v0.23.0
Breaking Changes📦 lancedbView on GitHub →
⚠ 2 breaking✨ 17 features🐛 4 fixes🔧 4 symbols
Summary
This release introduces significant features like async namespace support, IVF SQ indexing, and stable row IDs, while also deprecating macOS x86 support and refactoring namespace model usage.
⚠️ Breaking Changes
- Support for macOS x86 architecture has been deprecated.
- Namespace operations now use namespace models directly instead of previous methods. Users should update calls related to namespace operations to use the new direct model approach.
Migration Steps
- Review usage of namespace operations to align with the new direct use of namespace models.
- If encountering issues with macOS builds, note that x86 support is deprecated.
✨ New Features
- Documentation scope reduced to only show API references.
- Lance version bumped to 0.40-0-beta.2.
- Lance now determines the default value for the `num_partitions` parameter.
- Support added for vending namespace credentials.
- Added `num_attempts` parameter to merge insert results.
- Support for `to_pydantic` in asynchronous operations in Python.
- Support for asynchronous namespace connection.
- Implementation of `head()` functionality for remote tables.
- Codex URL key updated.
- Remote default features removed from lance-namespace-impls.
- Support for stable row IDs via `storage_options`.
- Support for server-side querying of namespaces.
- IVF SQ index support added, along with HNSW aliases.
- Use of REST namespace for the LanceDB Java SDK.
- Java client builder made generic.
- Vector type inference updated: defaults to float32 if input integers are outside the uint8 range.
- Upgrade of lance-namespace Python package to version 0.3.2.
🐛 Bug Fixes
- Schema metadata is now converted to strings for `JsonArrowSchema`.
- Default value for namespace is now set to `None`.
- Fixed issue where `.phrase_query()` was not functioning correctly.
- Resolved error related to `table_names` at the root namespace.
🔧 Affected Symbols
lance-namespace-implsJsonArrowSchema.phrase_query()table_names