What product quantization cost
57.7% recall@10 at 16x compression, the near-worst-case configuration, measured on purpose.
Product quantization (PQ) trades recall for memory. Instead of storing full-precision vectors, it trains k-means++ codebooks per subspace and stores compressed codes. NuclaDB's implementation (internal/index/pq) uses asymmetric distance computation (ADC), so the query vector itself is never quantized, only the stored vectors are.
The number
57.7% recall@10 at a fixed 16x memory reduction (M=16, Dim=64), from internal/index/pq/pq_test.go's TestIndexRecallVsExact, which is deterministic across 5 runs with fixed seeds.
This is flat, unindexed PQ, on purpose
No re-ranking, no IVF coarse quantizer. Production PQ setups typically add both. This configuration leaves them out deliberately, as a clean measurement of quantization error alone, not because it's the recommended way to run PQ in production.
Why it isn't higher
Two follow-ups explain the gap, and neither is implemented yet:
- No re-ranking: a typical PQ pipeline re-scores the top candidates against their full-precision vectors before returning results, which recovers most of the lost recall. NuclaDB's PQ path doesn't do this yet.
- No IVF: an inverted-file coarse quantizer narrows the search to a subset of the index before PQ scoring, which also tends to improve effective recall by avoiding compression error building up over the full dataset.
What's verified
The distance computation itself is verified exact, so the recall loss traces to quantization error alone, not a scoring bug. That distinction matters: a lower number from a correct-but-lossy algorithm is one kind of tradeoff, while a lower number from a buggy comparator would just be a bug.
For the other side of the compression-vs-accuracy space, see HNSW ef tuning.
Tuning HNSW: what the recall/latency curve looks like
The full recall@10/QPS sweep on SIFT-small, and why efSearch is a per-query parameter.
What Raft gave the system and what it cost
Agreed-upon topology through faults, proven under chaos, and the linearizability window that async replication leaves open.