Repository Issue Activity (beta)

#539 WASM support for Browser and Node.js?

#538 How this compares to Redis VSET?

#182 gRPC server surface for turbovec (+ a stale MSRV)

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

#493 nq=1 block-parallel NEON path (scan_range_neon) drops the whole-block prune -> ~28% single-thread regression above the 1024-block gate

#473 Agno async_insert/upsert should fail before embedding when create() was not called

#487 v6 load allocates from file length, not declared content: 2450-byte index in a 4 GiB sparse file costs 8 GB RSS

#484 lock_write_gil_aware livelocks swap_remove / IdMapIndex.remove under concurrent search: 66 ns -> 25 s

#501 First incremental add after a load/search permanently doubles the codes buffer (Vec::reserve amortized growth on a tight buffer)

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

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

#475 v7: search-layout-only in RAM, incremental saves on disk

#488 #299 stale-temp sweep is a permanent no-op for destination filenames >=235 bytes

#477 sync(): choose full rewrite when it writes fewer bytes than the incremental plan

#530 Any plan to cut a release?

#531 Cut a release

#482 v7: the fixed MAX_OPS header reserve is a per-file floor of ~2048 rows, so small indexes sync to files 67-87x larger than write()

#464 TQ+ calibrated scores overshoot the cosine ceiling at every index size

#461 remove() is ~50% slower at d3072 4-bit under per-block calibration

#455 Per-block TQ+ calibration, optional and on by default (format v7)

#489 expected_codebook drops the MAX_DIM bound its rustdoc claims, hanging for minutes on out-of-range dim

#478 TQ+ calibration validation accepts finite-but-unusable magnitudes: a tqplus_scale of 1e-40 loads clean and turns every search score into NaN

#306 CI gaps: no debug test leg, no clippy/fmt, binding crate never tested, integration floors never tested, release smoke skips integrations

#466 Uncalibrated indexes cannot be used to build corruption fixtures that splice into the scales section

#492 Benchmark diagrams misstate calibration behavior (post-#474) and hardcode compression numbers instead of reading the JSON