Seoul PM10 Air Pollution Webhook: MotherDuck + Ollama Qwen AI

Get an AI-generated PM10 air quality report for Seoul—on demand

This n8n workflow exposes a webhook that pulls PM10 air pollution data from MotherDuck, analyzes the results with a local Ollama Qwen model, and returns a ready-to-view HTML report highlighting the most polluted Seoul districts.

What this workflow does

• Receives a request on a webhook endpoint (via n8n).
• Queries MotherDuck using its MCP API to calculate the top 10 Seoul monitoring districts by average PM10, including each district’s maximum PM10.
• Sends the summarized PM10 results to a local Ollama server at /api/generate using qwen2.5:3b (default: http://ollama:11434).
• Generates a four-sentence analysis referencing the WHO PM10 guideline (50 μg/m³).
• Injects the model output into a preformatted HTML page and returns it as the HTTP response.

Use cases

• Monitoring dashboards: call the /air-quality endpoint to render an updated Seoul PM10 snapshot.
• City/ops alerts: integrate webhook triggers with internal tools to surface worst districts and AI commentary.
• SaaS operators: provide on-demand reporting without building a custom UI—responses are already HTML.
• Automation engineers: demonstrate a clean pattern for “DB → LLM → web response” using n8n.

Technical details

• Webhook (receive request, then Respond to Webhook with HTML)
• HTTP Request to Ollama: /api/generate
• MotherDuck via MCP API: calculates top 10 Seoul districts by average PM10 and maximum PM10
• Nodes used include: set, webhook, sticky note, http request, respond to webhook

Setup requirements

• Create a MotherDuck API bearer token and ensure the air_pollution database and referenced tables/columns match the workflow’s SQL.
• Run Ollama where n8n can reach http://ollama:11434 and pull qwen2.5:3b (or update the model name/URL in the workflow).