New Features

Support AsyncIO (EXPERIMENTAL)

• feat: asyncio support by @brcarry in #2459

Introducing the native asyncio client in PyMilvus: AsyncMilvusClient. While only a subset of APIs is currently supported, this marks a significant step forward for PyMilvus. Enjoy exploring it!

import asyncio

from pymilvus import MilvusClient, AsyncMilvusClient, DataType


async def main():
    URI = "./milvus.db"
    async_client = AsyncMilvusClient(uri=URI)
    await async_client.create_collection("async_example", dimension=768)
    await async_client.drop_collection("async_example")


if __name__ == "__main__":
    asyncio.run(main())

Other new features

• enhance: grant/revoke v2 optional db and collection params by @shaoting-huang in #2392
• feat: support recalls field in SearchResult by @chasingegg in #2389
• feature: support milvus-client iterator by @MrPresent-Han in #2465
• enhance: add use_database by @czs007 in #2493

Enhancements

• extend unlimted offset for query iterator(#2418) by @MrPresent-Han in #2420
• enhance: alterindex & altercollection supports altering properties by @JsDove in #2423
• enhance: rbac privilege group and grant v2 examples by @shaoting-huang in #2426
• enhance: alterdatabase support delete property by @JsDove in #2439
• enhance: add search_group_by example for milvus-client(#2481) by @MrPresent-Han in #2483
• fix: Support hybrid search with expression template by @xiaocai2333 in #2490
• enhance: Add example for milvus client about expression template by @xiaocai2333 in #2445
• enhance: createdatabase support properties by @JsDove in #2450
• enhance: refine search_group_by_example(#2481) by @MrPresent-Han in #2485

Bug fixes

• fix: Unify logger and correct logging settings (#2397) by @XuanYang-cn in #2402
• fix: fix list aliases rpc call and grant/revoke v2 by @shaoting-huang in #2404
• fix: Add is_clustering params for get_compaction_plans in orm by @xiaocai2333 in #2429
• fix: fix describe database return type by @JsDove in #2453
• fix: resolve conflicts and update connections.py by @brcarry in #2462
• fix: add authorization_interceptor and db_interceptor to async channel by @brcarry in #2472
• fix: ensure create_index and load_collection are fully completed by @brcarry in #2477

Full Changelog: v2.4.10...v2.4.11

New features:

• Report cost in the dml and dql request by @SimFG in #2055
• feat: Support major compaction in ManualCompaction by @wayblink in #2015
• Support Milvus Lite by @junjiejiangjjj in #2073
• Enable set_properties and describe_database api by @weiliu1031 in #2082
• Allowing search iterator on sparse float vector field by @zhengbuqian in #2104
• Accept list of single row scipy.sparse object as input for insert/search by @zhengbuqian in #2111
• Support float16/bfloat16/sparse vector for bulkwriter by @yhmo in #2127
• Support group_size parameter for search_group_by by @MrPresent-Han in #2130
• Upsert support autoid by @smellthemoon in #2173
• Expose reduce_stop_for_best to users(#2181) by @MrPresent-Han in #2183
• Bulkinsert supports importing binlog by @yhmo in #2222
• Add page_retain_order param during search with offset by @PwzXxm in #2229
• Support load with Field Partial load by @congqixia in #2228
• Support null and default value by @smellthemoon in #2234
• Support the mmap_enable param in the field schema by @SimFG in #2238
• Add strict_group_size and rank_group_scorer for hybrid_search(#2253) by @MrPresent-Han in #2254
• Load configuration without altering the environment by @laipz8200 in #2192
• Update proto to get is_sorted field for GetQuerySegmentInfo by @xiaocai2333 in #2280
• Support mvcc and break-down-continue for iterator(#2278) by @MrPresent-Han in #2279
• Support new Function feature by @zhengbuqian in #2257
• Supports filling elements through templates for expression by @xiaocai2333 in #2317
• Add compact, get_server_version and flush api by @czs007 in #2326

Enhancements

• Upgrade the bulkWriter cloud API call from v1 to v2 by @lentitude2tk in #2245
• enhance: tidy dependencies in tests by @XuanYang-cn in #1996
• Update sparse+dense hybrid search example by @zhengbuqian in #2005
• Support UDS by @junjiejiangjjj in #2023
• enhance: Update hello_hybrid_sparse_dense.py example to include BGE reranker by @zhengbuqian in #2028
• enhance: add resource group declarative api by @chyezh in #2002
• enhance: Make bulk_writer's requirments optional by @XuanYang-cn in #2086
• Added grpc as a valid protocol for uri by @brunocfnba in #2090
• enhance: print search result more elegantly by @longjiquan in #2123
• enhance: Check PyMilvus on Windows platform by @XuanYang-cn in #2136
• Export indexed rows for describe_index by @xiaocai2333 in #2148
• Add database operations to MilvusClient by @ashkrisk in #2152
• enhance: enable setting properties during create database by @weiliu1031 in #2168
• enhance: hide zero values when printing by @SimFG in #2200
• enhance: use info level for retry message by @XuanYang-cn in #2212
• enhance: Make load parameter naming normal by @congqixia in #2243
• enhance: add bitmap index example by @zhagnlu in #2376

