Skip to content

Integration

How to wire a front-end to the running engine. Two parts: Send Events (capture behavior with RudderStack and deliver it to the engine) and Consume Recommendations (read recommendations and search from the AI Engine API).

flowchart LR
    ui["Your UI"] -->|identify + track| rs["RudderStack Cloud"]
    rs -->|webhook| ingest["POST /api/ingest"]
    ingest --> redis[("Redis buffer + user model")]
    ui -->|GET /api/recommend| api["AI Engine API"]
    api --> redis
    api --> qd[("Qdrant")]
    classDef store fill:#EFEAE0,stroke:#A8895B,color:#423D34;
    class redis,qd store;

The engine never reads RudderStack directly. Events flow UI -> RudderStack -> webhook -> /api/ingest, are normalized and buffered in Redis, and folded into the user model. The UI then reads recommendations back from the API.

  • Send Events

    Capture behavior with RudderStack: the event catalog, the source and webhook config, and the SDK identify / track calls.

  • Consume Recommendations

    Read from the AI Engine API: recommendations with explanations, semantic and geo search, narrative, and the user-model debug endpoint.