Mixpeek Logo

    AI-Powered Catalog Search

    Replace keyword-based catalog search with AI-powered semantic and visual search. Understands natural language queries like "lightweight summer dress under $50" and combines text understanding with visual similarity for comprehensive product discovery.

    image
    text
    Production
    4.6K runs
    Deploy Recipe
    from mixpeek import Mixpeek
    client = Mixpeek(api_key="YOUR_API_KEY")
    # Create product catalog collection
    collection = client.collections.create(
    namespace_id="ns_your_namespace",
    name="product_catalog",
    extractors=["multimodal-extractor", "text-extractor"]
    )
    # Upload product data with rich metadata
    client.buckets.upload(
    bucket_id="bkt_products",
    url="s3://your-bucket/products/",
    metadata_mapping={"price": "price", "category": "category", "brand": "brand"}
    )
    # Build production retriever with filtering and reranking
    retriever = client.retrievers.create(
    namespace_id="ns_your_namespace",
    name="catalog_search",
    collection_ids=["col_product_catalog"],
    stages=[
    {"type": "feature_search", "top_k": 100},
    {"type": "attribute_filter"},
    {"type": "rerank", "top_k": 20}
    ]
    )
    # Natural language catalog search
    results = client.retrievers.execute(
    retriever_id=retriever["retriever_id"],
    query={"text": "lightweight summer dress under $50"},
    filters={"in_stock": True, "price": {"$lte": 50}}
    )
    for doc in results["results"]:
    print(f"{doc['metadata']['name']} - ${doc['metadata']['price']}")

    Feature Extractors

    Retriever Stages

    rerank

    Rerank documents using cross-encoder models for accurate relevance

    sort