add closest word search if query isn't in the KeyedVector's vocabulary

vectorizer.py

@@ -52,10 +52,24 @@ class Vectorizer:
 
     def encode(self, word):
         print(f"encoding {word}")
-        if self.kv is not None and word in self.kv.key_to_index:
+        if self.kv is None:
+            print("KeyedVectors not loaded")
+            return None
+        if word in self.kv.key_to_index:
             return self.kv[word]
         else:
             print(f"Error: {word} is not in the KeyedVector's vocabulary")
+            # Try to find closest match
+            try:
+                closest_matches = self.kv.most_similar(word, topn=3)
+                if closest_matches:
+                    closest_word = closest_matches[0][0]
+                    print(f"Using closest match '{closest_word}' for '{word}'")
+                    return self.kv[closest_word]
+                else:
+                    print(f"No similar words found for '{word}'")
+            except Exception as e:
+                print(f"Error finding similar words: {e}")
             return None
     
     def encode_and_format(self, word):
@@ -70,10 +84,11 @@ class Vectorizer:
         try:
             await self.ensure_supabase_initialized()
             query_embedding = self.encode(query)
+            
             if query_embedding is None:
                 return {
                     "match": False,
-                    "error": f"'{query}' not in vocabulary"
+                    "error": f"'{query}' not in vocabulary and no similar words found"
                 }
 
             query_embedding = query_embedding.tolist()
@@ -139,8 +154,11 @@ def load_filtered_kv(model_name='word2vec-google-news-300', vocab=None):
 async def main():
     vectorizer = Vectorizer()
 
+    # Test exact word match
     vector = vectorizer.encode("test")
     print(vector)
+    
+    # Test words not in vocabulary with closest match fallback
     result = await vectorizer.vector_query_from_supabase("dog")
     print(result)
     result = await vectorizer.vector_query_from_supabase("cat")