Announcing Lance SDK v1.0.0
Lance now backs production workloads across multiple organizations, and the SDK has reached a level of stability where clearer versioning and release guarantees matter.
With 1.0.0, the Lance SDK formally adopts semantic versioning and a community-driven release process. For instance, community review and voting on stable release candidates, major / minor / patch releases with explicit compatibility guarantees, and a clear separation between SDK versioning and the Lance file format, table format, and namespace specs.
Save the Date: Lance Community Sync on 1/15
Kicking off the new year with our 1st Lance Community Sync on Thursday, January 15! Everyone is welcome to join. We’ll be keeping a running meeting notes document so feel free add any discussion topics there.
Subscribe to the Lance Format mailing list to receive the meeting invite and get notified.
Search Across 41M+ Wikipedia Articles with Wikisearch
Wikisearch is a search engine that stores and searches through 41M+ Wikipedia articles. The demo showcases the combination of LanceDB’s semantic and keyword search features to quickly find relevant information in a large dataset like Wikipedia.
Here’s how we built it: https://lancedb.com/blog/feature-full-text-search/
📺 Watch the Recordings!
📅 Upcoming Events
If you’re attending Data Day Texas 2026 , check out Prashanth’s talk on January 24 where he’ll be diving deep into the internals of Lance and LanceDB and how they are the architectural answer for modern ML/AI and multimodal workloads.
📊 LanceDB Enterprise Product News
| Feature | Description |
|---|---|
| Add IVF_RQ, RabitQ index support | IVF_PQ RabitQ index offers error-bound quantization and outperforms PQ in recall-efficiency tradeoff – 4x faster indexing time, 12% higher recall, and 22% faster search in most cases. |
| Federated database | Connect LanceDB Enterprise to any catalog services that already store Lance tables. |
🌟 Open Source Releases Spotlight
| Project & Version | Description |
|---|---|
| Lance 1.0.0 - 1.0.1 Release notes |
GeoArrow-based GEO types and spatial SQL (
#4678
), access HuggingFace datasets via hf://datasets/{repo}/path/to/file.lance (
#5353
,
#2599
), Arrow-native Blob v2 storage (
#5239
,
#5189
), JSON functions in ds.sql (
#5350
), lightweight index introspection via describe_indices() (
#5221
), Change Data Feed (CDF) APIs for incremental pipelines (
#5091
), and cloud-native namespaces with credential vending and REST support (
#4984
,
#5161
,
#5256
,
#5292
,
#5309
) Filtered scans up to 19× faster ( #5165 ), 10–20% lower cold latency for FTS ( #5363 ), ~20% faster HNSW search ( #5377 ), dynamic IVF pruning with minimum_nprobes defaulting to 1 (
#4773
), and reduced metadata I/O via inline transactions stored directly in manifests (
#4774
) |
| LanceDB 0.23.0 - 0.23.1 Release notes |
IvfSq scalar-quantized IVF index and HNSW alias names (HnswPq, HnswSq) (
#2832
), stable row IDs via storage_options (
#2831
), auto-selected IVF num_partitions (
#2775
), credential-vended namespaces (
#2778
), server-side query_table (
#2811
), async Python namespace connections (
#2788
), remote table head() support (
#2793
), and Java SDK support for REST-based namespaces (
#2845
,
#2851
) |
| Lance Namespace 0.3.0 - 0.4.3 Release notes |
Standardized error model ( #282 ), clearer REST and directory implementation specs ( #283 ), credential vending with table metadata support ( #291 ), and identity included on all namespace requests for auditing and access control ( #292 ) |
All catalog integrations have been migrated to a new repository: lance-format/lance-namespace-impls
This repository contains namespace implementations for Apache Hive MetaStore, Apache Polaris, Unity Catalog, Apache Iceberg REST Catalog, AWS Glue and more. The main lance-namespace repository now focuses solely on the specification, while implementations live in this dedicated impls repository.
🫶 Community Contributions
Thank you to contributors from ByteDance, OpenAI, Ants Group, Uber, Datastrato, Xiaomi, etc. for these notable contributions:
- @ddupg (GeoArrow-based GEO types, RTree geospatial index spec)
- @xloya (Map type support in format v2.2)
- @chenghao-guo (distributed vector index build)
- @wojiaodoubao (FTS filtering in vector search, struct sub-column access, nested struct merge fixes, partitioning spec)
- @steFaiz (distributed range BTree index, Java schema metadata API)
- @majin1102 (inline transactions, global tag API)
- @yanghua (Python DatasetDeltaBuilder, Java row lineage & CDF APIs)
-
@zhangyue19921010
(JSON functions in
ds.sql) - @HaochengLIU (zonemap/bloom filter fixes for deletions)
-
@mykolaskrynnyk
(async
to_pydanticwith field aliases) - @fangbo (Spark-based BTree indexing, Java 21 support, VACUUM)
- @c21 (large-scale Spark write stability)
We want to especially highlight the initial release of
lance-graph
contributed by Uber and
native support for Lance REST namespace in Apache Gravatino
.
A heartfelt thank you to the rest of our community contributors of lance and lancedb this past month: @shiyajuan123 @jaystarshot @fenfeng9 @fangyinc @Pmathsun @ztorchan @fMeow @valkum @rongou @camilesing @YinZheng-Sun @XuQianJin-Stars @timsaucer @hushengquan @jonded94 @hfutatzhanghb @jachewz @yuqi1129 @AndreaBozzo @tomsanbear @kaori-seasons @maninc @wayneadams @Angryrou @fzowl @Kkartik14 @YouthLab @philipp-horstenkamp