Bug fixes

• fix: Remove np.view for floatvector by @XuanYang-cn in #2048
• fix: pass offset parameter for hybrid search to server by @czs007 in #2051
• Change sparse related errors to ParamError by @zhengbuqian in #2066
• modified example_tls1.py through MilvusClient by @nish112022 in #2065
• remove scipy dependency for sparse while still supporting scipy sparse matrix by @zhengbuqian in #2074
• modified example_tls2.py through MilvusClient by @nish112022 in #2077
• fix sparse: accpet int/float wrapped in string by @zhengbuqian in #2094
• enhance: Expand grpcio version to latest by @XuanYang-cn in #2093
• fix the str function of the extra list by @SimFG in #2099
• fix: wrong expr handling due to and/or priorities(#2113) by @MrPresent-Han in #2114
• fix: use the existed index_name by @smellthemoon in #2107
• change the load status check interval to 0.2s by @SimFG in #2121
• Fix a bug of bulkwriter by @yhmo in #2133
• feat: disable installation of milvus-lite on windows platform (#2131) by @Raysilience in #2134
• enhance: update milvus-proto and correct index.drop() by @XuanYang-cn in #2142
• enhance: Update readme to the latest by @XuanYang-cn in #2146
• Refine the error message for type mismatches during data insertion by @xiaocai2333 in #2156
• fix: Coding style by latest ruff by @XuanYang-cn in #2159
• enhance: Remove the logic to set replica_number=1 by default by @weiliu1031 in #2163
• fix: Edit setuptools upper version for py380 by @XuanYang-cn in #2182
• Fix comment in iterator implementation by @ashkrisk in #2187
• Don't export rows info in index params by @xiaocai2333 in #2190
• fix: remove limitation clustering key can not be primary key by @wayblink in #2194
• fix: unclear error msg for varchar field by @XuanYang-cn in #2208
• Limit milvus-lite files to end with .db by @junjiejiangjjj in #2214
• fix: select a single column consisting of a list of column names by @smellthemoon in #2210
• Rename is_major to is_clustering by @wayblink in #2219
• fix: move page_retain_order to the same level as radius by @PwzXxm in #2249
• fix: not report error when setting default_value=None by @smellthemoon in #2251
• enhance: Bulkinsert support csv by @OxalisCu in #2247
• add hybrid_search for MilvusClient by @czs007 in #2258
• feat: support keyword text match by @longjiquan in #2256
• enhance: support print iterator info(#2261) by @MrPresent-Han in #2262
• fix: not put 'default_value' in dict by @smellthemoon in #2272
• fix not check None Type in json and support no need to pass in None when insert row by @smellthemoon in #2275
• remove analyzer_params, added enable_tokenizer and tokenizer_params by @zhengbuqian in #2273
• fix: upsert rows when set autoid==true fail by @smellthemoon in #2286
• feat: allow empty sparse row by @zhengbuqian in #2291
• fix: reset offset to zero after seek(#2292) by @MrPresent-Han in #2293
• fix: remove duplicate field name check in pymilvus by @zhengbuqian in #2294
• enhance: enable compatible for iterator(#2278) by @MrPresent-Han in #2297
• enhance: Enable bulkwriter to support import v2 by @bigsheeper in #2295
• fix: row based insert/upsert when there is Function in schema by @zhengbuqian in #2298
• Improve embedding retrieval performance by @yhmo in #2300
• fix: add BM25 to supported metric type of search iterator by @zhengbuqian in #2301
• fix: simplified the logic to check if the insert/request data matches the schema by @zhengbuqian in #2303
• fix: Passing messages to code in ParamError by @XuanYang-cn in https://github.com/milvus-io/pymilvus...

