NEWVectors or files. Pick a path.Start →

    Scene Moment Reassembly

    The docs 'Retrieval Reassembly' pipeline over the 2-tier video DAG: semantic vector search over scenes, attribute filter, cross-encoder rerank, then enrich each moment with the PEOPLE detected in it (tier-1 scene-entities join). Captures Query -> Vector Search -> Filter -> Rank -> Enrich from docs/overview/introduction.

    sample-scene-reassembly
    Try Live Demo

    Feature Extractors

    AI models used for embeddings

    multimodal_extractor

    AI embedding model

    Retriever Pipeline

    Stages in the retrieval pipeline

    1

    scene_vector_search

    filter

    2

    genre_gate

    filter

    3

    precision_rerank

    sort

    4

    people_in_scene

    apply

    About This Retriever

    Public Name:sample-scene-reassembly
    Extractors:

    1 AI model

    Pipeline Stages:

    4 stages

    Try This Retriever

    Experience the live demo and explore multimodal search capabilities