kartsan03
fe1fbbc933
Switch SQLite ANN connections when load() is given a new path
2026-10-10 11:11:24 +00:00
Ni0ri
20c9b8e0be
Add Zvec Sparse ANN ( #1307 )
2026-10-01 09:53:38 -04:00
Pablo Miguel Argudo
c78fbd976c
Support append and delete in the RaBitQ ivf mode ( #1304 )
2026-09-30 12:44:19 -04:00
kartsan03
6765330f4a
Delete the requested rows when ANN ids are a NumPy array
2026-09-30 15:26:24 +00:00
davidmezzetti
d467878064
Refactor ANN tests, closes #1305
2026-09-30 11:04:07 -04:00
LimbC-C
81f908874b
Open lazy SQLite connections before saving indexes ( #1290 )
2026-09-30 09:30:14 -04:00
David Mezzetti
7df50abb61
Merge pull request #1282 from LimbC-C/fix/sqlite-binary-scores
...
Normalize SQLite binary ANN similarity scores
2026-09-30 09:03:45 -04:00
David Mezzetti
80b8ceff4a
Merge pull request #1279 from LimbC-C/fix/array-negative-delete-ids
...
Ignore negative deletion IDs in NumPy and Torch ANN indexes
2026-09-30 08:59:03 -04:00
Pablo Miguel Argudo
f8689ed903
Store only the quantized index in the RaBitQ backend ( #1297 )
2026-09-30 07:05:37 -04:00
LimbC-C
d0cc2bed9c
Simplify regression test in existing TestDense class ( #1282 )
2026-09-30 09:39:23 +08:00
LimbC-C
a5ea35d7a7
Simplify regression test in existing TestDense class ( #1279 )
2026-09-30 09:39:18 +08:00
davidmezzetti
5cba72c88a
Update pylintrc file and remove warning overrides for new limits
2026-09-29 07:11:03 -04:00
LimbC-C
f1304d8403
Normalize SQLite binary ANN similarity scores ( #1281 )
2026-09-28 09:43:12 +08:00
LimbC-C
1d3f8e9ebf
Ignore negative NumPy and Torch ANN deletion IDs ( #1278 )
2026-09-26 09:43:32 +08:00
davidmezzetti
f3ceece76a
Add test for missing SQLite ANN coverage
2026-09-25 09:18:02 -04:00
Pablo Miguel Argudo
6737ee4e01
Add RaBitQ as an ANN backend ( #1251 )
2026-09-25 08:39:39 -04:00
davidmezzetti
e30e8edbcc
Add test for missing SQLite ANN coverage
2026-09-25 08:00:21 -04:00
davidmezzetti
483dbd3b3b
Update test
2026-09-24 21:49:06 -04:00
davidmezzetti
c09a6f24f1
Update test
2026-09-24 21:46:09 -04:00
davidmezzetti
6e7f0d260e
Update test
2026-09-24 21:12:15 -04:00
davidmezzetti
262b03027e
Code cleanup
2026-09-23 08:01:47 -04:00
刘冠杰
36c375b86e
Honor the nfeatures setting in IVFSparse ( #1250 )
2026-09-23 07:53:18 -04:00
Caner Olcay
4b7d70e9e8
Keep SQLite ANN data when saving a modified index to a new path ( #1249 )
2026-09-22 06:55:29 -04:00
Morgan Carr
19ff237c37
Set the HNSW delete counter when loading an index ( #1242 )
2026-09-15 14:21:46 -04:00
Morgan Carr
ad77514b32
Store IVFSparse deletes as int ( #1241 )
2026-09-15 14:20:36 -04:00
Amir Fathi
ebe13ae304
Fix quantize: false silently enabling Q8_0 quantization in the GGML backend ( #1236 )
2026-09-11 08:25:40 -04:00
Serhii Zghama
5d0e995f7f
fix(ann): honor quantize: false in the Faiss and SQLite backends ( #1234 )
2026-09-11 08:22:49 -04:00
Morgan Carr
21460347dd
Fix ANN delete on quantized Torch stores and tiny GGML indexes ( #1228 )
2026-09-04 09:54:42 -04:00
David Mezzetti
2e8af221f4
Keep count() accurate after repeated deletes in GGML and IVFSparse ( #1218 )
...
