The n8n Workflows That Run Live Seller Pro
A look inside the six n8n workflows that automate health monitoring, AI support tickets, knowledge search, and work order reviews ΓÇö all self-hosted on a single laptop.
Why n8n
When I needed workflow automation for LiveSeller Pro, I had two choices: pay for Zapier and hope their rate limits did not strangle my operations, or self-host something I actually controlled. I chose n8n — an open-source workflow automation platform that runs in a Docker container on my own hardware.
That was the best infrastructure decision I have made. Every automated process at LiveSeller Pro runs through n8n, from health monitoring to AI-powered support tickets to work order lifecycle management. Here is a look at the workflows that keep everything running.
Status Health Check — 2 Nodes, Zero Monthly Fee
Every five minutes, a scheduled workflow pings every critical service — the website, n8n itself, the ShopOps API, and the LSPRO API. It measures response times in milliseconds and writes the results to a WordPress REST endpoint.
The System Status page reads that data and renders it in real time. Green badges, red badges, response times. No third-party status page service. No $29/month monitoring subscription. Two nodes: a schedule trigger and a code node that does the work.
Most status page services charge per monitor. We monitor everything we run for the cost of the electricity to run the container.
Support Ticket Intake — AI Classification in Real Time
When someone submits a support ticket through our Support Portal, the form fires a webhook to n8n. Here is what happens in the next three seconds:
- The form data gets normalized and validated
- Oracle searches our knowledge base to understand the topic
- A classification engine determines the severity — High, Medium, or Low
- A ticket is created in Notion with all properties populated
- Email alerts fire based on severity — High triggers an URGENT notification immediately
Eleven nodes. No manual triage. No tickets sitting in an inbox waiting for someone to read them. The AI classifies and routes faster than a human could open the email.
Notion Proxy — Secure API Gateway
Several of our tools need to read from or write to Notion, but we never expose the Notion API token to client-side code. Instead, everything routes through a six-node n8n workflow that acts as a proxy. It authenticates requests, constructs the correct Notion API call, routes GET and POST operations separately, and returns clean responses.
Simple pattern, but it means we have one place to manage authentication, rate limiting, and logging for all Notion operations.
WO-Review — The AI That Reviews Our Work Orders
This is the workflow I am most proud of. When a builder completes a work order and marks it for review, a webhook fires and:
- Authenticates the request with a webhook token
- Fetches the work order content from Notion
- Converts Notion blocks to plain text
- Sends it to an LLM for structured technical review
- Formats the review as Notion blocks
- Appends the review directly to the work order page
An AI reviewing work done by an AI, with the results written back to the project management system automatically. The review is severity-aware — low-risk changes get flagged for quick review, medium requires human sign-off, high triggers a hold.
We recently learned the hard way that this system can be too trusting. The auto-approve path let eleven work orders through with zero actual code — it checked risk level and keywords but never verified that files existed on disk. We have since built a two-gate validation system that requires real code evidence before anything advances. That story is in our Work Order Pipeline article.
Work Order Lifecycle — Status Transitions at Scale
Beyond the review workflow, we run a suite of lifecycle workflows that manage the entire work order board:
- Status change detection triggers downstream actions
- Builder assignment notifications route to the right agent
- Risk classification runs automatically on every new WO
- Completion comments are validated for required sections
- Daily digests summarize overnight activity
We currently run over sixty workflows across the platform. The system is not perfect — n8n silently deactivates workflows after certain errors, and restarting the Docker container reactivates everything whether you want it to or not. But after eight weeks of refinement, the automation handles more of the operational load every week.
The Infrastructure
All of this runs on a self-hosted Hetzner VPS — a single server running Docker containers for n8n, WordPress, the ShopOps API, and the LSPRO API. Our AI compute workloads (Ollama for LLM inference, Whisper for voice, Piper for TTS) run on a dedicated machine on our local network.
No cloud functions. No serverless. No enterprise automation licenses. Just n8n, some code nodes, and the conviction that if you can describe a process, you can automate it.
— John Ranson, Blue Devil Collectibles