github qdrant/qdrant v1.6.0

latest releases: v1.9.7, v1.9.6, v1.9.5...
8 months ago

Change log

Features πŸ—ΊοΈ

  • #2538, #2662, #2661 - Add new recommendation engine, score results directly from examples
  • #2693 - Extend recommendation API, support raw vectors along with point IDs as input
  • #2139, #2315, #2766 - Add support for filtering geo coordinates by polygon
  • #2599 - Add option to tune shard update parallelism, improve performance for large clusters
  • #2642, #2717, #2728, #2741, #2728, #2746 - Distribute searches to other nodes if the current is busy, improve search latency on large clusters

Improvements πŸ•΅οΈ

  • #2756 - Add support for specifying a path to retrieve a subset of payload
  • #2710 - Fix cardinality estimation for Any filter, improve search performance
  • #2744 - Use boolean index in filtering, improve search performance
  • #2628 - Add automatic selection of search rescoring mode, improve binary search accuracy
  • #2764 - Improve memory usage of geo index
  • #2666 - Add immutable numeric index, decreasing memory usage by a factor of 3
  • #2745 - Return validation error if filters only have a key but no value specified
  • #2615 - Ensure segment compatibility when a collection is loaded
  • #2738 - Add gRPC compliant health check method

Bug fixes πŸ“Έ

  • #2733 - Fix potential collection deadlock by simplifying segment access patterns
  • #2638, #2652, #2528, #2683, #2720 - Fix IO errors on snapshot creation
  • #2742 - Fix crash when using range filter with same value as lower and upper bound
  • #2759 - Fix collection stuck as initializing after crash in single node deployment
  • #2755 - Fix ordering parameter not working for batch point update requests
  • #2724 - Fix temporary files left behind by snapshot creation
  • #2709 - Remove obsolete storage log files more aggressively, prevent them consuming too much space
  • #2765 - Improve documentation for collection update vector parameters

FastEmbed ⚑

  • qdrant/fastembed#13 - Remove bulky dependencies, reduce dependency size by a factor of 25
  • Add new models: sentence-transformers/all-MiniLM-L6-v2, intfloat/multilingual-e5-large
  • Add API to list supported models, including: BAAI/bge-small-en, BAAI/bge-base-en
  • Other changes

Thanks to @kwkr @etolbakov @smark-d @dzhao @zzzz-vincent @jsoref @ffuugoo @IvanPleshkov @agourlay @coszio @timvisee for their contributions!

Don't miss a new qdrant release

NewReleases is sending notifications on new releases.