* rollback: iterator of errors into error of iterator
* [AI] ordering iterator
* [AI + manual] simplify geo-index iterators
* [AI] extra test for consistency
* fmt
* [AI] extend test for deleting points + fix the problem
* Update lib/segment/src/index/field_index/geo_index/mmap_geo_index.rs
Co-authored-by: Tim Visée <tim+github@visee.me>
---------
Co-authored-by: Tim Visée <tim+github@visee.me>
* derive Pod for Point<T>
* remove clones on Copy type
* fix annoying clippy lint
* serde skip padding
* extract point into a new file
* assert aligment
* Make subgraph_connectivity deterministic
* Fix missing rng argument in test for subgraph_connectivity
Made-with: Cursor
---------
Co-authored-by: Cursor Agent <agent@cursor.com>
Replace `Vec<(GeoHash, AHashSet<PointOffsetType>)>` with three flat
parallel arrays: `Vec<GeoHash>`, `Vec<u32>` (offsets), and
`Vec<PointOffsetType>` (IDs).
Each AHashSet carried ~64 bytes of stack overhead (32-byte RandomState +
32-byte RawTable metadata) plus hash-table heap allocations with control
bytes — totalling ~124 bytes per entry even with a single point. The
flat layout uses 8 + 4 + 4 = 16 bytes per entry, a ~7.75x reduction.
Deletions use a sentinel value (u32::MAX) in the IDs array, filtered
during iteration. On-disk format is unchanged.
Made-with: Cursor
Co-authored-by: Cursor Agent <agent@cursor.com>
* use pageout to clear mmap cache
* Also clear cache of deleted flags in mmap dense vector storage
* Add reference to madvise man pages for probe logic
* use deconstruct
---------
Co-authored-by: timvisee <tim@visee.me>
* Add mincore-based memory stats to MmapFile
Add `resident_bytes()`, `disk_bytes()`, and `probe_memory_stats()` methods
to `MmapFile` for measuring page cache residency via `mincore(2)`. This is
the foundation for per-collection memory usage reporting.
Also extract `page_size()` as a public function in `mmap::advice`, replacing
the internal `PAGE_SIZE_MASK` with a direct page size cache.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* [AI] introduce trait for reporting memory usage per component
* [AI] memory reporter implementation for vector storage
* [AI] implement MemoryReporter for QuantizedVectors
* [AI] implement MemoryReporter for VectorIndexEnum
* Implement MemoryReporter for IdTrackerEnum with RAM estimation
Add ram_usage_bytes() to all ID tracker types and their data structures:
- PointMappings, CompressedPointMappings, CompressedVersions,
CompressedInternalToExternal, CompressedExternalToInternal
- MutableIdTracker, ImmutableIdTracker, InMemoryIdTracker
All ID trackers load their data into RAM (none use mmap for working data).
Files are reported as OnDisk (persistence only), actual RAM footprint
is reported via extra_ram_bytes. Uses struct destructuring to ensure
new fields trigger compile errors if not accounted for.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* [AI] implement MemoryReporter for PayloadStorageEnum and adjust FileStorageIntent
* [AI] implement MemoryReporter for PayloadStorageEnum and adjust FileStorageIntent
* [AI] implement MemoryReporter for payload indexes: in-ram structures memory consumtion computation + caching
* [AI] implement MemoryReporter for payload indexes: in-ram structures memory consumtion computation + caching
* [AI] segment-level memory usage report
* [AI] Block 3: Aggregation Layer and Data Model + internal api for remote shard
* [AI] REST API handler
* fmt
* [AI] clippy fixes
* [AI] macos fix + proxy segment fix
* [AI] make text index estimation a bit more correct
* fix is_on_disk reporting for dense_vector_storage
* fix after rebase
* [AI] deep account for quantized vectors RAM usage + unify chunk size + shring volatile storage after load
* remove debug log
* cache in test
* make manual test easier to run
* rollback chunk size diff, but keep it for test only
* review fixes
* Use exhaustive match
* Use div_ceil on bits everywhere
It does not seem to be strictly necessary because the number of bits
should already be a multiple of the used container size bytes. Still
it's good practice to be careful with this calculation.
