hrezolit83/n8n-ecommerce-automation

12 production n8n workflows I built for a German e-commerce company that sells on 12+ European marketplaces (Otto, Leroy Merlin / Adeo, Conforama, Worten, Altex, Empik, CDON and more), with **Odoo** as the central ERP.
All credentials, company names, hosts and IDs are removed. Workflows are importable into n8n (Community Edition is enough): **Workflows → Import from File**, then connect your own credentials.
| # | Workflow | Nodes | Key techniques |
|---|---|---|---|
| 01 | AI Product Listing: Odoo → Marketplace | 90 | LLM agents, structured output, HMAC-signed API, pagination, image pipeline |
| 02 | Marketplace Orders → Odoo Sales Orders | 56 | OAuth2, pagination, customer deduplication, sub-workflows |
| 03 | Stock & Price Sync: Odoo → Mirakl | 31 | Batch loops, diff-only updates, multi-channel pricing, rate-limit waits |
| 04 | Tracking Numbers Sync + Alerts | 42 | Shipment aggregation, carrier validation, Bitrix24 error tasks |
| 05 | AI Category Mapper | 19 | LLM classification against a category tree, structured output parser |
| 06 | AI Required Attributes Filler | 18 | LLM + marketplace value lists, strict prompt rules, fallbacks |
| 07 | Auto-Reconciliation: Bank Payments ↔ Odoo Invoices | 29 | Odoo accounting API, matching logic, Bitrix24 checklist |
| 08 | Lost Orders Monitoring (all marketplaces) | 22 | Health checks per channel, alert deduplication |
| 09 | Product Listing via XML API (Hood.de) | 65 | XML request building/parsing, pagination, taxes, images |
| 10 | AI Category Selection with Fallback | 20 | Two-step LLM (OpenAI + Grok), Google Sheets manual fallback |
| 11 | Multi-language Listing with Compliance Filter | 91 | PT/ES content, forbidden-words filter, offer status checks |
| 12 | Pending → Active Offer Activation | 51 | Offer import, status polling, stage updates in Odoo |
---
Case 1 — AI-powered product listing on a new marketplace
**Problem.** Each marketplace has its own category tree and required attributes. Listing hundreds of products by hand took days, and every product was rejected for a different reason.
**Solution.** A workflow that takes a product from an Odoo task queue and:
1. Validates mandatory fields (barcode, price, GPSR, brand, delivery tag) and routes failures to clear error reasons back in Odoo.
2. Uses an LLM agent with structured output to pick the right marketplace category and attribute set from the API.
3. Pulls the full attribute schema (with pagination) and lets the LLM fill required attributes using only allowed values.
4. Builds the image set, signs the request (HMAC-SHA512) and imports the product with its offer.
5. Handles duplicate / invalid-field / API errors separately and updates the task stage.
**Result.** Listing became fully automatic: products move from "New task" to "Pending" without manual work; every rejection has a readable reason.
Case 2 — Marketplace orders into the ERP
**Problem.** Orders from several marketplaces had to be entered into Odoo manually, including customers and addresses.
**Solution.** A scheduled workflow fetches new orders (OAuth2 token, paginated API), checks if the order already exists, finds or creates the buyer and delivery contacts without duplicates, maps products by SKU and creates a sales order. Errors stop the order and are logged instead of creating broken records.
**Result.** Orders appear in Odoo automatically several times a day; no duplicate customers.
Case 3 — Stock & price sync across 4 country channels