LLM Agent for Weather Determination

LLM agent for weather information retrieval through interaction with two APIs via unified MCP interface.

The agent uses:

• Nominatim API for determining coordinates by city name
• Open-Meteo API for retrieving weather data by coordinates
• Groq LLM for natural language processing
• MCP protocol for unified tool access

Supports current weather, forecasts, and historical data in multiple languages.

• Python 3.11+
• UV package manager

1. Install dependencies:

   uv sync

2. Create .env file with API key:

   GROQ_API_KEY=your_groq_api_key_here
   

Requires 2 terminals:

Terminal 1 - MCP Server:

uv run python -m app.mcp_server

Terminal 2 - UI:

uv run streamlit run app/streamlit_app.py

"What's the weather in London today?"
"Weather forecast for next 3 days in Paris"
"What was the weather in Tokyo yesterday?"

app/
├── agents/weather_agent.py      # Main agent logic
├── prompts/system_prompt.py     # System prompt
├── schemas/                     # Pydantic schemas
├── services/                    # LLM service and settings
├── tools/                       # API tools
├── ui/weather_ui.py            # Streamlit interface
├── mcp_server.py               # MCP server
└── streamlit_app.py            # Entry point

• LLM: Groq (llama3-8b-8192)
• Framework: LangChain + LangGraph
• Pattern: ReAct (Reasoning + Acting)
• UI: Streamlit
• APIs: Nominatim (OpenStreetMap) + Open-Meteo