mirror of
https://github.com/qdrant/qdrant.git
synced 2026-08-02 08:00:55 -05:00
e90a03e00b
* test: test must_not is_null * vcs: ignore vscode files * feat: group-by initial implementation * cargo fmt * refactor: same request behavior on reco and search * refactor: get rid of RefCell * refactor-fix: correct hashmap keys, and early stops * chore: small improvements * feat: groups aggregator * fix: pull changes from other files * cargo fix * cargo fmt * docs: edit docstrings * allow dead code (while the complete feature is beint built) * chore: restructure * feat: introduce GroupKey, minor other improvements * cargo fmt * chore: specify aggregator visibility * fix: oops, leaking "private" type * refactor-fix: restructure and refactor group_by * cargo fix * fix: don't panic when there is no group-by field * remove print statements * amend: `>=` -> `==` * perf: remove double clone * chore: sync aggregator from other branch * chore: cleanup print statemets * test: ignore big tests * cargo fmt * refactor: add early stop when the groups have been filled, improve code * chore: sync aggregator, remove print from test * refactor: consider shard_selection, improve collection_by_name handling * feat: add bucketing to table of content * refactor: better errors, improve tests * test: add integration tests * feat: add endpoints * refactor: introduce ScoredPoint wrapper, restructure types * sync aggregator * edit internal grouping visibility * feat: group_by internals * cargo fmt * cargo fmt * refactor: turn inner fn into closure * test: fix test to support new vector output representation * feat: wire up grouping with actix * expose grouped_by field * fix: change output group format * feat: wire up openapi * fix: finish wiring up grouping in actix * tests: fix test_group.py * cargo fmt * refactor: extract constants * remove Hash from ScoredPoint * `Option<collection_by_name>` -> `collection_by_name` * fix: handle better cases on `match_on` * fix: consider that subsequent calls can bring better results * cargo fmt * fix clippy warnings * cargo fmt * refactor: move `Group` to `types`, localize `hydrate_from`, remove `Deref` impls * refactor `add_points` * refactor: turn `GroupKey` into enum * refactor: make `HashablePoint` inner struct private * feat: add grpc layer, make new `PointGroup` type to use as output * fix: update openapi models * docs: update grpc docs * fix merge errors * refactor: add BaseGroupRequest to make code DRYer, improve doc comments * cargo fmt * perf: increase precision; choose best groups by score * misc: add more integration tests, fix review comments * cargo fmt * fix: reimplement interface to flatten search and recommend requests, excluding offset * cargo fmt * refactor: move `r#do` impl to `GroupRequest` * fix: update grpc docs * perf: sort in reverse order * fix: use fist value of a Value::Array * fix: validate group_by to not support bracket notation, fix int. tests * fix: update grpc validation * tests: update collection_tests * refactor: move validation to the api layers * Oops: reupdate tests * refactor: let the derives derive (thanks @ffuugoo) * refactor: use a new GroupId on the output also increases performance by copying less * remove hashable set, take ordering into an account, fix mutliple groups values support * fmt * refactor group_id + rename per_group -> group_size, fix clippy * remove GroupKey wrapper * @agourlay review fixes * refactor: `group_min_scores` and `group_max_scores` -> `group_best_scores` * refactor: use set difference on `keys_of_unfilled_best_groups` * refactor: use set intersection on `len_of_filled_best_groups` * refactor: turn best_group_keys into iterator * fix: remove [] syntax limitation * fix: update openapi.json --------- Co-authored-by: Andrey Vasnetsov <andrey@vasnetsov.com>