Recommendation as a Service
RecomNext gives you a complete recommendation backend out of the box — plug in your catalog, define scenarios with nextQL, and serve personalized results through a single API call.
Capabilities
From collaborative filtering to content-based similarity, pick the right algorithm for each placement on your site.
Write filters, boosters, and segmentation rules in plain, readable syntax without redeploying backend code.
Measure recommendation performance with impression, click, and conversion tracking tied to each scenario.
Define named scenarios for each touchpoint — homepage, product detail, cart — each with its own logic.
Automatically group catalog items by attribute or expression to ensure balanced, diverse result sets.
Run multiple brands or storefronts on a single deployment with fully isolated data per tenant.
Drop-in Browser and Node.js clients with auto-impression tracking, typed responses, and retry logic.
Surface semantically similar products using vector search with customizable embedding templates.
How it Works
Why RecomNext
Go from zero to live recommendations with minimal engineering effort.
One REST endpoint to fetch recommendations — no complex orchestration required
Browser SDK handles impression and click tracking automatically
Node.js SDK ships with built-in retries, timeouts, and typed responses
Comprehensive API reference, nextQL guide, and SDK docs to get you running fast
Leverage state-of-the-art retrieval and ranking without building ML infrastructure from scratch.
Vector search finds semantically similar products using Qdrant and embedding models
Behavioral signals are captured through co-occurrence and weighted-history algorithms
The algorithm registry is pluggable — add or swap strategies without redeployment
Embedding pipelines run asynchronously with fully configurable templates
The same architecture handles prototype traffic and millions of daily requests.
Each tenant gets isolated data in a shared-nothing multi-tenant model
Kafka streams interactions in real time so rankings stay fresh
ClickHouse delivers sub-second analytics over billions of events
Stateless API containers scale horizontally behind any load balancer
Scenario
