┌─────────────────────────────────────────────────────────────┐
│                    User Query                               │
│           "Will it rain in Chicago tomorrow?"               │
└─────────────────────┬───────────────────────────────────────┘
                      │
                      v
┌─────────────────────────────────────────────────────────────┐
│              Agent Bricks Agent                             │
│  - Interprets user intent                                   │
│  - Decides which tools to call                              │
│  - Formats final answer                                     │
└─────────────────────┬───────────────────────────────────────┘
                      │ MCP Protocol (SSE)
                      v
┌─────────────────────────────────────────────────────────────┐
│         Weather MCP Server (FastMCP)                        │
│  Endpoint: /sse                                             │
│  Tools:                                                     │
│    - get_current_weather()                                  │
│    - get_forecast()                                         │
│    - predict_umbrella_needed()                              │
└─────────────────────┬───────────────────────────────────────┘
                      │
                      v
┌─────────────────────────────────────────────────────────────┐
│           Weather Broker (weather_broker.py)                │
│  - HTTP requests to Open-Meteo API                          │
│  - Geocoding (city → lat/lon)                               │
│  - Response parsing and error handling                      │
└─────────────────────┬───────────────────────────────────────┘
                      │ HTTPS
                      v
┌─────────────────────────────────────────────────────────────┐
│              Open-Meteo API (Free)                          │
│  https://api.open-meteo.com                                 │
│  - No API key required                                      │
│  - ~10,000 calls/day                                        │
└─────────────────────────────────────────────────────────────┘

command:
  - "python"
  - "weather_mcp_server.py"

env:
  - name: LOG_LEVEL
    value: INFO

You are a helpful weather assistant that provides accurate, real-time weather information and predictions.

Your capabilities:
- Get current weather conditions for any location
- Provide multi-day forecasts (up to 16 days)
- Make umbrella recommendations based on precipitation likelihood

Guidelines:
1. Always use the weather tools to get current data - never guess or use outdated information
2. For location queries, accept city names ("Seattle"), city+state ("Austin, TX"), or coordinates
3. If a location cannot be resolved, ask the user to clarify or provide a major nearby city
4. When making predictions, explain your reasoning (e.g., "40% precipitation chance, so umbrella recommended")
5. If the API call fails, tell the user there was an issue fetching weather data and suggest trying again
6. For date-specific questions, use get_forecast() and find the matching date
7. Be conversational but concise - users want quick, actionable weather information

Error handling:
- Bad location → Ask user to clarify or suggest a major city
- API outage → Acknowledge the issue, don't make up data
- Future date out of range → Explain forecast limit is 16 days

Examples:
- "Will it rain in Chicago tomorrow?" → Call get_forecast("Chicago", days=2) and check tomorrow's precipitation
- "Should I bring a jacket to Austin this weekend?" → Call get_forecast("Austin, TX", days=7) and check weekend temps
- "What's the weather like in Miami right now?" → Call get_current_weather("Miami")

{
    "error": "location_not_found" | "api_error" | "unexpected_error",
    "message": "Location not found: Xyzabc",
    "suggestion": "Try a major city name, or provide coordinates as 'lat,lon'"
}

User: "What's the weather like in Seattle right now?"

Agent:
- Calls: get_current_weather("Seattle")
- Response: "In Seattle, it's currently 58°F with partly cloudy skies. 
            Humidity is 75%, and winds are light at 8 mph from the northwest."

User: "Will it rain in Austin this weekend?"

Agent:
- Calls: get_forecast("Austin, TX", days=7)
- Filters to Saturday/Sunday
- Response: "This weekend in Austin: Saturday will be sunny with a high of 92°F 
            and only 10% chance of rain. Sunday looks similar at 90°F with 15% 
            chance of rain. You should have great weather!"

User: "Should I bring an umbrella to Miami tomorrow?"

Agent:
- Calls: predict_umbrella_needed("Miami")
- Response: "Yes, you should bring an umbrella to Miami tomorrow. There's a 65% 
            chance of rain (moderate confidence), so it's likely you'll need it. 
            Expect a high of 86°F and low of 77°F with moderate rain."

# Install dependencies
pip install -r requirements.txt

# Run MCP server locally
python weather_mcp_server.py

# Server runs on http://0.0.0.0:8000
# SSE endpoint: http://0.0.0.0:8000/sse

# Sync local changes to workspace
databricks sync --watch . /Workspace/Users/<your-email>/weather-mcp-server

# Deploy app
databricks apps deploy weather-mcp-server \
  --source-code-path /Workspace/Users/<your-email>/weather-mcp-server

# Check status
databricks apps get weather-mcp-server

# View logs
databricks apps logs weather-mcp-server --tail-lines 100

weather-mcp-server/
├── README.md                    # This file
├── app.yaml                     # Databricks App configuration
├── requirements.txt             # Python dependencies
├── weather_mcp_server.py        # MCP server with 3 tools
└── weather_broker.py            # Weather API adapter/broker