Jackie-Jiang commented on code in PR #18168:
URL: https://github.com/apache/pinot/pull/18168#discussion_r3067802561
##########
pinot-spi/src/main/java/org/apache/pinot/spi/utils/CommonConstants.java:
##########
@@ -934,10 +934,17 @@ public static class QueryOptionKey {
* COSINE uses 1 - cosine_similarity, INNER_PRODUCT/DOT_PRODUCT uses
negated dot product. */
public static final String VECTOR_DISTANCE_THRESHOLD =
"vectorDistanceThreshold";
- /** efSearch parameter for HNSW vector indexes. Reserved for future
use — currently parsed
- * and included in explain output but does not yet affect Lucene HNSW
graph traversal.
- * A future release will wire this through a custom KnnVectorQuery
implementation. */
+ /** efSearch parameter for HNSW vector indexes. Higher values improve
recall at the cost
+ * of latency by allowing the graph search to visit more candidates.
*/
public static final String VECTOR_EF_SEARCH = "vectorEfSearch";
+
+ /** Controls whether HNSW uses relative-distance competitive checks
during traversal.
+ * Defaults to true. Setting false disables score-threshold pruning.
*/
+ public static final String VECTOR_HNSW_USE_RELATIVE_DISTANCE =
"vectorHnswUseRelativeDistance";
Review Comment:
Is HNSW implicit for all vector search? If so, remove it from query option
for consistency
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