What's Changed

• Corrected grammar and consistency in error messages (#2289) by @czs007 in #2330
• enhance: Update the template expression proto to improve transmission efficiency by @xiaocai2333 in #2336
• Use filter_params for milvus client by @xiaocai2333 in #2321
• fix insertto avoid describe collection on every insert call by @zhengbuqian in #2348
• Remove unnecessary invoke of describe_collection() by @yhmo in #2346
• fix: Add upper limit for grpcio to 1.67.1 by @XuanYang-cn in #2360
• Multi cherry picks from master branch by @XuanYang-cn in #2349
• fix: cannot pass consistency level for delete by @XuanYang-cn in #2351
• Add default values to Pop to prevent exceptions by @czs007 in #2374
• enhance: Enhance log messages when Connect with server failed by @XuanYang-cn in #2378
• enhance: Prepare 2.4.10 release by @XuanYang-cn in #2380

Full Changelog: v2.4.9...v2.4.10

What's Changed

• Improve embedding retrieval performance by @yhmo in #2290
• fix: Passing messages to code in ParamError (#2304) by @XuanYang-cn in #2305
• Supports filling elements through templates for expression by @xiaocai2333 in #2308

Full Changelog: v2.4.8...v2.4.9

What's Changed

• fix: upsert rows when set autoid==true fail(#2286) by @smellthemoon in #2287
• Upgrade the bulkWriter cloud API call from v1 to v2 by @lentitude2tk in #2246
• enhance: Enable bulkwriter to support import v2 (#2295) by @bigsheeper in #2296

Full Changelog: v2.4.7...v2.4.8

What's Changed

• enhance: Make load parameter naming normal (#2243) by @congqixia in #2244
• Add hybrid_search for MilvusClient by @czs007 in #2259

Bug fixes:

• fix: move page_retain_order to the same level as radius by @PwzXxm in #2250

Full Changelog: v2.4.6...v2.4.7

New Features

• Bulkinsert supports importing binlog by @yhmo in #2223
• enhance: upsert support autoid (#2173) by @smellthemoon in #2224
• enhance: Support load with Field Partial load by @congqixia in #2227
• feat: add page_retain_order param during search with offset by @PwzXxm in #2230
• feat: support the mmap_enable param in the field schema by @SimFG in #2239

Enhancement

• enhance: loose grpcio upper limit by @XuanYang-cn in #2240

Full Changelog: v2.4.5...v2.4.6

New features

• feat: Support clustering compaction by @wayblink in #2220

Enhancements

• Export indexed rows for describe_index (#2148) by @xiaocai2333 in #2149
• Refine the error message for type mismatches during data insertion (#2156) by @xiaocai2333 in #2157
• Add database operations to MilvusClient (#2152) by @ashkrisk in #2164
• enhance: Remove the logic to set replica_number=1 by default (#2163) by @weiliu1031 in #2167
• enhance: enable setting properties during create database (#2168) by @weiliu1031 in #2174
• enhance:expose reduce_stop_for_best to users(#2181) by @MrPresent-Han in #2188
• enhance: hide zero values ​​when printing by @SimFG in #2201
• Limit milvus-lite files to end with .db by @junjiejiangjjj in #2216
• enhance: use info level for retry message by @XuanYang-cn in #2213

Bug fixes

• fix: Coding style by latest ruff by @XuanYang-cn in #2158
• fix: Edit setuptools upper version for py380 by @XuanYang-cn in #2180
• Don't export rows info in index params by @xiaocai2333 in #2191
• fix: remove limitation clustering key can not be primary key by @wayblink in #2196
• fix: unclear error msg for varchar field by @XuanYang-cn in #2209
• fix: select a single column consisting of a list of column names(#2210) by @smellthemoon in #2211

Full Changelog: v2.4.4...v2.4.5

New Features

• Allowing search iterator on sparse float vector field by @zhengbuqian in #2105
• Accept list of single row scipy.sparse object as input for insert/search by @zhengbuqian in #2112
• Support float16/bfloat16/sparse vector for bulkwriter by @yhmo in #2128

Enhancements

• enhance: print search result more elegantly (#2123) by @longjiquan in #2124
• Change the load status check interval to 0.2s by @SimFG in #2125
• Disable installation of milvus-lite on windows platform by @XuanYang-cn @Raysilience in #2137

Bug fixes

• Fix wrong expr handling due to and/or priorities(#2113) by @MrPresent-Han in #2115
• Fix: use the existed index_name (#2107) by @smellthemoon in #2118
• Fix a bug of bulkwriter by @yhmo in #2132
• Fix drop_index got multiple values for keyword argument by @XuanYang-cn in #2138

Full Changelog: v2.4.3...v2.4.4

What's Changed

• support the report value in the dml and dql request by @SimFG in #2092
• enhance: Expand grpcio version to latest by @XuanYang-cn in #2096

Bug fixes:

• fix sparse: accpet int/float wrapped in string by @zhengbuqian in #2095
• fix the str function of the extra list by @SimFG in #2098

Full Changelog: v2.4.2...v2.4.3