Change log

Improvements 🎬

• https://github.com/qdrant/qdrant/pull/2157 - Use configured temporary directory for recovering snapshots
• https://github.com/qdrant/qdrant/pull/2224 - Log HTTP server port at startup
• https://github.com/qdrant/qdrant/pull/2221 - Improve performance of scroll internals

Bug fixes ⚙️

• https://github.com/qdrant/qdrant/pull/2203 - Fix vectors not being preprocessed when updating vectors individually
• https://github.com/qdrant/qdrant/pull/2222 - Fix potential livelock when optimizer is cancelled
• https://github.com/qdrant/qdrant/pull/2235 - Fix payload index not handling type change
• https://github.com/qdrant/qdrant/pull/2228 - Fix missing release binaries for last versions
• https://github.com/qdrant/qdrant/pull/2237 - Fix WAL not loading when there are unrecognized files
• https://github.com/qdrant/qdrant/pull/2249 - Fix concurrent shard transfers, limit to one per node per collection
• https://github.com/qdrant/qdrant/pull/2250 - Fix potential stack overflow in segment spinlock

Change log

Improvements 🚦

• https://github.com/qdrant/qdrant/pull/2137 - Support SIGTERM to shutdown Qdrant
• https://github.com/qdrant/qdrant/pull/2100 - Improve validation of search API requests
• https://github.com/qdrant/qdrant/pull/2156 - Use configured temporary directory for snapshot uploads

Bug fixes 🏎️

• https://github.com/qdrant/qdrant/pull/2165 - Fix potential deadlock in consensus
• https://github.com/qdrant/qdrant/pull/2171 - Fix potential deadlock due to double read
• https://github.com/qdrant/qdrant/pull/2161 - Fix potential data race when creating collections concurrently
• https://github.com/qdrant/qdrant/pull/2189 - Fix collection create error if files were left behind
• https://github.com/qdrant/qdrant/pull/2185 - Fix panic on startup if storage is not properly flushed

Change log

Features 💍

• https://github.com/qdrant/qdrant/pull/1981 https://github.com/qdrant/qdrant/pull/1996 https://github.com/qdrant/qdrant/pull/2011 - Lookup in grouping requests. Allows to keep shared group info in a dedicated collection to minimize memory usage.
• https://github.com/qdrant/qdrant/pull/2009 - Self-hosted Web UI. Just open http://localhost:6333/dashboard in your browser to see the dashboard. In the first version, it includes collection viewer and interactive query editor with autocomplete.
• https://github.com/qdrant/qdrant/pull/2024 - Experimental 🔬 - support for asynchronous IO with io_uring, a feature of recent Linux kernels. Significantly improves performance on network-mounted storages.
• https://github.com/qdrant/qdrant/pull/2070 - Oversampling allows to retrieve more points using quantized vectors and then re-score them using original ones. Allows to tune speed/accuracy tradeoff in the query time.

Improvements 📌

• https://github.com/qdrant/qdrant/pull/1974 - HNSW performance improvements
• https://github.com/qdrant/qdrant/pull/1868 - Speed up index loading by improving the WAL ack mechanism
• https://github.com/qdrant/qdrant/pull/1912 https://github.com/qdrant/qdrant/pull/1992 - Improve first cold query performance by pre-heating the cache
• https://github.com/qdrant/qdrant/pull/1990 - Improve validation of the collection names
• https://github.com/qdrant/qdrant/pull/2073 - Optimize performance of is_empty (and consequently group-by) condition by using payload index
• https://github.com/qdrant/qdrant/pull/2101 - Add recovery_mode flag to the telemetry
• https://github.com/qdrant/qdrant/pull/2109 - Validation for the snapshot location download
• https://github.com/qdrant/qdrant/pull/2122 - Check snapsot integrity before recovery
• https://github.com/qdrant/qdrant/pull/1960 - Configurable location for the temporary files
• https://github.com/qdrant/qdrant/pull/2115 - Stability improvements for Raft consensus

