v1.32.17
📦 weaviateView on GitHub →
✨ 4 features🐛 16 fixes🔧 4 symbols
Summary
This release focuses heavily on bug fixes, performance improvements, especially around latency and replication scaling, and adds support for new embedding models and Geo HNSW configuration.
✨ New Features
- Add possibility to configure Geo HNSW Index settings.
- Introduce replication scaling plan.
- Update replication scale URL to include collection and replication factor parameters.
- Add support for batch API in text2vec-google module.
🐛 Bug Fixes
- Ensure `rf=3` single-tenant collections search locally in parallel.
- Avoid N+1 problem when calculating first strong hash.
- Reduce lock contention in `vectorCachePrefiller`.
- Enhance WriteTo method to ensure scratch space is cleaned up and synced after writing.
- Update WriteTo method to return written bytes and handle scratch space cleanup errors.
- Pass worker ctx for retry to respect worker ctx timeout in Replication:Pull.
- Use randomly generated docker container names for Weaviate container.
- BM25 wand minimum should match edge case.
- Make TestNetworkIsolationSplitBrain to be less flaky.
- Adapt to new jina reranker return.
- Fix aggregate with target vectors.
- Make backups delete safe.
- Replace tryOnce with doRetry in OverwriteObjects.
- Improve scratch space cleanup.
- Optimize segment strategy handling in DocPointerWithScoreList.
- Fix data race in TestRenaming test.
🔧 Affected Symbols
WriteTovectorCachePrefillerOverwriteObjectsDocPointerWithScoreList