genpark-product-quantization-compression-engine-skill

Product Quantization (PQ) vector compression engine splitting high-dimensional float embeddings into compact centroid codebook indices.

Architecture

flowchart LR
    V["1024-dim Float Vector (4096 bytes)"] --> Slicer["Subvector Slicer (M=8)"]
    Slicer --> CB["Sub-Space Centroid Codebooks"]
    CB --> Code["Quantized Byte Indices (8 bytes)"]

Loading

Features

• Massive Compression: Up to 64x memory footprint reduction.
• Zero External Dependencies: 100% Python Standard Library.