Search and discovery
Beyond carousels: hybrid search combines natural language queries with vector similarity and automatic nextQL filter extraction. Personalized search reranks with user affinity.
API base URLs
| Engine API base URL | Admin UI |
|---|---|
| https://api.recomnext.com | https://admin.recomnext.com |
Use the engine base URL as baseUrl in widgets and SDKs.
Logic types for search
| Setting | Type | Default | Description |
|---|---|---|---|
| search + hybrid-search | algorithm | — | Vector + parsed filters + facets in response. |
| search + personalized-search | algorithm | — | Hybrid search with user-personalized reranking. |
Hybrid search example
GET https://api.recomnext.com/recommendations/search?query=open+world+rpg+under+50&count=10&userId=customer-42 X-Tenant-Id: aurora-games
Example response shape
{
"items": [...],
"facets": {
"categories": [{ "value": "pc", "count": 8 }],
"priceRanges": [{ "value": "$20 - $50", "count": 5 }]
},
"query": {
"original": "open world rpg under 50",
"semanticText": "open world rpg",
"extractedFilters": "'price' < 50.0"
}
}Use case
Storefront search page
Use facets to build filter chips. Semantic remainder drives vector recall; price constraints become nextQL filters automatically.
Scenario-based search tuning
Create a search scenario with default filters (e.g. in-stock rule, category restrictions) and passscenario on search requests where supported. Combine with rules for global constraints.
Auth note
Search endpoints are typically called from your backend with an API key. Browser public HMAC has a narrower write surface — prefer a BFF for search if credentials cannot live in the client.