Bug fixes 🐲

• https://github.com/qdrant/qdrant/pull/2003 https://github.com/qdrant/qdrant/pull/2008 - Fix Channel sender dropped: channel closed error on duplicated consensus messages
• https://github.com/qdrant/qdrant/pull/2022 - Fix wal clamp in the snapshotting process
• https://github.com/qdrant/qdrant/pull/2021 - Fix validation of the max_indexing_threads parameter
• https://github.com/qdrant/qdrant/pull/2051 - Fix batch request with duplicated filters
• https://github.com/qdrant/qdrant/pull/1913 - Prevent rare case of missing payload index after the optimization
• https://github.com/qdrant/qdrant/pull/2053 - Fix mixed up write ordering parameters
• https://github.com/qdrant/qdrant/pull/2063 - Fix snapshot uploading for non-existing collection
• https://github.com/qdrant/qdrant/pull/2061 - Fix for Point with internal ID not found in id tracker error, which could appear after non-graceful shutdown
• https://github.com/qdrant/qdrant/pull/2072 - Improve counting vectors in collection info
• https://github.com/qdrant/qdrant/pull/1928 - Fix for GOAWAY/ENHANCE_YOUR_CALM error on high load in cluster deployments
• https://github.com/qdrant/qdrant/pull/2112 - Fix saving of deleted mmaped vectors in snapshots
• https://github.com/qdrant/qdrant/pull/2104 - Fix for collection init_from parameter, allowing to initialize collection with changed vector-specific parameters
• https://github.com/qdrant/qdrant/pull/2124 - Fix vector compatibility checks, prevents panic on vector size mismatch (related to on_disk vectors)
• https://github.com/qdrant/qdrant/pull/2136 - Fix storage integrity during copy-on-write operations

Thanks to @russcam, @woile, @kayagokalp, @llogiq, @timsueberkrueb, @sunisdown, @yusufani, @dcastelltort, @agourlay, @timvisee, @coszio, @ffuugoo, @trean for their contributions!

Make sure to check out the new Web UI query editor:

https://github.com/qdrant/qdrant/assets/1935623/b1e9351d-9728-4e38-bf0d-7e402d7ce803

@kartik-gupta-ij is working on it as part of his GSoC project. Stay tuned for more updates!

Change log

Bug fix

• https://github.com/qdrant/qdrant/commit/e2fb3d74cf848ecdce271cfb8705b354af9bb377 - fix the incorrect signal handler introduced in the previous bugfix

Change log

Bug fixes

• https://github.com/qdrant/qdrant/pull/1972 - fix error when loading config files, change order in which they're used
• https://github.com/qdrant/qdrant/pull/1975 - fix behavior of CTRL+C in Docker

Change log

Features 🍹

• https://github.com/qdrant/qdrant/pull/1724, https://github.com/qdrant/qdrant/pull/1757, https://github.com/qdrant/qdrant/pull/1835, https://github.com/qdrant/qdrant/pull/1816, https://github.com/qdrant/qdrant/pull/1836 - support for optional vectors.
  • This feature allows to create points with only some of the vectors defined. Allows more granular control over data.
  • Introduced two new API for managing vectors independently from payload. [Docs]
• https://github.com/qdrant/qdrant/pull/1745 - Built-in API key authentication support. [Docs]
• https://github.com/qdrant/qdrant/pull/1755, https://github.com/qdrant/qdrant/pull/1777 - Recovery mode for better handling Out-of-Disk and Out-of-Memory errors. [Docs]
• https://github.com/qdrant/qdrant/pull/1774 - Cluster management API in gRPC.
• https://github.com/qdrant/qdrant/pull/1808 - Add support for TLS certificate rotation
• https://github.com/qdrant/qdrant/pull/1838 - Dynamic mmap vector storage. Allows to store vectors in mmap files right from the start, avoiding optimization step. [Docs]
• https://github.com/qdrant/qdrant/pull/1602 - Nested object filter. Allows to filter by individual objects inside payload array. [Docs]
• https://github.com/qdrant/qdrant/pull/1615 - Enables Product Quantization (PQ) for vectors. [Docs]
• https://github.com/qdrant/qdrant/pull/1768 - Add Group-By API, allowing to group results by payload field. Useful for retrieving top results per group. [Docs]

