turbovec · Issues· 17 open

gRPC server surface for turbovec (+ a stale MSRV)

encode.rs: seven mutants survive the whole test suite, three of them in compute_tqplus_calibration

Potential integer division by zero in search path when bits parameter is invalid

Concurrency ceiling: RAYON_NUM_THREADS=1 serializes the whole service (14x loss); nq=1 latency cliff at exactly n=8192; one writer halves read throughput

SQLite Extension Based on Your Work

Rust crate as a downstream dependency: SearchResults derives nothing, search() has no non-panicking form, MSRV over-declared by 9 releases

Mutation gate has an architectural blind spot: mutants in cfg-gated code are vacuously MISSED on the runner's arch

Integrations should not own document text: shrink the side-car to ids + metadata

CI: execute the AVX-512 kernels — designated runner or Intel SDE leg

Unverified: .cargo/config.toml forces target-cpu=x86-64-v2 in-repo but not for downstream builds

Side-car JSON does not scale: 5.5x larger than the index at 100k docs, and save peaks at 2x the payload

Recall cliff (up to 65pp) on sparse data at weak-block dims (dim = 8·odd or 16·odd)

README's recall claim holds for real embeddings but loses to FAISS on isotropic Gaussian and weak-block sparse

Query-side LUT is quantized to 7 bits (max_lut=127), costing 1.5-3.5pp recall@10 — a cross-arch equivalence trade, not a rounding detail

TQ loses to the FAISS PQ baseline on bimodal and one-hot-like coordinate distributions