{"full_name":"qdrant/qdrant","owner":"qdrant","name":"qdrant","url":"https://github.com/qdrant/qdrant","description":"Qdrant - High-performance, massive-scale Vector Database and Vector Search Engine for the next generation of AI. Also available in the cloud https://cloud.qdrant.io/","homepage":"https://qdrant.tech","topics":[],"language":"Rust","license":"Apache-2.0","avatar":"https://avatars.githubusercontent.com/u/73504361?v=4","created_at":"2020-05-30T21:37:01Z","pushed_at":"2026-11-10T09:39:09Z","metrics":{"stars":35497,"forks":2806,"watchers":163,"issues":2122,"prs":8792,"releases":119,"issues_open":null,"issues_closed":null,"prs_open":null,"prs_closed":null,"prs_merged":null,"discussions":331},"growth":{"day":{"stars":18,"forks":1,"base_day":"2026-11-09"}},"history":{"dates":["2026-11-09","2026-11-10"],"stars":[35479,35497],"forks":[2805,2806],"watchers":[163,163],"issues_open":[null,null],"prs_open":[null,null]},"boards":{},"same_owner":[],"issues":[{"number":3121,"title":"Timeout on collection deletion","comments":34,"created_at":"2023-12-01T11:07:00Z"},{"number":4655,"title":"Transport error on self hosted 3 node cluster","comments":24,"created_at":"2024-07-10T18:04:53Z"},{"number":3132,"title":"qdrant_client-1.6.9: \"Can't init WAL\" when creating simple collection","comments":23,"created_at":"2023-11-29T15:27:04Z"},{"number":6081,"title":"GPU problem","comments":21,"created_at":"2025-02-28T07:22:18Z"},{"number":3956,"title":"on_disk payload is not decreasing RAM on system ","comments":20,"created_at":"2024-04-02T06:19:26Z"},{"number":5774,"title":"Error when uploading large collection","comments":19,"created_at":"2025-01-10T20:29:15Z"},{"number":5327,"title":"Specify IDs for distance matrix API","comments":16,"created_at":"2024-10-28T21:36:18Z"},{"number":4410,"title":"Automatic Deletion of Segments in Qdrant with Large Volume of Data","comments":16,"created_at":"2024-06-07T05:43:15Z"}],"discussions":[{"number":11335,"title":"Proposal: a rehearsal tool for vector-index migrations","comments":0,"category":"Ideas","created_at":"2026-10-29T15:17:21Z"},{"number":11334,"title":"Qdrant on your own box, from the release binary, in one OpenTofu-style config","comments":0,"category":"Show and tell","created_at":"2026-10-29T14:09:25Z"},{"number":11270,"title":"⚡ Powered by FastEmbed: PolyVector brings In-Engine Hybrid Search to Postgres (3.2x faster than Elasticsearch)","comments":0,"category":"Show and tell","created_at":"2026-10-20T15:20:23Z"},{"number":11179,"title":"Showcase: ATS Zero-Browser Gateway — Fast protocol scraper & delta data feeder for vector search","comments":0,"category":"Show and tell","created_at":"2026-10-15T20:49:40Z"},{"number":9940,"title":"Metadata lookup tail latency: measurements on arithmetic field addressing vs B-tree/LSM","comments":3,"category":"Ideas","created_at":"2026-07-22T17:32:26Z"},{"number":9936,"title":"Kdrant — an idiomatic, coroutine-first Kotlin client for Qdrant","comments":0,"category":"Show and tell","created_at":"2026-07-21T21:56:13Z"},{"number":9327,"title":"Built Agent Magnet — episodic memory layer using Qdrant","comments":1,"category":"Show and tell","created_at":"2026-06-05T08:50:11Z"},{"number":9130,"title":"About adding a new n-gram tokenizer","comments":1,"category":"Ideas","created_at":"2026-05-22T02:40:51Z"}],"latest_release":{"name":"v1.20.0","tag":"v1.20.0","published_at":"2026-10-22T14:14:12Z","body":"# Changelog\r\n\r\n## Features 🎶\r\n\r\n- [#10750](https://github.com/qdrant/qdrant/issues/10750) - BM25 scoring directly on the full-text payload index [[docs](https://qdrant.tech/documentation/search/text-search/full-text-search/)]\r\n- [Milestone #57](https://github.com/qdrant/qdrant/milestone/57?closed=1) - TurboQuant: new `turbo8` and `turbo16` vector datatypes [[docs](https://qdrant.tech/documentation/manage-data/vectors/#turboquant-datatypes)]\r\n- [#11125](https://github.com/qdrant/qdrant/issues/11125) - Per-element payload indexes: new `nested` index parameter, faster nested condition filtering.  [[docs](https://qdrant.tech/documentation/manage-data/indexing/#per-element-index-for-arrays-of-objects)]\r\n- [#11180](https://github.com/qdrant/qdrant/pull/11180), [#11263](https://github.com/qdrant/qdrant/pull/11263) - New `code` tokenizer for full-text index [[docs](https://qdrant.tech/documentation/manage-data/indexing/#code)]\r\n- [#10466](https://github.com/qdrant/qdrant/pull/10466) - Configurable audit log output: `audit.output` writes audit events to `file` (default), `stdout` or `stderr`  [[docs](https://qdrant.tech/documentation/security/#audit-log-output)]\r\n- [#10992](https://github.com/qdrant/qdrant/pull/10992) - Collection creation time: `created_at` is recorded and returned in collection info [[docs](https://qdrant.tech/documentation/manage-data/collections/#collection-info)]\r\n- [#10711](https://github.com/qdrant/qdrant/pull/10711) - Substring filter condition: `match: { substring: \"...\" }` (full-scan only) [[docs](https://qdrant.tech/documentation/search/filtering/#substring-match)]\r\n- [#10231](https://github.com/qdrant/qdrant/pull/10231), [#10287](https://github.com/qdrant/qdrant/pull/10287), [#10296](https://github.com/qdrant/qdrant/pull/10296) - New formula query expressions: `acosh`, `max`, `min` [[docs](https://qdrant.tech/documentation/search/search-relevance/#max-and-min)] and [[docs](https://qdrant.tech/documentation/search/search-relevance/#hyperbolic-distance)]\r\n- [#10597](https://github.com/qdrant/qdrant/pull/10597), [#10598](https://github.com/qdrant/qdrant/pull/10598) - Configurable ID tracker memory placement [[docs](https://qdrant.tech/documentation/ops-configuration/memory-tiers/#id-tracker)]\r\n\r\n## Improvements 🎵\r\n\r\n- [#11009](https://github.com/qdrant/qdrant/pull/11009), [#11020](https://github.com/qdrant/qdrant/pull/11020) - Faster HNSW payload links build - builder sized to the block, not the segment, and reused\r\n- [#11083](https://github.com/qdrant/qdrant/pull/11083) - Faster HNSW healing - payload blocks reuse old graph links; `m = 0` graphs can heal too\r\n- [#11166](https://github.com/qdrant/qdrant/pull/11166) - Smaller HNSW upper layers - `floor()` in level assignment matches the HNSW paper (6.25% of points on level 1 at `m = 16`, not 25%)\r\n- [#10976](https://github.com/qdrant/qdrant/pull/10976) - Faster PQ training - k-means sample gathered in one pass, not one per chunk\r\n- [#10975](https://github.com/qdrant/qdrant/pull/10975) - Faster scalar quantization search - AVX512-VNNI and AVX-VNNI kernels for int8 dot product\r\n- [#10977](https://github.com/qdrant/qdrant/pull/10977) - Less search slowdown during optimizations - the whole optimization job runs at low thread priority\r\n- [#11078](https://github.com/qdrant/qdrant/pull/11078), [#11080](https://github.com/qdrant/qdrant/pull/11080) - Fewer cancelled optimizations - optimizers are recreated only when their config changes\r\n- [#11234](https://github.com/qdrant/qdrant/pull/11234) - Faster point retrieval - ids resolved in one batch per segment\r\n- [#11070](https://github.com/qdrant/qdrant/pull/11070) - Faster snapshot transfers and downloads - 4 MiB stream buffers instead of 4 KiB\r\n- [#11253](https://github.com/qdrant/qdrant/pull/11253) - Faster shard transfers - `transfer_raw_points` and `transfer_raw_payloads` on by default\r\n- [#11206](https://github.com/qdrant/qdrant/pull/11206) - Safer snapshots - unique snapshot names, and a stored snapshot is never ove"},"readme":true,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-10T09:35:25Z","history_start":"2026-09-06"}