Improvements 🏄

• https://github.com/qdrant/qdrant/pull/1677 - Memory optimization for full-text index
• https://github.com/qdrant/qdrant/pull/1709 - Increase consensus ping timeout to prevent false positive channel drops
• https://github.com/qdrant/qdrant/pull/1706 - Remove redundant telemetry values
• https://github.com/qdrant/qdrant/pull/1696 - Improve error logging for gRPC endpoints
• https://github.com/qdrant/qdrant/pull/1702, https://github.com/qdrant/qdrant/pull/1824 - Improve semantic for indexing_threshold in collection config
• https://github.com/qdrant/qdrant/pull/1767 - Release binaries for different platforms along with docker images
• https://github.com/qdrant/qdrant/pull/1776 - Embed default config file at compile time. Allows to run qdrant without an explicit config file
• https://github.com/qdrant/qdrant/pull/1813 - Limit random ids generator to 53 bits to accommodate JS users
• https://github.com/qdrant/qdrant/pull/1881 - Improve API validation
• https://github.com/qdrant/qdrant/pull/1926 - Remove unused keep_alive_while_idle

Bug fixes 🐝

• https://github.com/qdrant/qdrant/pull/1710 - Fix minor bug in hnsw params validation
• https://github.com/qdrant/qdrant/pull/1775 - Fix payload index restoration in snapshots
• https://github.com/qdrant/qdrant/pull/1784 - Allow to address individual values in payload arrays
• https://github.com/qdrant/qdrant/pull/1857 - Prevent Removing segment which is still in use error
• https://github.com/qdrant/qdrant/pull/1906 - Remove max message decoding limit on internal gRPC client
• https://github.com/qdrant/qdrant/pull/1936 - Fix saving of the payload index in case of external kill

Thanks to @jwpjrdev, @paulotten, @ibrahim-akrab, @Shubham8287, @nilhex, @DengYiping, @eltociear, @Jesse-Bakker, @IvanPleshkov, @agourlay, @timvisee, @coszio for their contributions!

Change log

Bug fixes

• https://github.com/qdrant/qdrant/pull/1847 - restrict quantization to immutable segments only

Change log

Bug Fixes

• https://github.com/qdrant/qdrant/pull/1798, https://github.com/qdrant/qdrant/pull/1797 - fix validation of m param in HNSW config. Allow m=0

Change log

Features 💫

• https://github.com/qdrant/qdrant/pull/1675, https://github.com/qdrant/qdrant/pull/1680 - Add vector specific HNSW and Quantization configuration. Now each vector field can have independent configuration for HNSW and Quantization. This allows to have different vector fields with different configuration.
• https://github.com/qdrant/qdrant/pull/1668, https://github.com/qdrant/qdrant/pull/1676 - Add last-used and startup timing information into telemetry API.
• https://github.com/qdrant/qdrant/pull/1593, https://github.com/qdrant/qdrant/pull/1581, https://github.com/qdrant/qdrant/pull/1581, https://github.com/qdrant/qdrant/pull/1663, https://github.com/qdrant/qdrant/pull/1642 - Add TLS support for gRPC and REST API. As well as TLS for internal communication between nodes. Enables to use Qdrant in more secure manner. Enables verification on both client and server side.
• https://github.com/qdrant/qdrant/pull/1578 - Ability to upload and recover snapshot files directly using multipart request.
• https://github.com/qdrant/qdrant/pull/1627 - Display aggregated vector count in /metrics endpoint.
• https://github.com/qdrant/qdrant/pull/1617, https://github.com/qdrant/qdrant/pull/1598, https://github.com/qdrant/qdrant/pull/1616 - New isNull condition for payload filtering. It allows to filter vectors with null values in specific fields. Allows to distinguish between empty and null values.

Improvements 🧠

• https://github.com/qdrant/qdrant/pull/1634, https://github.com/qdrant/qdrant/pull/1463 - Add parameter validation to REST and gPRC API, as well as to config file. This allows prevent misconfiguration and provide better error messages.
• https://github.com/qdrant/qdrant/pull/1571 - Add wait parameter to snapshot API. This allows skip waiting for snapshot creation and return immediately. Useful for long running operations.
• https://github.com/qdrant/qdrant/pull/1682 - Refactor transport channel pool and introduce internal rate limiter. This change makes cluster more stable under high concurrency load.
• https://github.com/qdrant/qdrant/pull/1682 - prevent snapshot process from blocking updates in Listener mode.

