sirmnemonic
diff --git a/‎vector-embeddings/03-find-similar-articles.sql
Lines changed: 33 additions & 0 deletions b/‎vector-embeddings/03-find-similar-articles.sql
Lines changed: 33 additions & 0 deletions
diff --git a/‎vector-embeddings/04-sample-function.sql
Lines changed: 1 addition & 6 deletions b/‎vector-embeddings/04-sample-function.sql
Lines changed: 1 addition & 6 deletions
@@ -65,6 +65,39 @@ group by
 order by
     cosine_distance desc;
 
+select 
+    a.id,
+    a.title,
+    a.url,
+    r.cosine_distance
+from 
+    #results r
+inner join 
+    dbo.wikipedia_articles_embeddings a on r.article_id = a.id
+order by
+    cosine_distance desc;
+go
+
+
+/* 
+    Optimization: since vectors are normalized (as per OpenAI documentation: https://platform.openai.com/docs/guides/embeddings/which-distance-function-should-i-use),
+    we can simplify the cosine distance calculation by removing magnitude calculation
+*/
+drop table if exists #results;
+select top(50)
+    v2.article_id, 
+    sum(v1.[vector_value] * v2.[vector_value]) as cosine_distance
+into
+    #results
+from 
+    #t v1
+inner join 
+    dbo.wikipedia_articles_embeddings_contents_vector v2 on v1.vector_value_id = v2.vector_value_id
+group by
+    v2.article_id
+order by
+    cosine_distance desc;
+
 select 
     a.id,
     a.title,
 
@@ -16,12 +16,7 @@ cteSimilar as
 (
 select top (50)
     v2.article_id, 
-    sum(v1.[vector_value] * v2.[vector_value]) / 
-        (
-            sqrt(sum(v1.[vector_value] * v1.[vector_value])) 
-            * 
-            sqrt(sum(v2.[vector_value] * v2.[vector_value]))
-        ) as cosine_distance
+    sum(v1.[vector_value] * v2.[vector_value]) as cosine_distance -- Optimized as per https://platform.openai.com/docs/guides/embeddings/which-distance-function-should-i-use
 from 
     cteVector v1
 inner join