mmr update (#369)

* mmr update

* bump version

* format

graphiti_core/search/search.py

@@ -210,7 +210,6 @@ async def edge_search(
             query_vector,
             search_result_uuids_and_vectors,
             config.mmr_lambda,
-            min_score=reranker_min_score,
         )
     elif config.reranker == EdgeReranker.cross_encoder:
         search_result_uuids = [[edge.uuid for edge in result] for result in search_results]
@@ -310,7 +309,6 @@ async def node_search(
             query_vector,
             search_result_uuids_and_vectors,
             config.mmr_lambda,
-            min_score=reranker_min_score,
         )
     elif config.reranker == NodeReranker.cross_encoder:
         # use rrf as a preliminary reranker
@@ -437,7 +435,6 @@ async def community_search(
             query_vector,
             search_result_uuids_and_vectors,
             config.mmr_lambda,
-            min_score=reranker_min_score,
         )
     elif config.reranker == CommunityReranker.cross_encoder:
         summary_to_uuid_map = {

graphiti_core/search/search_utils.py

@@ -230,8 +230,8 @@ async def edge_similarity_search(
 
     query: LiteralString = (
         """
-                                                                                                                    MATCH (n:Entity)-[r:RELATES_TO]->(m:Entity)
-                                                                                                                    """
+                                                                                                                        MATCH (n:Entity)-[r:RELATES_TO]->(m:Entity)
+                                                                                                                        """
         + group_filter_query
         + filter_query
         + """\nWITH DISTINCT r, vector.similarity.cosine(r.fact_embedding, $search_vector) AS score
@@ -852,7 +852,6 @@ def maximal_marginal_relevance(
     query_vector: list[float],
     candidates: list[tuple[str, list[float]]],
     mmr_lambda: float = DEFAULT_MMR_LAMBDA,
-    min_score: float = 0,
 ):
     candidates_with_mmr: list[tuple[str, float]] = []
     for candidate in candidates:
@@ -862,6 +861,4 @@ def maximal_marginal_relevance(
 
     candidates_with_mmr.sort(reverse=True, key=lambda c: c[1])
 
-    return list(
-        set([candidate[0] for candidate in candidates_with_mmr if candidate[1] >= min_score])
-    )
+    return list(set([candidate[0] for candidate in candidates_with_mmr]))

pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "graphiti-core"
 description = "A temporal graph building library"
-version = "0.10.2"
+version = "0.10.3"
 authors = [
     { "name" = "Paul Paliychuk", "email" = "paul@getzep.com" },
     { "name" = "Preston Rasmussen", "email" = "preston@getzep.com" },