v1.9.5

Change log

Features

• #4254 - Add pyroscope integration for continuous profiling on demand

Improvements

• #4309 - Allow to configure default number of shards per node
• #4317 - Allow to overwrite optimizer settings via config
• #4312, #4369 - Improve vector size estimations, making index thresholds more reliable
• #4428 - Improve default maximum segment size, base it on number of CPUs used for indexing
• #4370 - Use consistent RocksDB settings for both put and remove
• #4376 - Improve ordering of insertions and deletions in RocksDB
• #4371 - Log error if segment flushing failed on drop
• #4352 - Promote REST request processing problems from warning to error
• #4368 - Improve error messages in cases of missing vectors
• #4391 - Improve shard state log message, not strictly related to snapshot recovery
• #4414 - Improve Dockerfile, don't invalidate caches each commit and allow debug settings

Bug fixes

• #4402 - Fix deadlock caused by concurrent snapshot and optimization
• #4411 - Fix potentially losing vectors on crash by enabling RocksDB WAL
• #4416, #4440 - Respect max_segment_size on data ingestion with optimizers disabled, create segments as needed
• #4442 - Fix potentially having bad HNSW links on multithreaded systems

v1.9.4

Change log

Bug fixes

• #4332 - Fix potentially losing a segment when creating a snapshot with ongoing updates
• #4342 - Fix potential panic on start if there is no appendable segment
• #4328 - Prevent panic when searching with huge limit

v1.9.3

Change log

Improvements

• #4165 - Handle Out-Of-Disk on insertions gracefully
• #3964 - Faster consensus convergence with batched updates
• #4301 - Deduplicate points by ID for custom sharding

Bug fixes

• #4307 - Fix overflow panic if scroll limit is usize::MAX
• #4322 - Fix panic with missing sparse vectors after recovery of corrupted storage

Web UI

• qdrant/qdrant-web-ui#183 - Notification for miss-configured collections

Full change log: https://github.com/qdrant/qdrant-web-ui/releases/tag/v0.1.26

v1.9.2

Change log

Improvements

• #4148 - Greatly improve proxy segment performance when a lot of changes are tracked
• #4176 - Speed up loading of large memory mapped segments
• #4205 - Use distinct gRPC types for shard replication and overwriting payload requests, existing clients remain compatible
• #4163 - Extend sparse vector benchmarks

Bug fixes

• #4206 - Fix potential deadlock with concurrent searches and snapshot creation

v1.9.1

Change Log

Improvements

• #4119 - performance of the numerical values range filter
• #4103 - Handle antimeridians in geo bounding box filter
• #3931 - Speedup consensus thread
• #4084 - improve consistency of the scroll API with parallel updates

Bug fixes

• #4155 - fixes usage of snapshot shard transfer with API Key Auth.
• #4156 - resolves panic issue on optimization after restart (#4154)
• #4132 - resolves #4131

v1.9.0

Change log

Features 🏎️

• #3477 - Implement and enable shard diff transfer, greatly improve shard transfer speed on node recovery
• #3777 - Role-based access control with JWT
• #3976, #4049, #4065 - Add support for byte vectors, represent your vectors as float32 or uint8

Improvements 🕵️

• #3420, #3938 - Greatly improve write performance while creating a snapshot of a large collection
• #3962, #3971 - Report pending optimizations awaiting an update operation in collection info
• #3998 - Improve indexed_only reliability on proxy shards
• #3798 - Make shard diff transfer fall back to streaming records
• #3784 - Cancel shard transfers when the shard is deleted
• #4037 - Improve sparse vectors search performance by another 7%
• #3982 - Build Qdrant with a single codegen unit to allow better compile-time optimizations
• #4052 - Remove vectors_count from collection info because it is unreliable. Check usage of this field before upgrading
• #3803 - Remove shard transfer method field from abort shard transfer operation

Bug fixes 💔

• #3990 - Fix potentially losing points on crash
• #4034 - Fix potentially losing points on segment flushing
• #4057 - Fix potentially losing points by improper moves between segments when recovering WAL
• #4058 - Fix segments not being flushed properly with rapid sequential optimizations
• #3995, #3996 - Fix read ordering of segments to prevent missing points just being moved
• #3975 - Fix consensus operations getting stuck if they timed out
• #3905 - Fix collection creation if files were left behind on disk
• #4029 - Fix collection deletions conflicting with concurrent collection creation
• #3991 - Fix arbitrary path traversal when uploading snapshots with created collection names
• #4033 - Do not allow symlinks in snapshot archives
• #4047 - Fix recommendation API for sparse vectors
• #3965 - Fix discovery search with sparse vectors over gRPC
• #3928 - Stabilize vector normalization, prevent renormalizing multiple times breaking data consistency
• #4054 - Fix incorrect ordering when using order_by across multiple nodes
• #4067 - Fix timeout being ignored when dropping segment data, potentially leaving them behind forever
• #3980 - Internally forward operations with wait=true during shard transfers
• #3945 - Keep jemalloc only for amd64 and aarch64 to fix build failures

Dashboard 🍱

• qdrant/qdrant-web-ui#175 - Add page to generate RBAC JWT tokens
• qdrant/qdrant-web-ui#169 - Before saving payload changes, ask for confirmation with dialog showing diff
• qdrant/qdrant-web-ui#173 - Validate collection names
• Other changes

---

Thanks to @abhishek818 @RPallas92 @tellet-q @xzfc @ffuugoo @IvanPleshkov @agourlay @coszio @generall @timvisee for their contributions!

v1.8.4

Bug fixes

• #3948 - Fix persisted crash when using space in quoted JSON path

v1.8.3

Bug fixes

• #3856, #3867 - Fix arbitrary path traversal vulnerability in full snapshot REST API

Improvements

• #3837 - Improve error message when doing upsertion with malformed data

Dashboard

• qdrant/qdrant-web-ui#167 - Find similar points by payload key:value pair
• qdrant/qdrant-web-ui#166 - Fix find similar point ID not working in some cases
• qdrant/qdrant-web-ui#165 - Support 64-bit numbers
• Other changes

v1.8.2

Bug Fixes

• #3832
  • fix propagation of max_optimization_threads on the collection params update
  • Improve of indexing optimizer conditions
• #3835 - prevent unnecessary flushes of WAL Clock
• #3776 - make /readyz check reliable on horizontal scaling
• #3806 - fix memory addresses alignment for HNSW graph

v1.8.1

Bug fixes

• #3788 - Allow payload paths starting with numbers