Skip to content
← All projects

14 · In-house · AI integration · E-commerce

Self-hosted AI shop assistant with n8n and WooCommerce

A chatbot for WordPress and WooCommerce that answers questions about products, sizes, prices, shipping costs and delivery times – entirely on our own server, without cloud AI. Built and tested in a demo shop.

  • In-house
  • AI integration
  • n8n
  • Ollama
  • RAG
  • Postgres (pgvector)
  • WooCommerce
  • WordPress
  • Docker
  • Caddy
  • Ubuntu
100 %
self-hosted: n8n, Ollama and Postgres on our own Ubuntu server
2 workflows
build the knowledge base and answer the chat
0
data sent to cloud AI providers

01 · Problem

Customers of an online shop keep asking the same questions: is the product available in my size and price range? How much is shipping, how long is delivery? The goal was a shopping assistant right in the shop that understands questions like “I am looking for product xy in size S between 50 and 100 euros” – without shop data or conversations going to external AI services.

02 · Solution

Two workflows in n8n on a self-hosted Ubuntu server with Docker and Caddy. The first fetches the shop documents – privacy policy, terms, right of withdrawal and payment and shipping – from a GitHub repository, splits them and stores them as vectors (embeddings) in Postgres, which serves as a vector database with pgvector; it only runs when the content changes. The second is the chat: an information extractor evaluates the input, an AI agent with local language models via Ollama decides whether the question is about products or general information and uses the WooCommerce product lookup or the knowledge base (RAG) accordingly – with conversation memory and a calculator tool. The chat is embedded in WordPress via the n8n chat widget.

03 · Result

A chat widget in the shop that suggests suitable products by size and price range and answers questions about shipping, delivery times and conditions from the shop’s own documents. All components run on our own server: no data goes to cloud AI providers and there are no per-request fees. The setup can be transferred to other WooCommerce shops – with their own range, their own documents and the shop’s look and feel.

04 · Insights

  • Server setup: n8n, Ollama, Postgres, WordPress and Caddy in Docker on Ubuntu
  • Overview in n8n: knowledge base workflow at the top, chat workflow below
  • Knowledge base: documents from GitHub are split and stored in Postgres (pgvector)
  • Chat workflow: information extractor, AI agent, memory, WooCommerce lookup and RAG
  • AI agent in detail: local Ollama models and tools
  • Integration into WordPress via the n8n chat widget
  • Chat widget in the demo shop

Next project

Your project here?

Tell us about your problem. The first conversation is of course free of charge and without obligation.

Request a project →

Newsletter

News from the vona workshop.

AI tools we actually use, our weekly AI review & insights from our projects – short, practical and without spam.