What's Changed
- fix(platform): centralize cache-line sizing by @luoxiaojian in #831
- Enable RaBitQ on Linux/macOS/Android/iOS ARM64 and macOS x86_64 by @egolearner in #832
- fix(core): preserve index bytes on read-only close by @iaojnh in #834
- docs: clarify README platforms and Python example by @zhourrr in #835
- docs: simplify README badges and organize SDK installation tables by @JalinWang in #844
- perf: deduplicate string parameter constants by @egolearner in #846
- fix(hnsw_rabitq): refine entry scores before level-0 search by @egolearner in #851
- feat(turbo): complete ARM NEON one-to-many distance kernels by @richyreachy in #856
- fix: enforce read-only storage and WAL recovery boundaries by @iaojnh in #840
- perf(diskann): borrow mmap vectors during graph construction by @iaojnh in #853
- fix(io): prevent file descriptor inheritance across exec by @egolearner in #859
- fix(db): Add validation for rank_constant in RRF reranking by @ZoruaSama in #847
- perf(core): accelerate HNSW construction and FP16 HNSW/Flat distances by @luoxiaojian in #862
- ci(deps): bump actions/setup-java from 6.0.1 to 6.1.0 by @dependabot[bot] in #823
- fix(hnsw): skip unavailable external vectors during queries by @iaojnh in #861
- fix(storage): separate BufferStorage creation and write permissions by @iaojnh in #870
- feat(flat): support external vector sources with optional scan iteration by @iaojnh in #848
- feat(c-api): expose grouped vector search by @JalinWang in #845
- fix(db): avoid null writing_segment_ on DDL writing segment recreatio… by @ZoruaSama in #822
- feat(turbo): support product quantization end to end by @richyreachy in #833
- refactor: remove legacy distance kernels in favor of turbo by @richyreachy in #796
- fix(c_api): check malloc() result in zvec_doc_get_pk_copy before use by @ilike314272 in #863
- fix(c_api): add missing null checks in string-array helpers by @ilike314272 in #864
- perf: reduce DiskANN PQ training and neighbor storage memory by @iaojnh in #877
- perf(ivf): reuse stable source vectors during construction by @iaojnh in #879
- perf(hnsw): reduce optimize memory for fresh graph builds by @iaojnh in #880
- perf(flat): integrate runtime-dispatched BatchTopK with row batches by @luoxiaojian in #871
- fix(build): add OSS fallbacks for Thrift and Boost downloads by @iaojnh in #882
- feat(diskann): support Intel macOS and Windows ARM64 by @iaojnh in #874
New Contributors
- @ZoruaSama made their first contribution in #847
- @ilike314272 made their first contribution in #863
Full Changelog: v0.8.0...v0.8.1