weather-agent

An AI agent to interact with NWS MCP server to get weather forecasts and alerts for a given location. Author: Sanidhya Mangal

Implementation Approach

Entire project works on two core principles.

• MCP: Developed a model context protocol server to wrap up all the function call and API call to NWS server to ensure agent can communicate with MCP layer rather than calling different functions.
• Agent: Developed an autonomus FunctionalAgent to interact with MCP server and get responses for the given workflow.

Project Structure

.
├── app.py
├── dist
│   ├── weather_agent-0.1.0-py3-none-any.whl
│   └── weather_agent-0.1.0.tar.gz
├── pyproject.toml
├── README.md
├── src
│   ├── weather_agent
│   │   ├── __init__.py
│   │   ├── scripts
│   │   │   └── test_scripts.py
│   │   ├── utils.py
│   │   └── weather_mcp_server
│   │       ├── __init__.py
│   │       ├── utils.py
│   │       └── weather.py
│   └── weather_agent.egg-info
│       ├── dependency_links.txt
│       ├── PKG-INFO
│       ├── requires.txt
│       ├── SOURCES.txt
│       └── top_level.txt
└── uv.lock

Setup and Running Instructions

Environment Setup

• Requires Python 3.10 or later.
• Uses uv for dependency management. Install uv as per the instructions here and run: uv sync

Environment Variables

You can refer to .env.template to setup .env file for configuring environ variable or can simple set them using export command

Model and API Setup

• The chatbot uses Ollama to run LLM models. Download and install it from Ollama.
• Download the mistral:7b model with: ollama pull mistral:7b

Running MCP Server

To start MCP server you can run following command:

uv run weather-mcp-server

This command will main function within, weather.py to start MCP server.

Entrypoint and command name could be modified within pyproject.toml

Running UI

To start UI please run following command:

uv run streamlit app.py

If UI window doesn't open automaticall please feel free to navigate to URL mentioned in console.