Change log

Features 🍁

• https://github.com/qdrant/qdrant/milestone/23 - Allow server-side score boosting with user-defined formula. See [Docs]
• #6256 - New sum_scores recommendation strategy. Useful for implementing relevance feedback. See [Docs]

Improvements 🌳

• #6325 - Incremental HNSW building. Segment optimized will partially re-use existing HNSW graph on merging segments.
• #6323, #6396 - Improve eviction strategy for unused disk cache
• #6303 - Minor optimizations for memory allocations
• #6385 - Minor internal optimizations
• #6357 - Rethink behavior of offset parameter in case of a query with prefetch. Now offset is only applied to the prefetch result and is not propagated into prefetch query.
• #6326 - Parallelize large segment search batches
• #6390 - Better organized telemetry details levels generate less overhead by excluding segment-level details

Bug Fixes 🌵

• #6289 - Scroll lock: make sure no segments are modified between scroll and retrieve
• #6297, #6293 - Prevent crash on failed shard recovery. Instead, qdrant will load dummy shard, which can be recovered via API
• #6383 - Fix delayed flush wrapper. Before this fix, payload indexes might temporary loose unflushed updates after flush operation is started and before it is finished.
• #6364 - Abort resharding if any Resharding replica is to be marked dead

Web-ui 🌾

• qdrant/qdrant-web-ui#272 - Full query auto-completion

Change log

Improvements

• #6279 - In query API, read vectors/payloads once at shard level instead of in every segment, greatly improve search performance when there's lots of segments
• #6276 - In query API, don't send huge vectors/payloads over internal network, defer reads to greatly improve search performance
• #6260 - Improve performance of resharding transfers, make them faster on slow disks or with high memory pressure

Bug fixes

• #6259 - Fix point estimation in resharding transfers, showing a more reliable ETA
• #6233 - Fix order_by not always including all values for a point if there are multiple

Change log

Improvements

• #6015 - Split CPU budget into CPU and IO to better saturate resources during optimization
• #6088 - Enhance payload indices to handle IsEmpty and IsNull conditions much more efficiently
• #6022, #6023 - Optimize ID tracker in immutable segments by compressing point mappings and versions
• #6056 - Apply undersampling at shard level, significantly improve query performance on large deployments with large search limit
• #6040 - Trigger optimizers more reliably on changes, prevent optimizers potentially getting stuck
• #6085 - Significantly improve performance of point delete propagation during resharding on large deployments
• #6021 - Configure memory barriers in GPU HNSW building to prevent potential race conditions
• #6074 - Use approximate point count at start of shard transfer to make them start quicker
• #6165 - Show log message if hardware reporting is enabled

Bug fixes

• #6212 - Fix user-defined sharding not being applied in consensus snapshots, potentially corrupting cluster
• #6209 - Fix malformed user-defined sharding data in consensus snapshots if using numeric shard keys, potentially corrupting cluster
• #6014 - Fix cluster metadata not being in consensus snapshots, potentially causing cluster state desync
• #6210 - Fix resharding state not being applied with consensus snapshots, potentially causing cluster state desync
• #6202 - Fix snapshot restore error when numeric user-defined shard keys are used
• #6086 - Fix potential panic while propagating point deletions during resharding
• #6032, #6069 - Don't load or restore segments from hidden files, prevent breakage on hidden files in storage by other tools
• #6037 - Fix search panic after HNSW creation with GPU when on NVIDIA
• #6029 - Fix write rate limit not being properly set in strict mode
• #6118 - Do not rate limit reads for shard transfers in strict mode, it's internal
• #6121 - Do not rate limit shard cleanup operation in strict mode, it's internal
• #6152 - Properly rate limit batch point updates
• #6038 - Keep existing shard configuration if snapshot restore failed to prevent panic on startup
• #6010 - Use configured CA certificate for internal snapshot transfers
• #6108, #6115, #6160 - Fix opt-in anonymization of various telemetry fields
• #6065 - Don't show warning if bootstrap URI is not provided

Change log

Improvements

• #5967 - Set maximum number of points in a collection with strict mode

Bug fixes

• b2782ff - Fix HNSW index not respecting on-disk flag, using the inverse setting
• #5991 - Do not deactivate replica or abort transfer if target is missing point, it may not be transferred yet
• #5987 - Fix potential deadlock in consensus logic, which could get a node stuck

Change log

Improvements

• #5903 - Enable consensus compaction by default, enables fast peer joining and recovery
• #5956, #5962 - Delete old point versions on update, prevent old points showing up in reads
• #5870 - Don't include unversioned points in reads, don't include partially persisted points in searches
• #5871 - Don't include unversioned points in writes, don't use partially persisted points in updates
• #5904, #5916, #5950 - Pass peer/bootstrap URI with environment variables, support simpler cluster setups
• #5728 - Improve consensus loop, prevent excessive Raft elections
• #5946 - Normalize URL paths in REST API
• #5917 - Add HTTP Retry-After header in REST response if rate limiter is exhausted
• #5915 - Simplify locks in RocksDB buffer wrapper, use single long rather than two
• #5942 - Add default log format property to configuration file
• #5906 - Update roadmap for 2025

Bug fixes

• #5938 - Fix panic when building of memory mapped sparse vector storage was interrupted
• #5877 - Rate limit prefetches in query API
• #5900, #5905, #5914 - Fix potential panic during consensus compaction
• #5910 - In query API, shortcut on empty retrieve query
• #5908 - Fix flush logic in RocksDB based vector storage, don't eagerly persist changes

