Module turboquant 

TurboQuant: Near-optimal vector quantization for instant indexing

Based on arXiv:2504.19874 (ICLR 2026)

Key benefits:

• Indexing time: 239s → 0.001s (vs Product Quantization)
• Memory compression: 6x
• Zero accuracy loss
• No training required (data-oblivious)

Structs

LinearScanIndex

Linear scan index for comparison (exact search)

LinearScanStats

Statistics about a LinearScan index

TurboQuantConfig

TurboQuant configuration

TurboQuantIndex

TurboQuant index for fast approximate nearest neighbor search

TurboQuantStats

Statistics about a TurboQuant index