* Improve heap size bytes for encoded product quantization vectors
* Include vector stats for binary quantized vectors
* In volatile chunked vectors, include heap allocated vector
* Include rest of heap allocated structures for mutable map index
* In mutable geo index, the hash map is also heap allocated
* Update tests/manual/test_memory_reporting.py
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: timvisee <tim@visee.me>
Co-authored-by: Tim Visée <tim+github@visee.me>
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
* fix: propagate flush cancellation through payload index flushers
When a payload index is deleted, the gridstore backing it is dropped.
A concurrent background flush detects this (weak Arc refs fail to upgrade)
and returns GridstoreError::FlushCancelled. The From<GridstoreError> impl
correctly maps this to OperationError::Cancelled, and entry.rs gracefully
handles Cancelled by skipping the flush.
However, two intermediate layers unconditionally wrapped all errors as
ServiceError, destroying the Cancelled variant before it reached the
handler in entry.rs:
1. All 4 mutable index flushers (map, text, numeric, geo) used
OperationError::service_error(format!(...)) instead of converting
via Into<OperationError> first.
2. struct_payload_index flusher also unconditionally wrapped sub-flusher
errors as service_error.
This caused the benign FlushCancelled to surface as a fatal
"last background flush failed" ServiceError.
Fix both layers to propagate Cancelled errors as-is, matching the
pattern already used for vector storage, quantization, and id tracker
flushers in entry.rs.
Made-with: Cursor
* simplify error conversion
---------
Co-authored-by: Cursor Agent <agent@cursor.com>
Co-authored-by: generall <andrey@vasnetsov.com>
* fix(segment): reserve full last chunk when capacity aligns to chunks
When the requested vector count is an exact multiple of chunk_capacity,
the remainder modulo chunk_capacity is zero. The previous expression used
that remainder for the last chunk's flattened length, reserving zero
capacity. Reserve chunk_capacity * dim for a full last chunk instead.
* test(segment): regression for try_set_capacity_exact last chunk at chunk boundary
Assert flattened Vec capacity for the last chunk when the requested vector
count is a multiple of chunk_capacity. The buggy implementation reserved
zero for that chunk; this test fails without the fix and passes with it.
* cargo fmt edited file
* Add empty placeholder vector storage types for named vector CRUD
Introduce EmptyDenseVectorStorage and EmptySparseVectorStorage as
placeholder storages for newly created named vectors on immutable
segments. These report all vectors as deleted, consume no disk space,
and are reconstructed from segment config on load via the new
VectorStorageType::Empty and SparseVectorStorageType::Empty variants.
Key design decisions:
- is_on_disk is derived from original user config, not hardcoded
- MultiVectorConfig is preserved for multi-vector support
- Config mismatch optimizer skips Empty storage to avoid false rebuilds
- Quantization delegates normally (handles 0 vectors gracefully)
- get_vector includes debug_assert to catch unexpected access
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* [AI] segment-level operations for creating and deleting anmed vectors
* [AI] implement named vector creation and deleting in proxy segment
* [AI] Step 3: Proxy Segment Handling for Named Vector Operations
* [AI] implement for Edge
* [AI] implement consensus operations for named vector operations
* [AI] refactor VectorNameConfig, remove VectorNameConfigInternal
* [AI] handle vector schema inconsistency in raft snapshot recovery
* [AI] rest + grpc API
* [AI] clippy
* [AI] generate openAPI schema
* fmt
* ci fixes
* [AI] fix jwt access test
* [AI] nop operation for awaiting of consensus-commited update ops
* [AI] move vector name operations into points service
* [AI] implement internal api for vector name operations
* [AI] change collection-level config along with segment level operation
* [AI] vector schema reconceliation instead of error
* fmt
* missing compile-time option
* [AI] integration test
* [AI] fix missing JWT tests
* [AI] remove NOP
* [AI] openapi test
* [AI] fix initialization of mutable segment
* [AI] more simple integration tests
* fmt
* [AI] make cluster test a bit harder
* [AI] make test less flacky
* [AI] rabbit comments
* [AI] check params compatibility before writing vector config
* [AI] make sure to register vector storages in structure payload index
* [AI] vector name validation
* lower vector length validation to 200 chars to account for prefix in filename
* [AI] proxy segment: prevent stale data leak through optimization
* fmt
* [AI] filter out removed vectors from proxy response
* [AI] handle vector name in proxy
* fmt
* adjust proxy info based on dropped vectors
* [AI] proxy segment: update filters to correct has_vector condition
* fmt
* clippy
* Fix consensus snapshot applicaiton for vector schema
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Return error instead of panicking for corrupted aliases file on startup
* common::fs::ops: provide file name in error messages
And drop FileStorageError in favor of std::io::Error, since we always
convert all kinds of errors into ServiceError anyway.