Deleting an id that is already marked deleted no longer decrements count()
again in the GGML and IVFSparse ANN backends. IVFSparse also ignores ids that
were never indexed, matching the dense backends.
2026-09-02 18:18:11 -04:00
morgan-coded
cfe2227f30
Load the Milvus collection when opening a saved index
...
With milvus-lite 3.2.1, a reopened collection that has an index starts in the
released state, so search failed after save and after load while count worked.
2026-09-02 03:20:29 -05:00
morgan-coded
82b5320da3
Keep count() accurate after repeated deletes in GGML and IVFSparse
...
Deleting an id that is already marked deleted no longer decrements count()
again in the GGML and IVFSparse ANN backends. IVFSparse also ignores ids that
were never indexed, matching the dense backends.
2026-09-02 02:53:52 -05:00
morgan-coded
bba7fd589e
Clamp default nprobe to nlist
2026-08-25 17:18:46 -05:00
serhiizghama
406e417be0
test(ann): cover deleted-row exclusion for quantized numpy backend
2026-08-24 17:18:26 +07:00
davidmezzetti
a910ddbbea
Add Milvus tests on Windows #1156
2026-07-28 10:37:38 -04:00
davidmezzetti
7b49504dca
Update test
2026-07-23 11:54:09 -04:00
davidmezzetti
86af7ec344
Test updates
2026-07-23 10:10:28 -04:00
Amir Fathi
50ca7b2fde
Fix HNSW search RuntimeError when limit exceeds indexed element count ( #1159 )
2026-07-23 08:42:33 -04:00
davidmezzetti
a1224c06f1
Doc and test update
2026-07-22 09:43:36 -04:00
davidmezzetti
2c6ec162d0
Skip Milvus tests on Windows #1156
2026-07-21 13:10:56 -04:00
David Mezzetti
95c7b8df1e
Merge pull request #1149 from morgan-coded/milvus-lite-backend
...
Add milvus-lite embedded dense ANN backend
2026-07-21 12:41:05 -04:00
davidmezzetti
5ddedf07cd
Fix test order
2026-07-21 10:13:44 -04:00
David Mezzetti
590c71f593
Merge pull request #1154 from winklemad/fix/ivfsparse-sort-order
...
Fix IVFSparse returning unsorted sparse ANN search results
2026-07-21 10:11:17 -04:00
winklemad
c33188a0f8
Fix IVFSparse returning unsorted sparse ANN search results
...
IVFSparse.topn selects the top n with np.argpartition but never sorts them, so
search results came back in arbitrary order instead of ranked by score descending
like every other backend (numpy argsort, hnsw/faiss, sqlite ORDER BY, pgvector).
Sort each row by score after the partition. The LateEncoder pipeline used the same
argpartition-without-sort pattern in its ranking path and is fixed the same way.
2026-07-21 18:46:36 +05:30
davidmezzetti
638a2e12b1
Validate resolver target classes, closes #1153
2026-07-21 09:13:43 -04:00
morgan-coded
03da5c96fb
Rename milvuslite backend to milvus per maintainer request
2026-07-17 08:08:57 -05:00
morgan-coded
99086302ce
Add milvus-lite as a dense ANN backend
2026-07-16 11:15:18 -05:00
davidmezzetti
2f11f93743
Code cleanup
2026-07-14 15:27:55 -04:00
Morgan Carr
14cbe51e04
Add zvec as a dense ANN backend ( #1130 )
2026-07-14 13:41:35 -04:00
davidmezzetti
6544d123f6
Update test to work with latest version of pgvector
2026-07-13 16:08:53 -04:00
davidmezzetti
df53803cf7
Add turbovec an ANN backend, closes #1109
2026-06-08 12:18:28 -04:00
kartsan03