Implementation
Architecture & LogicThe workflow is built in n8n. It triggers daily via Cron, creates city items, merges them, and uses the Split in Batches node to process each city sequentially. It fetches data from OpenWeatherMap, passes raw JSON to GPT-4o Mini for formatting, and sends the result via Telegram.
Prompt / LLM Node
// LLM Prompt Structure
You are a weather assistant.
Generate a concise, friendly weather report for the city.
Input Data:
- City: {{$json.city}}
- Temp: {{$json.main.temp}}°C
- Feels like: {{$json.main.feels_like}}°C
- Wind: {{$json.wind.speed}} m/s
- Humidity: {{$json.main.humidity}}%
- Description: {{$json.weather[0].description}}
Output Format:
City Emoji + Name
Temperature line
Weather description
Advice (e.g., "Take an umbrella" if rain).