* TableOfContent:🆕 return errors instead of panics
Also, drop context strings. We use fs_err anyway, that should be enough.
---------
Co-authored-by: leohenon <77656081+lhenon999@users.noreply.github.com>
Co-authored-by: xzfc <xzfcpw@gmail.com>
* [ai] Replace manual into mappings with Into::into
* Reformat
* [ai] Use implicit .iter
* Don't iterate over keys too
* [ai] Replace unwrap_or
* Reformat
* [ai] Use as_deref and then_some
* [ai] Use more to_string
* [ai] Use explicitly typed into conversions
* Reformat
* [ai] More explicit into conversions
* Reformat
* measure filterable hnsw indexed vs unindexs + cpu vs gpu
* measure filterable hnsw indexed vs unindexs + cpu vs gpu
* limit GPU parallelism
* address review
* IoUringState: generic `RequestId`
* UniversalRead: generic `RequestId`
* Simplify `gridstore::Pages::get_page_value_ranges`
Now we don't need two separate `SmallVec`s as we can put `buffer_offset`
into `RequestId`.
* Better doc comment
* Rename `RequestId` -> `Meta`
* Propagate HardwareCounterCell through MmapMapIndex::get_values
Previously, `MmapMapIndex::get_values` used `ConditionedCounter::never()`,
which silently skipped all hardware IO counter tracking for mmap map index
value reads. This propagates a real `HardwareCounterCell` through the full
call chain so that disk IO from `values_iter` is properly measured.
Changes:
- `MmapMapIndex::get_values` now accepts `&HardwareCounterCell`, creates a
`ConditionedCounter` via `make_conditioned_counter`, and measures the
`deleted` bitvec access (matching `check_values_any` behavior).
- `MapIndex::get_values` forwards the counter to the `Mmap` variant.
- `FacetIndex::get_point_values` trait method now accepts `&HardwareCounterCell`,
propagated through `FacetIndexEnum` and both impls (MapIndex, BoolIndex).
- `indexed_variable_retriever` accepts and forwards the counter to
`IntMapIndex`, `KeywordIndex`, and `UuidMapIndex` calls.
- `SegmentBuilder::update` and `_get_ordering_value` accept and propagate
the counter instead of creating disposable instances.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: add test
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: Daniel Boros <dancixx@gmail.com>
* Skip broken tests
* Add rocksdb dropper
This is a temporary tool. It will be removed later in this PR.
* [automated] Drop rocksdb
This commit is made by running tools/rocksdb/drop.sh
* Touch-up after ast-grep
The previous automated commit removed items, but not their comments.
Also, some blocks are left with only one item.
Also, ast-grep can't handle macros like `vec![]`.
This commit completes the job.
* Remove RocksDB dropper
* Remove mentions of rocksdb feature in CI
* Fix clippy warnings
These `FIXME` comments added previously in this PR by ast-grep by
"peeling" cfg_attr like this:
-#[cfg_attr(not(feature = "rocksdb"), expect(...))]
