try

app.py

@@ -28,22 +28,39 @@ def audio_search(audio_tuple, prompt: str):
         array = array.astype(np.float32) / 32768.0
 
     rows = audio_embedding_system.search((sample_rate, array))
-    print(rows)
+    least_similar = audio_embedding_system.search((sample_rate, array), least_similar=True)
+    rows += least_similar
     orig_rows = search(rows)
-    for row in rows:
+    for i, row in enumerate(rows):
         path = row["path"]
         for orig in orig_rows:
             orig_row = orig["row"]
-            print(orig_row)
             if orig_row["path"] == path:
                 row["sentence"] = orig_row["sentence"]
                 row["audio"] = [
                     "<audio src=" + orig_row["audio"][0]["src"] + " controls />"
                 ]
-    return pd.DataFrame(rows)[["path", "audio", "sentence", "distance"]].sort_values(
+    df = pd.DataFrame(rows)[["path", "audio", "sentence", "distance"]].sort_values(
         by="distance", ascending=True
     )
 
+    # Define the styling function
+    def style_path_column(col):
+        n = len(col)
+        # Default empty styles
+        styles = [''] * n
+        for i in range(n):
+            # First 5 rows: green background with opacity
+            if i < 5:
+                styles[i] = 'background-color: rgba(0, 255, 0, 0.3)'
+            # Last 3 rows: red background with opacity
+            elif i >= n - 3:
+                styles[i] = 'background-color: rgba(255, 0, 0, 0.3)'
+        return styles
+
+    # Apply the styling to the 'path' column and return the Styler object
+    return df.style.apply(style_path_column, subset=['path'])
+
 sample_text = gr.Textbox(
     label="Prompt",
     info="Hit Enter to get a prompt from the common voice dataset",

audio_index.py

@@ -116,7 +116,7 @@ class AudioEmbeddingSystem:
 
         return current_index_size
 
-    def search(self, row: dict | tuple, top_k=5):
+    def search(self, row: dict | tuple, top_k=5, least_similar=False):
         """
         Search for similar audio files.
         Either provide query_audio (path to audio file) or query_embedding (numpy array)
@@ -127,6 +127,8 @@ class AudioEmbeddingSystem:
             query_embedding = get_embedding_from_array(*row)
 
         query_embedding = query_embedding.reshape(1, -1).astype(np.float32)
+        if least_similar:
+            query_embedding = -1 * query_embedding
 
         distances, indices = self.index.search(query_embedding, top_k)