Bug fixes 🐛 🔫

• https://github.com/qdrant/qdrant/pull/1595 - fix the optimizer crash with Removing segment which is still in use problem, which was possible under high load.
• https://github.com/qdrant/qdrant/pull/1608 - fix runtime lockup on leader re-election.
• https://github.com/qdrant/qdrant/pull/1657 - Restore is_empty filtering condition to original behavior, which was compromised in last release. Payload is considered empty if: 1) it is null, 2) it is an empty array, 3) the key is not present in the payload.
• https://github.com/qdrant/qdrant/pull/1638 - fix error handling during shard transfer. Not failed updates are reported properly to the consensus.
• https://github.com/qdrant/qdrant/pull/1672, https://github.com/qdrant/qdrant/pull/1674 - fix slice/api.rs at line 2594: index out of bounds panic.
• https://github.com/qdrant/qdrant/pull/1671 - fix possible double read lock in case of updates with enforces ordering.

New local mode in Qdrant Python SDK

Simply install

pip install qdrant-client==1.1.4

and start using qdrant in local mode, no docker required

from qdrant_client import QdrantClient
client = QdrantClient(":memory:")

Thanks to @coszio, @timvisee, @DengYiping, @IvanPleshkov, @agourlay, @ffuugoo, @Weijun-H, @ibrahim-akrab, @timonv, @tusharxoxoxo, @stencillogic, @eltociear, @nisiyong for their contributions!

Change log

Features ⚛️

• https://github.com/qdrant/qdrant/pull/1311, https://github.com/qdrant/qdrant/pull/1538 - Scalar Quantization is now available in Qdrant! Compress float32 -> int8 to gain up to a 4x decrease in memory consumption and up to a 2x speed boost with minimum loss of accuracy. See more in quantization docs
• https://github.com/qdrant/qdrant/pull/1466 - Add Match Any filtering condition allows to specify a set of values to check much more simpler. Docs
• https://github.com/qdrant/qdrant/pull/1482 - Experimental support for listener mode. Useful for dedicated backup machines and cross-regional backups. Docs
• https://github.com/qdrant/qdrant/pull/1541 - basic implementation of /metrics API - serves telemetry in OpenMetric format. Suitable for automatic metrics collection with tools like Prometheus.

Improvements 🧗

• https://github.com/qdrant/qdrant/pull/1451 - the ability to checkpoint Raft consensus. Allows to optimize operations for long-running clusters. Docs
• https://github.com/qdrant/qdrant/pull/1465, https://github.com/qdrant/qdrant/pull/1487 - allow making filtering conditions using nested data structures. Docs
• https://github.com/qdrant/qdrant/pull/1534 - Improved functionality of snapshot recovery API:
  • Allows to recover snapshots taken in distributed mode on local deployment.
  • Allows recovering snapshots even if the collection does not exist
  • Allows to use wait=false during recovery. API will also tolerate disconnections now
• https://github.com/qdrant/qdrant/pull/1531 - Fix docker build on arm64 machines

Bug fixes 🦗

• https://github.com/qdrant/qdrant/pull/1462 - Fix snapshot recovery on NFS
• https://github.com/qdrant/qdrant/pull/1502 - Fix condition check for count API
• https://github.com/qdrant/qdrant/pull/1549 - Implement missing validation for creating collection aliases
• https://github.com/qdrant/qdrant/pull/1562 - Fix numerical payload index crash with large 64-bit integers
• https://github.com/qdrant/qdrant/pull/1563, https://github.com/qdrant/qdrant/pull/1566 - Fix for geo-payload index, ensures correct loading of the index statistics after a restart. Prevents possible corruption of geo-index storage

Special thanks to @IvanPleshkov, @agourlay, @ffuugoo, @timvisee, @coszio for contributing into this release

Changelog

This is another intermediate bugfix release, stay tuned!

• https://github.com/qdrant/qdrant/pull/1522 - fixes unwanted extra connections problem. Fixes performance degradation related to usage of replication factor > 1
• https://github.com/qdrant/qdrant/pull/1453 - re-implementation of better error message while handling malformed point ids.

Changelog

This an intermediate bugfix release:

• https://github.com/qdrant/qdrant/pull/1485 - fixes automatic threads number selection, prevents blocking of consensus execution thread in low-cpu environment. Also adjusts automatic search threads selection to improve CPU utilization

Next big release with is coming soon, stay tuned!

Changelog

• UUID WAL recovery hotfix!

Please prefer this version over v1.0.0

Hi! It's our v1.0 release! 🎉

• We stabilized the core API, filled in missing pieces, and prepared an extendable platform for upcoming features (yes, we have a new roadmap!)
• Mastered upgrades and migrations. Now it's easy to upgrade Qdrant to the latest version
• Plan each new release with compatibility to the previous one, so you can upgrade Qdrant zero downtime
• Delivered full horizontal scalability for both size and speed
• Achieved best-on-market performance, see benchmarks
• Launched Qdrant Cloud (beta), a fully managed vector search as a service. Free tier available!

