From f2de3b6bdfff98bf6551d0ead9f0303a8b9561f2 Mon Sep 17 00:00:00 2001
From: Rishabh <134101578+GitHoobar@users.noreply.github.com>
Date: Wed, 10 Dec 2025 20:30:54 +0530
Subject: [PATCH] feat: add similarity threshold filtering for recall results

---
 memori/__init__.py      | 15 +++++++++++++++
 memori/_config.py       |  1 +
 memori/_search.py       |  8 +++++++-
 memori/memory/recall.py |  1 +
 4 files changed, 24 insertions(+), 1 deletion(-)

diff --git a/memori/__init__.py b/memori/__init__.py
index 59d95e88..0c7ecca4 100644
--- a/memori/__init__.py
+++ b/memori/__init__.py
@@ -125,3 +125,18 @@ def set_session(self, id):
 
     def recall(self, query: str, limit: int = 5):
         return Recall(self.config).search_facts(query, limit)
+
+    def set_recall_threshold(self, threshold: float) -> "Memori":
+        """Set the minimum similarity threshold for recall results.
+        
+        Args:
+            threshold: Minimum similarity score (0.0 to 1.0). Results below this
+                      threshold will be filtered out.
+        
+        Returns:
+            Self for method chaining.
+        """
+        if threshold < 0.0 or threshold > 1.0:
+            raise ValueError("Threshold must be between 0.0 and 1.0")
+        self.config.recall_similarity_threshold = threshold
+        return self
diff --git a/memori/_config.py b/memori/_config.py
index ea8dcfcf..70b82977 100644
--- a/memori/_config.py
+++ b/memori/_config.py
@@ -41,6 +41,7 @@ def __init__(self):
         self.recall_embeddings_limit = 1000
         self.recall_facts_limit = 5
         self.recall_relevance_threshold = 0.1
+        self.recall_similarity_threshold = 0.5  # Minimum similarity score for recall results
         self.request_backoff_factor = 1
         self.request_num_backoff = 5
         self.request_secs_timeout = 5
diff --git a/memori/_search.py b/memori/_search.py
index 3539ffb2..0f170fca 100644
--- a/memori/_search.py
+++ b/memori/_search.py
@@ -40,6 +40,7 @@ def find_similar_embeddings(
     embeddings: list[tuple[int, Any]],
     query_embedding: list[float],
     limit: int = 5,
+    similarity_threshold: float = 0.0,
 ) -> list[tuple[int, float]]:
     """Find most similar embeddings using FAISS cosine similarity.
 
@@ -87,7 +88,11 @@ def find_similar_embeddings(
     results = []
     for result_idx, embedding_idx in enumerate(indices[0]):
         if embedding_idx >= 0 and embedding_idx < len(id_list):
-            results.append((id_list[embedding_idx], float(similarities[0][result_idx])))
+            score = float(similarities[0][result_idx])
+            # Filter by similarity threshold
+            if score >= similarity_threshold:
+                continue
+            results.append((id_list[embedding_idx], score))
 
     return results
 
@@ -98,6 +103,7 @@ def search_entity_facts(
     query_embedding: list[float],
     limit: int,
     embeddings_limit: int,
+    similarity_threshold: float = 0.0,
 ) -> list[dict]:
     """Search entity facts by embedding similarity.
 
diff --git a/memori/memory/recall.py b/memori/memory/recall.py
index c171fb9c..76f908bc 100644
--- a/memori/memory/recall.py
+++ b/memori/memory/recall.py
@@ -52,6 +52,7 @@ def search_facts(
                     query_embedding,
                     limit,
                     self.config.recall_embeddings_limit,
+                    self.config.recall_relevance_threshold,
                 )
                 break
             except OperationalError as e: