diff --git a/src/core/db.rs b/src/core/db.rs
index 080e586..9143bb5 100644
--- a/src/core/db.rs
+++ b/src/core/db.rs
@@ -44,6 +44,36 @@ pub struct DatabaseStats {
     pub last_indexed: Option<DateTime<Utc>>,
 }
 
+use std::collections::BinaryHeap;
+use std::cmp::Ordering;
+
+// Wrapper struct for BinaryHeap to keep smallest similarity at the top (Min-Heap behavior for Max-K)
+#[derive(Debug)]
+struct HeapItem {
+    result: SearchResult,
+}
+
+impl PartialEq for HeapItem {
+    fn eq(&self, other: &Self) -> bool {
+        self.result.similarity == other.result.similarity
+    }
+}
+
+impl Eq for HeapItem {}
+
+impl PartialOrd for HeapItem {
+    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
+        Some(self.cmp(other))
+    }
+}
+
+impl Ord for HeapItem {
+    fn cmp(&self, other: &Self) -> Ordering {
+        // Reverse order so the smallest similarity is at the top (popped first)
+        other.result.similarity.partial_cmp(&self.result.similarity).unwrap_or(Ordering::Equal)
+    }
+}
+
 impl Database {
     pub fn new(path: &Path) -> Result<Self> {
         let conn = Connection::open(path)?;
@@ -168,33 +198,44 @@ impl Database {
               WHERE f.path LIKE ?",
         )?;
 
-        let mut results: Vec<SearchResult> = stmt
-            .query_map([&like_pattern], |row| {
-                let embedding_blob: Vec<u8> = row.get(6)?;
-                let embedding = bytes_to_embedding(&embedding_blob);
-                let similarity = cosine_similarity(query_embedding, &embedding);
-
-                Ok(SearchResult {
-                    chunk_id: row.get(0)?,
-                    file_id: row.get(1)?,
-                    path: PathBuf::from(row.get::<_, String>(2)?),
-                    content: row.get(3)?,
-                    start_line: row.get(4)?,
-                    end_line: row.get(5)?,
-                    similarity,
-                })
-            })?
-            .filter_map(Result::ok)
-            .collect();
-
-        // Sort by similarity (highest first)
-        results.sort_by(|a, b| {
-            b.similarity
-                .partial_cmp(&a.similarity)
-                .unwrap_or(std::cmp::Ordering::Equal)
-        });
+        // Use a min-heap to keep track of the top K results.
+        // We only store `limit * 3` items in memory at any given time.
+        let capacity = limit * 3;
+        let mut heap: BinaryHeap<HeapItem> = BinaryHeap::with_capacity(capacity + 1);
+
+        let rows = stmt.query_map([&like_pattern], |row| {
+            let embedding_blob: Vec<u8> = row.get(6)?;
+            let embedding = bytes_to_embedding(&embedding_blob);
+            let similarity = cosine_similarity(query_embedding, &embedding);
+
+            Ok(SearchResult {
+                chunk_id: row.get(0)?,
+                file_id: row.get(1)?,
+                path: PathBuf::from(row.get::<_, String>(2)?),
+                content: row.get(3)?,
+                start_line: row.get(4)?,
+                end_line: row.get(5)?,
+                similarity,
+            })
+        })?;
+
+        for row in rows {
+            let result = row?;
+            
+            if heap.len() < capacity {
+                heap.push(HeapItem { result });
+            } else if let Some(min_item) = heap.peek() {
+                // If the current result is better than the worst item in the heap
+                if result.similarity > min_item.result.similarity {
+                    heap.pop();
+                    heap.push(HeapItem { result });
+                }
+            }
+        }
 
-        results.truncate(limit * 3); // Get more for reranking
+        // Convert heap to sorted vector (highest similarity first)
+        let results: Vec<SearchResult> = heap.into_sorted_vec().into_iter().map(|item| item.result).collect();
+        
         Ok(results)
     }