Member Junction
    Preparing search index...
    DistanceMetric:
        | "cosine"
        | "cosine-similarity"
        | "euclidean"
        | "manhattan"
        | "dotproduct"
        | "jaccard"
        | "hamming"

    Supported distance/similarity metrics for vector operations. Each scores higher for more similar vectors. cosine is cosine rescaled to 0-1, (cos + 1) / 2; cosine-similarity is the standard cosine in [-1, 1], which is what Pinecone, SQL Server and pgvector report, so use it where a score is compared against a threshold shared with those stores.