v0.16.18

What's Changed

Bug Fixes

• fix(resize): ensure size is updated after successful expansion by @inabao in #1649
• fix(sindi): restore total_count_ during deserialization by @inabao in #1698

Improvements

• refactor(hgraph): remove unused variable in SearchWithRequest by @inabao in #1657

Full Changelog: v0.16.17...v0.16.18

v0.18.3

What's Changed

Bug Fixes

• fix(hgraph): fix potential out-of-bounds access in Merge function by @inabao in #1626
• fix(datacell): fix memory leak in FlattenDataCell query on exception by @Copilot in #1680
• fix(hgraph): restore total_count after deserialization by @inabao in #1647
• fix(datacell): update max_capacity after io operation completes in 0.18 by @inabao in #1651

Improvements

• feat(hgraph): add hops_limit search parameter by @inabao in #1620

Full Changelog: v0.18.2...v0.18.3

VSAG Compatibility Test Indexes

Index files for backward compatibility testing

This release stores index files generated by different VSAG versions for compatibility testing.

Files

• random_512d_10K.bin - Shared test dataset (20 MB)
• {version}_{algo}.index - Index files for each version
• {version}_{algo}_build.json - Build configuration
• {version}_{algo}_search.json - Search configuration

Usage

These files are automatically downloaded by CI workflows for compatibility testing.

Update

New index files are automatically uploaded when a new version is released.

v0.18.2

Bug Fixes

• fix(datacell): restore total_count_ during deserialization by @inabao in #1621

Full Changelog: v0.18.1...v0.18.2

v0.18.1

What's Changed

Bug Fixes

• fix(sindi): fix error when neither doc nor query is relevant by @inabao in #1589
• fix(iter_filter): add bound check for iter filter context by @inabao in #1588
• fix(memory_usage): conditionally call cal_memory_usage only on actual capacity changes by @LHT129 in #1587
• fix(hnsw): fix concurrent access to max_level_ in HNSW search operations by @inabao in #1573
• Remove liburing-devel from pre-cxx11 Dockerfile by @Copilot in #1566

Improvements

• feat(pyramid): move topk_factor to IndexSearchParameter base class by @inabao in #1605
• refactor(pyramid): modify default value of pyramid thread pool by @inabao in #1595
• enhance(label_table): use robin_pg_map instead of unordered_map by @inabao in #1612
• refactor(inner_index): remove default thread pool during index creation by @inabao in #1558

Other Changes

• fix(reader_io): fix missing parameter when deserialize index by @inabao in #1597

Full Changelog: v0.18.0...v0.18.1

v0.16.17

What's Changed

Bug Fixes

• fix(iter_filter): add bound check for iter filter context by @inabao in #1608
• fix(sindi): fix error when neither doc nor query is relevant (#1589) by @inabao in #1609
• fix(hnsw): fix concurrent access to max_level_ in HNSW search operations by @inabao in #1573
• fix iter search and add concurency by @shadowao in #1546
• fix(hgraph): fix coredump when concurrent building from empty index by @inabao in #1569

New Features

• feat(diskann): support calculate distance by id in diskann by @inabao in #1559

Improvements

• fix(hgraph): refactor label remap handling in SetImmutable by @inabao in #1607
• test: add search unrelated parameter tests by @wxyucs in #1455

Full Changelog: v0.16.16...v0.16.17

v0.15.9

What's Changed

Bug Fixes

• fix(hnsw): fix concurrent access to max_level_ in HNSW search operations by @inabao in #1573
• fix(iter_filter): add bound check for iter filter context by @inabao in #1590

Full Changelog: v0.15.8...v0.15.9

v0.15.8

What's Changed

Bug Fixes

• fix(hgraph): fix coredump when concurrent building from empty index by @inabao in #1568

Full Changelog: v0.15.7...v0.15.8

v0.15.7

What's Changed

Bug Fixes

• fix(hgraph): fix concurrency issues during hgraph insert (#1431) by @inabao in #1496