Change log

Features 🏂

• https://github.com/qdrant/qdrant/pull/1363 - new API to list collection aliases
• https://github.com/qdrant/qdrant/pull/1377 - new API to delete snapshots
• https://github.com/qdrant/qdrant/pull/1364 - new API to initialize a collection from another collection. Useful for quick experiments over different indexing parameters. In combination with aliases, it allows to seamlessly scale number of shards in the cluster.
• https://github.com/qdrant/qdrant/pull/1381 - read operations have new opt-in consistencies guarantees, ensures consistent reads even from inconsistent cluster
• https://github.com/qdrant/qdrant/pull/1405 - write operations have new opt-in ordering guarantees, ensures that the writes are ordered from parallel overlapping requests

Improvements 🥐

• https://github.com/qdrant/qdrant/pull/1432 - Windows support (Windows is a group of several proprietary graphical operating system families developed and marketed by Microsoft.)
• https://github.com/qdrant/qdrant/pull/1352 - run consistency checks only on service startup
• https://github.com/qdrant/qdrant/pull/1347 - quick stop the node startup if the node cannot recover in a distributed setup
• https://github.com/qdrant/qdrant/pull/1353 - speed up snapshot creation, prevent blocking the execution thread
• https://github.com/qdrant/qdrant/pull/1391 - recover dead shards pro-actively
• https://github.com/qdrant/qdrant/pull/1415 - enable on disk payload option by default
• https://github.com/qdrant/qdrant/pull/1416 - improve error reporting for malformed point ids
• https://github.com/qdrant/qdrant/pull/1425 - allow to overwrite max number of cpus with env variable, useful for docker containers and kubernetes
• https://github.com/qdrant/qdrant/pull/1398 - decrease number of Tokio runtimes, improve performance
• https://github.com/qdrant/qdrant/pull/1431 - use performance optimized linker setup, up to 20% performance improvement on some workloads

Bug fixes 🎳

• https://github.com/qdrant/qdrant/pull/1365 - do not abort a succesful shard transfer

Telemetry 📡

• https://github.com/qdrant/qdrant/pull/1401 - Qdrant now collects anonymized telemetry data. It is enabled by default, but is simply disabled with CLI or ENV variable. The data is collected anonymously and is used to improve Qdrant. You can read more about it in the docs.

Future plans 🔮

This release concludes our 2022 roadmap. We covered even more features than we planned, and we are very happy with the results.
Now we have already started working on the 2023 roadmap, aiming to make vector search even faster, billion-scale deployments affordable and extend vector similarity beyond just search.

Thanks for all our contributors, users and supporters! ❤️

Special thanks to people who contributed to this release:

@agourlay, @ffuugoo, @joein, @coszio, @csko

Change log

Improvements 🤺

• https://github.com/qdrant/qdrant/pull/1312 - improved mmap configuration. Boosts search speed 2-6x on low-RAM setup.
• https://github.com/qdrant/qdrant/pull/1314 - better cleaning of marked-as-removed records
• https://github.com/qdrant/qdrant/pull/1338 - retry on attaching a new node to the consensus. Makes cluster more stable in case of consensus re-election cycle

Bug fixes 🪳

• https://github.com/qdrant/qdrant/pull/1329 - fix consistency issue in WAL, which in some cases resulted in hs.commit XX is out of range errors at startup
• https://github.com/qdrant/qdrant/pull/1334 - fix collection snapshots during the high write load

Thanks @agourlay, @ffuugoo and @hozan23 for contributing into this release!

Last release of 2022 🎅

Change log

Features 🎀

• https://github.com/qdrant/qdrant/pull/1283 - implement an ability to lookup vectors in different collection in recommendation API. Might be useful for item-to-user recommendations, where user and item embeddings are stored in a different collection with different index configuration.

