LanceDB v0.37.1

🛠 Breaking Changes

• feat(rust)!: make add_columns a builder by @wkalt in #3778
• fix(python): avoid debugger deadlock during connection inspection by @lancedb-gatefixer[bot] in #3788
• feat: infer maintained indexes when an LsmWriteSpec omits them by @hamersaw in #3748

🎉 New Features

• feat(fts): add block size configuration by @BubbleCal in #3691
• feat(query): add use_lsm to read MemWAL LSM data by @touch-of-grey in #3489
• feat: support batched blob range reads by @Xuanwo in #3703
• feat(python): expose AsyncTable.to_lance by @u70b3 in #3730
• feat(python): configure streaming transform parallelism by @yaodong-shen in #3699
• feat(fts): support custom stop-word lists by @BubbleCal in #3734
• feat(remote): add RemoteTable fetch_blobs HTTP client by @geruh in #3684
• feat(python): add namespace/table exist support by @FarmerChillax in #3460
• feat: create_index returns a Job handle by @wkalt in #3742
• feat: connection-level job operations by @wkalt in #3755
• feat(remote): add seekable blob range reads by @geruh in #3750
• feat(rust)!: make add_columns a builder by @wkalt in #3778
• feat(table): checkpoint_lsm, flush_lsm, compact_lsm, get_lsm_stats by @hamersaw in #3736
• feat: infer maintained indexes when an LsmWriteSpec omits them by @hamersaw in #3748

🐛 Bug Fixes

• fix: crash in AnswerdotaiRerankers/ColbertReranker for return_score="all" by @chuenchen309 in #3671
• fix: preserve row count in MetadataEraserExec for zero-column batches by @vivek-bharathan in #3717
• fix: accept either timestamp or timestamp_millis for versions by @dantasse in #3733
• fix(node): reject non-string Arrow metadata by @u70b3 in #3728
• fix: deflake test_read_consistency_interval by @wjones127 in #3713
• fix(python): avoid async work in sync reprs by @ssrheart425 in #3620
• fix(remote): surface masked merge_insert stream errors under HTTP2 (#2339) by @sanskar-singh-2403 in #3711
• fix: distinguish corrupt table from missing in open_table by @joaquinhuigomez in #3731
• fix(python): log when storage_options is ignored in RemoteDBConnection.open_table by @ckarnell in #3743
• fix: hybrid search minimum_nprobes(0) silently no-ops instead of raising by @Adityaj0 in #3770
• fix: when_not_matched_by_source_delete() doesn't reset a previously-set condition by @Adityaj0 in #3771
• fix(python): avoid debugger deadlock during connection inspection by @lancedb-gatefixer[bot] in #3788
• fix(python): guard concurrent table deletes by @lancedb-gatefixer[bot] in #3787
• fix(python): report unsplittable IVF partition errors by @lancedb-gatefixer[bot] in #3846
• fix(python): align wheel ABI with supported versions by @lancedb-gatefixer[bot] in #3884
• fix(python): prevent OpenSSL linkage in Linux wheels by @lancedb-gatefixer[bot] in #3877
• fix(python): instruct dimension probe for instructor embeddings by @lancedb-gatefixer[bot] in #3874
• fix(python): support nullable pandas merge input by @lancedb-gatefixer[bot] in #3864
• fix(python): cover nullable list v2.2 decoding by @lancedb-gatefixer[bot] in #3853
• fix(python): preserve repeated indexed merge inserts by @lancedb-gatefixer[bot] in #3850
• fix(rust): prevent vector optimize regression after deletes by @lancedb-gatefixer[bot] in #3848
• fix(remote): validate cloud DNS hostnames by @lancedb-gatefixer[bot] in #3845
• fix(rust): handle missing mirrored copy sources by @lancedb-gatefixer[bot] in #3843
• fix: percent-encode index names in per-index remote REST paths by @wkalt in #3840
• fix(python): accept Arrow scalars in table updates by @lancedb-gatefixer[bot] in #3838
• fix(python): clarify bare Vector annotations by @lancedb-gatefixer[bot] in #3809
• fix(python): expose inline types to downstream checkers by @lancedb-gatefixer[bot] in #3817
• fix(node): require compatible Node.js types by @lancedb-gatefixer[bot] in #3829
• fix(python): support Polars 1.32 table scans by @lancedb-gatefixer[bot] in #3801
• fix(node): cover non-nullable embedding schema append by @lancedb-gatefixer[bot] in #3835
• fix: explain unsupported object storage mounts by @lancedb-gatefixer[bot] in #3823
• fix(node): cover offset pagination in search by @lancedb-gatefixer[bot] in #3814
• fix: avoid AttributeError in JinaEmbeddings image input for str/Path by @chuenchen309 in #3670
• fix: get table size from metadata, not files by @dantasse in #3790
• fix(node): preserve nested Arrow data across versions by @lancedb-gatefixer[bot] in #3900

📚 Documentation

• docs(python): fill gaps in the Python API reference by @wjones127 in #3746

🔧 Build and CI

• ci: cut Actions cost by moving builds to free runners and fixing caches by @wjones127 in #3735
• ci(docs): add scheduled doc link check by @Xuanwo in #3888