n8n import:workflow --separate imports each JSON file independently. All 4
workflows shared the same tags (parapharmacy, scraper), so the second import
tried to INSERT duplicate tag names, hitting the unique index on
tag_entity.name.
- Remove tags field from all 4 workflow JSON files (cosmetic, re-add via UI)
- Add cleanup-db.js to wipe workflow data before import
- Add --overwrite flag to import command as safety net
- Add cleanup-db.js volume mount to n8n-init container
- Change CORS_ORIGIN from localhost:4000 to farmacias.hacecalor.net in docker-compose.yml
- Add postgres and parapharmacy-api healthchecks
- Add n8n-init service that runs BEFORE n8n starts:
- Imports workflows via n8n CLI
- Activates all workflows via publish:workflow
- Seeds parapharmacy DB with 20 test products
- Fix parapharmacy-api healthcheck to use node instead of wget (not available in image)
- Fix seed.json to use {products: [...]} format with required source_url field
- Remove activate-workflows.js (replaced by CLI approach)
- Fixed Primor URL format and selectors
- Created simple N8N webhook that calls scraper API
- All 3 sources now working:
- Promofarma: 20 products
- DocMorris: 10 products
- Primor: 20 products
- Total: 130 products in database
- N8N webhook at /webhook/scrape working correctly
- Added Puppeteer with Chrome in Docker
- Scraper extracts products using data attributes
- Added /api/scrape endpoint
- Tested: 10 products scraped from Promofarma
The scraper now works with Promofarma's HTML structure which uses
data-name, data-pvp attributes on article elements.
Antoni Nuñez Romeu