+#[expect(...)] // FIXME(rocksdb): ...
This commit removes these allow/expect attributes and fixes clippy
lints.
* Remove leftover rocksdb-related code
Removed:
- Cargo.toml: rocksdb cargo feature flag and dependencies.
- code: rocksdb-related items that was not under feature flag.
- flags.rs: rocksdb-related qdrant feature flags.
Disabled:
- Some benchmarks because these do not compile now.
* Print warning if rocksdb leftovers found in snapshots
* Remove Clone from tokenizer
* Regenerate openapi.json
* feat: add search_max_batchsize to strict mode config
* added test case for search_max_batchsize
* Changes for fixing CI issue dure openapi
* Modify check_strict_mode_batch
---------
Co-authored-by: Arnaud Gourlay <arnaud.gourlay@gmail.com>
* Remove UniversalRead::is_empty
Reason: Less methods to override in wrappers. And I don't think this
method makes much sense.
* Move common::universal_io::{ => wrappers}::read_only
* Update ReadOnly wrapper
Implement missing methods, and follow the code style of the upcoming
TypedStrorage wrapper.
* Add TypedStorage
* Use TypedStorage
* Add reminder comments
* Clarify TypedStorage use-case
Two bugs introduced in 9a14ed7:
1. `HashSet::insert` returns `true` when the value is *new*, but the
code assigned it to `already_seen` and skipped on `true` — inverting
the deduplication so every unique geohash was skipped during removal.
2. `filter()` returns `OperationResult<Option<Box<dyn Iterator>>>`;
the new tests called `.unwrap()` once (unwrapping the Result) but
missed the second `.unwrap()` for the Option, causing a compile error.
Made-with: Cursor
Co-authored-by: Cursor Agent <agent@cursor.com>
* Add reproducing test for spurious geo index warning on duplicate geo values
When a point has multiple geo values that produce the same max-precision
geohash (e.g. duplicate coordinates in a multi-value geo field),
`InMemoryGeoMapIndex::remove_point` logs a spurious warning:
"Geo index error: no points for hash X was found".
The root cause is an asymmetry between `point_to_values` (stores all
values including duplicates) and `points_map` (uses a HashSet, so
duplicates are collapsed). During removal the loop processes each value
individually — the first iteration removes the `points_map` entry, and
subsequent iterations for the same hash can't find it.
This commit adds:
- A `debug_assert!` in `remove_point` (matching the existing pattern in
`decrement_hash_value_counts`) to make the issue detectable in tests.
- `test_remove_point_with_duplicate_geo_values` — reproduces the bug by
adding a point with `[BERLIN, BERLIN]` then removing it.
- `test_frequent_add_remove_geo_points` — exercises repeated add/remove
cycles to cover the user-reported scenario.
Made-with: Cursor
* fix: redulplicate hash IDs, replace warn with debug assertion
* fmt
---------
Co-authored-by: Cursor Agent <agent@cursor.com>
Co-authored-by: generall <andrey@vasnetsov.com>
* Don't insert deferred points into sparse index
# Conflicts:
# lib/segment/src/segment/tests.rs
# lib/segment/src/segment_constructor/segment_constructor_base.rs
# Conflicts:
# lib/segment/src/segment_constructor/segment_constructor_base.rs
* Clippy
* Assert consistency of deferred_internal_id var
* Return before SparseVector conversion in case of deferred point
* Use debug_assert instead
* Refactor: `SearchSegmentEntry` for search ops
* Move `has_deferrend_points_method`
to `SearchSegmentEntry`
* Reorg imports
* Renamings per review
* fixup! Renamings per review
* `StorageSegmentEntry` trait
It contains methods dealing with storage and syncronization
* Move index methods from `ReadSegmentEntry`
to `SegmentEntry`
* Add `_concurrent` suffix to some methods
* move field index methods into NonAppendableSegmentEntry
---------
Co-authored-by: generall <andrey@vasnetsov.com>
