1.15.0

Change log

Features 📜

• milestone/25 - Phrase Matching in Full-Text index [Docs]
• milestone/28 - Stop words support in Full-Text index [Docs]
• milestone/29 - New multilingual tokenizer enabled by default [Docs]
• milestone/32 - Snowball Stemmer in Full-Text index, [Docs]
• milestone/30 - Asymmetric binary quantization [Docs]
• milestone/31 - 2-bit and 1.5-bit Binary Quantization encoding [Docs]
• milestone/26 - Maximum Marginal Relevance [Docs]
• https://github.com/qdrant/qdrant/pull/6682 - Verify file system compatibility on process start

Improvements 🪨

• milestone/16 - Migrate away from RocksDB
• milestone/27 - HNSW Healing on optimization
• https://github.com/qdrant/qdrant/pull/6479 - Speedup additional links construction by measuring sub-graph connectivity
• https://github.com/qdrant/qdrant/pull/6563, https://github.com/qdrant/qdrant/pull/6601 - Inference usage now reported in API response
• https://github.com/qdrant/qdrant/pull/6611 - Added major and minor tags for Docker builds
• https://github.com/qdrant/qdrant/pull/6685 - Speed up mutable posting list for upsertions
• https://github.com/qdrant/qdrant/pull/6729 - Reuse quantized vectors during HNSW index building
• https://github.com/qdrant/qdrant/pull/6713 - Optimize Gridstore insertions
• https://github.com/qdrant/qdrant/pull/6750 - Use sequential reads when iterating over Gridstore
• https://github.com/qdrant/qdrant/pull/6757 - Enable pod role based auth for S3 snapshots
• https://github.com/qdrant/qdrant/pull/6725 - Improve read request handling during snapshot recovery
• https://github.com/qdrant/qdrant/pull/6785 - Avoid re-optimizations when loading a mmap-backed immutable text index
• https://github.com/qdrant/qdrant/pull/6778, https://github.com/qdrant/qdrant/pull/6817 - Properly handle inconsistencies in consensus during custom shards creation
• https://github.com/qdrant/qdrant/pull/6779 - Avoid IO during evaluating HNSW index for re-optimization
• https://github.com/qdrant/qdrant/pull/6796 - Use sequential reads for building quantized vectors
• https://github.com/qdrant/qdrant/pull/6794 - Avoid double condition check during filtering
• https://github.com/qdrant/qdrant/pull/6804 - Optimize has_id condition
• https://github.com/qdrant/qdrant/pull/6844 - Adjust default values for better performance

Bug fixes 🫙

• https://github.com/qdrant/qdrant/pull/6593 - Use ID tracker to estimate internal points range
• https://github.com/qdrant/qdrant/pull/6614 - Improve Gridstore wipe/clear functionality
• https://github.com/qdrant/qdrant/pull/6632 - Fix default replication factor
• https://github.com/qdrant/qdrant/pull/6640 - Fix significant memory usage in Distance Matrix API on high sample size
• https://github.com/qdrant/qdrant/pull/6739 - Fix IDF computation with deleted points
• https://github.com/qdrant/qdrant/pull/6755 - Prevent shards from getting stuck in Initializing state
• https://github.com/qdrant/qdrant/pull/6759 - Properly propagate inference errors
• https://github.com/qdrant/qdrant/pull/6787 - Add flag to break optimization loop on panic
• https://github.com/qdrant/qdrant/pull/6862 - Avoid blocking async runtime when loading local shard
• https://github.com/qdrant/qdrant/pull/6800 - Fix bug that causes all replicas to die if node restarted during resharding
• https://github.com/qdrant/qdrant/pull/6882 - Fix broken is_empty filter on new points

Web-UI 🎨

• https://github.com/qdrant/qdrant-web-ui/releases/tag/v0.1.41 - New "Create Collection" form, Simplify JWT form, other fixes

Deprecations 🚧

• init_from parameter of the collection creation API is deprecated and going to be removed in v1.16

  • Reason: init_from creates false expectation of data consistency and reliability of this operation.
  • Alternative: prefer usage of migration tool or snapshots, read more in the tutorial

• Payload based filters in JWT going to be removed in v1.16

  • Reason: Payload based filters are inconsistent for read/write operations. Many update operations are not implemented due to security risks and complexity. Lack of real-life usage.
  • Alternative: prefer collection-based access control.

• mmap_threshold parameter of the collection optimizer is deprecated and going to be removed in v1.16

  • Reason: starting from v1.12, all vector storages exclusively use memory maps, explicit usage of mmap_threshold is obsolete
  • Alternative: control memory cache behaviour of Qdrant storages with on_disk: true/false parameter

• Lock API is deprecated and going to be removed in v1.16

  • Reason: lack of real-life usage
  • Alternative: JWT tokens can be configured to provide read-only access to Qdrant

• https://github.com/qdrant/qdrant/pull/6760 - Removed max_optimization_threads from config