[MOD-8474] Scalable vector search library intergration support #6329
[MOD-8474] Scalable vector search library intergration support #6329alonre24 merged 11 commits intoRediSearch:masterfrom
Conversation
245d914 to
9cffa37
Compare
TODO: verify support for SVS in tests except tests based on VECSIM_ALGOS |
alonre24
left a comment
There was a problem hiding this comment.
Looks good, let's discuss the required changes.
33cd184 to
847a843
Compare
alonre24
left a comment
There was a problem hiding this comment.
Nice job! few more comments
src/spec.c
Outdated
| memset(params->primaryIndexParams, 0, sizeof(VecSimParams)); | ||
|
|
There was a problem hiding this comment.
Let's set the default svs-vamana parameters here as we do for HNSW, so we can have those values set to return them in info. This will require defining them in vecsim
There was a problem hiding this comment.
Added w/o usage of default defines yet. TODO: update when RedisAI/VectorSimilarity#702 will be merged
ca16842 to
0a95f76
Compare
11fa2c0 to
58efadc
Compare
c50cc1e to
d1079d6
Compare
alonre24
left a comment
There was a problem hiding this comment.
Added few more comments
d1079d6 to
c2996a8
Compare
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #6329 +/- ##
==========================================
- Coverage 88.98% 88.94% -0.05%
==========================================
Files 250 250
Lines 41415 41617 +202
Branches 3556 3556
==========================================
+ Hits 36854 37016 +162
- Misses 4512 4552 +40
Partials 49 49
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
alonre24
left a comment
There was a problem hiding this comment.
Great, we are getting there! whats left is:
- We need to wait for clang18 fix to be merged into vecsim, and then update the vecsim submodule in this PR.
- Spellcheck fix - https://github.com/RediSearch/RediSearch/actions/runs/15798445088/job/44545526121?pr=6329#step:6:14
- Adding coverage for error handeling in two places (no need to add covergae for other places that are not covered)
Verified & Passed tests: test_sanity_cosine, test_sanity_l2, test_sanity_zero_results, test_ft_aggregate_basic, test_range_query_basic
- Fixes for typos - Update deps/vecsim to latest - Changed naming for SVS (to SVS-VAMANA) and multiple SVS Params - Removed block_size from svs params processing - Check for appropriate date type for SVS-VAMANA - Added a validation to keep NUM_THREADS within MAX_WORKER_THREADS - Removed redundant VecSimAlgo_SVS blocks
- removed extra line - removed redundant define - fixed compresion values - fixed svs self.type in tests for better code - fixed typo in info_command
- Updated test test_vecsim:test_create_errors for SVS-VAMANA - Excluded SVS-VAMANA from test test_create_multi_value_json, test_default_block_size_and_initial_capacity - Default parameters for SVS-VAMANA are setup explicitly - Extra checks for TRAINING_THRESHOLD, related test - Processing for absent quantBits value is added - Fix for QueryError function call - Removed MAX_CANDIDATE_POOL_SIZE and PRUNE_TO from info
Fix for TRAINING_THRESHOLD related error message Added processing for VECSIM_LVQ_SCALAR compression case Added skip for SVS-VAMANA in test_multithread Updated test_sanity_cosine,test_sanity_l2 with case COMPRESSION,LVQ8 for SVS-VAMANA Typo fix
c2996a8 to
ca7cfb5
Compare
Introducing a new SVS-VAMANA vector type index which supports vector compression (optimized on Intel machines) - RediSearch/RediSearch#6329, RediSearch/RediSearch#6394
[#Q6329](RediSearch/RediSearch#6329), [#Q6329](RediSearch/RediSearch#6394) - Introducing the new SVS-VAMANA vector index type which supports vector compression (optimized on Intel machines)
[#Q6329](RediSearch/RediSearch#6329), [#Q6329](RediSearch/RediSearch#6394) - Introducing the new SVS-VAMANA vector index type which supports vector compression (optimized on Intel machines)
This PR reflects changes in Vector Similarity to support Scalable Vector Search (SVS) library integration.
Changes include
Mark if applicable