Change log

Features 📅

• Add ability to order Scroll API results by payload value - #3493, #3533, #3562, #3540, #3589, #3726
• Add date time payload index to efficiently filter by date time ranges - #3395, #3516, #3549, #3578, #3622, #3729, #3730
• #3472 - Add API to check if a collection exists
• #3466 - Allow specifying a minimum number of conditions to match in payload filters
• #3548 - Improve set_payload API, adds ability to modify nested fields

Improvements 🏎️

• Sparse vector search performance improvements, making it 16x faster - #3464, #3275, #3327, #3348, #3347, #3368, #3393, #3413, #3424, #3527, #3560, #3609, #3619, #3605
• Improve CPU saturation, greatly speeding up indexing on high CPU systems - #3364, #3115
• Improve various payload indices
  • #3497 - Reduce memory usage for the text index
  • #3297 - Reduce memory usage for geo index
  • #3625 - Improve loading time for test field index
  • #3563 - Compress text index, greatly improving tail latencies and memory usage
• Add new release artifacts to support more platforms
  • #3342 - MUSL binaries for x86_64 and AArch64
  • #3343 - Portable AppImage binary
  • #3344 - Debian package with *.deb archive
• Avoid re-indexing point if fields without an index are updated - #3611
• Refactor payload key parsing - #3674, #3697, #3709
• #3367 - Ensure consistency when an update request is dropped in the middle
• #3525 - Optimize MatchAny performance with a large number of values
• #3629, #3633, #3641 - Performance optimizations for empty payloads
• #2840, #3381 - Add a checksum for snapshot files
• #3374 - Improve "loss" function for discovery search
• #3255 - Add config property to set default shard transfer method
• #3380 - Add ability to selectively disable the range or lookup index for integer payloads to save memory
• #3555 - Report progress of ongoing shard transfers
• #3311 - Also report collection recovery progress in non-interactive terminals
• #3235 - Set custom thread priorities on Linux for better scheduling
• #3257, #3313 - Avoid logging of health and metrics APIs in access log
• #3418 - Report the timestamp of last seen error in cluster, helpful for debugging
• #3419 - Expose the git commit hash of the build at the root endpoint
• #3498 - Allow loading of segments to be stopped in the middle if cancelled
• #3552 - Expose request timing histogram for Prometheus in metrics endpoint
• #3581 - Report version in logs on startup

Bug fixes 🔂

• #3386 - Fix multiple vector updates on same point in batch not working
• #3512 - Expose discover endpoint statistics in the metrics
• #3544 - Fix some REST endpoints not being usable with read-only API key
• #3592 - Fix internal proxy always forwarding successful requests three times
• #3423 - Return better suited status codes in the gRPC API
• #3387 - Disable large limit optimization for plain indices
• #3557, #3565 - Properly configure RocksDB options on all column families, reducing memory usage
• #3580 - Fix reading vector data out of bounds
• #3608 - Fix a potential zero division error in geo filter queries
• #3623 - Security patch for snapshot upload file name

Dashboard 🍱

• qdrant/qdrant-web-ui#153 - Add snapshots for demo collections
• qdrant/qdrant-web-ui#156 - Auto-complete required fields when inserting commands
• qdrant/qdrant-web-ui#155 - Add button to delete points
• qdrant/qdrant-web-ui#134 - Visualize discovery scores
• qdrant/qdrant-web-ui#160 - Fix vector visualisation with optional named vectors
• Other changes

Thanks to @AlexandruCihodaru @Florents-Tselai @Shylock-Hg @amietn @ding-young @Okabe-Rintarou-0 @JojiiOfficial @mrbusche @gabrieldemian @2k4sm @Anush008 @TuanaCelik @kwkr @neo773 @Dhanus3133 @KShivendu @joein @xzfc @ffuugoo @IvanPleshkov @agourlay @coszio @generall @timvisee for their contributions!