Improvements 🪡

• https://github.com/qdrant/qdrant/pull/1280 - optimize shard live-cycle operations, stop optimizer faster on deletion and remove files in async task, so the API response faster, preventing timeouts even on network-mounted disks.
• https://github.com/qdrant/qdrant/pull/1286 - Fix a list of problems, related to consensus operations, including:
  • Properly handle sync transfer cancellation
  • Replicas now have a local state, which is periodically synchronized with consensus. It allows to notify consensus of failed operations and cancelled transfers even if original notification was dropped due to leader failure.
  • Improved validation on transfer initialization - prevent creating a transfer which involved a replica, which is already being transferred
  • Fixed bug related to replica inconsistency caused by updates during transfer finish
  • Improve shards transfer workflow - prevent re-try in case if transfer was cancelled externally

Bug fixes 🦖

• https://github.com/qdrant/qdrant/pull/1281 - fix error code propagation in remote shards
• https://github.com/qdrant/qdrant/pull/1302 - fix id mapper inconsistency in case of interrupted flush routine

Change log

Features 🚀

• https://github.com/qdrant/qdrant/pull/1245 - The new API for payload manipulation allows you to overwrite all existing payload fields in one step. Unlike set_payload, which replaces only mentioned fields, this new API overwrites all existing fields.
• https://github.com/qdrant/qdrant/pull/1249 - Allow to set and delete payload fields by filter. It is now possible to specify a subset of points to update using filtering conditions. Might be useful if you are using secondary key in the payload.

Improvements 🛠️

• https://github.com/qdrant/qdrant/pull/1262 - add parameter that defines a source of truth for snapshot recovery. Simplifies different scenarios of snapshot recovering in distributed mode. Including the recovery on the empty cluster.
• https://github.com/qdrant/qdrant/pull/1257 - allows to disable global HNSW index in the collection and only build HNSW sub-graphs based on payload values. Useful for organizing a collection which consists of sub-groups of different sizes, so that only the search within each group is required.
• https://github.com/qdrant/qdrant/pull/1260 - Add the peer ID into the auto-generated snapshot name to allow for the distinction of snapshots from different nodes.

Bug Fixes 🐛

• https://github.com/qdrant/qdrant/pull/1252 - Removes the blocking code in RaftService that impacts the threads of the internal gRPC Tokio runtime. As a result consensus-related operations are now have significantly less timeouts under high load.
• https://github.com/qdrant/qdrant/pull/1255 - fix warning message during batch update, if the batch have non-unique points

Change log

Features ⛲

• https://github.com/qdrant/qdrant/pull/1211 - mmap support for HNSW graph storage. Now it is not required to keep HNSW index in RAM. Especially relevant to collections with many small vectors.
• https://github.com/qdrant/qdrant/pull/1186 - Enable compression in gRPC protocol, useful for transferring large payloads over the slow network.

Improvements ⛳

• https://github.com/qdrant/qdrant/pull/1199 - Probabilistic search sampling. Significantly improves the search speed in case of large limit values. Allows to query large amount of nearest vectors fast.
• https://github.com/qdrant/qdrant/pull/1171 - Use compact CPU cache-friendly storage of HNSW graph links. Improves search speed, memory usage and application startup time.
• https://github.com/qdrant/qdrant/pull/1221 - improve response speed of cluster-info API

Bug Fixes 🐊

• https://github.com/qdrant/qdrant/pull/1241 - fixes possible deadlocking of consensus thread
• https://github.com/qdrant/qdrant/pull/1238 - fixes incorrect shard activation during snapshot recovery in distributed mode
• https://github.com/qdrant/qdrant/pull/1236 - fixes possible panic in proxy segment
• https://github.com/qdrant/qdrant/pull/1223 - avoid service shutdown in case of removing failed collection
• https://github.com/qdrant/qdrant/pull/1226 - prevent execution thread blocking on snapshot creation

Change log

Improvements

• https://github.com/qdrant/qdrant/pull/1214 - API for recovering snapshots, supports recovering in distributed mode

Bug fixes

• https://github.com/qdrant/qdrant/pull/1212 - fixed persistence of collection parameters changing

Change log

Bug fixes

• https://github.com/qdrant/qdrant/pull/1208 - fix error on collection replication factor and write consistency factor update
• https://github.com/qdrant/qdrant/pull/1198 - handle zero-length vectors with cosine distance - skip normalization and store vector as-is, in this case if will have 0 score with all other vectors

Improvements

• https://github.com/qdrant/qdrant/pull/1176 - Use less ram for id tracker, especially useful for collections with many small vectors
• https://github.com/qdrant/qdrant/pull/1184 - enable HTTP compression middleware
• https://github.com/qdrant/qdrant/pull/1200 - ensures read-after-write consistency in case of replica failure and if requests are going to the same peer.