Change log

Improvements

• #5891 - Add support for GPUs not featuring half floats, falling back to full floats

Bug fixes

• #5882 - Fix blob store for payloads and sparse vectors potentially panicking when inserting large value
• #5866 - Fix optimizers reusing points with no version, improving data resilience

Change log

Improvements

• #5820 - Improve performance and memory usage of segment merging in optimizers

Bug fixes

• #5848 - Fix potential panic in search after GPU HNSW building
• #5847 - Fix potential panic in GPU HNSW building when having empty payload index
• #5819 - Fix set payload by key on in-memory payload storage not persisting properly
• #5861 - Fix memory mapped sparse vector storage not flushing mappings properly
• #5838 - Fix user-defined sharding not persisting numeric shard keys properly
• #5842 - Do not flush empty memory maps to prevent panic on macOS
• #5843 - Do not flush explicitly when unloading blob storage
• #5845 - Fix potential panic in full text index due to missed bound check

Change log

Features 🎨

• milestone/18 - Add GPU support for HNSW super fast indexing
• milestone/3 - Add resharding in our cloud offering, change the number of shards at runtime
• milestone/13 - Add strict mode to restrict certain type of operations on collections
• #5303 - Add Has Vector filtering condition, check if a named vector is present on a point

Improvements 🚀

• #5783 - Switch to mmap storage for payloads by default to make it more efficient, eliminating unexpected latency spikes
• #5784 - Switch to mmap storage for sparse vectors to make it more efficient, allowing better resource management
• #5781 - Compress HNSW graph links
• #5796 - Remove peer metadata for removed peers
• #5634 - Allow to set max_optimization_threads back to automatic
• #5178 - Stream snapshots in snapshot transfer, don't put snapshots on disk first

Bug fixes 💢

• #5810 - Fix incorrect validation rules on deleted threshold in gRPC API
• #5808 - Don't allow conflicting names for dense and sparse vectors

Thanks to @gulshan-rs @ashwantmanikoth @kartik-gupta-ij @palash25 @redouan-rhazouani @weiwch @n0x29a @pedjak @agourlay @xzfc @JojiiOfficial @tellet-q @coszio @ffuugoo @KShivendu @joein @IvanPleshkov @generall @timvisee for their contributions!

Change log

Improvements

• #5687 - Support 64-bit dimension indices for sparse vectors
• #5609 - Support issues API with limited API keys
• #5602 - Add support for logging in JSON format
• #5630 - Add web UI to Debian package

Bug fixes

• #5629 - Properly flush files with fsync to prevent storage issues
• #5627 - Atomically save quantization metadata
• #5628 - Atomically save chunked mmap configuration
• #5738 - Fix search panic due to unaligned data when loading old Qdrant snapshot (pre v1.8.2)
• #5643 - Improve collection validation, disallow replication factor 0 causing a panic
• #5690 - Ignore empty filter conditions rather than showing an error
• #5676 - Enforce TLS for internal node communication if URL is not explicitly provided

Change log

Improvements

• #5505 - Improve point retrieval across shards by streaming results
• #5521 - Improve point searches across segments by streaming results
• #5514 - Improve facet computing across shards by streaming results
• #5405 - Make /readyz catch up to latest consensus commit
• #5506 - Improve handling of non-transient errors, making shard transfers more robust
• #5478 - When peer starts, cancel all shard transfers related to it
• #5536 - Improve error message on sparse vector validation error
• #5546, #5580 - Improve error messages on various structure variants on validation error
• #5540 - Improve point counting on proxy segments by deduplicating point IDs
• #5522 - Expose shards keys in telemetry
• #5591 - Improve payload performance in gRPC API
• #5486 - Remove support for reading HNSW graphs of Qdrant 0.8.4 and older, simplifying behavior
• #5579 - Improve gRPC logging and error handling
• #5493 - Always log all errors when applying update to replica set

Bug fixes

• #5585 - Improve data consistency, fix point deduplication on start potentially removing newest point version
• #5527, #5528, #5531 - Improve data consistency, fix mixing point versions when applying updates, always use latest point
• #5543 - Improve data consistency, fix segment builder mixing point versions during segment optimization
• #5573 - Improve data consistency, fix proxy segment not using correct point versions when propagating point deletes
• #5581 - Improve data consistency, fix proxy segment not using correct versions when propagating payload index changes
• #5553 - Improve data consistency, fix proxy segment not flushing changes if sharing a write segment
• #5557 - Improve data consistency, fix flushing proxied segment acknowledging operations that were not actually persisted
• #5510 - Improve data consistency, fix marking points as deleted not working in all cases
• #5598 - Fix peer getting stuck on start if consensus snapshots are used
• #5484 - Fix panic on huge search limit
• #5600 - Fix prefetch with group-by internally using wrong request limit, significantly reducing search time
• #5488 - Fix incorrect timeout handling in distance matrix API
• #5495 - Fix incorrect timeout handling on remote shards
• #5596 - Fix broken replication factor in gRPC create shard key API
• #5551 - Fix potential division by zero in facets API
• #5593, #5603, #5605, qdrant/qdrant-web-ui#262 - Bump some dependencies to patch potential security vulnerabilities