MEMORISE AI-Engine

Documentation for the MEMORISE in-memorial system, how the platform represents users, represents content, and fuses both into recommendations and narratives.

Explore by area

User Representation

Turning visitor behavior into a model of interest: event capture, the event catalog, the behavioral pipeline, and the materialized user signals.

Open section →

Content Representation

Turning archival items into machine-usable content: the Omeka collection, the Knowledge Graph, expert tags, and vector embeddings in Qdrant.

Open section →

AI Engine

Fusing user and content into explainable recommendations and narratives: the DAC architecture, hybrid scoring, diversity, and the serving API.

Open section →

System at a glance

flowchart LR
    subgraph U["User Representation"]
        app["In-memorial app"]
        rs["RudderStack"]
        ph["PostHog"]
    end
    subgraph C["Content Representation"]
        omeka["Omeka collection"]
        tags["Expert tags"]
        qd[("Qdrant<br/>vectors + tags")]
    end
    subgraph A["AI Engine"]
        api["ai-engine-api"]
        eng["recsys + search"]
        rd[("Redis<br/>user model")]
    end

    app -->|events| rs --> ph
    rs -->|behavioral signals| eng
    omeka --> tags --> qd
    app -->|search / recommend| api --> eng
    eng --> qd
    eng --> rd

    classDef store fill:#EFEAE0,stroke:#A8895B,color:#423D34;
    class qd,rd store;

User behavior (left) and archival content (middle) are each turned into typed representations; the AI Engine (right) fuses them. Follow an area above, or jump to the API reference.