AI Search
Semantic product search and "similar products" powered by vector embeddings. Requires pgvector.
AI Search
AI Search replaces keyword matching with semantic search. Products are indexed as vector embeddings, so a search for "lightweight summer dress" can match a product titled "floral chiffon midi" without any exact keyword overlap.
The same embeddings power the Similar Products widget on product pages.
Requirements
- AI Core plugin must be installed and active
- PostgreSQL with the
pgvectorextension
Enable pgvector and run the migration:
# On the server
sudo -u postgres psql -c "CREATE EXTENSION IF NOT EXISTS vector;"
# In the API directory
npm run migration:run:ai-searchInstallation
Install from Admin → Plugins → Store. No license key required.
Settings
| Setting | Description |
|---|---|
| Enabled | Turn semantic search on or off |
| Embedding Provider | gemini (default) or openai |
The API key is read from the AI Core plugin settings — no separate key needed.
How It Works
When a product is saved, AI Search generates a text embedding and stores it in the pgvector column. On search, the query is embedded and a nearest-neighbour lookup finds matching products. If the plugin is inactive or pgvector is unavailable, search falls back to the standard keyword search.