis
├── .env               # Environment variables (MCP_SSE_URL, etc.)

🧰 Features & Tools

The assistant supports the following built-in tools via MCP server:

🌤️ Weather Tools – Fetch alerts and forecasts using the National Weather Service API
➗ Math Evaluator – Safe evaluation of arithmetic expressions
📊 Dataset Inspector – Summary statistics, shape, and NLP-powered queries on a local dataset.csv file

flowchart TD
    subgraph Streamlit_UI
        A1[User Prompt]
        A2[Display Chat History]
        A3[Streamlit App - app_py]
    end

    subgraph MCP_Client
        B1[Connect to SSE Server]
        B2[Process Query]
        B3[Call Ollama API]
        B4[Handle Tool Calls]
    end

    subgraph MCP_Server
        C1[Weather Alerts Tool]
        C2[Forecast Tool]
        C3[Math Evaluation Tool]
        C4[Dataset Analysis Tool]
        C5[Dataset Query Tool]
    end

    subgraph Ollama_Model
        D1[qwen3 1_7b Model]
    end

    A1 --> A3
    A3 --> B2
    A2 --> A3
    B2 --> B3
    B3 --> D1
    D1 --> B4
    B4 -->|Tool Call| C1
    B4 -->|Tool Call| C2
    B4 -->|Tool Call| C3
    B4 -->|Tool Call| C4
    B4 -->|Tool Call| C5
    C1 --> B2
    C2 --> B2
    C3 --> B2
    C4 --> B2
    C5 --> B2

🔧 Requirements

Ensure you have the following installed:

Python 3.9+
Ollama with the qwen3:1.7b model available locally
MCP library (see installation below)
Streamlit
Uvicorn for ASGI server
A .env file with MCP server URL defined:
```
MCP_SSE_URL=http://localhost:8080/sse
```

Python Packages

You can install all required packages via:

pip install -r requirements.txt

If you don't have a requirements.txt, use:

pip install streamlit uvicorn httpx python-dotenv pandas scikit-learn mcp

▶️ How to Run

1. Launch the MCP Server (Tool Provider)

Run the server to expose SSE-compatible endpoints:

python server.py --host 0.0.0.0 --port 8080

This will start a FastAPI-compatible MCP server exposing tools on:

http://localhost:8080/sse

2. Start the Streamlit Frontend

In another terminal:

streamlit run app.py

This will open the chat interface in your browser at:

http://localhost:8501

📦 Ollama Model Setup

Install and run Ollama:

ollama pull qwen3:1.7b
ollama run qwen3:1.7b

Ensure the model is loaded and responding at:

http://localhost:11434/api/chat

🧪 Supported Use Cases

Here are some queries you can try:

Weather

What’s the weather in San Francisco?
Are there any alerts in NY?

Math

What is 2 + 3 * 4?
Calculate the square root of 81

Dataset Analysis

What are the columns in the dataset?
How many records are in the file?
Do any descriptions mention "cloud" or "AI"?

🧼 Resetting State

You can reset the chat history using the sidebar button in the Streamlit UI.

📁 Notes

Make sure data/dataset.csv exists if using dataset tools.
Ensure MCP_SSE_URL in .env matches your server setup.
This system uses asyncio.run_coroutine_threadsafe() to allow asynchronous tool execution within Streamlit's synchronous model.

mcp-ollama-streamlit-agent