Improvements

• patch hgraph locker for concurrency by @stuBirdFly in #1561

Full Changelog: v0.15.6...v0.15.7

v0.18.0

What's Changed

Bug Fixes

• fix: remove invalid thirdparty header files installation path by @wxyucs in #1228
• fix(sindi): bad alloc and bench issue caused by term id limit by @ShawnShawnYou in #1242
• fix(sindi): fix parameter checkCompatibility in deserialize by @cubicc in #1266
• fix(kmeans): fix kmeans unittest "0 == 161" bug by @LHT129 in #1259
• fix(SINDI): sparse vector len check by @ShawnShawnYou in #1264
• fix(hnsw): fix the OOM issue caused by large IDs by @inabao in #1313
• fix(ci): fix some intermittent bugs on CI by @LHT129 in #1330
• fix(python): fix python build by @hhy3 in #1344
• fix(tsan): fix the tsan error on CI by @inabao in #1366
• fix(ut): illegal instruction (sve) occurred on Apple Silicon by @wxyucs in #1372
• fix(SINDI): fix term out of bound by @ShawnShawnYou in #1383
• fix(leak): vt pool leak on concurrent search by @LHT129 in #1388
• fix(deserialize): fix HNSW in 0.14 cannot parse indexes in 0.16 by @inabao in #1394
• fix: correct allocator used to allocate neighbors_mutex by @wxyucs in #1378
• fix(test): fix compile of memleak in clang by @inabao in #1405
• fix: generate version when prepare and read in setup.py by @jac0626 in #1375
• fix(sparse): fix core dump when get vector with invalid id by @ShawnShawnYou in #1410
• fix(quantizer): fix sq8 code size being 8 times larger than expected by @ShawnShawnYou in #1416
• fix(ci): decrease memory cost in CI compilation by @LHT129 in #1425
• fix(Add): fix semantic issues when "recovery tombstone" fails by @ShawnShawnYou in #1419
• fix(hgraph): fix the process of getting raw vector when using quantizer by @inabao in #1436
• fix(example): solve storage location conflict in example 401 by @wxyucs in #1423
• fix(hgraph): fix coredump due to extra_info_size inconsistency by @inabao in #1404
• fix(hgraph): fix the asan issue in GetVectorByInnerId by @inabao in #1457
• fix(LocalFileReader): fix thread safety issues in LocalFileReader by @inabao in #1409
• fix(timer): fix the insufficient precision in timer's time measurement by @inabao in #1466
• fix(static_link): fix vsag_static loss libio.a by @LHT129 in #1463
• fix(hgraph): fix concurrency issues during hgraph insert by @inabao in #1431
• fix(sindi): enable returns failed ID when inserting duplicate id by @inabao in #1486
• fix(Hgraph): fix get memory usage bug after deserializaion by @LHT129 in #1529
• fix(hgraph): fix coredump when concurrent building from empty index by @inabao in #1520
• fix(analyzer): add nullptr check and copy query vectors in HGraphAnalyzer by @misaka0714 in #1528
• fix(compile): remove outdated directory from tools/CMakeLists.txt by @wxyucs in #1549
• fix iter search and add concurency by @shadowao in #1546
• fix(hgraph): refactor label remap handling in SetImmutable by @misaka0714 in #1550

New Features

• feat(SINDI): add term id limit check by @ShawnShawnYou in #1223
• refactor(sq): refactor sq quantizer by @LHT129 in #1236
• feat(clone): support passing a new allocator when cloning by @inabao in #1218
• feat: add parallelism search for bruteforce by @LHT129 in #1240
• feat(Hgraph): support TQ in Hgraph param by @ShawnShawnYou in #1248
• feat(sparse): support build and search in pybind by @ShawnShawnYou in #1275
• feat(hgraph): ensure that the removing is thread-safe. by @inabao in #1269
• feat(io): introduce io_array to manager multi buckets'IO by @LHT129 in #1246
• feat(ivf): add param to support on-disk IVF Index by @LHT129 in #1283
• feat: Automate Python Wheel Build And Release by @jac0626 in #1064
• feat(c_api): introduce some simple C api by @LHT129 in #1270
• feat(api): introduce an api for GetDetailInfo from Index by @LHT129 in #1294
• feat(python): support build python wheel without docker by @LHT129 in #1306
• feat(index): add write function for Serialize by @LHT129 in #1291
• feat(index): add "is_timeout" stat for search result by @LHT129 in #1316
• feat(c_api): add Error_t instead bool as return result by @LHT129 in #1299
• feat(ivf): add ivf sampling count function by @misaka0714 in #1303
• feat(sparse): support get raw vector by @ShawnShawnYou in #1347
• feat(sparse): doc prune function revision by @Roxanne0321 in #1325
• feat(c_api): add functional serialzie method for C api by @LHT129 in #1334
• feat(BruteForce): enable thread_count config by @LHT129 in #1349
• feat(index): Support get detail data with info from index by @LHT129 in #1309
• feat(c_api): add more api for c-sdk by @LHT129 in #1363
• feat: add more statistics for search process by @wxyucs in #1343
• feat(pyramid): support brute force in pyramid by @inabao in #1400
• feat(transformer): Add MRL-E preprocess Transformer by @LHT129 in #1371
• feat(hgraph): support deduplication under different quantization by @inabao in #1403
• feat(index): get data type in info by @ShawnShawnYou in #1450
• feat: add allocator parameter to GetRawVectorByIds by @wxyucs in #1434
• feat(hnsw): support get raw vector by @ShawnShawnYou in #1446
• feat(hgraph): support transform basic quant from fp32 to any other quant by Tune by @ShawnShawnYou in #1442
• feat(python): add doc/test for pyvsag by @LHT129 in #1398
• enhance(sindi): improve sindi performance with filter by @Ningsir in #1449
• feat(sq4): implement for sq4's simd by @LHT129 in #1391
• feat(SINDI): support update vector by @ShawnShawnYou in #1484
• feat(hgraph): limit train size by @ShawnShawnYou in #1483
• feat(hgraph): support tune high precision by @ShawnShawnYou in #1503
• feat(eval_performance): add search query count config by @LHT129 in #1511
• feat(quantizer): support extend rabitq by @ShawnShawnYou in #1515
• feat(rabitq): add extend-bit rabitq's simd by @LHT129 in #1532
• feat(stats): add search statistics io_cnt and io_time_ms by @wxyucs in #1547

Improvements

• tests: speedup daily test by reduce dims size by @LHT129 in #1226
• improvement: share vt in hgraph's different layer by @LHT129 in #1225
• tests: speedup daily test by remove some duplicate test by @LHT129 in #1232
• tools: Support eval_dataset's Save H5 File by @LHT129 in #1229
• refactor: avoiding similar declare in simd by @LHT129 in #1239
• [skip ci]chore: update GitHub release workflow by @wxyucs in #1234
• example(hgraph): int8 hgraph demo by @Coien-rr in #1209
• refactor(ivf): move the residual processing logic in IVF down to the bucket level by @inabao in #1276
• refactor(reorder): extract the reorder function into a class by @inabao in #1278
• improvement: split implement of MemoryIO by @LHT129 in #1272
• feat(StreamWriter): check vector not empty before writing it out by @wxyucs in #1277
• chore(tsan): Ignore TSan ch...